entry_point
stringlengths 1
65
| original_triton_code
stringlengths 4.5k
619k
| python_code
stringlengths 208
60.9k
| triton_code
stringlengths 1.15k
275k
| repo_name
stringlengths 7
115
| module_name
stringlengths 1
65
| synthetic
bool 1
class | uuid
int64 0
18.5k
| licenses
listlengths 1
6
| stars
int64 0
19.8k
| sha
stringlengths 40
40
| repo_link
stringlengths 72
180
|
|---|---|---|---|---|---|---|---|---|---|---|---|
ComplexConv2d
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ss/cssphvn3aplsuvxcraiqbvuxfxzncuc3v2wfef6e6yugscpjj4ui.py
# Topologically Sorted Source Nodes: [conv2d, conv2d_1, sub, conv2d_2, conv2d_3, add], Original ATen: [aten.convolution, aten.sub, aten.add]
# Source node to ATen node mapping:
# add => add
# conv2d => convolution
# conv2d_1 => convolution_1
# conv2d_2 => convolution_2
# conv2d_3 => convolution_3
# sub => sub
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
# %convolution_1 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_6, %primals_4, %primals_5, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%convolution, %convolution_1), kwargs = {})
# %convolution_2 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_6, %primals_1, %primals_2, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
# %convolution_3 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_4, %primals_5, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%convolution_2, %convolution_3), kwargs = {})
triton_poi_fused_add_convolution_sub_0 = async_compile.triton('triton_poi_fused_add_convolution_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_convolution_sub_0', 'mutated_arg_names': ['in_out_ptr0', 'in_out_ptr1'], 'no_x_dim': False, 'num_load': 6, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_convolution_sub_0(in_out_ptr0, in_out_ptr1, in_ptr0, in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 4) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + (x3), xmask)
tmp4 = tl.load(in_ptr2 + (x1), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_out_ptr1 + (x3), xmask)
tmp9 = tl.load(in_ptr3 + (x3), xmask)
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 - tmp5
tmp8 = tmp7 + tmp1
tmp10 = tmp9 + tmp4
tmp11 = tmp8 + tmp10
tl.store(in_out_ptr0 + (x3), tmp6, xmask)
tl.store(in_out_ptr1 + (x3), tmp11, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 2, 2), (16, 4, 2, 1))
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(primals_6, primals_4, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 4, 2, 2), (16, 4, 2, 1))
# Topologically Sorted Source Nodes: [conv2d_2], Original ATen: [aten.convolution]
buf3 = extern_kernels.convolution(primals_6, primals_1, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 4, 2, 2), (16, 4, 2, 1))
# Topologically Sorted Source Nodes: [conv2d_3], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(primals_3, primals_4, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 2, 2), (16, 4, 2, 1))
buf2 = buf0; del buf0 # reuse
buf5 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [conv2d, conv2d_1, sub, conv2d_2, conv2d_3, add], Original ATen: [aten.convolution, aten.sub, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_convolution_sub_0.run(buf2, buf5, primals_2, buf1, primals_5, buf4, 64, grid=grid(64), stream=stream0)
del buf1
del buf4
del primals_2
del primals_5
return (buf2, buf5, primals_1, primals_3, primals_4, primals_6, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 3, 3), (36, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4, 3, 3), (36, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from torch.nn import Module
import torch
from torch.nn import Conv2d
class ComplexConv2d(Module):
def __init__(self, in_channels, out_channels, kernel_size=3, stride=1,
padding=0, dilation=1, groups=1, bias=True):
super(ComplexConv2d, self).__init__()
self.conv_r = Conv2d(in_channels, out_channels, kernel_size, stride,
padding, dilation, groups, bias)
self.conv_i = Conv2d(in_channels, out_channels, kernel_size, stride,
padding, dilation, groups, bias)
def forward(self, input_r, input_i):
return self.conv_r(input_r) - self.conv_i(input_i), self.conv_r(input_i
) + self.conv_i(input_r)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_channels': 4, 'out_channels': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch.nn import Module
from torch.nn import Conv2d
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
@triton.jit
def triton_poi_fused_add_convolution_sub_0(in_out_ptr0, in_out_ptr1,
in_ptr0, in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 4 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + x3, xmask)
tmp4 = tl.load(in_ptr2 + x1, xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_out_ptr1 + x3, xmask)
tmp9 = tl.load(in_ptr3 + x3, xmask)
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 - tmp5
tmp8 = tmp7 + tmp1
tmp10 = tmp9 + tmp4
tmp11 = tmp8 + tmp10
tl.store(in_out_ptr0 + x3, tmp6, xmask)
tl.store(in_out_ptr1 + x3, tmp11, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 2, 2), (16, 4, 2, 1))
buf1 = extern_kernels.convolution(primals_6, primals_4, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 4, 2, 2), (16, 4, 2, 1))
buf3 = extern_kernels.convolution(primals_6, primals_1, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 4, 2, 2), (16, 4, 2, 1))
buf4 = extern_kernels.convolution(primals_3, primals_4, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 2, 2), (16, 4, 2, 1))
buf2 = buf0
del buf0
buf5 = buf3
del buf3
get_raw_stream(0)
triton_poi_fused_add_convolution_sub_0[grid(64)](buf2, buf5,
primals_2, buf1, primals_5, buf4, 64, XBLOCK=64, num_warps=1,
num_stages=1)
del buf1
del buf4
del primals_2
del primals_5
return buf2, buf5, primals_1, primals_3, primals_4, primals_6
class ComplexConv2dNew(Module):
def __init__(self, in_channels, out_channels, kernel_size=3, stride=1,
padding=0, dilation=1, groups=1, bias=True):
super(ComplexConv2dNew, self).__init__()
self.conv_r = Conv2d(in_channels, out_channels, kernel_size, stride,
padding, dilation, groups, bias)
self.conv_i = Conv2d(in_channels, out_channels, kernel_size, stride,
padding, dilation, groups, bias)
def forward(self, input_0, input_1):
primals_1 = self.conv_r.weight
primals_2 = self.conv_r.bias
primals_4 = self.conv_i.weight
primals_5 = self.conv_i.bias
primals_3 = input_0
primals_6 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6])
return output[0], output[1]
|
drydenwiebe/complexPyTorch
|
ComplexConv2d
| false | 12,323 |
[
"MIT"
] | 0 |
cea88ba7ee5692dfa1b40f0ba609ef14160d5073
|
https://github.com/drydenwiebe/complexPyTorch/tree/cea88ba7ee5692dfa1b40f0ba609ef14160d5073
|
Transition
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/nh/cnhx37tsffx4r7taj3xi72s7yfpnnccem24fupfbht6b7bzliavu.py
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.gelu]
# Source node to ATen node mapping:
# x_1 => add, erf, mul, mul_1, mul_2
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_1, 0.5), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_1, 0.7071067811865476), kwargs = {})
# %erf : [num_users=1] = call_function[target=torch.ops.aten.erf.default](args = (%mul_1,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%erf, 1), kwargs = {})
# %mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%mul, %add), kwargs = {})
triton_poi_fused_gelu_0 = async_compile.triton('triton_poi_fused_gelu_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_gelu_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_gelu_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 0.5
tmp2 = tmp0 * tmp1
tmp3 = 0.7071067811865476
tmp4 = tmp0 * tmp3
tmp5 = libdevice.erf(tmp4)
tmp6 = 1.0
tmp7 = tmp5 + tmp6
tmp8 = tmp2 * tmp7
tl.store(out_ptr0 + (x0), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.gelu]
stream0 = get_raw_stream(0)
triton_poi_fused_gelu_0.run(buf0, buf1, 256, grid=grid(256), stream=stream0)
return (buf1, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.parallel
class Transition(nn.Module):
def __init__(self, in_features, out_features, act_layer=nn.GELU):
super(Transition, self).__init__()
self.act = act_layer()
self.linear = nn.Linear(in_features, out_features)
def forward(self, x):
x = self.linear(x)
x = self.act(x)
return x
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_features': 4, 'out_features': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
import torch.nn.parallel
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_gelu_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 0.5
tmp2 = tmp0 * tmp1
tmp3 = 0.7071067811865476
tmp4 = tmp0 * tmp3
tmp5 = libdevice.erf(tmp4)
tmp6 = 1.0
tmp7 = tmp5 + tmp6
tmp8 = tmp2 * tmp7
tl.store(out_ptr0 + x0, tmp8, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (64,
4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_gelu_0[grid(256)](buf0, buf1, 256, XBLOCK=256,
num_warps=4, num_stages=1)
return buf1, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), buf0
class TransitionNew(nn.Module):
def __init__(self, in_features, out_features, act_layer=nn.GELU):
super(TransitionNew, self).__init__()
self.act = act_layer()
self.linear = nn.Linear(in_features, out_features)
def forward(self, input_0):
primals_1 = self.linear.weight
primals_2 = self.linear.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
druzhkov-paul/T2T-ViT
|
Transition
| false | 12,324 |
[
"BSD-3-Clause-Clear"
] | 0 |
819c3ddc4cb6f464d4a9866d8713c7ace42ebf6c
|
https://github.com/druzhkov-paul/T2T-ViT/tree/819c3ddc4cb6f464d4a9866d8713c7ace42ebf6c
|
_TestNetStrided
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/yk/cykarwk3qy3pmem22bdbqxnd35nfbs5r3u5357r5uoderrncltwa.py
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# conv2d => convolution
# Graph fragment:
# %convolution : [num_users=2] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_0 = async_compile.triton('triton_poi_fused_convolution_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[524288],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 288000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 3600) % 20
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/3k/c3k57g5yhtujuoxgs7jwv6kipspabfze46wndcbnl5hbd4s7xuz5.py
# Topologically Sorted Source Nodes: [max_pool2d, x], Original ATen: [aten.max_pool2d_with_indices, aten.relu]
# Source node to ATen node mapping:
# max_pool2d => _low_memory_max_pool2d_with_offsets, getitem_1
# x => relu
# Graph fragment:
# %_low_memory_max_pool2d_with_offsets : [num_users=2] = call_function[target=torch.ops.prims._low_memory_max_pool2d_with_offsets.default](args = (%convolution, [2, 2], [2, 2], [0, 0], [1, 1], False), kwargs = {})
# %getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%_low_memory_max_pool2d_with_offsets, 1), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%getitem,), kwargs = {})
triton_poi_fused_max_pool2d_with_indices_relu_1 = async_compile.triton('triton_poi_fused_max_pool2d_with_indices_relu_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[131072],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i8', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_max_pool2d_with_indices_relu_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_max_pool2d_with_indices_relu_1(in_ptr0, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 72000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 30
x3 = (xindex // 30)
x2 = (xindex // 18000)
x4 = xindex % 18000
x5 = xindex
tmp0 = tl.load(in_ptr0 + ((2*x0) + (120*x3)), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + (2*x0) + (120*x3)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (60 + (2*x0) + (120*x3)), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr0 + (61 + (2*x0) + (120*x3)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 > tmp0
tmp3 = tl.full([1], 1, tl.int8)
tmp4 = tl.full([1], 0, tl.int8)
tmp5 = tl.where(tmp2, tmp3, tmp4)
tmp6 = triton_helpers.maximum(tmp1, tmp0)
tmp8 = tmp7 > tmp6
tmp9 = tl.full([1], 2, tl.int8)
tmp10 = tl.where(tmp8, tmp9, tmp5)
tmp11 = triton_helpers.maximum(tmp7, tmp6)
tmp13 = tmp12 > tmp11
tmp14 = tl.full([1], 3, tl.int8)
tmp15 = tl.where(tmp13, tmp14, tmp10)
tmp16 = triton_helpers.maximum(tmp12, tmp11)
tmp17 = tl.full([1], 0, tl.int32)
tmp18 = triton_helpers.maximum(tmp17, tmp16)
tl.store(out_ptr0 + (x4 + (18048*x2)), tmp15, xmask)
tl.store(out_ptr1 + (x5), tmp18, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ge/cgeibflzxp3re7kcuq7zwbnhqeymzvrmuddlen3bimagvhhn3gfr.py
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# conv2d_1 => convolution_1
# Graph fragment:
# %convolution_1 : [num_users=2] = call_function[target=torch.ops.aten.convolution.default](args = (%relu, %primals_4, %primals_5, [2, 2], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_2 = async_compile.triton('triton_poi_fused_convolution_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[65536],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_2', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_2(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 33800
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 169) % 50
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/sp/cspffc4gey5olor47qd5fg2diyed3c2tkeqb7ymxqm2nonzzmufl.py
# Topologically Sorted Source Nodes: [max_pool2d_1, x_1], Original ATen: [aten.max_pool2d_with_indices, aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# max_pool2d_1 => _low_memory_max_pool2d_with_offsets_1, getitem_3
# x_1 => relu_1
# Graph fragment:
# %_low_memory_max_pool2d_with_offsets_1 : [num_users=2] = call_function[target=torch.ops.prims._low_memory_max_pool2d_with_offsets.default](args = (%convolution_1, [2, 2], [2, 2], [0, 0], [1, 1], False), kwargs = {})
# %getitem_3 : [num_users=1] = call_function[target=operator.getitem](args = (%_low_memory_max_pool2d_with_offsets_1, 1), kwargs = {})
# %relu_1 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%getitem_2,), kwargs = {})
# %le_1 : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu_1, 0), kwargs = {})
triton_poi_fused_max_pool2d_with_indices_relu_threshold_backward_3 = async_compile.triton('triton_poi_fused_max_pool2d_with_indices_relu_threshold_backward_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8192],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i8', 2: '*fp32', 3: '*i1', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_max_pool2d_with_indices_relu_threshold_backward_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_max_pool2d_with_indices_relu_threshold_backward_3(in_ptr0, out_ptr0, out_ptr1, out_ptr2, xnumel, XBLOCK : tl.constexpr):
xnumel = 7200
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 6
x1 = (xindex // 6) % 6
x5 = (xindex // 36)
x3 = (xindex // 1800)
x4 = xindex % 1800
tmp0 = tl.load(in_ptr0 + ((2*x0) + (26*x1) + (169*x5)), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + (2*x0) + (26*x1) + (169*x5)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (13 + (2*x0) + (26*x1) + (169*x5)), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr0 + (14 + (2*x0) + (26*x1) + (169*x5)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 > tmp0
tmp3 = tl.full([1], 1, tl.int8)
tmp4 = tl.full([1], 0, tl.int8)
tmp5 = tl.where(tmp2, tmp3, tmp4)
tmp6 = triton_helpers.maximum(tmp1, tmp0)
tmp8 = tmp7 > tmp6
tmp9 = tl.full([1], 2, tl.int8)
tmp10 = tl.where(tmp8, tmp9, tmp5)
tmp11 = triton_helpers.maximum(tmp7, tmp6)
tmp13 = tmp12 > tmp11
tmp14 = tl.full([1], 3, tl.int8)
tmp15 = tl.where(tmp13, tmp14, tmp10)
tmp16 = triton_helpers.maximum(tmp12, tmp11)
tmp17 = tl.full([1], 0, tl.int32)
tmp18 = triton_helpers.maximum(tmp17, tmp16)
tmp19 = 0.0
tmp20 = tmp18 <= tmp19
tl.store(out_ptr0 + (x4 + (1920*x3)), tmp15, xmask)
tl.store(out_ptr1 + (x4 + (1824*x3)), tmp18, xmask)
tl.store(out_ptr2 + (x4 + (1920*x3)), tmp20, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ew/cew6u4uka7aoqfdk47cfa4ku6ujfpizha2puen2hsmbq7lpyto34.py
# Topologically Sorted Source Nodes: [max_pool2d_1, x_1, x_2], Original ATen: [aten.max_pool2d_with_indices, aten.relu, aten.view]
# Source node to ATen node mapping:
# max_pool2d_1 => _low_memory_max_pool2d_with_offsets_1
# x_1 => relu_1
# x_2 => view
# Graph fragment:
# %_low_memory_max_pool2d_with_offsets_1 : [num_users=2] = call_function[target=torch.ops.prims._low_memory_max_pool2d_with_offsets.default](args = (%convolution_1, [2, 2], [2, 2], [0, 0], [1, 1], False), kwargs = {})
# %relu_1 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%getitem_2,), kwargs = {})
# %view : [num_users=2] = call_function[target=torch.ops.aten.reshape.default](args = (%relu_1, [-1, 200]), kwargs = {})
triton_poi_fused_max_pool2d_with_indices_relu_view_4 = async_compile.triton('triton_poi_fused_max_pool2d_with_indices_relu_view_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8192],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_max_pool2d_with_indices_relu_view_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_max_pool2d_with_indices_relu_view_4(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 7200
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + ((1824*(x0 // 1800)) + (x0 % 1800)), xmask)
tl.store(out_ptr0 + (x0), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/7v/c7vocwkkblk6u27na653nnemx4hgdzzjqxtsnjomrz5tc757hk34.py
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.relu]
# Source node to ATen node mapping:
# x_3 => relu_2
# Graph fragment:
# %add_tensor : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mm_default, %primals_7), kwargs = {})
# %relu_2 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%add_tensor,), kwargs = {})
triton_poi_fused_relu_5 = async_compile.triton('triton_poi_fused_relu_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32768],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_5', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_5(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 18000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 500
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x2), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kf/ckfee3gq2oegm5e2jm5amkbvt2xob72z2qdyzobqqsdvop5jkwl6.py
# Topologically Sorted Source Nodes: [log_softmax], Original ATen: [aten._log_softmax]
# Source node to ATen node mapping:
# log_softmax => amax, exp, log, sub, sub_1, sum_1
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%addmm_1, [1], True), kwargs = {})
# %sub : [num_users=2] = call_function[target=torch.ops.aten.sub.Tensor](args = (%addmm_1, %amax), kwargs = {})
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [1], True), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%sum_1,), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sub, %log), kwargs = {})
triton_per_fused__log_softmax_6 = async_compile.triton('triton_per_fused__log_softmax_6', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[64, 16],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused__log_softmax_6', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 2, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused__log_softmax_6(in_ptr0, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 36
rnumel = 10
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = rindex < rnumel
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (10*x0)), rmask & xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(rmask & xmask, tmp1, float("-inf"))
tmp4 = triton_helpers.max2(tmp3, 1)[:, None]
tmp5 = tmp0 - tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK, RBLOCK])
tmp9 = tl.where(rmask & xmask, tmp7, 0)
tmp10 = tl.sum(tmp9, 1)[:, None]
tmp11 = tl_math.log(tmp10)
tmp12 = tmp5 - tmp11
tl.store(out_ptr2 + (r1 + (10*x0)), tmp12, rmask & xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9 = args
args.clear()
assert_size_stride(primals_1, (20, 1, 5, 5), (25, 25, 5, 1))
assert_size_stride(primals_2, (20, ), (1, ))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
assert_size_stride(primals_4, (50, 20, 5, 5), (500, 25, 5, 1))
assert_size_stride(primals_5, (50, ), (1, ))
assert_size_stride(primals_6, (500, 200), (200, 1))
assert_size_stride(primals_7, (500, ), (1, ))
assert_size_stride(primals_8, (10, 500), (500, 1))
assert_size_stride(primals_9, (10, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 20, 60, 60), (72000, 3600, 60, 1))
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_0.run(buf1, primals_2, 288000, grid=grid(288000), stream=stream0)
del primals_2
buf2 = empty_strided_cuda((4, 20, 30, 30), (18048, 900, 30, 1), torch.int8)
buf3 = empty_strided_cuda((4, 20, 30, 30), (18000, 900, 30, 1), torch.float32)
# Topologically Sorted Source Nodes: [max_pool2d, x], Original ATen: [aten.max_pool2d_with_indices, aten.relu]
triton_poi_fused_max_pool2d_with_indices_relu_1.run(buf1, buf2, buf3, 72000, grid=grid(72000), stream=stream0)
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(buf3, primals_4, stride=(2, 2), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 50, 13, 13), (8450, 169, 13, 1))
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
triton_poi_fused_convolution_2.run(buf5, primals_5, 33800, grid=grid(33800), stream=stream0)
del primals_5
buf6 = empty_strided_cuda((4, 50, 6, 6), (1920, 36, 6, 1), torch.int8)
buf7 = empty_strided_cuda((4, 50, 6, 6), (1824, 36, 6, 1), torch.float32)
buf15 = empty_strided_cuda((4, 50, 6, 6), (1920, 36, 6, 1), torch.bool)
# Topologically Sorted Source Nodes: [max_pool2d_1, x_1], Original ATen: [aten.max_pool2d_with_indices, aten.relu, aten.threshold_backward]
triton_poi_fused_max_pool2d_with_indices_relu_threshold_backward_3.run(buf5, buf6, buf7, buf15, 7200, grid=grid(7200), stream=stream0)
buf8 = empty_strided_cuda((36, 200), (200, 1), torch.float32)
# Topologically Sorted Source Nodes: [max_pool2d_1, x_1, x_2], Original ATen: [aten.max_pool2d_with_indices, aten.relu, aten.view]
triton_poi_fused_max_pool2d_with_indices_relu_view_4.run(buf7, buf8, 7200, grid=grid(7200), stream=stream0)
del buf7
buf9 = empty_strided_cuda((36, 500), (500, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(buf8, reinterpret_tensor(primals_6, (200, 500), (1, 200), 0), out=buf9)
buf10 = buf9; del buf9 # reuse
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.relu]
triton_poi_fused_relu_5.run(buf10, primals_7, 18000, grid=grid(18000), stream=stream0)
del primals_7
buf11 = empty_strided_cuda((36, 10), (10, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_4], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_9, buf10, reinterpret_tensor(primals_8, (500, 10), (1, 500), 0), alpha=1, beta=1, out=buf11)
del primals_9
buf14 = empty_strided_cuda((36, 10), (10, 1), torch.float32)
# Topologically Sorted Source Nodes: [log_softmax], Original ATen: [aten._log_softmax]
triton_per_fused__log_softmax_6.run(buf11, buf14, 36, 10, grid=grid(36), stream=stream0)
del buf11
return (buf14, primals_1, primals_3, primals_4, buf1, buf2, buf3, buf5, buf6, buf8, buf10, buf14, primals_8, primals_6, buf15, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((20, 1, 5, 5), (25, 25, 5, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((20, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 1, 64, 64), (4096, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((50, 20, 5, 5), (500, 25, 5, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((50, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((500, 200), (200, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((500, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((10, 500), (500, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((10, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn.functional as F
import torch.nn
import torch.utils.data
import torch.utils.tensorboard._pytorch_graph
import torch.onnx.symbolic_caffe2
class _TestNetStrided(torch.nn.Module):
def __init__(self):
super(_TestNetStrided, self).__init__()
self.conv1 = torch.nn.Conv2d(1, 20, kernel_size=5)
self.conv2 = torch.nn.Conv2d(20, 50, kernel_size=5, stride=(2, 2))
self.fc1 = torch.nn.Linear(200, 500)
self.fc2 = torch.nn.Linear(500, 10)
def forward(self, x):
x = F.relu(F.max_pool2d(self.conv1(x), 2))
x = F.relu(F.max_pool2d(self.conv2(x), 2))
x = x.view(-1, 200)
x = F.relu(self.fc1(x))
x = self.fc2(x)
return F.log_softmax(x, dim=1)
def get_inputs():
return [torch.rand([4, 1, 64, 64])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn
import torch.utils.data
import torch.utils.tensorboard._pytorch_graph
import torch.onnx.symbolic_caffe2
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 288000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 3600 % 20
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
@triton.jit
def triton_poi_fused_max_pool2d_with_indices_relu_1(in_ptr0, out_ptr0,
out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 72000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 30
x3 = xindex // 30
x2 = xindex // 18000
x4 = xindex % 18000
x5 = xindex
tmp0 = tl.load(in_ptr0 + (2 * x0 + 120 * x3), xmask, eviction_policy=
'evict_last')
tmp1 = tl.load(in_ptr0 + (1 + 2 * x0 + 120 * x3), xmask,
eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (60 + 2 * x0 + 120 * x3), xmask,
eviction_policy='evict_last')
tmp12 = tl.load(in_ptr0 + (61 + 2 * x0 + 120 * x3), xmask,
eviction_policy='evict_last')
tmp2 = tmp1 > tmp0
tmp3 = tl.full([1], 1, tl.int8)
tmp4 = tl.full([1], 0, tl.int8)
tmp5 = tl.where(tmp2, tmp3, tmp4)
tmp6 = triton_helpers.maximum(tmp1, tmp0)
tmp8 = tmp7 > tmp6
tmp9 = tl.full([1], 2, tl.int8)
tmp10 = tl.where(tmp8, tmp9, tmp5)
tmp11 = triton_helpers.maximum(tmp7, tmp6)
tmp13 = tmp12 > tmp11
tmp14 = tl.full([1], 3, tl.int8)
tmp15 = tl.where(tmp13, tmp14, tmp10)
tmp16 = triton_helpers.maximum(tmp12, tmp11)
tmp17 = tl.full([1], 0, tl.int32)
tmp18 = triton_helpers.maximum(tmp17, tmp16)
tl.store(out_ptr0 + (x4 + 18048 * x2), tmp15, xmask)
tl.store(out_ptr1 + x5, tmp18, xmask)
@triton.jit
def triton_poi_fused_convolution_2(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 33800
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 169 % 50
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
@triton.jit
def triton_poi_fused_max_pool2d_with_indices_relu_threshold_backward_3(in_ptr0,
out_ptr0, out_ptr1, out_ptr2, xnumel, XBLOCK: tl.constexpr):
xnumel = 7200
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 6
x1 = xindex // 6 % 6
x5 = xindex // 36
x3 = xindex // 1800
x4 = xindex % 1800
tmp0 = tl.load(in_ptr0 + (2 * x0 + 26 * x1 + 169 * x5), xmask,
eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + 2 * x0 + 26 * x1 + 169 * x5), xmask,
eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (13 + 2 * x0 + 26 * x1 + 169 * x5), xmask,
eviction_policy='evict_last')
tmp12 = tl.load(in_ptr0 + (14 + 2 * x0 + 26 * x1 + 169 * x5), xmask,
eviction_policy='evict_last')
tmp2 = tmp1 > tmp0
tmp3 = tl.full([1], 1, tl.int8)
tmp4 = tl.full([1], 0, tl.int8)
tmp5 = tl.where(tmp2, tmp3, tmp4)
tmp6 = triton_helpers.maximum(tmp1, tmp0)
tmp8 = tmp7 > tmp6
tmp9 = tl.full([1], 2, tl.int8)
tmp10 = tl.where(tmp8, tmp9, tmp5)
tmp11 = triton_helpers.maximum(tmp7, tmp6)
tmp13 = tmp12 > tmp11
tmp14 = tl.full([1], 3, tl.int8)
tmp15 = tl.where(tmp13, tmp14, tmp10)
tmp16 = triton_helpers.maximum(tmp12, tmp11)
tmp17 = tl.full([1], 0, tl.int32)
tmp18 = triton_helpers.maximum(tmp17, tmp16)
tmp19 = 0.0
tmp20 = tmp18 <= tmp19
tl.store(out_ptr0 + (x4 + 1920 * x3), tmp15, xmask)
tl.store(out_ptr1 + (x4 + 1824 * x3), tmp18, xmask)
tl.store(out_ptr2 + (x4 + 1920 * x3), tmp20, xmask)
@triton.jit
def triton_poi_fused_max_pool2d_with_indices_relu_view_4(in_ptr0, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xnumel = 7200
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (1824 * (x0 // 1800) + x0 % 1800), xmask)
tl.store(out_ptr0 + x0, tmp0, xmask)
@triton.jit
def triton_poi_fused_relu_5(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 18000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 500
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x2, tmp4, xmask)
@triton.jit
def triton_per_fused__log_softmax_6(in_ptr0, out_ptr2, xnumel, rnumel,
XBLOCK: tl.constexpr):
xnumel = 36
rnumel = 10
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
rmask = rindex < rnumel
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 10 * x0), rmask & xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(rmask & xmask, tmp1, float('-inf'))
tmp4 = triton_helpers.max2(tmp3, 1)[:, None]
tmp5 = tmp0 - tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK, RBLOCK])
tmp9 = tl.where(rmask & xmask, tmp7, 0)
tmp10 = tl.sum(tmp9, 1)[:, None]
tmp11 = tl_math.log(tmp10)
tmp12 = tmp5 - tmp11
tl.store(out_ptr2 + (r1 + 10 * x0), tmp12, rmask & xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9) = args
args.clear()
assert_size_stride(primals_1, (20, 1, 5, 5), (25, 25, 5, 1))
assert_size_stride(primals_2, (20,), (1,))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
assert_size_stride(primals_4, (50, 20, 5, 5), (500, 25, 5, 1))
assert_size_stride(primals_5, (50,), (1,))
assert_size_stride(primals_6, (500, 200), (200, 1))
assert_size_stride(primals_7, (500,), (1,))
assert_size_stride(primals_8, (10, 500), (500, 1))
assert_size_stride(primals_9, (10,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 20, 60, 60), (72000, 3600, 60, 1))
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_convolution_0[grid(288000)](buf1, primals_2,
288000, XBLOCK=512, num_warps=8, num_stages=1)
del primals_2
buf2 = empty_strided_cuda((4, 20, 30, 30), (18048, 900, 30, 1),
torch.int8)
buf3 = empty_strided_cuda((4, 20, 30, 30), (18000, 900, 30, 1),
torch.float32)
triton_poi_fused_max_pool2d_with_indices_relu_1[grid(72000)](buf1,
buf2, buf3, 72000, XBLOCK=512, num_warps=8, num_stages=1)
buf4 = extern_kernels.convolution(buf3, primals_4, stride=(2, 2),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 50, 13, 13), (8450, 169, 13, 1))
buf5 = buf4
del buf4
triton_poi_fused_convolution_2[grid(33800)](buf5, primals_5, 33800,
XBLOCK=512, num_warps=4, num_stages=1)
del primals_5
buf6 = empty_strided_cuda((4, 50, 6, 6), (1920, 36, 6, 1), torch.int8)
buf7 = empty_strided_cuda((4, 50, 6, 6), (1824, 36, 6, 1), torch.
float32)
buf15 = empty_strided_cuda((4, 50, 6, 6), (1920, 36, 6, 1), torch.bool)
triton_poi_fused_max_pool2d_with_indices_relu_threshold_backward_3[grid
(7200)](buf5, buf6, buf7, buf15, 7200, XBLOCK=256, num_warps=4,
num_stages=1)
buf8 = empty_strided_cuda((36, 200), (200, 1), torch.float32)
triton_poi_fused_max_pool2d_with_indices_relu_view_4[grid(7200)](buf7,
buf8, 7200, XBLOCK=256, num_warps=4, num_stages=1)
del buf7
buf9 = empty_strided_cuda((36, 500), (500, 1), torch.float32)
extern_kernels.mm(buf8, reinterpret_tensor(primals_6, (200, 500), (
1, 200), 0), out=buf9)
buf10 = buf9
del buf9
triton_poi_fused_relu_5[grid(18000)](buf10, primals_7, 18000,
XBLOCK=256, num_warps=4, num_stages=1)
del primals_7
buf11 = empty_strided_cuda((36, 10), (10, 1), torch.float32)
extern_kernels.addmm(primals_9, buf10, reinterpret_tensor(primals_8,
(500, 10), (1, 500), 0), alpha=1, beta=1, out=buf11)
del primals_9
buf14 = empty_strided_cuda((36, 10), (10, 1), torch.float32)
triton_per_fused__log_softmax_6[grid(36)](buf11, buf14, 36, 10,
XBLOCK=1, num_warps=2, num_stages=1)
del buf11
return (buf14, primals_1, primals_3, primals_4, buf1, buf2, buf3, buf5,
buf6, buf8, buf10, buf14, primals_8, primals_6, buf15)
class _TestNetStridedNew(torch.nn.Module):
def __init__(self):
super(_TestNetStridedNew, self).__init__()
self.conv1 = torch.nn.Conv2d(1, 20, kernel_size=5)
self.conv2 = torch.nn.Conv2d(20, 50, kernel_size=5, stride=(2, 2))
self.fc1 = torch.nn.Linear(200, 500)
self.fc2 = torch.nn.Linear(500, 10)
def forward(self, input_0):
primals_1 = self.conv1.weight
primals_2 = self.conv1.bias
primals_4 = self.conv2.weight
primals_5 = self.conv2.bias
primals_6 = self.fc1.weight
primals_7 = self.fc1.bias
primals_8 = self.fc2.weight
primals_9 = self.fc2.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9])
return output[0]
|
arjunsuresh/aimet
|
_TestNetStrided
| false | 12,325 |
[
"BSD-3-Clause"
] | 0 |
f6e09cb07a91eed3a5e6b8e19e6b065303af5a39
|
https://github.com/arjunsuresh/aimet/tree/f6e09cb07a91eed3a5e6b8e19e6b065303af5a39
|
Concat
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/m6/cm6nixhuxyf76p5x5wtskqwfrb5kdl2bfda4t34yltftpilyxauy.py
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cat => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%arg0_1, %arg1_1],), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[512],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 512
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 64)
x0 = xindex % 64
x2 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (64*x1)), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + (x0 + (64*((-4) + x1))), tmp6 & xmask, other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x2), tmp10, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((8, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(arg0_1, arg1_1, buf0, 512, grid=grid(512), stream=stream0)
del arg0_1
del arg1_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn
import torch.utils.data
import torch.utils.tensorboard._pytorch_graph
import torch.onnx.symbolic_caffe2
class Concat(torch.nn.Module):
""" Concat module for a functional concat"""
def __init__(self, axis: 'int'=0):
super(Concat, self).__init__()
self.axis = axis
def forward(self, x, y):
"""
Forward-pass routine for divide op
"""
return torch.cat((x, y), self.axis)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn
import torch.utils.data
import torch.utils.tensorboard._pytorch_graph
import torch.onnx.symbolic_caffe2
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 512
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 64
x0 = xindex % 64
x2 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 64 * x1), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (x0 + 64 * (-4 + x1)), tmp6 & xmask, other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x2, tmp10, xmask)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((8, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(512)](arg0_1, arg1_1, buf0, 512, XBLOCK
=256, num_warps=4, num_stages=1)
del arg0_1
del arg1_1
return buf0,
class ConcatNew(torch.nn.Module):
""" Concat module for a functional concat"""
def __init__(self, axis: 'int'=0):
super(ConcatNew, self).__init__()
self.axis = axis
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
arjunsuresh/aimet
|
Concat
| false | 12,326 |
[
"BSD-3-Clause"
] | 0 |
f6e09cb07a91eed3a5e6b8e19e6b065303af5a39
|
https://github.com/arjunsuresh/aimet/tree/f6e09cb07a91eed3a5e6b8e19e6b065303af5a39
|
EdgeCaseModel
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/lj/clj3mscf4tqz7wlbw6be5v4cyjekfwhnnfmmz6ngsoi4fvpbcy27.py
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# x => convolution
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_0 = async_compile.triton('triton_poi_fused_convolution_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[262144],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 144000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 3600) % 10
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (10, 1, 5, 5), (25, 25, 5, 1))
assert_size_stride(primals_2, (10, ), (1, ))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 10, 60, 60), (36000, 3600, 60, 1))
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.convolution]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_0.run(buf1, primals_2, 144000, grid=grid(144000), stream=stream0)
del primals_2
return (buf1, primals_1, primals_3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((10, 1, 5, 5), (25, 25, 5, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((10, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 1, 64, 64), (4096, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from typing import Any
import torch.nn as nn
class LayerWithRidiculouslyLongNameAndDoesntDoAnything(nn.Module):
""" Model with a very long name. """
def __init__(self) ->None:
super().__init__()
self.identity = nn.Identity()
def forward(self, x: 'Any') ->Any:
return self.identity(x)
class EdgeCaseModel(nn.Module):
""" Model that throws an exception when used. """
def __init__(self, throw_error: 'bool'=False, return_str: 'bool'=False,
return_class: 'bool'=False) ->None:
super().__init__()
self.throw_error = throw_error
self.return_str = return_str
self.return_class = return_class
self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
self.model = LayerWithRidiculouslyLongNameAndDoesntDoAnything()
def forward(self, x: 'torch.Tensor') ->torch.Tensor:
x = self.conv1(x)
x = self.model('string output' if self.return_str else x)
if self.throw_error:
x = self.conv1(x)
if self.return_class:
x = self.model(EdgeCaseModel)
return x
def get_inputs():
return [torch.rand([4, 1, 64, 64])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from typing import Any
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 144000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 3600 % 10
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (10, 1, 5, 5), (25, 25, 5, 1))
assert_size_stride(primals_2, (10,), (1,))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 10, 60, 60), (36000, 3600, 60, 1))
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_convolution_0[grid(144000)](buf1, primals_2,
144000, XBLOCK=1024, num_warps=4, num_stages=1)
del primals_2
return buf1, primals_1, primals_3
class LayerWithRidiculouslyLongNameAndDoesntDoAnything(nn.Module):
""" Model with a very long name. """
def __init__(self) ->None:
super().__init__()
self.identity = nn.Identity()
def forward(self, x: 'Any') ->Any:
return self.identity(x)
class EdgeCaseModelNew(nn.Module):
""" Model that throws an exception when used. """
def __init__(self, throw_error: 'bool'=False, return_str: 'bool'=False,
return_class: 'bool'=False) ->None:
super().__init__()
self.throw_error = throw_error
self.return_str = return_str
self.return_class = return_class
self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
self.model = LayerWithRidiculouslyLongNameAndDoesntDoAnything()
def forward(self, input_0):
primals_1 = self.conv1.weight
primals_2 = self.conv1.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
e-dorigatti/torchinfo
|
EdgeCaseModel
| false | 12,327 |
[
"MIT"
] | 0 |
9fa0e677fb7002e89afd5b1bb372fe8c1dd813d6
|
https://github.com/e-dorigatti/torchinfo/tree/9fa0e677fb7002e89afd5b1bb372fe8c1dd813d6
|
ComplexConvTranspose2d
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/yv/cyv6tiqnrs5v23ryzvbw2vt4qnufw6d22jee3c5qrlatmk4dks3w.py
# Topologically Sorted Source Nodes: [conv_transpose2d, conv_transpose2d_1, sub, conv_transpose2d_2, conv_transpose2d_3, add], Original ATen: [aten.convolution, aten.sub, aten.add]
# Source node to ATen node mapping:
# add => add
# conv_transpose2d => convolution
# conv_transpose2d_1 => convolution_1
# conv_transpose2d_2 => convolution_2
# conv_transpose2d_3 => convolution_3
# sub => sub
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [0, 0], [1, 1], True, [0, 0], 1), kwargs = {})
# %convolution_1 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_6, %primals_4, %primals_5, [1, 1], [0, 0], [1, 1], True, [0, 0], 1), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%convolution, %convolution_1), kwargs = {})
# %convolution_2 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_6, %primals_1, %primals_2, [1, 1], [0, 0], [1, 1], True, [0, 0], 1), kwargs = {})
# %convolution_3 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_4, %primals_5, [1, 1], [0, 0], [1, 1], True, [0, 0], 1), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%convolution_2, %convolution_3), kwargs = {})
triton_poi_fused_add_convolution_sub_0 = async_compile.triton('triton_poi_fused_add_convolution_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_convolution_sub_0', 'mutated_arg_names': ['in_out_ptr0', 'in_out_ptr1'], 'no_x_dim': False, 'num_load': 6, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_convolution_sub_0(in_out_ptr0, in_out_ptr1, in_ptr0, in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK : tl.constexpr):
xnumel = 784
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 49) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + (x3), xmask)
tmp4 = tl.load(in_ptr2 + (x1), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_out_ptr1 + (x3), xmask)
tmp9 = tl.load(in_ptr3 + (x3), xmask)
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 - tmp5
tmp8 = tmp7 + tmp1
tmp10 = tmp9 + tmp4
tmp11 = tmp8 + tmp10
tl.store(in_out_ptr0 + (x3), tmp6, xmask)
tl.store(in_out_ptr1 + (x3), tmp11, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv_transpose2d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=True, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 7, 7), (196, 49, 7, 1))
# Topologically Sorted Source Nodes: [conv_transpose2d_1], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(primals_6, primals_4, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=True, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 4, 7, 7), (196, 49, 7, 1))
# Topologically Sorted Source Nodes: [conv_transpose2d_2], Original ATen: [aten.convolution]
buf3 = extern_kernels.convolution(primals_6, primals_1, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=True, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 4, 7, 7), (196, 49, 7, 1))
# Topologically Sorted Source Nodes: [conv_transpose2d_3], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(primals_3, primals_4, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=True, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 7, 7), (196, 49, 7, 1))
buf2 = buf0; del buf0 # reuse
buf5 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [conv_transpose2d, conv_transpose2d_1, sub, conv_transpose2d_2, conv_transpose2d_3, add], Original ATen: [aten.convolution, aten.sub, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_convolution_sub_0.run(buf2, buf5, primals_2, buf1, primals_5, buf4, 784, grid=grid(784), stream=stream0)
del buf1
del buf4
del primals_2
del primals_5
return (buf2, buf5, primals_1, primals_3, primals_4, primals_6, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from torch.nn import Module
import torch
from torch.nn import ConvTranspose2d
class ComplexConvTranspose2d(Module):
def __init__(self, in_channels, out_channels, kernel_size, stride=1,
padding=0, output_padding=0, groups=1, bias=True, dilation=1,
padding_mode='zeros'):
super(ComplexConvTranspose2d, self).__init__()
self.conv_tran_r = ConvTranspose2d(in_channels, out_channels,
kernel_size, stride, padding, output_padding, groups, bias,
dilation, padding_mode)
self.conv_tran_i = ConvTranspose2d(in_channels, out_channels,
kernel_size, stride, padding, output_padding, groups, bias,
dilation, padding_mode)
def forward(self, input_r, input_i):
return self.conv_tran_r(input_r) - self.conv_tran_i(input_i
), self.conv_tran_r(input_i) + self.conv_tran_i(input_r)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_channels': 4, 'out_channels': 4, 'kernel_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch.nn import Module
from torch.nn import ConvTranspose2d
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
@triton.jit
def triton_poi_fused_add_convolution_sub_0(in_out_ptr0, in_out_ptr1,
in_ptr0, in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK: tl.constexpr):
xnumel = 784
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 49 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + x3, xmask)
tmp4 = tl.load(in_ptr2 + x1, xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_out_ptr1 + x3, xmask)
tmp9 = tl.load(in_ptr3 + x3, xmask)
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 - tmp5
tmp8 = tmp7 + tmp1
tmp10 = tmp9 + tmp4
tmp11 = tmp8 + tmp10
tl.store(in_out_ptr0 + x3, tmp6, xmask)
tl.store(in_out_ptr1 + x3, tmp11, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=True,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 7, 7), (196, 49, 7, 1))
buf1 = extern_kernels.convolution(primals_6, primals_4, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=True,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 4, 7, 7), (196, 49, 7, 1))
buf3 = extern_kernels.convolution(primals_6, primals_1, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=True,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 4, 7, 7), (196, 49, 7, 1))
buf4 = extern_kernels.convolution(primals_3, primals_4, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=True,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 7, 7), (196, 49, 7, 1))
buf2 = buf0
del buf0
buf5 = buf3
del buf3
get_raw_stream(0)
triton_poi_fused_add_convolution_sub_0[grid(784)](buf2, buf5,
primals_2, buf1, primals_5, buf4, 784, XBLOCK=256, num_warps=4,
num_stages=1)
del buf1
del buf4
del primals_2
del primals_5
return buf2, buf5, primals_1, primals_3, primals_4, primals_6
class ComplexConvTranspose2dNew(Module):
def __init__(self, in_channels, out_channels, kernel_size, stride=1,
padding=0, output_padding=0, groups=1, bias=True, dilation=1,
padding_mode='zeros'):
super(ComplexConvTranspose2dNew, self).__init__()
self.conv_tran_r = ConvTranspose2d(in_channels, out_channels,
kernel_size, stride, padding, output_padding, groups, bias,
dilation, padding_mode)
self.conv_tran_i = ConvTranspose2d(in_channels, out_channels,
kernel_size, stride, padding, output_padding, groups, bias,
dilation, padding_mode)
def forward(self, input_0, input_1):
primals_1 = self.conv_tran_r.weight
primals_2 = self.conv_tran_r.bias
primals_3 = self.conv_tran_i.weight
primals_5 = self.conv_tran_i.bias
primals_4 = input_0
primals_6 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6])
return output[0], output[1]
|
drydenwiebe/complexPyTorch
|
ComplexConvTranspose2d
| false | 12,328 |
[
"MIT"
] | 0 |
cea88ba7ee5692dfa1b40f0ba609ef14160d5073
|
https://github.com/drydenwiebe/complexPyTorch/tree/cea88ba7ee5692dfa1b40f0ba609ef14160d5073
|
OuterProductLayer
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/p6/cp66pqi5yaj5elbfua3qtyhcywwcmpiswdj7u3jzry2i62aro25j.py
# Topologically Sorted Source Nodes: [p_1, p_2], Original ATen: [aten.mul, aten.sum]
# Source node to ATen node mapping:
# p_1 => mul
# p_2 => sum_1
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%unsqueeze, %unsqueeze_1), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul, [-1]), kwargs = {})
triton_poi_fused_mul_sum_0 = async_compile.triton('triton_poi_fused_mul_sum_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_sum_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_sum_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 96
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 6
x2 = (xindex // 24)
x3 = xindex % 24
x4 = xindex
tmp18 = tl.load(in_ptr1 + (4*x3), xmask, eviction_policy='evict_last')
tmp21 = tl.load(in_ptr1 + (1 + (4*x3)), xmask, eviction_policy='evict_last')
tmp25 = tl.load(in_ptr1 + (2 + (4*x3)), xmask, eviction_policy='evict_last')
tmp29 = tl.load(in_ptr1 + (3 + (4*x3)), xmask, eviction_policy='evict_last')
tmp0 = x0
tmp1 = tl.full([1], 3, tl.int64)
tmp2 = tmp0 < tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.full([1], 2, tl.int64)
tmp6 = tmp0 < tmp5
tmp7 = tl.full([1], 0, tl.int64)
tmp8 = tl.where(tmp6, tmp7, tmp7)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tl.full([1], 4, tl.int64)
tmp11 = tmp0 < tmp10
tmp12 = tl.full([1], 5, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tl.where(tmp13, tmp3, tmp5)
tmp15 = tl.where(tmp11, tmp3, tmp14)
tmp16 = tl.where(tmp2, tmp9, tmp15)
tmp17 = tl.load(in_ptr0 + ((4*tmp16) + (16*x2)), xmask, eviction_policy='evict_last')
tmp19 = tmp17 * tmp18
tmp20 = tl.load(in_ptr0 + (1 + (4*tmp16) + (16*x2)), xmask, eviction_policy='evict_last')
tmp22 = tmp20 * tmp21
tmp23 = tmp19 + tmp22
tmp24 = tl.load(in_ptr0 + (2 + (4*tmp16) + (16*x2)), xmask, eviction_policy='evict_last')
tmp26 = tmp24 * tmp25
tmp27 = tmp23 + tmp26
tmp28 = tl.load(in_ptr0 + (3 + (4*tmp16) + (16*x2)), xmask, eviction_policy='evict_last')
tmp30 = tmp28 * tmp29
tmp31 = tmp27 + tmp30
tl.store(out_ptr0 + (x4), tmp31, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6a/c6asanugzoupldkrdkisv2ufakttxezcrdo7m3cqduqfcbra4cbj.py
# Topologically Sorted Source Nodes: [q, outer_product, sum_2], Original ATen: [aten.index, aten.mul, aten.sum]
# Source node to ATen node mapping:
# outer_product => mul_1
# q => index_1
# sum_2 => sum_2
# Graph fragment:
# %index_1 : [num_users=1] = call_function[target=torch.ops.aten.index.Tensor](args = (%primals_1, [None, %lift_fresh_copy_1]), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%permute, %index_1), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul_1, [-1]), kwargs = {})
triton_poi_fused_index_mul_sum_1 = async_compile.triton('triton_poi_fused_index_mul_sum_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_index_mul_sum_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_index_mul_sum_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 24
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 6
x1 = (xindex // 6)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (24*x1)), xmask)
tmp19 = tl.load(in_ptr0 + (6 + x0 + (24*x1)), xmask)
tmp23 = tl.load(in_ptr0 + (12 + x0 + (24*x1)), xmask)
tmp27 = tl.load(in_ptr0 + (18 + x0 + (24*x1)), xmask)
tmp1 = x0
tmp2 = tl.full([1], 3, tl.int64)
tmp3 = tmp1 < tmp2
tmp4 = tl.full([1], 1, tl.int64)
tmp5 = tmp1 < tmp4
tmp6 = tl.full([1], 2, tl.int64)
tmp7 = tmp1 < tmp6
tmp8 = tl.where(tmp7, tmp6, tmp2)
tmp9 = tl.where(tmp5, tmp4, tmp8)
tmp10 = tl.full([1], 4, tl.int64)
tmp11 = tmp1 < tmp10
tmp12 = tl.full([1], 5, tl.int64)
tmp13 = tmp1 < tmp12
tmp14 = tl.where(tmp13, tmp2, tmp2)
tmp15 = tl.where(tmp11, tmp6, tmp14)
tmp16 = tl.where(tmp3, tmp9, tmp15)
tmp17 = tl.load(in_ptr1 + ((4*tmp16) + (16*x1)), xmask, eviction_policy='evict_last')
tmp18 = tmp0 * tmp17
tmp20 = tl.load(in_ptr1 + (1 + (4*tmp16) + (16*x1)), xmask, eviction_policy='evict_last')
tmp21 = tmp19 * tmp20
tmp22 = tmp18 + tmp21
tmp24 = tl.load(in_ptr1 + (2 + (4*tmp16) + (16*x1)), xmask, eviction_policy='evict_last')
tmp25 = tmp23 * tmp24
tmp26 = tmp22 + tmp25
tmp28 = tl.load(in_ptr1 + (3 + (4*tmp16) + (16*x1)), xmask, eviction_policy='evict_last')
tmp29 = tmp27 * tmp28
tmp30 = tmp26 + tmp29
tl.store(out_ptr0 + (x2), tmp30, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 6, 4), (24, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 6), (24, 6, 1), torch.float32)
# Topologically Sorted Source Nodes: [p_1, p_2], Original ATen: [aten.mul, aten.sum]
stream0 = get_raw_stream(0)
triton_poi_fused_mul_sum_0.run(primals_1, primals_2, buf0, 96, grid=grid(96), stream=stream0)
del primals_2
buf1 = empty_strided_cuda((4, 6), (6, 1), torch.float32)
# Topologically Sorted Source Nodes: [q, outer_product, sum_2], Original ATen: [aten.index, aten.mul, aten.sum]
triton_poi_fused_index_mul_sum_1.run(buf0, primals_1, buf1, 24, grid=grid(24), stream=stream0)
del buf0
return (buf1, primals_1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 6, 4), (24, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class OuterProductLayer(nn.Module):
"""OuterProduct Layer used in PNN. This implementation is
adapted from code that the author of the paper published on https://github.com/Atomu2014/product-nets.
"""
def __init__(self, num_feature_field, embedding_size, device):
"""
Args:
num_feature_field(int) :number of feature fields.
embedding_size(int) :number of embedding size.
device(torch.device) : device object of the model.
"""
super(OuterProductLayer, self).__init__()
self.num_feature_field = num_feature_field
num_pairs = int(num_feature_field * (num_feature_field - 1) / 2)
embed_size = embedding_size
self.kernel = nn.Parameter(torch.rand(embed_size, num_pairs,
embed_size), requires_grad=True)
nn.init.xavier_uniform_(self.kernel)
self
def forward(self, feat_emb):
"""
Args:
feat_emb(torch.FloatTensor) :3D tensor with shape: [batch_size,num_pairs,embedding_size].
Returns:
outer_product(torch.FloatTensor): The outer product of input tensor. shape of [batch_size, num_pairs]
"""
row = []
col = []
for i in range(self.num_feature_field - 1):
for j in range(i + 1, self.num_feature_field):
row.append(i)
col.append(j)
p = feat_emb[:, row]
q = feat_emb[:, col]
p.unsqueeze_(dim=1)
p = torch.mul(p, self.kernel.unsqueeze(0))
p = torch.sum(p, dim=-1)
p = torch.transpose(p, 2, 1)
outer_product = p * q
return outer_product.sum(dim=-1)
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'num_feature_field': 4, 'embedding_size': 4, 'device': 0}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_mul_sum_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 96
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 6
x2 = xindex // 24
x3 = xindex % 24
x4 = xindex
tmp18 = tl.load(in_ptr1 + 4 * x3, xmask, eviction_policy='evict_last')
tmp21 = tl.load(in_ptr1 + (1 + 4 * x3), xmask, eviction_policy='evict_last'
)
tmp25 = tl.load(in_ptr1 + (2 + 4 * x3), xmask, eviction_policy='evict_last'
)
tmp29 = tl.load(in_ptr1 + (3 + 4 * x3), xmask, eviction_policy='evict_last'
)
tmp0 = x0
tmp1 = tl.full([1], 3, tl.int64)
tmp2 = tmp0 < tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.full([1], 2, tl.int64)
tmp6 = tmp0 < tmp5
tmp7 = tl.full([1], 0, tl.int64)
tmp8 = tl.where(tmp6, tmp7, tmp7)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tl.full([1], 4, tl.int64)
tmp11 = tmp0 < tmp10
tmp12 = tl.full([1], 5, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tl.where(tmp13, tmp3, tmp5)
tmp15 = tl.where(tmp11, tmp3, tmp14)
tmp16 = tl.where(tmp2, tmp9, tmp15)
tmp17 = tl.load(in_ptr0 + (4 * tmp16 + 16 * x2), xmask, eviction_policy
='evict_last')
tmp19 = tmp17 * tmp18
tmp20 = tl.load(in_ptr0 + (1 + 4 * tmp16 + 16 * x2), xmask,
eviction_policy='evict_last')
tmp22 = tmp20 * tmp21
tmp23 = tmp19 + tmp22
tmp24 = tl.load(in_ptr0 + (2 + 4 * tmp16 + 16 * x2), xmask,
eviction_policy='evict_last')
tmp26 = tmp24 * tmp25
tmp27 = tmp23 + tmp26
tmp28 = tl.load(in_ptr0 + (3 + 4 * tmp16 + 16 * x2), xmask,
eviction_policy='evict_last')
tmp30 = tmp28 * tmp29
tmp31 = tmp27 + tmp30
tl.store(out_ptr0 + x4, tmp31, xmask)
@triton.jit
def triton_poi_fused_index_mul_sum_1(in_ptr0, in_ptr1, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 24
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 6
x1 = xindex // 6
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 24 * x1), xmask)
tmp19 = tl.load(in_ptr0 + (6 + x0 + 24 * x1), xmask)
tmp23 = tl.load(in_ptr0 + (12 + x0 + 24 * x1), xmask)
tmp27 = tl.load(in_ptr0 + (18 + x0 + 24 * x1), xmask)
tmp1 = x0
tmp2 = tl.full([1], 3, tl.int64)
tmp3 = tmp1 < tmp2
tmp4 = tl.full([1], 1, tl.int64)
tmp5 = tmp1 < tmp4
tmp6 = tl.full([1], 2, tl.int64)
tmp7 = tmp1 < tmp6
tmp8 = tl.where(tmp7, tmp6, tmp2)
tmp9 = tl.where(tmp5, tmp4, tmp8)
tmp10 = tl.full([1], 4, tl.int64)
tmp11 = tmp1 < tmp10
tmp12 = tl.full([1], 5, tl.int64)
tmp13 = tmp1 < tmp12
tmp14 = tl.where(tmp13, tmp2, tmp2)
tmp15 = tl.where(tmp11, tmp6, tmp14)
tmp16 = tl.where(tmp3, tmp9, tmp15)
tmp17 = tl.load(in_ptr1 + (4 * tmp16 + 16 * x1), xmask, eviction_policy
='evict_last')
tmp18 = tmp0 * tmp17
tmp20 = tl.load(in_ptr1 + (1 + 4 * tmp16 + 16 * x1), xmask,
eviction_policy='evict_last')
tmp21 = tmp19 * tmp20
tmp22 = tmp18 + tmp21
tmp24 = tl.load(in_ptr1 + (2 + 4 * tmp16 + 16 * x1), xmask,
eviction_policy='evict_last')
tmp25 = tmp23 * tmp24
tmp26 = tmp22 + tmp25
tmp28 = tl.load(in_ptr1 + (3 + 4 * tmp16 + 16 * x1), xmask,
eviction_policy='evict_last')
tmp29 = tmp27 * tmp28
tmp30 = tmp26 + tmp29
tl.store(out_ptr0 + x2, tmp30, xmask)
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 6, 4), (24, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 6), (24, 6, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_mul_sum_0[grid(96)](primals_1, primals_2, buf0, 96,
XBLOCK=128, num_warps=4, num_stages=1)
del primals_2
buf1 = empty_strided_cuda((4, 6), (6, 1), torch.float32)
triton_poi_fused_index_mul_sum_1[grid(24)](buf0, primals_1, buf1,
24, XBLOCK=32, num_warps=1, num_stages=1)
del buf0
return buf1, primals_1
class OuterProductLayerNew(nn.Module):
"""OuterProduct Layer used in PNN. This implementation is
adapted from code that the author of the paper published on https://github.com/Atomu2014/product-nets.
"""
def __init__(self, num_feature_field, embedding_size, device):
"""
Args:
num_feature_field(int) :number of feature fields.
embedding_size(int) :number of embedding size.
device(torch.device) : device object of the model.
"""
super(OuterProductLayerNew, self).__init__()
self.num_feature_field = num_feature_field
num_pairs = int(num_feature_field * (num_feature_field - 1) / 2)
embed_size = embedding_size
self.kernel = nn.Parameter(torch.rand(embed_size, num_pairs,
embed_size), requires_grad=True)
nn.init.xavier_uniform_(self.kernel)
self
def forward(self, input_0):
primals_2 = self.kernel
primals_1 = input_0
output = call([primals_1, primals_2])
return output[0]
|
dreaming-qin/RecBole
|
OuterProductLayer
| false | 12,329 |
[
"MIT"
] | 0 |
d6de39521484ded60c387ca604abaf86310acdbe
|
https://github.com/dreaming-qin/RecBole/tree/d6de39521484ded60c387ca604abaf86310acdbe
|
Net
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ky/ckydmp6lolzbvnwispw5t2yckxjd3gyyjcoho6yi2npgo3vxvg3x.py
# Topologically Sorted Source Nodes: [out_1], Original ATen: [aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# out_1 => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_1,), kwargs = {})
# %le_1 : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_relu_threshold_backward_0 = async_compile.triton('triton_poi_fused_relu_threshold_backward_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*i1', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_threshold_backward_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_ptr0, in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 16320
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x0 = xindex % 255
x2 = (xindex // 4080)
x3 = xindex % 4080
tmp0 = tl.load(in_ptr0 + (x4), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x3 + (4096*x2)), tmp4, xmask)
tl.store(out_ptr1 + (x3 + (4096*x2)), tmp6, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/iq/ciqh5ggtyjeepj6foripuzjpqlyfqcua26cpgf6bome5fqvhzm7l.py
# Topologically Sorted Source Nodes: [out_1, out_2], Original ATen: [aten.relu, aten.view]
# Source node to ATen node mapping:
# out_1 => relu
# out_2 => view_2
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_1,), kwargs = {})
# %view_2 : [num_users=2] = call_function[target=torch.ops.aten.reshape.default](args = (%relu, [64, 255]), kwargs = {})
triton_poi_fused_relu_view_1 = async_compile.triton('triton_poi_fused_relu_view_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_view_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_view_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16320
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 255
x1 = (xindex // 255)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (255*(x1 % 16)) + (4096*(x1 // 16))), xmask)
tl.store(out_ptr0 + (x2), tmp0, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7 = args
args.clear()
assert_size_stride(primals_1, (255, 4), (4, 1))
assert_size_stride(primals_2, (255, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (255, 255), (255, 1))
assert_size_stride(primals_5, (255, ), (1, ))
assert_size_stride(primals_6, (4, 255), (255, 1))
assert_size_stride(primals_7, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 255), (255, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 255), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((4, 4, 4, 255), (4096, 1020, 255, 1), torch.float32)
buf8 = empty_strided_cuda((4, 4, 4, 255), (4096, 1020, 255, 1), torch.bool)
# Topologically Sorted Source Nodes: [out_1], Original ATen: [aten.relu, aten.threshold_backward]
stream0 = get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0.run(buf0, primals_2, buf1, buf8, 16320, grid=grid(16320), stream=stream0)
del primals_2
buf2 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [out_1, out_2], Original ATen: [aten.relu, aten.view]
triton_poi_fused_relu_view_1.run(buf1, buf2, 16320, grid=grid(16320), stream=stream0)
buf3 = empty_strided_cuda((64, 255), (255, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(buf2, reinterpret_tensor(primals_4, (255, 255), (1, 255), 0), out=buf3)
buf4 = buf1; del buf1 # reuse
buf7 = empty_strided_cuda((4, 4, 4, 255), (4096, 1020, 255, 1), torch.bool)
# Topologically Sorted Source Nodes: [out_3], Original ATen: [aten.relu, aten.threshold_backward]
triton_poi_fused_relu_threshold_backward_0.run(buf3, primals_5, buf4, buf7, 16320, grid=grid(16320), stream=stream0)
del primals_5
buf5 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [out_3, out_4], Original ATen: [aten.relu, aten.view]
triton_poi_fused_relu_view_1.run(buf4, buf5, 16320, grid=grid(16320), stream=stream0)
del buf4
buf6 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [out_4], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_7, buf5, reinterpret_tensor(primals_6, (255, 4), (1, 255), 0), alpha=1, beta=1, out=buf6)
del primals_7
return (reinterpret_tensor(buf6, (4, 4, 4, 4), (64, 16, 4, 1), 0), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), buf2, buf5, primals_6, buf7, primals_4, buf8, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((255, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((255, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((255, 255), (255, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((255, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 255), (255, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class Net(nn.Module):
def __init__(self, input_placeholder, output_size):
super(Net, self).__init__()
self.fc1 = nn.Linear(input_placeholder, 255)
self.relu1 = nn.ReLU()
self.fc2 = nn.Linear(255, 255)
self.relu2 = nn.ReLU()
self.fc3 = nn.Linear(255, output_size)
def forward(self, x):
out = self.fc1(x)
out = self.relu1(out)
out = self.fc2(out)
out = self.relu2(out)
out = self.fc3(out)
return out
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'input_placeholder': 4, 'output_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_ptr0, in_ptr1, out_ptr0,
out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 16320
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x0 = xindex % 255
x2 = xindex // 4080
x3 = xindex % 4080
tmp0 = tl.load(in_ptr0 + x4, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x3 + 4096 * x2), tmp4, xmask)
tl.store(out_ptr1 + (x3 + 4096 * x2), tmp6, xmask)
@triton.jit
def triton_poi_fused_relu_view_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 16320
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 255
x1 = xindex // 255
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 255 * (x1 % 16) + 4096 * (x1 // 16)), xmask)
tl.store(out_ptr0 + x2, tmp0, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7) = args
args.clear()
assert_size_stride(primals_1, (255, 4), (4, 1))
assert_size_stride(primals_2, (255,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (255, 255), (255, 1))
assert_size_stride(primals_5, (255,), (1,))
assert_size_stride(primals_6, (4, 255), (255, 1))
assert_size_stride(primals_7, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 255), (255, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 255), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((4, 4, 4, 255), (4096, 1020, 255, 1),
torch.float32)
buf8 = empty_strided_cuda((4, 4, 4, 255), (4096, 1020, 255, 1),
torch.bool)
get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0[grid(16320)](buf0,
primals_2, buf1, buf8, 16320, XBLOCK=256, num_warps=4, num_stages=1
)
del primals_2
buf2 = buf0
del buf0
triton_poi_fused_relu_view_1[grid(16320)](buf1, buf2, 16320, XBLOCK
=256, num_warps=4, num_stages=1)
buf3 = empty_strided_cuda((64, 255), (255, 1), torch.float32)
extern_kernels.mm(buf2, reinterpret_tensor(primals_4, (255, 255), (
1, 255), 0), out=buf3)
buf4 = buf1
del buf1
buf7 = empty_strided_cuda((4, 4, 4, 255), (4096, 1020, 255, 1),
torch.bool)
triton_poi_fused_relu_threshold_backward_0[grid(16320)](buf3,
primals_5, buf4, buf7, 16320, XBLOCK=256, num_warps=4, num_stages=1
)
del primals_5
buf5 = buf3
del buf3
triton_poi_fused_relu_view_1[grid(16320)](buf4, buf5, 16320, XBLOCK
=256, num_warps=4, num_stages=1)
del buf4
buf6 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_7, buf5, reinterpret_tensor(primals_6,
(255, 4), (1, 255), 0), alpha=1, beta=1, out=buf6)
del primals_7
return reinterpret_tensor(buf6, (4, 4, 4, 4), (64, 16, 4, 1), 0
), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0
), buf2, buf5, primals_6, buf7, primals_4, buf8
class NetNew(nn.Module):
def __init__(self, input_placeholder, output_size):
super(NetNew, self).__init__()
self.fc1 = nn.Linear(input_placeholder, 255)
self.relu1 = nn.ReLU()
self.fc2 = nn.Linear(255, 255)
self.relu2 = nn.ReLU()
self.fc3 = nn.Linear(255, output_size)
def forward(self, input_0):
primals_1 = self.fc1.weight
primals_2 = self.fc1.bias
primals_4 = self.fc2.weight
primals_5 = self.fc2.bias
primals_6 = self.fc3.weight
primals_7 = self.fc3.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7])
return output[0]
|
dylan-albertazzi/Berkely_DeepRL
|
Net
| false | 12,330 |
[
"MIT"
] | 0 |
997d066df7b429f6ad365dca8105490dae8f978e
|
https://github.com/dylan-albertazzi/Berkely_DeepRL/tree/997d066df7b429f6ad365dca8105490dae8f978e
|
MPNetAttention
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/x2/cx2hdvwyo7m5jvhhvtugzxqvmy6z4nsfhkkjhvgzbbm3cb6dsum2.py
# Topologically Sorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
# Graph fragment:
# %mul_scalar : [num_users=1] = call_function[target=torch.ops.aten.mul.Scalar](args = (%permute_default, 1.0), kwargs = {})
# %clone_default : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%expand_default,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_0 = async_compile.triton('triton_poi_fused_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 4], tile_hint=TileHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_0(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = (yindex // 4)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x2) + (16*y1)), xmask & ymask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (y0), ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = 1.0
tmp4 = tmp2 * tmp3
tl.store(out_ptr0 + (x2 + (4*y3)), tmp4, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/5j/c5jll3kxtd32cl7pwubrb5oky2mtzckfgip2xbwad7crvvp4zk4r.py
# Topologically Sorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
# Graph fragment:
# %amax_default : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%view_default_2, [-1], True), kwargs = {})
# %sub_tensor : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view_default_2, %amax_default), kwargs = {})
# %exp_default : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_tensor,), kwargs = {})
triton_poi_fused_1 = async_compile.triton('triton_poi_fused_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tmp9 = tl_math.exp(tmp8)
tl.store(out_ptr0 + (x2), tmp9, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kt/cktnex5febczl2ac6zugjmcksgsd5kjdufazv65vtepuwob3cb7a.py
# Topologically Sorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
# Graph fragment:
# %sum_dim_int_list : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_default, [-1], True), kwargs = {})
# %div_tensor : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp_default, %sum_dim_int_list), kwargs = {})
# %eq_scalar : [num_users=1] = call_function[target=torch.ops.aten.eq.Scalar](args = (%view_default_2, -inf), kwargs = {})
# %logical_not_default : [num_users=1] = call_function[target=torch.ops.aten.logical_not.default](args = (%eq_scalar,), kwargs = {})
# %any_dim : [num_users=1] = call_function[target=torch.ops.aten.any.dim](args = (%logical_not_default, -1, True), kwargs = {})
# %logical_not_default_1 : [num_users=1] = call_function[target=torch.ops.aten.logical_not.default](args = (%any_dim,), kwargs = {})
# %full_default : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([4, 4, 4, 4], 0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %where_self : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%logical_not_default_1, %full_default, %div_tensor), kwargs = {})
triton_poi_fused_2 = async_compile.triton('triton_poi_fused_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 9, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp18 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp25 = tl.load(in_ptr1 + (x2), xmask)
tmp26 = tl.load(in_ptr1 + (4*x1), xmask, eviction_policy='evict_last')
tmp27 = tl.load(in_ptr1 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp29 = tl.load(in_ptr1 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp31 = tl.load(in_ptr1 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp1 = float("-inf")
tmp2 = tmp0 == tmp1
tmp3 = tmp2 == 0
tmp4 = tmp3.to(tl.int64)
tmp5 = (tmp4 != 0)
tmp7 = tmp6 == tmp1
tmp8 = tmp7 == 0
tmp9 = tmp8.to(tl.int64)
tmp10 = (tmp9 != 0)
tmp11 = tmp5 | tmp10
tmp13 = tmp12 == tmp1
tmp14 = tmp13 == 0
tmp15 = tmp14.to(tl.int64)
tmp16 = (tmp15 != 0)
tmp17 = tmp11 | tmp16
tmp19 = tmp18 == tmp1
tmp20 = tmp19 == 0
tmp21 = tmp20.to(tl.int64)
tmp22 = (tmp21 != 0)
tmp23 = tmp17 | tmp22
tmp24 = tmp23 == 0
tmp28 = tmp26 + tmp27
tmp30 = tmp28 + tmp29
tmp32 = tmp30 + tmp31
tmp33 = tmp25 / tmp32
tmp34 = 0.0
tmp35 = tl.where(tmp24, tmp34, tmp33)
tl.store(out_ptr0 + (x2), tmp35, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/vv/cvvnhithjvmvhfjufxwwzclfobkrgbyyteg66hp24r675f7elw4c.py
# Topologically Sorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
# Graph fragment:
# %clone_default_2 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%expand_default_3,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_3 = async_compile.triton('triton_poi_fused_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 4], tile_hint=TileHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_3(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = (yindex // 4)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x2) + (16*y1)), xmask & ymask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (y0), ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (x2 + (4*y3)), tmp2, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6t/c6t5a5ere3lqjiu7zh3uu4oxmpdoujdaqqmeunxqapgzo4m74uav.py
# Topologically Sorted Source Nodes: [c_1], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# c_1 => clone_4
# Graph fragment:
# %clone_4 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_7,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_4 = async_compile.triton('triton_poi_fused_clone_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 4], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_4(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = (yindex // 4)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x2) + (16*y1)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + (4*y3)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6m/c6mhj5zwirfhy5e4o45uaeov72uwfby4udubpm2fcz42iqvs2g57.py
# Topologically Sorted Source Nodes: [add, attention_output], Original ATen: [aten.add, aten.native_layer_norm]
# Source node to ATen node mapping:
# add => add
# attention_output => var_mean
# Graph fragment:
# %add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%view_17, %primals_3), kwargs = {})
# %var_mean : [num_users=2] = call_function[target=torch.ops.aten.var_mean.correction](args = (%add, [2]), kwargs = {correction: 0, keepdim: True})
triton_poi_fused_add_native_layer_norm_5 = async_compile.triton('triton_poi_fused_add_native_layer_norm_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_native_layer_norm_5', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_native_layer_norm_5(in_ptr0, in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (4*x0), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (4*x0), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr1 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 + tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 + tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp2 - tmp16
tmp18 = tmp17 * tmp17
tmp19 = tmp5 - tmp16
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp22 = tmp9 - tmp16
tmp23 = tmp22 * tmp22
tmp24 = tmp21 + tmp23
tmp25 = tmp13 - tmp16
tmp26 = tmp25 * tmp25
tmp27 = tmp24 + tmp26
tmp28 = tmp27 / tmp15
tl.store(out_ptr0 + (x0), tmp16, xmask)
tl.store(out_ptr1 + (x0), tmp28, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/l3/cl3bnd5pv2p4ydfmlj74bv4mbiwr2ntrdvbubnjubetyhosmxag6.py
# Topologically Sorted Source Nodes: [add, attention_output], Original ATen: [aten.add, aten.native_layer_norm]
# Source node to ATen node mapping:
# add => add
# attention_output => add_1, add_2, mul, mul_1, rsqrt, sub_1
# Graph fragment:
# %add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%view_17, %primals_3), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%getitem, 1.0), kwargs = {})
# %rsqrt : [num_users=1] = call_function[target=torch.ops.aten.rsqrt.default](args = (%add_1,), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add, %getitem_1), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_1, %rsqrt), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%mul, %primals_10), kwargs = {})
# %add_2 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul_1, %primals_11), kwargs = {})
triton_poi_fused_add_native_layer_norm_6 = async_compile.triton('triton_poi_fused_add_native_layer_norm_6', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: '*fp32', 7: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_native_layer_norm_6', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 6, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_native_layer_norm_6(in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp3 = tl.load(in_ptr2 + (x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr3 + (x1), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr4 + (x0), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr5 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp4 = tmp2 - tmp3
tmp6 = 1.0
tmp7 = tmp5 + tmp6
tmp8 = libdevice.rsqrt(tmp7)
tmp9 = tmp4 * tmp8
tmp11 = tmp9 * tmp10
tmp13 = tmp11 + tmp12
tl.store(out_ptr0 + (x2), tmp13, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4, ), (1, ))
assert_size_stride(primals_8, (4, 4), (4, 1))
assert_size_stride(primals_9, (4, ), (1, ))
assert_size_stride(primals_10, (4, ), (1, ))
assert_size_stride(primals_11, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf2)
del primals_6
buf3 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
stream0 = get_raw_stream(0)
triton_poi_fused_0.run(buf0, primals_2, buf3, 16, 4, grid=grid(16, 4), stream=stream0)
del primals_2
buf4 = reinterpret_tensor(buf0, (4, 4, 1, 4), (16, 4, 4, 1), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
triton_poi_fused_0.run(buf1, primals_5, buf4, 16, 4, grid=grid(16, 4), stream=stream0)
del primals_5
buf5 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 0), 0), reinterpret_tensor(buf4, (16, 1, 4), (4, 0, 1), 0), out=buf5)
buf6 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
triton_poi_fused_1.run(buf5, buf6, 256, grid=grid(256), stream=stream0)
buf7 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
triton_poi_fused_2.run(buf5, buf6, buf7, 256, grid=grid(256), stream=stream0)
del buf5
del buf6
buf8 = reinterpret_tensor(buf1, (4, 4, 4, 1), (16, 4, 1, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
triton_poi_fused_3.run(buf2, primals_7, buf8, 16, 4, grid=grid(16, 4), stream=stream0)
del primals_7
buf9 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.bmm(reinterpret_tensor(buf7, (16, 4, 4), (16, 4, 1), 0), reinterpret_tensor(buf8, (16, 4, 1), (4, 1, 0), 0), out=buf9)
buf10 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [c_1], Original ATen: [aten.clone]
triton_poi_fused_clone_4.run(buf9, buf10, 16, 4, grid=grid(16, 4), stream=stream0)
buf11 = reinterpret_tensor(buf9, (16, 4), (4, 1), 0); del buf9 # reuse
# Topologically Sorted Source Nodes: [o], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_9, reinterpret_tensor(buf10, (16, 4), (4, 1), 0), reinterpret_tensor(primals_8, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf11)
del primals_9
buf12 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
buf13 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
# Topologically Sorted Source Nodes: [add, attention_output], Original ATen: [aten.add, aten.native_layer_norm]
triton_poi_fused_add_native_layer_norm_5.run(buf11, primals_3, buf12, buf13, 16, grid=grid(16), stream=stream0)
buf14 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [add, attention_output], Original ATen: [aten.add, aten.native_layer_norm]
triton_poi_fused_add_native_layer_norm_6.run(buf11, primals_3, buf12, buf13, primals_10, primals_11, buf14, 64, grid=grid(64), stream=stream0)
del buf12
del buf13
del primals_11
return (buf14, primals_3, primals_10, buf7, reinterpret_tensor(buf8, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf4, (16, 4, 1), (4, 1, 4), 0), reinterpret_tensor(buf10, (16, 4), (4, 1), 0), buf11, primals_8, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from _paritybench_helpers import _mock_config
import math
import torch
from typing import List
from typing import Tuple
from torch import nn
from typing import Set
import torch.utils.checkpoint
def find_pruneable_heads_and_indices(heads: 'List[int]', n_heads: 'int',
head_size: 'int', already_pruned_heads: 'Set[int]') ->Tuple[Set[int],
torch.LongTensor]:
"""
Finds the heads and their indices taking :obj:`already_pruned_heads` into account.
Args:
heads (:obj:`List[int]`): List of the indices of heads to prune.
n_heads (:obj:`int`): The number of heads in the model.
head_size (:obj:`int`): The size of each head.
already_pruned_heads (:obj:`Set[int]`): A set of already pruned heads.
Returns:
:obj:`Tuple[Set[int], torch.LongTensor]`: A tuple with the remaining heads and their corresponding indices.
"""
mask = torch.ones(n_heads, head_size)
heads = set(heads) - already_pruned_heads
for head in heads:
head = head - sum(1 if h < head else 0 for h in already_pruned_heads)
mask[head] = 0
mask = mask.view(-1).contiguous().eq(1)
index: 'torch.LongTensor' = torch.arange(len(mask))[mask].long()
return heads, index
def prune_linear_layer(layer: 'nn.Linear', index: 'torch.LongTensor', dim:
'int'=0) ->nn.Linear:
"""
Prune a linear layer to keep only entries in index.
Used to remove heads.
Args:
layer (:obj:`torch.nn.Linear`): The layer to prune.
index (:obj:`torch.LongTensor`): The indices to keep in the layer.
dim (:obj:`int`, `optional`, defaults to 0): The dimension on which to keep the indices.
Returns:
:obj:`torch.nn.Linear`: The pruned layer as a new layer with :obj:`requires_grad=True`.
"""
index = index
W = layer.weight.index_select(dim, index).clone().detach()
if layer.bias is not None:
if dim == 1:
b = layer.bias.clone().detach()
else:
b = layer.bias[index].clone().detach()
new_size = list(layer.weight.size())
new_size[dim] = len(index)
new_layer = nn.Linear(new_size[1], new_size[0], bias=layer.bias is not None
)
new_layer.weight.requires_grad = False
new_layer.weight.copy_(W.contiguous())
new_layer.weight.requires_grad = True
if layer.bias is not None:
new_layer.bias.requires_grad = False
new_layer.bias.copy_(b.contiguous())
new_layer.bias.requires_grad = True
return new_layer
class MPNetSelfAttention(nn.Module):
def __init__(self, config):
super().__init__()
if (config.hidden_size % config.num_attention_heads != 0 and not
hasattr(config, 'embedding_size')):
raise ValueError(
f'The hidden size ({config.hidden_size}) is not a multiple of the number of attention heads ({config.num_attention_heads})'
)
self.num_attention_heads = config.num_attention_heads
self.attention_head_size = int(config.hidden_size / config.
num_attention_heads)
self.all_head_size = (self.num_attention_heads * self.
attention_head_size)
self.q = nn.Linear(config.hidden_size, self.all_head_size)
self.k = nn.Linear(config.hidden_size, self.all_head_size)
self.v = nn.Linear(config.hidden_size, self.all_head_size)
self.o = nn.Linear(config.hidden_size, config.hidden_size)
self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
def transpose_for_scores(self, x):
new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.
attention_head_size)
x = x.view(*new_x_shape)
return x.permute(0, 2, 1, 3)
def forward(self, hidden_states, attention_mask=None, head_mask=None,
position_bias=None, output_attentions=False, **kwargs):
q = self.q(hidden_states)
k = self.k(hidden_states)
v = self.v(hidden_states)
q = self.transpose_for_scores(q)
k = self.transpose_for_scores(k)
v = self.transpose_for_scores(v)
attention_scores = torch.matmul(q, k.transpose(-1, -2))
attention_scores = attention_scores / math.sqrt(self.
attention_head_size)
if position_bias is not None:
attention_scores += position_bias
if attention_mask is not None:
attention_scores = attention_scores + attention_mask
attention_probs = nn.Softmax(dim=-1)(attention_scores)
attention_probs = self.dropout(attention_probs)
if head_mask is not None:
attention_probs = attention_probs * head_mask
c = torch.matmul(attention_probs, v)
c = c.permute(0, 2, 1, 3).contiguous()
new_c_shape = c.size()[:-2] + (self.all_head_size,)
c = c.view(*new_c_shape)
o = self.o(c)
outputs = (o, attention_probs) if output_attentions else (o,)
return outputs
class MPNetAttention(nn.Module):
def __init__(self, config):
super().__init__()
self.attn = MPNetSelfAttention(config)
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.
layer_norm_eps)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
self.pruned_heads = set()
def prune_heads(self, heads):
if len(heads) == 0:
return
heads, index = find_pruneable_heads_and_indices(heads, self.attn.
num_attention_heads, self.attn.attention_head_size, self.
pruned_heads)
self.attn.q = prune_linear_layer(self.attn.q, index)
self.attn.k = prune_linear_layer(self.attn.k, index)
self.attn.v = prune_linear_layer(self.attn.v, index)
self.attn.o = prune_linear_layer(self.attn.o, index, dim=1)
self.attn.num_attention_heads = self.attn.num_attention_heads - len(
heads)
self.attn.all_head_size = (self.attn.attention_head_size * self.
attn.num_attention_heads)
self.pruned_heads = self.pruned_heads.union(heads)
def forward(self, hidden_states, attention_mask=None, head_mask=None,
position_bias=None, output_attentions=False, **kwargs):
self_outputs = self.attn(hidden_states, attention_mask, head_mask,
position_bias, output_attentions=output_attentions)
attention_output = self.LayerNorm(self.dropout(self_outputs[0]) +
hidden_states)
outputs = (attention_output,) + self_outputs[1:]
return outputs
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'config': _mock_config(hidden_size=4, num_attention_heads=
4, attention_probs_dropout_prob=0.5, layer_norm_eps=1,
hidden_dropout_prob=0.5)}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import math
from typing import List
from typing import Tuple
from torch import nn
from typing import Set
import torch.utils.checkpoint
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_0(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel, YBLOCK:
tl.constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = yindex // 4
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x2 + 16 * y1), xmask & ymask,
eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + y0, ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = 1.0
tmp4 = tmp2 * tmp3
tl.store(out_ptr0 + (x2 + 4 * y3), tmp4, xmask & ymask)
@triton.jit
def triton_poi_fused_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tmp9 = tl_math.exp(tmp8)
tl.store(out_ptr0 + x2, tmp9, xmask)
@triton.jit
def triton_poi_fused_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4
x2 = xindex
tmp0 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp18 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp25 = tl.load(in_ptr1 + x2, xmask)
tmp26 = tl.load(in_ptr1 + 4 * x1, xmask, eviction_policy='evict_last')
tmp27 = tl.load(in_ptr1 + (1 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp29 = tl.load(in_ptr1 + (2 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp31 = tl.load(in_ptr1 + (3 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp1 = float('-inf')
tmp2 = tmp0 == tmp1
tmp3 = tmp2 == 0
tmp4 = tmp3.to(tl.int64)
tmp5 = tmp4 != 0
tmp7 = tmp6 == tmp1
tmp8 = tmp7 == 0
tmp9 = tmp8.to(tl.int64)
tmp10 = tmp9 != 0
tmp11 = tmp5 | tmp10
tmp13 = tmp12 == tmp1
tmp14 = tmp13 == 0
tmp15 = tmp14.to(tl.int64)
tmp16 = tmp15 != 0
tmp17 = tmp11 | tmp16
tmp19 = tmp18 == tmp1
tmp20 = tmp19 == 0
tmp21 = tmp20.to(tl.int64)
tmp22 = tmp21 != 0
tmp23 = tmp17 | tmp22
tmp24 = tmp23 == 0
tmp28 = tmp26 + tmp27
tmp30 = tmp28 + tmp29
tmp32 = tmp30 + tmp31
tmp33 = tmp25 / tmp32
tmp34 = 0.0
tmp35 = tl.where(tmp24, tmp34, tmp33)
tl.store(out_ptr0 + x2, tmp35, xmask)
@triton.jit
def triton_poi_fused_3(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel, YBLOCK:
tl.constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = yindex // 4
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x2 + 16 * y1), xmask & ymask,
eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + y0, ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (x2 + 4 * y3), tmp2, xmask & ymask)
@triton.jit
def triton_poi_fused_clone_4(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = yindex // 4
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x2 + 16 * y1), xmask & ymask,
eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + 4 * y3), tmp0, xmask & ymask)
@triton.jit
def triton_poi_fused_add_native_layer_norm_5(in_ptr0, in_ptr1, out_ptr0,
out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 4 * x0, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + 4 * x0, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (2 + 4 * x0), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (2 + 4 * x0), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp12 = tl.load(in_ptr1 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 + tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 + tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp2 - tmp16
tmp18 = tmp17 * tmp17
tmp19 = tmp5 - tmp16
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp22 = tmp9 - tmp16
tmp23 = tmp22 * tmp22
tmp24 = tmp21 + tmp23
tmp25 = tmp13 - tmp16
tmp26 = tmp25 * tmp25
tmp27 = tmp24 + tmp26
tmp28 = tmp27 / tmp15
tl.store(out_ptr0 + x0, tmp16, xmask)
tl.store(out_ptr1 + x0, tmp28, xmask)
@triton.jit
def triton_poi_fused_add_native_layer_norm_6(in_ptr0, in_ptr1, in_ptr2,
in_ptr3, in_ptr4, in_ptr5, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp3 = tl.load(in_ptr2 + x1, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr3 + x1, xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr4 + x0, xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr5 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp4 = tmp2 - tmp3
tmp6 = 1.0
tmp7 = tmp5 + tmp6
tmp8 = libdevice.rsqrt(tmp7)
tmp9 = tmp4 * tmp8
tmp11 = tmp9 * tmp10
tmp13 = tmp11 + tmp12
tl.store(out_ptr0 + x2, tmp13, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4,), (1,))
assert_size_stride(primals_8, (4, 4), (4, 1))
assert_size_stride(primals_9, (4,), (1,))
assert_size_stride(primals_10, (4,), (1,))
assert_size_stride(primals_11, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf2)
del primals_6
buf3 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_0[grid(16, 4)](buf0, primals_2, buf3, 16, 4,
XBLOCK=4, YBLOCK=8, num_warps=1, num_stages=1)
del primals_2
buf4 = reinterpret_tensor(buf0, (4, 4, 1, 4), (16, 4, 4, 1), 0)
del buf0
triton_poi_fused_0[grid(16, 4)](buf1, primals_5, buf4, 16, 4,
XBLOCK=4, YBLOCK=8, num_warps=1, num_stages=1)
del primals_5
buf5 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 0),
0), reinterpret_tensor(buf4, (16, 1, 4), (4, 0, 1), 0), out=buf5)
buf6 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_1[grid(256)](buf5, buf6, 256, XBLOCK=128,
num_warps=4, num_stages=1)
buf7 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_2[grid(256)](buf5, buf6, buf7, 256, XBLOCK=256,
num_warps=4, num_stages=1)
del buf5
del buf6
buf8 = reinterpret_tensor(buf1, (4, 4, 4, 1), (16, 4, 1, 1), 0)
del buf1
triton_poi_fused_3[grid(16, 4)](buf2, primals_7, buf8, 16, 4,
XBLOCK=4, YBLOCK=8, num_warps=1, num_stages=1)
del primals_7
buf9 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0)
del buf2
extern_kernels.bmm(reinterpret_tensor(buf7, (16, 4, 4), (16, 4, 1),
0), reinterpret_tensor(buf8, (16, 4, 1), (4, 1, 0), 0), out=buf9)
buf10 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 1), torch.float32)
triton_poi_fused_clone_4[grid(16, 4)](buf9, buf10, 16, 4, XBLOCK=4,
YBLOCK=16, num_warps=1, num_stages=1)
buf11 = reinterpret_tensor(buf9, (16, 4), (4, 1), 0)
del buf9
extern_kernels.addmm(primals_9, reinterpret_tensor(buf10, (16, 4),
(4, 1), 0), reinterpret_tensor(primals_8, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf11)
del primals_9
buf12 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
buf13 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
triton_poi_fused_add_native_layer_norm_5[grid(16)](buf11, primals_3,
buf12, buf13, 16, XBLOCK=16, num_warps=1, num_stages=1)
buf14 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_add_native_layer_norm_6[grid(64)](buf11, primals_3,
buf12, buf13, primals_10, primals_11, buf14, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del buf12
del buf13
del primals_11
return buf14, primals_3, primals_10, buf7, reinterpret_tensor(buf8, (16,
1, 4), (4, 1, 1), 0), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0
), reinterpret_tensor(buf4, (16, 4, 1), (4, 1, 4), 0
), reinterpret_tensor(buf10, (16, 4), (4, 1), 0), buf11, primals_8
def find_pruneable_heads_and_indices(heads: 'List[int]', n_heads: 'int',
head_size: 'int', already_pruned_heads: 'Set[int]') ->Tuple[Set[int],
torch.LongTensor]:
"""
Finds the heads and their indices taking :obj:`already_pruned_heads` into account.
Args:
heads (:obj:`List[int]`): List of the indices of heads to prune.
n_heads (:obj:`int`): The number of heads in the model.
head_size (:obj:`int`): The size of each head.
already_pruned_heads (:obj:`Set[int]`): A set of already pruned heads.
Returns:
:obj:`Tuple[Set[int], torch.LongTensor]`: A tuple with the remaining heads and their corresponding indices.
"""
mask = torch.ones(n_heads, head_size)
heads = set(heads) - already_pruned_heads
for head in heads:
head = head - sum(1 if h < head else 0 for h in already_pruned_heads)
mask[head] = 0
mask = mask.view(-1).contiguous().eq(1)
index: 'torch.LongTensor' = torch.arange(len(mask))[mask].long()
return heads, index
def prune_linear_layer(layer: 'nn.Linear', index: 'torch.LongTensor', dim:
'int'=0) ->nn.Linear:
"""
Prune a linear layer to keep only entries in index.
Used to remove heads.
Args:
layer (:obj:`torch.nn.Linear`): The layer to prune.
index (:obj:`torch.LongTensor`): The indices to keep in the layer.
dim (:obj:`int`, `optional`, defaults to 0): The dimension on which to keep the indices.
Returns:
:obj:`torch.nn.Linear`: The pruned layer as a new layer with :obj:`requires_grad=True`.
"""
index = index
W = layer.weight.index_select(dim, index).clone().detach()
if layer.bias is not None:
if dim == 1:
b = layer.bias.clone().detach()
else:
b = layer.bias[index].clone().detach()
new_size = list(layer.weight.size())
new_size[dim] = len(index)
new_layer = nn.Linear(new_size[1], new_size[0], bias=layer.bias is not None
)
new_layer.weight.requires_grad = False
new_layer.weight.copy_(W.contiguous())
new_layer.weight.requires_grad = True
if layer.bias is not None:
new_layer.bias.requires_grad = False
new_layer.bias.copy_(b.contiguous())
new_layer.bias.requires_grad = True
return new_layer
class MPNetSelfAttention(nn.Module):
def __init__(self, config):
super().__init__()
if (config.hidden_size % config.num_attention_heads != 0 and not
hasattr(config, 'embedding_size')):
raise ValueError(
f'The hidden size ({config.hidden_size}) is not a multiple of the number of attention heads ({config.num_attention_heads})'
)
self.num_attention_heads = config.num_attention_heads
self.attention_head_size = int(config.hidden_size / config.
num_attention_heads)
self.all_head_size = (self.num_attention_heads * self.
attention_head_size)
self.q = nn.Linear(config.hidden_size, self.all_head_size)
self.k = nn.Linear(config.hidden_size, self.all_head_size)
self.v = nn.Linear(config.hidden_size, self.all_head_size)
self.o = nn.Linear(config.hidden_size, config.hidden_size)
self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
def transpose_for_scores(self, x):
new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.
attention_head_size)
x = x.view(*new_x_shape)
return x.permute(0, 2, 1, 3)
def forward(self, hidden_states, attention_mask=None, head_mask=None,
position_bias=None, output_attentions=False, **kwargs):
q = self.q(hidden_states)
k = self.k(hidden_states)
v = self.v(hidden_states)
q = self.transpose_for_scores(q)
k = self.transpose_for_scores(k)
v = self.transpose_for_scores(v)
attention_scores = torch.matmul(q, k.transpose(-1, -2))
attention_scores = attention_scores / math.sqrt(self.
attention_head_size)
if position_bias is not None:
attention_scores += position_bias
if attention_mask is not None:
attention_scores = attention_scores + attention_mask
attention_probs = nn.Softmax(dim=-1)(attention_scores)
attention_probs = self.dropout(attention_probs)
if head_mask is not None:
attention_probs = attention_probs * head_mask
c = torch.matmul(attention_probs, v)
c = c.permute(0, 2, 1, 3).contiguous()
new_c_shape = c.size()[:-2] + (self.all_head_size,)
c = c.view(*new_c_shape)
o = self.o(c)
outputs = (o, attention_probs) if output_attentions else (o,)
return outputs
class MPNetAttentionNew(nn.Module):
def __init__(self, config):
super().__init__()
self.attn = MPNetSelfAttention(config)
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.
layer_norm_eps)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
self.pruned_heads = set()
def prune_heads(self, heads):
if len(heads) == 0:
return
heads, index = find_pruneable_heads_and_indices(heads, self.attn.
num_attention_heads, self.attn.attention_head_size, self.
pruned_heads)
self.attn.q = prune_linear_layer(self.attn.q, index)
self.attn.k = prune_linear_layer(self.attn.k, index)
self.attn.v = prune_linear_layer(self.attn.v, index)
self.attn.o = prune_linear_layer(self.attn.o, index, dim=1)
self.attn.num_attention_heads = self.attn.num_attention_heads - len(
heads)
self.attn.all_head_size = (self.attn.attention_head_size * self.
attn.num_attention_heads)
self.pruned_heads = self.pruned_heads.union(heads)
def forward(self, input_0):
primals_1 = self.attn.q.weight
primals_2 = self.attn.q.bias
primals_4 = self.attn.k.weight
primals_5 = self.attn.k.bias
primals_6 = self.attn.v.weight
primals_7 = self.attn.v.bias
primals_8 = self.attn.o.weight
primals_9 = self.attn.o.bias
primals_10 = self.LayerNorm.weight
primals_11 = self.LayerNorm.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11])
return output[0]
|
Clemens123/transformers
|
MPNetAttention
| false | 12,331 |
[
"Apache-2.0"
] | 0 |
22abe7bbc587c16ec30f9d1aa549dcbeba6e9e26
|
https://github.com/Clemens123/transformers/tree/22abe7bbc587c16ec30f9d1aa549dcbeba6e9e26
|
NoiseLayer
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/hy/chy2gdvcdlpxio6r2ezu76sbqj6jaum4snkq6izi5zxkdqh3u2yj.py
# Topologically Sorted Source Nodes: [mul, x], Original ATen: [aten.mul, aten.add]
# Source node to ATen node mapping:
# mul => mul
# x => add
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view, %randn), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%primals_1, %mul), kwargs = {})
triton_poi_fused_add_mul_0 = async_compile.triton('triton_poi_fused_add_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_mul_0(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 16) % 4
x0 = xindex % 16
x2 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr1 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr2 + (x0 + (16*x2)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 * tmp2
tmp4 = tmp0 + tmp3
tl.store(out_ptr0 + (x3), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [noise], Original ATen: [aten.randn]
buf0 = torch.ops.aten.randn.default([4, 1, 4, 4], dtype=torch.float32, device=device(type='cuda', index=0), pin_memory=False)
buf1 = buf0
del buf0
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, x], Original ATen: [aten.mul, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_mul_0.run(primals_1, primals_2, buf1, buf2, 256, grid=grid(256), stream=stream0)
del primals_1
del primals_2
return (buf2, buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class NoiseLayer(nn.Module):
"""adds noise. noise is per pixel (constant over channels) with per-channel weight"""
def __init__(self, channels):
super().__init__()
self.weight = nn.Parameter(torch.zeros(channels))
self.noise = None
def forward(self, x, noise=None):
if noise is None and self.noise is None:
noise = torch.randn(x.size(0), 1, x.size(2), x.size(3), device=
x.device, dtype=x.dtype)
elif noise is None:
noise = self.noise
x = x + self.weight.view(1, -1, 1, 1) * noise
return x
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'channels': 4}]
|
import torch
from torch import device
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_mul_0(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 16 % 4
x0 = xindex % 16
x2 = xindex // 64
tmp0 = tl.load(in_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr1 + x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr2 + (x0 + 16 * x2), xmask, eviction_policy=
'evict_last')
tmp3 = tmp1 * tmp2
tmp4 = tmp0 + tmp3
tl.store(out_ptr0 + x3, tmp4, xmask)
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = torch.ops.aten.randn.default([4, 1, 4, 4], dtype=torch.
float32, device=device(type='cuda', index=0), pin_memory=False)
buf1 = buf0
del buf0
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_mul_0[grid(256)](primals_1, primals_2, buf1,
buf2, 256, XBLOCK=128, num_warps=4, num_stages=1)
del primals_1
del primals_2
return buf2, buf1
class NoiseLayerNew(nn.Module):
"""adds noise. noise is per pixel (constant over channels) with per-channel weight"""
def __init__(self, channels):
super().__init__()
self.weight = nn.Parameter(torch.zeros(channels))
self.noise = None
def forward(self, input_0):
primals_2 = self.weight
primals_1 = input_0
output = call([primals_1, primals_2])
return output[0]
|
eitanrich/ganspace-manifold
|
NoiseLayer
| false | 12,332 |
[
"Apache-2.0"
] | 0 |
148d5d30001c43794a40bbed885601e7816f5d7d
|
https://github.com/eitanrich/ganspace-manifold/tree/148d5d30001c43794a40bbed885601e7816f5d7d
|
KnowledgeDistillationLoss
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/td/ctdj5kazgiki6gdaadhqtp2x7tq2ee5ey5hqqdcoqmp54jyhf74f.py
# Topologically Sorted Source Nodes: [outputs], Original ATen: [aten._log_softmax]
# Source node to ATen node mapping:
# outputs => amax, sub
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%slice_1, [1], True), kwargs = {})
# %sub : [num_users=2] = call_function[target=torch.ops.aten.sub.Tensor](args = (%slice_1, %amax), kwargs = {})
triton_poi_fused__log_softmax_0 = async_compile.triton('triton_poi_fused__log_softmax_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__log_softmax_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__log_softmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 16
x2 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (16 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (32 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (48 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tl.store(out_ptr0 + (x3), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/k4/ck4jcdo7pholu6j7hyu5a5wtavlli6c52bnxxxqwlta7ah2dps4s.py
# Topologically Sorted Source Nodes: [labels], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# labels => exp_1
# Graph fragment:
# %mul_tensor : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg1_1, 1), kwargs = {})
# %amax_default : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%mul_tensor, [1], True), kwargs = {})
# %sub_tensor : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_tensor, %amax_default), kwargs = {})
# %mul_tensor_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_tensor, 1.0), kwargs = {})
# %exp_1 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%mul_tensor_1,), kwargs = {})
triton_poi_fused__softmax_1 = async_compile.triton('triton_poi_fused__softmax_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 16
x2 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x3), xmask)
tmp3 = tl.load(in_ptr0 + (x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (16 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (32 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (48 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = tmp14 * tmp1
tmp16 = tl_math.exp(tmp15)
tl.store(out_ptr0 + (x3), tmp16, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/tf/ctfwzouxcntmhueaa2wrmeoexhrsrlrgnisvw3gxzpnbe6tqizfn.py
# Topologically Sorted Source Nodes: [outputs, labels, mul_1], Original ATen: [aten._log_softmax, aten._softmax, aten.mul]
# Source node to ATen node mapping:
# labels => div, sum_2
# mul_1 => mul_1
# outputs => exp, log, sub_1, sum_1
# Graph fragment:
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [1], True), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%sum_1,), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sub, %log), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_1, [1], True), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp_1, %sum_2), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_1, %div), kwargs = {})
triton_poi_fused__log_softmax__softmax_mul_2 = async_compile.triton('triton_poi_fused__log_softmax__softmax_mul_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__log_softmax__softmax_mul_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 10, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__log_softmax__softmax_mul_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 16
x2 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (16 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (32 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp9 = tl.load(in_ptr0 + (48 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp14 = tl.load(in_ptr1 + (x3), xmask)
tmp15 = tl.load(in_ptr1 + (x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp16 = tl.load(in_ptr1 + (16 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp18 = tl.load(in_ptr1 + (32 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp20 = tl.load(in_ptr1 + (48 + x0 + (64*x2)), xmask, eviction_policy='evict_last')
tmp2 = tl_math.exp(tmp1)
tmp4 = tl_math.exp(tmp3)
tmp5 = tmp2 + tmp4
tmp7 = tl_math.exp(tmp6)
tmp8 = tmp5 + tmp7
tmp10 = tl_math.exp(tmp9)
tmp11 = tmp8 + tmp10
tmp12 = tl_math.log(tmp11)
tmp13 = tmp0 - tmp12
tmp17 = tmp15 + tmp16
tmp19 = tmp17 + tmp18
tmp21 = tmp19 + tmp20
tmp22 = tmp14 / tmp21
tmp23 = tmp13 * tmp22
tl.store(out_ptr0 + (x3), tmp23, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/xh/cxhnxhufk74s4p7gma545bi7je2zowpnn6bhbh645dch7pmafdxd.py
# Topologically Sorted Source Nodes: [loss, mean_1, outputs_1], Original ATen: [aten.mean, aten.neg]
# Source node to ATen node mapping:
# loss => mean
# mean_1 => mean_1
# outputs_1 => neg
# Graph fragment:
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%mul_1, [1]), kwargs = {})
# %mean_1 : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%mean,), kwargs = {})
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%mean_1,), kwargs = {})
triton_per_fused_mean_neg_3 = async_compile.triton('triton_per_fused_mean_neg_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 64],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {2: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 3), equal_to_1=(2,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_mean_neg_3', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_mean_neg_3(in_out_ptr0, in_ptr0, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 1
rnumel = 64
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex % 16
r1 = (rindex // 16)
tmp0 = tl.load(in_ptr0 + (r0 + (64*r1)), None)
tmp1 = tl.load(in_ptr0 + (16 + r0 + (64*r1)), None)
tmp3 = tl.load(in_ptr0 + (32 + r0 + (64*r1)), None)
tmp5 = tl.load(in_ptr0 + (48 + r0 + (64*r1)), None)
tmp2 = tmp0 + tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp7 = 4.0
tmp8 = tmp6 / tmp7
tmp9 = tl.broadcast_to(tmp8, [XBLOCK, RBLOCK])
tmp11 = tl.sum(tmp9, 1)[:, None]
tmp12 = 64.0
tmp13 = tmp11 / tmp12
tmp14 = -tmp13
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([XBLOCK, 1], 0, tl.int32)), tmp14, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [outputs], Original ATen: [aten._log_softmax]
stream0 = get_raw_stream(0)
triton_poi_fused__log_softmax_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [labels], Original ATen: [aten._softmax]
triton_poi_fused__softmax_1.run(arg1_1, buf1, 256, grid=grid(256), stream=stream0)
del arg1_1
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [outputs, labels, mul_1], Original ATen: [aten._log_softmax, aten._softmax, aten.mul]
triton_poi_fused__log_softmax__softmax_mul_2.run(buf0, buf1, buf2, 256, grid=grid(256), stream=stream0)
del buf0
del buf1
buf3 = empty_strided_cuda((), (), torch.float32)
buf4 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [loss, mean_1, outputs_1], Original ATen: [aten.mean, aten.neg]
triton_per_fused_mean_neg_3.run(buf4, buf2, 1, 64, grid=grid(1), stream=stream0)
del buf2
return (buf4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
class KnowledgeDistillationLoss(nn.Module):
def __init__(self, reduction='mean', alpha=1.0):
super().__init__()
self.reduction = reduction
self.alpha = alpha
def forward(self, inputs, targets, mask=None):
inputs = inputs.narrow(1, 0, targets.shape[1])
outputs = torch.log_softmax(inputs, dim=1)
labels = torch.softmax(targets * self.alpha, dim=1)
loss = (outputs * labels).mean(dim=1)
if mask is not None:
loss = loss * mask.float()
if self.reduction == 'mean':
outputs = -torch.mean(loss)
elif self.reduction == 'sum':
outputs = -torch.sum(loss)
else:
outputs = -loss
return outputs
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused__log_softmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 16
x2 = xindex // 64
tmp0 = tl.load(in_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + (x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp2 = tl.load(in_ptr0 + (16 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp4 = tl.load(in_ptr0 + (32 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp6 = tl.load(in_ptr0 + (48 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tl.store(out_ptr0 + x3, tmp8, xmask)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 16
x2 = xindex // 64
tmp0 = tl.load(in_ptr0 + x3, xmask)
tmp3 = tl.load(in_ptr0 + (x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp5 = tl.load(in_ptr0 + (16 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp8 = tl.load(in_ptr0 + (32 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp11 = tl.load(in_ptr0 + (48 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = tmp14 * tmp1
tmp16 = tl_math.exp(tmp15)
tl.store(out_ptr0 + x3, tmp16, xmask)
@triton.jit
def triton_poi_fused__log_softmax__softmax_mul_2(in_ptr0, in_ptr1, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 16
x2 = xindex // 64
tmp0 = tl.load(in_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + (x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp3 = tl.load(in_ptr0 + (16 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp6 = tl.load(in_ptr0 + (32 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp9 = tl.load(in_ptr0 + (48 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp14 = tl.load(in_ptr1 + x3, xmask)
tmp15 = tl.load(in_ptr1 + (x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp16 = tl.load(in_ptr1 + (16 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp18 = tl.load(in_ptr1 + (32 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp20 = tl.load(in_ptr1 + (48 + x0 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp2 = tl_math.exp(tmp1)
tmp4 = tl_math.exp(tmp3)
tmp5 = tmp2 + tmp4
tmp7 = tl_math.exp(tmp6)
tmp8 = tmp5 + tmp7
tmp10 = tl_math.exp(tmp9)
tmp11 = tmp8 + tmp10
tmp12 = tl_math.log(tmp11)
tmp13 = tmp0 - tmp12
tmp17 = tmp15 + tmp16
tmp19 = tmp17 + tmp18
tmp21 = tmp19 + tmp20
tmp22 = tmp14 / tmp21
tmp23 = tmp13 * tmp22
tl.store(out_ptr0 + x3, tmp23, xmask)
@triton.jit
def triton_per_fused_mean_neg_3(in_out_ptr0, in_ptr0, xnumel, rnumel,
XBLOCK: tl.constexpr):
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex % 16
r1 = rindex // 16
tmp0 = tl.load(in_ptr0 + (r0 + 64 * r1), None)
tmp1 = tl.load(in_ptr0 + (16 + r0 + 64 * r1), None)
tmp3 = tl.load(in_ptr0 + (32 + r0 + 64 * r1), None)
tmp5 = tl.load(in_ptr0 + (48 + r0 + 64 * r1), None)
tmp2 = tmp0 + tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp7 = 4.0
tmp8 = tmp6 / tmp7
tmp9 = tl.broadcast_to(tmp8, [XBLOCK, RBLOCK])
tmp11 = tl.sum(tmp9, 1)[:, None]
tmp12 = 64.0
tmp13 = tmp11 / tmp12
tmp14 = -tmp13
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([XBLOCK, 1], 0, tl.int32), tmp14, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused__log_softmax_0[grid(256)](arg0_1, buf0, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused__softmax_1[grid(256)](arg1_1, buf1, 256, XBLOCK=
128, num_warps=4, num_stages=1)
del arg1_1
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused__log_softmax__softmax_mul_2[grid(256)](buf0, buf1,
buf2, 256, XBLOCK=128, num_warps=4, num_stages=1)
del buf0
del buf1
buf3 = empty_strided_cuda((), (), torch.float32)
buf4 = buf3
del buf3
triton_per_fused_mean_neg_3[grid(1)](buf4, buf2, 1, 64, XBLOCK=1,
num_warps=2, num_stages=1)
del buf2
return buf4,
class KnowledgeDistillationLossNew(nn.Module):
def __init__(self, reduction='mean', alpha=1.0):
super().__init__()
self.reduction = reduction
self.alpha = alpha
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
edoardofantolino/MLDLproject4
|
KnowledgeDistillationLoss
| false | 12,333 |
[
"MIT"
] | 0 |
fed0cfd51f5984bbf21205a43ea43dc49f4d289a
|
https://github.com/edoardofantolino/MLDLproject4/tree/fed0cfd51f5984bbf21205a43ea43dc49f4d289a
|
SubpixelConvolutionLayer
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/3x/c3xs42gpn3grgaejyguafrbyhbbi4mky2tdfmfgpjkvjjdcx64dk.py
# Unsorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
triton_poi_fused_0 = async_compile.triton('triton_poi_fused_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384, 16], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_0(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16384
xnumel = 9
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = tl.full([XBLOCK, YBLOCK], True, tl.int1)
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 64
y1 = (yindex // 64)
tmp0 = tl.load(in_ptr0 + (x2 + (9*y3)), xmask, eviction_policy='evict_last')
tl.store(out_ptr0 + (y0 + (64*x2) + (576*y1)), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/nq/cnqioqtc5smqmnt22pzdujcgch6iuo4ayzdajy2hr5awqxgsqhdm.py
# Unsorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
triton_poi_fused_1 = async_compile.triton('triton_poi_fused_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256, 4096], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 256
xnumel = 4096
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, YBLOCK], True, tl.int1)
x2 = xindex
y3 = yindex
y0 = yindex % 64
y1 = (yindex // 64)
tmp0 = tl.load(in_ptr0 + (x2 + (4096*y3)), ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (y0 + (64*x2) + (262144*y1)), tmp0, ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kx/ckx2rnidyqb5tiful7teg7va3kagir3uo3nzo2l5v3pauvflocqb.py
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# out => convolution
# Graph fragment:
# %convolution : [num_users=2] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_2 = async_compile.triton('triton_poi_fused_convolution_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4194304],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_2', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_2(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4194304
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 256
tmp0 = tl.load(in_out_ptr0 + (x2), None)
tmp1 = tl.load(in_ptr0 + (x0), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x2), tmp2, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/mp/cmpqzc6swhfwsh5sgmdsboqplylzump4cxyfwiiz7zoavqd6ac6v.py
# Topologically Sorted Source Nodes: [out_2], Original ATen: [aten._prelu_kernel]
# Source node to ATen node mapping:
# out_2 => gt, mul, where
# Graph fragment:
# %gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%view_1, 0), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_2, %view_1), kwargs = {})
# %where : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt, %view_1, %mul), kwargs = {})
triton_poi_fused__prelu_kernel_3 = async_compile.triton('triton_poi_fused__prelu_kernel_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4194304],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__prelu_kernel_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__prelu_kernel_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4194304
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x0 = xindex % 128
x1 = (xindex // 128) % 128
x2 = (xindex // 16384) % 64
x3 = (xindex // 1048576)
x4 = xindex
tmp0 = tl.load(in_ptr0 + ((2*(x1 % 2)) + (4*x2) + (256*(x0 // 2)) + (16384*(x1 // 2)) + (1048576*x3) + (x0 % 2)), None)
tmp3 = tl.load(in_ptr1 + (0))
tmp4 = tl.broadcast_to(tmp3, [XBLOCK])
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp5 = tmp4 * tmp0
tmp6 = tl.where(tmp2, tmp0, tmp5)
tl.store(out_ptr0 + (x4), tmp6, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (256, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_2, (256, ), (1, ))
assert_size_stride(primals_3, (4, 64, 64, 64), (262144, 4096, 64, 1))
assert_size_stride(primals_4, (1, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((256, 64, 3, 3), (576, 1, 192, 64), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
stream0 = get_raw_stream(0)
triton_poi_fused_0.run(primals_1, buf0, 16384, 9, grid=grid(16384, 9), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((4, 64, 64, 64), (262144, 1, 4096, 64), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
triton_poi_fused_1.run(primals_3, buf1, 256, 4096, grid=grid(256, 4096), stream=stream0)
del primals_3
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
buf2 = extern_kernels.convolution(buf1, buf0, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 256, 64, 64), (1048576, 1, 16384, 256))
buf3 = buf2; del buf2 # reuse
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
triton_poi_fused_convolution_2.run(buf3, primals_2, 4194304, grid=grid(4194304), stream=stream0)
del primals_2
buf4 = empty_strided_cuda((4, 64, 128, 128), (1048576, 16384, 128, 1), torch.float32)
# Topologically Sorted Source Nodes: [out_2], Original ATen: [aten._prelu_kernel]
triton_poi_fused__prelu_kernel_3.run(buf3, primals_4, buf4, 4194304, grid=grid(4194304), stream=stream0)
return (buf4, buf0, buf1, primals_4, buf3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((256, 64, 3, 3), (576, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((256, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 64, 64, 64), (262144, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class SubpixelConvolutionLayer(nn.Module):
def __init__(self, channels: 'int'=64) ->None:
"""
Args:
channels (int): Number of channels in the input image. (Default: 64)
"""
super(SubpixelConvolutionLayer, self).__init__()
self.conv = nn.Conv2d(channels, channels * 4, kernel_size=3, stride
=1, padding=1)
self.pixel_shuffle = nn.PixelShuffle(upscale_factor=2)
self.prelu = nn.PReLU()
def forward(self, x: 'torch.Tensor') ->torch.Tensor:
out = self.conv(x)
out = self.pixel_shuffle(out)
out = self.prelu(out)
return out
def get_inputs():
return [torch.rand([4, 64, 64, 64])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_0(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
xnumel = 9
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
tl.full([XBLOCK, YBLOCK], True, tl.int1)
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 64
y1 = yindex // 64
tmp0 = tl.load(in_ptr0 + (x2 + 9 * y3), xmask, eviction_policy='evict_last'
)
tl.store(out_ptr0 + (y0 + 64 * x2 + 576 * y1), tmp0, xmask)
@triton.jit
def triton_poi_fused_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 256
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, YBLOCK], True, tl.int1)
x2 = xindex
y3 = yindex
y0 = yindex % 64
y1 = yindex // 64
tmp0 = tl.load(in_ptr0 + (x2 + 4096 * y3), ymask, eviction_policy=
'evict_last')
tl.store(out_ptr0 + (y0 + 64 * x2 + 262144 * y1), tmp0, ymask)
@triton.jit
def triton_poi_fused_convolution_2(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 256
tmp0 = tl.load(in_out_ptr0 + x2, None)
tmp1 = tl.load(in_ptr0 + x0, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x2, tmp2, None)
@triton.jit
def triton_poi_fused__prelu_kernel_3(in_ptr0, in_ptr1, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x0 = xindex % 128
x1 = xindex // 128 % 128
x2 = xindex // 16384 % 64
x3 = xindex // 1048576
x4 = xindex
tmp0 = tl.load(in_ptr0 + (2 * (x1 % 2) + 4 * x2 + 256 * (x0 // 2) +
16384 * (x1 // 2) + 1048576 * x3 + x0 % 2), None)
tmp3 = tl.load(in_ptr1 + 0)
tmp4 = tl.broadcast_to(tmp3, [XBLOCK])
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp5 = tmp4 * tmp0
tmp6 = tl.where(tmp2, tmp0, tmp5)
tl.store(out_ptr0 + x4, tmp6, None)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (256, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_2, (256,), (1,))
assert_size_stride(primals_3, (4, 64, 64, 64), (262144, 4096, 64, 1))
assert_size_stride(primals_4, (1,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((256, 64, 3, 3), (576, 1, 192, 64), torch
.float32)
get_raw_stream(0)
triton_poi_fused_0[grid(16384, 9)](primals_1, buf0, 16384, 9,
XBLOCK=16, YBLOCK=64, num_warps=4, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((4, 64, 64, 64), (262144, 1, 4096, 64),
torch.float32)
triton_poi_fused_1[grid(256, 4096)](primals_3, buf1, 256, 4096,
XBLOCK=32, YBLOCK=32, num_warps=4, num_stages=1)
del primals_3
buf2 = extern_kernels.convolution(buf1, buf0, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 256, 64, 64), (1048576, 1, 16384, 256))
buf3 = buf2
del buf2
triton_poi_fused_convolution_2[grid(4194304)](buf3, primals_2,
4194304, XBLOCK=1024, num_warps=4, num_stages=1)
del primals_2
buf4 = empty_strided_cuda((4, 64, 128, 128), (1048576, 16384, 128,
1), torch.float32)
triton_poi_fused__prelu_kernel_3[grid(4194304)](buf3, primals_4,
buf4, 4194304, XBLOCK=512, num_warps=8, num_stages=1)
return buf4, buf0, buf1, primals_4, buf3
class SubpixelConvolutionLayerNew(nn.Module):
def __init__(self, channels: 'int'=64) ->None:
"""
Args:
channels (int): Number of channels in the input image. (Default: 64)
"""
super(SubpixelConvolutionLayerNew, self).__init__()
self.conv = nn.Conv2d(channels, channels * 4, kernel_size=3, stride
=1, padding=1)
self.pixel_shuffle = nn.PixelShuffle(upscale_factor=2)
self.prelu = nn.PReLU()
def forward(self, input_0):
primals_1 = self.conv.weight
primals_2 = self.conv.bias
primals_4 = self.prelu.weight
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
duylebkHCM/Anime-Face-Generator-
|
SubpixelConvolutionLayer
| false | 12,334 |
[
"MIT"
] | 0 |
ffcbe22f2073971e81b1bbc61b7ef7970889f8a2
|
https://github.com/duylebkHCM/Anime-Face-Generator-/tree/ffcbe22f2073971e81b1bbc61b7ef7970889f8a2
|
RecursiveNet
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ox/cox2ybkqjuarvohngrkwnqr4ehcw652sn4xc4nhy6qfdms3qyhfe.py
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# out => convolution
# Graph fragment:
# %convolution : [num_users=2] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_1, %primals_2, %primals_3, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_0 = async_compile.triton('triton_poi_fused_convolution_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1048576],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1048576
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = (xindex // 4096) % 64
tmp0 = tl.load(in_out_ptr0 + (x3), None)
tmp1 = tl.load(in_ptr0 + (x1), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 64, 64, 64), (262144, 4096, 64, 1))
assert_size_stride(primals_2, (64, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_3, (64, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_1, primals_2, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_0.run(buf1, primals_3, 1048576, grid=grid(1048576), stream=stream0)
# Topologically Sorted Source Nodes: [out_1], Original ATen: [aten.convolution]
buf2 = extern_kernels.convolution(buf1, primals_2, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf3 = buf2; del buf2 # reuse
# Topologically Sorted Source Nodes: [out_1], Original ATen: [aten.convolution]
triton_poi_fused_convolution_0.run(buf3, primals_3, 1048576, grid=grid(1048576), stream=stream0)
# Topologically Sorted Source Nodes: [out_2], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(buf3, primals_2, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [out_2], Original ATen: [aten.convolution]
triton_poi_fused_convolution_0.run(buf5, primals_3, 1048576, grid=grid(1048576), stream=stream0)
# Topologically Sorted Source Nodes: [out_3], Original ATen: [aten.convolution]
buf6 = extern_kernels.convolution(buf5, primals_2, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf7 = buf6; del buf6 # reuse
# Topologically Sorted Source Nodes: [out_3], Original ATen: [aten.convolution]
triton_poi_fused_convolution_0.run(buf7, primals_3, 1048576, grid=grid(1048576), stream=stream0)
# Topologically Sorted Source Nodes: [out_4], Original ATen: [aten.convolution]
buf8 = extern_kernels.convolution(buf7, primals_2, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf8, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf9 = buf8; del buf8 # reuse
# Topologically Sorted Source Nodes: [out_4], Original ATen: [aten.convolution]
triton_poi_fused_convolution_0.run(buf9, primals_3, 1048576, grid=grid(1048576), stream=stream0)
# Topologically Sorted Source Nodes: [out_5], Original ATen: [aten.convolution]
buf10 = extern_kernels.convolution(buf9, primals_2, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf10, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf11 = buf10; del buf10 # reuse
# Topologically Sorted Source Nodes: [out_5], Original ATen: [aten.convolution]
triton_poi_fused_convolution_0.run(buf11, primals_3, 1048576, grid=grid(1048576), stream=stream0)
del primals_3
return (buf11, primals_1, primals_2, buf1, buf3, buf5, buf7, buf9, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 64, 64, 64), (262144, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((64, 64, 3, 3), (576, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((64, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from typing import Any
import torch.nn as nn
class RecursiveNet(nn.Module):
""" Model that uses a layer recursively in computation. """
def __init__(self) ->None:
super().__init__()
self.conv1 = nn.Conv2d(64, 64, 3, 1, 1)
def forward(self, x: 'torch.Tensor', args1: 'Any'=None, args2: 'Any'=None
) ->torch.Tensor:
del args1, args2
out = x
for _ in range(3):
out = self.conv1(out)
out = self.conv1(out)
return out
def get_inputs():
return [torch.rand([4, 64, 64, 64])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = xindex // 4096 % 64
tmp0 = tl.load(in_out_ptr0 + x3, None)
tmp1 = tl.load(in_ptr0 + x1, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, None)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 64, 64, 64), (262144, 4096, 64, 1))
assert_size_stride(primals_2, (64, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_3, (64,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_1, primals_2, stride=(1,
1), padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_convolution_0[grid(1048576)](buf1, primals_3,
1048576, XBLOCK=512, num_warps=8, num_stages=1)
buf2 = extern_kernels.convolution(buf1, primals_2, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf3 = buf2
del buf2
triton_poi_fused_convolution_0[grid(1048576)](buf3, primals_3,
1048576, XBLOCK=512, num_warps=8, num_stages=1)
buf4 = extern_kernels.convolution(buf3, primals_2, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf5 = buf4
del buf4
triton_poi_fused_convolution_0[grid(1048576)](buf5, primals_3,
1048576, XBLOCK=512, num_warps=8, num_stages=1)
buf6 = extern_kernels.convolution(buf5, primals_2, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf7 = buf6
del buf6
triton_poi_fused_convolution_0[grid(1048576)](buf7, primals_3,
1048576, XBLOCK=512, num_warps=8, num_stages=1)
buf8 = extern_kernels.convolution(buf7, primals_2, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf8, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf9 = buf8
del buf8
triton_poi_fused_convolution_0[grid(1048576)](buf9, primals_3,
1048576, XBLOCK=512, num_warps=8, num_stages=1)
buf10 = extern_kernels.convolution(buf9, primals_2, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf10, (4, 64, 64, 64), (262144, 4096, 64, 1))
buf11 = buf10
del buf10
triton_poi_fused_convolution_0[grid(1048576)](buf11, primals_3,
1048576, XBLOCK=512, num_warps=8, num_stages=1)
del primals_3
return buf11, primals_1, primals_2, buf1, buf3, buf5, buf7, buf9
class RecursiveNetNew(nn.Module):
""" Model that uses a layer recursively in computation. """
def __init__(self) ->None:
super().__init__()
self.conv1 = nn.Conv2d(64, 64, 3, 1, 1)
def forward(self, input_0):
primals_2 = self.conv1.weight
primals_3 = self.conv1.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
e-dorigatti/torchinfo
|
RecursiveNet
| false | 12,335 |
[
"MIT"
] | 0 |
9fa0e677fb7002e89afd5b1bb372fe8c1dd813d6
|
https://github.com/e-dorigatti/torchinfo/tree/9fa0e677fb7002e89afd5b1bb372fe8c1dd813d6
|
MyLinear
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/fl/cflw6zjzdk2wqtau7m6nsei5vavjfijzxhb37zaa3xp4yxpw5yb2.py
# Topologically Sorted Source Nodes: [mul_1], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# mul_1 => mul_1
# Graph fragment:
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, 1), kwargs = {})
triton_poi_fused_mul_0 = async_compile.triton('triton_poi_fused_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2o/c2oqkq7zaubqmw7vuixxlseb2ff5jzqqbyczicxlmsahuxwdpdyp.py
# Topologically Sorted Source Nodes: [bias], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# bias => mul
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_1, 1), kwargs = {})
triton_poi_fused_mul_1 = async_compile.triton('triton_poi_fused_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, ), (1, ))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul_1], Original ATen: [aten.mul]
stream0 = get_raw_stream(0)
triton_poi_fused_mul_0.run(primals_2, buf0, 16, grid=grid(16), stream=stream0)
del primals_2
buf1 = empty_strided_cuda((4, ), (1, ), torch.float32)
# Topologically Sorted Source Nodes: [bias], Original ATen: [aten.mul]
triton_poi_fused_mul_1.run(primals_1, buf1, 4, grid=grid(4), stream=stream0)
del primals_1
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [bias, linear], Original ATen: [aten.mul, aten.addmm]
extern_kernels.addmm(buf1, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(buf0, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf2)
del buf0
del buf1
return (reinterpret_tensor(buf2, (4, 4, 4, 4), (64, 16, 4, 1), 0), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class MyLinear(nn.Module):
"""Linear layer with equalized learning rate and custom learning rate multiplier."""
def __init__(self, input_size, output_size, gain=2 ** 0.5, use_wscale=
False, lrmul=1, bias=True):
super().__init__()
he_std = gain * input_size ** -0.5
if use_wscale:
init_std = 1.0 / lrmul
self.w_mul = he_std * lrmul
else:
init_std = he_std / lrmul
self.w_mul = lrmul
self.weight = torch.nn.Parameter(torch.randn(output_size,
input_size) * init_std)
if bias:
self.bias = torch.nn.Parameter(torch.zeros(output_size))
self.b_mul = lrmul
else:
self.bias = None
def forward(self, x):
bias = self.bias
if bias is not None:
bias = bias * self.b_mul
return F.linear(x, self.weight * self.w_mul, bias)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'input_size': 4, 'output_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x0, tmp2, xmask)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x0, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4,), (1,))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_mul_0[grid(16)](primals_2, buf0, 16, XBLOCK=16,
num_warps=1, num_stages=1)
del primals_2
buf1 = empty_strided_cuda((4,), (1,), torch.float32)
triton_poi_fused_mul_1[grid(4)](primals_1, buf1, 4, XBLOCK=4,
num_warps=1, num_stages=1)
del primals_1
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(buf1, reinterpret_tensor(primals_3, (64, 4), (
4, 1), 0), reinterpret_tensor(buf0, (4, 4), (1, 4), 0), alpha=1,
beta=1, out=buf2)
del buf0
del buf1
return reinterpret_tensor(buf2, (4, 4, 4, 4), (64, 16, 4, 1), 0
), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0)
class MyLinearNew(nn.Module):
"""Linear layer with equalized learning rate and custom learning rate multiplier."""
def __init__(self, input_size, output_size, gain=2 ** 0.5, use_wscale=
False, lrmul=1, bias=True):
super().__init__()
he_std = gain * input_size ** -0.5
if use_wscale:
init_std = 1.0 / lrmul
self.w_mul = he_std * lrmul
else:
init_std = he_std / lrmul
self.w_mul = lrmul
self.weight = torch.nn.Parameter(torch.randn(output_size,
input_size) * init_std)
if bias:
self.bias = torch.nn.Parameter(torch.zeros(output_size))
self.b_mul = lrmul
else:
self.bias = None
def forward(self, input_0):
primals_2 = self.weight
primals_1 = self.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
eitanrich/ganspace-manifold
|
MyLinear
| false | 12,336 |
[
"Apache-2.0"
] | 0 |
148d5d30001c43794a40bbed885601e7816f5d7d
|
https://github.com/eitanrich/ganspace-manifold/tree/148d5d30001c43794a40bbed885601e7816f5d7d
|
BCELoss2d
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/jl/cjlfb7i5sda6iuixuufqhuq6ofisyq3co2j2rvs4i66aznhxymc2.py
# Topologically Sorted Source Nodes: [binary_cross_entropy], Original ATen: [aten.binary_cross_entropy]
# Source node to ATen node mapping:
# binary_cross_entropy => full_default, full_default_1, log, log1p, maximum, maximum_1, mean, mul, mul_1, neg, sub, sub_1
# Graph fragment:
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view_1, 1), kwargs = {})
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%view,), kwargs = {})
# %log1p : [num_users=1] = call_function[target=torch.ops.aten.log1p.default](args = (%neg,), kwargs = {})
# %full_default : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([], -100), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %maximum : [num_users=1] = call_function[target=torch.ops.aten.maximum.default](args = (%log1p, %full_default), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub, %maximum), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%view,), kwargs = {})
# %full_default_1 : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([], -100), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %maximum_1 : [num_users=1] = call_function[target=torch.ops.aten.maximum.default](args = (%log, %full_default_1), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_1, %maximum_1), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul, %mul_1), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%sub_1,), kwargs = {})
triton_per_fused_binary_cross_entropy_0 = async_compile.triton('triton_per_fused_binary_cross_entropy_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_binary_cross_entropy_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_binary_cross_entropy_0(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp3 = tl.load(in_ptr1 + (r0), None)
tmp1 = 1.0
tmp2 = tmp0 - tmp1
tmp4 = tl.sigmoid(tmp3)
tmp5 = -tmp4
tmp6 = libdevice.log1p(tmp5)
tmp7 = -100.0
tmp8 = triton_helpers.maximum(tmp6, tmp7)
tmp9 = tmp2 * tmp8
tmp10 = tl_math.log(tmp4)
tmp11 = triton_helpers.maximum(tmp10, tmp7)
tmp12 = tmp0 * tmp11
tmp13 = tmp9 - tmp12
tmp14 = tl.broadcast_to(tmp13, [RBLOCK])
tmp16 = triton_helpers.promote_to_tensor(tl.sum(tmp14, 0))
tmp17 = 256.0
tmp18 = tmp16 / tmp17
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp18, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [binary_cross_entropy], Original ATen: [aten.binary_cross_entropy]
stream0 = get_raw_stream(0)
triton_per_fused_binary_cross_entropy_0.run(buf1, arg1_1, arg0_1, 1, 256, grid=grid(1), stream=stream0)
del arg0_1
del arg1_1
return (buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn.functional as F
import torch.nn as nn
class BCELoss2d(nn.Module):
def __init__(self, weight=None, size_average=True):
"""
Imlements Binary Cross Entropy loss function.
"""
super(BCELoss2d, self).__init__()
self.bce_loss = nn.BCELoss(weight, size_average)
def forward(self, logits, targets):
probs = F.sigmoid(logits)
probs_flat = probs.view(-1)
targets_flat = targets.view(-1)
return self.bce_loss(probs_flat, targets_flat)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_binary_cross_entropy_0(in_out_ptr0, in_ptr0, in_ptr1,
xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp3 = tl.load(in_ptr1 + r0, None)
tmp1 = 1.0
tmp2 = tmp0 - tmp1
tmp4 = tl.sigmoid(tmp3)
tmp5 = -tmp4
tmp6 = libdevice.log1p(tmp5)
tmp7 = -100.0
tmp8 = triton_helpers.maximum(tmp6, tmp7)
tmp9 = tmp2 * tmp8
tmp10 = tl_math.log(tmp4)
tmp11 = triton_helpers.maximum(tmp10, tmp7)
tmp12 = tmp0 * tmp11
tmp13 = tmp9 - tmp12
tmp14 = tl.broadcast_to(tmp13, [RBLOCK])
tmp16 = triton_helpers.promote_to_tensor(tl.sum(tmp14, 0))
tmp17 = 256.0
tmp18 = tmp16 / tmp17
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp18, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0
del buf0
get_raw_stream(0)
triton_per_fused_binary_cross_entropy_0[grid(1)](buf1, arg1_1,
arg0_1, 1, 256, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
return buf1,
class BCELoss2dNew(nn.Module):
def __init__(self, weight=None, size_average=True):
"""
Imlements Binary Cross Entropy loss function.
"""
super(BCELoss2dNew, self).__init__()
self.bce_loss = nn.BCELoss(weight, size_average)
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
ekalyashov/segmentation-unet
|
BCELoss2d
| false | 12,337 |
[
"MIT"
] | 0 |
59dc95419481b2535a52332e0be92b15c7450674
|
https://github.com/ekalyashov/segmentation-unet/tree/59dc95419481b2535a52332e0be92b15c7450674
|
StyleMod
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/yg/cygzi6evso6kefobgrwgjcxh5qgbn7zouyf7to742xeh6bsflplb.py
# Topologically Sorted Source Nodes: [mul_1], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# mul_1 => mul_1
# Graph fragment:
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, 0.5), kwargs = {})
triton_poi_fused_mul_0 = async_compile.triton('triton_poi_fused_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 32
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 0.5
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2n/c2npkiljqvyo5boces3kqpxvha7qfatrulgz7qm2oitxblzqbabd.py
# Topologically Sorted Source Nodes: [add, mul_2, x], Original ATen: [aten.add, aten.mul]
# Source node to ATen node mapping:
# add => add
# mul_2 => mul_2
# x => add_1
# Graph fragment:
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%select, 1.0), kwargs = {})
# %mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_4, %add), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul_2, %select_1), kwargs = {})
triton_poi_fused_add_mul_1 = async_compile.triton('triton_poi_fused_add_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4096],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_mul_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_mul_1(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4096
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = (xindex // 16) % 4
x2 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x3), None)
tmp1 = tl.load(in_ptr1 + (x1 + (8*x2)), None, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr2 + (x1), None, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (4 + x1 + (8*x2)), None, eviction_policy='evict_last')
tmp9 = tl.load(in_ptr2 + (4 + x1), None, eviction_policy='evict_last')
tmp3 = 1.0
tmp4 = tmp2 * tmp3
tmp5 = tmp1 + tmp4
tmp6 = tmp5 + tmp3
tmp7 = tmp0 * tmp6
tmp10 = tmp9 * tmp3
tmp11 = tmp8 + tmp10
tmp12 = tmp7 + tmp11
tl.store(out_ptr0 + (x3), tmp12, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (8, ), (1, ))
assert_size_stride(primals_2, (8, 4), (4, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (64, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((8, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul_1], Original ATen: [aten.mul]
stream0 = get_raw_stream(0)
triton_poi_fused_mul_0.run(primals_2, buf0, 32, grid=grid(32), stream=stream0)
del primals_2
buf1 = empty_strided_cuda((64, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(buf0, (4, 8), (1, 4), 0), out=buf1)
del buf0
buf2 = empty_strided_cuda((64, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [add, mul_2, x], Original ATen: [aten.add, aten.mul]
triton_poi_fused_add_mul_1.run(primals_4, buf1, primals_1, buf2, 4096, grid=grid(4096), stream=stream0)
del buf1
del primals_1
return (buf2, primals_4, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((8, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((8, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((64, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class MyLinear(nn.Module):
"""Linear layer with equalized learning rate and custom learning rate multiplier."""
def __init__(self, input_size, output_size, gain=2 ** 0.5, use_wscale=
False, lrmul=1, bias=True):
super().__init__()
he_std = gain * input_size ** -0.5
if use_wscale:
init_std = 1.0 / lrmul
self.w_mul = he_std * lrmul
else:
init_std = he_std / lrmul
self.w_mul = lrmul
self.weight = torch.nn.Parameter(torch.randn(output_size,
input_size) * init_std)
if bias:
self.bias = torch.nn.Parameter(torch.zeros(output_size))
self.b_mul = lrmul
else:
self.bias = None
def forward(self, x):
bias = self.bias
if bias is not None:
bias = bias * self.b_mul
return F.linear(x, self.weight * self.w_mul, bias)
class StyleMod(nn.Module):
def __init__(self, latent_size, channels, use_wscale):
super(StyleMod, self).__init__()
self.lin = MyLinear(latent_size, channels * 2, gain=1.0, use_wscale
=use_wscale)
def forward(self, x, latent):
style = self.lin(latent)
shape = [-1, 2, x.size(1)] + (x.dim() - 2) * [1]
style = style.view(shape)
x = x * (style[:, 0] + 1.0) + style[:, 1]
return x
def get_inputs():
return [torch.rand([64, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'latent_size': 4, 'channels': 4, 'use_wscale': 1.0}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
import torch.nn.functional as F
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 32
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 0.5
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x0, tmp2, xmask)
@triton.jit
def triton_poi_fused_add_mul_1(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = xindex // 16 % 4
x2 = xindex // 64
tmp0 = tl.load(in_ptr0 + x3, None)
tmp1 = tl.load(in_ptr1 + (x1 + 8 * x2), None, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr2 + x1, None, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (4 + x1 + 8 * x2), None, eviction_policy=
'evict_last')
tmp9 = tl.load(in_ptr2 + (4 + x1), None, eviction_policy='evict_last')
tmp3 = 1.0
tmp4 = tmp2 * tmp3
tmp5 = tmp1 + tmp4
tmp6 = tmp5 + tmp3
tmp7 = tmp0 * tmp6
tmp10 = tmp9 * tmp3
tmp11 = tmp8 + tmp10
tmp12 = tmp7 + tmp11
tl.store(out_ptr0 + x3, tmp12, None)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (8,), (1,))
assert_size_stride(primals_2, (8, 4), (4, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (64, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((8, 4), (4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_mul_0[grid(32)](primals_2, buf0, 32, XBLOCK=32,
num_warps=1, num_stages=1)
del primals_2
buf1 = empty_strided_cuda((64, 8), (8, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0),
reinterpret_tensor(buf0, (4, 8), (1, 4), 0), out=buf1)
del buf0
buf2 = empty_strided_cuda((64, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_add_mul_1[grid(4096)](primals_4, buf1, primals_1,
buf2, 4096, XBLOCK=256, num_warps=4, num_stages=1)
del buf1
del primals_1
return buf2, primals_4, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0)
class MyLinear(nn.Module):
"""Linear layer with equalized learning rate and custom learning rate multiplier."""
def __init__(self, input_size, output_size, gain=2 ** 0.5, use_wscale=
False, lrmul=1, bias=True):
super().__init__()
he_std = gain * input_size ** -0.5
if use_wscale:
init_std = 1.0 / lrmul
self.w_mul = he_std * lrmul
else:
init_std = he_std / lrmul
self.w_mul = lrmul
self.weight = torch.nn.Parameter(torch.randn(output_size,
input_size) * init_std)
if bias:
self.bias = torch.nn.Parameter(torch.zeros(output_size))
self.b_mul = lrmul
else:
self.bias = None
def forward(self, x):
bias = self.bias
if bias is not None:
bias = bias * self.b_mul
return F.linear(x, self.weight * self.w_mul, bias)
class StyleModNew(nn.Module):
def __init__(self, latent_size, channels, use_wscale):
super(StyleModNew, self).__init__()
self.lin = MyLinear(latent_size, channels * 2, gain=1.0, use_wscale
=use_wscale)
def forward(self, input_0, input_1):
primals_2 = self.lin.weight
primals_1 = self.lin.bias
primals_4 = input_0
primals_3 = input_1
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
eitanrich/ganspace-manifold
|
StyleMod
| false | 12,338 |
[
"Apache-2.0"
] | 0 |
148d5d30001c43794a40bbed885601e7816f5d7d
|
https://github.com/eitanrich/ganspace-manifold/tree/148d5d30001c43794a40bbed885601e7816f5d7d
|
ProteinResNetPooler
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/rq/crq67an2qbnihu4lgijwmlkbxtlerrf4ah6jnhhssto73bmo5zvy.py
# Topologically Sorted Source Nodes: [attention_weights], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# attention_weights => amax, div, exp, sub, sum_1
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%view_1, [-1], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view_1, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [-1], True), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_0 = async_compile.triton('triton_poi_fused__softmax_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tmp0 - tmp0
tmp2 = tl_math.exp(tmp1)
tmp3 = tmp2 / tmp2
tl.store(out_ptr0 + (x0), tmp3, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2k/c2kiox2wvshockbbzjlycxwhjeigavlrfwuvcpbcbxpipbm7d7k6.py
# Topologically Sorted Source Nodes: [pooled_output_1], Original ATen: [aten.tanh]
# Source node to ATen node mapping:
# pooled_output_1 => tanh
# Graph fragment:
# %add_tensor : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mm_default, %primals_5), kwargs = {})
# %tanh : [num_users=1] = call_function[target=torch.ops.aten.tanh.default](args = (%add_tensor,), kwargs = {})
triton_poi_fused_tanh_1 = async_compile.triton('triton_poi_fused_tanh_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_tanh_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_tanh_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x2), tmp3, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (1, 4), (4, 1))
assert_size_stride(primals_2, (1, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf1 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [attention_scores], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 1), (1, 4), 0), alpha=1, beta=1, out=buf1)
del primals_1
del primals_2
buf2 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
# Topologically Sorted Source Nodes: [attention_weights], Original ATen: [aten._softmax]
stream0 = get_raw_stream(0)
triton_poi_fused__softmax_0.run(buf1, buf2, 16, grid=grid(16), stream=stream0)
buf3 = empty_strided_cuda((4, 4, 1), (4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [attention_weights, matmul], Original ATen: [aten._softmax, aten.bmm]
extern_kernels.bmm(reinterpret_tensor(primals_3, (4, 4, 4), (16, 1, 4), 0), buf2, out=buf3)
buf4 = reinterpret_tensor(buf2, (4, 4), (4, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf3, (4, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf4)
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [pooled_output_1], Original ATen: [aten.tanh]
triton_poi_fused_tanh_1.run(buf5, primals_5, 16, grid=grid(16), stream=stream0)
del primals_5
return (buf5, primals_3, buf1, reinterpret_tensor(buf3, (4, 4), (4, 1), 0), buf5, primals_4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from _paritybench_helpers import _mock_config
import torch
import torch.nn as nn
class ProteinResNetPooler(nn.Module):
def __init__(self, config):
super().__init__()
self.attention_weights = nn.Linear(config.hidden_size, 1)
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
self.activation = nn.Tanh()
def forward(self, hidden_states, mask=None):
attention_scores = self.attention_weights(hidden_states)
if mask is not None:
attention_scores += -10000.0 * (1 - mask)
attention_weights = torch.softmax(attention_scores, -1)
weighted_mean_embedding = torch.matmul(hidden_states.transpose(1, 2
), attention_weights).squeeze(2)
pooled_output = self.dense(weighted_mean_embedding)
pooled_output = self.activation(pooled_output)
return pooled_output
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'config': _mock_config(hidden_size=4)}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused__softmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tmp0 - tmp0
tmp2 = tl_math.exp(tmp1)
tmp3 = tmp2 / tmp2
tl.store(out_ptr0 + x0, tmp3, xmask)
@triton.jit
def triton_poi_fused_tanh_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x2, tmp3, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (1, 4), (4, 1))
assert_size_stride(primals_2, (1,), (1,))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf1 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (16,
4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 1), (1, 4), 0
), alpha=1, beta=1, out=buf1)
del primals_1
del primals_2
buf2 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
get_raw_stream(0)
triton_poi_fused__softmax_0[grid(16)](buf1, buf2, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf3 = empty_strided_cuda((4, 4, 1), (4, 1, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(primals_3, (4, 4, 4), (16, 1,
4), 0), buf2, out=buf3)
buf4 = reinterpret_tensor(buf2, (4, 4), (4, 1), 0)
del buf2
extern_kernels.mm(reinterpret_tensor(buf3, (4, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf4)
buf5 = buf4
del buf4
triton_poi_fused_tanh_1[grid(16)](buf5, primals_5, 16, XBLOCK=16,
num_warps=1, num_stages=1)
del primals_5
return buf5, primals_3, buf1, reinterpret_tensor(buf3, (4, 4), (4, 1), 0
), buf5, primals_4
class ProteinResNetPoolerNew(nn.Module):
def __init__(self, config):
super().__init__()
self.attention_weights = nn.Linear(config.hidden_size, 1)
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
self.activation = nn.Tanh()
def forward(self, input_0):
primals_1 = self.attention_weights.weight
primals_2 = self.attention_weights.bias
primals_4 = self.dense.weight
primals_5 = self.dense.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4, primals_5])
return output[0]
|
ekvall93/tape
|
ProteinResNetPooler
| false | 12,339 |
[
"BSD-3-Clause"
] | 0 |
1ca4d5a39c72f806f23a36fb7a7c7325f06096ae
|
https://github.com/ekvall93/tape/tree/1ca4d5a39c72f806f23a36fb7a7c7325f06096ae
|
Accuracy
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/jd/cjdnaobowqernajenb4axacuvfjpuu3bnchwmpvbem4jevqy6y4v.py
# Topologically Sorted Source Nodes: [predictions], Original ATen: [aten.argmax]
# Source node to ATen node mapping:
# predictions => argmax
# Graph fragment:
# %argmax : [num_users=1] = call_function[target=torch.ops.aten.argmax.default](args = (%arg0_1, -1), kwargs = {})
triton_poi_fused_argmax_0 = async_compile.triton('triton_poi_fused_argmax_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i64', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_argmax_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_argmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (4*x0), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp17 = tl.load(in_ptr0 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp32 = tl.load(in_ptr0 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp2 = tmp0 > tmp1
tmp3 = tmp0 == tmp1
tmp4 = tmp0 != tmp0
tmp5 = tmp1 != tmp1
tmp6 = tmp4 > tmp5
tmp7 = tmp2 | tmp6
tmp8 = tmp4 & tmp5
tmp9 = tmp3 | tmp8
tmp10 = tl.full([1], 0, tl.int64)
tmp11 = tl.full([1], 1, tl.int64)
tmp12 = tmp10 < tmp11
tmp13 = tmp9 & tmp12
tmp14 = tmp7 | tmp13
tmp15 = tl.where(tmp14, tmp0, tmp1)
tmp16 = tl.where(tmp14, tmp10, tmp11)
tmp18 = tmp15 > tmp17
tmp19 = tmp15 == tmp17
tmp20 = tmp15 != tmp15
tmp21 = tmp17 != tmp17
tmp22 = tmp20 > tmp21
tmp23 = tmp18 | tmp22
tmp24 = tmp20 & tmp21
tmp25 = tmp19 | tmp24
tmp26 = tl.full([1], 2, tl.int64)
tmp27 = tmp16 < tmp26
tmp28 = tmp25 & tmp27
tmp29 = tmp23 | tmp28
tmp30 = tl.where(tmp29, tmp15, tmp17)
tmp31 = tl.where(tmp29, tmp16, tmp26)
tmp33 = tmp30 > tmp32
tmp34 = tmp30 == tmp32
tmp35 = tmp30 != tmp30
tmp36 = tmp32 != tmp32
tmp37 = tmp35 > tmp36
tmp38 = tmp33 | tmp37
tmp39 = tmp35 & tmp36
tmp40 = tmp34 | tmp39
tmp41 = tl.full([1], 3, tl.int64)
tmp42 = tmp31 < tmp41
tmp43 = tmp40 & tmp42
tmp44 = tmp38 | tmp43
tmp45 = tl.where(tmp44, tmp30, tmp32)
tmp46 = tl.where(tmp44, tmp31, tmp41)
tl.store(out_ptr0 + (x0), tmp46, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/vd/cvdkoxf2wkd5oryi23fngtr24eavrkiw46qraujejvcpyvdhsfen.py
# Topologically Sorted Source Nodes: [eq, valid_mask, correct, sum_1, float_2, sum_2, float_3, truediv], Original ATen: [aten.eq, aten.ne, aten.mul, aten.sum, aten._to_copy, aten.div]
# Source node to ATen node mapping:
# correct => mul
# eq => eq
# float_2 => convert_element_type
# float_3 => convert_element_type_1
# sum_1 => sum_1
# sum_2 => sum_2
# truediv => div
# valid_mask => ne
# Graph fragment:
# %eq : [num_users=1] = call_function[target=torch.ops.aten.eq.Tensor](args = (%argmax, %arg1_1), kwargs = {})
# %ne : [num_users=2] = call_function[target=torch.ops.aten.ne.Scalar](args = (%arg1_1, -100), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%eq, %ne), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%mul,), kwargs = {})
# %convert_element_type : [num_users=1] = call_function[target=torch.ops.prims.convert_element_type.default](args = (%sum_1, torch.float32), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%ne,), kwargs = {})
# %convert_element_type_1 : [num_users=1] = call_function[target=torch.ops.prims.convert_element_type.default](args = (%sum_2, torch.float32), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%convert_element_type, %convert_element_type_1), kwargs = {})
triton_per_fused__to_copy_div_eq_mul_ne_sum_1 = async_compile.triton('triton_per_fused__to_copy_div_eq_mul_ne_sum_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*i64', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused__to_copy_div_eq_mul_ne_sum_1', 'mutated_arg_names': [], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 2, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused__to_copy_div_eq_mul_ne_sum_1(in_ptr0, in_ptr1, out_ptr2, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex % 64
r2 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr1 + (r2), None)
tmp1 = tmp0.to(tl.float32)
tmp3 = tmp1 == tmp2
tmp4 = -100.0
tmp5 = tmp2 != tmp4
tmp6 = tmp3 & tmp5
tmp7 = tmp6.to(tl.int64)
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = tmp5.to(tl.int64)
tmp12 = tl.broadcast_to(tmp11, [RBLOCK])
tmp14 = triton_helpers.promote_to_tensor(tl.sum(tmp12, 0))
tmp15 = tmp10.to(tl.float32)
tmp16 = tmp14.to(tl.float32)
tmp17 = tmp15 / tmp16
tl.store(out_ptr2 + (tl.full([1], 0, tl.int32)), tmp17, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.int64)
# Topologically Sorted Source Nodes: [predictions], Original ATen: [aten.argmax]
stream0 = get_raw_stream(0)
triton_poi_fused_argmax_0.run(arg0_1, buf0, 64, grid=grid(64), stream=stream0)
del arg0_1
buf3 = empty_strided_cuda((), (), torch.float32)
# Topologically Sorted Source Nodes: [eq, valid_mask, correct, sum_1, float_2, sum_2, float_3, truediv], Original ATen: [aten.eq, aten.ne, aten.mul, aten.sum, aten._to_copy, aten.div]
triton_per_fused__to_copy_div_eq_mul_ne_sum_1.run(buf0, arg1_1, buf3, 1, 256, grid=grid(1), stream=stream0)
del arg1_1
del buf0
return (buf3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
def accuracy(logits, labels, ignore_index: 'int'=-100):
with torch.no_grad():
valid_mask = labels != ignore_index
predictions = logits.float().argmax(-1)
correct = (predictions == labels) * valid_mask
return correct.sum().float() / valid_mask.sum().float()
class Accuracy(nn.Module):
def __init__(self, ignore_index: 'int'=-100):
super().__init__()
self.ignore_index = ignore_index
def forward(self, inputs, target):
return accuracy(inputs, target, self.ignore_index)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_argmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 4 * x0, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp17 = tl.load(in_ptr0 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp32 = tl.load(in_ptr0 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp2 = tmp0 > tmp1
tmp3 = tmp0 == tmp1
tmp4 = tmp0 != tmp0
tmp5 = tmp1 != tmp1
tmp6 = tmp4 > tmp5
tmp7 = tmp2 | tmp6
tmp8 = tmp4 & tmp5
tmp9 = tmp3 | tmp8
tmp10 = tl.full([1], 0, tl.int64)
tmp11 = tl.full([1], 1, tl.int64)
tmp12 = tmp10 < tmp11
tmp13 = tmp9 & tmp12
tmp14 = tmp7 | tmp13
tmp15 = tl.where(tmp14, tmp0, tmp1)
tmp16 = tl.where(tmp14, tmp10, tmp11)
tmp18 = tmp15 > tmp17
tmp19 = tmp15 == tmp17
tmp20 = tmp15 != tmp15
tmp21 = tmp17 != tmp17
tmp22 = tmp20 > tmp21
tmp23 = tmp18 | tmp22
tmp24 = tmp20 & tmp21
tmp25 = tmp19 | tmp24
tmp26 = tl.full([1], 2, tl.int64)
tmp27 = tmp16 < tmp26
tmp28 = tmp25 & tmp27
tmp29 = tmp23 | tmp28
tmp30 = tl.where(tmp29, tmp15, tmp17)
tmp31 = tl.where(tmp29, tmp16, tmp26)
tmp33 = tmp30 > tmp32
tmp34 = tmp30 == tmp32
tmp35 = tmp30 != tmp30
tmp36 = tmp32 != tmp32
tmp37 = tmp35 > tmp36
tmp38 = tmp33 | tmp37
tmp39 = tmp35 & tmp36
tmp40 = tmp34 | tmp39
tmp41 = tl.full([1], 3, tl.int64)
tmp42 = tmp31 < tmp41
tmp43 = tmp40 & tmp42
tmp44 = tmp38 | tmp43
tl.where(tmp44, tmp30, tmp32)
tmp46 = tl.where(tmp44, tmp31, tmp41)
tl.store(out_ptr0 + x0, tmp46, xmask)
@triton.jit
def triton_per_fused__to_copy_div_eq_mul_ne_sum_1(in_ptr0, in_ptr1,
out_ptr2, xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex % 64
r2 = rindex
tmp0 = tl.load(in_ptr0 + r0, None, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr1 + r2, None)
tmp1 = tmp0.to(tl.float32)
tmp3 = tmp1 == tmp2
tmp4 = -100.0
tmp5 = tmp2 != tmp4
tmp6 = tmp3 & tmp5
tmp7 = tmp6.to(tl.int64)
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = tmp5.to(tl.int64)
tmp12 = tl.broadcast_to(tmp11, [RBLOCK])
tmp14 = triton_helpers.promote_to_tensor(tl.sum(tmp12, 0))
tmp15 = tmp10.to(tl.float32)
tmp16 = tmp14.to(tl.float32)
tmp17 = tmp15 / tmp16
tl.store(out_ptr2 + tl.full([1], 0, tl.int32), tmp17, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.int64)
get_raw_stream(0)
triton_poi_fused_argmax_0[grid(64)](arg0_1, buf0, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del arg0_1
buf3 = empty_strided_cuda((), (), torch.float32)
triton_per_fused__to_copy_div_eq_mul_ne_sum_1[grid(1)](buf0, arg1_1,
buf3, 1, 256, num_warps=2, num_stages=1)
del arg1_1
del buf0
return buf3,
def accuracy(logits, labels, ignore_index: 'int'=-100):
with torch.no_grad():
valid_mask = labels != ignore_index
predictions = logits.float().argmax(-1)
correct = (predictions == labels) * valid_mask
return correct.sum().float() / valid_mask.sum().float()
class AccuracyNew(nn.Module):
def __init__(self, ignore_index: 'int'=-100):
super().__init__()
self.ignore_index = ignore_index
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
ekvall93/tape
|
Accuracy
| false | 12,340 |
[
"BSD-3-Clause"
] | 0 |
1ca4d5a39c72f806f23a36fb7a7c7325f06096ae
|
https://github.com/ekvall93/tape/tree/1ca4d5a39c72f806f23a36fb7a7c7325f06096ae
|
DNNModel
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/r3/cr3febcwm3t44fuoitsx3ou2p6xg4sk4f7unagmmrvffasxf47te.py
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# x => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_1,), kwargs = {})
# %le_2 : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_relu_threshold_backward_0 = async_compile.triton('triton_poi_fused_relu_threshold_backward_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i1', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_threshold_backward_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_out_ptr0, in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(in_out_ptr0 + (x2), tmp4, xmask)
tl.store(out_ptr0 + (x2), tmp6, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4, ), (1, ))
assert_size_stride(primals_8, (3, 4), (4, 1))
assert_size_stride(primals_9, (3, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf0 # reuse
buf9 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.relu, aten.threshold_backward]
stream0 = get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0.run(buf1, primals_2, buf9, 256, grid=grid(256), stream=stream0)
del primals_2
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf1, (64, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf2)
buf3 = reinterpret_tensor(buf2, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf2 # reuse
buf8 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.relu, aten.threshold_backward]
triton_poi_fused_relu_threshold_backward_0.run(buf3, primals_5, buf8, 256, grid=grid(256), stream=stream0)
del primals_5
buf4 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf4)
buf5 = reinterpret_tensor(buf4, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf4 # reuse
buf7 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [x_2], Original ATen: [aten.relu, aten.threshold_backward]
triton_poi_fused_relu_threshold_backward_0.run(buf5, primals_7, buf7, 256, grid=grid(256), stream=stream0)
del primals_7
buf6 = empty_strided_cuda((64, 3), (3, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_9, reinterpret_tensor(buf5, (64, 4), (4, 1), 0), reinterpret_tensor(primals_8, (4, 3), (1, 4), 0), alpha=1, beta=1, out=buf6)
del primals_9
return (reinterpret_tensor(buf6, (4, 4, 4, 3), (48, 12, 3, 1), 0), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(buf1, (64, 4), (4, 1), 0), reinterpret_tensor(buf3, (64, 4), (4, 1), 0), reinterpret_tensor(buf5, (64, 4), (4, 1), 0), primals_8, buf7, primals_6, buf8, primals_4, buf9, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((3, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((3, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
class DNNModel(nn.Module):
def __init__(self, dropout=0.2):
super(DNNModel, self).__init__()
self.fc1 = nn.Linear(4, 4)
self.relu1 = nn.ReLU()
self.dropout1 = nn.Dropout(p=dropout)
self.fc2 = nn.Linear(4, 4)
self.relu2 = nn.ReLU()
self.dropout2 = nn.Dropout(p=dropout)
self.fc3 = nn.Linear(4, 4)
self.relu3 = nn.ReLU()
self.dropout3 = nn.Dropout(p=dropout)
self.relu_last = nn.ReLU()
self.fc_last = nn.Linear(4, 3)
def forward(self, x):
x = self.relu1(self.fc1(x))
x = self.relu2(self.fc2(x))
x = self.relu3(self.fc3(x))
x = self.fc_last(x)
return x
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_out_ptr0, in_ptr0,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(in_out_ptr0 + x2, tmp4, xmask)
tl.store(out_ptr0 + x2, tmp6, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4,), (1,))
assert_size_stride(primals_8, (3, 4), (4, 1))
assert_size_stride(primals_9, (3,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf0
buf9 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0[grid(256)](buf1,
primals_2, buf9, 256, XBLOCK=128, num_warps=4, num_stages=1)
del primals_2
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf1, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf2)
buf3 = reinterpret_tensor(buf2, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf2
buf8 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
triton_poi_fused_relu_threshold_backward_0[grid(256)](buf3,
primals_5, buf8, 256, XBLOCK=128, num_warps=4, num_stages=1)
del primals_5
buf4 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf3, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf4)
buf5 = reinterpret_tensor(buf4, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf4
buf7 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
triton_poi_fused_relu_threshold_backward_0[grid(256)](buf5,
primals_7, buf7, 256, XBLOCK=128, num_warps=4, num_stages=1)
del primals_7
buf6 = empty_strided_cuda((64, 3), (3, 1), torch.float32)
extern_kernels.addmm(primals_9, reinterpret_tensor(buf5, (64, 4), (
4, 1), 0), reinterpret_tensor(primals_8, (4, 3), (1, 4), 0),
alpha=1, beta=1, out=buf6)
del primals_9
return reinterpret_tensor(buf6, (4, 4, 4, 3), (48, 12, 3, 1), 0
), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0
), reinterpret_tensor(buf1, (64, 4), (4, 1), 0), reinterpret_tensor(
buf3, (64, 4), (4, 1), 0), reinterpret_tensor(buf5, (64, 4), (4, 1), 0
), primals_8, buf7, primals_6, buf8, primals_4, buf9
class DNNModelNew(nn.Module):
def __init__(self, dropout=0.2):
super(DNNModelNew, self).__init__()
self.fc1 = nn.Linear(4, 4)
self.relu1 = nn.ReLU()
self.dropout1 = nn.Dropout(p=dropout)
self.fc2 = nn.Linear(4, 4)
self.relu2 = nn.ReLU()
self.dropout2 = nn.Dropout(p=dropout)
self.fc3 = nn.Linear(4, 4)
self.relu3 = nn.ReLU()
self.dropout3 = nn.Dropout(p=dropout)
self.relu_last = nn.ReLU()
self.fc_last = nn.Linear(4, 3)
def forward(self, input_0):
primals_1 = self.fc1.weight
primals_2 = self.fc1.bias
primals_4 = self.fc2.weight
primals_5 = self.fc2.bias
primals_6 = self.fc3.weight
primals_7 = self.fc3.bias
primals_8 = self.fc_last.weight
primals_9 = self.fc_last.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9])
return output[0]
|
ehsangolshani/workload-to-metric-mapper
|
DNNModel
| false | 12,341 |
[
"Apache-2.0"
] | 0 |
4c2825696200748382247909f2f777f49bf62cf0
|
https://github.com/ehsangolshani/workload-to-metric-mapper/tree/4c2825696200748382247909f2f777f49bf62cf0
|
SoftDiceLoss
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/gd/cgdtd7ki7lurypoeyfwjebdfquygdeupjef4ltfbbbdk5u7owcpl.py
# Topologically Sorted Source Nodes: [intersection, sum_1, sum_2, sum_3], Original ATen: [aten.mul, aten.sum]
# Source node to ATen node mapping:
# intersection => mul
# sum_1 => sum_1
# sum_2 => sum_2
# sum_3 => sum_3
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view, %view_1), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul, [1]), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%view, [1]), kwargs = {})
# %sum_3 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%view_1, [1]), kwargs = {})
triton_per_fused_mul_sum_0 = async_compile.triton('triton_per_fused_mul_sum_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[4, 64],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: 'i32', 6: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 6), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_mul_sum_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 3, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_mul_sum_0(in_ptr0, in_ptr1, out_ptr0, out_ptr1, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 4
rnumel = 64
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (64*x0)), xmask, other=0.0)
tmp2 = tl.load(in_ptr1 + (r1 + (64*x0)), xmask, other=0.0)
tmp1 = tl.sigmoid(tmp0)
tmp3 = tmp1 * tmp2
tmp4 = tl.broadcast_to(tmp3, [XBLOCK, RBLOCK])
tmp6 = tl.where(xmask, tmp4, 0)
tmp7 = tl.sum(tmp6, 1)[:, None]
tmp8 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp10 = tl.where(xmask, tmp8, 0)
tmp11 = tl.sum(tmp10, 1)[:, None]
tmp12 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp14 = tl.where(xmask, tmp12, 0)
tmp15 = tl.sum(tmp14, 1)[:, None]
tl.store(out_ptr0 + (x0), tmp7, xmask)
tl.store(out_ptr1 + (x0), tmp11, xmask)
tl.store(out_ptr2 + (x0), tmp15, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/wf/cwfp66pdiuvpu7ia4v5ln2nftqomh4gdpfux44irhboxg3weyrzz.py
# Topologically Sorted Source Nodes: [mul_1, add, add_1, add_2, score, sum_4, truediv_1, score_1], Original ATen: [aten.mul, aten.add, aten.div, aten.sum, aten.rsub]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# add_2 => add_2
# mul_1 => mul_1
# score => div
# score_1 => sub
# sum_4 => sum_4
# truediv_1 => div_1
# Graph fragment:
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sum_1, 2.0), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul_1, 1), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sum_2, %sum_3), kwargs = {})
# %add_2 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%add_1, 1), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%add, %add_2), kwargs = {})
# %sum_4 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%div,), kwargs = {})
# %div_1 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sum_4, 4), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (1, %div_1), kwargs = {})
triton_per_fused_add_div_mul_rsub_sum_1 = async_compile.triton('triton_per_fused_add_div_mul_rsub_sum_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 4],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {4: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=(4,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_div_mul_rsub_sum_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_div_mul_rsub_sum_1(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 1
rnumel = 4
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp5 = tl.load(in_ptr1 + (r0), None)
tmp6 = tl.load(in_ptr2 + (r0), None)
tmp1 = 2.0
tmp2 = tmp0 * tmp1
tmp3 = 1.0
tmp4 = tmp2 + tmp3
tmp7 = tmp5 + tmp6
tmp8 = tmp7 + tmp3
tmp9 = tmp4 / tmp8
tmp10 = tl.broadcast_to(tmp9, [XBLOCK, RBLOCK])
tmp12 = tl.sum(tmp10, 1)[:, None]
tmp13 = 0.25
tmp14 = tmp12 * tmp13
tmp15 = tmp3 - tmp14
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([XBLOCK, 1], 0, tl.int32)), tmp15, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, ), (1, ), torch.float32)
buf1 = empty_strided_cuda((4, ), (1, ), torch.float32)
buf2 = empty_strided_cuda((4, ), (1, ), torch.float32)
# Topologically Sorted Source Nodes: [intersection, sum_1, sum_2, sum_3], Original ATen: [aten.mul, aten.sum]
stream0 = get_raw_stream(0)
triton_per_fused_mul_sum_0.run(arg1_1, arg0_1, buf0, buf1, buf2, 4, 64, grid=grid(4), stream=stream0)
del arg0_1
del arg1_1
buf3 = empty_strided_cuda((), (), torch.float32)
buf4 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [mul_1, add, add_1, add_2, score, sum_4, truediv_1, score_1], Original ATen: [aten.mul, aten.add, aten.div, aten.sum, aten.rsub]
triton_per_fused_add_div_mul_rsub_sum_1.run(buf4, buf0, buf1, buf2, 1, 4, grid=grid(1), stream=stream0)
del buf0
del buf1
del buf2
return (buf4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn.functional as F
import torch.nn as nn
class SoftDiceLoss(nn.Module):
def __init__(self, weight=None, size_average=True):
"""
Imlements Dice loss function (using Sørensen–Dice coefficient).
"""
super(SoftDiceLoss, self).__init__()
def forward(self, logits, targets):
num = targets.size(0)
probs = F.sigmoid(logits)
m1 = probs.view(num, -1)
m2 = targets.view(num, -1)
intersection = m1 * m2
score = (2.0 * intersection.sum(1) + 1) / (m1.sum(1) + m2.sum(1) + 1)
score = 1 - score.sum() / num
return score
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_mul_sum_0(in_ptr0, in_ptr1, out_ptr0, out_ptr1,
out_ptr2, xnumel, rnumel, XBLOCK: tl.constexpr):
xnumel = 4
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 64 * x0), xmask, other=0.0)
tmp2 = tl.load(in_ptr1 + (r1 + 64 * x0), xmask, other=0.0)
tmp1 = tl.sigmoid(tmp0)
tmp3 = tmp1 * tmp2
tmp4 = tl.broadcast_to(tmp3, [XBLOCK, RBLOCK])
tmp6 = tl.where(xmask, tmp4, 0)
tmp7 = tl.sum(tmp6, 1)[:, None]
tmp8 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp10 = tl.where(xmask, tmp8, 0)
tmp11 = tl.sum(tmp10, 1)[:, None]
tmp12 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp14 = tl.where(xmask, tmp12, 0)
tmp15 = tl.sum(tmp14, 1)[:, None]
tl.store(out_ptr0 + x0, tmp7, xmask)
tl.store(out_ptr1 + x0, tmp11, xmask)
tl.store(out_ptr2 + x0, tmp15, xmask)
@triton.jit
def triton_per_fused_add_div_mul_rsub_sum_1(in_out_ptr0, in_ptr0, in_ptr1,
in_ptr2, xnumel, rnumel, XBLOCK: tl.constexpr):
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp5 = tl.load(in_ptr1 + r0, None)
tmp6 = tl.load(in_ptr2 + r0, None)
tmp1 = 2.0
tmp2 = tmp0 * tmp1
tmp3 = 1.0
tmp4 = tmp2 + tmp3
tmp7 = tmp5 + tmp6
tmp8 = tmp7 + tmp3
tmp9 = tmp4 / tmp8
tmp10 = tl.broadcast_to(tmp9, [XBLOCK, RBLOCK])
tmp12 = tl.sum(tmp10, 1)[:, None]
tmp13 = 0.25
tmp14 = tmp12 * tmp13
tmp15 = tmp3 - tmp14
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([XBLOCK, 1], 0, tl.int32), tmp15, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4,), (1,), torch.float32)
buf1 = empty_strided_cuda((4,), (1,), torch.float32)
buf2 = empty_strided_cuda((4,), (1,), torch.float32)
get_raw_stream(0)
triton_per_fused_mul_sum_0[grid(4)](arg1_1, arg0_1, buf0, buf1,
buf2, 4, 64, XBLOCK=1, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
buf3 = empty_strided_cuda((), (), torch.float32)
buf4 = buf3
del buf3
triton_per_fused_add_div_mul_rsub_sum_1[grid(1)](buf4, buf0, buf1,
buf2, 1, 4, XBLOCK=1, num_warps=2, num_stages=1)
del buf0
del buf1
del buf2
return buf4,
class SoftDiceLossNew(nn.Module):
def __init__(self, weight=None, size_average=True):
"""
Imlements Dice loss function (using Sørensen–Dice coefficient).
"""
super(SoftDiceLossNew, self).__init__()
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
ekalyashov/segmentation-unet
|
SoftDiceLoss
| false | 12,342 |
[
"MIT"
] | 0 |
59dc95419481b2535a52332e0be92b15c7450674
|
https://github.com/ekalyashov/segmentation-unet/tree/59dc95419481b2535a52332e0be92b15c7450674
|
LeNet_300_100
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/gi/cgisfp7cjgya6wz3zy6wgsfrei7pmk7oec6olmpydu3ckl4g7flh.py
# Topologically Sorted Source Nodes: [x1], Original ATen: [aten.relu]
# Source node to ATen node mapping:
# x1 => relu
# Graph fragment:
# %add_tensor_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mm_default_1, %primals_3), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%add_tensor_1,), kwargs = {})
triton_poi_fused_relu_0 = async_compile.triton('triton_poi_fused_relu_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[2048],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1200
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 300
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x2), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/d7/cd7rilnjd42cirsc5dhnnwlficmjz5omrtsdfojgouhplcpynn4n.py
# Topologically Sorted Source Nodes: [x2], Original ATen: [aten.relu]
# Source node to ATen node mapping:
# x2 => relu_1
# Graph fragment:
# %add_tensor : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mm_default, %primals_5), kwargs = {})
# %relu_1 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%add_tensor,), kwargs = {})
triton_poi_fused_relu_1 = async_compile.triton('triton_poi_fused_relu_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[512],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 400
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 100
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x2), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/fq/cfqeaxpwvkwok3wdzj4x74nonwbnkdl7jv6hborhhhugwqldix3l.py
# Topologically Sorted Source Nodes: [log_softmax], Original ATen: [aten._log_softmax]
# Source node to ATen node mapping:
# log_softmax => amax, exp, log, sub, sub_1, sum_1
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%addmm_2, [1], True), kwargs = {})
# %sub : [num_users=2] = call_function[target=torch.ops.aten.sub.Tensor](args = (%addmm_2, %amax), kwargs = {})
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [1], True), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%sum_1,), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sub, %log), kwargs = {})
triton_per_fused__log_softmax_2 = async_compile.triton('triton_per_fused__log_softmax_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[4, 16],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused__log_softmax_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 2, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused__log_softmax_2(in_ptr0, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 4
rnumel = 10
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = rindex < rnumel
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (10*x0)), rmask & xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(rmask & xmask, tmp1, float("-inf"))
tmp4 = triton_helpers.max2(tmp3, 1)[:, None]
tmp5 = tmp0 - tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK, RBLOCK])
tmp9 = tl.where(rmask & xmask, tmp7, 0)
tmp10 = tl.sum(tmp9, 1)[:, None]
tmp11 = tl_math.log(tmp10)
tmp12 = tmp5 - tmp11
tl.store(out_ptr2 + (r1 + (10*x0)), tmp12, rmask & xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7 = args
args.clear()
assert_size_stride(primals_1, (4, 784), (784, 1))
assert_size_stride(primals_2, (300, 784), (784, 1))
assert_size_stride(primals_3, (300, ), (1, ))
assert_size_stride(primals_4, (100, 300), (300, 1))
assert_size_stride(primals_5, (100, ), (1, ))
assert_size_stride(primals_6, (10, 100), (100, 1))
assert_size_stride(primals_7, (10, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 300), (300, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(primals_1, reinterpret_tensor(primals_2, (784, 300), (1, 784), 0), out=buf0)
del primals_2
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [x1], Original ATen: [aten.relu]
stream0 = get_raw_stream(0)
triton_poi_fused_relu_0.run(buf1, primals_3, 1200, grid=grid(1200), stream=stream0)
del primals_3
buf2 = empty_strided_cuda((4, 100), (100, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(buf1, reinterpret_tensor(primals_4, (300, 100), (1, 300), 0), out=buf2)
buf3 = buf2; del buf2 # reuse
# Topologically Sorted Source Nodes: [x2], Original ATen: [aten.relu]
triton_poi_fused_relu_1.run(buf3, primals_5, 400, grid=grid(400), stream=stream0)
del primals_5
buf4 = empty_strided_cuda((4, 10), (10, 1), torch.float32)
# Topologically Sorted Source Nodes: [x3], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_7, buf3, reinterpret_tensor(primals_6, (100, 10), (1, 100), 0), alpha=1, beta=1, out=buf4)
del primals_7
buf7 = empty_strided_cuda((4, 10), (10, 1), torch.float32)
# Topologically Sorted Source Nodes: [log_softmax], Original ATen: [aten._log_softmax]
triton_per_fused__log_softmax_2.run(buf4, buf7, 4, 10, grid=grid(4), stream=stream0)
del buf4
return (buf7, primals_1, buf1, buf3, buf7, primals_6, primals_4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 784), (784, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((300, 784), (784, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((300, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((100, 300), (300, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((100, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((10, 100), (100, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((10, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.nn.parallel
import torch.optim
import torch.utils.data
import torch.utils.data.distributed
class LeNet_300_100(nn.Module):
"""Simple NN with hidden layers [300, 100]
Based on https://github.com/mi-lad/snip/blob/master/train.py
by Milad Alizadeh.
"""
def __init__(self, save_features=None, bench_model=False):
super(LeNet_300_100, self).__init__()
self.fc1 = nn.Linear(28 * 28, 300, bias=True)
self.fc2 = nn.Linear(300, 100, bias=True)
self.fc3 = nn.Linear(100, 10, bias=True)
self.mask = None
def forward(self, x):
x0 = x.view(-1, 28 * 28)
x1 = F.relu(self.fc1(x0))
x2 = F.relu(self.fc2(x1))
x3 = self.fc3(x2)
return F.log_softmax(x3, dim=1)
def get_inputs():
return [torch.rand([4, 784])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
import torch.nn.parallel
import torch.optim
import torch.utils.data
import torch.utils.data.distributed
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_relu_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 1200
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 300
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x2, tmp4, xmask)
@triton.jit
def triton_poi_fused_relu_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 400
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 100
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x2, tmp4, xmask)
@triton.jit
def triton_per_fused__log_softmax_2(in_ptr0, out_ptr2, xnumel, rnumel,
XBLOCK: tl.constexpr):
xnumel = 4
rnumel = 10
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
rmask = rindex < rnumel
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 10 * x0), rmask & xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(rmask & xmask, tmp1, float('-inf'))
tmp4 = triton_helpers.max2(tmp3, 1)[:, None]
tmp5 = tmp0 - tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK, RBLOCK])
tmp9 = tl.where(rmask & xmask, tmp7, 0)
tmp10 = tl.sum(tmp9, 1)[:, None]
tmp11 = tl_math.log(tmp10)
tmp12 = tmp5 - tmp11
tl.store(out_ptr2 + (r1 + 10 * x0), tmp12, rmask & xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7) = args
args.clear()
assert_size_stride(primals_1, (4, 784), (784, 1))
assert_size_stride(primals_2, (300, 784), (784, 1))
assert_size_stride(primals_3, (300,), (1,))
assert_size_stride(primals_4, (100, 300), (300, 1))
assert_size_stride(primals_5, (100,), (1,))
assert_size_stride(primals_6, (10, 100), (100, 1))
assert_size_stride(primals_7, (10,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 300), (300, 1), torch.float32)
extern_kernels.mm(primals_1, reinterpret_tensor(primals_2, (784,
300), (1, 784), 0), out=buf0)
del primals_2
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_relu_0[grid(1200)](buf1, primals_3, 1200, XBLOCK=
128, num_warps=4, num_stages=1)
del primals_3
buf2 = empty_strided_cuda((4, 100), (100, 1), torch.float32)
extern_kernels.mm(buf1, reinterpret_tensor(primals_4, (300, 100), (
1, 300), 0), out=buf2)
buf3 = buf2
del buf2
triton_poi_fused_relu_1[grid(400)](buf3, primals_5, 400, XBLOCK=256,
num_warps=4, num_stages=1)
del primals_5
buf4 = empty_strided_cuda((4, 10), (10, 1), torch.float32)
extern_kernels.addmm(primals_7, buf3, reinterpret_tensor(primals_6,
(100, 10), (1, 100), 0), alpha=1, beta=1, out=buf4)
del primals_7
buf7 = empty_strided_cuda((4, 10), (10, 1), torch.float32)
triton_per_fused__log_softmax_2[grid(4)](buf4, buf7, 4, 10, XBLOCK=
1, num_warps=2, num_stages=1)
del buf4
return buf7, primals_1, buf1, buf3, buf7, primals_6, primals_4
class LeNet_300_100New(nn.Module):
"""Simple NN with hidden layers [300, 100]
Based on https://github.com/mi-lad/snip/blob/master/train.py
by Milad Alizadeh.
"""
def __init__(self, save_features=None, bench_model=False):
super(LeNet_300_100New, self).__init__()
self.fc1 = nn.Linear(28 * 28, 300, bias=True)
self.fc2 = nn.Linear(300, 100, bias=True)
self.fc3 = nn.Linear(100, 10, bias=True)
self.mask = None
def forward(self, input_0):
primals_2 = self.fc1.weight
primals_3 = self.fc1.bias
primals_4 = self.fc2.weight
primals_5 = self.fc2.bias
primals_6 = self.fc3.weight
primals_7 = self.fc3.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7])
return output[0]
|
elony314/sparse_learning
|
LeNet_300_100
| false | 12,343 |
[
"MIT"
] | 0 |
fff9ea0267016bda747f2882ef8de508ac1369e7
|
https://github.com/elony314/sparse_learning/tree/fff9ea0267016bda747f2882ef8de508ac1369e7
|
Attention
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/um/cum65j23qchrjf5dndblqgbw6zomhgwfj2obfidtgy7b5j3zwklm.py
# Topologically Sorted Source Nodes: [attn_weights_2], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# attn_weights_2 => amax, exp, sub
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%squeeze, [1], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%squeeze, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
triton_poi_fused__softmax_0 = async_compile.triton('triton_poi_fused__softmax_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tmp9 = tl_math.exp(tmp8)
tl.store(out_ptr0 + (x2), tmp9, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/wk/cwk2wao7opapqbjj7klnqrd6tgist3ts3nc5veryzhzstwpx7d4l.py
# Topologically Sorted Source Nodes: [attn_weights_2], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# attn_weights_2 => div, sum_1
# Graph fragment:
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [1], True), kwargs = {})
# %div : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_1 = async_compile.triton('triton_poi_fused__softmax_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/zd/czdeq2ohbgubcyeps2ukquvfhigxtyega57i24ketclusfgmyedi.py
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cat => cat
# Graph fragment:
# %cat : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%squeeze_1, %primals_4], 1), kwargs = {})
triton_poi_fused_cat_2 = async_compile.triton('triton_poi_fused_cat_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 32
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x1 = (xindex // 8)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x1) + x0), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + ((4*x1) + ((-4) + x0)), tmp6 & xmask, eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x2), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/h6/ch6bafozqfkjjkpo4c4dyiba65qeun72gg5z3pz456zdtwhj66ke.py
# Topologically Sorted Source Nodes: [attn_hidden], Original ATen: [aten.tanh]
# Source node to ATen node mapping:
# attn_hidden => tanh
# Graph fragment:
# %add_tensor : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mm_default, %primals_6), kwargs = {})
# %tanh : [num_users=1] = call_function[target=torch.ops.aten.tanh.default](args = (%add_tensor,), kwargs = {})
triton_poi_fused_tanh_3 = async_compile.triton('triton_poi_fused_tanh_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_tanh_3', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_tanh_3(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x2), tmp3, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (4, ), (1, ))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, 8), (8, 1))
assert_size_stride(primals_6, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [attn_weights], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_3, reinterpret_tensor(primals_1, (16, 4), (4, 1), 0), reinterpret_tensor(primals_2, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf0)
del primals_2
del primals_3
buf1 = empty_strided_cuda((4, 4, 1), (4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [attn_weights_1], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf0, (4, 4, 4), (16, 4, 1), 0), reinterpret_tensor(primals_4, (4, 4, 1), (4, 1, 1), 0), out=buf1)
del buf0
buf2 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [attn_weights_2], Original ATen: [aten._softmax]
stream0 = get_raw_stream(0)
triton_poi_fused__softmax_0.run(buf1, buf2, 16, grid=grid(16), stream=stream0)
buf3 = reinterpret_tensor(buf1, (4, 4), (4, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [attn_weights_2], Original ATen: [aten._softmax]
triton_poi_fused__softmax_1.run(buf2, buf3, 16, grid=grid(16), stream=stream0)
buf4 = reinterpret_tensor(buf2, (4, 4, 1), (4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [bmm_1], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(primals_1, (4, 4, 4), (16, 1, 4), 0), reinterpret_tensor(buf3, (4, 4, 1), (4, 1, 1), 0), out=buf4)
buf5 = empty_strided_cuda((4, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
triton_poi_fused_cat_2.run(buf4, primals_4, buf5, 32, grid=grid(32), stream=stream0)
buf6 = reinterpret_tensor(buf4, (4, 4), (4, 1), 0); del buf4 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(buf5, reinterpret_tensor(primals_5, (8, 4), (1, 8), 0), out=buf6)
buf7 = buf6; del buf6 # reuse
# Topologically Sorted Source Nodes: [attn_hidden], Original ATen: [aten.tanh]
triton_poi_fused_tanh_3.run(buf7, primals_6, 16, grid=grid(16), stream=stream0)
del primals_6
return (buf7, buf3, primals_1, buf3, buf5, buf7, primals_5, reinterpret_tensor(primals_4, (4, 1, 4), (4, 1, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, 8), (8, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class Attention(nn.Module):
def __init__(self, device, hidden_size):
super(Attention, self).__init__()
self.device = device
self.hidden_size = hidden_size
self.concat_linear = nn.Linear(self.hidden_size * 2, self.hidden_size)
self.attn = nn.Linear(self.hidden_size, hidden_size)
def forward(self, rnn_outputs, final_hidden_state):
_batch_size, _seq_len, _ = rnn_outputs.shape
attn_weights = self.attn(rnn_outputs)
attn_weights = torch.bmm(attn_weights, final_hidden_state.unsqueeze(2))
attn_weights = F.softmax(attn_weights.squeeze(2), dim=1)
context = torch.bmm(rnn_outputs.transpose(1, 2), attn_weights.
unsqueeze(2)).squeeze(2)
attn_hidden = torch.tanh(self.concat_linear(torch.cat((context,
final_hidden_state), dim=1)))
return attn_hidden, attn_weights
def get_inputs():
return [torch.rand([4, 4, 4]), torch.rand([4, 4])]
def get_init_inputs():
return [[], {'device': 0, 'hidden_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused__softmax_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tmp9 = tl_math.exp(tmp8)
tl.store(out_ptr0 + x2, tmp9, xmask)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_cat_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 32
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x1 = xindex // 8
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x1 + x0), tmp4 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (4 * x1 + (-4 + x0)), tmp6 & xmask,
eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x2, tmp10, xmask)
@triton.jit
def triton_poi_fused_tanh_3(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x2, tmp3, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (4,), (1,))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, 8), (8, 1))
assert_size_stride(primals_6, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_3, reinterpret_tensor(primals_1, (16,
4), (4, 1), 0), reinterpret_tensor(primals_2, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf0)
del primals_2
del primals_3
buf1 = empty_strided_cuda((4, 4, 1), (4, 1, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf0, (4, 4, 4), (16, 4, 1),
0), reinterpret_tensor(primals_4, (4, 4, 1), (4, 1, 1), 0), out
=buf1)
del buf0
buf2 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused__softmax_0[grid(16)](buf1, buf2, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf3 = reinterpret_tensor(buf1, (4, 4), (4, 1), 0)
del buf1
triton_poi_fused__softmax_1[grid(16)](buf2, buf3, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf4 = reinterpret_tensor(buf2, (4, 4, 1), (4, 1, 1), 0)
del buf2
extern_kernels.bmm(reinterpret_tensor(primals_1, (4, 4, 4), (16, 1,
4), 0), reinterpret_tensor(buf3, (4, 4, 1), (4, 1, 1), 0), out=buf4
)
buf5 = empty_strided_cuda((4, 8), (8, 1), torch.float32)
triton_poi_fused_cat_2[grid(32)](buf4, primals_4, buf5, 32, XBLOCK=
32, num_warps=1, num_stages=1)
buf6 = reinterpret_tensor(buf4, (4, 4), (4, 1), 0)
del buf4
extern_kernels.mm(buf5, reinterpret_tensor(primals_5, (8, 4), (1, 8
), 0), out=buf6)
buf7 = buf6
del buf6
triton_poi_fused_tanh_3[grid(16)](buf7, primals_6, 16, XBLOCK=16,
num_warps=1, num_stages=1)
del primals_6
return (buf7, buf3, primals_1, buf3, buf5, buf7, primals_5,
reinterpret_tensor(primals_4, (4, 1, 4), (4, 1, 1), 0))
class AttentionNew(nn.Module):
def __init__(self, device, hidden_size):
super(AttentionNew, self).__init__()
self.device = device
self.hidden_size = hidden_size
self.concat_linear = nn.Linear(self.hidden_size * 2, self.hidden_size)
self.attn = nn.Linear(self.hidden_size, hidden_size)
def forward(self, input_0, input_1):
primals_5 = self.concat_linear.weight
primals_3 = self.concat_linear.bias
primals_2 = self.attn.weight
primals_6 = self.attn.bias
primals_1 = input_0
primals_4 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6])
return output[0], output[1]
|
ekvall93/tape
|
Attention
| false | 12,344 |
[
"BSD-3-Clause"
] | 0 |
1ca4d5a39c72f806f23a36fb7a7c7325f06096ae
|
https://github.com/ekvall93/tape/tree/1ca4d5a39c72f806f23a36fb7a7c7325f06096ae
|
L2Norm
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/az/cazayodov2pejo62rnac6rludmval2wsivw6phwx6vee62yxa77i.py
# Topologically Sorted Source Nodes: [pow_1, sum_1], Original ATen: [aten.pow, aten.sum]
# Source node to ATen node mapping:
# pow_1 => pow_1
# sum_1 => sum_1
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_2, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [1], True), kwargs = {})
triton_red_fused_pow_sum_0 = async_compile.triton('triton_red_fused_pow_sum_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.reduction(
size_hints=[256, 128],
reduction_hint=ReductionHint.OUTER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_red_fused_pow_sum_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_red_fused_pow_sum_0(in_ptr0, out_ptr0, xnumel, rnumel, XBLOCK : tl.constexpr, RBLOCK : tl.constexpr):
xnumel = 256
rnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rbase = tl.arange(0, RBLOCK)[None, :]
x0 = xindex % 16
x1 = (xindex // 16)
_tmp3 = tl.full([XBLOCK, RBLOCK], 0, tl.float32)
x3 = xindex
for roffset in range(0, rnumel, RBLOCK):
rindex = roffset + rbase
rmask = rindex < rnumel
r2 = rindex
tmp0 = tl.load(in_ptr0 + (x0 + (16*r2) + (2048*x1)), rmask & xmask, eviction_policy='evict_last', other=0.0)
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = _tmp3 + tmp2
_tmp3 = tl.where(rmask & xmask, tmp4, _tmp3)
tmp3 = tl.sum(_tmp3, 1)[:, None]
tl.store(out_ptr0 + (x3), tmp3, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/yb/cybccy6ik3u5wdvejbe7sswchqjx66t2tiskszpk5e3v4xddkzk3.py
# Topologically Sorted Source Nodes: [pow_1, sum_1, clamp, rsqrt], Original ATen: [aten.pow, aten.sum, aten.clamp, aten.rsqrt]
# Source node to ATen node mapping:
# clamp => clamp_min
# pow_1 => pow_1
# rsqrt => rsqrt
# sum_1 => sum_1
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_2, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [1], True), kwargs = {})
# %clamp_min : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%sum_1, 1e-12), kwargs = {})
# %rsqrt : [num_users=2] = call_function[target=torch.ops.aten.rsqrt.default](args = (%clamp_min,), kwargs = {})
triton_per_fused_clamp_pow_rsqrt_sum_1 = async_compile.triton('triton_per_fused_clamp_pow_rsqrt_sum_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[64, 4],
reduction_hint=ReductionHint.OUTER_TINY,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_clamp_pow_rsqrt_sum_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_clamp_pow_rsqrt_sum_1(in_out_ptr0, in_ptr0, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 64
rnumel = 4
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r2 = rindex
x0 = xindex % 16
x1 = (xindex // 16)
x3 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (16*r2) + (64*x1)), xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(xmask, tmp1, 0)
tmp4 = tl.sum(tmp3, 1)[:, None]
tmp5 = 1e-12
tmp6 = triton_helpers.maximum(tmp4, tmp5)
tmp7 = libdevice.rsqrt(tmp6)
tl.debug_barrier()
tl.store(in_out_ptr0 + (x3), tmp7, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/go/cgo6toghmmfwug7mqg2i6moqp5yn5zkdrlyxi73gxriq4acxdhks.py
# Topologically Sorted Source Nodes: [mul, mul_1], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# mul => mul
# mul_1 => mul_1
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_1, %primals_2), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%mul, %rsqrt), kwargs = {})
triton_poi_fused_mul_2 = async_compile.triton('triton_poi_fused_mul_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32768],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_2(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 32768
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x1 = (xindex // 16) % 512
x3 = xindex
x0 = xindex % 16
x2 = (xindex // 8192)
tmp0 = tl.load(in_ptr0 + (x1), None, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x3), None)
tmp3 = tl.load(in_ptr2 + (x0 + (16*x2)), None, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp4 = tmp2 * tmp3
tl.store(out_ptr0 + (x3), tmp4, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (1, 512, 1, 1), (512, 1, 1, 1))
assert_size_stride(primals_2, (4, 512, 4, 4), (8192, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 4, 4), (64, 256, 4, 1, 16), torch.float32)
# Topologically Sorted Source Nodes: [pow_1, sum_1], Original ATen: [aten.pow, aten.sum]
stream0 = get_raw_stream(0)
triton_red_fused_pow_sum_0.run(primals_2, buf0, 256, 128, grid=grid(256), stream=stream0)
buf1 = empty_strided_cuda((4, 1, 4, 4), (16, 64, 4, 1), torch.float32)
buf2 = reinterpret_tensor(buf1, (4, 1, 4, 4), (16, 16, 4, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [pow_1, sum_1, clamp, rsqrt], Original ATen: [aten.pow, aten.sum, aten.clamp, aten.rsqrt]
triton_per_fused_clamp_pow_rsqrt_sum_1.run(buf2, buf0, 64, 4, grid=grid(64), stream=stream0)
del buf0
buf3 = empty_strided_cuda((4, 512, 4, 4), (8192, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, mul_1], Original ATen: [aten.mul]
triton_poi_fused_mul_2.run(primals_1, primals_2, buf2, buf3, 32768, grid=grid(32768), stream=stream0)
del primals_1
return (buf3, primals_2, buf2, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 512, 1, 1), (512, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 512, 4, 4), (8192, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.onnx
class L2Norm(nn.Module):
"""
Scale shall be learnable according to original paper
scale: initial scale number
chan_num: L2Norm channel number (norm over all channels)
"""
def __init__(self, scale=20, chan_num=512):
super(L2Norm, self).__init__()
self.scale = nn.Parameter(torch.Tensor([scale] * chan_num).view(1,
chan_num, 1, 1))
def forward(self, data):
return self.scale * data * data.pow(2).sum(dim=1, keepdim=True).clamp(
min=1e-12).rsqrt()
def get_inputs():
return [torch.rand([4, 512, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
import torch.onnx
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_red_fused_pow_sum_0(in_ptr0, out_ptr0, xnumel, rnumel, XBLOCK:
tl.constexpr, RBLOCK: tl.constexpr):
xnumel = 256
rnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rbase = tl.arange(0, RBLOCK)[None, :]
x0 = xindex % 16
x1 = xindex // 16
_tmp3 = tl.full([XBLOCK, RBLOCK], 0, tl.float32)
x3 = xindex
for roffset in range(0, rnumel, RBLOCK):
rindex = roffset + rbase
rmask = rindex < rnumel
r2 = rindex
tmp0 = tl.load(in_ptr0 + (x0 + 16 * r2 + 2048 * x1), rmask & xmask,
eviction_policy='evict_last', other=0.0)
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = _tmp3 + tmp2
_tmp3 = tl.where(rmask & xmask, tmp4, _tmp3)
tmp3 = tl.sum(_tmp3, 1)[:, None]
tl.store(out_ptr0 + x3, tmp3, xmask)
@triton.jit
def triton_per_fused_clamp_pow_rsqrt_sum_1(in_out_ptr0, in_ptr0, xnumel,
rnumel, XBLOCK: tl.constexpr):
xnumel = 64
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r2 = rindex
x0 = xindex % 16
x1 = xindex // 16
x3 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 16 * r2 + 64 * x1), xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(xmask, tmp1, 0)
tmp4 = tl.sum(tmp3, 1)[:, None]
tmp5 = 1e-12
tmp6 = triton_helpers.maximum(tmp4, tmp5)
tmp7 = libdevice.rsqrt(tmp6)
tl.debug_barrier()
tl.store(in_out_ptr0 + x3, tmp7, xmask)
@triton.jit
def triton_poi_fused_mul_2(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x1 = xindex // 16 % 512
x3 = xindex
x0 = xindex % 16
x2 = xindex // 8192
tmp0 = tl.load(in_ptr0 + x1, None, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + x3, None)
tmp3 = tl.load(in_ptr2 + (x0 + 16 * x2), None, eviction_policy='evict_last'
)
tmp2 = tmp0 * tmp1
tmp4 = tmp2 * tmp3
tl.store(out_ptr0 + x3, tmp4, None)
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (1, 512, 1, 1), (512, 1, 1, 1))
assert_size_stride(primals_2, (4, 512, 4, 4), (8192, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 4, 4), (64, 256, 4, 1, 16),
torch.float32)
get_raw_stream(0)
triton_red_fused_pow_sum_0[grid(256)](primals_2, buf0, 256, 128,
XBLOCK=64, RBLOCK=8, num_warps=4, num_stages=1)
buf1 = empty_strided_cuda((4, 1, 4, 4), (16, 64, 4, 1), torch.float32)
buf2 = reinterpret_tensor(buf1, (4, 1, 4, 4), (16, 16, 4, 1), 0)
del buf1
triton_per_fused_clamp_pow_rsqrt_sum_1[grid(64)](buf2, buf0, 64, 4,
XBLOCK=64, num_warps=2, num_stages=1)
del buf0
buf3 = empty_strided_cuda((4, 512, 4, 4), (8192, 16, 4, 1), torch.
float32)
triton_poi_fused_mul_2[grid(32768)](primals_1, primals_2, buf2,
buf3, 32768, XBLOCK=256, num_warps=4, num_stages=1)
del primals_1
return buf3, primals_2, buf2
class L2NormNew(nn.Module):
"""
Scale shall be learnable according to original paper
scale: initial scale number
chan_num: L2Norm channel number (norm over all channels)
"""
def __init__(self, scale=20, chan_num=512):
super(L2NormNew, self).__init__()
self.scale = nn.Parameter(torch.Tensor([scale] * chan_num).view(1,
chan_num, 1, 1))
def forward(self, input_0):
primals_1 = self.scale
primals_2 = input_0
output = call([primals_1, primals_2])
return output[0]
|
ephrem-git/inference
|
L2Norm
| false | 12,345 |
[
"Apache-2.0"
] | 0 |
bfbda5fc419364c3f71b5b1640f6c00e7675b212
|
https://github.com/ephrem-git/inference/tree/bfbda5fc419364c3f71b5b1640f6c00e7675b212
|
Get_gradient_nopadding
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/oi/coispgrpvordht4pc36porvrnlkw2qp26a2e4lnlycgqgz24vmr5.py
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# x => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%sqrt, %sqrt_1, %sqrt_2, %sqrt_3], 1), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: '*fp32', 7: '*fp32', 8: '*fp32', 9: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, in_ptr6, in_ptr7, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 16) % 4
x0 = xindex % 16
x2 = (xindex // 64)
x3 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (16*x2)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp5 * tmp5
tmp7 = tl.load(in_ptr1 + (x0 + (16*x2)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp8 = tmp7 * tmp7
tmp9 = tmp6 + tmp8
tmp10 = 1e-06
tmp11 = tmp9 + tmp10
tmp12 = libdevice.sqrt(tmp11)
tmp13 = tl.full(tmp12.shape, 0.0, tmp12.dtype)
tmp14 = tl.where(tmp4, tmp12, tmp13)
tmp15 = tmp0 >= tmp3
tmp16 = tl.full([1], 2, tl.int64)
tmp17 = tmp0 < tmp16
tmp18 = tmp15 & tmp17
tmp19 = tl.load(in_ptr2 + (x0 + (16*x2)), tmp18 & xmask, eviction_policy='evict_last', other=0.0)
tmp20 = tmp19 * tmp19
tmp21 = tl.load(in_ptr3 + (x0 + (16*x2)), tmp18 & xmask, eviction_policy='evict_last', other=0.0)
tmp22 = tmp21 * tmp21
tmp23 = tmp20 + tmp22
tmp24 = tmp23 + tmp10
tmp25 = libdevice.sqrt(tmp24)
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp18, tmp25, tmp26)
tmp28 = tmp0 >= tmp16
tmp29 = tl.full([1], 3, tl.int64)
tmp30 = tmp0 < tmp29
tmp31 = tmp28 & tmp30
tmp32 = tl.load(in_ptr4 + (x0 + (16*x2)), tmp31 & xmask, eviction_policy='evict_last', other=0.0)
tmp33 = tmp32 * tmp32
tmp34 = tl.load(in_ptr5 + (x0 + (16*x2)), tmp31 & xmask, eviction_policy='evict_last', other=0.0)
tmp35 = tmp34 * tmp34
tmp36 = tmp33 + tmp35
tmp37 = tmp36 + tmp10
tmp38 = libdevice.sqrt(tmp37)
tmp39 = tl.full(tmp38.shape, 0.0, tmp38.dtype)
tmp40 = tl.where(tmp31, tmp38, tmp39)
tmp41 = tmp0 >= tmp29
tmp42 = tl.full([1], 4, tl.int64)
tmp43 = tmp0 < tmp42
tmp44 = tl.load(in_ptr6 + (x0 + (16*x2)), tmp41 & xmask, eviction_policy='evict_last', other=0.0)
tmp45 = tmp44 * tmp44
tmp46 = tl.load(in_ptr7 + (x0 + (16*x2)), tmp41 & xmask, eviction_policy='evict_last', other=0.0)
tmp47 = tmp46 * tmp46
tmp48 = tmp45 + tmp47
tmp49 = tmp48 + tmp10
tmp50 = libdevice.sqrt(tmp49)
tmp51 = tl.full(tmp50.shape, 0.0, tmp50.dtype)
tmp52 = tl.where(tmp41, tmp50, tmp51)
tmp53 = tl.where(tmp31, tmp40, tmp52)
tmp54 = tl.where(tmp18, tmp27, tmp53)
tmp55 = tl.where(tmp4, tmp14, tmp54)
tl.store(out_ptr0 + (x3), tmp55, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1, arg2_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (1, 1, 3, 3), (9, 9, 3, 1))
assert_size_stride(arg2_1, (1, 1, 3, 3), (9, 9, 3, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [x_i_v], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 0), arg1_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 1, 4, 4), (16, 16, 4, 1))
# Topologically Sorted Source Nodes: [x_i_h], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 0), arg2_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 1, 4, 4), (16, 16, 4, 1))
# Topologically Sorted Source Nodes: [x_i_v_1], Original ATen: [aten.convolution]
buf2 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 16), arg1_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 1, 4, 4), (16, 16, 4, 1))
# Topologically Sorted Source Nodes: [x_i_h_1], Original ATen: [aten.convolution]
buf3 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 16), arg2_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 1, 4, 4), (16, 16, 4, 1))
# Topologically Sorted Source Nodes: [x_i_v_2], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 32), arg1_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 1, 4, 4), (16, 16, 4, 1))
# Topologically Sorted Source Nodes: [x_i_h_2], Original ATen: [aten.convolution]
buf5 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 32), arg2_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf5, (4, 1, 4, 4), (16, 16, 4, 1))
# Topologically Sorted Source Nodes: [x_i_v_3], Original ATen: [aten.convolution]
buf6 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 48), arg1_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 1, 4, 4), (16, 16, 4, 1))
del arg1_1
# Topologically Sorted Source Nodes: [x_i_h_3], Original ATen: [aten.convolution]
buf7 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1, 4, 4), (64, 0, 4, 1), 48), arg2_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf7, (4, 1, 4, 4), (16, 16, 4, 1))
del arg0_1
del arg2_1
buf8 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(buf0, buf1, buf2, buf3, buf4, buf5, buf6, buf7, buf8, 256, grid=grid(256), stream=stream0)
del buf0
del buf1
del buf2
del buf3
del buf4
del buf5
del buf6
del buf7
return (buf8, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((1, 1, 3, 3), (9, 9, 3, 1), device='cuda:0', dtype=torch.float32)
arg2_1 = rand_strided((1, 1, 3, 3), (9, 9, 3, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1, arg2_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class Get_gradient_nopadding(nn.Module):
def __init__(self):
super(Get_gradient_nopadding, self).__init__()
kernel_v = [[0, -1, 0], [0, 0, 0], [0, 1, 0]]
kernel_h = [[0, 0, 0], [-1, 0, 1], [0, 0, 0]]
kernel_h = torch.FloatTensor(kernel_h).unsqueeze(0).unsqueeze(0)
kernel_v = torch.FloatTensor(kernel_v).unsqueeze(0).unsqueeze(0)
self.weight_h = nn.Parameter(data=kernel_h, requires_grad=False)
self.weight_v = nn.Parameter(data=kernel_v, requires_grad=False)
def forward(self, x):
x_list = []
for i in range(x.shape[1]):
x_i = x[:, i]
x_i_v = F.conv2d(x_i.unsqueeze(1), self.weight_v, padding=1)
x_i_h = F.conv2d(x_i.unsqueeze(1), self.weight_h, padding=1)
x_i = torch.sqrt(torch.pow(x_i_v, 2) + torch.pow(x_i_h, 2) + 1e-06)
x_list.append(x_i)
x = torch.cat(x_list, dim=1)
return x
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4,
in_ptr5, in_ptr6, in_ptr7, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 16 % 4
x0 = xindex % 16
x2 = xindex // 64
x3 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 16 * x2), tmp4 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp6 = tmp5 * tmp5
tmp7 = tl.load(in_ptr1 + (x0 + 16 * x2), tmp4 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp8 = tmp7 * tmp7
tmp9 = tmp6 + tmp8
tmp10 = 1e-06
tmp11 = tmp9 + tmp10
tmp12 = libdevice.sqrt(tmp11)
tmp13 = tl.full(tmp12.shape, 0.0, tmp12.dtype)
tmp14 = tl.where(tmp4, tmp12, tmp13)
tmp15 = tmp0 >= tmp3
tmp16 = tl.full([1], 2, tl.int64)
tmp17 = tmp0 < tmp16
tmp18 = tmp15 & tmp17
tmp19 = tl.load(in_ptr2 + (x0 + 16 * x2), tmp18 & xmask,
eviction_policy='evict_last', other=0.0)
tmp20 = tmp19 * tmp19
tmp21 = tl.load(in_ptr3 + (x0 + 16 * x2), tmp18 & xmask,
eviction_policy='evict_last', other=0.0)
tmp22 = tmp21 * tmp21
tmp23 = tmp20 + tmp22
tmp24 = tmp23 + tmp10
tmp25 = libdevice.sqrt(tmp24)
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp18, tmp25, tmp26)
tmp28 = tmp0 >= tmp16
tmp29 = tl.full([1], 3, tl.int64)
tmp30 = tmp0 < tmp29
tmp31 = tmp28 & tmp30
tmp32 = tl.load(in_ptr4 + (x0 + 16 * x2), tmp31 & xmask,
eviction_policy='evict_last', other=0.0)
tmp33 = tmp32 * tmp32
tmp34 = tl.load(in_ptr5 + (x0 + 16 * x2), tmp31 & xmask,
eviction_policy='evict_last', other=0.0)
tmp35 = tmp34 * tmp34
tmp36 = tmp33 + tmp35
tmp37 = tmp36 + tmp10
tmp38 = libdevice.sqrt(tmp37)
tmp39 = tl.full(tmp38.shape, 0.0, tmp38.dtype)
tmp40 = tl.where(tmp31, tmp38, tmp39)
tmp41 = tmp0 >= tmp29
tl.full([1], 4, tl.int64)
tmp44 = tl.load(in_ptr6 + (x0 + 16 * x2), tmp41 & xmask,
eviction_policy='evict_last', other=0.0)
tmp45 = tmp44 * tmp44
tmp46 = tl.load(in_ptr7 + (x0 + 16 * x2), tmp41 & xmask,
eviction_policy='evict_last', other=0.0)
tmp47 = tmp46 * tmp46
tmp48 = tmp45 + tmp47
tmp49 = tmp48 + tmp10
tmp50 = libdevice.sqrt(tmp49)
tmp51 = tl.full(tmp50.shape, 0.0, tmp50.dtype)
tmp52 = tl.where(tmp41, tmp50, tmp51)
tmp53 = tl.where(tmp31, tmp40, tmp52)
tmp54 = tl.where(tmp18, tmp27, tmp53)
tmp55 = tl.where(tmp4, tmp14, tmp54)
tl.store(out_ptr0 + x3, tmp55, xmask)
def call(args):
arg0_1, arg1_1, arg2_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (1, 1, 3, 3), (9, 9, 3, 1))
assert_size_stride(arg2_1, (1, 1, 3, 3), (9, 9, 3, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 0), arg1_1, stride=(1, 1), padding=(1, 1),
dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf0, (4, 1, 4, 4), (16, 16, 4, 1))
buf1 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 0), arg2_1, stride=(1, 1), padding=(1, 1),
dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf1, (4, 1, 4, 4), (16, 16, 4, 1))
buf2 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 16), arg1_1, stride=(1, 1), padding=(1, 1
), dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf2, (4, 1, 4, 4), (16, 16, 4, 1))
buf3 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 16), arg2_1, stride=(1, 1), padding=(1, 1
), dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf3, (4, 1, 4, 4), (16, 16, 4, 1))
buf4 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 32), arg1_1, stride=(1, 1), padding=(1, 1
), dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf4, (4, 1, 4, 4), (16, 16, 4, 1))
buf5 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 32), arg2_1, stride=(1, 1), padding=(1, 1
), dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf5, (4, 1, 4, 4), (16, 16, 4, 1))
buf6 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 48), arg1_1, stride=(1, 1), padding=(1, 1
), dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf6, (4, 1, 4, 4), (16, 16, 4, 1))
del arg1_1
buf7 = extern_kernels.convolution(reinterpret_tensor(arg0_1, (4, 1,
4, 4), (64, 0, 4, 1), 48), arg2_1, stride=(1, 1), padding=(1, 1
), dilation=(1, 1), transposed=False, output_padding=(0, 0),
groups=1, bias=None)
assert_size_stride(buf7, (4, 1, 4, 4), (16, 16, 4, 1))
del arg0_1
del arg2_1
buf8 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(256)](buf0, buf1, buf2, buf3, buf4,
buf5, buf6, buf7, buf8, 256, XBLOCK=128, num_warps=4, num_stages=1)
del buf0
del buf1
del buf2
del buf3
del buf4
del buf5
del buf6
del buf7
return buf8,
class Get_gradient_nopaddingNew(nn.Module):
def __init__(self):
super(Get_gradient_nopaddingNew, self).__init__()
kernel_v = [[0, -1, 0], [0, 0, 0], [0, 1, 0]]
kernel_h = [[0, 0, 0], [-1, 0, 1], [0, 0, 0]]
kernel_h = torch.FloatTensor(kernel_h).unsqueeze(0).unsqueeze(0)
kernel_v = torch.FloatTensor(kernel_v).unsqueeze(0).unsqueeze(0)
self.weight_h = nn.Parameter(data=kernel_h, requires_grad=False)
self.weight_v = nn.Parameter(data=kernel_v, requires_grad=False)
def forward(self, input_0):
arg1_1 = self.weight_h
arg2_1 = self.weight_v
arg0_1 = input_0
output = call([arg0_1, arg1_1, arg2_1])
return output[0]
|
eqprog/ESRGAN
|
Get_gradient_nopadding
| false | 12,346 |
[
"Apache-2.0"
] | 0 |
d5eb02531cf0ce4e8df93793f3012486bac8d87a
|
https://github.com/eqprog/ESRGAN/tree/d5eb02531cf0ce4e8df93793f3012486bac8d87a
|
MultiheadAttention
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/5w/c5wnubyijcgstpnbhnht5ommr737mwfx67lgpfc6mvwlwmhzfkmq.py
# Topologically Sorted Source Nodes: [q_1], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# q_1 => mul
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_1, 1.0), kwargs = {})
triton_poi_fused_mul_0 = async_compile.triton('triton_poi_fused_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = 1.0
tmp4 = tmp2 * tmp3
tl.store(in_out_ptr0 + (x2), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ko/ckow7ci7f3mygm6ujdzdisip6tet25h4hj6uestesqalhkarwrrw.py
# Topologically Sorted Source Nodes: [softmax], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# softmax => amax, div, exp, sub, sum_1
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%bmm, [-1], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%bmm, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [-1], True), kwargs = {})
# %div : [num_users=3] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_per_fused__softmax_1 = async_compile.triton('triton_per_fused__softmax_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[64, 16],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused__softmax_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 2, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused__softmax_1(in_ptr0, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 64
rnumel = 16
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (16*x0)), xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(xmask, tmp1, float("-inf"))
tmp4 = triton_helpers.max2(tmp3, 1)[:, None]
tmp5 = tmp0 - tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK, RBLOCK])
tmp9 = tl.where(xmask, tmp7, 0)
tmp10 = tl.sum(tmp9, 1)[:, None]
tmp11 = tmp6 / tmp10
tl.store(out_ptr2 + (r1 + (16*x0)), tmp11, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/qa/cqazar4hg4rdjbxm7zr5mix2w3dkhfmvvjksn7c6lktr5yfe6ndy.py
# Topologically Sorted Source Nodes: [contiguous_3], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# contiguous_3 => clone_1
# Graph fragment:
# %clone_1 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_8,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_2 = async_compile.triton('triton_poi_fused_clone_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4, 16], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 4
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x1 = xindex
y0 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x1)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x1 + (16*y0)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/c2/cc2wsialcqiknwetscnqy3fzaqmmib3cxfb7tsfjx7hdlsxbdq7s.py
# Topologically Sorted Source Nodes: [sum_1, attn_weights_4], Original ATen: [aten.sum, aten.div]
# Source node to ATen node mapping:
# attn_weights_4 => div_1
# sum_1 => sum_2
# Graph fragment:
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%view_17, [1]), kwargs = {})
# %div_1 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sum_2, 4), kwargs = {})
triton_poi_fused_div_sum_3 = async_compile.triton('triton_poi_fused_div_sum_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_sum_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_sum_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 64
x1 = (xindex // 64)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (256*x1)), xmask)
tmp1 = tl.load(in_ptr0 + (64 + x0 + (256*x1)), xmask)
tmp3 = tl.load(in_ptr0 + (128 + x0 + (256*x1)), xmask)
tmp5 = tl.load(in_ptr0 + (192 + x0 + (256*x1)), xmask)
tmp2 = tmp0 + tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (12, 4), (4, 1))
assert_size_stride(primals_5, (12, ), (1, ))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_1, (16, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf0)
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [k], Original ATen: [aten.addmm]
extern_kernels.addmm(reinterpret_tensor(primals_5, (4, ), (1, ), 4), reinterpret_tensor(primals_2, (64, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 16), alpha=1, beta=1, out=buf1)
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [v], Original ATen: [aten.addmm]
extern_kernels.addmm(reinterpret_tensor(primals_5, (4, ), (1, ), 8), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 32), alpha=1, beta=1, out=buf2)
del primals_4
buf3 = reinterpret_tensor(buf0, (4, 4, 4), (16, 4, 1), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [q_1], Original ATen: [aten.mul]
stream0 = get_raw_stream(0)
triton_poi_fused_mul_0.run(buf3, primals_5, 64, grid=grid(64), stream=stream0)
del primals_5
buf4 = empty_strided_cuda((16, 4, 16), (64, 16, 1), torch.float32)
# Topologically Sorted Source Nodes: [attn_weights], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 1), (1, 16, 0), 0), reinterpret_tensor(buf1, (16, 1, 16), (1, 1, 16), 0), out=buf4)
buf7 = empty_strided_cuda((16, 4, 16), (64, 16, 1), torch.float32)
# Topologically Sorted Source Nodes: [softmax], Original ATen: [aten._softmax]
triton_per_fused__softmax_1.run(buf4, buf7, 64, 16, grid=grid(64), stream=stream0)
del buf4
buf8 = empty_strided_cuda((16, 4, 1), (4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [attn], Original ATen: [aten.bmm]
extern_kernels.bmm(buf7, reinterpret_tensor(buf2, (16, 16, 1), (1, 16, 1), 0), out=buf8)
buf9 = empty_strided_cuda((4, 16, 1), (16, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [contiguous_3], Original ATen: [aten.clone]
triton_poi_fused_clone_2.run(buf8, buf9, 4, 16, grid=grid(4, 16), stream=stream0)
buf10 = reinterpret_tensor(buf8, (16, 4), (4, 1), 0); del buf8 # reuse
# Topologically Sorted Source Nodes: [attn_2], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_7, reinterpret_tensor(buf9, (16, 4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf10)
del primals_7
buf11 = empty_strided_cuda((4, 4, 16), (64, 16, 1), torch.float32)
# Topologically Sorted Source Nodes: [sum_1, attn_weights_4], Original ATen: [aten.sum, aten.div]
triton_poi_fused_div_sum_3.run(buf7, buf11, 256, grid=grid(256), stream=stream0)
return (reinterpret_tensor(buf10, (4, 4, 4), (16, 4, 1), 0), buf11, reinterpret_tensor(primals_1, (16, 4), (4, 1), 0), reinterpret_tensor(primals_2, (64, 4), (4, 1), 0), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), buf7, reinterpret_tensor(buf9, (16, 4), (4, 1), 0), primals_6, reinterpret_tensor(buf2, (16, 1, 16), (1, 1, 16), 0), reinterpret_tensor(buf3, (16, 1, 4), (1, 1, 16), 0), reinterpret_tensor(buf1, (16, 16, 1), (1, 16, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((12, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((12, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn.functional as F
from torch import nn
import torch.utils.data
from torch.nn import Parameter
import torch.onnx.operators
import torch.optim
import torch.optim.lr_scheduler
class MultiheadAttention(nn.Module):
"""Multi-headed attention.
See "Attention Is All You Need" for more details.
"""
def __init__(self, embed_dim, num_heads, dropout=0.0, bias=True,
add_bias_kv=False, add_zero_attn=False):
super().__init__()
self.embed_dim = embed_dim
self.num_heads = num_heads
self.dropout = dropout
self.head_dim = embed_dim // num_heads
assert self.head_dim * num_heads == self.embed_dim, 'embed_dim must be divisible by num_heads'
self.scaling = self.head_dim ** -0.5
self.in_proj_weight = Parameter(torch.Tensor(3 * embed_dim, embed_dim))
if bias:
self.in_proj_bias = Parameter(torch.Tensor(3 * embed_dim))
else:
self.register_parameter('in_proj_bias', None)
self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
if add_bias_kv:
self.bias_k = Parameter(torch.Tensor(1, 1, embed_dim))
self.bias_v = Parameter(torch.Tensor(1, 1, embed_dim))
else:
self.bias_k = self.bias_v = None
self.add_zero_attn = add_zero_attn
self.reset_parameters()
self.onnx_trace = False
def prepare_for_onnx_export_(self):
self.onnx_trace = True
def reset_parameters(self):
nn.init.xavier_uniform_(self.in_proj_weight)
nn.init.xavier_uniform_(self.out_proj.weight)
if self.in_proj_bias is not None:
nn.init.constant_(self.in_proj_bias, 0.0)
nn.init.constant_(self.out_proj.bias, 0.0)
if self.bias_k is not None:
nn.init.xavier_normal_(self.bias_k)
if self.bias_v is not None:
nn.init.xavier_normal_(self.bias_v)
def forward(self, query, key, value, key_padding_mask=None,
incremental_state=None, need_weights=True, static_kv=False,
attn_mask=None):
"""Input shape: Time x Batch x Channel
Self-attention can be implemented by passing in the same arguments for
query, key and value. Timesteps can be masked by supplying a T x T mask in the
`attn_mask` argument. Padding elements can be excluded from
the key by passing a binary ByteTensor (`key_padding_mask`) with shape:
batch x src_len, where padding elements are indicated by 1s.
"""
qkv_same = query.data_ptr() == key.data_ptr() == value.data_ptr()
kv_same = key.data_ptr() == value.data_ptr()
tgt_len, bsz, embed_dim = query.size()
assert embed_dim == self.embed_dim
assert list(query.size()) == [tgt_len, bsz, embed_dim]
assert key.size() == value.size()
if incremental_state is not None:
saved_state = self._get_input_buffer(incremental_state)
if 'prev_key' in saved_state:
if static_kv:
assert kv_same and not qkv_same
key = value = None
else:
saved_state = None
if qkv_same:
q, k, v = self.in_proj_qkv(query)
elif kv_same:
q = self.in_proj_q(query)
if key is None:
assert value is None
k = v = None
else:
k, v = self.in_proj_kv(key)
else:
q = self.in_proj_q(query)
k = self.in_proj_k(key)
v = self.in_proj_v(value)
q *= self.scaling
if self.bias_k is not None:
assert self.bias_v is not None
k = torch.cat([k, self.bias_k.repeat(1, bsz, 1)])
v = torch.cat([v, self.bias_v.repeat(1, bsz, 1)])
if attn_mask is not None:
attn_mask = torch.cat([attn_mask, attn_mask.new_zeros(
attn_mask.size(0), 1)], dim=1)
if key_padding_mask is not None:
key_padding_mask = torch.cat([key_padding_mask,
key_padding_mask.new_zeros(key_padding_mask.size(0), 1)
], dim=1)
q = q.contiguous().view(tgt_len, bsz * self.num_heads, self.head_dim
).transpose(0, 1)
if k is not None:
k = k.contiguous().view(-1, bsz * self.num_heads, self.head_dim
).transpose(0, 1)
if v is not None:
v = v.contiguous().view(-1, bsz * self.num_heads, self.head_dim
).transpose(0, 1)
if saved_state is not None:
if 'prev_key' in saved_state:
prev_key = saved_state['prev_key'].view(bsz * self.
num_heads, -1, self.head_dim)
if static_kv:
k = prev_key
else:
k = torch.cat((prev_key, k), dim=1)
if 'prev_value' in saved_state:
prev_value = saved_state['prev_value'].view(bsz * self.
num_heads, -1, self.head_dim)
if static_kv:
v = prev_value
else:
v = torch.cat((prev_value, v), dim=1)
saved_state['prev_key'] = k.view(bsz, self.num_heads, -1, self.
head_dim)
saved_state['prev_value'] = v.view(bsz, self.num_heads, -1,
self.head_dim)
self._set_input_buffer(incremental_state, saved_state)
src_len = k.size(1)
if key_padding_mask is not None:
assert key_padding_mask.size(0) == bsz
assert key_padding_mask.size(1) == src_len
if self.add_zero_attn:
src_len += 1
k = torch.cat([k, k.new_zeros((k.size(0), 1) + k.size()[2:])],
dim=1)
v = torch.cat([v, v.new_zeros((v.size(0), 1) + v.size()[2:])],
dim=1)
if attn_mask is not None:
attn_mask = torch.cat([attn_mask, attn_mask.new_zeros(
attn_mask.size(0), 1)], dim=1)
if key_padding_mask is not None:
key_padding_mask = torch.cat([key_padding_mask, torch.zeros
(key_padding_mask.size(0), 1).type_as(key_padding_mask)
], dim=1)
attn_weights = torch.bmm(q, k.transpose(1, 2))
assert list(attn_weights.size()) == [bsz * self.num_heads, tgt_len,
src_len]
if attn_mask is not None:
attn_mask = attn_mask.unsqueeze(0)
if self.onnx_trace:
attn_mask = attn_mask.repeat(attn_weights.size(0), 1, 1)
attn_weights += attn_mask
if key_padding_mask is not None:
attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len,
src_len)
if self.onnx_trace:
attn_weights = torch.where(key_padding_mask.unsqueeze(1).
unsqueeze(2), torch.Tensor([float('-Inf')]),
attn_weights.float()).type_as(attn_weights)
else:
attn_weights = attn_weights.float().masked_fill(
key_padding_mask.unsqueeze(1).unsqueeze(2), float('-inf')
).type_as(attn_weights)
attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len,
src_len)
attn_weights = F.softmax(attn_weights.float(), dim=-1).type_as(
attn_weights)
attn_weights = F.dropout(attn_weights, p=self.dropout, training=
self.training)
attn = torch.bmm(attn_weights, v)
assert list(attn.size()) == [bsz * self.num_heads, tgt_len, self.
head_dim]
if self.onnx_trace and attn.size(1) == 1:
attn = attn.contiguous().view(tgt_len, bsz, embed_dim)
else:
attn = attn.transpose(0, 1).contiguous().view(tgt_len, bsz,
embed_dim)
attn = self.out_proj(attn)
if need_weights:
attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len,
src_len)
attn_weights = attn_weights.sum(dim=1) / self.num_heads
else:
attn_weights = None
return attn, attn_weights
def in_proj_qkv(self, query):
return self._in_proj(query).chunk(3, dim=-1)
def in_proj_kv(self, key):
return self._in_proj(key, start=self.embed_dim).chunk(2, dim=-1)
def in_proj_q(self, query):
return self._in_proj(query, end=self.embed_dim)
def in_proj_k(self, key):
return self._in_proj(key, start=self.embed_dim, end=2 * self.embed_dim)
def in_proj_v(self, value):
return self._in_proj(value, start=2 * self.embed_dim)
def _in_proj(self, input, start=0, end=None):
weight = self.in_proj_weight
bias = self.in_proj_bias
weight = weight[start:end, :]
if bias is not None:
bias = bias[start:end]
return F.linear(input, weight, bias)
def reorder_incremental_state(self, incremental_state, new_order):
"""Reorder buffered internal state (for incremental generation)."""
input_buffer = self._get_input_buffer(incremental_state)
if input_buffer is not None:
for k in input_buffer.keys():
input_buffer[k] = input_buffer[k].index_select(0, new_order)
self._set_input_buffer(incremental_state, input_buffer)
def _get_input_buffer(self, incremental_state):
return utils.get_incremental_state(self, incremental_state,
'attn_state') or {}
def _set_input_buffer(self, incremental_state, buffer):
utils.set_incremental_state(self, incremental_state, 'attn_state',
buffer)
def get_inputs():
return [torch.rand([4, 4, 4]), torch.rand([4, 4, 4, 4]), torch.rand([4,
4, 4, 4])]
def get_init_inputs():
return [[], {'embed_dim': 4, 'num_heads': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn.functional as F
from torch import nn
import torch.utils.data
from torch.nn import Parameter
import torch.onnx.operators
import torch.optim
import torch.optim.lr_scheduler
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_mul_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = 1.0
tmp4 = tmp2 * tmp3
tl.store(in_out_ptr0 + x2, tmp4, xmask)
@triton.jit
def triton_per_fused__softmax_1(in_ptr0, out_ptr2, xnumel, rnumel, XBLOCK:
tl.constexpr):
xnumel = 64
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 16 * x0), xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(xmask, tmp1, float('-inf'))
tmp4 = triton_helpers.max2(tmp3, 1)[:, None]
tmp5 = tmp0 - tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK, RBLOCK])
tmp9 = tl.where(xmask, tmp7, 0)
tmp10 = tl.sum(tmp9, 1)[:, None]
tmp11 = tmp6 / tmp10
tl.store(out_ptr2 + (r1 + 16 * x0), tmp11, xmask)
@triton.jit
def triton_poi_fused_clone_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 4
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x1 = xindex
y0 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x1), xmask & ymask, eviction_policy=
'evict_last')
tl.store(out_ptr0 + (x1 + 16 * y0), tmp0, xmask & ymask)
@triton.jit
def triton_poi_fused_div_sum_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 64
x1 = xindex // 64
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 256 * x1), xmask)
tmp1 = tl.load(in_ptr0 + (64 + x0 + 256 * x1), xmask)
tmp3 = tl.load(in_ptr0 + (128 + x0 + 256 * x1), xmask)
tmp5 = tl.load(in_ptr0 + (192 + x0 + 256 * x1), xmask)
tmp2 = tmp0 + tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7) = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (12, 4), (4, 1))
assert_size_stride(primals_5, (12,), (1,))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_1, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf0)
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(reinterpret_tensor(primals_5, (4,), (1,), 4),
reinterpret_tensor(primals_2, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 16), alpha=1,
beta=1, out=buf1)
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(reinterpret_tensor(primals_5, (4,), (1,), 8),
reinterpret_tensor(primals_3, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 32), alpha=1,
beta=1, out=buf2)
del primals_4
buf3 = reinterpret_tensor(buf0, (4, 4, 4), (16, 4, 1), 0)
del buf0
get_raw_stream(0)
triton_poi_fused_mul_0[grid(64)](buf3, primals_5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del primals_5
buf4 = empty_strided_cuda((16, 4, 16), (64, 16, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 1), (1, 16, 0),
0), reinterpret_tensor(buf1, (16, 1, 16), (1, 1, 16), 0), out=buf4)
buf7 = empty_strided_cuda((16, 4, 16), (64, 16, 1), torch.float32)
triton_per_fused__softmax_1[grid(64)](buf4, buf7, 64, 16, XBLOCK=32,
num_warps=4, num_stages=1)
del buf4
buf8 = empty_strided_cuda((16, 4, 1), (4, 1, 1), torch.float32)
extern_kernels.bmm(buf7, reinterpret_tensor(buf2, (16, 16, 1), (1,
16, 1), 0), out=buf8)
buf9 = empty_strided_cuda((4, 16, 1), (16, 1, 1), torch.float32)
triton_poi_fused_clone_2[grid(4, 16)](buf8, buf9, 4, 16, XBLOCK=16,
YBLOCK=4, num_warps=1, num_stages=1)
buf10 = reinterpret_tensor(buf8, (16, 4), (4, 1), 0)
del buf8
extern_kernels.addmm(primals_7, reinterpret_tensor(buf9, (16, 4), (
4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf10)
del primals_7
buf11 = empty_strided_cuda((4, 4, 16), (64, 16, 1), torch.float32)
triton_poi_fused_div_sum_3[grid(256)](buf7, buf11, 256, XBLOCK=256,
num_warps=4, num_stages=1)
return reinterpret_tensor(buf10, (4, 4, 4), (16, 4, 1), 0
), buf11, reinterpret_tensor(primals_1, (16, 4), (4, 1), 0
), reinterpret_tensor(primals_2, (64, 4), (4, 1), 0
), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0
), buf7, reinterpret_tensor(buf9, (16, 4), (4, 1), 0
), primals_6, reinterpret_tensor(buf2, (16, 1, 16), (1, 1, 16), 0
), reinterpret_tensor(buf3, (16, 1, 4), (1, 1, 16), 0
), reinterpret_tensor(buf1, (16, 16, 1), (1, 16, 1), 0)
class MultiheadAttentionNew(nn.Module):
"""Multi-headed attention.
See "Attention Is All You Need" for more details.
"""
def __init__(self, embed_dim, num_heads, dropout=0.0, bias=True,
add_bias_kv=False, add_zero_attn=False):
super().__init__()
self.embed_dim = embed_dim
self.num_heads = num_heads
self.dropout = dropout
self.head_dim = embed_dim // num_heads
assert self.head_dim * num_heads == self.embed_dim, 'embed_dim must be divisible by num_heads'
self.scaling = self.head_dim ** -0.5
self.in_proj_weight = Parameter(torch.Tensor(3 * embed_dim, embed_dim))
if bias:
self.in_proj_bias = Parameter(torch.Tensor(3 * embed_dim))
else:
self.register_parameter('in_proj_bias', None)
self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
if add_bias_kv:
self.bias_k = Parameter(torch.Tensor(1, 1, embed_dim))
self.bias_v = Parameter(torch.Tensor(1, 1, embed_dim))
else:
self.bias_k = self.bias_v = None
self.add_zero_attn = add_zero_attn
self.reset_parameters()
self.onnx_trace = False
def prepare_for_onnx_export_(self):
self.onnx_trace = True
def reset_parameters(self):
nn.init.xavier_uniform_(self.in_proj_weight)
nn.init.xavier_uniform_(self.out_proj.weight)
if self.in_proj_bias is not None:
nn.init.constant_(self.in_proj_bias, 0.0)
nn.init.constant_(self.out_proj.bias, 0.0)
if self.bias_k is not None:
nn.init.xavier_normal_(self.bias_k)
if self.bias_v is not None:
nn.init.xavier_normal_(self.bias_v)
def in_proj_qkv(self, query):
return self._in_proj(query).chunk(3, dim=-1)
def in_proj_kv(self, key):
return self._in_proj(key, start=self.embed_dim).chunk(2, dim=-1)
def in_proj_q(self, query):
return self._in_proj(query, end=self.embed_dim)
def in_proj_k(self, key):
return self._in_proj(key, start=self.embed_dim, end=2 * self.embed_dim)
def in_proj_v(self, value):
return self._in_proj(value, start=2 * self.embed_dim)
def _in_proj(self, input, start=0, end=None):
weight = self.in_proj_weight
bias = self.in_proj_bias
weight = weight[start:end, :]
if bias is not None:
bias = bias[start:end]
return F.linear(input, weight, bias)
def reorder_incremental_state(self, incremental_state, new_order):
"""Reorder buffered internal state (for incremental generation)."""
input_buffer = self._get_input_buffer(incremental_state)
if input_buffer is not None:
for k in input_buffer.keys():
input_buffer[k] = input_buffer[k].index_select(0, new_order)
self._set_input_buffer(incremental_state, input_buffer)
def _get_input_buffer(self, incremental_state):
return utils.get_incremental_state(self, incremental_state,
'attn_state') or {}
def _set_input_buffer(self, incremental_state, buffer):
utils.set_incremental_state(self, incremental_state, 'attn_state',
buffer)
def forward(self, input_0, input_1, input_2):
primals_4 = self.in_proj_weight
primals_5 = self.in_proj_bias
primals_6 = self.out_proj.weight
primals_7 = self.out_proj.bias
primals_1 = input_0
primals_2 = input_1
primals_3 = input_2
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7])
return output[0], output[1]
|
edbltn/fairseq
|
MultiheadAttention
| false | 12,347 |
[
"BSD-3-Clause"
] | 0 |
e4d25fd96f1e38190400dbbdbc77eeda71ac50a0
|
https://github.com/edbltn/fairseq/tree/e4d25fd96f1e38190400dbbdbc77eeda71ac50a0
|
StackTime
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/aq/caqlpmcvl2fgbcutymqnd6wdh2kdoybf6voaz6hssdn2hgciacdm.py
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cat => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%arg0_1, %slice_scatter_default, %slice_scatter_default_1, %slice_scatter_default_2], 2), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4) % 16
x0 = xindex % 4
x4 = (xindex // 64)
x3 = (xindex // 256)
x2 = (xindex // 64) % 4
x5 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (4*x1) + (16*x4)), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = x3
tmp11 = tl.full([1], 3, tl.int64)
tmp12 = tmp10 < tmp11
tmp13 = tmp12 & tmp9
tmp14 = tl.load(in_ptr0 + (64 + x0 + (4*((-4) + x1)) + (16*x4)), tmp13 & xmask, other=0.0)
tmp15 = 0.0
tmp16 = tl.where(tmp12, tmp14, tmp15)
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp9, tmp16, tmp17)
tmp19 = tmp0 >= tmp7
tmp20 = tl.full([1], 12, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.full([1], 2, tl.int64)
tmp24 = tmp10 < tmp23
tmp25 = tmp24 & tmp22
tmp26 = tl.load(in_ptr0 + (128 + x0 + (4*((-8) + x1)) + (16*x4)), tmp25 & xmask, other=0.0)
tmp27 = tl.where(tmp24, tmp26, tmp15)
tmp28 = tl.full(tmp27.shape, 0.0, tmp27.dtype)
tmp29 = tl.where(tmp22, tmp27, tmp28)
tmp30 = tmp0 >= tmp20
tmp31 = tl.full([1], 16, tl.int64)
tmp32 = tmp0 < tmp31
tmp33 = tl.full([1], 1, tl.int64)
tmp34 = tmp10 < tmp33
tmp35 = tmp34 & tmp30
tmp36 = tl.load(in_ptr0 + (192 + x0 + (4*((-12) + x1)) + (16*x2)), tmp35 & xmask, eviction_policy='evict_last', other=0.0)
tmp37 = tl.where(tmp34, tmp36, tmp15)
tmp38 = tl.full(tmp37.shape, 0.0, tmp37.dtype)
tmp39 = tl.where(tmp30, tmp37, tmp38)
tmp40 = tl.where(tmp22, tmp29, tmp39)
tmp41 = tl.where(tmp9, tmp18, tmp40)
tmp42 = tl.where(tmp4, tmp5, tmp41)
tl.store(out_ptr0 + (x5), tmp42, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/vc/cvchxawa6ugiimgj4cxrq33nbcizxy4qbfaisgrluclblyp36bzs.py
# Topologically Sorted Source Nodes: [truediv, ceil, x_lens], Original ATen: [aten.div, aten.ceil, aten._to_copy]
# Source node to ATen node mapping:
# ceil => ceil
# truediv => div
# x_lens => convert_element_type
# Graph fragment:
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%arg1_1, 4), kwargs = {})
# %ceil : [num_users=1] = call_function[target=torch.ops.aten.ceil.default](args = (%div,), kwargs = {})
# %convert_element_type : [num_users=1] = call_function[target=torch.ops.prims.convert_element_type.default](args = (%ceil, torch.int32), kwargs = {})
triton_poi_fused__to_copy_ceil_div_1 = async_compile.triton('triton_poi_fused__to_copy_ceil_div_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__to_copy_ceil_div_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__to_copy_ceil_div_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tmp3 = libdevice.ceil(tmp2)
tmp4 = tmp3.to(tl.int32)
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 16, 4), (256, 64, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(arg0_1, buf0, 1024, grid=grid(1024), stream=stream0)
del arg0_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.int32)
# Topologically Sorted Source Nodes: [truediv, ceil, x_lens], Original ATen: [aten.div, aten.ceil, aten._to_copy]
triton_poi_fused__to_copy_ceil_div_1.run(arg1_1, buf1, 256, grid=grid(256), stream=stream0)
del arg1_1
return (reinterpret_tensor(buf0, (1, 4, 16, 4), (1024, 64, 4, 1), 0), buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.onnx
class StackTime(torch.nn.Module):
__constants__ = ['factor']
def __init__(self, factor):
super().__init__()
self.factor = int(factor)
def forward(self, x, x_lens):
seq = [x]
for i in range(1, self.factor):
tmp = torch.zeros_like(x)
tmp[:-i, :, :] = x[i:, :, :]
seq.append(tmp)
x_lens = torch.ceil(x_lens.float() / self.factor).int()
return torch.cat(seq, dim=2)[::self.factor, :, :], x_lens
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'factor': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.onnx
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4 % 16
x0 = xindex % 4
x4 = xindex // 64
x3 = xindex // 256
x2 = xindex // 64 % 4
x5 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 4 * x1 + 16 * x4), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = x3
tmp11 = tl.full([1], 3, tl.int64)
tmp12 = tmp10 < tmp11
tmp13 = tmp12 & tmp9
tmp14 = tl.load(in_ptr0 + (64 + x0 + 4 * (-4 + x1) + 16 * x4), tmp13 &
xmask, other=0.0)
tmp15 = 0.0
tmp16 = tl.where(tmp12, tmp14, tmp15)
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp9, tmp16, tmp17)
tmp19 = tmp0 >= tmp7
tmp20 = tl.full([1], 12, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.full([1], 2, tl.int64)
tmp24 = tmp10 < tmp23
tmp25 = tmp24 & tmp22
tmp26 = tl.load(in_ptr0 + (128 + x0 + 4 * (-8 + x1) + 16 * x4), tmp25 &
xmask, other=0.0)
tmp27 = tl.where(tmp24, tmp26, tmp15)
tmp28 = tl.full(tmp27.shape, 0.0, tmp27.dtype)
tmp29 = tl.where(tmp22, tmp27, tmp28)
tmp30 = tmp0 >= tmp20
tl.full([1], 16, tl.int64)
tmp33 = tl.full([1], 1, tl.int64)
tmp34 = tmp10 < tmp33
tmp35 = tmp34 & tmp30
tmp36 = tl.load(in_ptr0 + (192 + x0 + 4 * (-12 + x1) + 16 * x2), tmp35 &
xmask, eviction_policy='evict_last', other=0.0)
tmp37 = tl.where(tmp34, tmp36, tmp15)
tmp38 = tl.full(tmp37.shape, 0.0, tmp37.dtype)
tmp39 = tl.where(tmp30, tmp37, tmp38)
tmp40 = tl.where(tmp22, tmp29, tmp39)
tmp41 = tl.where(tmp9, tmp18, tmp40)
tmp42 = tl.where(tmp4, tmp5, tmp41)
tl.store(out_ptr0 + x5, tmp42, xmask)
@triton.jit
def triton_poi_fused__to_copy_ceil_div_1(in_ptr0, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tmp3 = libdevice.ceil(tmp2)
tmp4 = tmp3.to(tl.int32)
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 16, 4), (256, 64, 4, 1), torch.float32
)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(1024)](arg0_1, buf0, 1024, XBLOCK=128,
num_warps=4, num_stages=1)
del arg0_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.int32)
triton_poi_fused__to_copy_ceil_div_1[grid(256)](arg1_1, buf1, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg1_1
return reinterpret_tensor(buf0, (1, 4, 16, 4), (1024, 64, 4, 1), 0), buf1
class StackTimeNew(torch.nn.Module):
__constants__ = ['factor']
def __init__(self, factor):
super().__init__()
self.factor = int(factor)
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0], output[1]
|
ephrem-git/inference
|
StackTime
| false | 12,348 |
[
"Apache-2.0"
] | 0 |
bfbda5fc419364c3f71b5b1640f6c00e7675b212
|
https://github.com/ephrem-git/inference/tree/bfbda5fc419364c3f71b5b1640f6c00e7675b212
|
SharpenedCosineSimilarity
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/k2/ck2ifjln7mgxcq7f4l4ctfrzcqh5srlzt3rsj2xkjsdz5mju6rli.py
# Topologically Sorted Source Nodes: [square, square_sum, add, sqrt, truediv, square_1, x_norm, x_2, sign, abs_1, x_3, x_4, mul], Original ATen: [aten.pow, aten.sum, aten.add, aten.sqrt, aten.div, aten.mul, aten.sign, aten.abs]
# Source node to ATen node mapping:
# abs_1 => abs_1
# add => add
# mul => mul_1
# sign => sign
# sqrt => sqrt
# square => pow_1
# square_1 => pow_2
# square_sum => sum_1
# truediv => div
# x_2 => mul, sum_3
# x_3 => add_4
# x_4 => pow_6
# x_norm => add_1
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%view, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [1, 4], True), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sum_1, 1e-12), kwargs = {})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%add,), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_2, 100), kwargs = {})
# %pow_2 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%div, 2), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sqrt, %pow_2), kwargs = {})
# %sum_3 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%permute, [4], True), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sum_3, %permute_1), kwargs = {})
# %sign : [num_users=1] = call_function[target=torch.ops.aten.sign.default](args = (%view_1,), kwargs = {})
# %abs_1 : [num_users=1] = call_function[target=torch.ops.aten.abs.default](args = (%view_1,), kwargs = {})
# %add_4 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%abs_1, 1e-12), kwargs = {})
# %pow_6 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Tensor](args = (%add_4, %view_2), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sign, %pow_6), kwargs = {})
triton_poi_fused_abs_add_div_mul_pow_sign_sqrt_sum_0 = async_compile.triton('triton_poi_fused_abs_add_div_mul_pow_sign_sqrt_sum_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_abs_add_div_mul_pow_sign_sqrt_sum_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 7, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_abs_add_div_mul_pow_sign_sqrt_sum_0(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = (xindex // 16)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (64*x1)), xmask)
tmp2 = tl.load(in_ptr0 + (16 + x0 + (64*x1)), xmask)
tmp5 = tl.load(in_ptr0 + (32 + x0 + (64*x1)), xmask)
tmp8 = tl.load(in_ptr0 + (48 + x0 + (64*x1)), xmask)
tmp14 = tl.load(in_ptr1 + (0))
tmp15 = tl.broadcast_to(tmp14, [XBLOCK])
tmp27 = tl.load(in_ptr2 + (0))
tmp28 = tl.broadcast_to(tmp27, [XBLOCK])
tmp44 = tl.load(in_ptr3 + (0))
tmp45 = tl.broadcast_to(tmp44, [XBLOCK])
tmp1 = tmp0 * tmp0
tmp3 = tmp2 * tmp2
tmp4 = tmp1 + tmp3
tmp6 = tmp5 * tmp5
tmp7 = tmp4 + tmp6
tmp9 = tmp8 * tmp8
tmp10 = tmp7 + tmp9
tmp11 = 1e-12
tmp12 = tmp10 + tmp11
tmp13 = libdevice.sqrt(tmp12)
tmp16 = 0.01
tmp17 = tmp15 * tmp16
tmp18 = tmp17 * tmp17
tmp19 = tmp13 + tmp18
tmp20 = tmp0 / tmp19
tmp21 = tmp2 / tmp19
tmp22 = tmp20 + tmp21
tmp23 = tmp5 / tmp19
tmp24 = tmp22 + tmp23
tmp25 = tmp8 / tmp19
tmp26 = tmp24 + tmp25
tmp29 = tmp28 * tmp28
tmp30 = tmp29 + tmp11
tmp31 = libdevice.sqrt(tmp30)
tmp32 = tmp31 + tmp18
tmp33 = tmp28 / tmp32
tmp34 = tmp26 * tmp33
tmp35 = tl.full([1], 0, tl.int32)
tmp36 = tmp35 < tmp34
tmp37 = tmp36.to(tl.int8)
tmp38 = tmp34 < tmp35
tmp39 = tmp38.to(tl.int8)
tmp40 = tmp37 - tmp39
tmp41 = tmp40.to(tmp34.dtype)
tmp42 = tl_math.abs(tmp34)
tmp43 = tmp42 + tmp11
tmp46 = 0.1
tmp47 = tmp45 * tmp46
tmp48 = tmp47 * tmp47
tmp49 = libdevice.pow(tmp43, tmp48)
tmp50 = tmp41 * tmp49
tl.store(in_out_ptr0 + (x2), tmp50, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (1, ), (1, ))
assert_size_stride(primals_3, (1, 1, 1), (1, 1, 1))
assert_size_stride(primals_4, (1, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 4, 1), (16, 64, 4, 1, 64), torch.float32)
buf1 = reinterpret_tensor(buf0, (4, 1, 4, 4, 1, 1), (16, 64, 4, 1, 64, 64), 0); del buf0 # reuse
buf2 = reinterpret_tensor(buf1, (4, 1, 4, 4), (16, 16, 4, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [square, square_sum, add, sqrt, truediv, square_1, x_norm, x_2, sign, abs_1, x_3, x_4, mul], Original ATen: [aten.pow, aten.sum, aten.add, aten.sqrt, aten.div, aten.mul, aten.sign, aten.abs]
stream0 = get_raw_stream(0)
triton_poi_fused_abs_add_div_mul_pow_sign_sqrt_sum_0.run(buf2, primals_1, primals_2, primals_3, primals_4, 64, grid=grid(64), stream=stream0)
return (buf2, primals_1, primals_2, primals_3, primals_4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((1, 1, 1), (1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
def unfold2d(x, kernel_size: 'int', stride: 'int', padding: 'int'):
x = F.pad(x, [padding] * 4)
bs, in_c, h, w = x.size()
ks = kernel_size
strided_x = x.as_strided((bs, in_c, (h - ks) // stride + 1, (w - ks) //
stride + 1, ks, ks), (in_c * h * w, h * w, stride * w, stride, w, 1))
return strided_x
class SharpenedCosineSimilarity(nn.Module):
def __init__(self, in_channels=1, out_channels=1, kernel_size=1, stride
=1, padding=0, eps=1e-12):
super(SharpenedCosineSimilarity, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.eps = eps
self.padding = int(padding)
w = torch.empty(out_channels, in_channels, kernel_size, kernel_size)
nn.init.xavier_uniform_(w)
self.w = nn.Parameter(w.view(out_channels, in_channels, -1),
requires_grad=True)
self.p_scale = 10
p_init = 2 ** 0.5 * self.p_scale
self.register_parameter('p', nn.Parameter(torch.empty(out_channels)))
nn.init.constant_(self.p, p_init)
self.q_scale = 100
self.register_parameter('q', nn.Parameter(torch.empty(1)))
nn.init.constant_(self.q, 10)
def forward(self, x):
x = unfold2d(x, kernel_size=self.kernel_size, stride=self.stride,
padding=self.padding)
n, c, h, w, _, _ = x.shape
x = x.reshape(n, c, h, w, -1)
square_sum = torch.sum(torch.square(x), [1, 4], keepdim=True)
x_norm = torch.add(torch.sqrt(square_sum + self.eps), torch.square(
self.q / self.q_scale))
square_sum = torch.sum(torch.square(self.w), [1, 2], keepdim=True)
w_norm = torch.add(torch.sqrt(square_sum + self.eps), torch.square(
self.q / self.q_scale))
x = torch.einsum('nchwl,vcl->nvhw', x / x_norm, self.w / w_norm)
sign = torch.sign(x)
x = torch.abs(x) + self.eps
x = x.pow(torch.square(self.p / self.p_scale).view(1, -1, 1, 1))
return sign * x
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
import torch.nn.functional as F
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_abs_add_div_mul_pow_sign_sqrt_sum_0(in_out_ptr0,
in_ptr0, in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = xindex // 16
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 64 * x1), xmask)
tmp2 = tl.load(in_ptr0 + (16 + x0 + 64 * x1), xmask)
tmp5 = tl.load(in_ptr0 + (32 + x0 + 64 * x1), xmask)
tmp8 = tl.load(in_ptr0 + (48 + x0 + 64 * x1), xmask)
tmp14 = tl.load(in_ptr1 + 0)
tmp15 = tl.broadcast_to(tmp14, [XBLOCK])
tmp27 = tl.load(in_ptr2 + 0)
tmp28 = tl.broadcast_to(tmp27, [XBLOCK])
tmp44 = tl.load(in_ptr3 + 0)
tmp45 = tl.broadcast_to(tmp44, [XBLOCK])
tmp1 = tmp0 * tmp0
tmp3 = tmp2 * tmp2
tmp4 = tmp1 + tmp3
tmp6 = tmp5 * tmp5
tmp7 = tmp4 + tmp6
tmp9 = tmp8 * tmp8
tmp10 = tmp7 + tmp9
tmp11 = 1e-12
tmp12 = tmp10 + tmp11
tmp13 = libdevice.sqrt(tmp12)
tmp16 = 0.01
tmp17 = tmp15 * tmp16
tmp18 = tmp17 * tmp17
tmp19 = tmp13 + tmp18
tmp20 = tmp0 / tmp19
tmp21 = tmp2 / tmp19
tmp22 = tmp20 + tmp21
tmp23 = tmp5 / tmp19
tmp24 = tmp22 + tmp23
tmp25 = tmp8 / tmp19
tmp26 = tmp24 + tmp25
tmp29 = tmp28 * tmp28
tmp30 = tmp29 + tmp11
tmp31 = libdevice.sqrt(tmp30)
tmp32 = tmp31 + tmp18
tmp33 = tmp28 / tmp32
tmp34 = tmp26 * tmp33
tmp35 = tl.full([1], 0, tl.int32)
tmp36 = tmp35 < tmp34
tmp37 = tmp36.to(tl.int8)
tmp38 = tmp34 < tmp35
tmp39 = tmp38.to(tl.int8)
tmp40 = tmp37 - tmp39
tmp41 = tmp40.to(tmp34.dtype)
tmp42 = tl_math.abs(tmp34)
tmp43 = tmp42 + tmp11
tmp46 = 0.1
tmp47 = tmp45 * tmp46
tmp48 = tmp47 * tmp47
tmp49 = libdevice.pow(tmp43, tmp48)
tmp50 = tmp41 * tmp49
tl.store(in_out_ptr0 + x2, tmp50, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (1,), (1,))
assert_size_stride(primals_3, (1, 1, 1), (1, 1, 1))
assert_size_stride(primals_4, (1,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 4, 1), (16, 64, 4, 1, 64),
torch.float32)
buf1 = reinterpret_tensor(buf0, (4, 1, 4, 4, 1, 1), (16, 64, 4, 1,
64, 64), 0)
del buf0
buf2 = reinterpret_tensor(buf1, (4, 1, 4, 4), (16, 16, 4, 1), 0)
del buf1
get_raw_stream(0)
triton_poi_fused_abs_add_div_mul_pow_sign_sqrt_sum_0[grid(64)](buf2,
primals_1, primals_2, primals_3, primals_4, 64, XBLOCK=64,
num_warps=1, num_stages=1)
return buf2, primals_1, primals_2, primals_3, primals_4
def unfold2d(x, kernel_size: 'int', stride: 'int', padding: 'int'):
x = F.pad(x, [padding] * 4)
bs, in_c, h, w = x.size()
ks = kernel_size
strided_x = x.as_strided((bs, in_c, (h - ks) // stride + 1, (w - ks) //
stride + 1, ks, ks), (in_c * h * w, h * w, stride * w, stride, w, 1))
return strided_x
class SharpenedCosineSimilarityNew(nn.Module):
def __init__(self, in_channels=1, out_channels=1, kernel_size=1, stride
=1, padding=0, eps=1e-12):
super(SharpenedCosineSimilarityNew, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.eps = eps
self.padding = int(padding)
w = torch.empty(out_channels, in_channels, kernel_size, kernel_size)
nn.init.xavier_uniform_(w)
self.w = nn.Parameter(w.view(out_channels, in_channels, -1),
requires_grad=True)
self.p_scale = 10
p_init = 2 ** 0.5 * self.p_scale
self.register_parameter('p', nn.Parameter(torch.empty(out_channels)))
nn.init.constant_(self.p, p_init)
self.q_scale = 100
self.register_parameter('q', nn.Parameter(torch.empty(1)))
nn.init.constant_(self.q, 10)
def forward(self, input_0):
primals_3 = self.w
primals_2 = self.p
primals_4 = self.q
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
enzokro/sharpened_cosine_similarity_torch
|
SharpenedCosineSimilarity
| false | 12,349 |
[
"MIT"
] | 0 |
150c84f5cf81721baf097abdc0d4ac772fb39fc4
|
https://github.com/enzokro/sharpened_cosine_similarity_torch/tree/150c84f5cf81721baf097abdc0d4ac772fb39fc4
|
PredictionHead
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ff/cff2cdwjkv2nl2npzg2amb6bbqjuqegcu5trpyf6uikstnfivpbv.py
# Topologically Sorted Source Nodes: [class_logits_1], Original ATen: [aten.clone, aten._unsafe_view]
# Source node to ATen node mapping:
# class_logits_1 => clone, view
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute,), kwargs = {memory_format: torch.contiguous_format})
# %view : [num_users=1] = call_function[target=torch.ops.aten.reshape.default](args = (%clone, [4, 64, 4]), kwargs = {})
triton_poi_fused__unsafe_view_clone_0 = async_compile.triton('triton_poi_fused__unsafe_view_clone_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64, 16], tile_hint=TileHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__unsafe_view_clone_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__unsafe_view_clone_0(in_out_ptr0, in_ptr0, in_ptr1, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 64
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = (yindex // 16)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (16*x2) + (256*y1)), xmask & ymask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x2), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.debug_barrier()
tl.store(in_out_ptr0 + (x2 + (16*y3)), tmp2, xmask & ymask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (16, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_3, (16, ), (1, ))
assert_size_stride(primals_4, (16, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_5, (16, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [class_logits], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_1, primals_2, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 16, 4, 4), (256, 16, 4, 1))
# Topologically Sorted Source Nodes: [box_regression], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(primals_1, primals_4, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 16, 4, 4), (256, 16, 4, 1))
buf2 = empty_strided_cuda((4, 4, 4, 16), (256, 64, 16, 1), torch.float32)
buf3 = reinterpret_tensor(buf2, (4, 64, 4), (256, 4, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [class_logits_1], Original ATen: [aten.clone, aten._unsafe_view]
stream0 = get_raw_stream(0)
triton_poi_fused__unsafe_view_clone_0.run(buf3, buf0, primals_3, 64, 16, grid=grid(64, 16), stream=stream0)
del primals_3
buf4 = reinterpret_tensor(buf0, (4, 4, 4, 16), (256, 64, 16, 1), 0); del buf0 # reuse
buf5 = reinterpret_tensor(buf4, (4, 64, 4), (256, 4, 1), 0); del buf4 # reuse
# Topologically Sorted Source Nodes: [box_regression_1], Original ATen: [aten.clone, aten._unsafe_view]
triton_poi_fused__unsafe_view_clone_0.run(buf5, buf1, primals_5, 64, 16, grid=grid(64, 16), stream=stream0)
del buf1
del primals_5
return (buf3, buf5, primals_1, primals_2, primals_4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((16, 4, 1, 1), (4, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((16, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((16, 4, 1, 1), (4, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((16, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.onnx
class PredictionHead(nn.Module):
def __init__(self, in_channels, num_classes, num_anchors):
super(PredictionHead, self).__init__()
self.classification = nn.Conv2d(in_channels, num_classes *
num_anchors, kernel_size=1)
self.regression = nn.Conv2d(in_channels, num_anchors * 4, kernel_size=1
)
self.num_classes = num_classes
self.num_anchors = num_anchors
def forward(self, x):
bs = x.shape[0]
class_logits = self.classification(x)
box_regression = self.regression(x)
class_logits = class_logits.permute(0, 2, 3, 1).reshape(bs, -1,
self.num_classes)
box_regression = box_regression.permute(0, 2, 3, 1).reshape(bs, -1, 4)
return class_logits, box_regression
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_channels': 4, 'num_classes': 4, 'num_anchors': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
import torch.onnx
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused__unsafe_view_clone_0(in_out_ptr0, in_ptr0, in_ptr1,
ynumel, xnumel, YBLOCK: tl.constexpr, XBLOCK: tl.constexpr):
ynumel = 64
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = yindex // 16
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 16 * x2 + 256 * y1), xmask & ymask,
eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + x2, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.debug_barrier()
tl.store(in_out_ptr0 + (x2 + 16 * y3), tmp2, xmask & ymask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (16, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_3, (16,), (1,))
assert_size_stride(primals_4, (16, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_5, (16,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_1, primals_2, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 16, 4, 4), (256, 16, 4, 1))
buf1 = extern_kernels.convolution(primals_1, primals_4, stride=(1,
1), padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 16, 4, 4), (256, 16, 4, 1))
buf2 = empty_strided_cuda((4, 4, 4, 16), (256, 64, 16, 1), torch.
float32)
buf3 = reinterpret_tensor(buf2, (4, 64, 4), (256, 4, 1), 0)
del buf2
get_raw_stream(0)
triton_poi_fused__unsafe_view_clone_0[grid(64, 16)](buf3, buf0,
primals_3, 64, 16, XBLOCK=16, YBLOCK=16, num_warps=4, num_stages=1)
del primals_3
buf4 = reinterpret_tensor(buf0, (4, 4, 4, 16), (256, 64, 16, 1), 0)
del buf0
buf5 = reinterpret_tensor(buf4, (4, 64, 4), (256, 4, 1), 0)
del buf4
triton_poi_fused__unsafe_view_clone_0[grid(64, 16)](buf5, buf1,
primals_5, 64, 16, XBLOCK=16, YBLOCK=16, num_warps=4, num_stages=1)
del buf1
del primals_5
return buf3, buf5, primals_1, primals_2, primals_4
class PredictionHeadNew(nn.Module):
def __init__(self, in_channels, num_classes, num_anchors):
super(PredictionHeadNew, self).__init__()
self.classification = nn.Conv2d(in_channels, num_classes *
num_anchors, kernel_size=1)
self.regression = nn.Conv2d(in_channels, num_anchors * 4, kernel_size=1
)
self.num_classes = num_classes
self.num_anchors = num_anchors
def forward(self, input_0):
primals_2 = self.classification.weight
primals_3 = self.classification.bias
primals_4 = self.regression.weight
primals_5 = self.regression.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4, primals_5])
return output[0], output[1]
|
ephrem-git/inference
|
PredictionHead
| false | 12,350 |
[
"Apache-2.0"
] | 0 |
bfbda5fc419364c3f71b5b1640f6c00e7675b212
|
https://github.com/ephrem-git/inference/tree/bfbda5fc419364c3f71b5b1640f6c00e7675b212
|
ExtClassifier
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/f3/cf3wjo3codglmel3mdjaodbq3s3viwdoc74iaz5e3kntwsnjtjqi.py
# Topologically Sorted Source Nodes: [sent_scores], Original ATen: [aten.sigmoid]
# Source node to ATen node mapping:
# sent_scores => sigmoid
# Graph fragment:
# %sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%squeeze,), kwargs = {})
triton_poi_fused_sigmoid_0 = async_compile.triton('triton_poi_fused_sigmoid_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_sigmoid_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_sigmoid_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + (x0), xmask)
tmp1 = tl.load(in_ptr0 + (0))
tmp2 = tl.broadcast_to(tmp1, [XBLOCK])
tmp3 = tmp0 + tmp2
tmp4 = tl.sigmoid(tmp3)
tl.store(in_out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (1, 4), (4, 1))
assert_size_stride(primals_2, (1, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 1), (1, 4), 0), out=buf0)
del primals_1
buf1 = reinterpret_tensor(buf0, (4, 4, 4), (16, 4, 1), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [sent_scores], Original ATen: [aten.sigmoid]
stream0 = get_raw_stream(0)
triton_poi_fused_sigmoid_0.run(buf1, primals_2, 64, grid=grid(64), stream=stream0)
del primals_2
return (buf1, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.cuda
import torch.distributed
class ExtClassifier(nn.Module):
def __init__(self, hidden_size):
super(ExtClassifier, self).__init__()
self.linear1 = nn.Linear(hidden_size, 1)
self.sigmoid = nn.Sigmoid()
def forward(self, x, mask=None):
h = self.linear1(x).squeeze(-1)
if mask:
sent_scores = self.sigmoid(h) * mask.float()
else:
sent_scores = self.sigmoid(h)
return sent_scores
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'hidden_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
import torch.cuda
import torch.distributed
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_sigmoid_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + x0, xmask)
tmp1 = tl.load(in_ptr0 + 0)
tmp2 = tl.broadcast_to(tmp1, [XBLOCK])
tmp3 = tmp0 + tmp2
tmp4 = tl.sigmoid(tmp3)
tl.store(in_out_ptr0 + x0, tmp4, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (1, 4), (4, 1))
assert_size_stride(primals_2, (1,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 1), (1, 4), 0), out=buf0)
del primals_1
buf1 = reinterpret_tensor(buf0, (4, 4, 4), (16, 4, 1), 0)
del buf0
get_raw_stream(0)
triton_poi_fused_sigmoid_0[grid(64)](buf1, primals_2, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del primals_2
return buf1, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), buf1
class ExtClassifierNew(nn.Module):
def __init__(self, hidden_size):
super(ExtClassifierNew, self).__init__()
self.linear1 = nn.Linear(hidden_size, 1)
self.sigmoid = nn.Sigmoid()
def forward(self, input_0):
primals_1 = self.linear1.weight
primals_2 = self.linear1.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
eric-zhizu/OpenNMT-kpg-release
|
ExtClassifier
| false | 12,351 |
[
"MIT"
] | 0 |
9f15dea6f663425eef2157845c4c8042ad845c11
|
https://github.com/eric-zhizu/OpenNMT-kpg-release/tree/9f15dea6f663425eef2157845c4c8042ad845c11
|
SharpenedCosineSimilarity_ConvImpl
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/6z/c6zdymjqbcg6tzw4vjf7cx2zk6qvfiusar4uc4wjnopf5siefhn7.py
# Topologically Sorted Source Nodes: [w_norm, truediv, q_sqr, add, add_1, w_normed], Original ATen: [aten.linalg_vector_norm, aten.div, aten.pow, aten.add]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# q_sqr => pow_3
# truediv => div
# w_norm => pow_1, pow_2, sum_1
# w_normed => div_1
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%view, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [1, 2, 3], True), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_1, 0.5), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_2, 100), kwargs = {})
# %pow_3 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%div, 2), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%pow_2, 1e-12), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%add, %pow_3), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%view, %add_1), kwargs = {})
triton_poi_fused_add_div_linalg_vector_norm_pow_0 = async_compile.triton('triton_poi_fused_add_div_linalg_vector_norm_pow_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_linalg_vector_norm_pow_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_linalg_vector_norm_pow_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
tmp0 = tl.load(in_ptr0 + (0))
tmp1 = tl.broadcast_to(tmp0, [XBLOCK])
tmp6 = tl.load(in_ptr1 + (0))
tmp7 = tl.broadcast_to(tmp6, [XBLOCK])
tmp2 = tmp1 * tmp1
tmp3 = libdevice.sqrt(tmp2)
tmp4 = 1e-12
tmp5 = tmp3 + tmp4
tmp8 = 0.01
tmp9 = tmp7 * tmp8
tmp10 = tmp9 * tmp9
tmp11 = tmp5 + tmp10
tmp12 = tmp1 / tmp11
tl.store(out_ptr0 + (tl.full([XBLOCK], 0, tl.int32)), tmp12, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ur/curr2hysdbhckrnsu3xyxh2x7dtqbz755nv7utpsbca5nio22rla.py
# Topologically Sorted Source Nodes: [truediv, q_sqr, pow_2, avg_pool2d, x_norm_squared, add_2, sqrt, add_3, y, sign, abs_1, y_1, y_2, mul], Original ATen: [aten.div, aten.pow, aten.avg_pool2d, aten.sum, aten.add, aten.sqrt, aten.sign, aten.abs, aten.mul]
# Source node to ATen node mapping:
# abs_1 => abs_1
# add_2 => add_2
# add_3 => add_3
# avg_pool2d => avg_pool2d
# mul => mul
# pow_2 => pow_4
# q_sqr => pow_3
# sign => sign
# sqrt => sqrt
# truediv => div
# x_norm_squared => sum_2
# y => div_2
# y_1 => add_4
# y_2 => pow_6
# Graph fragment:
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_2, 100), kwargs = {})
# %pow_3 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%div, 2), kwargs = {})
# %pow_4 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_3, 2), kwargs = {})
# %avg_pool2d : [num_users=1] = call_function[target=torch.ops.aten.avg_pool2d.default](args = (%pow_4, [1, 1], [1, 1], [0, 0], False, True, 1), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%avg_pool2d, [1], True), kwargs = {})
# %add_2 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sum_2, 1e-12), kwargs = {})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%add_2,), kwargs = {})
# %add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%sqrt, %pow_3), kwargs = {})
# %div_2 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%convolution, %add_3), kwargs = {})
# %sign : [num_users=1] = call_function[target=torch.ops.aten.sign.default](args = (%div_2,), kwargs = {})
# %abs_1 : [num_users=1] = call_function[target=torch.ops.aten.abs.default](args = (%div_2,), kwargs = {})
# %add_4 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%abs_1, 1e-12), kwargs = {})
# %pow_6 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Tensor](args = (%add_4, %view_1), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sign, %pow_6), kwargs = {})
triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1 = async_compile.triton('triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 16384
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), None)
tmp7 = tl.load(in_ptr1 + (0))
tmp8 = tl.broadcast_to(tmp7, [XBLOCK])
tmp13 = tl.load(in_ptr2 + (x0), None)
tmp24 = tl.load(in_ptr3 + (0))
tmp25 = tl.broadcast_to(tmp24, [XBLOCK])
tmp1 = tmp0 * tmp0
tmp2 = 1.0
tmp3 = tmp1 * tmp2
tmp4 = 1e-12
tmp5 = tmp3 + tmp4
tmp6 = libdevice.sqrt(tmp5)
tmp9 = 0.01
tmp10 = tmp8 * tmp9
tmp11 = tmp10 * tmp10
tmp12 = tmp6 + tmp11
tmp14 = tmp13 / tmp12
tmp15 = tl.full([1], 0, tl.int32)
tmp16 = tmp15 < tmp14
tmp17 = tmp16.to(tl.int8)
tmp18 = tmp14 < tmp15
tmp19 = tmp18.to(tl.int8)
tmp20 = tmp17 - tmp19
tmp21 = tmp20.to(tmp14.dtype)
tmp22 = tl_math.abs(tmp14)
tmp23 = tmp22 + tmp4
tmp26 = 0.1
tmp27 = tmp25 * tmp26
tmp28 = tmp27 * tmp27
tmp29 = libdevice.pow(tmp23, tmp28)
tmp30 = tmp21 * tmp29
tl.store(out_ptr0 + (x0), tmp12, None)
tl.store(out_ptr1 + (x0), tmp30, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 1, 1), (1, 1, 1))
assert_size_stride(primals_2, (1, ), (1, ))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
assert_size_stride(primals_4, (1, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 1, 1, 1), (1, 1, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [w_norm, truediv, q_sqr, add, add_1, w_normed], Original ATen: [aten.linalg_vector_norm, aten.div, aten.pow, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_linalg_vector_norm_pow_0.run(primals_1, primals_2, buf0, 1, grid=grid(1), stream=stream0)
# Topologically Sorted Source Nodes: [y_denorm], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(primals_3, buf0, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 1, 64, 64), (4096, 4096, 64, 1))
buf2 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1), torch.float32)
buf3 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1), torch.float32)
# Topologically Sorted Source Nodes: [truediv, q_sqr, pow_2, avg_pool2d, x_norm_squared, add_2, sqrt, add_3, y, sign, abs_1, y_1, y_2, mul], Original ATen: [aten.div, aten.pow, aten.avg_pool2d, aten.sum, aten.add, aten.sqrt, aten.sign, aten.abs, aten.mul]
triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1.run(primals_3, primals_2, buf1, primals_4, buf2, buf3, 16384, grid=grid(16384), stream=stream0)
return (buf3, primals_1, primals_2, primals_3, primals_4, buf0, buf1, buf2, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 1, 1), (1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 1, 64, 64), (4096, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class SharpenedCosineSimilarity_ConvImpl(nn.Module):
def __init__(self, in_channels=1, out_channels=1, kernel_size=1, stride
=1, padding=0, eps=1e-12):
super(SharpenedCosineSimilarity_ConvImpl, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.eps = eps
self.padding = int(padding)
w = torch.empty(out_channels, in_channels, kernel_size, kernel_size)
nn.init.xavier_uniform_(w)
self.w = nn.Parameter(w.view(out_channels, in_channels, -1),
requires_grad=True)
self.p_scale = 10
p_init = 2 ** 0.5 * self.p_scale
self.register_parameter('p', nn.Parameter(torch.empty(out_channels)))
nn.init.constant_(self.p, p_init)
self.q_scale = 100
self.register_parameter('q', nn.Parameter(torch.empty(1)))
nn.init.constant_(self.q, 10)
def forward(self, x):
w = self.w.reshape(self.out_channels, self.in_channels, self.
kernel_size, self.kernel_size)
w_norm = torch.linalg.vector_norm(w, dim=(1, 2, 3), keepdim=True)
q_sqr = (self.q / self.q_scale) ** 2
w_normed = w / (w_norm + self.eps + q_sqr)
x_norm_squared = F.avg_pool2d(x ** 2, kernel_size=self.kernel_size,
stride=self.stride, padding=self.padding, divisor_override=1).sum(
dim=1, keepdim=True)
y_denorm = F.conv2d(x, w_normed, bias=None, stride=self.stride,
padding=self.padding)
y = y_denorm / ((x_norm_squared + self.eps).sqrt() + q_sqr)
sign = torch.sign(y)
y = torch.abs(y) + self.eps
p_sqr = (self.p / self.p_scale) ** 2
y = y.pow(p_sqr.reshape(1, -1, 1, 1))
return sign * y
def get_inputs():
return [torch.rand([4, 1, 64, 64])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_div_linalg_vector_norm_pow_0(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
tmp0 = tl.load(in_ptr0 + 0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK])
tmp6 = tl.load(in_ptr1 + 0)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK])
tmp2 = tmp1 * tmp1
tmp3 = libdevice.sqrt(tmp2)
tmp4 = 1e-12
tmp5 = tmp3 + tmp4
tmp8 = 0.01
tmp9 = tmp7 * tmp8
tmp10 = tmp9 * tmp9
tmp11 = tmp5 + tmp10
tmp12 = tmp1 / tmp11
tl.store(out_ptr0 + tl.full([XBLOCK], 0, tl.int32), tmp12, None)
@triton.jit
def triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1(in_ptr0,
in_ptr1, in_ptr2, in_ptr3, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr
):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, None)
tmp7 = tl.load(in_ptr1 + 0)
tmp8 = tl.broadcast_to(tmp7, [XBLOCK])
tmp13 = tl.load(in_ptr2 + x0, None)
tmp24 = tl.load(in_ptr3 + 0)
tmp25 = tl.broadcast_to(tmp24, [XBLOCK])
tmp1 = tmp0 * tmp0
tmp2 = 1.0
tmp3 = tmp1 * tmp2
tmp4 = 1e-12
tmp5 = tmp3 + tmp4
tmp6 = libdevice.sqrt(tmp5)
tmp9 = 0.01
tmp10 = tmp8 * tmp9
tmp11 = tmp10 * tmp10
tmp12 = tmp6 + tmp11
tmp14 = tmp13 / tmp12
tmp15 = tl.full([1], 0, tl.int32)
tmp16 = tmp15 < tmp14
tmp17 = tmp16.to(tl.int8)
tmp18 = tmp14 < tmp15
tmp19 = tmp18.to(tl.int8)
tmp20 = tmp17 - tmp19
tmp21 = tmp20.to(tmp14.dtype)
tmp22 = tl_math.abs(tmp14)
tmp23 = tmp22 + tmp4
tmp26 = 0.1
tmp27 = tmp25 * tmp26
tmp28 = tmp27 * tmp27
tmp29 = libdevice.pow(tmp23, tmp28)
tmp30 = tmp21 * tmp29
tl.store(out_ptr0 + x0, tmp12, None)
tl.store(out_ptr1 + x0, tmp30, None)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 1, 1), (1, 1, 1))
assert_size_stride(primals_2, (1,), (1,))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
assert_size_stride(primals_4, (1,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 1, 1, 1), (1, 1, 1, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_linalg_vector_norm_pow_0[grid(1)](primals_1,
primals_2, buf0, 1, XBLOCK=1, num_warps=1, num_stages=1)
buf1 = extern_kernels.convolution(primals_3, buf0, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 1, 64, 64), (4096, 4096, 64, 1))
buf2 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1),
torch.float32)
buf3 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1),
torch.float32)
triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1[grid
(16384)](primals_3, primals_2, buf1, primals_4, buf2, buf3,
16384, XBLOCK=256, num_warps=4, num_stages=1)
return buf3, primals_1, primals_2, primals_3, primals_4, buf0, buf1, buf2
class SharpenedCosineSimilarity_ConvImplNew(nn.Module):
def __init__(self, in_channels=1, out_channels=1, kernel_size=1, stride
=1, padding=0, eps=1e-12):
super(SharpenedCosineSimilarity_ConvImplNew, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.eps = eps
self.padding = int(padding)
w = torch.empty(out_channels, in_channels, kernel_size, kernel_size)
nn.init.xavier_uniform_(w)
self.w = nn.Parameter(w.view(out_channels, in_channels, -1),
requires_grad=True)
self.p_scale = 10
p_init = 2 ** 0.5 * self.p_scale
self.register_parameter('p', nn.Parameter(torch.empty(out_channels)))
nn.init.constant_(self.p, p_init)
self.q_scale = 100
self.register_parameter('q', nn.Parameter(torch.empty(1)))
nn.init.constant_(self.q, 10)
def forward(self, input_0):
primals_1 = self.w
primals_2 = self.p
primals_4 = self.q
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
enzokro/sharpened_cosine_similarity_torch
|
SharpenedCosineSimilarity_ConvImpl
| false | 12,352 |
[
"MIT"
] | 0 |
150c84f5cf81721baf097abdc0d4ac772fb39fc4
|
https://github.com/enzokro/sharpened_cosine_similarity_torch/tree/150c84f5cf81721baf097abdc0d4ac772fb39fc4
|
SharpenedCosineSimilarityAnnotated
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/6z/c6zdymjqbcg6tzw4vjf7cx2zk6qvfiusar4uc4wjnopf5siefhn7.py
# Topologically Sorted Source Nodes: [w_norm, truediv, q_sqr, add, add_1, w_normed], Original ATen: [aten.linalg_vector_norm, aten.div, aten.pow, aten.add]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# q_sqr => pow_3
# truediv => div
# w_norm => pow_1, pow_2, sum_1
# w_normed => div_1
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_1, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [1, 2, 3], True), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_1, 0.5), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_2, 100), kwargs = {})
# %pow_3 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%div, 2), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%pow_2, 1e-12), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%add, %pow_3), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_1, %add_1), kwargs = {})
triton_poi_fused_add_div_linalg_vector_norm_pow_0 = async_compile.triton('triton_poi_fused_add_div_linalg_vector_norm_pow_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_linalg_vector_norm_pow_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_linalg_vector_norm_pow_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
tmp0 = tl.load(in_ptr0 + (0))
tmp1 = tl.broadcast_to(tmp0, [XBLOCK])
tmp6 = tl.load(in_ptr1 + (0))
tmp7 = tl.broadcast_to(tmp6, [XBLOCK])
tmp2 = tmp1 * tmp1
tmp3 = libdevice.sqrt(tmp2)
tmp4 = 1e-12
tmp5 = tmp3 + tmp4
tmp8 = 0.01
tmp9 = tmp7 * tmp8
tmp10 = tmp9 * tmp9
tmp11 = tmp5 + tmp10
tmp12 = tmp1 / tmp11
tl.store(out_ptr0 + (tl.full([XBLOCK], 0, tl.int32)), tmp12, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/vm/cvm6p2twaljr6l7mxczpn3cb4yxjhcxr7hulcliee2piilm4galp.py
# Topologically Sorted Source Nodes: [truediv, q_sqr, add_2, pow_2, avg_pool2d, x_norm_squared, sqrt, add_3, y, sign, abs_1, y_1, y_2, mul], Original ATen: [aten.div, aten.pow, aten.add, aten.avg_pool2d, aten.sum, aten.sqrt, aten.sign, aten.abs, aten.mul]
# Source node to ATen node mapping:
# abs_1 => abs_1
# add_2 => add_2
# add_3 => add_3
# avg_pool2d => avg_pool2d
# mul => mul
# pow_2 => pow_4
# q_sqr => pow_3
# sign => sign
# sqrt => sqrt
# truediv => div
# x_norm_squared => sum_2
# y => div_2
# y_1 => add_4
# y_2 => pow_6
# Graph fragment:
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_2, 100), kwargs = {})
# %pow_3 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%div, 2), kwargs = {})
# %add_2 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%primals_3, 1e-12), kwargs = {})
# %pow_4 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%add_2, 2), kwargs = {})
# %avg_pool2d : [num_users=1] = call_function[target=torch.ops.aten.avg_pool2d.default](args = (%pow_4, [1, 1], [1, 1], [0, 0], False, True, 1), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%avg_pool2d, [1], True), kwargs = {})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%sum_2,), kwargs = {})
# %add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%sqrt, %pow_3), kwargs = {})
# %div_2 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%convolution, %add_3), kwargs = {})
# %sign : [num_users=1] = call_function[target=torch.ops.aten.sign.default](args = (%div_2,), kwargs = {})
# %abs_1 : [num_users=1] = call_function[target=torch.ops.aten.abs.default](args = (%div_2,), kwargs = {})
# %add_4 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%abs_1, 1e-12), kwargs = {})
# %pow_6 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Tensor](args = (%add_4, %view), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sign, %pow_6), kwargs = {})
triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1 = async_compile.triton('triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 16384
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), None)
tmp7 = tl.load(in_ptr1 + (0))
tmp8 = tl.broadcast_to(tmp7, [XBLOCK])
tmp13 = tl.load(in_ptr2 + (x0), None)
tmp24 = tl.load(in_ptr3 + (0))
tmp25 = tl.broadcast_to(tmp24, [XBLOCK])
tmp1 = 1e-12
tmp2 = tmp0 + tmp1
tmp3 = tmp2 * tmp2
tmp4 = 1.0
tmp5 = tmp3 * tmp4
tmp6 = libdevice.sqrt(tmp5)
tmp9 = 0.01
tmp10 = tmp8 * tmp9
tmp11 = tmp10 * tmp10
tmp12 = tmp6 + tmp11
tmp14 = tmp13 / tmp12
tmp15 = tl.full([1], 0, tl.int32)
tmp16 = tmp15 < tmp14
tmp17 = tmp16.to(tl.int8)
tmp18 = tmp14 < tmp15
tmp19 = tmp18.to(tl.int8)
tmp20 = tmp17 - tmp19
tmp21 = tmp20.to(tmp14.dtype)
tmp22 = tl_math.abs(tmp14)
tmp23 = tmp22 + tmp1
tmp26 = 0.1
tmp27 = tmp25 * tmp26
tmp28 = tmp27 * tmp27
tmp29 = libdevice.pow(tmp23, tmp28)
tmp30 = tmp21 * tmp29
tl.store(out_ptr0 + (x0), tmp12, None)
tl.store(out_ptr1 + (x0), tmp30, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 1, 1, 1), (1, 1, 1, 1))
assert_size_stride(primals_2, (1, ), (1, ))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
assert_size_stride(primals_4, (1, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 1, 1, 1), (1, 1, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [w_norm, truediv, q_sqr, add, add_1, w_normed], Original ATen: [aten.linalg_vector_norm, aten.div, aten.pow, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_linalg_vector_norm_pow_0.run(primals_1, primals_2, buf0, 1, grid=grid(1), stream=stream0)
# Topologically Sorted Source Nodes: [y_denorm], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(primals_3, buf0, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 1, 64, 64), (4096, 4096, 64, 1))
buf2 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1), torch.float32)
buf3 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1), torch.float32)
# Topologically Sorted Source Nodes: [truediv, q_sqr, add_2, pow_2, avg_pool2d, x_norm_squared, sqrt, add_3, y, sign, abs_1, y_1, y_2, mul], Original ATen: [aten.div, aten.pow, aten.add, aten.avg_pool2d, aten.sum, aten.sqrt, aten.sign, aten.abs, aten.mul]
triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1.run(primals_3, primals_2, buf1, primals_4, buf2, buf3, 16384, grid=grid(16384), stream=stream0)
return (buf3, primals_1, primals_2, primals_3, primals_4, buf0, buf1, buf2, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 1, 1, 1), (1, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 1, 64, 64), (4096, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class SharpenedCosineSimilarityAnnotated(nn.Module):
def __init__(self, in_channels=1, out_channels=1, kernel_size=1, stride
=1, padding=0, eps=1e-12):
super(SharpenedCosineSimilarityAnnotated, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.eps = eps
self.padding = int(padding)
w = torch.empty(out_channels, in_channels, kernel_size, kernel_size)
self.register_parameter('w', nn.Parameter(w))
nn.init.xavier_uniform_(self.w)
self.p_scale = 10
p_init = 2 ** 0.5 * self.p_scale
self.register_parameter('p', nn.Parameter(torch.empty(out_channels)))
nn.init.constant_(self.p, p_init)
self.q_scale = 100
self.register_parameter('q', nn.Parameter(torch.empty(1)))
nn.init.constant_(self.q, 10)
def forward(self, x):
w_norm = torch.linalg.vector_norm(self.w, dim=(1, 2, 3), keepdim=True)
q_sqr = (self.q / self.q_scale) ** 2
w_normed = self.w / (w_norm + self.eps + q_sqr)
x_norm_squared = F.avg_pool2d((x + self.eps) ** 2, kernel_size=self
.kernel_size, stride=self.stride, padding=self.padding,
divisor_override=1).sum(dim=1, keepdim=True)
y_denorm = F.conv2d(x, w_normed, bias=None, stride=self.stride,
padding=self.padding)
y = y_denorm / (x_norm_squared.sqrt() + q_sqr)
sign = torch.sign(y)
y = torch.abs(y) + self.eps
p_sqr = (self.p / self.p_scale) ** 2
y = y.pow(p_sqr.reshape(1, -1, 1, 1))
return sign * y
def get_inputs():
return [torch.rand([4, 1, 64, 64])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_div_linalg_vector_norm_pow_0(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
tmp0 = tl.load(in_ptr0 + 0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK])
tmp6 = tl.load(in_ptr1 + 0)
tmp7 = tl.broadcast_to(tmp6, [XBLOCK])
tmp2 = tmp1 * tmp1
tmp3 = libdevice.sqrt(tmp2)
tmp4 = 1e-12
tmp5 = tmp3 + tmp4
tmp8 = 0.01
tmp9 = tmp7 * tmp8
tmp10 = tmp9 * tmp9
tmp11 = tmp5 + tmp10
tmp12 = tmp1 / tmp11
tl.store(out_ptr0 + tl.full([XBLOCK], 0, tl.int32), tmp12, None)
@triton.jit
def triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1(in_ptr0,
in_ptr1, in_ptr2, in_ptr3, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr
):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, None)
tmp7 = tl.load(in_ptr1 + 0)
tmp8 = tl.broadcast_to(tmp7, [XBLOCK])
tmp13 = tl.load(in_ptr2 + x0, None)
tmp24 = tl.load(in_ptr3 + 0)
tmp25 = tl.broadcast_to(tmp24, [XBLOCK])
tmp1 = 1e-12
tmp2 = tmp0 + tmp1
tmp3 = tmp2 * tmp2
tmp4 = 1.0
tmp5 = tmp3 * tmp4
tmp6 = libdevice.sqrt(tmp5)
tmp9 = 0.01
tmp10 = tmp8 * tmp9
tmp11 = tmp10 * tmp10
tmp12 = tmp6 + tmp11
tmp14 = tmp13 / tmp12
tmp15 = tl.full([1], 0, tl.int32)
tmp16 = tmp15 < tmp14
tmp17 = tmp16.to(tl.int8)
tmp18 = tmp14 < tmp15
tmp19 = tmp18.to(tl.int8)
tmp20 = tmp17 - tmp19
tmp21 = tmp20.to(tmp14.dtype)
tmp22 = tl_math.abs(tmp14)
tmp23 = tmp22 + tmp1
tmp26 = 0.1
tmp27 = tmp25 * tmp26
tmp28 = tmp27 * tmp27
tmp29 = libdevice.pow(tmp23, tmp28)
tmp30 = tmp21 * tmp29
tl.store(out_ptr0 + x0, tmp12, None)
tl.store(out_ptr1 + x0, tmp30, None)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 1, 1, 1), (1, 1, 1, 1))
assert_size_stride(primals_2, (1,), (1,))
assert_size_stride(primals_3, (4, 1, 64, 64), (4096, 4096, 64, 1))
assert_size_stride(primals_4, (1,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 1, 1, 1), (1, 1, 1, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_linalg_vector_norm_pow_0[grid(1)](primals_1,
primals_2, buf0, 1, XBLOCK=1, num_warps=1, num_stages=1)
buf1 = extern_kernels.convolution(primals_3, buf0, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 1, 64, 64), (4096, 4096, 64, 1))
buf2 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1),
torch.float32)
buf3 = empty_strided_cuda((4, 1, 64, 64), (4096, 4096, 64, 1),
torch.float32)
triton_poi_fused_abs_add_avg_pool2d_div_mul_pow_sign_sqrt_sum_1[grid
(16384)](primals_3, primals_2, buf1, primals_4, buf2, buf3,
16384, XBLOCK=256, num_warps=4, num_stages=1)
return buf3, primals_1, primals_2, primals_3, primals_4, buf0, buf1, buf2
class SharpenedCosineSimilarityAnnotatedNew(nn.Module):
def __init__(self, in_channels=1, out_channels=1, kernel_size=1, stride
=1, padding=0, eps=1e-12):
super(SharpenedCosineSimilarityAnnotatedNew, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.eps = eps
self.padding = int(padding)
w = torch.empty(out_channels, in_channels, kernel_size, kernel_size)
self.register_parameter('w', nn.Parameter(w))
nn.init.xavier_uniform_(self.w)
self.p_scale = 10
p_init = 2 ** 0.5 * self.p_scale
self.register_parameter('p', nn.Parameter(torch.empty(out_channels)))
nn.init.constant_(self.p, p_init)
self.q_scale = 100
self.register_parameter('q', nn.Parameter(torch.empty(1)))
nn.init.constant_(self.q, 10)
def forward(self, input_0):
primals_1 = self.w
primals_2 = self.p
primals_4 = self.q
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
enzokro/sharpened_cosine_similarity_torch
|
SharpenedCosineSimilarityAnnotated
| false | 12,353 |
[
"MIT"
] | 0 |
150c84f5cf81721baf097abdc0d4ac772fb39fc4
|
https://github.com/enzokro/sharpened_cosine_similarity_torch/tree/150c84f5cf81721baf097abdc0d4ac772fb39fc4
|
Divide
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/i3/ci3wf6sq5envi5qqy36dg5xyilb5fmbazlijbvgfe3ony6kt56ir.py
# Topologically Sorted Source Nodes: [div], Original ATen: [aten.div]
# Source node to ATen node mapping:
# div => div
# Graph fragment:
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%arg1_1, %arg0_1), kwargs = {})
triton_poi_fused_div_0 = async_compile.triton('triton_poi_fused_div_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask)
tmp2 = tmp0 / tmp1
tl.store(out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [div], Original ATen: [aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_div_0.run(arg1_1, arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
del arg1_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn
import torch.utils.data
import torch.utils.tensorboard._pytorch_graph
import torch.onnx.symbolic_caffe2
class Divide(torch.nn.Module):
""" Divide module for a functional divide"""
def forward(self, x, y):
"""
Forward-pass routine for divide op
"""
return torch.div(x, y)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn
import torch.utils.data
import torch.utils.tensorboard._pytorch_graph
import torch.onnx.symbolic_caffe2
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_div_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask)
tmp2 = tmp0 / tmp1
tl.store(out_ptr0 + x0, tmp2, xmask)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_div_0[grid(256)](arg1_1, arg0_1, buf0, 256, XBLOCK
=256, num_warps=4, num_stages=1)
del arg0_1
del arg1_1
return buf0,
class DivideNew(torch.nn.Module):
""" Divide module for a functional divide"""
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
arjunsuresh/aimet
|
Divide
| false | 12,354 |
[
"BSD-3-Clause"
] | 0 |
f6e09cb07a91eed3a5e6b8e19e6b065303af5a39
|
https://github.com/arjunsuresh/aimet/tree/f6e09cb07a91eed3a5e6b8e19e6b065303af5a39
|
EqualLinear
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/fl/cflw6zjzdk2wqtau7m6nsei5vavjfijzxhb37zaa3xp4yxpw5yb2.py
# Topologically Sorted Source Nodes: [mul], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# mul => mul
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_1, 1), kwargs = {})
triton_poi_fused_mul_0 = async_compile.triton('triton_poi_fused_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2o/c2oqkq7zaubqmw7vuixxlseb2ff5jzqqbyczicxlmsahuxwdpdyp.py
# Topologically Sorted Source Nodes: [mul_1], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# mul_1 => mul_1
# Graph fragment:
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, 1), kwargs = {})
triton_poi_fused_mul_1 = async_compile.triton('triton_poi_fused_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul], Original ATen: [aten.mul]
stream0 = get_raw_stream(0)
triton_poi_fused_mul_0.run(primals_1, buf0, 16, grid=grid(16), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((4, ), (1, ), torch.float32)
# Topologically Sorted Source Nodes: [mul_1], Original ATen: [aten.mul]
triton_poi_fused_mul_1.run(primals_2, buf1, 4, grid=grid(4), stream=stream0)
del primals_2
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul_1, linear], Original ATen: [aten.mul, aten.addmm]
extern_kernels.addmm(buf1, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(buf0, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf2)
del buf0
del buf1
return (reinterpret_tensor(buf2, (4, 4, 4, 4), (64, 16, 4, 1), 0), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn.functional as F
from torch import nn
class EqualLinear(nn.Module):
def __init__(self, in_dim, out_dim, lr_mul=1, bias=True):
super().__init__()
self.weight = nn.Parameter(torch.randn(out_dim, in_dim))
if bias:
self.bias = nn.Parameter(torch.zeros(out_dim))
self.lr_mul = lr_mul
def forward(self, input):
return F.linear(input, self.weight * self.lr_mul, bias=self.bias *
self.lr_mul)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_dim': 4, 'out_dim': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x0, tmp2, xmask)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x0, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_mul_0[grid(16)](primals_1, buf0, 16, XBLOCK=16,
num_warps=1, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((4,), (1,), torch.float32)
triton_poi_fused_mul_1[grid(4)](primals_2, buf1, 4, XBLOCK=4,
num_warps=1, num_stages=1)
del primals_2
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(buf1, reinterpret_tensor(primals_3, (64, 4), (
4, 1), 0), reinterpret_tensor(buf0, (4, 4), (1, 4), 0), alpha=1,
beta=1, out=buf2)
del buf0
del buf1
return reinterpret_tensor(buf2, (4, 4, 4, 4), (64, 16, 4, 1), 0
), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0)
class EqualLinearNew(nn.Module):
def __init__(self, in_dim, out_dim, lr_mul=1, bias=True):
super().__init__()
self.weight = nn.Parameter(torch.randn(out_dim, in_dim))
if bias:
self.bias = nn.Parameter(torch.zeros(out_dim))
self.lr_mul = lr_mul
def forward(self, input_0):
primals_1 = self.weight
primals_2 = self.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
ericguizzo/stylegan2-pytorch
|
EqualLinear
| false | 12,355 |
[
"MIT"
] | 0 |
d6e5cf4e30247e12d330537676f9ba63867cfaa0
|
https://github.com/ericguizzo/stylegan2-pytorch/tree/d6e5cf4e30247e12d330537676f9ba63867cfaa0
|
ScaleNorm
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/5i/c5iachkxsftg5m5zniolyz4aeojcm42muwsrb2a34ygecaxk5f47.py
# Topologically Sorted Source Nodes: [norm, clamp, norm_1], Original ATen: [aten.linalg_vector_norm, aten.clamp, aten.div]
# Source node to ATen node mapping:
# clamp => clamp_min
# norm => pow_1, pow_2, sum_1
# norm_1 => div
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_2, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [-1], True), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_1, 0.5), kwargs = {})
# %clamp_min : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%pow_2, 1e-05), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_1, %clamp_min), kwargs = {})
triton_poi_fused_clamp_div_linalg_vector_norm_0 = async_compile.triton('triton_poi_fused_clamp_div_linalg_vector_norm_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clamp_div_linalg_vector_norm_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clamp_div_linalg_vector_norm_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (0))
tmp1 = tl.broadcast_to(tmp0, [XBLOCK])
tmp2 = tl.load(in_ptr1 + (4*x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr1 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp3 = tmp2 * tmp2
tmp5 = tmp4 * tmp4
tmp6 = tmp3 + tmp5
tmp8 = tmp7 * tmp7
tmp9 = tmp6 + tmp8
tmp11 = tmp10 * tmp10
tmp12 = tmp9 + tmp11
tmp13 = libdevice.sqrt(tmp12)
tmp14 = 1e-05
tmp15 = triton_helpers.maximum(tmp13, tmp14)
tmp16 = tmp1 / tmp15
tl.store(out_ptr0 + (x0), tmp16, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/pq/cpqdtx3g3pyjzwakdnj54l47sm3lyvtzkr3iagg7am3bkbp6ommo.py
# Topologically Sorted Source Nodes: [norm, clamp, norm_1, mul], Original ATen: [aten.linalg_vector_norm, aten.clamp, aten.div, aten.mul]
# Source node to ATen node mapping:
# clamp => clamp_min
# mul => mul
# norm => pow_1, pow_2, sum_1
# norm_1 => div
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_2, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [-1], True), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_1, 0.5), kwargs = {})
# %clamp_min : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%pow_2, 1e-05), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_1, %clamp_min), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %div), kwargs = {})
triton_poi_fused_clamp_div_linalg_vector_norm_mul_1 = async_compile.triton('triton_poi_fused_clamp_div_linalg_vector_norm_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clamp_div_linalg_vector_norm_mul_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clamp_div_linalg_vector_norm_mul_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x2), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (), ())
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 64), torch.float32)
# Topologically Sorted Source Nodes: [norm, clamp, norm_1], Original ATen: [aten.linalg_vector_norm, aten.clamp, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_clamp_div_linalg_vector_norm_0.run(primals_1, primals_2, buf0, 64, grid=grid(64), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [norm, clamp, norm_1, mul], Original ATen: [aten.linalg_vector_norm, aten.clamp, aten.div, aten.mul]
triton_poi_fused_clamp_div_linalg_vector_norm_mul_1.run(primals_2, buf0, buf1, 256, grid=grid(256), stream=stream0)
del buf0
return (buf1, primals_2, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((), (), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.nn as nn
class ScaleNorm(nn.Module):
"""ScaleNorm"""
"""All g’s in SCALE NORM are initialized to sqrt(d)"""
def __init__(self, scale, eps=1e-05):
super(ScaleNorm, self).__init__()
self.scale = nn.Parameter(torch.tensor(math.sqrt(scale)))
self.eps = eps
def forward(self, x):
norm = self.scale / torch.norm(x, dim=-1, keepdim=True).clamp(min=
self.eps)
return x * norm
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'scale': 1.0}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice
import math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_clamp_div_linalg_vector_norm_0(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK])
tmp2 = tl.load(in_ptr1 + 4 * x0, xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (2 + 4 * x0), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr1 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp3 = tmp2 * tmp2
tmp5 = tmp4 * tmp4
tmp6 = tmp3 + tmp5
tmp8 = tmp7 * tmp7
tmp9 = tmp6 + tmp8
tmp11 = tmp10 * tmp10
tmp12 = tmp9 + tmp11
tmp13 = libdevice.sqrt(tmp12)
tmp14 = 1e-05
tmp15 = triton_helpers.maximum(tmp13, tmp14)
tmp16 = tmp1 / tmp15
tl.store(out_ptr0 + x0, tmp16, xmask)
@triton.jit
def triton_poi_fused_clamp_div_linalg_vector_norm_mul_1(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x2, tmp2, xmask)
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (), ())
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 64), torch.float32)
get_raw_stream(0)
triton_poi_fused_clamp_div_linalg_vector_norm_0[grid(64)](primals_1,
primals_2, buf0, 64, XBLOCK=64, num_warps=1, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_clamp_div_linalg_vector_norm_mul_1[grid(256)](
primals_2, buf0, buf1, 256, XBLOCK=128, num_warps=4, num_stages=1)
del buf0
return buf1, primals_2
class ScaleNormNew(nn.Module):
"""ScaleNorm"""
"""All g’s in SCALE NORM are initialized to sqrt(d)"""
def __init__(self, scale, eps=1e-05):
super(ScaleNormNew, self).__init__()
self.scale = nn.Parameter(torch.tensor(math.sqrt(scale)))
self.eps = eps
def forward(self, input_0):
primals_1 = self.scale
primals_2 = input_0
output = call([primals_1, primals_2])
return output[0]
|
eweiner/MAT_Extension
|
ScaleNorm
| false | 12,356 |
[
"MIT"
] | 0 |
505884a67f97bf54e1198077d15a48531fcac7a5
|
https://github.com/eweiner/MAT_Extension/tree/505884a67f97bf54e1198077d15a48531fcac7a5
|
Generator
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/pj/cpjtaps4dusmzrrvkgal3gobvvp6pxqqx5zvwpjyzgdjffmdihfr.py
# Topologically Sorted Source Nodes: [out_masked, out_sum, mask_sum, out_avg_pooling], Original ATen: [aten.mul, aten.sum, aten.div]
# Source node to ATen node mapping:
# mask_sum => sum_2
# out_avg_pooling => div
# out_masked => mul
# out_sum => sum_1
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul, [1]), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%unsqueeze, [1]), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sum_1, %sum_2), kwargs = {})
triton_poi_fused_div_mul_sum_0 = async_compile.triton('triton_poi_fused_div_mul_sum_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_mul_sum_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_mul_sum_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex % 64
x1 = (xindex // 4) % 16
x2 = (xindex // 64)
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x3), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x1 + (64*x2)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (64 + x3), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (16 + x1 + (64*x2)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (128 + x3), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (32 + x1 + (64*x2)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (192 + x3), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr1 + (48 + x1 + (64*x2)), xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tmp15 = tmp1 + tmp4
tmp16 = tmp15 + tmp8
tmp17 = tmp16 + tmp12
tmp18 = tmp14 / tmp17
tl.store(out_ptr0 + (x4), tmp18, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (1, 4), (4, 1))
assert_size_stride(primals_4, (1, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [out_masked, out_sum, mask_sum, out_avg_pooling], Original ATen: [aten.mul, aten.sum, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_div_mul_sum_0.run(primals_2, primals_1, buf0, 256, grid=grid(256), stream=stream0)
del primals_1
del primals_2
buf2 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [projected], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_4, reinterpret_tensor(buf0, (64, 4), (4, 1), 0), reinterpret_tensor(primals_3, (4, 1), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_3
del primals_4
return (reinterpret_tensor(buf2, (4, 4, 4, 1), (16, 4, 1, 1), 0), reinterpret_tensor(buf0, (64, 4), (4, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((1, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.nn as nn
class LayerNorm(nn.Module):
"""Construct a layernorm module (See citation for details)."""
def __init__(self, features, eps=1e-06):
super(LayerNorm, self).__init__()
self.a_2 = nn.Parameter(torch.ones(features))
self.b_2 = nn.Parameter(torch.zeros(features))
self.eps = eps
def forward(self, x):
mean = x.mean(-1, keepdim=True)
std = x.std(-1, keepdim=True)
return self.a_2 * (x - mean) / (std + self.eps) + self.b_2
class ScaleNorm(nn.Module):
"""ScaleNorm"""
"""All g’s in SCALE NORM are initialized to sqrt(d)"""
def __init__(self, scale, eps=1e-05):
super(ScaleNorm, self).__init__()
self.scale = nn.Parameter(torch.tensor(math.sqrt(scale)))
self.eps = eps
def forward(self, x):
norm = self.scale / torch.norm(x, dim=-1, keepdim=True).clamp(min=
self.eps)
return x * norm
class Generator(nn.Module):
"""Define standard linear + softmax generation step."""
def __init__(self, d_model, aggregation_type='mean', n_output=1,
n_layers=1, leaky_relu_slope=0.01, dropout=0.0, scale_norm=False):
super(Generator, self).__init__()
if n_layers == 1:
self.proj = nn.Linear(d_model, n_output)
else:
self.proj = []
for i in range(n_layers - 1):
self.proj.append(nn.Linear(d_model, d_model))
self.proj.append(nn.LeakyReLU(leaky_relu_slope))
self.proj.append(ScaleNorm(d_model) if scale_norm else
LayerNorm(d_model))
self.proj.append(nn.Dropout(dropout))
self.proj.append(nn.Linear(d_model, n_output))
self.proj = torch.nn.Sequential(*self.proj)
self.aggregation_type = aggregation_type
def forward(self, x, mask):
mask = mask.unsqueeze(-1).float()
out_masked = x * mask
if self.aggregation_type == 'mean':
out_sum = out_masked.sum(dim=1)
mask_sum = mask.sum(dim=1)
out_avg_pooling = out_sum / mask_sum
elif self.aggregation_type == 'sum':
out_sum = out_masked.sum(dim=1)
out_avg_pooling = out_sum
elif self.aggregation_type == 'dummy_node':
out_avg_pooling = out_masked[:, 0]
projected = self.proj(out_avg_pooling)
return projected
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'d_model': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_div_mul_sum_0(in_ptr0, in_ptr1, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex % 64
x1 = xindex // 4 % 16
x2 = xindex // 64
x4 = xindex
tmp0 = tl.load(in_ptr0 + x3, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x1 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp3 = tl.load(in_ptr0 + (64 + x3), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (16 + x1 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp7 = tl.load(in_ptr0 + (128 + x3), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (32 + x1 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp11 = tl.load(in_ptr0 + (192 + x3), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr1 + (48 + x1 + 64 * x2), xmask, eviction_policy=
'evict_last')
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tmp15 = tmp1 + tmp4
tmp16 = tmp15 + tmp8
tmp17 = tmp16 + tmp12
tmp18 = tmp14 / tmp17
tl.store(out_ptr0 + x4, tmp18, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (1, 4), (4, 1))
assert_size_stride(primals_4, (1,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_div_mul_sum_0[grid(256)](primals_2, primals_1,
buf0, 256, XBLOCK=128, num_warps=4, num_stages=1)
del primals_1
del primals_2
buf2 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
extern_kernels.addmm(primals_4, reinterpret_tensor(buf0, (64, 4), (
4, 1), 0), reinterpret_tensor(primals_3, (4, 1), (1, 4), 0),
alpha=1, beta=1, out=buf2)
del primals_3
del primals_4
return reinterpret_tensor(buf2, (4, 4, 4, 1), (16, 4, 1, 1), 0
), reinterpret_tensor(buf0, (64, 4), (4, 1), 0)
class LayerNorm(nn.Module):
"""Construct a layernorm module (See citation for details)."""
def __init__(self, features, eps=1e-06):
super(LayerNorm, self).__init__()
self.a_2 = nn.Parameter(torch.ones(features))
self.b_2 = nn.Parameter(torch.zeros(features))
self.eps = eps
def forward(self, x):
mean = x.mean(-1, keepdim=True)
std = x.std(-1, keepdim=True)
return self.a_2 * (x - mean) / (std + self.eps) + self.b_2
class ScaleNorm(nn.Module):
"""ScaleNorm"""
"""All g’s in SCALE NORM are initialized to sqrt(d)"""
def __init__(self, scale, eps=1e-05):
super(ScaleNorm, self).__init__()
self.scale = nn.Parameter(torch.tensor(math.sqrt(scale)))
self.eps = eps
def forward(self, x):
norm = self.scale / torch.norm(x, dim=-1, keepdim=True).clamp(min=
self.eps)
return x * norm
class GeneratorNew(nn.Module):
"""Define standard linear + softmax generation step."""
def __init__(self, d_model, aggregation_type='mean', n_output=1,
n_layers=1, leaky_relu_slope=0.01, dropout=0.0, scale_norm=False):
super(GeneratorNew, self).__init__()
if n_layers == 1:
self.proj = nn.Linear(d_model, n_output)
else:
self.proj = []
for i in range(n_layers - 1):
self.proj.append(nn.Linear(d_model, d_model))
self.proj.append(nn.LeakyReLU(leaky_relu_slope))
self.proj.append(ScaleNorm(d_model) if scale_norm else
LayerNorm(d_model))
self.proj.append(nn.Dropout(dropout))
self.proj.append(nn.Linear(d_model, n_output))
self.proj = torch.nn.Sequential(*self.proj)
self.aggregation_type = aggregation_type
def forward(self, input_0, input_1):
primals_3 = self.proj.weight
primals_4 = self.proj.bias
primals_1 = input_0
primals_2 = input_1
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
eweiner/MAT_Extension
|
Generator
| false | 12,357 |
[
"MIT"
] | 0 |
505884a67f97bf54e1198077d15a48531fcac7a5
|
https://github.com/eweiner/MAT_Extension/tree/505884a67f97bf54e1198077d15a48531fcac7a5
|
SAB
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/p6/cp62j5tv2tk4ynnj52fvxnzwu3qhxoiovlhltiksafkjm6cs5wpz.py
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# Q_ => cat
# Graph fragment:
# %cat : [num_users=3] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem, %getitem_1, %getitem_2, %getitem_3],), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4)
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x0) + (16*x1)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + (4*x0) + (16*((-4) + x1))), tmp9 & xmask, eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + (4*x0) + (16*((-8) + x1))), tmp14 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tmp17 = tl.full([1], 16, tl.int64)
tmp18 = tmp0 < tmp17
tmp19 = tl.load(in_ptr0 + (3 + (4*x0) + (16*((-12) + x1))), tmp16 & xmask, eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + (x2), tmp22, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/eh/ceheq5ns3kg3p6tebb47gdy475c5v4keklf245jl4fgbpvugznm5.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => exp
# Graph fragment:
# %mul_tensor : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%bmm, 1), kwargs = {})
# %amax_default : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%mul_tensor, [2], True), kwargs = {})
# %sub_tensor : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_tensor, %amax_default), kwargs = {})
# %div_tensor : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub_tensor, 2.0), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%div_tensor,), kwargs = {})
triton_poi_fused__softmax_1 = async_compile.triton('triton_poi_fused__softmax_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp3 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + (x2), tmp17, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/3f/c3fx6bzkalkw7u7askqdnz4rzlcoyqiec4r434sjc5x3axxgkrmr.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => div_1, sum_1
# Graph fragment:
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [2], True), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_2 = async_compile.triton('triton_poi_fused__softmax_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2n/c2nnsl6rh3uwvjzhffvx73mazqy356d27sevembkk4zhua32vcxd.py
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# O => cat_3
# Graph fragment:
# %cat_3 : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem_12, %getitem_13, %getitem_14, %getitem_15], 2), kwargs = {})
triton_poi_fused_cat_3 = async_compile.triton('triton_poi_fused_cat_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = (xindex // 4)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tmp29 = tl.full([1], 4, tl.int64)
tmp30 = tmp0 < tmp29
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + (x2), tmp38, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ws/cwsvc5z65iswwv5a6qo3vtrk3rtb7xslw5l36aatdxiyao2vymh7.py
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
# Source node to ATen node mapping:
# O_1 => add_1
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_7,), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%cat_3, %relu), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_add_relu_threshold_backward_4 = async_compile.triton('triton_poi_fused_add_relu_threshold_backward_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*i1', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_relu_threshold_backward_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_4(in_ptr0, in_ptr1, in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp2 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + (x2), tmp6, xmask)
tl.store(out_ptr1 + (x2), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4, ), (1, ))
assert_size_stride(primals_8, (4, 4), (4, 1))
assert_size_stride(primals_9, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [Q], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [K], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_5, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf1)
del primals_4
del primals_5
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [V], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_7, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_6
del primals_7
buf3 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(buf0, buf3, 64, grid=grid(64), stream=stream0)
buf4 = reinterpret_tensor(buf0, (16, 4, 1), (4, 1, 64), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [V_], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf2, buf4, 64, grid=grid(64), stream=stream0)
buf5 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [K_], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf1, buf5, 64, grid=grid(64), stream=stream0)
buf6 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [bmm], Original ATen: [aten.bmm]
extern_kernels.bmm(buf3, reinterpret_tensor(buf5, (16, 1, 4), (4, 0, 1), 0), out=buf6)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_1.run(buf6, buf7, 256, grid=grid(256), stream=stream0)
buf8 = buf6; del buf6 # reuse
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_2.run(buf7, buf8, 256, grid=grid(256), stream=stream0)
del buf7
buf9 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [bmm_1], Original ATen: [aten.bmm]
extern_kernels.bmm(buf8, buf4, out=buf9)
buf10 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
triton_poi_fused_cat_3.run(buf3, buf9, buf10, 64, grid=grid(64), stream=stream0)
buf11 = reinterpret_tensor(buf9, (16, 4), (4, 1), 0); del buf9 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf10, (16, 4), (4, 1), 0), reinterpret_tensor(primals_8, (4, 4), (1, 4), 0), out=buf11)
buf12 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf13 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
triton_poi_fused_add_relu_threshold_backward_4.run(buf10, buf11, primals_9, buf12, buf13, 64, grid=grid(64), stream=stream0)
del buf11
del primals_9
return (buf12, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), buf8, reinterpret_tensor(buf10, (16, 4), (4, 1), 0), buf13, primals_8, reinterpret_tensor(buf4, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0), buf5, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
class MAB(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MAB, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, Q, K):
Q = self.fc_q(Q)
K, V = self.fc_k(K), self.fc_v(K)
dim_split = self.dim_V // self.num_heads
Q_ = torch.cat(Q.split(dim_split, 2), 0)
K_ = torch.cat(K.split(dim_split, 2), 0)
V_ = torch.cat(V.split(dim_split, 2), 0)
A = torch.softmax(Q_.bmm(K_.transpose(1, 2)) / math.sqrt(self.dim_V), 2
)
O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
O = O + F.relu(self.fc_o(O))
O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
return O
class SAB(nn.Module):
def __init__(self, dim_in, dim_out, num_heads, ln=False):
super(SAB, self).__init__()
self.mab = MAB(dim_in, dim_in, dim_out, num_heads, ln=ln)
def forward(self, X):
return self.mab(X, X)
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'dim_in': 4, 'dim_out': 4, 'num_heads': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import math
import torch.nn as nn
import torch.nn.functional as F
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x0 + 16 * x1), tmp4 & xmask,
eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + 4 * x0 + 16 * (-4 + x1)), tmp9 & xmask,
eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + 4 * x0 + 16 * (-8 + x1)), tmp14 & xmask,
eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tl.full([1], 16, tl.int64)
tmp19 = tl.load(in_ptr0 + (3 + 4 * x0 + 16 * (-12 + x1)), tmp16 & xmask,
eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + x2, tmp22, xmask)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp3 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + x2, tmp17, xmask)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_cat_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = xindex // 4
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp6 = tl.load(in_ptr1 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tl.full([1], 4, tl.int64)
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + x2, tmp38, xmask)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_4(in_ptr0, in_ptr1,
in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp2 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + x2, tmp6, xmask)
tl.store(out_ptr1 + x2, tmp8, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4,), (1,))
assert_size_stride(primals_8, (4, 4), (4, 1))
assert_size_stride(primals_9, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (16,
4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_5, reinterpret_tensor(primals_3, (16,
4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf1)
del primals_4
del primals_5
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_7, reinterpret_tensor(primals_3, (16,
4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf2)
del primals_6
del primals_7
buf3 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(64)](buf0, buf3, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf4 = reinterpret_tensor(buf0, (16, 4, 1), (4, 1, 64), 0)
del buf0
triton_poi_fused_cat_0[grid(64)](buf2, buf4, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf5 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0)
del buf2
triton_poi_fused_cat_0[grid(64)](buf1, buf5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf6 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(buf3, reinterpret_tensor(buf5, (16, 1, 4), (4, 0,
1), 0), out=buf6)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_1[grid(256)](buf6, buf7, 256, XBLOCK=256,
num_warps=4, num_stages=1)
buf8 = buf6
del buf6
triton_poi_fused__softmax_2[grid(256)](buf7, buf8, 256, XBLOCK=256,
num_warps=4, num_stages=1)
del buf7
buf9 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 1), 0)
del buf1
extern_kernels.bmm(buf8, buf4, out=buf9)
buf10 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_cat_3[grid(64)](buf3, buf9, buf10, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf11 = reinterpret_tensor(buf9, (16, 4), (4, 1), 0)
del buf9
extern_kernels.mm(reinterpret_tensor(buf10, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_8, (4, 4), (1, 4), 0), out=buf11)
buf12 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf13 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
triton_poi_fused_add_relu_threshold_backward_4[grid(64)](buf10,
buf11, primals_9, buf12, buf13, 64, XBLOCK=64, num_warps=1,
num_stages=1)
del buf11
del primals_9
return buf12, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0
), buf8, reinterpret_tensor(buf10, (16, 4), (4, 1), 0
), buf13, primals_8, reinterpret_tensor(buf4, (16, 1, 4), (4, 1, 1), 0
), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0), buf5
class MAB(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MAB, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, Q, K):
Q = self.fc_q(Q)
K, V = self.fc_k(K), self.fc_v(K)
dim_split = self.dim_V // self.num_heads
Q_ = torch.cat(Q.split(dim_split, 2), 0)
K_ = torch.cat(K.split(dim_split, 2), 0)
V_ = torch.cat(V.split(dim_split, 2), 0)
A = torch.softmax(Q_.bmm(K_.transpose(1, 2)) / math.sqrt(self.dim_V), 2
)
O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
O = O + F.relu(self.fc_o(O))
O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
return O
class SABNew(nn.Module):
def __init__(self, dim_in, dim_out, num_heads, ln=False):
super(SABNew, self).__init__()
self.mab = MAB(dim_in, dim_in, dim_out, num_heads, ln=ln)
def forward(self, input_0):
primals_1 = self.mab.fc_q.weight
primals_2 = self.mab.fc_q.bias
primals_4 = self.mab.fc_k.weight
primals_5 = self.mab.fc_k.bias
primals_6 = self.mab.fc_v.weight
primals_7 = self.mab.fc_v.bias
primals_8 = self.mab.fc_o.weight
primals_9 = self.mab.fc_o.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9])
return output[0]
|
ernoult/set_transformer
|
SAB
| false | 12,358 |
[
"MIT"
] | 0 |
4b380106e1f43b7eb6315624c57d4d1d38737b78
|
https://github.com/ernoult/set_transformer/tree/4b380106e1f43b7eb6315624c57d4d1d38737b78
|
MAB
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/p6/cp62j5tv2tk4ynnj52fvxnzwu3qhxoiovlhltiksafkjm6cs5wpz.py
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# Q_ => cat
# Graph fragment:
# %cat : [num_users=3] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem, %getitem_1, %getitem_2, %getitem_3],), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4)
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x0) + (16*x1)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + (4*x0) + (16*((-4) + x1))), tmp9 & xmask, eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + (4*x0) + (16*((-8) + x1))), tmp14 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tmp17 = tl.full([1], 16, tl.int64)
tmp18 = tmp0 < tmp17
tmp19 = tl.load(in_ptr0 + (3 + (4*x0) + (16*((-12) + x1))), tmp16 & xmask, eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + (x2), tmp22, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/eh/ceheq5ns3kg3p6tebb47gdy475c5v4keklf245jl4fgbpvugznm5.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => exp
# Graph fragment:
# %mul_tensor : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%bmm, 1), kwargs = {})
# %amax_default : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%mul_tensor, [2], True), kwargs = {})
# %sub_tensor : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_tensor, %amax_default), kwargs = {})
# %div_tensor : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub_tensor, 2.0), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%div_tensor,), kwargs = {})
triton_poi_fused__softmax_1 = async_compile.triton('triton_poi_fused__softmax_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp3 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + (x2), tmp17, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/3f/c3fx6bzkalkw7u7askqdnz4rzlcoyqiec4r434sjc5x3axxgkrmr.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => div_1, sum_1
# Graph fragment:
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [2], True), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_2 = async_compile.triton('triton_poi_fused__softmax_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2n/c2nnsl6rh3uwvjzhffvx73mazqy356d27sevembkk4zhua32vcxd.py
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# O => cat_3
# Graph fragment:
# %cat_3 : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem_12, %getitem_13, %getitem_14, %getitem_15], 2), kwargs = {})
triton_poi_fused_cat_3 = async_compile.triton('triton_poi_fused_cat_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = (xindex // 4)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tmp29 = tl.full([1], 4, tl.int64)
tmp30 = tmp0 < tmp29
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + (x2), tmp38, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ws/cwsvc5z65iswwv5a6qo3vtrk3rtb7xslw5l36aatdxiyao2vymh7.py
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
# Source node to ATen node mapping:
# O_1 => add_1
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_7,), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%cat_3, %relu), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_add_relu_threshold_backward_4 = async_compile.triton('triton_poi_fused_add_relu_threshold_backward_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*i1', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_relu_threshold_backward_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_4(in_ptr0, in_ptr1, in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp2 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + (x2), tmp6, xmask)
tl.store(out_ptr1 + (x2), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4, ), (1, ))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [Q], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [K], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_5, reinterpret_tensor(primals_6, (16, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf1)
del primals_4
del primals_5
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [V], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_6, (16, 4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_7
del primals_8
buf3 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(buf0, buf3, 64, grid=grid(64), stream=stream0)
buf4 = reinterpret_tensor(buf0, (16, 4, 1), (4, 1, 64), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [V_], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf2, buf4, 64, grid=grid(64), stream=stream0)
buf5 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [K_], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf1, buf5, 64, grid=grid(64), stream=stream0)
buf6 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [bmm], Original ATen: [aten.bmm]
extern_kernels.bmm(buf3, reinterpret_tensor(buf5, (16, 1, 4), (4, 0, 1), 0), out=buf6)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_1.run(buf6, buf7, 256, grid=grid(256), stream=stream0)
buf8 = buf6; del buf6 # reuse
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_2.run(buf7, buf8, 256, grid=grid(256), stream=stream0)
del buf7
buf9 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [bmm_1], Original ATen: [aten.bmm]
extern_kernels.bmm(buf8, buf4, out=buf9)
buf10 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
triton_poi_fused_cat_3.run(buf3, buf9, buf10, 64, grid=grid(64), stream=stream0)
buf11 = reinterpret_tensor(buf9, (16, 4), (4, 1), 0); del buf9 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf10, (16, 4), (4, 1), 0), reinterpret_tensor(primals_9, (4, 4), (1, 4), 0), out=buf11)
buf12 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf13 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
triton_poi_fused_add_relu_threshold_backward_4.run(buf10, buf11, primals_10, buf12, buf13, 64, grid=grid(64), stream=stream0)
del buf11
del primals_10
return (buf12, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_6, (16, 4), (4, 1), 0), buf8, reinterpret_tensor(buf10, (16, 4), (4, 1), 0), buf13, primals_9, reinterpret_tensor(buf4, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0), buf5, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
class MAB(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MAB, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, Q, K):
Q = self.fc_q(Q)
K, V = self.fc_k(K), self.fc_v(K)
dim_split = self.dim_V // self.num_heads
Q_ = torch.cat(Q.split(dim_split, 2), 0)
K_ = torch.cat(K.split(dim_split, 2), 0)
V_ = torch.cat(V.split(dim_split, 2), 0)
A = torch.softmax(Q_.bmm(K_.transpose(1, 2)) / math.sqrt(self.dim_V), 2
)
O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
O = O + F.relu(self.fc_o(O))
O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
return O
def get_inputs():
return [torch.rand([4, 4, 4]), torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'dim_Q': 4, 'dim_K': 4, 'dim_V': 4, 'num_heads': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x0 + 16 * x1), tmp4 & xmask,
eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + 4 * x0 + 16 * (-4 + x1)), tmp9 & xmask,
eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + 4 * x0 + 16 * (-8 + x1)), tmp14 & xmask,
eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tl.full([1], 16, tl.int64)
tmp19 = tl.load(in_ptr0 + (3 + 4 * x0 + 16 * (-12 + x1)), tmp16 & xmask,
eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + x2, tmp22, xmask)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp3 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + x2, tmp17, xmask)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_cat_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = xindex // 4
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp6 = tl.load(in_ptr1 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tl.full([1], 4, tl.int64)
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + x2, tmp38, xmask)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_4(in_ptr0, in_ptr1,
in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp2 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + x2, tmp6, xmask)
tl.store(out_ptr1 + x2, tmp8, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4,), (1,))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_2, reinterpret_tensor(primals_3, (16,
4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_5, reinterpret_tensor(primals_6, (16,
4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf1)
del primals_4
del primals_5
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_6, (16,
4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf2)
del primals_7
del primals_8
buf3 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(64)](buf0, buf3, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf4 = reinterpret_tensor(buf0, (16, 4, 1), (4, 1, 64), 0)
del buf0
triton_poi_fused_cat_0[grid(64)](buf2, buf4, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf5 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0)
del buf2
triton_poi_fused_cat_0[grid(64)](buf1, buf5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf6 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(buf3, reinterpret_tensor(buf5, (16, 1, 4), (4, 0,
1), 0), out=buf6)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_1[grid(256)](buf6, buf7, 256, XBLOCK=256,
num_warps=4, num_stages=1)
buf8 = buf6
del buf6
triton_poi_fused__softmax_2[grid(256)](buf7, buf8, 256, XBLOCK=256,
num_warps=4, num_stages=1)
del buf7
buf9 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 1), 0)
del buf1
extern_kernels.bmm(buf8, buf4, out=buf9)
buf10 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_cat_3[grid(64)](buf3, buf9, buf10, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf11 = reinterpret_tensor(buf9, (16, 4), (4, 1), 0)
del buf9
extern_kernels.mm(reinterpret_tensor(buf10, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_9, (4, 4), (1, 4), 0), out=buf11)
buf12 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf13 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
triton_poi_fused_add_relu_threshold_backward_4[grid(64)](buf10,
buf11, primals_10, buf12, buf13, 64, XBLOCK=64, num_warps=1,
num_stages=1)
del buf11
del primals_10
return buf12, reinterpret_tensor(primals_3, (16, 4), (4, 1), 0
), reinterpret_tensor(primals_6, (16, 4), (4, 1), 0
), buf8, reinterpret_tensor(buf10, (16, 4), (4, 1), 0
), buf13, primals_9, reinterpret_tensor(buf4, (16, 1, 4), (4, 1, 1), 0
), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0), buf5
class MABNew(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MABNew, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, input_0, input_1):
primals_1 = self.fc_q.weight
primals_2 = self.fc_q.bias
primals_4 = self.fc_k.weight
primals_5 = self.fc_k.bias
primals_7 = self.fc_v.weight
primals_8 = self.fc_v.bias
primals_9 = self.fc_o.weight
primals_10 = self.fc_o.bias
primals_3 = input_0
primals_6 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9, primals_10])
return output[0]
|
ernoult/set_transformer
|
MAB
| false | 12,359 |
[
"MIT"
] | 0 |
4b380106e1f43b7eb6315624c57d4d1d38737b78
|
https://github.com/ernoult/set_transformer/tree/4b380106e1f43b7eb6315624c57d4d1d38737b78
|
EdgeFeaturesLayer
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/3u/c3ub52l73zdv4klgqzgxmtzrzxvztuyczv2jksnvrjr7erq7guxd.py
# Topologically Sorted Source Nodes: [linear], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# linear => clone
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_0 = async_compile.triton('triton_poi_fused_clone_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64, 4], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_0(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 64
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = (yindex // 16)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (16*x2) + (64*y1)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + (4*y3)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/lx/clxenpwkl4qtcky22cudzrb6ruwgm2vjfzwtegj2siml77dc4lga.py
# Topologically Sorted Source Nodes: [relu], Original ATen: [aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%permute_2,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_relu_threshold_backward_1 = async_compile.triton('triton_poi_fused_relu_threshold_backward_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i1', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_threshold_backward_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_threshold_backward_1(in_out_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + (x0), xmask)
tmp1 = tl.full([1], 0, tl.int32)
tmp2 = triton_helpers.maximum(tmp1, tmp0)
tmp3 = 0.0
tmp4 = tmp2 <= tmp3
tl.store(in_out_ptr0 + (x0), tmp2, xmask)
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (1, 4), (4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear], Original ATen: [aten.clone]
stream0 = get_raw_stream(0)
triton_poi_fused_clone_0.run(primals_1, buf0, 64, 4, grid=grid(64, 4), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear], Original ATen: [aten.mm]
extern_kernels.mm(reinterpret_tensor(buf0, (64, 4), (4, 1), 0), reinterpret_tensor(primals_2, (4, 1), (1, 4), 0), out=buf1)
del primals_2
buf2 = reinterpret_tensor(buf1, (4, 1, 4, 4), (16, 1, 4, 1), 0); del buf1 # reuse
buf3 = empty_strided_cuda((4, 1, 4, 4), (16, 1, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu], Original ATen: [aten.relu, aten.threshold_backward]
triton_poi_fused_relu_threshold_backward_1.run(buf2, buf3, 64, grid=grid(64), stream=stream0)
return (buf2, reinterpret_tensor(buf0, (64, 4), (4, 1), 0), buf3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((1, 4), (4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class EdgeFeaturesLayer(nn.Module):
def __init__(self, d_model, d_edge, h, dropout):
super(EdgeFeaturesLayer, self).__init__()
assert d_model % h == 0
d_model // h
self.linear = nn.Linear(d_edge, 1, bias=False)
with torch.no_grad():
self.linear.weight.fill_(0.25)
def forward(self, x):
p_edge = x.permute(0, 2, 3, 1)
p_edge = self.linear(p_edge).permute(0, 3, 1, 2)
return torch.relu(p_edge)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'d_model': 4, 'd_edge': 4, 'h': 4, 'dropout': 0.5}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_clone_0(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 64
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = yindex // 16
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 16 * x2 + 64 * y1), xmask & ymask,
eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + 4 * y3), tmp0, xmask & ymask)
@triton.jit
def triton_poi_fused_relu_threshold_backward_1(in_out_ptr0, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + x0, xmask)
tmp1 = tl.full([1], 0, tl.int32)
tmp2 = triton_helpers.maximum(tmp1, tmp0)
tmp3 = 0.0
tmp4 = tmp2 <= tmp3
tl.store(in_out_ptr0 + x0, tmp2, xmask)
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (1, 4), (4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_clone_0[grid(64, 4)](primals_1, buf0, 64, 4,
XBLOCK=4, YBLOCK=32, num_warps=4, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf0, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_2, (4, 1), (1, 4), 0), out=buf1)
del primals_2
buf2 = reinterpret_tensor(buf1, (4, 1, 4, 4), (16, 1, 4, 1), 0)
del buf1
buf3 = empty_strided_cuda((4, 1, 4, 4), (16, 1, 4, 1), torch.bool)
triton_poi_fused_relu_threshold_backward_1[grid(64)](buf2, buf3, 64,
XBLOCK=64, num_warps=1, num_stages=1)
return buf2, reinterpret_tensor(buf0, (64, 4), (4, 1), 0), buf3
class EdgeFeaturesLayerNew(nn.Module):
def __init__(self, d_model, d_edge, h, dropout):
super(EdgeFeaturesLayerNew, self).__init__()
assert d_model % h == 0
d_model // h
self.linear = nn.Linear(d_edge, 1, bias=False)
with torch.no_grad():
self.linear.weight.fill_(0.25)
def forward(self, input_0):
primals_2 = self.linear.weight
primals_1 = input_0
output = call([primals_1, primals_2])
return output[0]
|
eweiner/MAT_Extension
|
EdgeFeaturesLayer
| false | 12,360 |
[
"MIT"
] | 0 |
505884a67f97bf54e1198077d15a48531fcac7a5
|
https://github.com/eweiner/MAT_Extension/tree/505884a67f97bf54e1198077d15a48531fcac7a5
|
Actor
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/6o/c6o7ainbzocsswla76yvmdsc5donraaar3dzlx2icwrueb7fc46u.py
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# x_1 => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_1,), kwargs = {})
# %le_1 : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_relu_threshold_backward_0 = async_compile.triton('triton_poi_fused_relu_threshold_backward_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i1', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_threshold_backward_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_out_ptr0, in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16384
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 256
tmp0 = tl.load(in_out_ptr0 + (x2), None)
tmp1 = tl.load(in_ptr0 + (x0), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(in_out_ptr0 + (x2), tmp4, None)
tl.store(out_ptr0 + (x2), tmp6, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/hj/chjzotk5iydxvuetxetlv36s7car7cdb24whkuqihxwcy5kkr4o2.py
# Topologically Sorted Source Nodes: [tanh], Original ATen: [aten.tanh]
# Source node to ATen node mapping:
# tanh => tanh
# Graph fragment:
# %tanh : [num_users=1] = call_function[target=torch.ops.aten.tanh.default](args = (%view_5,), kwargs = {})
triton_poi_fused_tanh_1 = async_compile.triton('triton_poi_fused_tanh_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_tanh_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_tanh_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x2), tmp3, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (256, 4), (4, 1))
assert_size_stride(primals_3, (256, ), (1, ))
assert_size_stride(primals_4, (256, 256), (256, 1))
assert_size_stride(primals_5, (256, ), (1, ))
assert_size_stride(primals_6, (4, 256), (256, 1))
assert_size_stride(primals_7, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 256), (256, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_1, (64, 4), (4, 1), 0), reinterpret_tensor(primals_2, (4, 256), (1, 4), 0), out=buf0)
del primals_2
buf1 = reinterpret_tensor(buf0, (4, 4, 4, 256), (4096, 1024, 256, 1), 0); del buf0 # reuse
buf7 = empty_strided_cuda((4, 4, 4, 256), (4096, 1024, 256, 1), torch.bool)
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.relu, aten.threshold_backward]
stream0 = get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0.run(buf1, primals_3, buf7, 16384, grid=grid(16384), stream=stream0)
del primals_3
buf2 = empty_strided_cuda((64, 256), (256, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf1, (64, 256), (256, 1), 0), reinterpret_tensor(primals_4, (256, 256), (1, 256), 0), out=buf2)
buf3 = reinterpret_tensor(buf2, (4, 4, 4, 256), (4096, 1024, 256, 1), 0); del buf2 # reuse
buf6 = empty_strided_cuda((4, 4, 4, 256), (4096, 1024, 256, 1), torch.bool)
# Topologically Sorted Source Nodes: [x_2], Original ATen: [aten.relu, aten.threshold_backward]
triton_poi_fused_relu_threshold_backward_0.run(buf3, primals_5, buf6, 16384, grid=grid(16384), stream=stream0)
del primals_5
buf4 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf3, (64, 256), (256, 1), 0), reinterpret_tensor(primals_6, (256, 4), (1, 256), 0), out=buf4)
buf5 = reinterpret_tensor(buf4, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf4 # reuse
# Topologically Sorted Source Nodes: [tanh], Original ATen: [aten.tanh]
triton_poi_fused_tanh_1.run(buf5, primals_7, 256, grid=grid(256), stream=stream0)
del primals_7
return (buf5, reinterpret_tensor(primals_1, (64, 4), (4, 1), 0), reinterpret_tensor(buf1, (64, 256), (256, 1), 0), reinterpret_tensor(buf3, (64, 256), (256, 1), 0), buf5, primals_6, buf6, primals_4, buf7, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((256, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((256, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((256, 256), (256, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((256, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 256), (256, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import numpy as np
import torch.nn.functional as F
import torch.nn as nn
class Actor(nn.Module):
def __init__(self, device, action_size, observation_size):
super(Actor, self).__init__()
self.device = device
self.fc1 = nn.Linear(np.array((observation_size,)).prod(), 256)
self.fc2 = nn.Linear(256, 256)
self.fc_mu = nn.Linear(256, np.prod((action_size,)))
def forward(self, x):
x = torch.Tensor(x)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
return torch.tanh(self.fc_mu(x))
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'device': 0, 'action_size': 4, 'observation_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice
import numpy as np
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_out_ptr0, in_ptr0,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 256
tmp0 = tl.load(in_out_ptr0 + x2, None)
tmp1 = tl.load(in_ptr0 + x0, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(in_out_ptr0 + x2, tmp4, None)
tl.store(out_ptr0 + x2, tmp6, None)
@triton.jit
def triton_poi_fused_tanh_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x2, tmp3, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7) = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (256, 4), (4, 1))
assert_size_stride(primals_3, (256,), (1,))
assert_size_stride(primals_4, (256, 256), (256, 1))
assert_size_stride(primals_5, (256,), (1,))
assert_size_stride(primals_6, (4, 256), (256, 1))
assert_size_stride(primals_7, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 256), (256, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_1, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_2, (4, 256), (1, 4), 0), out=buf0)
del primals_2
buf1 = reinterpret_tensor(buf0, (4, 4, 4, 256), (4096, 1024, 256, 1), 0
)
del buf0
buf7 = empty_strided_cuda((4, 4, 4, 256), (4096, 1024, 256, 1),
torch.bool)
get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0[grid(16384)](buf1,
primals_3, buf7, 16384, XBLOCK=128, num_warps=4, num_stages=1)
del primals_3
buf2 = empty_strided_cuda((64, 256), (256, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf1, (64, 256), (256, 1), 0),
reinterpret_tensor(primals_4, (256, 256), (1, 256), 0), out=buf2)
buf3 = reinterpret_tensor(buf2, (4, 4, 4, 256), (4096, 1024, 256, 1), 0
)
del buf2
buf6 = empty_strided_cuda((4, 4, 4, 256), (4096, 1024, 256, 1),
torch.bool)
triton_poi_fused_relu_threshold_backward_0[grid(16384)](buf3,
primals_5, buf6, 16384, XBLOCK=128, num_warps=4, num_stages=1)
del primals_5
buf4 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf3, (64, 256), (256, 1), 0),
reinterpret_tensor(primals_6, (256, 4), (1, 256), 0), out=buf4)
buf5 = reinterpret_tensor(buf4, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf4
triton_poi_fused_tanh_1[grid(256)](buf5, primals_7, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del primals_7
return buf5, reinterpret_tensor(primals_1, (64, 4), (4, 1), 0
), reinterpret_tensor(buf1, (64, 256), (256, 1), 0
), reinterpret_tensor(buf3, (64, 256), (256, 1), 0
), buf5, primals_6, buf6, primals_4, buf7
class ActorNew(nn.Module):
def __init__(self, device, action_size, observation_size):
super(ActorNew, self).__init__()
self.device = device
self.fc1 = nn.Linear(np.array((observation_size,)).prod(), 256)
self.fc2 = nn.Linear(256, 256)
self.fc_mu = nn.Linear(256, np.prod((action_size,)))
def forward(self, input_0):
primals_2 = self.fc1.weight
primals_3 = self.fc1.bias
primals_4 = self.fc2.weight
primals_5 = self.fc2.bias
primals_6 = self.fc_mu.weight
primals_7 = self.fc_mu.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7])
return output[0]
|
faisman/deep-reinforcement-learning-projects
|
Actor
| false | 12,361 |
[
"MIT"
] | 0 |
cef102ec4019069a22f95d798f6694dce73655ae
|
https://github.com/faisman/deep-reinforcement-learning-projects/tree/cef102ec4019069a22f95d798f6694dce73655ae
|
ISAB
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/cm/ccmcgo4hhocf76otuns232vkfdobmiyhbrbzce7zxp7kc5eree6u.py
# Topologically Sorted Source Nodes: [repeat], Original ATen: [aten.repeat]
# Source node to ATen node mapping:
# repeat => repeat
# Graph fragment:
# %repeat : [num_users=1] = call_function[target=torch.ops.aten.repeat.default](args = (%primals_1, [4, 1, 1]), kwargs = {})
triton_poi_fused_repeat_0 = async_compile.triton('triton_poi_fused_repeat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_repeat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_repeat_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x2), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ev/cevijwbbqed46icacbs6dadxexw32cw7j44rnr3vfdt7cyd63ztr.py
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# Q_ => cat
# Graph fragment:
# %cat : [num_users=3] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem, %getitem_1, %getitem_2, %getitem_3],), kwargs = {})
triton_poi_fused_cat_1 = async_compile.triton('triton_poi_fused_cat_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4)
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x0) + (16*x1)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + (4*x0) + (16*((-4) + x1))), tmp9 & xmask, eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + (4*x0) + (16*((-8) + x1))), tmp14 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tmp17 = tl.full([1], 16, tl.int64)
tmp18 = tmp0 < tmp17
tmp19 = tl.load(in_ptr0 + (3 + (4*x0) + (16*((-12) + x1))), tmp16 & xmask, eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + (x2), tmp22, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/7n/c7nhob3voygbzg73wacfq2bbcquxe5ozjclkx7uda5hunwppje36.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => exp
# Graph fragment:
# %mul_tensor_1 : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%bmm, 1), kwargs = {})
# %amax_default_1 : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%mul_tensor_1, [2], True), kwargs = {})
# %sub_tensor_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_tensor_1, %amax_default_1), kwargs = {})
# %div_tensor_1 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub_tensor_1, 2.0), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%div_tensor_1,), kwargs = {})
triton_poi_fused__softmax_2 = async_compile.triton('triton_poi_fused__softmax_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp3 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + (x2), tmp17, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ry/cryn7ntc2gpkbfzbre3xh7lffx7zkbskw6oihbzsekkgajmdbki6.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => div_1, sum_1
# Graph fragment:
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [2], True), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_3 = async_compile.triton('triton_poi_fused__softmax_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ho/chonql74cp63ou2wjfbbxvxkifexbc3cs3owc7nmvjuuhx6k6oqc.py
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# O => cat_3
# Graph fragment:
# %cat_3 : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem_12, %getitem_13, %getitem_14, %getitem_15], 2), kwargs = {})
triton_poi_fused_cat_4 = async_compile.triton('triton_poi_fused_cat_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = (xindex // 4)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tmp29 = tl.full([1], 4, tl.int64)
tmp30 = tmp0 < tmp29
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + (x2), tmp38, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/3j/c3jlveurtxkoy4v4xyl2fgccp7ieaeliub6jaslea3rhtsnp7cxd.py
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
# Source node to ATen node mapping:
# O_1 => add_1
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_7,), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%cat_3, %relu), kwargs = {})
# %le_1 : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_add_relu_threshold_backward_5 = async_compile.triton('triton_poi_fused_add_relu_threshold_backward_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*i1', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_relu_threshold_backward_5', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_5(in_ptr0, in_ptr1, in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp2 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + (x2), tmp6, xmask)
tl.store(out_ptr1 + (x2), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12, primals_13, primals_14, primals_15, primals_16, primals_17, primals_18 = args
args.clear()
assert_size_stride(primals_1, (1, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_3, (4, 4), (4, 1))
assert_size_stride(primals_4, (4, ), (1, ))
assert_size_stride(primals_5, (4, 4), (4, 1))
assert_size_stride(primals_6, (4, ), (1, ))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4, ), (1, ))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4, ), (1, ))
assert_size_stride(primals_11, (4, 4), (4, 1))
assert_size_stride(primals_12, (4, ), (1, ))
assert_size_stride(primals_13, (4, 4), (4, 1))
assert_size_stride(primals_14, (4, ), (1, ))
assert_size_stride(primals_15, (4, 4), (4, 1))
assert_size_stride(primals_16, (4, ), (1, ))
assert_size_stride(primals_17, (4, 4), (4, 1))
assert_size_stride(primals_18, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [repeat], Original ATen: [aten.repeat]
stream0 = get_raw_stream(0)
triton_poi_fused_repeat_0.run(primals_1, buf0, 64, grid=grid(64), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [Q], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_4, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), reinterpret_tensor(primals_3, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [K], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_6, reinterpret_tensor(primals_2, (16, 4), (4, 1), 0), reinterpret_tensor(primals_5, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_5
del primals_6
buf3 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [V], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_2, (16, 4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf3)
del primals_7
del primals_8
buf4 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf1, buf4, 64, grid=grid(64), stream=stream0)
buf5 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 64), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [V_], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf3, buf5, 64, grid=grid(64), stream=stream0)
buf6 = reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 1), 0); del buf3 # reuse
# Topologically Sorted Source Nodes: [K_], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf2, buf6, 64, grid=grid(64), stream=stream0)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [bmm], Original ATen: [aten.bmm]
extern_kernels.bmm(buf4, reinterpret_tensor(buf6, (16, 1, 4), (4, 0, 1), 0), out=buf7)
buf8 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_2.run(buf7, buf8, 256, grid=grid(256), stream=stream0)
buf9 = buf7; del buf7 # reuse
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_3.run(buf8, buf9, 256, grid=grid(256), stream=stream0)
buf10 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [bmm_1], Original ATen: [aten.bmm]
extern_kernels.bmm(buf9, buf5, out=buf10)
buf11 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
triton_poi_fused_cat_4.run(buf4, buf10, buf11, 64, grid=grid(64), stream=stream0)
buf12 = reinterpret_tensor(buf10, (16, 4), (4, 1), 0); del buf10 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf11, (16, 4), (4, 1), 0), reinterpret_tensor(primals_9, (4, 4), (1, 4), 0), out=buf12)
buf13 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [Q_1], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_12, reinterpret_tensor(primals_2, (16, 4), (4, 1), 0), reinterpret_tensor(primals_11, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf13)
del primals_11
del primals_12
buf14 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf28 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
triton_poi_fused_add_relu_threshold_backward_5.run(buf11, buf12, primals_10, buf14, buf28, 64, grid=grid(64), stream=stream0)
del primals_10
buf15 = buf12; del buf12 # reuse
# Topologically Sorted Source Nodes: [K_1], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_14, reinterpret_tensor(buf14, (16, 4), (4, 1), 0), reinterpret_tensor(primals_13, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf15)
del primals_14
buf16 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [V_1], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_16, reinterpret_tensor(buf14, (16, 4), (4, 1), 0), reinterpret_tensor(primals_15, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf16)
del primals_16
buf17 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
# Topologically Sorted Source Nodes: [Q__1], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf13, buf17, 64, grid=grid(64), stream=stream0)
buf18 = reinterpret_tensor(buf13, (16, 4, 1), (4, 1, 64), 0); del buf13 # reuse
# Topologically Sorted Source Nodes: [V__1], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf16, buf18, 64, grid=grid(64), stream=stream0)
buf19 = reinterpret_tensor(buf16, (16, 4, 1), (4, 1, 1), 0); del buf16 # reuse
# Topologically Sorted Source Nodes: [K__1], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf15, buf19, 64, grid=grid(64), stream=stream0)
buf20 = buf8; del buf8 # reuse
# Topologically Sorted Source Nodes: [bmm_2], Original ATen: [aten.bmm]
extern_kernels.bmm(buf17, reinterpret_tensor(buf19, (16, 1, 4), (4, 0, 1), 0), out=buf20)
buf21 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [A_1], Original ATen: [aten._softmax]
triton_poi_fused__softmax_2.run(buf20, buf21, 256, grid=grid(256), stream=stream0)
buf22 = buf20; del buf20 # reuse
# Topologically Sorted Source Nodes: [A_1], Original ATen: [aten._softmax]
triton_poi_fused__softmax_3.run(buf21, buf22, 256, grid=grid(256), stream=stream0)
del buf21
buf23 = reinterpret_tensor(buf15, (16, 4, 1), (4, 1, 1), 0); del buf15 # reuse
# Topologically Sorted Source Nodes: [bmm_3], Original ATen: [aten.bmm]
extern_kernels.bmm(buf22, buf18, out=buf23)
buf24 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [O_2], Original ATen: [aten.cat]
triton_poi_fused_cat_4.run(buf17, buf23, buf24, 64, grid=grid(64), stream=stream0)
buf25 = reinterpret_tensor(buf23, (16, 4), (4, 1), 0); del buf23 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf24, (16, 4), (4, 1), 0), reinterpret_tensor(primals_17, (4, 4), (1, 4), 0), out=buf25)
buf26 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf27 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu_1, O_3], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
triton_poi_fused_add_relu_threshold_backward_5.run(buf24, buf25, primals_18, buf26, buf27, 64, grid=grid(64), stream=stream0)
del buf25
del primals_18
return (buf26, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), reinterpret_tensor(primals_2, (16, 4), (4, 1), 0), buf9, reinterpret_tensor(buf11, (16, 4), (4, 1), 0), reinterpret_tensor(buf14, (16, 4), (4, 1), 0), buf22, reinterpret_tensor(buf24, (16, 4), (4, 1), 0), buf27, primals_17, reinterpret_tensor(buf18, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf17, (16, 1, 4), (4, 1, 1), 0), buf19, primals_15, primals_13, buf28, primals_9, reinterpret_tensor(buf5, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf4, (16, 1, 4), (4, 1, 1), 0), buf6, primals_3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_12 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_13 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_14 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_15 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_16 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_17 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_18 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12, primals_13, primals_14, primals_15, primals_16, primals_17, primals_18])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
class MAB(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MAB, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, Q, K):
Q = self.fc_q(Q)
K, V = self.fc_k(K), self.fc_v(K)
dim_split = self.dim_V // self.num_heads
Q_ = torch.cat(Q.split(dim_split, 2), 0)
K_ = torch.cat(K.split(dim_split, 2), 0)
V_ = torch.cat(V.split(dim_split, 2), 0)
A = torch.softmax(Q_.bmm(K_.transpose(1, 2)) / math.sqrt(self.dim_V), 2
)
O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
O = O + F.relu(self.fc_o(O))
O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
return O
class ISAB(nn.Module):
def __init__(self, dim_in, dim_out, num_heads, num_inds, ln=False):
super(ISAB, self).__init__()
self.I = nn.Parameter(torch.Tensor(1, num_inds, dim_out))
nn.init.xavier_uniform_(self.I)
self.mab0 = MAB(dim_out, dim_in, dim_out, num_heads, ln=ln)
self.mab1 = MAB(dim_in, dim_out, dim_out, num_heads, ln=ln)
def forward(self, X):
H = self.mab0(self.I.repeat(X.size(0), 1, 1), X)
return self.mab1(X, H)
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'dim_in': 4, 'dim_out': 4, 'num_heads': 4, 'num_inds': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import math
import torch.nn as nn
import torch.nn.functional as F
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_repeat_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x2 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tl.store(out_ptr0 + x2, tmp0, xmask)
@triton.jit
def triton_poi_fused_cat_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x0 + 16 * x1), tmp4 & xmask,
eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + 4 * x0 + 16 * (-4 + x1)), tmp9 & xmask,
eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + 4 * x0 + 16 * (-8 + x1)), tmp14 & xmask,
eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tl.full([1], 16, tl.int64)
tmp19 = tl.load(in_ptr0 + (3 + 4 * x0 + 16 * (-12 + x1)), tmp16 & xmask,
eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + x2, tmp22, xmask)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp3 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + x2, tmp17, xmask)
@triton.jit
def triton_poi_fused__softmax_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = xindex // 4
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp6 = tl.load(in_ptr1 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tl.full([1], 4, tl.int64)
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + x2, tmp38, xmask)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_5(in_ptr0, in_ptr1,
in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp2 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + x2, tmp6, xmask)
tl.store(out_ptr1 + x2, tmp8, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11, primals_12,
primals_13, primals_14, primals_15, primals_16, primals_17, primals_18
) = args
args.clear()
assert_size_stride(primals_1, (1, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_3, (4, 4), (4, 1))
assert_size_stride(primals_4, (4,), (1,))
assert_size_stride(primals_5, (4, 4), (4, 1))
assert_size_stride(primals_6, (4,), (1,))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4,), (1,))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4,), (1,))
assert_size_stride(primals_11, (4, 4), (4, 1))
assert_size_stride(primals_12, (4,), (1,))
assert_size_stride(primals_13, (4, 4), (4, 1))
assert_size_stride(primals_14, (4,), (1,))
assert_size_stride(primals_15, (4, 4), (4, 1))
assert_size_stride(primals_16, (4,), (1,))
assert_size_stride(primals_17, (4, 4), (4, 1))
assert_size_stride(primals_18, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_repeat_0[grid(64)](primals_1, buf0, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_4, reinterpret_tensor(buf0, (16, 4), (
4, 1), 0), reinterpret_tensor(primals_3, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_6, reinterpret_tensor(primals_2, (16,
4), (4, 1), 0), reinterpret_tensor(primals_5, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf2)
del primals_5
del primals_6
buf3 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_2, (16,
4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf3)
del primals_7
del primals_8
buf4 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
triton_poi_fused_cat_1[grid(64)](buf1, buf4, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf5 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 64), 0)
del buf1
triton_poi_fused_cat_1[grid(64)](buf3, buf5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf6 = reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 1), 0)
del buf3
triton_poi_fused_cat_1[grid(64)](buf2, buf6, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(buf4, reinterpret_tensor(buf6, (16, 1, 4), (4, 0,
1), 0), out=buf7)
buf8 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_2[grid(256)](buf7, buf8, 256, XBLOCK=256,
num_warps=4, num_stages=1)
buf9 = buf7
del buf7
triton_poi_fused__softmax_3[grid(256)](buf8, buf9, 256, XBLOCK=128,
num_warps=4, num_stages=1)
buf10 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0)
del buf2
extern_kernels.bmm(buf9, buf5, out=buf10)
buf11 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_cat_4[grid(64)](buf4, buf10, buf11, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf12 = reinterpret_tensor(buf10, (16, 4), (4, 1), 0)
del buf10
extern_kernels.mm(reinterpret_tensor(buf11, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_9, (4, 4), (1, 4), 0), out=buf12)
buf13 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_12, reinterpret_tensor(primals_2, (16,
4), (4, 1), 0), reinterpret_tensor(primals_11, (4, 4), (1, 4),
0), alpha=1, beta=1, out=buf13)
del primals_11
del primals_12
buf14 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf28 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
triton_poi_fused_add_relu_threshold_backward_5[grid(64)](buf11,
buf12, primals_10, buf14, buf28, 64, XBLOCK=64, num_warps=1,
num_stages=1)
del primals_10
buf15 = buf12
del buf12
extern_kernels.addmm(primals_14, reinterpret_tensor(buf14, (16, 4),
(4, 1), 0), reinterpret_tensor(primals_13, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf15)
del primals_14
buf16 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_16, reinterpret_tensor(buf14, (16, 4),
(4, 1), 0), reinterpret_tensor(primals_15, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf16)
del primals_16
buf17 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
triton_poi_fused_cat_1[grid(64)](buf13, buf17, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf18 = reinterpret_tensor(buf13, (16, 4, 1), (4, 1, 64), 0)
del buf13
triton_poi_fused_cat_1[grid(64)](buf16, buf18, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf19 = reinterpret_tensor(buf16, (16, 4, 1), (4, 1, 1), 0)
del buf16
triton_poi_fused_cat_1[grid(64)](buf15, buf19, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf20 = buf8
del buf8
extern_kernels.bmm(buf17, reinterpret_tensor(buf19, (16, 1, 4), (4,
0, 1), 0), out=buf20)
buf21 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_2[grid(256)](buf20, buf21, 256, XBLOCK=
256, num_warps=4, num_stages=1)
buf22 = buf20
del buf20
triton_poi_fused__softmax_3[grid(256)](buf21, buf22, 256, XBLOCK=
128, num_warps=4, num_stages=1)
del buf21
buf23 = reinterpret_tensor(buf15, (16, 4, 1), (4, 1, 1), 0)
del buf15
extern_kernels.bmm(buf22, buf18, out=buf23)
buf24 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_cat_4[grid(64)](buf17, buf23, buf24, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf25 = reinterpret_tensor(buf23, (16, 4), (4, 1), 0)
del buf23
extern_kernels.mm(reinterpret_tensor(buf24, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_17, (4, 4), (1, 4), 0), out=buf25)
buf26 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf27 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
triton_poi_fused_add_relu_threshold_backward_5[grid(64)](buf24,
buf25, primals_18, buf26, buf27, 64, XBLOCK=64, num_warps=1,
num_stages=1)
del buf25
del primals_18
return buf26, reinterpret_tensor(buf0, (16, 4), (4, 1), 0
), reinterpret_tensor(primals_2, (16, 4), (4, 1), 0
), buf9, reinterpret_tensor(buf11, (16, 4), (4, 1), 0
), reinterpret_tensor(buf14, (16, 4), (4, 1), 0
), buf22, reinterpret_tensor(buf24, (16, 4), (4, 1), 0
), buf27, primals_17, reinterpret_tensor(buf18, (16, 1, 4), (4, 1,
1), 0), reinterpret_tensor(buf17, (16, 1, 4), (4, 1, 1), 0
), buf19, primals_15, primals_13, buf28, primals_9, reinterpret_tensor(
buf5, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf4, (16, 1, 4
), (4, 1, 1), 0), buf6, primals_3
class MAB(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MAB, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, Q, K):
Q = self.fc_q(Q)
K, V = self.fc_k(K), self.fc_v(K)
dim_split = self.dim_V // self.num_heads
Q_ = torch.cat(Q.split(dim_split, 2), 0)
K_ = torch.cat(K.split(dim_split, 2), 0)
V_ = torch.cat(V.split(dim_split, 2), 0)
A = torch.softmax(Q_.bmm(K_.transpose(1, 2)) / math.sqrt(self.dim_V), 2
)
O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
O = O + F.relu(self.fc_o(O))
O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
return O
class ISABNew(nn.Module):
def __init__(self, dim_in, dim_out, num_heads, num_inds, ln=False):
super(ISABNew, self).__init__()
self.I = nn.Parameter(torch.Tensor(1, num_inds, dim_out))
nn.init.xavier_uniform_(self.I)
self.mab0 = MAB(dim_out, dim_in, dim_out, num_heads, ln=ln)
self.mab1 = MAB(dim_in, dim_out, dim_out, num_heads, ln=ln)
def forward(self, input_0):
primals_1 = self.I
primals_3 = self.mab0.fc_q.weight
primals_4 = self.mab0.fc_q.bias
primals_5 = self.mab0.fc_k.weight
primals_6 = self.mab0.fc_k.bias
primals_7 = self.mab0.fc_v.weight
primals_8 = self.mab0.fc_v.bias
primals_9 = self.mab0.fc_o.weight
primals_10 = self.mab0.fc_o.bias
primals_11 = self.mab1.fc_q.weight
primals_12 = self.mab1.fc_q.bias
primals_13 = self.mab1.fc_k.weight
primals_14 = self.mab1.fc_k.bias
primals_15 = self.mab1.fc_v.weight
primals_16 = self.mab1.fc_v.bias
primals_17 = self.mab1.fc_o.weight
primals_18 = self.mab1.fc_o.bias
primals_2 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11, primals_12, primals_13, primals_14,
primals_15, primals_16, primals_17, primals_18])
return output[0]
|
ernoult/set_transformer
|
ISAB
| false | 12,362 |
[
"MIT"
] | 0 |
4b380106e1f43b7eb6315624c57d4d1d38737b78
|
https://github.com/ernoult/set_transformer/tree/4b380106e1f43b7eb6315624c57d4d1d38737b78
|
PMA
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/cm/ccmcgo4hhocf76otuns232vkfdobmiyhbrbzce7zxp7kc5eree6u.py
# Topologically Sorted Source Nodes: [repeat], Original ATen: [aten.repeat]
# Source node to ATen node mapping:
# repeat => repeat
# Graph fragment:
# %repeat : [num_users=1] = call_function[target=torch.ops.aten.repeat.default](args = (%primals_1, [4, 1, 1]), kwargs = {})
triton_poi_fused_repeat_0 = async_compile.triton('triton_poi_fused_repeat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_repeat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_repeat_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x2), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ev/cevijwbbqed46icacbs6dadxexw32cw7j44rnr3vfdt7cyd63ztr.py
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# Q_ => cat
# Graph fragment:
# %cat : [num_users=3] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem, %getitem_1, %getitem_2, %getitem_3],), kwargs = {})
triton_poi_fused_cat_1 = async_compile.triton('triton_poi_fused_cat_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4)
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x0) + (16*x1)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + (4*x0) + (16*((-4) + x1))), tmp9 & xmask, eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + (4*x0) + (16*((-8) + x1))), tmp14 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tmp17 = tl.full([1], 16, tl.int64)
tmp18 = tmp0 < tmp17
tmp19 = tl.load(in_ptr0 + (3 + (4*x0) + (16*((-12) + x1))), tmp16 & xmask, eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + (x2), tmp22, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/7n/c7nhob3voygbzg73wacfq2bbcquxe5ozjclkx7uda5hunwppje36.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => exp
# Graph fragment:
# %mul_tensor : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%bmm, 1), kwargs = {})
# %amax_default : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%mul_tensor, [2], True), kwargs = {})
# %sub_tensor : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_tensor, %amax_default), kwargs = {})
# %div_tensor : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub_tensor, 2.0), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%div_tensor,), kwargs = {})
triton_poi_fused__softmax_2 = async_compile.triton('triton_poi_fused__softmax_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp3 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + (x2), tmp17, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ry/cryn7ntc2gpkbfzbre3xh7lffx7zkbskw6oihbzsekkgajmdbki6.py
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# A => div_1, sum_1
# Graph fragment:
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [2], True), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_3 = async_compile.triton('triton_poi_fused__softmax_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ho/chonql74cp63ou2wjfbbxvxkifexbc3cs3owc7nmvjuuhx6k6oqc.py
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# O => cat_3
# Graph fragment:
# %cat_3 : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%getitem_12, %getitem_13, %getitem_14, %getitem_15], 2), kwargs = {})
triton_poi_fused_cat_4 = async_compile.triton('triton_poi_fused_cat_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = (xindex // 4)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + (x1), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy='evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy='evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tmp29 = tl.full([1], 4, tl.int64)
tmp30 = tmp0 < tmp29
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy='evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + (x2), tmp38, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/3j/c3jlveurtxkoy4v4xyl2fgccp7ieaeliub6jaslea3rhtsnp7cxd.py
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
# Source node to ATen node mapping:
# O_1 => add_1
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_7,), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%cat_3, %relu), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_add_relu_threshold_backward_5 = async_compile.triton('triton_poi_fused_add_relu_threshold_backward_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*i1', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_relu_threshold_backward_5', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_5(in_ptr0, in_ptr1, in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp2 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + (x2), tmp6, xmask)
tl.store(out_ptr1 + (x2), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10 = args
args.clear()
assert_size_stride(primals_1, (1, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_3, (4, 4), (4, 1))
assert_size_stride(primals_4, (4, ), (1, ))
assert_size_stride(primals_5, (4, 4), (4, 1))
assert_size_stride(primals_6, (4, ), (1, ))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4, ), (1, ))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [repeat], Original ATen: [aten.repeat]
stream0 = get_raw_stream(0)
triton_poi_fused_repeat_0.run(primals_1, buf0, 64, grid=grid(64), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [Q], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_4, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), reinterpret_tensor(primals_3, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [K], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_6, reinterpret_tensor(primals_2, (16, 4), (4, 1), 0), reinterpret_tensor(primals_5, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_5
del primals_6
buf3 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [V], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_2, (16, 4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf3)
del primals_7
del primals_8
buf4 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
# Topologically Sorted Source Nodes: [Q_], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf1, buf4, 64, grid=grid(64), stream=stream0)
buf5 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 64), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [V_], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf3, buf5, 64, grid=grid(64), stream=stream0)
buf6 = reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 1), 0); del buf3 # reuse
# Topologically Sorted Source Nodes: [K_], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(buf2, buf6, 64, grid=grid(64), stream=stream0)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [bmm], Original ATen: [aten.bmm]
extern_kernels.bmm(buf4, reinterpret_tensor(buf6, (16, 1, 4), (4, 0, 1), 0), out=buf7)
buf8 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_2.run(buf7, buf8, 256, grid=grid(256), stream=stream0)
buf9 = buf7; del buf7 # reuse
# Topologically Sorted Source Nodes: [A], Original ATen: [aten._softmax]
triton_poi_fused__softmax_3.run(buf8, buf9, 256, grid=grid(256), stream=stream0)
del buf8
buf10 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [bmm_1], Original ATen: [aten.bmm]
extern_kernels.bmm(buf9, buf5, out=buf10)
buf11 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [O], Original ATen: [aten.cat]
triton_poi_fused_cat_4.run(buf4, buf10, buf11, 64, grid=grid(64), stream=stream0)
buf12 = reinterpret_tensor(buf10, (16, 4), (4, 1), 0); del buf10 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf11, (16, 4), (4, 1), 0), reinterpret_tensor(primals_9, (4, 4), (1, 4), 0), out=buf12)
buf13 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf14 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu, O_1], Original ATen: [aten.relu, aten.add, aten.threshold_backward]
triton_poi_fused_add_relu_threshold_backward_5.run(buf11, buf12, primals_10, buf13, buf14, 64, grid=grid(64), stream=stream0)
del buf12
del primals_10
return (buf13, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), reinterpret_tensor(primals_2, (16, 4), (4, 1), 0), buf9, reinterpret_tensor(buf11, (16, 4), (4, 1), 0), buf14, primals_9, reinterpret_tensor(buf5, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf4, (16, 1, 4), (4, 1, 1), 0), buf6, primals_3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
class MAB(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MAB, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, Q, K):
Q = self.fc_q(Q)
K, V = self.fc_k(K), self.fc_v(K)
dim_split = self.dim_V // self.num_heads
Q_ = torch.cat(Q.split(dim_split, 2), 0)
K_ = torch.cat(K.split(dim_split, 2), 0)
V_ = torch.cat(V.split(dim_split, 2), 0)
A = torch.softmax(Q_.bmm(K_.transpose(1, 2)) / math.sqrt(self.dim_V), 2
)
O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
O = O + F.relu(self.fc_o(O))
O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
return O
class PMA(nn.Module):
def __init__(self, dim, num_heads, num_seeds, ln=False):
super(PMA, self).__init__()
self.S = nn.Parameter(torch.Tensor(1, num_seeds, dim))
nn.init.xavier_uniform_(self.S)
self.mab = MAB(dim, dim, dim, num_heads, ln=ln)
def forward(self, X):
return self.mab(self.S.repeat(X.size(0), 1, 1), X)
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'dim': 4, 'num_heads': 4, 'num_seeds': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import math
import torch.nn as nn
import torch.nn.functional as F
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_repeat_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x2 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tl.store(out_ptr0 + x2, tmp0, xmask)
@triton.jit
def triton_poi_fused_cat_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4
x0 = xindex % 4
x2 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x0 + 16 * x1), tmp4 & xmask,
eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tl.load(in_ptr0 + (1 + 4 * x0 + 16 * (-4 + x1)), tmp9 & xmask,
eviction_policy='evict_last', other=0.0)
tmp11 = tmp0 >= tmp7
tmp12 = tl.full([1], 12, tl.int64)
tmp13 = tmp0 < tmp12
tmp14 = tmp11 & tmp13
tmp15 = tl.load(in_ptr0 + (2 + 4 * x0 + 16 * (-8 + x1)), tmp14 & xmask,
eviction_policy='evict_last', other=0.0)
tmp16 = tmp0 >= tmp12
tl.full([1], 16, tl.int64)
tmp19 = tl.load(in_ptr0 + (3 + 4 * x0 + 16 * (-12 + x1)), tmp16 & xmask,
eviction_policy='evict_last', other=0.0)
tmp20 = tl.where(tmp14, tmp15, tmp19)
tmp21 = tl.where(tmp9, tmp10, tmp20)
tmp22 = tl.where(tmp4, tmp5, tmp21)
tl.store(out_ptr0 + x2, tmp22, xmask)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp3 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = 0.5
tmp16 = tmp14 * tmp15
tmp17 = tl_math.exp(tmp16)
tl.store(out_ptr0 + x2, tmp17, xmask)
@triton.jit
def triton_poi_fused__softmax_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = xindex // 4
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp6 = tl.load(in_ptr1 + x1, tmp4 & xmask, eviction_policy='evict_last',
other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full(tmp7.shape, 0.0, tmp7.dtype)
tmp9 = tl.where(tmp4, tmp7, tmp8)
tmp10 = tmp0 >= tmp3
tmp11 = tl.full([1], 2, tl.int64)
tmp12 = tmp0 < tmp11
tmp13 = tmp10 & tmp12
tmp14 = tl.load(in_ptr0 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp15 = tl.load(in_ptr1 + (16 + x1), tmp13 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp16 = tmp14 + tmp15
tmp17 = tl.full(tmp16.shape, 0.0, tmp16.dtype)
tmp18 = tl.where(tmp13, tmp16, tmp17)
tmp19 = tmp0 >= tmp11
tmp20 = tl.full([1], 3, tl.int64)
tmp21 = tmp0 < tmp20
tmp22 = tmp19 & tmp21
tmp23 = tl.load(in_ptr0 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp24 = tl.load(in_ptr1 + (32 + x1), tmp22 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp25 = tmp23 + tmp24
tmp26 = tl.full(tmp25.shape, 0.0, tmp25.dtype)
tmp27 = tl.where(tmp22, tmp25, tmp26)
tmp28 = tmp0 >= tmp20
tl.full([1], 4, tl.int64)
tmp31 = tl.load(in_ptr0 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp32 = tl.load(in_ptr1 + (48 + x1), tmp28 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp33 = tmp31 + tmp32
tmp34 = tl.full(tmp33.shape, 0.0, tmp33.dtype)
tmp35 = tl.where(tmp28, tmp33, tmp34)
tmp36 = tl.where(tmp22, tmp27, tmp35)
tmp37 = tl.where(tmp13, tmp18, tmp36)
tmp38 = tl.where(tmp4, tmp9, tmp37)
tl.store(out_ptr0 + x2, tmp38, xmask)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_5(in_ptr0, in_ptr1,
in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp2 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp4 = tl.full([1], 0, tl.int32)
tmp5 = triton_helpers.maximum(tmp4, tmp3)
tmp6 = tmp0 + tmp5
tmp7 = 0.0
tmp8 = tmp5 <= tmp7
tl.store(out_ptr0 + x2, tmp6, xmask)
tl.store(out_ptr1 + x2, tmp8, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10) = args
args.clear()
assert_size_stride(primals_1, (1, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_3, (4, 4), (4, 1))
assert_size_stride(primals_4, (4,), (1,))
assert_size_stride(primals_5, (4, 4), (4, 1))
assert_size_stride(primals_6, (4,), (1,))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4,), (1,))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_repeat_0[grid(64)](primals_1, buf0, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_4, reinterpret_tensor(buf0, (16, 4), (
4, 1), 0), reinterpret_tensor(primals_3, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_6, reinterpret_tensor(primals_2, (16,
4), (4, 1), 0), reinterpret_tensor(primals_5, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf2)
del primals_5
del primals_6
buf3 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_2, (16,
4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf3)
del primals_7
del primals_8
buf4 = empty_strided_cuda((16, 4, 1), (4, 1, 64), torch.float32)
triton_poi_fused_cat_1[grid(64)](buf1, buf4, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf5 = reinterpret_tensor(buf1, (16, 4, 1), (4, 1, 64), 0)
del buf1
triton_poi_fused_cat_1[grid(64)](buf3, buf5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf6 = reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 1), 0)
del buf3
triton_poi_fused_cat_1[grid(64)](buf2, buf6, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf7 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(buf4, reinterpret_tensor(buf6, (16, 1, 4), (4, 0,
1), 0), out=buf7)
buf8 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_2[grid(256)](buf7, buf8, 256, XBLOCK=256,
num_warps=4, num_stages=1)
buf9 = buf7
del buf7
triton_poi_fused__softmax_3[grid(256)](buf8, buf9, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del buf8
buf10 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0)
del buf2
extern_kernels.bmm(buf9, buf5, out=buf10)
buf11 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_cat_4[grid(64)](buf4, buf10, buf11, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf12 = reinterpret_tensor(buf10, (16, 4), (4, 1), 0)
del buf10
extern_kernels.mm(reinterpret_tensor(buf11, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_9, (4, 4), (1, 4), 0), out=buf12)
buf13 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf14 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.bool)
triton_poi_fused_add_relu_threshold_backward_5[grid(64)](buf11,
buf12, primals_10, buf13, buf14, 64, XBLOCK=64, num_warps=1,
num_stages=1)
del buf12
del primals_10
return buf13, reinterpret_tensor(buf0, (16, 4), (4, 1), 0
), reinterpret_tensor(primals_2, (16, 4), (4, 1), 0
), buf9, reinterpret_tensor(buf11, (16, 4), (4, 1), 0
), buf14, primals_9, reinterpret_tensor(buf5, (16, 1, 4), (4, 1, 1), 0
), reinterpret_tensor(buf4, (16, 1, 4), (4, 1, 1), 0), buf6, primals_3
class MAB(nn.Module):
def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
super(MAB, self).__init__()
self.dim_V = dim_V
self.num_heads = num_heads
self.fc_q = nn.Linear(dim_Q, dim_V)
self.fc_k = nn.Linear(dim_K, dim_V)
self.fc_v = nn.Linear(dim_K, dim_V)
if ln:
self.ln0 = nn.LayerNorm(dim_V)
self.ln1 = nn.LayerNorm(dim_V)
self.fc_o = nn.Linear(dim_V, dim_V)
def forward(self, Q, K):
Q = self.fc_q(Q)
K, V = self.fc_k(K), self.fc_v(K)
dim_split = self.dim_V // self.num_heads
Q_ = torch.cat(Q.split(dim_split, 2), 0)
K_ = torch.cat(K.split(dim_split, 2), 0)
V_ = torch.cat(V.split(dim_split, 2), 0)
A = torch.softmax(Q_.bmm(K_.transpose(1, 2)) / math.sqrt(self.dim_V), 2
)
O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
O = O + F.relu(self.fc_o(O))
O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
return O
class PMANew(nn.Module):
def __init__(self, dim, num_heads, num_seeds, ln=False):
super(PMANew, self).__init__()
self.S = nn.Parameter(torch.Tensor(1, num_seeds, dim))
nn.init.xavier_uniform_(self.S)
self.mab = MAB(dim, dim, dim, num_heads, ln=ln)
def forward(self, input_0):
primals_1 = self.S
primals_3 = self.mab.fc_q.weight
primals_4 = self.mab.fc_q.bias
primals_5 = self.mab.fc_k.weight
primals_6 = self.mab.fc_k.bias
primals_7 = self.mab.fc_v.weight
primals_8 = self.mab.fc_v.bias
primals_9 = self.mab.fc_o.weight
primals_10 = self.mab.fc_o.bias
primals_2 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9, primals_10])
return output[0]
|
ernoult/set_transformer
|
PMA
| false | 12,363 |
[
"MIT"
] | 0 |
4b380106e1f43b7eb6315624c57d4d1d38737b78
|
https://github.com/ernoult/set_transformer/tree/4b380106e1f43b7eb6315624c57d4d1d38737b78
|
QNetwork
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ms/cmsuzohbg5nq52jnvirovzkvykrzzko5xomu7zyu5e5u2lhegppw.py
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# x_1 => cat
# Graph fragment:
# %cat : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%primals_1, %primals_2], 1), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 32
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x1 = (xindex // 8)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x1) + x0), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + ((4*x1) + ((-4) + x0)), tmp6 & xmask, eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x2), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/y2/cy2lwgz7dq2q2z4ifepdde4l7vyyvrwcx4zjn2ezmtzcanvhv374.py
# Topologically Sorted Source Nodes: [x_2], Original ATen: [aten.relu]
# Source node to ATen node mapping:
# x_2 => relu
# Graph fragment:
# %add_tensor_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mm_default_1, %primals_4), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%add_tensor_1,), kwargs = {})
triton_poi_fused_relu_1 = async_compile.triton('triton_poi_fused_relu_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 256
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x2), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (256, 8), (8, 1))
assert_size_stride(primals_4, (256, ), (1, ))
assert_size_stride(primals_5, (256, 256), (256, 1))
assert_size_stride(primals_6, (256, ), (1, ))
assert_size_stride(primals_7, (1, 256), (256, 1))
assert_size_stride(primals_8, (1, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(primals_1, primals_2, buf0, 32, grid=grid(32), stream=stream0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((4, 256), (256, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(buf0, reinterpret_tensor(primals_3, (8, 256), (1, 8), 0), out=buf1)
del primals_3
buf2 = buf1; del buf1 # reuse
# Topologically Sorted Source Nodes: [x_2], Original ATen: [aten.relu]
triton_poi_fused_relu_1.run(buf2, primals_4, 1024, grid=grid(1024), stream=stream0)
del primals_4
buf3 = empty_strided_cuda((4, 256), (256, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(buf2, reinterpret_tensor(primals_5, (256, 256), (1, 256), 0), out=buf3)
buf4 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.relu]
triton_poi_fused_relu_1.run(buf4, primals_6, 1024, grid=grid(1024), stream=stream0)
del primals_6
buf6 = empty_strided_cuda((4, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_4], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_8, buf4, reinterpret_tensor(primals_7, (256, 1), (1, 256), 0), alpha=1, beta=1, out=buf6)
del primals_8
return (buf6, buf0, buf2, buf4, primals_7, primals_5, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((256, 8), (8, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((256, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((256, 256), (256, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((256, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((1, 256), (256, 1), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((1, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import numpy as np
import torch.nn.functional as F
import torch.nn as nn
class QNetwork(nn.Module):
def __init__(self, device, action_size, observation_size):
super(QNetwork, self).__init__()
self.device = device
self.fc1 = nn.Linear(np.array((observation_size,)).prod() + np.prod
((action_size,)), 256)
self.fc2 = nn.Linear(256, 256)
self.fc3 = nn.Linear(256, 1)
def forward(self, x, a):
x = torch.Tensor(x)
x = torch.cat([x, a], 1)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
def get_inputs():
return [torch.rand([4, 4]), torch.rand([4, 4])]
def get_init_inputs():
return [[], {'device': 0, 'action_size': 4, 'observation_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import numpy as np
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 32
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x1 = xindex // 8
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x1 + x0), tmp4 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (4 * x1 + (-4 + x0)), tmp6 & xmask,
eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x2, tmp10, xmask)
@triton.jit
def triton_poi_fused_relu_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 256
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x2, tmp4, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (256, 8), (8, 1))
assert_size_stride(primals_4, (256,), (1,))
assert_size_stride(primals_5, (256, 256), (256, 1))
assert_size_stride(primals_6, (256,), (1,))
assert_size_stride(primals_7, (1, 256), (256, 1))
assert_size_stride(primals_8, (1,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 8), (8, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(32)](primals_1, primals_2, buf0, 32,
XBLOCK=32, num_warps=1, num_stages=1)
del primals_1
del primals_2
buf1 = empty_strided_cuda((4, 256), (256, 1), torch.float32)
extern_kernels.mm(buf0, reinterpret_tensor(primals_3, (8, 256), (1,
8), 0), out=buf1)
del primals_3
buf2 = buf1
del buf1
triton_poi_fused_relu_1[grid(1024)](buf2, primals_4, 1024, XBLOCK=
256, num_warps=4, num_stages=1)
del primals_4
buf3 = empty_strided_cuda((4, 256), (256, 1), torch.float32)
extern_kernels.mm(buf2, reinterpret_tensor(primals_5, (256, 256), (
1, 256), 0), out=buf3)
buf4 = buf3
del buf3
triton_poi_fused_relu_1[grid(1024)](buf4, primals_6, 1024, XBLOCK=
256, num_warps=4, num_stages=1)
del primals_6
buf6 = empty_strided_cuda((4, 1), (1, 1), torch.float32)
extern_kernels.addmm(primals_8, buf4, reinterpret_tensor(primals_7,
(256, 1), (1, 256), 0), alpha=1, beta=1, out=buf6)
del primals_8
return buf6, buf0, buf2, buf4, primals_7, primals_5
class QNetworkNew(nn.Module):
def __init__(self, device, action_size, observation_size):
super(QNetworkNew, self).__init__()
self.device = device
self.fc1 = nn.Linear(np.array((observation_size,)).prod() + np.prod
((action_size,)), 256)
self.fc2 = nn.Linear(256, 256)
self.fc3 = nn.Linear(256, 1)
def forward(self, input_0, input_1):
primals_3 = self.fc1.weight
primals_4 = self.fc1.bias
primals_5 = self.fc2.weight
primals_6 = self.fc2.bias
primals_7 = self.fc3.weight
primals_8 = self.fc3.bias
primals_1 = input_0
primals_2 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8])
return output[0]
|
faisman/deep-reinforcement-learning-projects
|
QNetwork
| false | 12,364 |
[
"MIT"
] | 0 |
cef102ec4019069a22f95d798f6694dce73655ae
|
https://github.com/faisman/deep-reinforcement-learning-projects/tree/cef102ec4019069a22f95d798f6694dce73655ae
|
MultiHeadedAttention
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/kw/ckwxyj3tnxjcjymmm5nl44ibsc366pu6pfs6wcbv6bj4bdxha34f.py
# Topologically Sorted Source Nodes: [new_cache], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# new_cache => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%permute_4, %permute_5], -1), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[512],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 512
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4) % 2
x0 = xindex % 4
x2 = (xindex // 8)
x3 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (4*x2)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 2, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + (x0 + (4*x2)), tmp6 & xmask, eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x3), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/a3/ca3c423eelt6wuklohapyjrb6r2xhk3je6u3r7sngu36te4ofgnb.py
# Topologically Sorted Source Nodes: [matmul], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# matmul => clone
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%expand,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_1 = async_compile.triton('triton_poi_fused_clone_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 16], tile_hint=TileHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = (yindex // 4)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x2) + (64*y1)), xmask & ymask)
tmp1 = tl.load(in_ptr1 + (y0), ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (x2 + (16*y3)), tmp2, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/7a/c7a5kf2lbc7npgaim3dwx342dxc6rsfexsi3o3hjbv7uivdat2sc.py
# Topologically Sorted Source Nodes: [matmul], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# matmul => clone_1
# Graph fragment:
# %clone_1 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%expand_1,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_2 = async_compile.triton('triton_poi_fused_clone_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 16], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = (yindex // 4)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x2) + (64*y1)), xmask & ymask)
tl.store(out_ptr0 + (x2 + (16*y3)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/uz/cuzyl6q2skpu5xk7xyc3peb6slmlfoalyna7ahvmpx4oqhhop73b.py
# Topologically Sorted Source Nodes: [attn], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# attn => div_1, exp, sum_1
# Graph fragment:
# %mul_tensor : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_11, 1), kwargs = {})
# %amax_default : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%mul_tensor, [-1], True), kwargs = {})
# %sub_tensor : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_tensor, %amax_default), kwargs = {})
# %div_tensor : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub_tensor, 1.0), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%div_tensor,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [-1], True), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_per_fused__softmax_3 = async_compile.triton('triton_per_fused__softmax_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[256, 16],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused__softmax_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 2, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused__softmax_3(in_ptr0, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 256
rnumel = 16
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (16*x0)), xmask, other=0.0)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp3 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5 = tl.where(xmask, tmp3, float("-inf"))
tmp6 = triton_helpers.max2(tmp5, 1)[:, None]
tmp7 = tmp2 - tmp6
tmp8 = tmp7 * tmp1
tmp9 = tl_math.exp(tmp8)
tmp10 = tl.broadcast_to(tmp9, [XBLOCK, RBLOCK])
tmp12 = tl.where(xmask, tmp10, 0)
tmp13 = tl.sum(tmp12, 1)[:, None]
tmp14 = tmp9 / tmp13
tl.store(out_ptr2 + (r1 + (16*x0)), tmp14, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/5j/c5jbbs6rjuscr2km33ndvlmtkgcup7curz3fm3tk7stvjquhtikm.py
# Topologically Sorted Source Nodes: [contiguous], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# contiguous => clone_4
# Graph fragment:
# %clone_4 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_7,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_4 = async_compile.triton('triton_poi_fused_clone_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64, 4], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_4(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 64
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = (yindex // 16)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (16*x2) + (64*y1)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + (4*y3)), tmp0, xmask & ymask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (4, ), (1, ))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4, ), (1, ))
assert_size_stride(primals_9, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_10, (4, 4), (4, 1))
assert_size_stride(primals_11, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_1, (64, 4), (4, 1), 0), reinterpret_tensor(primals_2, (4, 4), (1, 4), 0), out=buf0)
del primals_2
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear_1], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_5, reinterpret_tensor(primals_6, (64, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf1)
del primals_4
del primals_5
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear_2], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_9, (64, 4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_7
del primals_8
buf3 = empty_strided_cuda((4, 4, 16, 2), (128, 1, 8, 4), torch.float32)
# Topologically Sorted Source Nodes: [new_cache], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(buf1, buf2, buf3, 512, grid=grid(512), stream=stream0)
buf4 = empty_strided_cuda((4, 4, 16, 1), (64, 16, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [matmul], Original ATen: [aten.clone]
triton_poi_fused_clone_1.run(buf0, primals_3, buf4, 16, 16, grid=grid(16, 16), stream=stream0)
del primals_3
buf5 = reinterpret_tensor(buf0, (4, 4, 1, 16), (64, 16, 16, 1), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [matmul], Original ATen: [aten.clone]
triton_poi_fused_clone_2.run(buf1, buf5, 16, 16, grid=grid(16, 16), stream=stream0)
buf6 = empty_strided_cuda((16, 16, 16), (256, 16, 1), torch.float32)
# Topologically Sorted Source Nodes: [matmul], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf4, (16, 16, 1), (16, 1, 0), 0), reinterpret_tensor(buf5, (16, 1, 16), (16, 0, 1), 0), out=buf6)
buf9 = empty_strided_cuda((4, 4, 16, 16), (1024, 256, 16, 1), torch.float32)
# Topologically Sorted Source Nodes: [attn], Original ATen: [aten._softmax]
triton_per_fused__softmax_3.run(buf6, buf9, 256, 16, grid=grid(256), stream=stream0)
del buf6
buf10 = reinterpret_tensor(buf1, (4, 4, 16, 1), (64, 16, 1, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.clone]
triton_poi_fused_clone_2.run(buf2, buf10, 16, 16, grid=grid(16, 16), stream=stream0)
buf11 = reinterpret_tensor(buf2, (16, 16, 1), (16, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf9, (16, 16, 16), (256, 16, 1), 0), reinterpret_tensor(buf10, (16, 16, 1), (16, 1, 0), 0), out=buf11)
buf12 = empty_strided_cuda((4, 16, 4, 1), (64, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [contiguous], Original ATen: [aten.clone]
triton_poi_fused_clone_4.run(buf11, buf12, 64, 4, grid=grid(64, 4), stream=stream0)
buf13 = reinterpret_tensor(buf11, (64, 4), (4, 1), 0); del buf11 # reuse
# Topologically Sorted Source Nodes: [linear_3], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_11, reinterpret_tensor(buf12, (64, 4), (4, 1), 0), reinterpret_tensor(primals_10, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf13)
del primals_11
return (reinterpret_tensor(buf13, (4, 16, 4), (64, 4, 1), 0), buf3, reinterpret_tensor(primals_1, (64, 4), (4, 1), 0), reinterpret_tensor(primals_6, (64, 4), (4, 1), 0), reinterpret_tensor(primals_9, (64, 4), (4, 1), 0), buf9, reinterpret_tensor(buf12, (64, 4), (4, 1), 0), primals_10, reinterpret_tensor(buf10, (16, 1, 16), (16, 1, 1), 0), reinterpret_tensor(buf4, (16, 1, 16), (16, 1, 1), 0), reinterpret_tensor(buf5, (16, 16, 1), (16, 1, 16), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
from typing import Tuple
from torch import nn
class MultiHeadedAttention(nn.Module):
"""Multi-Head Attention layer.
Args:
n_head (int): The number of heads.
n_feat (int): The number of features.
dropout_rate (float): Dropout rate.
"""
def __init__(self, n_head: 'int', n_feat: 'int', dropout_rate: 'float'):
"""Construct an MultiHeadedAttention object."""
super().__init__()
assert n_feat % n_head == 0
self.d_k = n_feat // n_head
self.h = n_head
self.linear_q = nn.Linear(n_feat, n_feat)
self.linear_k = nn.Linear(n_feat, n_feat)
self.linear_v = nn.Linear(n_feat, n_feat)
self.linear_out = nn.Linear(n_feat, n_feat)
self.dropout = nn.Dropout(p=dropout_rate)
def forward_qkv(self, query: 'torch.Tensor', key: 'torch.Tensor', value:
'torch.Tensor') ->Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
"""Transform query, key and value.
Args:
query (torch.Tensor): Query tensor (#batch, time1, size).
key (torch.Tensor): Key tensor (#batch, time2, size).
value (torch.Tensor): Value tensor (#batch, time2, size).
Returns:
torch.Tensor: Transformed query tensor, size
(#batch, n_head, time1, d_k).
torch.Tensor: Transformed key tensor, size
(#batch, n_head, time2, d_k).
torch.Tensor: Transformed value tensor, size
(#batch, n_head, time2, d_k).
"""
n_batch = query.size(0)
q = self.linear_q(query).view(n_batch, -1, self.h, self.d_k)
k = self.linear_k(key).view(n_batch, -1, self.h, self.d_k)
v = self.linear_v(value).view(n_batch, -1, self.h, self.d_k)
q = q.transpose(1, 2)
k = k.transpose(1, 2)
v = v.transpose(1, 2)
return q, k, v
def forward_attention(self, value: 'torch.Tensor', scores:
'torch.Tensor', mask: 'torch.Tensor'=torch.ones((0, 0, 0), dtype=
torch.bool)) ->torch.Tensor:
"""Compute attention context vector.
Args:
value (torch.Tensor): Transformed value, size
(#batch, n_head, time2, d_k).
scores (torch.Tensor): Attention score, size
(#batch, n_head, time1, time2).
mask (torch.Tensor): Mask, size (#batch, 1, time2) or
(#batch, time1, time2), (0, 0, 0) means fake mask.
Returns:
torch.Tensor: Transformed value (#batch, time1, d_model)
weighted by the attention score (#batch, time1, time2).
"""
n_batch = value.size(0)
if mask.size(2) > 0:
mask = mask.unsqueeze(1).eq(0)
mask = mask[:, :, :, :scores.size(-1)]
scores = scores.masked_fill(mask, -float('inf'))
attn = torch.softmax(scores, dim=-1).masked_fill(mask, 0.0)
else:
attn = torch.softmax(scores, dim=-1)
p_attn = self.dropout(attn)
x = torch.matmul(p_attn, value)
x = x.transpose(1, 2).contiguous().view(n_batch, -1, self.h * self.d_k)
return self.linear_out(x)
def forward(self, query: 'torch.Tensor', key: 'torch.Tensor', value:
'torch.Tensor', mask: 'torch.Tensor'=torch.ones((0, 0, 0), dtype=
torch.bool), pos_emb: 'torch.Tensor'=torch.empty(0), cache:
'torch.Tensor'=torch.zeros((0, 0, 0, 0))) ->Tuple[torch.Tensor,
torch.Tensor]:
"""Compute scaled dot product attention.
Args:
query (torch.Tensor): Query tensor (#batch, time1, size).
key (torch.Tensor): Key tensor (#batch, time2, size).
value (torch.Tensor): Value tensor (#batch, time2, size).
mask (torch.Tensor): Mask tensor (#batch, 1, time2) or
(#batch, time1, time2).
1.When applying cross attention between decoder and encoder,
the batch padding mask for input is in (#batch, 1, T) shape.
2.When applying self attention of encoder,
the mask is in (#batch, T, T) shape.
3.When applying self attention of decoder,
the mask is in (#batch, L, L) shape.
4.If the different position in decoder see different block
of the encoder, such as Mocha, the passed in mask could be
in (#batch, L, T) shape. But there is no such case in current
Wenet.
cache (torch.Tensor): Cache tensor (1, head, cache_t, d_k * 2),
where `cache_t == chunk_size * num_decoding_left_chunks`
and `head * d_k == size`
Returns:
torch.Tensor: Output tensor (#batch, time1, d_model).
torch.Tensor: Cache tensor (1, head, cache_t + time1, d_k * 2)
where `cache_t == chunk_size * num_decoding_left_chunks`
and `head * d_k == size`
"""
q, k, v = self.forward_qkv(query, key, value)
if cache.size(0) > 0:
key_cache, value_cache = torch.split(cache, cache.size(-1) // 2,
dim=-1)
k = torch.cat([key_cache, k], dim=2)
v = torch.cat([value_cache, v], dim=2)
new_cache = torch.cat((k, v), dim=-1)
scores = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(self.d_k)
return self.forward_attention(v, scores, mask), new_cache
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4]), torch.rand(
[4, 4, 4, 4])]
def get_init_inputs():
return [[], {'n_head': 4, 'n_feat': 4, 'dropout_rate': 0.5}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
from typing import Tuple
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 512
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4 % 2
x0 = xindex % 4
x2 = xindex // 8
x3 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 1, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 4 * x2), tmp4 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 2, tl.int64)
tmp9 = tl.load(in_ptr1 + (x0 + 4 * x2), tmp6 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x3, tmp10, xmask)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel,
YBLOCK: tl.constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = yindex // 4
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x2 + 64 * y1), xmask & ymask)
tmp1 = tl.load(in_ptr1 + y0, ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (x2 + 16 * y3), tmp2, xmask & ymask)
@triton.jit
def triton_poi_fused_clone_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = yindex // 4
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x2 + 64 * y1), xmask & ymask)
tl.store(out_ptr0 + (x2 + 16 * y3), tmp0, xmask & ymask)
@triton.jit
def triton_per_fused__softmax_3(in_ptr0, out_ptr2, xnumel, rnumel, XBLOCK:
tl.constexpr):
xnumel = 256
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 16 * x0), xmask, other=0.0)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp3 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5 = tl.where(xmask, tmp3, float('-inf'))
tmp6 = triton_helpers.max2(tmp5, 1)[:, None]
tmp7 = tmp2 - tmp6
tmp8 = tmp7 * tmp1
tmp9 = tl_math.exp(tmp8)
tmp10 = tl.broadcast_to(tmp9, [XBLOCK, RBLOCK])
tmp12 = tl.where(xmask, tmp10, 0)
tmp13 = tl.sum(tmp12, 1)[:, None]
tmp14 = tmp9 / tmp13
tl.store(out_ptr2 + (r1 + 16 * x0), tmp14, xmask)
@triton.jit
def triton_poi_fused_clone_4(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 64
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = yindex // 16
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 16 * x2 + 64 * y1), xmask & ymask,
eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + 4 * y3), tmp0, xmask & ymask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11) = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (4,), (1,))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (4,), (1,))
assert_size_stride(primals_9, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_10, (4, 4), (4, 1))
assert_size_stride(primals_11, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_1, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_2, (4, 4), (1, 4), 0), out=buf0)
del primals_2
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_5, reinterpret_tensor(primals_6, (64,
4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf1)
del primals_4
del primals_5
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_8, reinterpret_tensor(primals_9, (64,
4), (4, 1), 0), reinterpret_tensor(primals_7, (4, 4), (1, 4), 0
), alpha=1, beta=1, out=buf2)
del primals_7
del primals_8
buf3 = empty_strided_cuda((4, 4, 16, 2), (128, 1, 8, 4), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(512)](buf1, buf2, buf3, 512, XBLOCK=256,
num_warps=4, num_stages=1)
buf4 = empty_strided_cuda((4, 4, 16, 1), (64, 16, 1, 1), torch.float32)
triton_poi_fused_clone_1[grid(16, 16)](buf0, primals_3, buf4, 16,
16, XBLOCK=16, YBLOCK=16, num_warps=4, num_stages=1)
del primals_3
buf5 = reinterpret_tensor(buf0, (4, 4, 1, 16), (64, 16, 16, 1), 0)
del buf0
triton_poi_fused_clone_2[grid(16, 16)](buf1, buf5, 16, 16, XBLOCK=
16, YBLOCK=16, num_warps=4, num_stages=1)
buf6 = empty_strided_cuda((16, 16, 16), (256, 16, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf4, (16, 16, 1), (16, 1, 0),
0), reinterpret_tensor(buf5, (16, 1, 16), (16, 0, 1), 0), out=buf6)
buf9 = empty_strided_cuda((4, 4, 16, 16), (1024, 256, 16, 1), torch
.float32)
triton_per_fused__softmax_3[grid(256)](buf6, buf9, 256, 16, XBLOCK=
32, num_warps=4, num_stages=1)
del buf6
buf10 = reinterpret_tensor(buf1, (4, 4, 16, 1), (64, 16, 1, 1), 0)
del buf1
triton_poi_fused_clone_2[grid(16, 16)](buf2, buf10, 16, 16, XBLOCK=
16, YBLOCK=16, num_warps=4, num_stages=1)
buf11 = reinterpret_tensor(buf2, (16, 16, 1), (16, 1, 1), 0)
del buf2
extern_kernels.bmm(reinterpret_tensor(buf9, (16, 16, 16), (256, 16,
1), 0), reinterpret_tensor(buf10, (16, 16, 1), (16, 1, 0), 0),
out=buf11)
buf12 = empty_strided_cuda((4, 16, 4, 1), (64, 4, 1, 1), torch.float32)
triton_poi_fused_clone_4[grid(64, 4)](buf11, buf12, 64, 4, XBLOCK=4,
YBLOCK=32, num_warps=4, num_stages=1)
buf13 = reinterpret_tensor(buf11, (64, 4), (4, 1), 0)
del buf11
extern_kernels.addmm(primals_11, reinterpret_tensor(buf12, (64, 4),
(4, 1), 0), reinterpret_tensor(primals_10, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf13)
del primals_11
return reinterpret_tensor(buf13, (4, 16, 4), (64, 4, 1), 0
), buf3, reinterpret_tensor(primals_1, (64, 4), (4, 1), 0
), reinterpret_tensor(primals_6, (64, 4), (4, 1), 0
), reinterpret_tensor(primals_9, (64, 4), (4, 1), 0
), buf9, reinterpret_tensor(buf12, (64, 4), (4, 1), 0
), primals_10, reinterpret_tensor(buf10, (16, 1, 16), (16, 1, 1), 0
), reinterpret_tensor(buf4, (16, 1, 16), (16, 1, 1), 0
), reinterpret_tensor(buf5, (16, 16, 1), (16, 1, 16), 0)
class MultiHeadedAttentionNew(nn.Module):
"""Multi-Head Attention layer.
Args:
n_head (int): The number of heads.
n_feat (int): The number of features.
dropout_rate (float): Dropout rate.
"""
def __init__(self, n_head: 'int', n_feat: 'int', dropout_rate: 'float'):
"""Construct an MultiHeadedAttention object."""
super().__init__()
assert n_feat % n_head == 0
self.d_k = n_feat // n_head
self.h = n_head
self.linear_q = nn.Linear(n_feat, n_feat)
self.linear_k = nn.Linear(n_feat, n_feat)
self.linear_v = nn.Linear(n_feat, n_feat)
self.linear_out = nn.Linear(n_feat, n_feat)
self.dropout = nn.Dropout(p=dropout_rate)
def forward_qkv(self, query: 'torch.Tensor', key: 'torch.Tensor', value:
'torch.Tensor') ->Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
"""Transform query, key and value.
Args:
query (torch.Tensor): Query tensor (#batch, time1, size).
key (torch.Tensor): Key tensor (#batch, time2, size).
value (torch.Tensor): Value tensor (#batch, time2, size).
Returns:
torch.Tensor: Transformed query tensor, size
(#batch, n_head, time1, d_k).
torch.Tensor: Transformed key tensor, size
(#batch, n_head, time2, d_k).
torch.Tensor: Transformed value tensor, size
(#batch, n_head, time2, d_k).
"""
n_batch = query.size(0)
q = self.linear_q(query).view(n_batch, -1, self.h, self.d_k)
k = self.linear_k(key).view(n_batch, -1, self.h, self.d_k)
v = self.linear_v(value).view(n_batch, -1, self.h, self.d_k)
q = q.transpose(1, 2)
k = k.transpose(1, 2)
v = v.transpose(1, 2)
return q, k, v
def forward_attention(self, value: 'torch.Tensor', scores:
'torch.Tensor', mask: 'torch.Tensor'=torch.ones((0, 0, 0), dtype=
torch.bool)) ->torch.Tensor:
"""Compute attention context vector.
Args:
value (torch.Tensor): Transformed value, size
(#batch, n_head, time2, d_k).
scores (torch.Tensor): Attention score, size
(#batch, n_head, time1, time2).
mask (torch.Tensor): Mask, size (#batch, 1, time2) or
(#batch, time1, time2), (0, 0, 0) means fake mask.
Returns:
torch.Tensor: Transformed value (#batch, time1, d_model)
weighted by the attention score (#batch, time1, time2).
"""
n_batch = value.size(0)
if mask.size(2) > 0:
mask = mask.unsqueeze(1).eq(0)
mask = mask[:, :, :, :scores.size(-1)]
scores = scores.masked_fill(mask, -float('inf'))
attn = torch.softmax(scores, dim=-1).masked_fill(mask, 0.0)
else:
attn = torch.softmax(scores, dim=-1)
p_attn = self.dropout(attn)
x = torch.matmul(p_attn, value)
x = x.transpose(1, 2).contiguous().view(n_batch, -1, self.h * self.d_k)
return self.linear_out(x)
def forward(self, input_0, input_1, input_2):
primals_2 = self.linear_q.weight
primals_3 = self.linear_q.bias
primals_4 = self.linear_k.weight
primals_5 = self.linear_k.bias
primals_7 = self.linear_v.weight
primals_8 = self.linear_v.bias
primals_10 = self.linear_out.weight
primals_11 = self.linear_out.bias
primals_1 = input_0
primals_6 = input_1
primals_9 = input_2
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11])
return output[0], output[1]
|
eschmidbauer/wenet
|
MultiHeadedAttention
| false | 12,365 |
[
"Apache-2.0"
] | 0 |
f0bbf6af16fa92d26a7f68ac21e0354a7500a025
|
https://github.com/eschmidbauer/wenet/tree/f0bbf6af16fa92d26a7f68ac21e0354a7500a025
|
core_network
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/sf/csfkqsknybmr6wr4qghsq2q5fkm44yw64vallsxwzclanv6zrtso.py
# Topologically Sorted Source Nodes: [add, h_t], Original ATen: [aten.add, aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# add => add
# h_t => relu
# Graph fragment:
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%view_1, %view_3), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%add,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_add_relu_threshold_backward_0 = async_compile.triton('triton_poi_fused_add_relu_threshold_backward_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*i1', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_relu_threshold_backward_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_0(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + (x2), xmask)
tmp4 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 + tmp5
tmp7 = tl.full([1], 0, tl.int32)
tmp8 = triton_helpers.maximum(tmp7, tmp6)
tmp9 = 0.0
tmp10 = tmp8 <= tmp9
tl.store(in_out_ptr0 + (x2), tmp8, xmask)
tl.store(out_ptr0 + (x2), tmp10, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_6, (64, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf1)
del primals_4
buf2 = reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf0 # reuse
buf3 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [add, h_t], Original ATen: [aten.add, aten.relu, aten.threshold_backward]
stream0 = get_raw_stream(0)
triton_poi_fused_add_relu_threshold_backward_0.run(buf2, primals_2, buf1, primals_5, buf3, 256, grid=grid(256), stream=stream0)
del buf1
del primals_2
del primals_5
return (buf2, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_6, (64, 4), (4, 1), 0), buf3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class core_network(nn.Module):
"""
An RNN that maintains an internal state that integrates
information extracted from the history of past observations.
It encodes the agent's knowledge of the environment through
a state vector `h_t` that gets updated at every time step `t`.
Concretely, it takes the glimpse representation `g_t` as input,
and combines it with its internal state `h_t_prev` at the previous
time step, to produce the new internal state `h_t` at the current
time step.
In other words:
`h_t = relu( fc(h_t_prev) + fc(g_t) )`
Args
----
- input_size: input size of the rnn.
- hidden_size: hidden size of the rnn.
- g_t: a 2D tensor of shape (B, hidden_size). The glimpse
representation returned by the glimpse network for the
current timestep `t`.
- h_t_prev: a 2D tensor of shape (B, hidden_size). The
hidden state vector for the previous timestep `t-1`.
Returns
-------
- h_t: a 2D tensor of shape (B, hidden_size). The hidden
state vector for the current timestep `t`.
"""
def __init__(self, input_size, hidden_size):
super(core_network, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.i2h = nn.Linear(input_size, hidden_size)
self.h2h = nn.Linear(hidden_size, hidden_size)
def forward(self, g_t, h_t_prev):
h1 = self.i2h(g_t)
h2 = self.h2h(h_t_prev)
h_t = F.relu(h1 + h2)
return h_t
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'input_size': 4, 'hidden_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_add_relu_threshold_backward_0(in_out_ptr0, in_ptr0,
in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + x2, xmask)
tmp4 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 + tmp5
tmp7 = tl.full([1], 0, tl.int32)
tmp8 = triton_helpers.maximum(tmp7, tmp6)
tmp9 = 0.0
tmp10 = tmp8 <= tmp9
tl.store(in_out_ptr0 + x2, tmp8, xmask)
tl.store(out_ptr0 + x2, tmp10, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_6, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf1)
del primals_4
buf2 = reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf0
buf3 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.bool)
get_raw_stream(0)
triton_poi_fused_add_relu_threshold_backward_0[grid(256)](buf2,
primals_2, buf1, primals_5, buf3, 256, XBLOCK=128, num_warps=4,
num_stages=1)
del buf1
del primals_2
del primals_5
return buf2, reinterpret_tensor(primals_3, (64, 4), (4, 1), 0
), reinterpret_tensor(primals_6, (64, 4), (4, 1), 0), buf3
class core_networkNew(nn.Module):
"""
An RNN that maintains an internal state that integrates
information extracted from the history of past observations.
It encodes the agent's knowledge of the environment through
a state vector `h_t` that gets updated at every time step `t`.
Concretely, it takes the glimpse representation `g_t` as input,
and combines it with its internal state `h_t_prev` at the previous
time step, to produce the new internal state `h_t` at the current
time step.
In other words:
`h_t = relu( fc(h_t_prev) + fc(g_t) )`
Args
----
- input_size: input size of the rnn.
- hidden_size: hidden size of the rnn.
- g_t: a 2D tensor of shape (B, hidden_size). The glimpse
representation returned by the glimpse network for the
current timestep `t`.
- h_t_prev: a 2D tensor of shape (B, hidden_size). The
hidden state vector for the previous timestep `t-1`.
Returns
-------
- h_t: a 2D tensor of shape (B, hidden_size). The hidden
state vector for the current timestep `t`.
"""
def __init__(self, input_size, hidden_size):
super(core_networkNew, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.i2h = nn.Linear(input_size, hidden_size)
self.h2h = nn.Linear(hidden_size, hidden_size)
def forward(self, input_0, input_1):
primals_1 = self.i2h.weight
primals_2 = self.i2h.bias
primals_4 = self.h2h.weight
primals_5 = self.h2h.bias
primals_3 = input_0
primals_6 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6])
return output[0]
|
felixnon/foveated-visual-attention
|
core_network
| false | 12,366 |
[
"MIT"
] | 0 |
7e7d9a5ef24ec42eb76ba72f783bb2227bdb4851
|
https://github.com/felixnon/foveated-visual-attention/tree/7e7d9a5ef24ec42eb76ba72f783bb2227bdb4851
|
PositionGenerator
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/df/cdfcie57v6pcdd6oeaz4mvlgksxgyuxzmlv5bklwemyulqhtcxta.py
# Topologically Sorted Source Nodes: [mean, std, sub, mul, add, truediv, add_1], Original ATen: [aten.mean, aten.std, aten.sub, aten.mul, aten.add, aten.div]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# mean => mean
# mul => mul
# std => sqrt, var
# sub => sub
# truediv => div
# Graph fragment:
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%primals_2, [-1], True), kwargs = {})
# %var : [num_users=1] = call_function[target=torch.ops.aten.var.correction](args = (%primals_2, [-1]), kwargs = {correction: 1.0, keepdim: True})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%var,), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%primals_2, %mean), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_3, %sub), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sqrt, 1e-06), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%mul, %add), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%div, %primals_4), kwargs = {})
triton_poi_fused_add_div_mean_mul_std_sub_0 = async_compile.triton('triton_poi_fused_add_div_mean_mul_std_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_mean_mul_std_sub_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 7, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_mean_mul_std_sub_0(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp2 = tl.load(in_ptr1 + (4*x1), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr1 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp30 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp8 = tmp6 + tmp7
tmp9 = 4.0
tmp10 = tmp8 / tmp9
tmp11 = tmp1 - tmp10
tmp12 = tmp0 * tmp11
tmp13 = tmp2 - tmp10
tmp14 = tmp13 * tmp13
tmp15 = tmp3 - tmp10
tmp16 = tmp15 * tmp15
tmp17 = tmp14 + tmp16
tmp18 = tmp5 - tmp10
tmp19 = tmp18 * tmp18
tmp20 = tmp17 + tmp19
tmp21 = tmp7 - tmp10
tmp22 = tmp21 * tmp21
tmp23 = tmp20 + tmp22
tmp24 = 3.0
tmp25 = tmp23 / tmp24
tmp26 = libdevice.sqrt(tmp25)
tmp27 = 1e-06
tmp28 = tmp26 + tmp27
tmp29 = tmp12 / tmp28
tmp31 = tmp29 + tmp30
tl.store(out_ptr0 + (x2), tmp31, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/tt/ctttzgnguystw446sgeisgiema54yz5d5wmngsz4pinf7chzr4i7.py
# Topologically Sorted Source Nodes: [out_masked], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# out_masked => mul_1
# Graph fragment:
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_1, %unsqueeze), kwargs = {})
triton_poi_fused_mul_1 = async_compile.triton('triton_poi_fused_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex % 256
x4 = (xindex // 4)
x5 = xindex
tmp0 = tl.load(in_ptr0 + (x3), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x4), xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x5), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4, ), (1, ))
assert_size_stride(primals_4, (4, ), (1, ))
assert_size_stride(primals_5, (3, 4), (4, 1))
assert_size_stride(primals_6, (3, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mean, std, sub, mul, add, truediv, add_1], Original ATen: [aten.mean, aten.std, aten.sub, aten.mul, aten.add, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_mean_mul_std_sub_0.run(primals_3, primals_2, primals_4, buf0, 256, grid=grid(256), stream=stream0)
del primals_3
del primals_4
buf1 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [out_masked], Original ATen: [aten.mul]
triton_poi_fused_mul_1.run(buf0, primals_1, buf1, 1024, grid=grid(1024), stream=stream0)
del buf0
buf2 = empty_strided_cuda((256, 3), (3, 1), torch.float32)
# Topologically Sorted Source Nodes: [projected], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_6, reinterpret_tensor(buf1, (256, 4), (4, 1), 0), reinterpret_tensor(primals_5, (4, 3), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_6
return (reinterpret_tensor(buf2, (4, 4, 4, 4, 3), (192, 48, 12, 3, 1), 0), primals_1, primals_2, reinterpret_tensor(buf1, (256, 4), (4, 1), 0), primals_5, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((3, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((3, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class LayerNorm(nn.Module):
"""Construct a layernorm module (See citation for details)."""
def __init__(self, features, eps=1e-06):
super(LayerNorm, self).__init__()
self.a_2 = nn.Parameter(torch.ones(features))
self.b_2 = nn.Parameter(torch.zeros(features))
self.eps = eps
def forward(self, x):
mean = x.mean(-1, keepdim=True)
std = x.std(-1, keepdim=True)
return self.a_2 * (x - mean) / (std + self.eps) + self.b_2
class PositionGenerator(nn.Module):
"""Define standard linear + softmax generation step."""
def __init__(self, d_model):
super(PositionGenerator, self).__init__()
self.norm = LayerNorm(d_model)
self.proj = nn.Linear(d_model, 3)
def forward(self, x, mask):
mask = mask.unsqueeze(-1).float()
out_masked = self.norm(x) * mask
projected = self.proj(out_masked)
return projected
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'d_model': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_add_div_mean_mul_std_sub_0(in_ptr0, in_ptr1, in_ptr2,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp2 = tl.load(in_ptr1 + 4 * x1, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr1 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp30 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp8 = tmp6 + tmp7
tmp9 = 4.0
tmp10 = tmp8 / tmp9
tmp11 = tmp1 - tmp10
tmp12 = tmp0 * tmp11
tmp13 = tmp2 - tmp10
tmp14 = tmp13 * tmp13
tmp15 = tmp3 - tmp10
tmp16 = tmp15 * tmp15
tmp17 = tmp14 + tmp16
tmp18 = tmp5 - tmp10
tmp19 = tmp18 * tmp18
tmp20 = tmp17 + tmp19
tmp21 = tmp7 - tmp10
tmp22 = tmp21 * tmp21
tmp23 = tmp20 + tmp22
tmp24 = 3.0
tmp25 = tmp23 / tmp24
tmp26 = libdevice.sqrt(tmp25)
tmp27 = 1e-06
tmp28 = tmp26 + tmp27
tmp29 = tmp12 / tmp28
tmp31 = tmp29 + tmp30
tl.store(out_ptr0 + x2, tmp31, xmask)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex % 256
x4 = xindex // 4
x5 = xindex
tmp0 = tl.load(in_ptr0 + x3, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + x4, xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x5, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4,), (1,))
assert_size_stride(primals_4, (4,), (1,))
assert_size_stride(primals_5, (3, 4), (4, 1))
assert_size_stride(primals_6, (3,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_mean_mul_std_sub_0[grid(256)](primals_3,
primals_2, primals_4, buf0, 256, XBLOCK=128, num_warps=4,
num_stages=1)
del primals_3
del primals_4
buf1 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1),
torch.float32)
triton_poi_fused_mul_1[grid(1024)](buf0, primals_1, buf1, 1024,
XBLOCK=128, num_warps=4, num_stages=1)
del buf0
buf2 = empty_strided_cuda((256, 3), (3, 1), torch.float32)
extern_kernels.addmm(primals_6, reinterpret_tensor(buf1, (256, 4),
(4, 1), 0), reinterpret_tensor(primals_5, (4, 3), (1, 4), 0),
alpha=1, beta=1, out=buf2)
del primals_6
return reinterpret_tensor(buf2, (4, 4, 4, 4, 3), (192, 48, 12, 3, 1), 0
), primals_1, primals_2, reinterpret_tensor(buf1, (256, 4), (4, 1), 0
), primals_5
class LayerNorm(nn.Module):
"""Construct a layernorm module (See citation for details)."""
def __init__(self, features, eps=1e-06):
super(LayerNorm, self).__init__()
self.a_2 = nn.Parameter(torch.ones(features))
self.b_2 = nn.Parameter(torch.zeros(features))
self.eps = eps
def forward(self, x):
mean = x.mean(-1, keepdim=True)
std = x.std(-1, keepdim=True)
return self.a_2 * (x - mean) / (std + self.eps) + self.b_2
class PositionGeneratorNew(nn.Module):
"""Define standard linear + softmax generation step."""
def __init__(self, d_model):
super(PositionGeneratorNew, self).__init__()
self.norm = LayerNorm(d_model)
self.proj = nn.Linear(d_model, 3)
def forward(self, input_0, input_1):
primals_3 = self.norm.a_2
primals_4 = self.norm.b_2
primals_5 = self.proj.weight
primals_6 = self.proj.bias
primals_1 = input_0
primals_2 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6])
return output[0]
|
eweiner/MAT_Extension
|
PositionGenerator
| false | 12,367 |
[
"MIT"
] | 0 |
505884a67f97bf54e1198077d15a48531fcac7a5
|
https://github.com/eweiner/MAT_Extension/tree/505884a67f97bf54e1198077d15a48531fcac7a5
|
CausalConv2d
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/2a/c2aucoxj5ek5nl3pl5n67fq4vdl55x2rw66tvv2aq23fsb3jvf5q.py
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.constant_pad_nd]
# Source node to ATen node mapping:
# out => constant_pad_nd
# Graph fragment:
# %constant_pad_nd : [num_users=2] = call_function[target=torch.ops.aten.constant_pad_nd.default](args = (%primals_1, [3, 0, 3, 0], 0.0), kwargs = {})
triton_poi_fused_constant_pad_nd_0 = async_compile.triton('triton_poi_fused_constant_pad_nd_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_constant_pad_nd_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_constant_pad_nd_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 784
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 7) % 7
x0 = xindex % 7
x2 = (xindex // 49)
x4 = xindex
tmp0 = (-3) + x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = (-3) + x0
tmp4 = tmp3 >= tmp1
tmp5 = tmp2 & tmp4
tmp6 = tl.load(in_ptr0 + ((-15) + x0 + (4*x1) + (16*x2)), tmp5 & xmask, other=0.0)
tl.store(out_ptr0 + (x4), tmp6, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6r/c6rjd3rxthhw2ub6d2gtgjazrjeoyhalmj36ujwzgwexsis27e73.py
# Topologically Sorted Source Nodes: [_weight_norm], Original ATen: [aten._weight_norm_interface]
# Source node to ATen node mapping:
# _weight_norm => div, mul, pow_1, pow_2, sum_1
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_3, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [1, 2, 3], True), kwargs = {})
# %pow_2 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_1, 0.5), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_2, %pow_2), kwargs = {})
# %mul : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_3, %div), kwargs = {})
triton_per_fused__weight_norm_interface_1 = async_compile.triton('triton_per_fused__weight_norm_interface_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[4, 64],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused__weight_norm_interface_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused__weight_norm_interface_1(in_out_ptr0, in_ptr0, in_ptr1, out_ptr0, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 4
rnumel = 64
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (64*x0)), xmask, other=0.0)
tmp7 = tl.load(in_ptr1 + (x0), xmask, eviction_policy='evict_last')
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.where(xmask, tmp2, 0)
tmp5 = tl.sum(tmp4, 1)[:, None]
tmp6 = libdevice.sqrt(tmp5)
tmp8 = tmp7 / tmp6
tmp9 = tmp0 * tmp8
tl.debug_barrier()
tl.store(in_out_ptr0 + (x0), tmp6, xmask)
tl.store(out_ptr0 + (r1 + (64*x0)), tmp9, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/w5/cw5gytijzzkwnfpq2a2axdsj4pfxgxmwiuzizuyd4bw5uwnanzw7.py
# Topologically Sorted Source Nodes: [out_1], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# out_1 => convolution
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%constant_pad_nd, %mul, %primals_4, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_2 = async_compile.triton('triton_poi_fused_convolution_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_2', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_2(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 16) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 1, 1, 1), (1, 1, 1, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 7, 7), (196, 49, 7, 1), torch.float32)
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.constant_pad_nd]
stream0 = get_raw_stream(0)
triton_poi_fused_constant_pad_nd_0.run(primals_1, buf0, 784, grid=grid(784), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((4, 1, 1, 1), (1, 4, 4, 4), torch.float32)
buf2 = reinterpret_tensor(buf1, (4, 1, 1, 1), (1, 1, 1, 1), 0); del buf1 # reuse
buf3 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [_weight_norm], Original ATen: [aten._weight_norm_interface]
triton_per_fused__weight_norm_interface_1.run(buf2, primals_3, primals_2, buf3, 4, 64, grid=grid(4), stream=stream0)
# Topologically Sorted Source Nodes: [out_1], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(buf0, buf3, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 4, 4), (64, 16, 4, 1))
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [out_1], Original ATen: [aten.convolution]
triton_poi_fused_convolution_2.run(buf5, primals_4, 256, grid=grid(256), stream=stream0)
del primals_4
return (buf5, buf3, primals_2, primals_3, buf0, buf2, buf3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 1, 1, 1), (1, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
class WNConv2d(nn.Module):
def __init__(self, in_channel, out_channel, kernel_size, stride=1,
padding=0, bias=True, activation=None):
super().__init__()
self.conv = nn.utils.weight_norm(nn.Conv2d(in_channel, out_channel,
kernel_size, stride=stride, padding=padding, bias=bias))
self.out_channel = out_channel
if isinstance(kernel_size, int):
kernel_size = [kernel_size, kernel_size]
self.kernel_size = kernel_size
self.activation = activation
def forward(self, input):
out = self.conv(input)
if self.activation is not None:
out = self.activation(out)
return out
class CausalConv2d(nn.Module):
def __init__(self, in_channel, out_channel, kernel_size, stride=1,
padding='downright', activation=None):
super().__init__()
if isinstance(kernel_size, int):
kernel_size = [kernel_size] * 2
self.kernel_size = kernel_size
if padding == 'downright':
pad = [kernel_size[1] - 1, 0, kernel_size[0] - 1, 0]
elif padding == 'down' or padding == 'causal':
pad = kernel_size[1] // 2
pad = [pad, pad, kernel_size[0] - 1, 0]
self.causal = 0
if padding == 'causal':
self.causal = kernel_size[1] // 2
self.pad = nn.ZeroPad2d(pad)
self.conv = WNConv2d(in_channel, out_channel, kernel_size, stride=
stride, padding=0, activation=activation)
def forward(self, input):
out = self.pad(input)
if self.causal > 0:
self.conv.conv.weight_v.data[:, :, -1, self.causal:].zero_()
out = self.conv(out)
return out
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_channel': 4, 'out_channel': 4, 'kernel_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_constant_pad_nd_0(in_ptr0, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 784
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 7 % 7
x0 = xindex % 7
x2 = xindex // 49
x4 = xindex
tmp0 = -3 + x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = -3 + x0
tmp4 = tmp3 >= tmp1
tmp5 = tmp2 & tmp4
tmp6 = tl.load(in_ptr0 + (-15 + x0 + 4 * x1 + 16 * x2), tmp5 & xmask,
other=0.0)
tl.store(out_ptr0 + x4, tmp6, xmask)
@triton.jit
def triton_per_fused__weight_norm_interface_1(in_out_ptr0, in_ptr0, in_ptr1,
out_ptr0, xnumel, rnumel, XBLOCK: tl.constexpr):
xnumel = 4
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 64 * x0), xmask, other=0.0)
tmp7 = tl.load(in_ptr1 + x0, xmask, eviction_policy='evict_last')
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.where(xmask, tmp2, 0)
tmp5 = tl.sum(tmp4, 1)[:, None]
tmp6 = libdevice.sqrt(tmp5)
tmp8 = tmp7 / tmp6
tmp9 = tmp0 * tmp8
tl.debug_barrier()
tl.store(in_out_ptr0 + x0, tmp6, xmask)
tl.store(out_ptr0 + (r1 + 64 * x0), tmp9, xmask)
@triton.jit
def triton_poi_fused_convolution_2(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 16 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 1, 1, 1), (1, 1, 1, 1))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 7, 7), (196, 49, 7, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_constant_pad_nd_0[grid(784)](primals_1, buf0, 784,
XBLOCK=256, num_warps=4, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((4, 1, 1, 1), (1, 4, 4, 4), torch.float32)
buf2 = reinterpret_tensor(buf1, (4, 1, 1, 1), (1, 1, 1, 1), 0)
del buf1
buf3 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_per_fused__weight_norm_interface_1[grid(4)](buf2, primals_3,
primals_2, buf3, 4, 64, XBLOCK=1, num_warps=2, num_stages=1)
buf4 = extern_kernels.convolution(buf0, buf3, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 4, 4), (64, 16, 4, 1))
buf5 = buf4
del buf4
triton_poi_fused_convolution_2[grid(256)](buf5, primals_4, 256,
XBLOCK=128, num_warps=4, num_stages=1)
del primals_4
return buf5, buf3, primals_2, primals_3, buf0, buf2, buf3
class WNConv2d(nn.Module):
def __init__(self, in_channel, out_channel, kernel_size, stride=1,
padding=0, bias=True, activation=None):
super().__init__()
self.conv = nn.utils.weight_norm(nn.Conv2d(in_channel, out_channel,
kernel_size, stride=stride, padding=padding, bias=bias))
self.out_channel = out_channel
if isinstance(kernel_size, int):
kernel_size = [kernel_size, kernel_size]
self.kernel_size = kernel_size
self.activation = activation
def forward(self, input):
out = self.conv(input)
if self.activation is not None:
out = self.activation(out)
return out
class CausalConv2dNew(nn.Module):
def __init__(self, in_channel, out_channel, kernel_size, stride=1,
padding='downright', activation=None):
super().__init__()
if isinstance(kernel_size, int):
kernel_size = [kernel_size] * 2
self.kernel_size = kernel_size
if padding == 'downright':
pad = [kernel_size[1] - 1, 0, kernel_size[0] - 1, 0]
elif padding == 'down' or padding == 'causal':
pad = kernel_size[1] // 2
pad = [pad, pad, kernel_size[0] - 1, 0]
self.causal = 0
if padding == 'causal':
self.causal = kernel_size[1] // 2
self.pad = nn.ZeroPad2d(pad)
self.conv = WNConv2d(in_channel, out_channel, kernel_size, stride=
stride, padding=0, activation=activation)
def forward(self, input_0):
primals_4 = self.conv.conv.bias
primals_2 = self.conv.conv.weight_g
primals_1 = self.conv.conv.weight_v
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
eric11220/vq-vae-2-pytorch
|
CausalConv2d
| false | 12,368 |
[
"MIT"
] | 0 |
ac455ec8873428e16a361d49bf1dda30472ece13
|
https://github.com/eric11220/vq-vae-2-pytorch/tree/ac455ec8873428e16a361d49bf1dda30472ece13
|
WNConv2d
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/4m/c4munax4bhqei64mhriszwqd42q6bjyh4nv3jazxhymu3f7wtucw.py
# Topologically Sorted Source Nodes: [_weight_norm], Original ATen: [aten._weight_norm_interface]
# Source node to ATen node mapping:
# _weight_norm => div, mul, pow_1, pow_2, sum_1
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%primals_2, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [1, 2, 3], True), kwargs = {})
# %pow_2 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_1, 0.5), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_1, %pow_2), kwargs = {})
# %mul : [num_users=2] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %div), kwargs = {})
triton_per_fused__weight_norm_interface_0 = async_compile.triton('triton_per_fused__weight_norm_interface_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[4, 64],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused__weight_norm_interface_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused__weight_norm_interface_0(in_out_ptr0, in_ptr0, in_ptr1, out_ptr0, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 4
rnumel = 64
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (64*x0)), xmask, other=0.0)
tmp7 = tl.load(in_ptr1 + (x0), xmask, eviction_policy='evict_last')
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.where(xmask, tmp2, 0)
tmp5 = tl.sum(tmp4, 1)[:, None]
tmp6 = libdevice.sqrt(tmp5)
tmp8 = tmp7 / tmp6
tmp9 = tmp0 * tmp8
tl.debug_barrier()
tl.store(in_out_ptr0 + (x0), tmp6, xmask)
tl.store(out_ptr0 + (r1 + (64*x0)), tmp9, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/tc/ctcagp37ljugm52zu6ckorigrppqo67voefe2f2odg5r6hyllhyu.py
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# out => convolution
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_4, %mul, %primals_3, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_1 = async_compile.triton('triton_poi_fused_convolution_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x2), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 1, 1, 1), (1, 1, 1, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4, ), (1, ))
assert_size_stride(primals_4, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 1, 1), (1, 4, 4, 4), torch.float32)
buf1 = reinterpret_tensor(buf0, (4, 1, 1, 1), (1, 1, 1, 1), 0); del buf0 # reuse
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [_weight_norm], Original ATen: [aten._weight_norm_interface]
stream0 = get_raw_stream(0)
triton_per_fused__weight_norm_interface_0.run(buf1, primals_2, primals_1, buf2, 4, 64, grid=grid(4), stream=stream0)
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
buf3 = extern_kernels.convolution(primals_4, buf2, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 4, 1, 1), (4, 1, 1, 1))
buf4 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.convolution]
triton_poi_fused_convolution_1.run(buf4, primals_3, 16, grid=grid(16), stream=stream0)
del primals_3
return (buf4, buf2, primals_1, primals_2, primals_4, buf1, buf2, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 1, 1, 1), (1, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
class WNConv2d(nn.Module):
def __init__(self, in_channel, out_channel, kernel_size, stride=1,
padding=0, bias=True, activation=None):
super().__init__()
self.conv = nn.utils.weight_norm(nn.Conv2d(in_channel, out_channel,
kernel_size, stride=stride, padding=padding, bias=bias))
self.out_channel = out_channel
if isinstance(kernel_size, int):
kernel_size = [kernel_size, kernel_size]
self.kernel_size = kernel_size
self.activation = activation
def forward(self, input):
out = self.conv(input)
if self.activation is not None:
out = self.activation(out)
return out
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_channel': 4, 'out_channel': 4, 'kernel_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_per_fused__weight_norm_interface_0(in_out_ptr0, in_ptr0, in_ptr1,
out_ptr0, xnumel, rnumel, XBLOCK: tl.constexpr):
xnumel = 4
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 64 * x0), xmask, other=0.0)
tmp7 = tl.load(in_ptr1 + x0, xmask, eviction_policy='evict_last')
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.where(xmask, tmp2, 0)
tmp5 = tl.sum(tmp4, 1)[:, None]
tmp6 = libdevice.sqrt(tmp5)
tmp8 = tmp7 / tmp6
tmp9 = tmp0 * tmp8
tl.debug_barrier()
tl.store(in_out_ptr0 + x0, tmp6, xmask)
tl.store(out_ptr0 + (r1 + 64 * x0), tmp9, xmask)
@triton.jit
def triton_poi_fused_convolution_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x2, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (4, 1, 1, 1), (1, 1, 1, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4,), (1,))
assert_size_stride(primals_4, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 1, 1), (1, 4, 4, 4), torch.float32)
buf1 = reinterpret_tensor(buf0, (4, 1, 1, 1), (1, 1, 1, 1), 0)
del buf0
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_per_fused__weight_norm_interface_0[grid(4)](buf1, primals_2,
primals_1, buf2, 4, 64, XBLOCK=1, num_warps=2, num_stages=1)
buf3 = extern_kernels.convolution(primals_4, buf2, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 4, 1, 1), (4, 1, 1, 1))
buf4 = buf3
del buf3
triton_poi_fused_convolution_1[grid(16)](buf4, primals_3, 16,
XBLOCK=16, num_warps=1, num_stages=1)
del primals_3
return buf4, buf2, primals_1, primals_2, primals_4, buf1, buf2
class WNConv2dNew(nn.Module):
def __init__(self, in_channel, out_channel, kernel_size, stride=1,
padding=0, bias=True, activation=None):
super().__init__()
self.conv = nn.utils.weight_norm(nn.Conv2d(in_channel, out_channel,
kernel_size, stride=stride, padding=padding, bias=bias))
self.out_channel = out_channel
if isinstance(kernel_size, int):
kernel_size = [kernel_size, kernel_size]
self.kernel_size = kernel_size
self.activation = activation
def forward(self, input_0):
primals_3 = self.conv.bias
primals_1 = self.conv.weight_g
primals_2 = self.conv.weight_v
primals_4 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
eric11220/vq-vae-2-pytorch
|
WNConv2d
| false | 12,369 |
[
"MIT"
] | 0 |
ac455ec8873428e16a361d49bf1dda30472ece13
|
https://github.com/eric11220/vq-vae-2-pytorch/tree/ac455ec8873428e16a361d49bf1dda30472ece13
|
MultiHeadAttention
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/dk/cdk4odz276xorciau5ehgl7f3s2mgkf3hrye6xep6kzubczdeqqy.py
# Topologically Sorted Source Nodes: [attention_scores], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# attention_scores => clone
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%expand,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_0 = async_compile.triton('triton_poi_fused_clone_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 4], tile_hint=TileHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_0(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = (yindex // 4)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x2) + (16*y1)), xmask & ymask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (y0), ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (x2 + (4*y3)), tmp2, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/fh/cfhydrwaejluqurspx2dvattjq4qiiexqog4gfsdmfob43r5rnk5.py
# Topologically Sorted Source Nodes: [attention_scores_1, attention_scores_2, attention_probs], Original ATen: [aten.div, aten.add, aten._softmax]
# Source node to ATen node mapping:
# attention_probs => amax, exp, sub, sum_1
# attention_scores_1 => div
# attention_scores_2 => add
# Graph fragment:
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%view_11, 1.0), kwargs = {})
# %add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%div, %primals_8), kwargs = {})
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%add, [-1], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [-1], True), kwargs = {})
triton_poi_fused__softmax_add_div_1 = async_compile.triton('triton_poi_fused__softmax_add_div_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_add_div_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_add_div_1(in_ptr0, in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 16
tmp0 = tl.load(in_ptr0 + (4*x2), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + (4*x0), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + (4*x2)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (2 + (4*x2)), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr1 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp15 = tl.load(in_ptr0 + (3 + (4*x2)), xmask, eviction_policy='evict_last')
tmp17 = tl.load(in_ptr1 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp5 * tmp1
tmp8 = tmp6 + tmp7
tmp9 = triton_helpers.maximum(tmp4, tmp8)
tmp11 = tmp10 * tmp1
tmp13 = tmp11 + tmp12
tmp14 = triton_helpers.maximum(tmp9, tmp13)
tmp16 = tmp15 * tmp1
tmp18 = tmp16 + tmp17
tmp19 = triton_helpers.maximum(tmp14, tmp18)
tmp20 = tmp4 - tmp19
tmp21 = tl_math.exp(tmp20)
tmp22 = tmp8 - tmp19
tmp23 = tl_math.exp(tmp22)
tmp24 = tmp21 + tmp23
tmp25 = tmp13 - tmp19
tmp26 = tl_math.exp(tmp25)
tmp27 = tmp24 + tmp26
tmp28 = tmp18 - tmp19
tmp29 = tl_math.exp(tmp28)
tmp30 = tmp27 + tmp29
tl.store(out_ptr0 + (x2), tmp19, xmask)
tl.store(out_ptr1 + (x2), tmp30, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/g6/cg65jkqmchbdahuikqawep5p32pz7fmy6cfvgzohspjy5l4iqhrs.py
# Topologically Sorted Source Nodes: [attention_scores_1, attention_scores_2, attention_probs], Original ATen: [aten.div, aten.add, aten._softmax]
# Source node to ATen node mapping:
# attention_probs => amax, div_1, exp, sub
# attention_scores_1 => div
# attention_scores_2 => add
# Graph fragment:
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%view_11, 1.0), kwargs = {})
# %add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%div, %primals_8), kwargs = {})
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%add, [-1], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_add_div_2 = async_compile.triton('triton_poi_fused__softmax_add_div_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_add_div_2', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_add_div_2(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x4 = xindex % 64
x5 = (xindex // 4)
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp3 = tl.load(in_ptr0 + (x4), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr1 + (x5), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr2 + (x5), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 - tmp5
tmp7 = tl_math.exp(tmp6)
tmp9 = tmp7 / tmp8
tl.store(in_out_ptr0 + (x3), tmp9, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/xt/cxtkkmujo4ytg6ycpz5lk5livtstr63pg5nsf5ijewjbtrfrqx6k.py
# Topologically Sorted Source Nodes: [context_layer_1], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# context_layer_1 => clone_4
# Graph fragment:
# %clone_4 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_6,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_3 = async_compile.triton('triton_poi_fused_clone_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 4], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_3(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = (yindex // 4)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x2) + (16*y1)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + (4*y3)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/t2/ct2fj4cxakdevxwd5upea4iyfznuislybj5p4wd6jgtx5ayzurnk.py
# Topologically Sorted Source Nodes: [add_1, hidden_states_2], Original ATen: [aten.add, aten.native_layer_norm]
# Source node to ATen node mapping:
# add_1 => add_1
# hidden_states_2 => var_mean
# Graph fragment:
# %add_1 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%view_17, %primals_3), kwargs = {})
# %var_mean : [num_users=2] = call_function[target=torch.ops.aten.var_mean.correction](args = (%add_1, [2]), kwargs = {correction: 0, keepdim: True})
triton_poi_fused_add_native_layer_norm_4 = async_compile.triton('triton_poi_fused_add_native_layer_norm_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_native_layer_norm_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_native_layer_norm_4(in_ptr0, in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (4*x0), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (4*x0), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr1 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 + tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 + tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp2 - tmp16
tmp18 = tmp17 * tmp17
tmp19 = tmp5 - tmp16
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp22 = tmp9 - tmp16
tmp23 = tmp22 * tmp22
tmp24 = tmp21 + tmp23
tmp25 = tmp13 - tmp16
tmp26 = tmp25 * tmp25
tmp27 = tmp24 + tmp26
tmp28 = tmp27 / tmp15
tl.store(out_ptr0 + (x0), tmp16, xmask)
tl.store(out_ptr1 + (x0), tmp28, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/mj/cmjciemk3whm7x2gsfwjg4qezvxvfib7vu5qk2zxl4iggsr3bnb5.py
# Topologically Sorted Source Nodes: [add_1, hidden_states_2], Original ATen: [aten.add, aten.native_layer_norm]
# Source node to ATen node mapping:
# add_1 => add_1
# hidden_states_2 => add_2, add_3, mul, mul_1, rsqrt, sub_1
# Graph fragment:
# %add_1 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%view_17, %primals_3), kwargs = {})
# %add_2 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%getitem, 1.0), kwargs = {})
# %rsqrt : [num_users=1] = call_function[target=torch.ops.aten.rsqrt.default](args = (%add_2,), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add_1, %getitem_1), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_1, %rsqrt), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%mul, %primals_11), kwargs = {})
# %add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul_1, %primals_12), kwargs = {})
triton_poi_fused_add_native_layer_norm_5 = async_compile.triton('triton_poi_fused_add_native_layer_norm_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: '*fp32', 7: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_native_layer_norm_5', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 6, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_native_layer_norm_5(in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp3 = tl.load(in_ptr2 + (x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr3 + (x1), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr4 + (x0), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr5 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp4 = tmp2 - tmp3
tmp6 = 1.0
tmp7 = tmp5 + tmp6
tmp8 = libdevice.rsqrt(tmp7)
tmp9 = tmp4 * tmp8
tmp11 = tmp9 * tmp10
tmp13 = tmp11 + tmp12
tl.store(out_ptr0 + (x2), tmp13, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4, ), (1, ))
assert_size_stride(primals_8, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4, ), (1, ))
assert_size_stride(primals_11, (4, ), (1, ))
assert_size_stride(primals_12, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf2)
del primals_6
buf3 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [attention_scores], Original ATen: [aten.clone]
stream0 = get_raw_stream(0)
triton_poi_fused_clone_0.run(buf0, primals_2, buf3, 16, 4, grid=grid(16, 4), stream=stream0)
del primals_2
buf4 = reinterpret_tensor(buf0, (4, 4, 1, 4), (16, 4, 4, 1), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [attention_scores], Original ATen: [aten.clone]
triton_poi_fused_clone_0.run(buf1, primals_5, buf4, 16, 4, grid=grid(16, 4), stream=stream0)
del primals_5
buf5 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [attention_scores], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 0), 0), reinterpret_tensor(buf4, (16, 1, 4), (4, 0, 1), 0), out=buf5)
buf6 = reinterpret_tensor(buf1, (4, 4, 4, 1), (16, 4, 1, 64), 0); del buf1 # reuse
buf7 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 64), torch.float32)
# Topologically Sorted Source Nodes: [attention_scores_1, attention_scores_2, attention_probs], Original ATen: [aten.div, aten.add, aten._softmax]
triton_poi_fused__softmax_add_div_1.run(buf5, primals_8, buf6, buf7, 64, grid=grid(64), stream=stream0)
buf8 = reinterpret_tensor(buf5, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf5 # reuse
# Topologically Sorted Source Nodes: [attention_scores_1, attention_scores_2, attention_probs], Original ATen: [aten.div, aten.add, aten._softmax]
triton_poi_fused__softmax_add_div_2.run(buf8, primals_8, buf6, buf7, 256, grid=grid(256), stream=stream0)
del primals_8
buf9 = reinterpret_tensor(buf7, (4, 4, 4, 1), (16, 4, 1, 1), 0); del buf7 # reuse
# Topologically Sorted Source Nodes: [context_layer], Original ATen: [aten.clone]
triton_poi_fused_clone_0.run(buf2, primals_7, buf9, 16, 4, grid=grid(16, 4), stream=stream0)
del primals_7
buf10 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [context_layer], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf8, (16, 4, 4), (16, 4, 1), 0), reinterpret_tensor(buf9, (16, 4, 1), (4, 1, 0), 0), out=buf10)
buf11 = reinterpret_tensor(buf6, (4, 4, 4, 1), (16, 4, 1, 1), 0); del buf6 # reuse
# Topologically Sorted Source Nodes: [context_layer_1], Original ATen: [aten.clone]
triton_poi_fused_clone_3.run(buf10, buf11, 16, 4, grid=grid(16, 4), stream=stream0)
buf12 = reinterpret_tensor(buf10, (16, 4), (4, 1), 0); del buf10 # reuse
# Topologically Sorted Source Nodes: [hidden_states], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_10, reinterpret_tensor(buf11, (16, 4), (4, 1), 0), reinterpret_tensor(primals_9, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf12)
del primals_10
buf13 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
buf14 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
# Topologically Sorted Source Nodes: [add_1, hidden_states_2], Original ATen: [aten.add, aten.native_layer_norm]
triton_poi_fused_add_native_layer_norm_4.run(buf12, primals_3, buf13, buf14, 16, grid=grid(16), stream=stream0)
buf15 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [add_1, hidden_states_2], Original ATen: [aten.add, aten.native_layer_norm]
triton_poi_fused_add_native_layer_norm_5.run(buf12, primals_3, buf13, buf14, primals_11, primals_12, buf15, 64, grid=grid(64), stream=stream0)
del buf13
del buf14
del primals_12
return (buf15, primals_3, primals_11, buf8, reinterpret_tensor(buf11, (16, 4), (4, 1), 0), buf12, primals_9, reinterpret_tensor(buf9, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0), reinterpret_tensor(buf4, (16, 4, 1), (4, 1, 4), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_12 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.nn as nn
class MultiHeadAttention(nn.Module):
"""
Multi-head Self-attention layers, a attention score dropout layer is introduced.
Args:
input_tensor (torch.Tensor): the input of the multi-head self-attention layer
attention_mask (torch.Tensor): the attention mask for input tensor
Returns:
hidden_states (torch.Tensor): the output of the multi-head self-attention layer
"""
def __init__(self, n_heads, hidden_size, hidden_dropout_prob,
attn_dropout_prob, layer_norm_eps):
super(MultiHeadAttention, self).__init__()
if hidden_size % n_heads != 0:
raise ValueError(
'The hidden size (%d) is not a multiple of the number of attention heads (%d)'
% (hidden_size, n_heads))
self.num_attention_heads = n_heads
self.attention_head_size = int(hidden_size / n_heads)
self.all_head_size = (self.num_attention_heads * self.
attention_head_size)
self.sqrt_attention_head_size = math.sqrt(self.attention_head_size)
self.query = nn.Linear(hidden_size, self.all_head_size)
self.key = nn.Linear(hidden_size, self.all_head_size)
self.value = nn.Linear(hidden_size, self.all_head_size)
self.softmax = nn.Softmax(dim=-1)
self.attn_dropout = nn.Dropout(attn_dropout_prob)
self.dense = nn.Linear(hidden_size, hidden_size)
self.LayerNorm = nn.LayerNorm(hidden_size, eps=layer_norm_eps)
self.out_dropout = nn.Dropout(hidden_dropout_prob)
def transpose_for_scores(self, x):
new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.
attention_head_size)
x = x.view(*new_x_shape)
return x
def forward(self, input_tensor, attention_mask):
mixed_query_layer = self.query(input_tensor)
mixed_key_layer = self.key(input_tensor)
mixed_value_layer = self.value(input_tensor)
query_layer = self.transpose_for_scores(mixed_query_layer).permute(
0, 2, 1, 3)
key_layer = self.transpose_for_scores(mixed_key_layer).permute(0, 2,
3, 1)
value_layer = self.transpose_for_scores(mixed_value_layer).permute(
0, 2, 1, 3)
attention_scores = torch.matmul(query_layer, key_layer)
attention_scores = attention_scores / self.sqrt_attention_head_size
attention_scores = attention_scores + attention_mask
attention_probs = self.softmax(attention_scores)
attention_probs = self.attn_dropout(attention_probs)
context_layer = torch.matmul(attention_probs, value_layer)
context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
new_context_layer_shape = context_layer.size()[:-2] + (self.
all_head_size,)
context_layer = context_layer.view(*new_context_layer_shape)
hidden_states = self.dense(context_layer)
hidden_states = self.out_dropout(hidden_states)
hidden_states = self.LayerNorm(hidden_states + input_tensor)
return hidden_states
def get_inputs():
return [torch.rand([4, 4, 4]), torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'n_heads': 4, 'hidden_size': 4, 'hidden_dropout_prob': 0.5,
'attn_dropout_prob': 0.5, 'layer_norm_eps': 1}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_clone_0(in_ptr0, in_ptr1, out_ptr0, ynumel, xnumel,
YBLOCK: tl.constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = yindex // 4
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x2 + 16 * y1), xmask & ymask,
eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + y0, ymask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (x2 + 4 * y3), tmp2, xmask & ymask)
@triton.jit
def triton_poi_fused__softmax_add_div_1(in_ptr0, in_ptr1, out_ptr0,
out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 16
tmp0 = tl.load(in_ptr0 + 4 * x2, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr1 + 4 * x0, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + 4 * x2), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (2 + 4 * x2), xmask, eviction_policy='evict_last'
)
tmp12 = tl.load(in_ptr1 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp15 = tl.load(in_ptr0 + (3 + 4 * x2), xmask, eviction_policy='evict_last'
)
tmp17 = tl.load(in_ptr1 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp5 * tmp1
tmp8 = tmp6 + tmp7
tmp9 = triton_helpers.maximum(tmp4, tmp8)
tmp11 = tmp10 * tmp1
tmp13 = tmp11 + tmp12
tmp14 = triton_helpers.maximum(tmp9, tmp13)
tmp16 = tmp15 * tmp1
tmp18 = tmp16 + tmp17
tmp19 = triton_helpers.maximum(tmp14, tmp18)
tmp20 = tmp4 - tmp19
tmp21 = tl_math.exp(tmp20)
tmp22 = tmp8 - tmp19
tmp23 = tl_math.exp(tmp22)
tmp24 = tmp21 + tmp23
tmp25 = tmp13 - tmp19
tmp26 = tl_math.exp(tmp25)
tmp27 = tmp24 + tmp26
tmp28 = tmp18 - tmp19
tmp29 = tl_math.exp(tmp28)
tmp30 = tmp27 + tmp29
tl.store(out_ptr0 + x2, tmp19, xmask)
tl.store(out_ptr1 + x2, tmp30, xmask)
@triton.jit
def triton_poi_fused__softmax_add_div_2(in_out_ptr0, in_ptr0, in_ptr1,
in_ptr2, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x4 = xindex % 64
x5 = xindex // 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp3 = tl.load(in_ptr0 + x4, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr1 + x5, xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr2 + x5, xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 - tmp5
tmp7 = tl_math.exp(tmp6)
tmp9 = tmp7 / tmp8
tl.store(in_out_ptr0 + x3, tmp9, xmask)
@triton.jit
def triton_poi_fused_clone_3(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 4
y1 = yindex // 4
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x2 + 16 * y1), xmask & ymask,
eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + 4 * y3), tmp0, xmask & ymask)
@triton.jit
def triton_poi_fused_add_native_layer_norm_4(in_ptr0, in_ptr1, out_ptr0,
out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 4 * x0, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + 4 * x0, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (2 + 4 * x0), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (2 + 4 * x0), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp12 = tl.load(in_ptr1 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp2 = tmp0 + tmp1
tmp5 = tmp3 + tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 + tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 + tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp2 - tmp16
tmp18 = tmp17 * tmp17
tmp19 = tmp5 - tmp16
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp22 = tmp9 - tmp16
tmp23 = tmp22 * tmp22
tmp24 = tmp21 + tmp23
tmp25 = tmp13 - tmp16
tmp26 = tmp25 * tmp25
tmp27 = tmp24 + tmp26
tmp28 = tmp27 / tmp15
tl.store(out_ptr0 + x0, tmp16, xmask)
tl.store(out_ptr1 + x0, tmp28, xmask)
@triton.jit
def triton_poi_fused_add_native_layer_norm_5(in_ptr0, in_ptr1, in_ptr2,
in_ptr3, in_ptr4, in_ptr5, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp3 = tl.load(in_ptr2 + x1, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr3 + x1, xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr4 + x0, xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr5 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp4 = tmp2 - tmp3
tmp6 = 1.0
tmp7 = tmp5 + tmp6
tmp8 = libdevice.rsqrt(tmp7)
tmp9 = tmp4 * tmp8
tmp11 = tmp9 * tmp10
tmp13 = tmp11 + tmp12
tl.store(out_ptr0 + x2, tmp13, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11, primals_12
) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4,), (1,))
assert_size_stride(primals_8, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (4,), (1,))
assert_size_stride(primals_11, (4,), (1,))
assert_size_stride(primals_12, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_4, (4, 4), (1, 4), 0), out=buf1)
del primals_4
buf2 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf2)
del primals_6
buf3 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_clone_0[grid(16, 4)](buf0, primals_2, buf3, 16, 4,
XBLOCK=4, YBLOCK=8, num_warps=1, num_stages=1)
del primals_2
buf4 = reinterpret_tensor(buf0, (4, 4, 1, 4), (16, 4, 4, 1), 0)
del buf0
triton_poi_fused_clone_0[grid(16, 4)](buf1, primals_5, buf4, 16, 4,
XBLOCK=4, YBLOCK=8, num_warps=1, num_stages=1)
del primals_5
buf5 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 1), (4, 1, 0),
0), reinterpret_tensor(buf4, (16, 1, 4), (4, 0, 1), 0), out=buf5)
buf6 = reinterpret_tensor(buf1, (4, 4, 4, 1), (16, 4, 1, 64), 0)
del buf1
buf7 = empty_strided_cuda((4, 4, 4, 1), (16, 4, 1, 64), torch.float32)
triton_poi_fused__softmax_add_div_1[grid(64)](buf5, primals_8, buf6,
buf7, 64, XBLOCK=64, num_warps=1, num_stages=1)
buf8 = reinterpret_tensor(buf5, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf5
triton_poi_fused__softmax_add_div_2[grid(256)](buf8, primals_8,
buf6, buf7, 256, XBLOCK=256, num_warps=4, num_stages=1)
del primals_8
buf9 = reinterpret_tensor(buf7, (4, 4, 4, 1), (16, 4, 1, 1), 0)
del buf7
triton_poi_fused_clone_0[grid(16, 4)](buf2, primals_7, buf9, 16, 4,
XBLOCK=4, YBLOCK=8, num_warps=1, num_stages=1)
del primals_7
buf10 = reinterpret_tensor(buf2, (16, 4, 1), (4, 1, 1), 0)
del buf2
extern_kernels.bmm(reinterpret_tensor(buf8, (16, 4, 4), (16, 4, 1),
0), reinterpret_tensor(buf9, (16, 4, 1), (4, 1, 0), 0), out=buf10)
buf11 = reinterpret_tensor(buf6, (4, 4, 4, 1), (16, 4, 1, 1), 0)
del buf6
triton_poi_fused_clone_3[grid(16, 4)](buf10, buf11, 16, 4, XBLOCK=4,
YBLOCK=16, num_warps=1, num_stages=1)
buf12 = reinterpret_tensor(buf10, (16, 4), (4, 1), 0)
del buf10
extern_kernels.addmm(primals_10, reinterpret_tensor(buf11, (16, 4),
(4, 1), 0), reinterpret_tensor(primals_9, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf12)
del primals_10
buf13 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
buf14 = empty_strided_cuda((4, 4, 1), (4, 1, 16), torch.float32)
triton_poi_fused_add_native_layer_norm_4[grid(16)](buf12, primals_3,
buf13, buf14, 16, XBLOCK=16, num_warps=1, num_stages=1)
buf15 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_add_native_layer_norm_5[grid(64)](buf12, primals_3,
buf13, buf14, primals_11, primals_12, buf15, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del buf13
del buf14
del primals_12
return buf15, primals_3, primals_11, buf8, reinterpret_tensor(buf11, (
16, 4), (4, 1), 0), buf12, primals_9, reinterpret_tensor(buf9, (16,
1, 4), (4, 1, 1), 0), reinterpret_tensor(buf3, (16, 1, 4), (4, 1, 1), 0
), reinterpret_tensor(buf4, (16, 4, 1), (4, 1, 4), 0)
class MultiHeadAttentionNew(nn.Module):
"""
Multi-head Self-attention layers, a attention score dropout layer is introduced.
Args:
input_tensor (torch.Tensor): the input of the multi-head self-attention layer
attention_mask (torch.Tensor): the attention mask for input tensor
Returns:
hidden_states (torch.Tensor): the output of the multi-head self-attention layer
"""
def __init__(self, n_heads, hidden_size, hidden_dropout_prob,
attn_dropout_prob, layer_norm_eps):
super(MultiHeadAttentionNew, self).__init__()
if hidden_size % n_heads != 0:
raise ValueError(
'The hidden size (%d) is not a multiple of the number of attention heads (%d)'
% (hidden_size, n_heads))
self.num_attention_heads = n_heads
self.attention_head_size = int(hidden_size / n_heads)
self.all_head_size = (self.num_attention_heads * self.
attention_head_size)
self.sqrt_attention_head_size = math.sqrt(self.attention_head_size)
self.query = nn.Linear(hidden_size, self.all_head_size)
self.key = nn.Linear(hidden_size, self.all_head_size)
self.value = nn.Linear(hidden_size, self.all_head_size)
self.softmax = nn.Softmax(dim=-1)
self.attn_dropout = nn.Dropout(attn_dropout_prob)
self.dense = nn.Linear(hidden_size, hidden_size)
self.LayerNorm = nn.LayerNorm(hidden_size, eps=layer_norm_eps)
self.out_dropout = nn.Dropout(hidden_dropout_prob)
def transpose_for_scores(self, x):
new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.
attention_head_size)
x = x.view(*new_x_shape)
return x
def forward(self, input_0, input_1):
primals_1 = self.query.weight
primals_2 = self.query.bias
primals_4 = self.key.weight
primals_5 = self.key.bias
primals_6 = self.value.weight
primals_7 = self.value.bias
primals_9 = self.dense.weight
primals_10 = self.dense.bias
primals_11 = self.LayerNorm.weight
primals_12 = self.LayerNorm.bias
primals_3 = input_0
primals_8 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11, primals_12])
return output[0]
|
dreaming-qin/RecBole
|
MultiHeadAttention
| false | 12,370 |
[
"MIT"
] | 0 |
d6de39521484ded60c387ca604abaf86310acdbe
|
https://github.com/dreaming-qin/RecBole/tree/d6de39521484ded60c387ca604abaf86310acdbe
|
location_network
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/z5/cz5bgdo2gmhnnmtf6w7lrjkvliacxo7nomq7mbmjquxqyxqgt5bj.py
# Topologically Sorted Source Nodes: [feat], Original ATen: [aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# feat => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_1,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_relu_threshold_backward_0 = async_compile.triton('triton_poi_fused_relu_threshold_backward_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i1', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_threshold_backward_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_out_ptr0, in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 2
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(in_out_ptr0 + (x2), tmp4, xmask)
tl.store(out_ptr0 + (x2), tmp6, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/dd/cddxwex7cwfprervbfeycuuplj36x3rh67cnh5kc3z7udvmaw5a5.py
# Topologically Sorted Source Nodes: [mu, mul, l_t, log_scale, sub, pow_2, neg, mul_1, truediv, sub_1, log_pi, log_pi_1], Original ATen: [aten.tanh, aten.mul, aten.add, aten.log, aten.sub, aten.pow, aten.neg, aten.div, aten.sum]
# Source node to ATen node mapping:
# l_t => add
# log_pi => sub_2
# log_pi_1 => sum_1
# log_scale => full_default
# mu => tanh
# mul => mul
# mul_1 => full_default_1
# neg => neg
# pow_2 => pow_2
# sub => sub
# sub_1 => sub_1
# truediv => div
# Graph fragment:
# %tanh : [num_users=2] = call_function[target=torch.ops.aten.tanh.default](args = (%view_3,), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%normal_functional, %expand), kwargs = {})
# %add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%tanh, %mul), kwargs = {})
# %full_default : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([4, 4, 4, 4], 1.3862943649291992), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add, %tanh), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub, 2), kwargs = {})
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%pow_2,), kwargs = {})
# %full_default_1 : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([4, 4, 4, 4], 32.0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%neg, %full_default_1), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%div, %full_default), kwargs = {})
# %sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sub_1, 0.9189385332046727), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%sub_2, [1]), kwargs = {})
triton_poi_fused_add_div_log_mul_neg_pow_sub_sum_tanh_1 = async_compile.triton('triton_poi_fused_add_div_log_mul_neg_pow_sub_sum_tanh_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_log_mul_neg_pow_sub_sum_tanh_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_log_mul_neg_pow_sub_sum_tanh_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = (xindex // 16)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (64*x1)), xmask)
tmp2 = tl.load(in_ptr1 + (x0 + (64*x1)), xmask)
tmp15 = tl.load(in_ptr0 + (16 + x0 + (64*x1)), xmask)
tmp17 = tl.load(in_ptr1 + (16 + x0 + (64*x1)), xmask)
tmp27 = tl.load(in_ptr0 + (32 + x0 + (64*x1)), xmask)
tmp29 = tl.load(in_ptr1 + (32 + x0 + (64*x1)), xmask)
tmp39 = tl.load(in_ptr0 + (48 + x0 + (64*x1)), xmask)
tmp41 = tl.load(in_ptr1 + (48 + x0 + (64*x1)), xmask)
tmp1 = libdevice.tanh(tmp0)
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tmp1 + tmp4
tmp6 = tmp5 - tmp1
tmp7 = tmp6 * tmp6
tmp8 = -tmp7
tmp9 = 0.03125
tmp10 = tmp8 * tmp9
tmp11 = 1.3862943649291992
tmp12 = tmp10 - tmp11
tmp13 = 0.9189385332046727
tmp14 = tmp12 - tmp13
tmp16 = libdevice.tanh(tmp15)
tmp18 = tmp17 * tmp3
tmp19 = tmp16 + tmp18
tmp20 = tmp19 - tmp16
tmp21 = tmp20 * tmp20
tmp22 = -tmp21
tmp23 = tmp22 * tmp9
tmp24 = tmp23 - tmp11
tmp25 = tmp24 - tmp13
tmp26 = tmp14 + tmp25
tmp28 = libdevice.tanh(tmp27)
tmp30 = tmp29 * tmp3
tmp31 = tmp28 + tmp30
tmp32 = tmp31 - tmp28
tmp33 = tmp32 * tmp32
tmp34 = -tmp33
tmp35 = tmp34 * tmp9
tmp36 = tmp35 - tmp11
tmp37 = tmp36 - tmp13
tmp38 = tmp26 + tmp37
tmp40 = libdevice.tanh(tmp39)
tmp42 = tmp41 * tmp3
tmp43 = tmp40 + tmp42
tmp44 = tmp43 - tmp40
tmp45 = tmp44 * tmp44
tmp46 = -tmp45
tmp47 = tmp46 * tmp9
tmp48 = tmp47 - tmp11
tmp49 = tmp48 - tmp13
tmp50 = tmp38 + tmp49
tl.store(out_ptr0 + (x2), tmp50, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/wa/cwablyx4r3ifgi7p3ugkmqxfsyq2clouhpzo5gpw6wrj2fjncf4s.py
# Topologically Sorted Source Nodes: [mu, mul, l_t, l_t_2], Original ATen: [aten.tanh, aten.mul, aten.add, aten.clamp]
# Source node to ATen node mapping:
# l_t => add
# l_t_2 => clamp_max, clamp_min
# mu => tanh
# mul => mul
# Graph fragment:
# %tanh : [num_users=2] = call_function[target=torch.ops.aten.tanh.default](args = (%view_3,), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%normal_functional, %expand), kwargs = {})
# %add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%tanh, %mul), kwargs = {})
# %clamp_min : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%add, -1), kwargs = {})
# %clamp_max : [num_users=1] = call_function[target=torch.ops.aten.clamp_max.default](args = (%clamp_min, 1), kwargs = {})
triton_poi_fused_add_clamp_mul_tanh_2 = async_compile.triton('triton_poi_fused_add_clamp_mul_tanh_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_clamp_mul_tanh_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_clamp_mul_tanh_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp2 = tl.load(in_ptr1 + (x0), xmask)
tmp1 = libdevice.tanh(tmp0)
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tmp1 + tmp4
tmp6 = -1.0
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = 1.0
tmp9 = triton_helpers.minimum(tmp7, tmp8)
tl.store(out_ptr0 + (x0), tmp9, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (2, 4), (4, 1))
assert_size_stride(primals_3, (2, ), (1, ))
assert_size_stride(primals_4, (4, 2), (2, 1))
assert_size_stride(primals_5, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 2), (2, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_1, (64, 4), (4, 1), 0), reinterpret_tensor(primals_2, (4, 2), (1, 4), 0), out=buf0)
del primals_2
buf1 = reinterpret_tensor(buf0, (4, 4, 4, 2), (32, 8, 2, 1), 0); del buf0 # reuse
buf8 = empty_strided_cuda((4, 4, 4, 2), (32, 8, 2, 1), torch.bool)
# Topologically Sorted Source Nodes: [feat], Original ATen: [aten.relu, aten.threshold_backward]
stream0 = get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0.run(buf1, primals_3, buf8, 128, grid=grid(128), stream=stream0)
del primals_3
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear_1], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_5, reinterpret_tensor(buf1, (64, 2), (2, 1), 0), reinterpret_tensor(primals_4, (2, 4), (1, 2), 0), alpha=1, beta=1, out=buf2)
del primals_5
buf3 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [eps], Original ATen: [aten.normal_functional]
buf4 = torch.ops.aten.normal_functional.default(buf3)
buf5 = buf4
del buf4
buf6 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mu, mul, l_t, log_scale, sub, pow_2, neg, mul_1, truediv, sub_1, log_pi, log_pi_1], Original ATen: [aten.tanh, aten.mul, aten.add, aten.log, aten.sub, aten.pow, aten.neg, aten.div, aten.sum]
triton_poi_fused_add_div_log_mul_neg_pow_sub_sum_tanh_1.run(buf2, buf5, buf6, 64, grid=grid(64), stream=stream0)
buf7 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [mu, mul, l_t, l_t_2], Original ATen: [aten.tanh, aten.mul, aten.add, aten.clamp]
triton_poi_fused_add_clamp_mul_tanh_2.run(buf2, buf5, buf7, 256, grid=grid(256), stream=stream0)
return (buf6, buf7, reinterpret_tensor(primals_1, (64, 4), (4, 1), 0), reinterpret_tensor(buf1, (64, 2), (2, 1), 0), buf2, buf5, primals_4, buf8, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((2, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((2, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 2), (2, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.distributions import Normal
class location_network(nn.Module):
"""
Uses the internal state `h_t` of the core network to
produce the location coordinates `l_t` for the next
time step.
Concretely, feeds the hidden state `h_t` through a fc
layer followed by a tanh to clamp the output beween
[-1, 1]. This produces a 2D vector of means used to
parametrize a two-component Gaussian with a fixed
variance from which the location coordinates `l_t`
for the next time step are sampled.
Hence, the location `l_t` is chosen stochastically
from a distribution conditioned on an affine
transformation of the hidden state vector `h_t`.
Args
----
- input_size: input size of the fc layer.
- output_size: output size of the fc layer.
- std: standard deviation of the normal distribution.
- h_t: the hidden state vector of the core network for
the current time step `t`.
Returns
-------
- mu: a 2D vector of shape (B, 2).
- l_t: a 2D vector of shape (B, 2).
"""
def __init__(self, input_size, output_size, std):
super().__init__()
self.std = std
hid_size = input_size // 2
self.fc = nn.Linear(input_size, hid_size)
self.fc_lt = nn.Linear(hid_size, output_size)
def forward(self, h_t):
feat = F.relu(self.fc(h_t.detach()))
mu = torch.tanh(self.fc_lt(feat))
l_t = torch.distributions.Normal(mu, self.std).rsample()
l_t = l_t.detach()
log_pi = Normal(mu, self.std).log_prob(l_t)
log_pi = torch.sum(log_pi, dim=1)
l_t = torch.clamp(l_t, -1, 1)
return log_pi, l_t
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'input_size': 4, 'output_size': 4, 'std': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_out_ptr0, in_ptr0,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 2
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(in_out_ptr0 + x2, tmp4, xmask)
tl.store(out_ptr0 + x2, tmp6, xmask)
@triton.jit
def triton_poi_fused_add_div_log_mul_neg_pow_sub_sum_tanh_1(in_ptr0,
in_ptr1, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = xindex // 16
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 64 * x1), xmask)
tmp2 = tl.load(in_ptr1 + (x0 + 64 * x1), xmask)
tmp15 = tl.load(in_ptr0 + (16 + x0 + 64 * x1), xmask)
tmp17 = tl.load(in_ptr1 + (16 + x0 + 64 * x1), xmask)
tmp27 = tl.load(in_ptr0 + (32 + x0 + 64 * x1), xmask)
tmp29 = tl.load(in_ptr1 + (32 + x0 + 64 * x1), xmask)
tmp39 = tl.load(in_ptr0 + (48 + x0 + 64 * x1), xmask)
tmp41 = tl.load(in_ptr1 + (48 + x0 + 64 * x1), xmask)
tmp1 = libdevice.tanh(tmp0)
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tmp1 + tmp4
tmp6 = tmp5 - tmp1
tmp7 = tmp6 * tmp6
tmp8 = -tmp7
tmp9 = 0.03125
tmp10 = tmp8 * tmp9
tmp11 = 1.3862943649291992
tmp12 = tmp10 - tmp11
tmp13 = 0.9189385332046727
tmp14 = tmp12 - tmp13
tmp16 = libdevice.tanh(tmp15)
tmp18 = tmp17 * tmp3
tmp19 = tmp16 + tmp18
tmp20 = tmp19 - tmp16
tmp21 = tmp20 * tmp20
tmp22 = -tmp21
tmp23 = tmp22 * tmp9
tmp24 = tmp23 - tmp11
tmp25 = tmp24 - tmp13
tmp26 = tmp14 + tmp25
tmp28 = libdevice.tanh(tmp27)
tmp30 = tmp29 * tmp3
tmp31 = tmp28 + tmp30
tmp32 = tmp31 - tmp28
tmp33 = tmp32 * tmp32
tmp34 = -tmp33
tmp35 = tmp34 * tmp9
tmp36 = tmp35 - tmp11
tmp37 = tmp36 - tmp13
tmp38 = tmp26 + tmp37
tmp40 = libdevice.tanh(tmp39)
tmp42 = tmp41 * tmp3
tmp43 = tmp40 + tmp42
tmp44 = tmp43 - tmp40
tmp45 = tmp44 * tmp44
tmp46 = -tmp45
tmp47 = tmp46 * tmp9
tmp48 = tmp47 - tmp11
tmp49 = tmp48 - tmp13
tmp50 = tmp38 + tmp49
tl.store(out_ptr0 + x2, tmp50, xmask)
@triton.jit
def triton_poi_fused_add_clamp_mul_tanh_2(in_ptr0, in_ptr1, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp2 = tl.load(in_ptr1 + x0, xmask)
tmp1 = libdevice.tanh(tmp0)
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tmp1 + tmp4
tmp6 = -1.0
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = 1.0
tmp9 = triton_helpers.minimum(tmp7, tmp8)
tl.store(out_ptr0 + x0, tmp9, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (2, 4), (4, 1))
assert_size_stride(primals_3, (2,), (1,))
assert_size_stride(primals_4, (4, 2), (2, 1))
assert_size_stride(primals_5, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 2), (2, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_1, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_2, (4, 2), (1, 4), 0), out=buf0)
del primals_2
buf1 = reinterpret_tensor(buf0, (4, 4, 4, 2), (32, 8, 2, 1), 0)
del buf0
buf8 = empty_strided_cuda((4, 4, 4, 2), (32, 8, 2, 1), torch.bool)
get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0[grid(128)](buf1,
primals_3, buf8, 128, XBLOCK=128, num_warps=4, num_stages=1)
del primals_3
buf2 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_5, reinterpret_tensor(buf1, (64, 2), (
2, 1), 0), reinterpret_tensor(primals_4, (2, 4), (1, 2), 0),
alpha=1, beta=1, out=buf2)
del primals_5
buf3 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
buf4 = torch.ops.aten.normal_functional.default(buf3)
buf5 = buf4
del buf4
buf6 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_add_div_log_mul_neg_pow_sub_sum_tanh_1[grid(64)](buf2,
buf5, buf6, 64, XBLOCK=64, num_warps=1, num_stages=1)
buf7 = buf3
del buf3
triton_poi_fused_add_clamp_mul_tanh_2[grid(256)](buf2, buf5, buf7,
256, XBLOCK=128, num_warps=4, num_stages=1)
return buf6, buf7, reinterpret_tensor(primals_1, (64, 4), (4, 1), 0
), reinterpret_tensor(buf1, (64, 2), (2, 1), 0
), buf2, buf5, primals_4, buf8
class location_networkNew(nn.Module):
"""
Uses the internal state `h_t` of the core network to
produce the location coordinates `l_t` for the next
time step.
Concretely, feeds the hidden state `h_t` through a fc
layer followed by a tanh to clamp the output beween
[-1, 1]. This produces a 2D vector of means used to
parametrize a two-component Gaussian with a fixed
variance from which the location coordinates `l_t`
for the next time step are sampled.
Hence, the location `l_t` is chosen stochastically
from a distribution conditioned on an affine
transformation of the hidden state vector `h_t`.
Args
----
- input_size: input size of the fc layer.
- output_size: output size of the fc layer.
- std: standard deviation of the normal distribution.
- h_t: the hidden state vector of the core network for
the current time step `t`.
Returns
-------
- mu: a 2D vector of shape (B, 2).
- l_t: a 2D vector of shape (B, 2).
"""
def __init__(self, input_size, output_size, std):
super().__init__()
self.std = std
hid_size = input_size // 2
self.fc = nn.Linear(input_size, hid_size)
self.fc_lt = nn.Linear(hid_size, output_size)
def forward(self, input_0):
primals_2 = self.fc.weight
primals_3 = self.fc.bias
primals_4 = self.fc_lt.weight
primals_5 = self.fc_lt.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4, primals_5])
return output[0], output[1]
|
felixnon/foveated-visual-attention
|
location_network
| false | 12,371 |
[
"MIT"
] | 0 |
7e7d9a5ef24ec42eb76ba72f783bb2227bdb4851
|
https://github.com/felixnon/foveated-visual-attention/tree/7e7d9a5ef24ec42eb76ba72f783bb2227bdb4851
|
MLPAttention
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/iq/ciqftckmgqckpktocc4vnzvzxloichfo465idr3x42oonjq74tp7.py
# Topologically Sorted Source Nodes: [inner_sum, tanh], Original ATen: [aten.add, aten.tanh]
# Source node to ATen node mapping:
# inner_sum => add
# tanh => tanh
# Graph fragment:
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%view_1, %view_3), kwargs = {})
# %tanh : [num_users=2] = call_function[target=torch.ops.aten.tanh.default](args = (%add,), kwargs = {})
triton_poi_fused_add_tanh_0 = async_compile.triton('triton_poi_fused_add_tanh_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_tanh_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_tanh_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + (x0), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask)
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x0), tmp3, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/dq/cdqbbqy7obiw5h7sjtb2zv2fvtkcjtt2cb3ee5uvgfirualk4m2f.py
# Topologically Sorted Source Nodes: [alpha], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# alpha => amax, exp, sub
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%squeeze, [0], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%squeeze, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
triton_poi_fused__softmax_1 = async_compile.triton('triton_poi_fused__softmax_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 16
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp3 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (16 + x0), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (32 + x0), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (48 + x0), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = tl_math.exp(tmp14)
tl.store(out_ptr0 + (x2), tmp15, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/f5/cf5347orlkm5ylmh4iouw6qvowwebewwg66dnqrhzt7tg7a4irp5.py
# Topologically Sorted Source Nodes: [alpha], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# alpha => div_1, sum_1
# Graph fragment:
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [0], True), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_2 = async_compile.triton('triton_poi_fused__softmax_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 16
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (16 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (32 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (48 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/nt/cntzaspueq6fx2yeimpeg4uzm6tlchzf2wwq7mhodo7u3mdomp62.py
# Topologically Sorted Source Nodes: [mul, sum_1], Original ATen: [aten.mul, aten.sum]
# Source node to ATen node mapping:
# mul => mul
# sum_1 => sum_2
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%unsqueeze, %primals_2), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul, [0]), kwargs = {})
triton_poi_fused_mul_sum_3 = async_compile.triton('triton_poi_fused_mul_sum_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_sum_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_sum_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x2), xmask)
tmp3 = tl.load(in_ptr0 + (16 + x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (64 + x2), xmask)
tmp7 = tl.load(in_ptr0 + (32 + x1), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (128 + x2), xmask)
tmp11 = tl.load(in_ptr0 + (48 + x1), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr1 + (192 + x2), xmask)
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tl.store(out_ptr0 + (x2), tmp14, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4, 4), (4, 1))
assert_size_stride(primals_4, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_5, (1, 4), (4, 1))
assert_size_stride(primals_6, (4, 4), (4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear], Original ATen: [aten.mm]
extern_kernels.mm(reinterpret_tensor(primals_2, (64, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear_1], Original ATen: [aten.mm]
extern_kernels.mm(reinterpret_tensor(primals_4, (64, 4), (4, 1), 0), reinterpret_tensor(primals_3, (4, 4), (1, 4), 0), out=buf1)
del primals_3
buf2 = reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [inner_sum, tanh], Original ATen: [aten.add, aten.tanh]
stream0 = get_raw_stream(0)
triton_poi_fused_add_tanh_0.run(buf2, buf1, 256, grid=grid(256), stream=stream0)
del buf1
buf3 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear_2], Original ATen: [aten.mm]
extern_kernels.mm(reinterpret_tensor(buf2, (64, 4), (4, 1), 0), reinterpret_tensor(primals_5, (4, 1), (1, 4), 0), out=buf3)
buf4 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [alpha], Original ATen: [aten._softmax]
triton_poi_fused__softmax_1.run(buf3, buf4, 64, grid=grid(64), stream=stream0)
buf5 = reinterpret_tensor(buf3, (4, 4, 4), (16, 4, 1), 0); del buf3 # reuse
# Topologically Sorted Source Nodes: [alpha], Original ATen: [aten._softmax]
triton_poi_fused__softmax_2.run(buf4, buf5, 64, grid=grid(64), stream=stream0)
buf6 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [mul, sum_1], Original ATen: [aten.mul, aten.sum]
triton_poi_fused_mul_sum_3.run(buf5, primals_2, buf6, 64, grid=grid(64), stream=stream0)
buf7 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear_3], Original ATen: [aten.mm]
extern_kernels.mm(reinterpret_tensor(buf6, (16, 4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf7)
return (buf5, reinterpret_tensor(buf7, (4, 4, 4), (16, 4, 1), 0), primals_2, reinterpret_tensor(primals_4, (64, 4), (4, 1), 0), buf2, buf5, reinterpret_tensor(buf6, (16, 4), (4, 1), 0), primals_6, primals_5, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((1, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
import torch.nn.functional as F
import torch.optim
def get_activation_fn(name):
"""Returns a callable activation function from torch."""
if name in (None, 'linear'):
return lambda x: x
elif name in ('sigmoid', 'tanh'):
return getattr(torch, name)
else:
return getattr(F, name)
class DotAttention(nn.Module):
"""Attention layer with dot product."""
def __init__(self, ctx_dim, hid_dim, att_bottleneck='ctx',
transform_ctx=True, att_activ='tanh', temp=1.0, ctx2hid=True,
mlp_bias=None):
super().__init__()
self.ctx_dim = ctx_dim
self.hid_dim = hid_dim
self._ctx2hid = ctx2hid
self.temperature = temp
self.activ = get_activation_fn(att_activ)
if isinstance(att_bottleneck, int):
self.mid_dim = att_bottleneck
else:
self.mid_dim = getattr(self, '{}_dim'.format(att_bottleneck))
self.hid2ctx = nn.Linear(self.hid_dim, self.mid_dim, bias=False)
if transform_ctx or self.mid_dim != self.ctx_dim:
self.ctx2ctx = nn.Linear(self.ctx_dim, self.mid_dim, bias=False)
else:
self.ctx2ctx = lambda x: x
if self._ctx2hid:
self.ctx2hid = nn.Linear(self.ctx_dim, self.hid_dim, bias=False)
else:
self.ctx2hid = lambda x: x
def forward(self, hid, ctx, ctx_mask=None):
"""Computes attention probabilities and final context using
decoder's hidden state and source annotations.
Arguments:
hid(Tensor): A set of decoder hidden states of shape `T*B*H`
where `T` == 1, `B` is batch dim and `H` is hidden state dim.
ctx(Tensor): A set of annotations of shape `S*B*C` where `S`
is the source timestep dim, `B` is batch dim and `C`
is annotation dim.
ctx_mask(FloatTensor): A binary mask of shape `S*B` with zeroes
in the padded positions.
Returns:
scores(Tensor): A tensor of shape `S*B` containing normalized
attention scores for each position and sample.
z_t(Tensor): A tensor of shape `B*H` containing the final
attended context vector for this target decoding timestep.
Notes:
This will only work when `T==1` for now.
"""
ctx_ = self.ctx2ctx(ctx)
hid_ = self.hid2ctx(hid)
scores = torch.bmm(hid_.permute(1, 0, 2), ctx_.permute(1, 2, 0)).div(
self.temperature).squeeze(1).t()
if ctx_mask is not None:
scores.masked_fill_((1 - ctx_mask).byte(), -100000000.0)
alpha = F.softmax(scores, dim=0)
return alpha, self.ctx2hid((alpha.unsqueeze(-1) * ctx).sum(0))
class MLPAttention(DotAttention):
"""Attention layer with feed-forward layer."""
def __init__(self, ctx_dim, hid_dim, att_bottleneck='ctx',
transform_ctx=True, att_activ='tanh', mlp_bias=False, temp=1.0,
ctx2hid=True):
super().__init__(ctx_dim, hid_dim, att_bottleneck, transform_ctx,
att_activ, temp, ctx2hid)
if mlp_bias:
self.bias = nn.Parameter(torch.Tensor(self.mid_dim))
self.bias.data.zero_()
else:
self.register_parameter('bias', None)
self.mlp = nn.Linear(self.mid_dim, 1, bias=False)
def forward(self, hid, ctx, ctx_mask=None):
"""Computes attention probabilities and final context using
decoder's hidden state and source annotations.
Arguments:
hid(Tensor): A set of decoder hidden states of shape `T*B*H`
where `T` == 1, `B` is batch dim and `H` is hidden state dim.
ctx(Tensor): A set of annotations of shape `S*B*C` where `S`
is the source timestep dim, `B` is batch dim and `C`
is annotation dim.
ctx_mask(FloatTensor): A binary mask of shape `S*B` with zeroes
in the padded positions.
Returns:
scores(Tensor): A tensor of shape `S*B` containing normalized
attention scores for each position and sample.
z_t(Tensor): A tensor of shape `B*H` containing the final
attended context vector for this target decoding timestep.
Notes:
This will only work when `T==1` for now.
"""
inner_sum = self.ctx2ctx(ctx) + self.hid2ctx(hid)
if self.bias is not None:
inner_sum.add_(self.bias)
scores = self.mlp(self.activ(inner_sum)).div(self.temperature).squeeze(
-1)
if ctx_mask is not None:
scores.masked_fill_((1 - ctx_mask).byte(), -100000000.0)
alpha = F.softmax(scores, dim=0)
return alpha, self.ctx2hid((alpha.unsqueeze(-1) * ctx).sum(0))
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'ctx_dim': 4, 'hid_dim': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch import nn
import torch.nn.functional as F
import torch.optim
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_add_tanh_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + x0, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask)
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x0, tmp3, xmask)
@triton.jit
def triton_poi_fused__softmax_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 16
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp3 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (16 + x0), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (32 + x0), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (48 + x0), xmask, eviction_policy='evict_last')
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = tl_math.exp(tmp14)
tl.store(out_ptr0 + x2, tmp15, xmask)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 16
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (16 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (32 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (48 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_mul_sum_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4
x2 = xindex
tmp0 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + x2, xmask)
tmp3 = tl.load(in_ptr0 + (16 + x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr1 + (64 + x2), xmask)
tmp7 = tl.load(in_ptr0 + (32 + x1), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr1 + (128 + x2), xmask)
tmp11 = tl.load(in_ptr0 + (48 + x1), xmask, eviction_policy='evict_last')
tmp12 = tl.load(in_ptr1 + (192 + x2), xmask)
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tl.store(out_ptr0 + x2, tmp14, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (4, 4), (4, 1))
assert_size_stride(primals_4, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_5, (1, 4), (4, 1))
assert_size_stride(primals_6, (4, 4), (4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_2, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_4, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_3, (4, 4), (1, 4), 0), out=buf1)
del primals_3
buf2 = reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf0
get_raw_stream(0)
triton_poi_fused_add_tanh_0[grid(256)](buf2, buf1, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del buf1
buf3 = empty_strided_cuda((64, 1), (1, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf2, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_5, (4, 1), (1, 4), 0), out=buf3)
buf4 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_1[grid(64)](buf3, buf4, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf5 = reinterpret_tensor(buf3, (4, 4, 4), (16, 4, 1), 0)
del buf3
triton_poi_fused__softmax_2[grid(64)](buf4, buf5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf6 = buf4
del buf4
triton_poi_fused_mul_sum_3[grid(64)](buf5, primals_2, buf6, 64,
XBLOCK=64, num_warps=1, num_stages=1)
buf7 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf6, (16, 4), (4, 1), 0),
reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), out=buf7)
return buf5, reinterpret_tensor(buf7, (4, 4, 4), (16, 4, 1), 0
), primals_2, reinterpret_tensor(primals_4, (64, 4), (4, 1), 0
), buf2, buf5, reinterpret_tensor(buf6, (16, 4), (4, 1), 0
), primals_6, primals_5
def get_activation_fn(name):
"""Returns a callable activation function from torch."""
if name in (None, 'linear'):
return lambda x: x
elif name in ('sigmoid', 'tanh'):
return getattr(torch, name)
else:
return getattr(F, name)
class DotAttention(nn.Module):
"""Attention layer with dot product."""
def __init__(self, ctx_dim, hid_dim, att_bottleneck='ctx',
transform_ctx=True, att_activ='tanh', temp=1.0, ctx2hid=True,
mlp_bias=None):
super().__init__()
self.ctx_dim = ctx_dim
self.hid_dim = hid_dim
self._ctx2hid = ctx2hid
self.temperature = temp
self.activ = get_activation_fn(att_activ)
if isinstance(att_bottleneck, int):
self.mid_dim = att_bottleneck
else:
self.mid_dim = getattr(self, '{}_dim'.format(att_bottleneck))
self.hid2ctx = nn.Linear(self.hid_dim, self.mid_dim, bias=False)
if transform_ctx or self.mid_dim != self.ctx_dim:
self.ctx2ctx = nn.Linear(self.ctx_dim, self.mid_dim, bias=False)
else:
self.ctx2ctx = lambda x: x
if self._ctx2hid:
self.ctx2hid = nn.Linear(self.ctx_dim, self.hid_dim, bias=False)
else:
self.ctx2hid = lambda x: x
def forward(self, hid, ctx, ctx_mask=None):
"""Computes attention probabilities and final context using
decoder's hidden state and source annotations.
Arguments:
hid(Tensor): A set of decoder hidden states of shape `T*B*H`
where `T` == 1, `B` is batch dim and `H` is hidden state dim.
ctx(Tensor): A set of annotations of shape `S*B*C` where `S`
is the source timestep dim, `B` is batch dim and `C`
is annotation dim.
ctx_mask(FloatTensor): A binary mask of shape `S*B` with zeroes
in the padded positions.
Returns:
scores(Tensor): A tensor of shape `S*B` containing normalized
attention scores for each position and sample.
z_t(Tensor): A tensor of shape `B*H` containing the final
attended context vector for this target decoding timestep.
Notes:
This will only work when `T==1` for now.
"""
ctx_ = self.ctx2ctx(ctx)
hid_ = self.hid2ctx(hid)
scores = torch.bmm(hid_.permute(1, 0, 2), ctx_.permute(1, 2, 0)).div(
self.temperature).squeeze(1).t()
if ctx_mask is not None:
scores.masked_fill_((1 - ctx_mask).byte(), -100000000.0)
alpha = F.softmax(scores, dim=0)
return alpha, self.ctx2hid((alpha.unsqueeze(-1) * ctx).sum(0))
class MLPAttentionNew(DotAttention):
"""Attention layer with feed-forward layer."""
def __init__(self, ctx_dim, hid_dim, att_bottleneck='ctx',
transform_ctx=True, att_activ='tanh', mlp_bias=False, temp=1.0,
ctx2hid=True):
super().__init__(ctx_dim, hid_dim, att_bottleneck, transform_ctx,
att_activ, temp, ctx2hid)
if mlp_bias:
self.bias = nn.Parameter(torch.Tensor(self.mid_dim))
self.bias.data.zero_()
else:
self.register_parameter('bias', None)
self.mlp = nn.Linear(self.mid_dim, 1, bias=False)
def forward(self, input_0, input_1):
primals_1 = self.hid2ctx.weight
primals_3 = self.ctx2ctx.weight
primals_6 = self.ctx2hid.weight
primals_5 = self.mlp.weight
primals_2 = input_0
primals_4 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6])
return output[0], output[1]
|
fmetze/nmtpytorch
|
MLPAttention
| false | 12,372 |
[
"MIT"
] | 0 |
658a39a2c50e4e9e2fde69b520ddac7efc083257
|
https://github.com/fmetze/nmtpytorch/tree/658a39a2c50e4e9e2fde69b520ddac7efc083257
|
BipolarSigmoid
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/tx/ctxrys2qrzzuoqfwp26krojzrmm5vxaskh6edcvo4j6ozm65jaux.py
# Topologically Sorted Source Nodes: [neg, exp, sub, neg_1, exp_1, add, truediv], Original ATen: [aten.neg, aten.exp, aten.rsub, aten.add, aten.div]
# Source node to ATen node mapping:
# add => add
# exp => exp
# exp_1 => exp_1
# neg => neg
# neg_1 => neg_1
# sub => sub
# truediv => div
# Graph fragment:
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%arg0_1,), kwargs = {})
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%neg,), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (1.0, %exp), kwargs = {})
# %neg_1 : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%arg0_1,), kwargs = {})
# %exp_1 : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%neg_1,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%exp_1, 1.0), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub, %add), kwargs = {})
triton_poi_fused_add_div_exp_neg_rsub_0 = async_compile.triton('triton_poi_fused_add_div_exp_neg_rsub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_exp_neg_rsub_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_exp_neg_rsub_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = -tmp0
tmp2 = tl_math.exp(tmp1)
tmp3 = 1.0
tmp4 = tmp3 - tmp2
tmp5 = tmp2 + tmp3
tmp6 = tmp4 / tmp5
tl.store(out_ptr0 + (x0), tmp6, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [neg, exp, sub, neg_1, exp_1, add, truediv], Original ATen: [aten.neg, aten.exp, aten.rsub, aten.add, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_exp_neg_rsub_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class BipolarSigmoid(nn.Module):
def forward(self, x):
return (1.0 - torch.exp(-x)) / (1.0 + torch.exp(-x))
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_div_exp_neg_rsub_0(in_ptr0, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = -tmp0
tmp2 = tl_math.exp(tmp1)
tmp3 = 1.0
tmp4 = tmp3 - tmp2
tmp5 = tmp2 + tmp3
tmp6 = tmp4 / tmp5
tl.store(out_ptr0 + x0, tmp6, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_exp_neg_rsub_0[grid(256)](arg0_1, buf0,
256, XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class BipolarSigmoidNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
BipolarSigmoid
| false | 12,373 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
InverseSqrt
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/at/cato32233qv3tqtf6htbqu3dr22uwifzkajkou5ew6uihfue6mhe.py
# Topologically Sorted Source Nodes: [mul, mul_1, add, sqrt, truediv], Original ATen: [aten.mul, aten.add, aten.sqrt, aten.div]
# Source node to ATen node mapping:
# add => add
# mul => mul
# mul_1 => mul_1
# sqrt => sqrt
# truediv => div
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, 1.0), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%mul, %arg0_1), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul_1, 1.0), kwargs = {})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%add,), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%arg0_1, %sqrt), kwargs = {})
triton_poi_fused_add_div_mul_sqrt_0 = async_compile.triton('triton_poi_fused_add_div_mul_sqrt_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_mul_sqrt_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_mul_sqrt_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp3 = tmp2 * tmp0
tmp4 = tmp3 + tmp1
tmp5 = libdevice.sqrt(tmp4)
tmp6 = tmp0 / tmp5
tl.store(out_ptr0 + (x0), tmp6, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, mul_1, add, sqrt, truediv], Original ATen: [aten.mul, aten.add, aten.sqrt, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_mul_sqrt_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class InverseSqrt(nn.Module):
def forward(self, x, alpha=1.0):
return x / torch.sqrt(1.0 + alpha * x * x)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_div_mul_sqrt_0(in_ptr0, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 1.0
tmp2 = tmp0 * tmp1
tmp3 = tmp2 * tmp0
tmp4 = tmp3 + tmp1
tmp5 = libdevice.sqrt(tmp4)
tmp6 = tmp0 / tmp5
tl.store(out_ptr0 + x0, tmp6, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_mul_sqrt_0[grid(256)](arg0_1, buf0, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class InverseSqrtNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
InverseSqrt
| false | 12,374 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
Gaussian
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/cu/ccuyglcilujzp5wkgsbpgat45t262w7p7xvi25g7vvmzlry4xegy.py
# Topologically Sorted Source Nodes: [neg, mul, truediv, exp], Original ATen: [aten.neg, aten.mul, aten.div, aten.exp]
# Source node to ATen node mapping:
# exp => exp
# mul => mul
# neg => neg
# truediv => div
# Graph fragment:
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%arg0_1,), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%neg, %arg0_1), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%mul, 2.0), kwargs = {})
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%div,), kwargs = {})
triton_poi_fused_div_exp_mul_neg_0 = async_compile.triton('triton_poi_fused_div_exp_mul_neg_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_exp_mul_neg_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_exp_mul_neg_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = -tmp0
tmp2 = tmp1 * tmp0
tmp3 = 0.5
tmp4 = tmp2 * tmp3
tmp5 = tl_math.exp(tmp4)
tl.store(out_ptr0 + (x0), tmp5, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [neg, mul, truediv, exp], Original ATen: [aten.neg, aten.mul, aten.div, aten.exp]
stream0 = get_raw_stream(0)
triton_poi_fused_div_exp_mul_neg_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class Gaussian(nn.Module):
def forward(self, x):
return torch.exp(-x * x / 2.0)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_div_exp_mul_neg_0(in_ptr0, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = -tmp0
tmp2 = tmp1 * tmp0
tmp3 = 0.5
tmp4 = tmp2 * tmp3
tmp5 = tl_math.exp(tmp4)
tl.store(out_ptr0 + x0, tmp5, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_div_exp_mul_neg_0[grid(256)](arg0_1, buf0, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class GaussianNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
Gaussian
| false | 12,375 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
BentIdentity
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/2d/c2dbbcdmapihcui2dnfgaqd2axwpdnfhkaew3twbjfflllmgd5at.py
# Topologically Sorted Source Nodes: [mul, add, sqrt, sub, truediv, add_1], Original ATen: [aten.mul, aten.add, aten.sqrt, aten.sub, aten.div]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# mul => mul
# sqrt => sqrt
# sub => sub
# truediv => div
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, %arg0_1), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul, 1.0), kwargs = {})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%add,), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sqrt, 1.0), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub, 2.0), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arg0_1, %div), kwargs = {})
triton_poi_fused_add_div_mul_sqrt_sub_0 = async_compile.triton('triton_poi_fused_add_div_mul_sqrt_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_mul_sqrt_sub_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_mul_sqrt_sub_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tmp0 * tmp0
tmp2 = 1.0
tmp3 = tmp1 + tmp2
tmp4 = libdevice.sqrt(tmp3)
tmp5 = tmp4 - tmp2
tmp6 = 0.5
tmp7 = tmp5 * tmp6
tmp8 = tmp0 + tmp7
tl.store(out_ptr0 + (x0), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, add, sqrt, sub, truediv, add_1], Original ATen: [aten.mul, aten.add, aten.sqrt, aten.sub, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_mul_sqrt_sub_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class BentIdentity(nn.Module):
def forward(self, x, alpha=1.0):
return x + (torch.sqrt(1.0 + x * x) - 1.0) / 2.0
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_div_mul_sqrt_sub_0(in_ptr0, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tmp0 * tmp0
tmp2 = 1.0
tmp3 = tmp1 + tmp2
tmp4 = libdevice.sqrt(tmp3)
tmp5 = tmp4 - tmp2
tmp6 = 0.5
tmp7 = tmp5 * tmp6
tmp8 = tmp0 + tmp7
tl.store(out_ptr0 + x0, tmp8, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_mul_sqrt_sub_0[grid(256)](arg0_1, buf0,
256, XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class BentIdentityNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
BentIdentity
| false | 12,376 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
Softmax
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/tp/ctpmhw5dm5lilgirvytgpux5dv4fc3lrjtb4iskbaeenz6etvmdg.py
# Topologically Sorted Source Nodes: [y, sum_1, truediv], Original ATen: [aten.exp, aten.sum, aten.div]
# Source node to ATen node mapping:
# sum_1 => sum_1
# truediv => div
# y => exp
# Graph fragment:
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%arg0_1,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [0]), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused_div_exp_sum_0 = async_compile.triton('triton_poi_fused_div_exp_sum_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_exp_sum_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_exp_sum_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp2 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp1 = tl_math.exp(tmp0)
tmp3 = tl_math.exp(tmp2)
tmp5 = tl_math.exp(tmp4)
tmp6 = tmp3 + tmp5
tmp8 = tl_math.exp(tmp7)
tmp9 = tmp6 + tmp8
tmp11 = tl_math.exp(tmp10)
tmp12 = tmp9 + tmp11
tmp13 = tmp1 / tmp12
tl.store(out_ptr0 + (x2), tmp13, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [y, sum_1, truediv], Original ATen: [aten.exp, aten.sum, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_div_exp_sum_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class Softmax(nn.Module):
def forward(self, x):
y = torch.exp(x)
return y / torch.sum(y, dim=0)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_div_exp_sum_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp2 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp1 = tl_math.exp(tmp0)
tmp3 = tl_math.exp(tmp2)
tmp5 = tl_math.exp(tmp4)
tmp6 = tmp3 + tmp5
tmp8 = tl_math.exp(tmp7)
tmp9 = tmp6 + tmp8
tmp11 = tl_math.exp(tmp10)
tmp12 = tmp9 + tmp11
tmp13 = tmp1 / tmp12
tl.store(out_ptr0 + x2, tmp13, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_div_exp_sum_0[grid(256)](arg0_1, buf0, 256, XBLOCK
=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class SoftmaxNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
Softmax
| false | 12,377 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
BasicMotionEncoder
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/7k/c7klnf45hofe2rzqy3twad4emot2skmrgdktcktkddqzesb47nny.py
# Unsorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
triton_poi_fused_0 = async_compile.triton('triton_poi_fused_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256, 4096], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_0(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 144
xnumel = 4096
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, YBLOCK], True, tl.int1)
x2 = xindex
y3 = yindex
y0 = yindex % 36
y1 = (yindex // 36)
tmp0 = tl.load(in_ptr0 + (x2 + (4096*y3)), ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (y0 + (36*x2) + (147456*y1)), tmp0, ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/cg/ccgjbjra22jzcgrej4cregu4qnsuv4valsbpzazlnyx6d4zadmny.py
# Unsorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
triton_poi_fused_1 = async_compile.triton('triton_poi_fused_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4096, 16], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 4096
xnumel = 9
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = tl.full([XBLOCK, YBLOCK], True, tl.int1)
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 64
y1 = (yindex // 64)
tmp0 = tl.load(in_ptr0 + (x2 + (9*y3)), xmask, eviction_policy='evict_last')
tl.store(out_ptr0 + (y0 + (64*x2) + (576*y1)), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6c/c6cncuvxxhqnmxps5zcsamewoq74gve7wfzt5tsiyaefaumvnizq.py
# Unsorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
triton_poi_fused_2 = async_compile.triton('triton_poi_fused_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128, 64], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 128
xnumel = 49
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 2
y1 = (yindex // 2)
tmp0 = tl.load(in_ptr0 + (x2 + (49*y3)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (y0 + (2*x2) + (98*y1)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/64/c64ftfrtg7wzrrafqmf23jsfazdeeslgcihlzd3ir6gzqnsqouzt.py
# Unsorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
triton_poi_fused_3 = async_compile.triton('triton_poi_fused_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8, 4096], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_3(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 8
xnumel = 4096
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, YBLOCK], True, tl.int1)
x2 = xindex
y3 = yindex
y0 = yindex % 2
y1 = (yindex // 2)
tmp0 = tl.load(in_ptr0 + (x2 + (4096*y3)), ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (y0 + (2*x2) + (8192*y1)), tmp0, ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/zz/czzry3mptvotkg2o63ggvxe23r6aakp7p6x6j4z6chr5g2jg64yo.py
# Unsorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
triton_poi_fused_4 = async_compile.triton('triton_poi_fused_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384, 16], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_4(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16128
xnumel = 9
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 128
y1 = (yindex // 128)
tmp0 = tl.load(in_ptr0 + (x2 + (9*y3)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (y0 + (128*x2) + (1152*y1)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/7c/c7caq6stn5xhqphn2xnmwbpvxspyfj5wahntqw4tlpltw5xu6ktg.py
# Topologically Sorted Source Nodes: [conv2d, cor], Original ATen: [aten.convolution, aten.relu]
# Source node to ATen node mapping:
# conv2d => convolution
# cor => relu
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution,), kwargs = {})
triton_poi_fused_convolution_relu_5 = async_compile.triton('triton_poi_fused_convolution_relu_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1048576],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_5', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_5(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1048576
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_out_ptr0 + (x2), None)
tmp1 = tl.load(in_ptr0 + (x0), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x2), tmp4, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/cg/ccgem2d3minyz4cphcb5m5gr254xubtv62ypjgbd2mjaj7ew6ps5.py
# Topologically Sorted Source Nodes: [cor_flo], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cor_flo => cat
# Graph fragment:
# %cat : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%relu_1, %relu_3], 1), kwargs = {})
triton_poi_fused_cat_6 = async_compile.triton('triton_poi_fused_cat_6', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[2097152],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_6', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_6(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 2097152
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x0 = xindex % 128
x1 = (xindex // 128)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 64, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((64*x1) + x0), tmp4, eviction_policy='evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + (x0), tmp4, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full([1], 0, tl.int32)
tmp9 = triton_helpers.maximum(tmp8, tmp7)
tmp10 = tl.full(tmp9.shape, 0.0, tmp9.dtype)
tmp11 = tl.where(tmp4, tmp9, tmp10)
tmp12 = tmp0 >= tmp3
tmp13 = tl.full([1], 128, tl.int64)
tmp14 = tmp0 < tmp13
tmp15 = tl.load(in_ptr2 + ((64*x1) + ((-64) + x0)), tmp12, eviction_policy='evict_last', other=0.0)
tmp16 = tl.load(in_ptr3 + ((-64) + x0), tmp12, eviction_policy='evict_last', other=0.0)
tmp17 = tmp15 + tmp16
tmp18 = triton_helpers.maximum(tmp8, tmp17)
tmp19 = tl.full(tmp18.shape, 0.0, tmp18.dtype)
tmp20 = tl.where(tmp12, tmp18, tmp19)
tmp21 = tl.where(tmp4, tmp11, tmp20)
tl.store(out_ptr0 + (x2), tmp21, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6h/c6hm7sfjm5znh6zaavrhemfdbagmwselk6wdf455zjm4m6hri6gr.py
# Topologically Sorted Source Nodes: [cat_1], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cat_1 => cat_1
# Graph fragment:
# %cat_1 : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%relu_4, %primals_8], 1), kwargs = {})
triton_poi_fused_cat_7 = async_compile.triton('triton_poi_fused_cat_7', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[2097152],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_7', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_7(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 2097152
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x1 = (xindex // 4096) % 128
x0 = xindex % 4096
x2 = (xindex // 524288)
x3 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 126, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((126*x0) + (516096*x2) + x1), tmp4, eviction_policy='evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + (x1), tmp4, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full([1], 0, tl.int32)
tmp9 = triton_helpers.maximum(tmp8, tmp7)
tmp10 = tl.full(tmp9.shape, 0.0, tmp9.dtype)
tmp11 = tl.where(tmp4, tmp9, tmp10)
tmp12 = tmp0 >= tmp3
tmp13 = tl.full([1], 128, tl.int64)
tmp14 = tmp0 < tmp13
tmp15 = tl.load(in_ptr2 + ((2*x0) + (8192*x2) + ((-126) + x1)), tmp12, eviction_policy='evict_last', other=0.0)
tmp16 = tl.where(tmp4, tmp11, tmp15)
tl.store(out_ptr0 + (x3), tmp16, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/oq/coqahmle73rziabtyvejxffgqwjt4fpmsrxv2jzdeffn5sl7uomz.py
# Topologically Sorted Source Nodes: [conv2d_4, out], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# conv2d_4 => convolution_4
# out => relu_4
# Graph fragment:
# %convolution_4 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%cat, %primals_11, %primals_12, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
# %relu_4 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_4,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu_4, 0), kwargs = {})
triton_poi_fused_convolution_relu_threshold_backward_8 = async_compile.triton('triton_poi_fused_convolution_relu_threshold_backward_8', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[2097152],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i1', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_threshold_backward_8', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_8(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 2064384
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 126
tmp0 = tl.load(in_ptr0 + (x2), None)
tmp1 = tl.load(in_ptr1 + (x0), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x2), tmp6, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/25/c25ulitk3enogdxsfm25azyl53hdfygxtei67fnidtwtwfmjifcq.py
# Topologically Sorted Source Nodes: [conv2d_3, flo_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# conv2d_3 => convolution_3
# flo_1 => relu_3
# Graph fragment:
# %convolution_3 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%relu_2, %primals_9, %primals_10, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
# %relu_3 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_3,), kwargs = {})
# %le_1 : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu_3, 0), kwargs = {})
triton_poi_fused_convolution_relu_threshold_backward_9 = async_compile.triton('triton_poi_fused_convolution_relu_threshold_backward_9', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1048576],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i1', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_threshold_backward_9', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_9(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1048576
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + (x2), None)
tmp1 = tl.load(in_ptr1 + (x0), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x2), tmp6, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12 = args
args.clear()
assert_size_stride(primals_1, (64, 36, 1, 1), (36, 1, 1, 1))
assert_size_stride(primals_2, (64, ), (1, ))
assert_size_stride(primals_3, (4, 36, 64, 64), (147456, 4096, 64, 1))
assert_size_stride(primals_4, (64, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_5, (64, ), (1, ))
assert_size_stride(primals_6, (64, 2, 7, 7), (98, 49, 7, 1))
assert_size_stride(primals_7, (64, ), (1, ))
assert_size_stride(primals_8, (4, 2, 64, 64), (8192, 4096, 64, 1))
assert_size_stride(primals_9, (64, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_10, (64, ), (1, ))
assert_size_stride(primals_11, (126, 128, 3, 3), (1152, 9, 3, 1))
assert_size_stride(primals_12, (126, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 36, 64, 64), (147456, 1, 2304, 36), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
stream0 = get_raw_stream(0)
triton_poi_fused_0.run(primals_3, buf0, 144, 4096, grid=grid(144, 4096), stream=stream0)
del primals_3
buf1 = empty_strided_cuda((64, 64, 3, 3), (576, 1, 192, 64), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
triton_poi_fused_1.run(primals_4, buf1, 4096, 9, grid=grid(4096, 9), stream=stream0)
del primals_4
buf2 = empty_strided_cuda((64, 2, 7, 7), (98, 1, 14, 2), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
triton_poi_fused_2.run(primals_6, buf2, 128, 49, grid=grid(128, 49), stream=stream0)
del primals_6
buf3 = empty_strided_cuda((4, 2, 64, 64), (8192, 1, 128, 2), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
triton_poi_fused_3.run(primals_8, buf3, 8, 4096, grid=grid(8, 4096), stream=stream0)
del primals_8
buf4 = empty_strided_cuda((64, 64, 3, 3), (576, 1, 192, 64), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
triton_poi_fused_1.run(primals_9, buf4, 4096, 9, grid=grid(4096, 9), stream=stream0)
del primals_9
buf5 = empty_strided_cuda((126, 128, 3, 3), (1152, 1, 384, 128), torch.float32)
# Unsorted Source Nodes: [], Original ATen: []
triton_poi_fused_4.run(primals_11, buf5, 16128, 9, grid=grid(16128, 9), stream=stream0)
del primals_11
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
buf6 = extern_kernels.convolution(buf0, primals_1, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 64, 64, 64), (262144, 1, 4096, 64))
buf7 = buf6; del buf6 # reuse
# Topologically Sorted Source Nodes: [conv2d, cor], Original ATen: [aten.convolution, aten.relu]
triton_poi_fused_convolution_relu_5.run(buf7, primals_2, 1048576, grid=grid(1048576), stream=stream0)
del primals_2
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
buf8 = extern_kernels.convolution(buf7, buf1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf8, (4, 64, 64, 64), (262144, 1, 4096, 64))
# Topologically Sorted Source Nodes: [conv2d_2], Original ATen: [aten.convolution]
buf9 = extern_kernels.convolution(buf3, buf2, stride=(1, 1), padding=(3, 3), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf9, (4, 64, 64, 64), (262144, 1, 4096, 64))
buf10 = buf9; del buf9 # reuse
# Topologically Sorted Source Nodes: [conv2d_2, flo], Original ATen: [aten.convolution, aten.relu]
triton_poi_fused_convolution_relu_5.run(buf10, primals_7, 1048576, grid=grid(1048576), stream=stream0)
del primals_7
# Topologically Sorted Source Nodes: [conv2d_3], Original ATen: [aten.convolution]
buf11 = extern_kernels.convolution(buf10, buf4, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf11, (4, 64, 64, 64), (262144, 1, 4096, 64))
buf12 = empty_strided_cuda((4, 128, 64, 64), (524288, 1, 8192, 128), torch.float32)
# Topologically Sorted Source Nodes: [cor_flo], Original ATen: [aten.cat]
triton_poi_fused_cat_6.run(buf8, primals_5, buf11, primals_10, buf12, 2097152, grid=grid(2097152), stream=stream0)
# Topologically Sorted Source Nodes: [conv2d_4], Original ATen: [aten.convolution]
buf13 = extern_kernels.convolution(buf12, buf5, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf13, (4, 126, 64, 64), (516096, 1, 8064, 126))
buf14 = empty_strided_cuda((4, 128, 64, 64), (524288, 4096, 64, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat_1], Original ATen: [aten.cat]
triton_poi_fused_cat_7.run(buf13, primals_12, buf3, buf14, 2097152, grid=grid(2097152), stream=stream0)
buf15 = empty_strided_cuda((4, 126, 64, 64), (516096, 1, 8064, 126), torch.bool)
# Topologically Sorted Source Nodes: [conv2d_4, out], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
triton_poi_fused_convolution_relu_threshold_backward_8.run(buf13, primals_12, buf15, 2064384, grid=grid(2064384), stream=stream0)
del buf13
del primals_12
buf16 = empty_strided_cuda((4, 64, 64, 64), (262144, 1, 4096, 64), torch.bool)
# Topologically Sorted Source Nodes: [conv2d_3, flo_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
triton_poi_fused_convolution_relu_threshold_backward_9.run(buf11, primals_10, buf16, 1048576, grid=grid(1048576), stream=stream0)
del buf11
del primals_10
buf17 = empty_strided_cuda((4, 64, 64, 64), (262144, 1, 4096, 64), torch.bool)
# Topologically Sorted Source Nodes: [conv2d_1, cor_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
triton_poi_fused_convolution_relu_threshold_backward_9.run(buf8, primals_5, buf17, 1048576, grid=grid(1048576), stream=stream0)
del buf8
del primals_5
return (buf14, primals_1, buf0, buf1, buf2, buf3, buf4, buf5, buf7, buf10, buf12, buf15, buf16, buf17, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((64, 36, 1, 1), (36, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((64, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 36, 64, 64), (147456, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((64, 64, 3, 3), (576, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((64, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((64, 2, 7, 7), (98, 49, 7, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((64, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, 2, 64, 64), (8192, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((64, 64, 3, 3), (576, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((64, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((126, 128, 3, 3), (1152, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_12 = rand_strided((126, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from _paritybench_helpers import _mock_config
import torch
import torch.nn.functional as F
import torch.nn as nn
class BasicMotionEncoder(nn.Module):
def __init__(self, args):
super(BasicMotionEncoder, self).__init__()
self.args = args
cor_planes = args.corr_levels * (2 * args.corr_radius + 1)
self.convc1 = nn.Conv2d(cor_planes, 64, 1, padding=0)
self.convc2 = nn.Conv2d(64, 64, 3, padding=1)
self.convf1 = nn.Conv2d(2, 64, 7, padding=3)
self.convf2 = nn.Conv2d(64, 64, 3, padding=1)
self.conv = nn.Conv2d(64 + 64, 128 - 2, 3, padding=1)
def forward(self, flow, corr):
cor = F.relu(self.convc1(corr))
cor = F.relu(self.convc2(cor))
flo = F.relu(self.convf1(flow))
flo = F.relu(self.convf2(flo))
cor_flo = torch.cat([cor, flo], dim=1)
out = F.relu(self.conv(cor_flo))
return torch.cat([out, flow], dim=1)
def get_inputs():
return [torch.rand([4, 2, 64, 64]), torch.rand([4, 36, 64, 64])]
def get_init_inputs():
return [[], {'args': _mock_config(corr_levels=4, corr_radius=4)}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_0(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 144
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, YBLOCK], True, tl.int1)
x2 = xindex
y3 = yindex
y0 = yindex % 36
y1 = yindex // 36
tmp0 = tl.load(in_ptr0 + (x2 + 4096 * y3), ymask, eviction_policy=
'evict_last')
tl.store(out_ptr0 + (y0 + 36 * x2 + 147456 * y1), tmp0, ymask)
@triton.jit
def triton_poi_fused_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
xnumel = 9
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
tl.full([XBLOCK, YBLOCK], True, tl.int1)
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 64
y1 = yindex // 64
tmp0 = tl.load(in_ptr0 + (x2 + 9 * y3), xmask, eviction_policy='evict_last'
)
tl.store(out_ptr0 + (y0 + 64 * x2 + 576 * y1), tmp0, xmask)
@triton.jit
def triton_poi_fused_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 128
xnumel = 49
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 2
y1 = yindex // 2
tmp0 = tl.load(in_ptr0 + (x2 + 49 * y3), xmask & ymask, eviction_policy
='evict_last')
tl.store(out_ptr0 + (y0 + 2 * x2 + 98 * y1), tmp0, xmask & ymask)
@triton.jit
def triton_poi_fused_3(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 8
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, YBLOCK], True, tl.int1)
x2 = xindex
y3 = yindex
y0 = yindex % 2
y1 = yindex // 2
tmp0 = tl.load(in_ptr0 + (x2 + 4096 * y3), ymask, eviction_policy=
'evict_last')
tl.store(out_ptr0 + (y0 + 2 * x2 + 8192 * y1), tmp0, ymask)
@triton.jit
def triton_poi_fused_4(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 16128
xnumel = 9
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 128
y1 = yindex // 128
tmp0 = tl.load(in_ptr0 + (x2 + 9 * y3), xmask & ymask, eviction_policy=
'evict_last')
tl.store(out_ptr0 + (y0 + 128 * x2 + 1152 * y1), tmp0, xmask & ymask)
@triton.jit
def triton_poi_fused_convolution_relu_5(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_out_ptr0 + x2, None)
tmp1 = tl.load(in_ptr0 + x0, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x2, tmp4, None)
@triton.jit
def triton_poi_fused_cat_6(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x0 = xindex % 128
x1 = xindex // 128
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 64, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (64 * x1 + x0), tmp4, eviction_policy=
'evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + x0, tmp4, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full([1], 0, tl.int32)
tmp9 = triton_helpers.maximum(tmp8, tmp7)
tmp10 = tl.full(tmp9.shape, 0.0, tmp9.dtype)
tmp11 = tl.where(tmp4, tmp9, tmp10)
tmp12 = tmp0 >= tmp3
tl.full([1], 128, tl.int64)
tmp15 = tl.load(in_ptr2 + (64 * x1 + (-64 + x0)), tmp12,
eviction_policy='evict_last', other=0.0)
tmp16 = tl.load(in_ptr3 + (-64 + x0), tmp12, eviction_policy=
'evict_last', other=0.0)
tmp17 = tmp15 + tmp16
tmp18 = triton_helpers.maximum(tmp8, tmp17)
tmp19 = tl.full(tmp18.shape, 0.0, tmp18.dtype)
tmp20 = tl.where(tmp12, tmp18, tmp19)
tmp21 = tl.where(tmp4, tmp11, tmp20)
tl.store(out_ptr0 + x2, tmp21, None)
@triton.jit
def triton_poi_fused_cat_7(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x1 = xindex // 4096 % 128
x0 = xindex % 4096
x2 = xindex // 524288
x3 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 126, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (126 * x0 + 516096 * x2 + x1), tmp4,
eviction_policy='evict_last', other=0.0)
tmp6 = tl.load(in_ptr1 + x1, tmp4, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full([1], 0, tl.int32)
tmp9 = triton_helpers.maximum(tmp8, tmp7)
tmp10 = tl.full(tmp9.shape, 0.0, tmp9.dtype)
tmp11 = tl.where(tmp4, tmp9, tmp10)
tmp12 = tmp0 >= tmp3
tl.full([1], 128, tl.int64)
tmp15 = tl.load(in_ptr2 + (2 * x0 + 8192 * x2 + (-126 + x1)), tmp12,
eviction_policy='evict_last', other=0.0)
tmp16 = tl.where(tmp4, tmp11, tmp15)
tl.store(out_ptr0 + x3, tmp16, None)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_8(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 126
tmp0 = tl.load(in_ptr0 + x2, None)
tmp1 = tl.load(in_ptr1 + x0, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + x2, tmp6, None)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_9(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + x2, None)
tmp1 = tl.load(in_ptr1 + x0, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + x2, tmp6, None)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11, primals_12
) = args
args.clear()
assert_size_stride(primals_1, (64, 36, 1, 1), (36, 1, 1, 1))
assert_size_stride(primals_2, (64,), (1,))
assert_size_stride(primals_3, (4, 36, 64, 64), (147456, 4096, 64, 1))
assert_size_stride(primals_4, (64, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_5, (64,), (1,))
assert_size_stride(primals_6, (64, 2, 7, 7), (98, 49, 7, 1))
assert_size_stride(primals_7, (64,), (1,))
assert_size_stride(primals_8, (4, 2, 64, 64), (8192, 4096, 64, 1))
assert_size_stride(primals_9, (64, 64, 3, 3), (576, 9, 3, 1))
assert_size_stride(primals_10, (64,), (1,))
assert_size_stride(primals_11, (126, 128, 3, 3), (1152, 9, 3, 1))
assert_size_stride(primals_12, (126,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 36, 64, 64), (147456, 1, 2304, 36),
torch.float32)
get_raw_stream(0)
triton_poi_fused_0[grid(144, 4096)](primals_3, buf0, 144, 4096,
XBLOCK=32, YBLOCK=32, num_warps=4, num_stages=1)
del primals_3
buf1 = empty_strided_cuda((64, 64, 3, 3), (576, 1, 192, 64), torch.
float32)
triton_poi_fused_1[grid(4096, 9)](primals_4, buf1, 4096, 9, XBLOCK=
16, YBLOCK=64, num_warps=4, num_stages=1)
del primals_4
buf2 = empty_strided_cuda((64, 2, 7, 7), (98, 1, 14, 2), torch.float32)
triton_poi_fused_2[grid(128, 49)](primals_6, buf2, 128, 49, XBLOCK=
32, YBLOCK=32, num_warps=4, num_stages=1)
del primals_6
buf3 = empty_strided_cuda((4, 2, 64, 64), (8192, 1, 128, 2), torch.
float32)
triton_poi_fused_3[grid(8, 4096)](primals_8, buf3, 8, 4096, XBLOCK=
128, YBLOCK=8, num_warps=4, num_stages=1)
del primals_8
buf4 = empty_strided_cuda((64, 64, 3, 3), (576, 1, 192, 64), torch.
float32)
triton_poi_fused_1[grid(4096, 9)](primals_9, buf4, 4096, 9, XBLOCK=
16, YBLOCK=64, num_warps=4, num_stages=1)
del primals_9
buf5 = empty_strided_cuda((126, 128, 3, 3), (1152, 1, 384, 128),
torch.float32)
triton_poi_fused_4[grid(16128, 9)](primals_11, buf5, 16128, 9,
XBLOCK=16, YBLOCK=64, num_warps=4, num_stages=1)
del primals_11
buf6 = extern_kernels.convolution(buf0, primals_1, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 64, 64, 64), (262144, 1, 4096, 64))
buf7 = buf6
del buf6
triton_poi_fused_convolution_relu_5[grid(1048576)](buf7, primals_2,
1048576, XBLOCK=1024, num_warps=4, num_stages=1)
del primals_2
buf8 = extern_kernels.convolution(buf7, buf1, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf8, (4, 64, 64, 64), (262144, 1, 4096, 64))
buf9 = extern_kernels.convolution(buf3, buf2, stride=(1, 1),
padding=(3, 3), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf9, (4, 64, 64, 64), (262144, 1, 4096, 64))
buf10 = buf9
del buf9
triton_poi_fused_convolution_relu_5[grid(1048576)](buf10, primals_7,
1048576, XBLOCK=1024, num_warps=4, num_stages=1)
del primals_7
buf11 = extern_kernels.convolution(buf10, buf4, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf11, (4, 64, 64, 64), (262144, 1, 4096, 64))
buf12 = empty_strided_cuda((4, 128, 64, 64), (524288, 1, 8192, 128),
torch.float32)
triton_poi_fused_cat_6[grid(2097152)](buf8, primals_5, buf11,
primals_10, buf12, 2097152, XBLOCK=1024, num_warps=4, num_stages=1)
buf13 = extern_kernels.convolution(buf12, buf5, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf13, (4, 126, 64, 64), (516096, 1, 8064, 126))
buf14 = empty_strided_cuda((4, 128, 64, 64), (524288, 4096, 64, 1),
torch.float32)
triton_poi_fused_cat_7[grid(2097152)](buf13, primals_12, buf3,
buf14, 2097152, XBLOCK=512, num_warps=8, num_stages=1)
buf15 = empty_strided_cuda((4, 126, 64, 64), (516096, 1, 8064, 126),
torch.bool)
triton_poi_fused_convolution_relu_threshold_backward_8[grid(2064384)](
buf13, primals_12, buf15, 2064384, XBLOCK=1024, num_warps=4,
num_stages=1)
del buf13
del primals_12
buf16 = empty_strided_cuda((4, 64, 64, 64), (262144, 1, 4096, 64),
torch.bool)
triton_poi_fused_convolution_relu_threshold_backward_9[grid(1048576)](
buf11, primals_10, buf16, 1048576, XBLOCK=1024, num_warps=4,
num_stages=1)
del buf11
del primals_10
buf17 = empty_strided_cuda((4, 64, 64, 64), (262144, 1, 4096, 64),
torch.bool)
triton_poi_fused_convolution_relu_threshold_backward_9[grid(1048576)](
buf8, primals_5, buf17, 1048576, XBLOCK=1024, num_warps=4,
num_stages=1)
del buf8
del primals_5
return (buf14, primals_1, buf0, buf1, buf2, buf3, buf4, buf5, buf7,
buf10, buf12, buf15, buf16, buf17)
class BasicMotionEncoderNew(nn.Module):
def __init__(self, args):
super(BasicMotionEncoderNew, self).__init__()
self.args = args
cor_planes = args.corr_levels * (2 * args.corr_radius + 1)
self.convc1 = nn.Conv2d(cor_planes, 64, 1, padding=0)
self.convc2 = nn.Conv2d(64, 64, 3, padding=1)
self.convf1 = nn.Conv2d(2, 64, 7, padding=3)
self.convf2 = nn.Conv2d(64, 64, 3, padding=1)
self.conv = nn.Conv2d(64 + 64, 128 - 2, 3, padding=1)
def forward(self, input_0, input_1):
primals_1 = self.convc1.weight
primals_2 = self.convc1.bias
primals_4 = self.convc2.weight
primals_5 = self.convc2.bias
primals_6 = self.convf1.weight
primals_7 = self.convf1.bias
primals_9 = self.convf2.weight
primals_10 = self.convf2.bias
primals_11 = self.conv.weight
primals_12 = self.conv.bias
primals_8 = input_0
primals_3 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11, primals_12])
return output[0]
|
eyecan-ai/RAFT-Stereo
|
BasicMotionEncoder
| false | 12,378 |
[
"MIT"
] | 0 |
dda04d8ca4345922947009cfc6f7deb8aaf2cb67
|
https://github.com/eyecan-ai/RAFT-Stereo/tree/dda04d8ca4345922947009cfc6f7deb8aaf2cb67
|
Sinc
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/e6/ce6t322f2zovhgueyidqf4fawj6rk3xbgauhvwphy6cczoyv65ds.py
# Topologically Sorted Source Nodes: [add, sin, add_1, truediv], Original ATen: [aten.add, aten.sin, aten.div]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# sin => sin
# truediv => div
# Graph fragment:
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arg0_1, 1e-09), kwargs = {})
# %sin : [num_users=1] = call_function[target=torch.ops.aten.sin.default](args = (%add,), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arg0_1, 1e-09), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sin, %add_1), kwargs = {})
triton_poi_fused_add_div_sin_0 = async_compile.triton('triton_poi_fused_add_div_sin_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_sin_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_sin_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 1e-09
tmp2 = tmp0 + tmp1
tmp3 = tl_math.sin(tmp2)
tmp4 = tmp3 / tmp2
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [add, sin, add_1, truediv], Original ATen: [aten.add, aten.sin, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_sin_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class Sinc(nn.Module):
def forward(self, x, epsilon=1e-09):
return torch.sin(x + epsilon) / (x + epsilon)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_div_sin_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 1e-09
tmp2 = tmp0 + tmp1
tmp3 = tl_math.sin(tmp2)
tmp4 = tmp3 / tmp2
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_sin_0[grid(256)](arg0_1, buf0, 256, XBLOCK
=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class SincNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
Sinc
| false | 12,379 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
SinReLU
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/lm/clmgwtsqo7qjdvmwh6a54f7ogim3aexgns2pthyoy7xxeo7iu35i.py
# Topologically Sorted Source Nodes: [sin, relu, add], Original ATen: [aten.sin, aten.relu, aten.add]
# Source node to ATen node mapping:
# add => add
# relu => relu
# sin => sin
# Graph fragment:
# %sin : [num_users=1] = call_function[target=torch.ops.aten.sin.default](args = (%arg0_1,), kwargs = {})
# %relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%arg0_1,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sin, %relu), kwargs = {})
triton_poi_fused_add_relu_sin_0 = async_compile.triton('triton_poi_fused_add_relu_sin_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_relu_sin_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_relu_sin_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl_math.sin(tmp0)
tmp2 = tl.full([1], 0, tl.int32)
tmp3 = triton_helpers.maximum(tmp2, tmp0)
tmp4 = tmp1 + tmp3
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [sin, relu, add], Original ATen: [aten.sin, aten.relu, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_relu_sin_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class SinReLU(nn.Module):
def forward(self, x):
return torch.sin(x) + torch.relu(x)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_relu_sin_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl_math.sin(tmp0)
tmp2 = tl.full([1], 0, tl.int32)
tmp3 = triton_helpers.maximum(tmp2, tmp0)
tmp4 = tmp1 + tmp3
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_relu_sin_0[grid(256)](arg0_1, buf0, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class SinReLUNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
SinReLU
| false | 12,380 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
Atan
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/6k/c6k2dulixmqdqegpc7pqn2gwbrpsqp2axnhmqt3isqoayinv5emy.py
# Topologically Sorted Source Nodes: [atan], Original ATen: [aten.atan]
# Source node to ATen node mapping:
# atan => atan
# Graph fragment:
# %atan : [num_users=1] = call_function[target=torch.ops.aten.atan.default](args = (%arg0_1,), kwargs = {})
triton_poi_fused_atan_0 = async_compile.triton('triton_poi_fused_atan_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_atan_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_atan_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = libdevice.atan(tmp0)
tl.store(out_ptr0 + (x0), tmp1, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [atan], Original ATen: [aten.atan]
stream0 = get_raw_stream(0)
triton_poi_fused_atan_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class Atan(nn.Module):
def forward(self, x):
return torch.atan(x)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_atan_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = libdevice.atan(tmp0)
tl.store(out_ptr0 + x0, tmp1, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_atan_0[grid(256)](arg0_1, buf0, 256, XBLOCK=256,
num_warps=4, num_stages=1)
del arg0_1
return buf0,
class AtanNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
Atan
| false | 12,381 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
CosReLU
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/tr/ctrq3c2557fbgqaj6jdrpt62dxk7r66etmcgw4ofxi7jwf7djelv.py
# Topologically Sorted Source Nodes: [cos, relu, add], Original ATen: [aten.cos, aten.relu, aten.add]
# Source node to ATen node mapping:
# add => add
# cos => cos
# relu => relu
# Graph fragment:
# %cos : [num_users=1] = call_function[target=torch.ops.aten.cos.default](args = (%arg0_1,), kwargs = {})
# %relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%arg0_1,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%cos, %relu), kwargs = {})
triton_poi_fused_add_cos_relu_0 = async_compile.triton('triton_poi_fused_add_cos_relu_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_cos_relu_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_cos_relu_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl_math.cos(tmp0)
tmp2 = tl.full([1], 0, tl.int32)
tmp3 = triton_helpers.maximum(tmp2, tmp0)
tmp4 = tmp1 + tmp3
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [cos, relu, add], Original ATen: [aten.cos, aten.relu, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_cos_relu_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class CosReLU(nn.Module):
def forward(self, x):
return torch.cos(x) + torch.relu(x)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_cos_relu_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl_math.cos(tmp0)
tmp2 = tl.full([1], 0, tl.int32)
tmp3 = triton_helpers.maximum(tmp2, tmp0)
tmp4 = tmp1 + tmp3
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_cos_relu_0[grid(256)](arg0_1, buf0, 256,
XBLOCK=128, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class CosReLUNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
CosReLU
| false | 12,382 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
CoAttention
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/zf/czfm4p56xpruypzwgub55e2whrxr6swk6vmccy46jkzo6sfg34mm.py
# Topologically Sorted Source Nodes: [catted], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# catted => cat
# Graph fragment:
# %cat : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%repeat, %repeat_1], 1), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[512],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 512
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = (xindex // 16) % 8
x1 = (xindex // 4) % 4
x3 = (xindex // 128)
x0 = xindex % 4
x4 = xindex
tmp0 = x2
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x3) + (16*x1) + x2), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + ((4*x3) + (16*x0) + ((-4) + x2)), tmp6 & xmask, eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x4), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/te/cteebxatvsuxuudpg5duimxcpdrm5lwq3i3afq43wpa3erd7oxfz.py
# Topologically Sorted Source Nodes: [conv2d, hidden], Original ATen: [aten.convolution, aten.tanh]
# Source node to ATen node mapping:
# conv2d => convolution
# hidden => tanh
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%cat, %primals_3, %primals_4, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
# %tanh : [num_users=2] = call_function[target=torch.ops.aten.tanh.default](args = (%convolution,), kwargs = {})
triton_poi_fused_convolution_tanh_1 = async_compile.triton('triton_poi_fused_convolution_tanh_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_tanh_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_tanh_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 16) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x3), tmp3, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ac/caca5c3nm47ub7lkcm5xltmycty2iduotmf6iyfmx2paytmzdt24.py
# Topologically Sorted Source Nodes: [dist_1_to_2, softmax_1], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# dist_1_to_2 => amax, exp, sub
# softmax_1 => amax_1, exp_1, sub_1
# Graph fragment:
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%squeeze, [2], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%squeeze, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %amax_1 : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%squeeze, [1], True), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%squeeze, %amax_1), kwargs = {})
# %exp_1 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_1,), kwargs = {})
triton_poi_fused__softmax_2 = async_compile.triton('triton_poi_fused__softmax_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 9, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x1 = (xindex // 4)
x0 = xindex % 4
x3 = (xindex // 16)
tmp0 = tl.load(in_ptr0 + (x4), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (x0 + (16*x3)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (4 + x0 + (16*x3)), xmask, eviction_policy='evict_last')
tmp13 = tl.load(in_ptr0 + (8 + x0 + (16*x3)), xmask, eviction_policy='evict_last')
tmp15 = tl.load(in_ptr0 + (12 + x0 + (16*x3)), xmask, eviction_policy='evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tmp9 = tl_math.exp(tmp8)
tmp12 = triton_helpers.maximum(tmp10, tmp11)
tmp14 = triton_helpers.maximum(tmp12, tmp13)
tmp16 = triton_helpers.maximum(tmp14, tmp15)
tmp17 = tmp0 - tmp16
tmp18 = tl_math.exp(tmp17)
tl.store(out_ptr0 + (x4), tmp9, xmask)
tl.store(out_ptr1 + (x4), tmp18, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/zu/czuvep3dmpmqmhiiliwubh4ghdt2qr27va67sszkua7trziinwov.py
# Topologically Sorted Source Nodes: [dist_1_to_2], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# dist_1_to_2 => div, sum_1
# Graph fragment:
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [2], True), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
triton_poi_fused__softmax_3 = async_compile.triton('triton_poi_fused__softmax_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/gk/cgksnje3jnvnppkznomwzxrmazk4ahkfzjas5hxl647g4l2vp27q.py
# Topologically Sorted Source Nodes: [cat_1], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cat_1 => cat_1
# Graph fragment:
# %cat_1 : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%primals_2, %permute_3], -1), kwargs = {})
triton_poi_fused_cat_4 = async_compile.triton('triton_poi_fused_cat_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x3 = (xindex // 8)
x1 = (xindex // 8) % 4
x2 = (xindex // 32)
x4 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x3) + x0), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + (x2 + (4*((-4) + x0)) + (16*x1)), tmp6 & xmask, eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x4), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/dj/cdjgukk64mq3js7jr5cdzsg7gkg7hzg6uiexzdnyk7oxapn3deus.py
# Topologically Sorted Source Nodes: [seq_1_to_2], Original ATen: [aten.tanh]
# Source node to ATen node mapping:
# seq_1_to_2 => tanh_1
# Graph fragment:
# %tanh_1 : [num_users=1] = call_function[target=torch.ops.aten.tanh.default](args = (%view_4,), kwargs = {})
triton_poi_fused_tanh_5 = async_compile.triton('triton_poi_fused_tanh_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_tanh_5', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_tanh_5(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x2), tmp3, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/cg/ccg5e776j77ye72qtmo5nfcxjaz6zv34474xpm34f62r6hfxzo6g.py
# Topologically Sorted Source Nodes: [softmax_1], Original ATen: [aten._softmax]
# Source node to ATen node mapping:
# softmax_1 => div_1, sum_2
# Graph fragment:
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_1, [1], True), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp_1, %sum_2), kwargs = {})
triton_poi_fused__softmax_6 = async_compile.triton('triton_poi_fused__softmax_6', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_6', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_6(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 4
x2 = (xindex // 16)
tmp0 = tl.load(in_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x0 + (16*x2)), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (4 + x0 + (16*x2)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (8 + x0 + (16*x2)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (12 + x0 + (16*x2)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x3), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_3, (4, 8, 1, 1), (8, 1, 1, 1))
assert_size_stride(primals_4, (4, ), (1, ))
assert_size_stride(primals_5, (1, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_6, (4, 8), (8, 1))
assert_size_stride(primals_7, (4, ), (1, ))
assert_size_stride(primals_8, (4, 8), (8, 1))
assert_size_stride(primals_9, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 8, 4, 4), (128, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [catted], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(primals_1, primals_2, buf0, 512, grid=grid(512), stream=stream0)
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(buf0, primals_3, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 4, 4, 4), (64, 16, 4, 1))
buf2 = buf1; del buf1 # reuse
# Topologically Sorted Source Nodes: [conv2d, hidden], Original ATen: [aten.convolution, aten.tanh]
triton_poi_fused_convolution_tanh_1.run(buf2, primals_4, 256, grid=grid(256), stream=stream0)
del primals_4
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
buf3 = extern_kernels.convolution(buf2, primals_5, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 1, 4, 4), (16, 16, 4, 1))
buf4 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf10 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [dist_1_to_2, softmax_1], Original ATen: [aten._softmax]
triton_poi_fused__softmax_2.run(buf3, buf4, buf10, 64, grid=grid(64), stream=stream0)
buf5 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [dist_1_to_2], Original ATen: [aten._softmax]
triton_poi_fused__softmax_3.run(buf4, buf5, 64, grid=grid(64), stream=stream0)
buf6 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [dist_1_to_2, matmul], Original ATen: [aten._softmax, aten.bmm]
extern_kernels.bmm(reinterpret_tensor(primals_1, (4, 4, 4), (4, 1, 16), 0), buf5, out=buf6)
buf7 = empty_strided_cuda((4, 4, 8), (32, 8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat_1], Original ATen: [aten.cat]
triton_poi_fused_cat_4.run(primals_2, buf6, buf7, 128, grid=grid(128), stream=stream0)
buf8 = reinterpret_tensor(buf6, (16, 4), (4, 1), 0); del buf6 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf7, (16, 8), (8, 1), 0), reinterpret_tensor(primals_6, (8, 4), (1, 8), 0), out=buf8)
buf9 = reinterpret_tensor(buf8, (4, 4, 4), (16, 4, 1), 0); del buf8 # reuse
# Topologically Sorted Source Nodes: [seq_1_to_2], Original ATen: [aten.tanh]
triton_poi_fused_tanh_5.run(buf9, primals_7, 64, grid=grid(64), stream=stream0)
del primals_7
buf11 = buf5; del buf5 # reuse
# Topologically Sorted Source Nodes: [softmax_1], Original ATen: [aten._softmax]
triton_poi_fused__softmax_6.run(buf10, buf11, 64, grid=grid(64), stream=stream0)
buf12 = buf10; del buf10 # reuse
# Topologically Sorted Source Nodes: [matmul_1], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(primals_2, (4, 4, 4), (4, 1, 16), 0), reinterpret_tensor(buf11, (4, 4, 4), (16, 1, 4), 0), out=buf12)
buf13 = empty_strided_cuda((4, 4, 8), (32, 8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat_2], Original ATen: [aten.cat]
triton_poi_fused_cat_4.run(primals_1, buf12, buf13, 128, grid=grid(128), stream=stream0)
buf14 = reinterpret_tensor(buf12, (16, 4), (4, 1), 0); del buf12 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf13, (16, 8), (8, 1), 0), reinterpret_tensor(primals_8, (8, 4), (1, 8), 0), out=buf14)
buf15 = reinterpret_tensor(buf14, (4, 4, 4), (16, 4, 1), 0); del buf14 # reuse
# Topologically Sorted Source Nodes: [seq_2_to_1], Original ATen: [aten.tanh]
triton_poi_fused_tanh_5.run(buf15, primals_9, 64, grid=grid(64), stream=stream0)
del primals_9
return (buf15, buf9, primals_3, primals_5, buf0, buf2, buf3, reinterpret_tensor(buf7, (16, 8), (8, 1), 0), buf9, buf11, reinterpret_tensor(buf13, (16, 8), (8, 1), 0), buf15, primals_8, reinterpret_tensor(primals_2, (4, 4, 4), (4, 16, 1), 0), primals_6, reinterpret_tensor(primals_1, (4, 4, 4), (4, 16, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 8, 1, 1), (8, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((1, 4, 1, 1), (4, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 8), (8, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, 8), (8, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
import torch.nn.functional as F
import torch.optim
def get_activation_fn(name):
"""Returns a callable activation function from torch."""
if name in (None, 'linear'):
return lambda x: x
elif name in ('sigmoid', 'tanh'):
return getattr(torch, name)
else:
return getattr(F, name)
class CoAttention(nn.Module):
"""Co-attention between two sequences.
Uses one hidden layer to compute an affinity matrix between two sequences.
This can be then normalized in two direction which gives us 1->2 and 2->1
attentions.
The co-attention is computed using a single feed-forward layer as in
Bahdanau's attention.
"""
def __init__(self, ctx_1_dim, ctx_2_dim, bottleneck, att_activ='tanh',
mlp_bias=False):
super().__init__()
self.mlp_hid = nn.Conv2d(ctx_1_dim + ctx_2_dim, bottleneck, 1)
self.mlp_out = nn.Conv2d(bottleneck, 1, 1, bias=mlp_bias)
self.activ = get_activation_fn(att_activ)
self.project_1_to_2 = nn.Linear(ctx_1_dim + ctx_2_dim, bottleneck)
self.project_2_to_1 = nn.Linear(ctx_1_dim + ctx_2_dim, bottleneck)
def forward(self, ctx_1, ctx_2, ctx_1_mask=None, ctx_2_mask=None):
if ctx_2_mask is not None:
ctx_2_neg_mask = (1.0 - ctx_2_mask.transpose(0, 1).unsqueeze(1)
) * -1000000000000.0
ctx_1_len = ctx_1.size(0)
ctx_2_len = ctx_2.size(0)
b_ctx_1 = ctx_1.permute(1, 2, 0).unsqueeze(3).repeat(1, 1, 1, ctx_2_len
)
b_ctx_2 = ctx_2.permute(1, 2, 0).unsqueeze(2).repeat(1, 1, ctx_1_len, 1
)
catted = torch.cat([b_ctx_1, b_ctx_2], dim=1)
hidden = self.activ(self.mlp_hid(catted))
affinity_matrix = self.mlp_out(hidden).squeeze(1)
if ctx_1_mask is not None:
ctx_1_neg_mask = (1.0 - ctx_1_mask.transpose(0, 1).unsqueeze(2)
) * -1000000000000.0
affinity_matrix += ctx_1_neg_mask
if ctx_2_mask is not None:
ctx_2_neg_mask = (1.0 - ctx_2_mask.transpose(0, 1).unsqueeze(1)
) * -1000000000000.0
affinity_matrix += ctx_2_neg_mask
dist_1_to_2 = F.softmax(affinity_matrix, dim=2)
context_1_to_2 = ctx_1.permute(1, 2, 0).matmul(dist_1_to_2).permute(
2, 0, 1)
seq_1_to_2 = self.activ(self.project_1_to_2(torch.cat([ctx_2,
context_1_to_2], dim=-1)))
dist_2_to_1 = F.softmax(affinity_matrix, dim=1).transpose(1, 2)
context_2_to_1 = ctx_2.permute(1, 2, 0).matmul(dist_2_to_1).permute(
2, 0, 1)
seq_2_to_1 = self.activ(self.project_2_to_1(torch.cat([ctx_1,
context_2_to_1], dim=-1)))
return seq_2_to_1, seq_1_to_2
def get_inputs():
return [torch.rand([4, 4, 4]), torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'ctx_1_dim': 4, 'ctx_2_dim': 4, 'bottleneck': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch import nn
import torch.nn.functional as F
import torch.optim
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 512
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex // 16 % 8
x1 = xindex // 4 % 4
x3 = xindex // 128
x0 = xindex % 4
x4 = xindex
tmp0 = x2
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x3 + 16 * x1 + x2), tmp4 & xmask,
eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (4 * x3 + 16 * x0 + (-4 + x2)), tmp6 & xmask,
eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x4, tmp10, xmask)
@triton.jit
def triton_poi_fused_convolution_tanh_1(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 16 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x3, tmp3, xmask)
@triton.jit
def triton_poi_fused__softmax_2(in_ptr0, out_ptr0, out_ptr1, xnumel, XBLOCK:
tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x1 = xindex // 4
x0 = xindex % 4
x3 = xindex // 16
tmp0 = tl.load(in_ptr0 + x4, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (x0 + 16 * x3), xmask, eviction_policy=
'evict_last')
tmp11 = tl.load(in_ptr0 + (4 + x0 + 16 * x3), xmask, eviction_policy=
'evict_last')
tmp13 = tl.load(in_ptr0 + (8 + x0 + 16 * x3), xmask, eviction_policy=
'evict_last')
tmp15 = tl.load(in_ptr0 + (12 + x0 + 16 * x3), xmask, eviction_policy=
'evict_last')
tmp3 = triton_helpers.maximum(tmp1, tmp2)
tmp5 = triton_helpers.maximum(tmp3, tmp4)
tmp7 = triton_helpers.maximum(tmp5, tmp6)
tmp8 = tmp0 - tmp7
tmp9 = tl_math.exp(tmp8)
tmp12 = triton_helpers.maximum(tmp10, tmp11)
tmp14 = triton_helpers.maximum(tmp12, tmp13)
tmp16 = triton_helpers.maximum(tmp14, tmp15)
tmp17 = tmp0 - tmp16
tmp18 = tl_math.exp(tmp17)
tl.store(out_ptr0 + x4, tmp9, xmask)
tl.store(out_ptr1 + x4, tmp18, xmask)
@triton.jit
def triton_poi_fused__softmax_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x3 = xindex // 8
x1 = xindex // 8 % 4
x2 = xindex // 32
x4 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x3 + x0), tmp4 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (x2 + 4 * (-4 + x0) + 16 * x1), tmp6 & xmask,
eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x4, tmp10, xmask)
@triton.jit
def triton_poi_fused_tanh_5(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x2, tmp3, xmask)
@triton.jit
def triton_poi_fused__softmax_6(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x0 = xindex % 4
x2 = xindex // 16
tmp0 = tl.load(in_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + (x0 + 16 * x2), xmask, eviction_policy=
'evict_last')
tmp2 = tl.load(in_ptr0 + (4 + x0 + 16 * x2), xmask, eviction_policy=
'evict_last')
tmp4 = tl.load(in_ptr0 + (8 + x0 + 16 * x2), xmask, eviction_policy=
'evict_last')
tmp6 = tl.load(in_ptr0 + (12 + x0 + 16 * x2), xmask, eviction_policy=
'evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x3, tmp8, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9) = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_3, (4, 8, 1, 1), (8, 1, 1, 1))
assert_size_stride(primals_4, (4,), (1,))
assert_size_stride(primals_5, (1, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_6, (4, 8), (8, 1))
assert_size_stride(primals_7, (4,), (1,))
assert_size_stride(primals_8, (4, 8), (8, 1))
assert_size_stride(primals_9, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 8, 4, 4), (128, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(512)](primals_1, primals_2, buf0, 512,
XBLOCK=256, num_warps=4, num_stages=1)
buf1 = extern_kernels.convolution(buf0, primals_3, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 4, 4, 4), (64, 16, 4, 1))
buf2 = buf1
del buf1
triton_poi_fused_convolution_tanh_1[grid(256)](buf2, primals_4, 256,
XBLOCK=128, num_warps=4, num_stages=1)
del primals_4
buf3 = extern_kernels.convolution(buf2, primals_5, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 1, 4, 4), (16, 16, 4, 1))
buf4 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf10 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_2[grid(64)](buf3, buf4, buf10, 64, XBLOCK
=64, num_warps=1, num_stages=1)
buf5 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused__softmax_3[grid(64)](buf4, buf5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf6 = buf4
del buf4
extern_kernels.bmm(reinterpret_tensor(primals_1, (4, 4, 4), (4, 1,
16), 0), buf5, out=buf6)
buf7 = empty_strided_cuda((4, 4, 8), (32, 8, 1), torch.float32)
triton_poi_fused_cat_4[grid(128)](primals_2, buf6, buf7, 128,
XBLOCK=128, num_warps=4, num_stages=1)
buf8 = reinterpret_tensor(buf6, (16, 4), (4, 1), 0)
del buf6
extern_kernels.mm(reinterpret_tensor(buf7, (16, 8), (8, 1), 0),
reinterpret_tensor(primals_6, (8, 4), (1, 8), 0), out=buf8)
buf9 = reinterpret_tensor(buf8, (4, 4, 4), (16, 4, 1), 0)
del buf8
triton_poi_fused_tanh_5[grid(64)](buf9, primals_7, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del primals_7
buf11 = buf5
del buf5
triton_poi_fused__softmax_6[grid(64)](buf10, buf11, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf12 = buf10
del buf10
extern_kernels.bmm(reinterpret_tensor(primals_2, (4, 4, 4), (4, 1,
16), 0), reinterpret_tensor(buf11, (4, 4, 4), (16, 1, 4), 0),
out=buf12)
buf13 = empty_strided_cuda((4, 4, 8), (32, 8, 1), torch.float32)
triton_poi_fused_cat_4[grid(128)](primals_1, buf12, buf13, 128,
XBLOCK=128, num_warps=4, num_stages=1)
buf14 = reinterpret_tensor(buf12, (16, 4), (4, 1), 0)
del buf12
extern_kernels.mm(reinterpret_tensor(buf13, (16, 8), (8, 1), 0),
reinterpret_tensor(primals_8, (8, 4), (1, 8), 0), out=buf14)
buf15 = reinterpret_tensor(buf14, (4, 4, 4), (16, 4, 1), 0)
del buf14
triton_poi_fused_tanh_5[grid(64)](buf15, primals_9, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del primals_9
return (buf15, buf9, primals_3, primals_5, buf0, buf2, buf3,
reinterpret_tensor(buf7, (16, 8), (8, 1), 0), buf9, buf11,
reinterpret_tensor(buf13, (16, 8), (8, 1), 0), buf15, primals_8,
reinterpret_tensor(primals_2, (4, 4, 4), (4, 16, 1), 0), primals_6,
reinterpret_tensor(primals_1, (4, 4, 4), (4, 16, 1), 0))
def get_activation_fn(name):
"""Returns a callable activation function from torch."""
if name in (None, 'linear'):
return lambda x: x
elif name in ('sigmoid', 'tanh'):
return getattr(torch, name)
else:
return getattr(F, name)
class CoAttentionNew(nn.Module):
"""Co-attention between two sequences.
Uses one hidden layer to compute an affinity matrix between two sequences.
This can be then normalized in two direction which gives us 1->2 and 2->1
attentions.
The co-attention is computed using a single feed-forward layer as in
Bahdanau's attention.
"""
def __init__(self, ctx_1_dim, ctx_2_dim, bottleneck, att_activ='tanh',
mlp_bias=False):
super().__init__()
self.mlp_hid = nn.Conv2d(ctx_1_dim + ctx_2_dim, bottleneck, 1)
self.mlp_out = nn.Conv2d(bottleneck, 1, 1, bias=mlp_bias)
self.activ = get_activation_fn(att_activ)
self.project_1_to_2 = nn.Linear(ctx_1_dim + ctx_2_dim, bottleneck)
self.project_2_to_1 = nn.Linear(ctx_1_dim + ctx_2_dim, bottleneck)
def forward(self, input_0, input_1):
primals_3 = self.mlp_hid.weight
primals_4 = self.mlp_hid.bias
primals_5 = self.mlp_out.weight
primals_6 = self.project_1_to_2.weight
primals_7 = self.project_1_to_2.bias
primals_8 = self.project_2_to_1.weight
primals_9 = self.project_2_to_1.bias
primals_1 = input_0
primals_2 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9])
return output[0], output[1]
|
fmetze/nmtpytorch
|
CoAttention
| false | 12,383 |
[
"MIT"
] | 0 |
658a39a2c50e4e9e2fde69b520ddac7efc083257
|
https://github.com/fmetze/nmtpytorch/tree/658a39a2c50e4e9e2fde69b520ddac7efc083257
|
ExpMSE
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/2a/c2a5t6lnprxskvluqln3si6lrnz3uzfisoewtwdmxxlhvs5h6psw.py
# Topologically Sorted Source Nodes: [sub, loss, mul, exp, exp_loss, mean], Original ATen: [aten.sub, aten.pow, aten.mul, aten.exp, aten.mean]
# Source node to ATen node mapping:
# exp => exp
# exp_loss => mul_1
# loss => pow_1
# mean => mean
# mul => mul
# sub => sub
# Graph fragment:
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg1_1, %arg0_1), kwargs = {})
# %pow_1 : [num_users=2] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub, 2), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%pow_1, 4), kwargs = {})
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%mul,), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%pow_1, %exp), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%mul_1,), kwargs = {})
triton_per_fused_exp_mean_mul_pow_sub_0 = async_compile.triton('triton_per_fused_exp_mean_mul_pow_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_exp_mean_mul_pow_sub_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_exp_mean_mul_pow_sub_0(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp1 = tl.load(in_ptr1 + (r0), None)
tmp2 = tmp0 - tmp1
tmp3 = tmp2 * tmp2
tmp4 = 4.0
tmp5 = tmp3 * tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tmp3 * tmp6
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = 256.0
tmp12 = tmp10 / tmp11
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp12, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [sub, loss, mul, exp, exp_loss, mean], Original ATen: [aten.sub, aten.pow, aten.mul, aten.exp, aten.mean]
stream0 = get_raw_stream(0)
triton_per_fused_exp_mean_mul_pow_sub_0.run(buf1, arg1_1, arg0_1, 1, 256, grid=grid(1), stream=stream0)
del arg0_1
del arg1_1
return (buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class ExpMSE(nn.Module):
def __init__(self, lam):
super().__init__()
self.lam = lam
def forward(self, output, target):
_assert_no_grad(target)
loss = (output - target).pow(2)
exp_loss = loss * torch.exp(self.lam * loss)
return exp_loss.mean()
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'lam': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_exp_mean_mul_pow_sub_0(in_out_ptr0, in_ptr0, in_ptr1,
xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp1 = tl.load(in_ptr1 + r0, None)
tmp2 = tmp0 - tmp1
tmp3 = tmp2 * tmp2
tmp4 = 4.0
tmp5 = tmp3 * tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tmp3 * tmp6
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = 256.0
tmp12 = tmp10 / tmp11
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp12, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0
del buf0
get_raw_stream(0)
triton_per_fused_exp_mean_mul_pow_sub_0[grid(1)](buf1, arg1_1,
arg0_1, 1, 256, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
return buf1,
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class ExpMSENew(nn.Module):
def __init__(self, lam):
super().__init__()
self.lam = lam
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
dattientran/attorch
|
ExpMSE
| false | 12,384 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
LogLog
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/7r/c7re6ctix6g7lsmr3nupxvpdjizwr2rsj6dhqwqidpxhmmh4er6s.py
# Topologically Sorted Source Nodes: [exp, neg, exp_1, sub], Original ATen: [aten.exp, aten.neg, aten.rsub]
# Source node to ATen node mapping:
# exp => exp
# exp_1 => exp_1
# neg => neg
# sub => sub
# Graph fragment:
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%arg0_1,), kwargs = {})
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%exp,), kwargs = {})
# %exp_1 : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%neg,), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (1.0, %exp_1), kwargs = {})
triton_poi_fused_exp_neg_rsub_0 = async_compile.triton('triton_poi_fused_exp_neg_rsub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_exp_neg_rsub_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_exp_neg_rsub_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl_math.exp(tmp0)
tmp2 = -tmp1
tmp3 = tl_math.exp(tmp2)
tmp4 = 1.0
tmp5 = tmp4 - tmp3
tl.store(out_ptr0 + (x0), tmp5, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [exp, neg, exp_1, sub], Original ATen: [aten.exp, aten.neg, aten.rsub]
stream0 = get_raw_stream(0)
triton_poi_fused_exp_neg_rsub_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class LogLog(nn.Module):
def forward(self, x):
return 1.0 - torch.exp(-torch.exp(x))
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_exp_neg_rsub_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl_math.exp(tmp0)
tmp2 = -tmp1
tmp3 = tl_math.exp(tmp2)
tmp4 = 1.0
tmp5 = tmp4 - tmp3
tl.store(out_ptr0 + x0, tmp5, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_exp_neg_rsub_0[grid(256)](arg0_1, buf0, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class LogLogNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
LogLog
| false | 12,385 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
AdjustedElu
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ne/cneryon7gkc2praqyswdbwgqcbt2xwdplgelkjlisb5r47n7ia36.py
# Topologically Sorted Source Nodes: [sub, elu, add], Original ATen: [aten.sub, aten.elu, aten.add]
# Source node to ATen node mapping:
# add => add
# elu => expm1, gt, mul, mul_1, mul_2, where
# sub => sub
# Graph fragment:
# %sub : [num_users=3] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg0_1, 1.0), kwargs = {})
# %gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sub, 0), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub, 1.0), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub, 1.0), kwargs = {})
# %expm1 : [num_users=1] = call_function[target=torch.ops.aten.expm1.default](args = (%mul_1,), kwargs = {})
# %mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%expm1, 1.0), kwargs = {})
# %where : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt, %mul, %mul_2), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%where, 1.0), kwargs = {})
triton_poi_fused_add_elu_sub_0 = async_compile.triton('triton_poi_fused_add_elu_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_elu_sub_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_elu_sub_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 1.0
tmp2 = tmp0 - tmp1
tmp3 = 0.0
tmp4 = tmp2 > tmp3
tmp5 = tmp2 * tmp1
tmp6 = libdevice.expm1(tmp5)
tmp7 = tmp6 * tmp1
tmp8 = tl.where(tmp4, tmp5, tmp7)
tmp9 = tmp8 + tmp1
tl.store(out_ptr0 + (x0), tmp9, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [sub, elu, add], Original ATen: [aten.sub, aten.elu, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_elu_sub_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn import functional as F
class AdjustedElu(nn.Module):
"""
Elu activation function that's adjusted to:
1) ensure that all outputs are positive and
2) f(x) = x for x >= 1
"""
def forward(self, x):
return F.elu(x - 1.0) + 1.0
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_elu_sub_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 1.0
tmp2 = tmp0 - tmp1
tmp3 = 0.0
tmp4 = tmp2 > tmp3
tmp5 = tmp2 * tmp1
tmp6 = libdevice.expm1(tmp5)
tmp7 = tmp6 * tmp1
tmp8 = tl.where(tmp4, tmp5, tmp7)
tmp9 = tmp8 + tmp1
tl.store(out_ptr0 + x0, tmp9, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_elu_sub_0[grid(256)](arg0_1, buf0, 256, XBLOCK
=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class AdjustedEluNew(nn.Module):
"""
Elu activation function that's adjusted to:
1) ensure that all outputs are positive and
2) f(x) = x for x >= 1
"""
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
dattientran/attorch
|
AdjustedElu
| false | 12,386 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
Log1Exp
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/mh/cmhyq4x4i6pwnc5hicb7gcbca4dhq7s2fsdwerlyyztp64z5hxgf.py
# Topologically Sorted Source Nodes: [exp, add, log], Original ATen: [aten.exp, aten.add, aten.log]
# Source node to ATen node mapping:
# add => add
# exp => exp
# log => log
# Graph fragment:
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%arg0_1,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%exp, 1.0), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%add,), kwargs = {})
triton_poi_fused_add_exp_log_0 = async_compile.triton('triton_poi_fused_add_exp_log_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_exp_log_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_exp_log_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl_math.exp(tmp0)
tmp2 = 1.0
tmp3 = tmp1 + tmp2
tmp4 = tl_math.log(tmp3)
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [exp, add, log], Original ATen: [aten.exp, aten.add, aten.log]
stream0 = get_raw_stream(0)
triton_poi_fused_add_exp_log_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
def log1exp(x):
return torch.log(1.0 + torch.exp(x))
class Log1Exp(nn.Module):
def forward(self, x):
return log1exp(x)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import math as tl_math
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_exp_log_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl_math.exp(tmp0)
tmp2 = 1.0
tmp3 = tmp1 + tmp2
tmp4 = tl_math.log(tmp3)
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_exp_log_0[grid(256)](arg0_1, buf0, 256, XBLOCK
=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
def log1exp(x):
return torch.log(1.0 + torch.exp(x))
class Log1ExpNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
dattientran/attorch
|
Log1Exp
| false | 12,387 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
ExponentialMSE
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/zo/czoxn5bht43wgqy5cqttg472ua5ibdptaijkekzl4n72hncwijhy.py
# Topologically Sorted Source Nodes: [sub, loss, mul, exp, exp_loss, mean], Original ATen: [aten.sub, aten.pow, aten.mul, aten.exp, aten.mean]
# Source node to ATen node mapping:
# exp => exp
# exp_loss => mul_1
# loss => pow_1
# mean => mean
# mul => mul
# sub => sub
# Graph fragment:
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg1_1, %arg0_1), kwargs = {})
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub, 2), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, 4), kwargs = {})
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%mul,), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%pow_1, %exp), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%mul_1,), kwargs = {})
triton_per_fused_exp_mean_mul_pow_sub_0 = async_compile.triton('triton_per_fused_exp_mean_mul_pow_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_exp_mean_mul_pow_sub_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_exp_mean_mul_pow_sub_0(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp1 = tl.load(in_ptr1 + (r0), None)
tmp2 = tmp0 - tmp1
tmp3 = tmp2 * tmp2
tmp4 = 4.0
tmp5 = tmp1 * tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tmp3 * tmp6
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = 256.0
tmp12 = tmp10 / tmp11
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp12, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [sub, loss, mul, exp, exp_loss, mean], Original ATen: [aten.sub, aten.pow, aten.mul, aten.exp, aten.mean]
stream0 = get_raw_stream(0)
triton_per_fused_exp_mean_mul_pow_sub_0.run(buf1, arg1_1, arg0_1, 1, 256, grid=grid(1), stream=stream0)
del arg0_1
del arg1_1
return (buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class ExponentialMSE(nn.Module):
def __init__(self, lam):
super().__init__()
self.lam = lam
def forward(self, output, target):
_assert_no_grad(target)
loss = (output - target).pow(2)
exp_loss = loss * torch.exp(self.lam * target)
return exp_loss.mean()
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'lam': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_exp_mean_mul_pow_sub_0(in_out_ptr0, in_ptr0, in_ptr1,
xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp1 = tl.load(in_ptr1 + r0, None)
tmp2 = tmp0 - tmp1
tmp3 = tmp2 * tmp2
tmp4 = 4.0
tmp5 = tmp1 * tmp4
tmp6 = tl_math.exp(tmp5)
tmp7 = tmp3 * tmp6
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = 256.0
tmp12 = tmp10 / tmp11
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp12, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0
del buf0
get_raw_stream(0)
triton_per_fused_exp_mean_mul_pow_sub_0[grid(1)](buf1, arg1_1,
arg0_1, 1, 256, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
return buf1,
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class ExponentialMSENew(nn.Module):
def __init__(self, lam):
super().__init__()
self.lam = lam
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
dattientran/attorch
|
ExponentialMSE
| false | 12,388 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
LogCosh
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/zx/czxmxzopctcgackydle44efrurdnxzu5wrykeqxx3yz7yy6vfch5.py
# Topologically Sorted Source Nodes: [sub, add, cosh, log, mean], Original ATen: [aten.sub, aten.add, aten.cosh, aten.log, aten.mean]
# Source node to ATen node mapping:
# add => add
# cosh => cosh
# log => log
# mean => mean
# sub => sub
# Graph fragment:
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg0_1, %arg1_1), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub, 1e-12), kwargs = {})
# %cosh : [num_users=1] = call_function[target=torch.ops.aten.cosh.default](args = (%add,), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%cosh,), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%log,), kwargs = {})
triton_per_fused_add_cosh_log_mean_sub_0 = async_compile.triton('triton_per_fused_add_cosh_log_mean_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_cosh_log_mean_sub_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_cosh_log_mean_sub_0(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp1 = tl.load(in_ptr1 + (r0), None)
tmp2 = tmp0 - tmp1
tmp3 = 1e-12
tmp4 = tmp2 + tmp3
tmp5 = libdevice.cosh(tmp4)
tmp6 = tl_math.log(tmp5)
tmp7 = tl.broadcast_to(tmp6, [RBLOCK])
tmp9 = triton_helpers.promote_to_tensor(tl.sum(tmp7, 0))
tmp10 = 256.0
tmp11 = tmp9 / tmp10
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp11, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [sub, add, cosh, log, mean], Original ATen: [aten.sub, aten.add, aten.cosh, aten.log, aten.mean]
stream0 = get_raw_stream(0)
triton_per_fused_add_cosh_log_mean_sub_0.run(buf1, arg0_1, arg1_1, 1, 256, grid=grid(1), stream=stream0)
del arg0_1
del arg1_1
return (buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class LogCosh(nn.Module):
def __init__(self, bias=1e-12):
super().__init__()
self.bias = bias
def forward(self, output, target):
_assert_no_grad(target)
return torch.mean(torch.log(torch.cosh(target - output + self.bias)))
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_add_cosh_log_mean_sub_0(in_out_ptr0, in_ptr0, in_ptr1,
xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp1 = tl.load(in_ptr1 + r0, None)
tmp2 = tmp0 - tmp1
tmp3 = 1e-12
tmp4 = tmp2 + tmp3
tmp5 = libdevice.cosh(tmp4)
tmp6 = tl_math.log(tmp5)
tmp7 = tl.broadcast_to(tmp6, [RBLOCK])
tmp9 = triton_helpers.promote_to_tensor(tl.sum(tmp7, 0))
tmp10 = 256.0
tmp11 = tmp9 / tmp10
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp11, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0
del buf0
get_raw_stream(0)
triton_per_fused_add_cosh_log_mean_sub_0[grid(1)](buf1, arg0_1,
arg1_1, 1, 256, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
return buf1,
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class LogCoshNew(nn.Module):
def __init__(self, bias=1e-12):
super().__init__()
self.bias = bias
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
dattientran/attorch
|
LogCosh
| false | 12,389 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
GroupSort
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/6h/c6hxblyogt5wxvcpiruhezavaic5v7sb5wrsvnhh4vkvnrwnb4nj.py
# Topologically Sorted Source Nodes: [sort], Original ATen: [aten.sort]
# Source node to ATen node mapping:
# sort => sort
# Graph fragment:
# %sort : [num_users=1] = call_function[target=torch.ops.aten.sort.default](args = (%view,), kwargs = {})
triton_per_fused_sort_0 = async_compile.triton('triton_per_fused_sort_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[64, 4],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_sort_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_sort_0(in_ptr0, out_ptr0, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 64
rnumel = 4
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + (4*x0)), xmask, other=0.0)
tmp1 = r1
tmp2 = tmp1.to(tl.int16)
tmp3 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp4 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5, tmp6, = triton_helpers.sort_with_index(tmp3, tmp4, None, 1, stable=False, descending=False)
tl.store(out_ptr0 + (r1 + (4*x0)), tmp5, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 1, 4), (64, 16, 4, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [sort], Original ATen: [aten.sort]
stream0 = get_raw_stream(0)
triton_per_fused_sort_0.run(arg0_1, buf0, 64, 4, grid=grid(64), stream=stream0)
del arg0_1
return (reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
def process_group_size(x, group_size, axis=-1):
size = list(x.size())
num_channels = size[axis]
if num_channels % group_size:
raise ValueError(
'number of features({}) is not a multiple of group_size({})'.
format(num_channels, num_units))
size[axis] = -1
if axis == -1:
size += [group_size]
else:
size.insert(axis + 1, group_size)
return size
def group_sort(x, group_size, axis=-1):
size = process_group_size(x, group_size, axis)
grouped_x = x.view(*size)
sort_dim = axis if axis == -1 else axis + 1
sorted_grouped_x, _ = grouped_x.sort(dim=sort_dim)
sorted_x = sorted_grouped_x.view(*list(x.shape))
return sorted_x
class GroupSort(nn.Module):
def __init__(self, group_size, axis=-1):
super(GroupSort, self).__init__()
self.group_size = group_size
self.axis = axis
def forward(self, x):
group_sorted = group_sort(x, self.group_size, self.axis)
return group_sorted
def extra_repr(self):
return 'num_groups: {}'.format(self.num_units)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'group_size': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_per_fused_sort_0(in_ptr0, out_ptr0, xnumel, rnumel, XBLOCK: tl.
constexpr):
xnumel = 64
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r1 = rindex
x0 = xindex
tmp0 = tl.load(in_ptr0 + (r1 + 4 * x0), xmask, other=0.0)
tmp1 = r1
tmp2 = tmp1.to(tl.int16)
tmp3 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp4 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5, _tmp6 = triton_helpers.sort_with_index(tmp3, tmp4, None, 1,
stable=False, descending=False)
tl.store(out_ptr0 + (r1 + 4 * x0), tmp5, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 1, 4), (64, 16, 4, 4, 1), torch
.float32)
get_raw_stream(0)
triton_per_fused_sort_0[grid(64)](arg0_1, buf0, 64, 4, XBLOCK=1,
num_warps=2, num_stages=1)
del arg0_1
return reinterpret_tensor(buf0, (4, 4, 4, 4), (64, 16, 4, 1), 0),
def process_group_size(x, group_size, axis=-1):
size = list(x.size())
num_channels = size[axis]
if num_channels % group_size:
raise ValueError(
'number of features({}) is not a multiple of group_size({})'.
format(num_channels, num_units))
size[axis] = -1
if axis == -1:
size += [group_size]
else:
size.insert(axis + 1, group_size)
return size
def group_sort(x, group_size, axis=-1):
size = process_group_size(x, group_size, axis)
grouped_x = x.view(*size)
sort_dim = axis if axis == -1 else axis + 1
sorted_grouped_x, _ = grouped_x.sort(dim=sort_dim)
sorted_x = sorted_grouped_x.view(*list(x.shape))
return sorted_x
class GroupSortNew(nn.Module):
def __init__(self, group_size, axis=-1):
super(GroupSortNew, self).__init__()
self.group_size = group_size
self.axis = axis
def extra_repr(self):
return 'num_groups: {}'.format(self.num_units)
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
dattientran/attorch
|
GroupSort
| false | 12,390 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
AvgCorr
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/pm/cpmnwgjehgeioaapsw7n3blodric4lxkefozxm5qgof54pgi43ed.py
# Topologically Sorted Source Nodes: [delta_out], Original ATen: [aten.sub]
# Source node to ATen node mapping:
# delta_out => sub
# Graph fragment:
# %sub : [num_users=2] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg1_1, %expand), kwargs = {})
triton_poi_fused_sub_0 = async_compile.triton('triton_poi_fused_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_sub_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_sub_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tl.store(out_ptr0 + (x2), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/up/cupbvlxmuoljtjvmt7ybhiqucxwkyemdkfbwx6ur6nmhovkfyxcj.py
# Topologically Sorted Source Nodes: [mul, mean_4, pow_1, var_out, add, pow_2, var_target, add_1, mul_1, sqrt, corrs, mean_5], Original ATen: [aten.mul, aten.mean, aten.pow, aten.add, aten.sqrt, aten.div]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# corrs => div
# mean_4 => mean_4
# mean_5 => mean_5
# mul => mul
# mul_1 => mul_1
# pow_1 => pow_1
# pow_2 => pow_2
# sqrt => sqrt
# var_out => mean_2
# var_target => mean_3
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub, %sub_1), kwargs = {})
# %mean_4 : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%mul, [0], True), kwargs = {})
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub, 2), kwargs = {})
# %mean_2 : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%pow_1, [0], True), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mean_2, 1e-12), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub_1, 2), kwargs = {})
# %mean_3 : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%pow_2, [0], True), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mean_3, 1e-12), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add, %add_1), kwargs = {})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%mul_1,), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%mean_4, %sqrt), kwargs = {})
# %mean_5 : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%div,), kwargs = {})
triton_per_fused_add_div_mean_mul_pow_sqrt_1 = async_compile.triton('triton_per_fused_add_div_mean_mul_pow_sqrt_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 64],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_div_mean_mul_pow_sqrt_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_div_mean_mul_pow_sqrt_1(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 1
rnumel = 64
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp1 = tl.load(in_ptr1 + (r0), None)
tmp3 = tl.load(in_ptr0 + (64 + r0), None)
tmp4 = tl.load(in_ptr1 + (64 + r0), None)
tmp7 = tl.load(in_ptr0 + (128 + r0), None)
tmp8 = tl.load(in_ptr1 + (128 + r0), None)
tmp11 = tl.load(in_ptr0 + (192 + r0), None)
tmp12 = tl.load(in_ptr1 + (192 + r0), None)
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp0 * tmp0
tmp18 = tmp3 * tmp3
tmp19 = tmp17 + tmp18
tmp20 = tmp7 * tmp7
tmp21 = tmp19 + tmp20
tmp22 = tmp11 * tmp11
tmp23 = tmp21 + tmp22
tmp24 = tmp23 / tmp15
tmp25 = 1e-12
tmp26 = tmp24 + tmp25
tmp27 = tmp1 * tmp1
tmp28 = tmp4 * tmp4
tmp29 = tmp27 + tmp28
tmp30 = tmp8 * tmp8
tmp31 = tmp29 + tmp30
tmp32 = tmp12 * tmp12
tmp33 = tmp31 + tmp32
tmp34 = tmp33 / tmp15
tmp35 = tmp34 + tmp25
tmp36 = tmp26 * tmp35
tmp37 = libdevice.sqrt(tmp36)
tmp38 = tmp16 / tmp37
tmp39 = tl.broadcast_to(tmp38, [XBLOCK, RBLOCK])
tmp41 = tl.sum(tmp39, 1)[:, None]
tmp42 = 64.0
tmp43 = tmp41 / tmp42
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([XBLOCK, 1], 0, tl.int32)), tmp43, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [delta_out], Original ATen: [aten.sub]
stream0 = get_raw_stream(0)
triton_poi_fused_sub_0.run(arg1_1, buf0, 256, grid=grid(256), stream=stream0)
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [delta_target], Original ATen: [aten.sub]
triton_poi_fused_sub_0.run(arg0_1, buf1, 256, grid=grid(256), stream=stream0)
del arg0_1
buf3 = empty_strided_cuda((), (), torch.float32)
buf4 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [mul, mean_4, pow_1, var_out, add, pow_2, var_target, add_1, mul_1, sqrt, corrs, mean_5], Original ATen: [aten.mul, aten.mean, aten.pow, aten.add, aten.sqrt, aten.div]
triton_per_fused_add_div_mean_mul_pow_sqrt_1.run(buf4, buf0, buf1, 1, 64, grid=grid(1), stream=stream0)
del buf0
del buf1
return (buf4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class AvgCorr(nn.Module):
def __init__(self, eps=1e-12):
self.eps = eps
super().__init__()
def forward(self, output, target):
_assert_no_grad(target)
delta_out = output - output.mean(0, keepdim=True).expand_as(output)
delta_target = target - target.mean(0, keepdim=True).expand_as(target)
var_out = delta_out.pow(2).mean(0, keepdim=True)
var_target = delta_target.pow(2).mean(0, keepdim=True)
corrs = (delta_out * delta_target).mean(0, keepdim=True) / ((
var_out + self.eps) * (var_target + self.eps)).sqrt()
return corrs.mean()
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_sub_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tl.store(out_ptr0 + x2, tmp10, xmask)
@triton.jit
def triton_per_fused_add_div_mean_mul_pow_sqrt_1(in_out_ptr0, in_ptr0,
in_ptr1, xnumel, rnumel, XBLOCK: tl.constexpr):
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp1 = tl.load(in_ptr1 + r0, None)
tmp3 = tl.load(in_ptr0 + (64 + r0), None)
tmp4 = tl.load(in_ptr1 + (64 + r0), None)
tmp7 = tl.load(in_ptr0 + (128 + r0), None)
tmp8 = tl.load(in_ptr1 + (128 + r0), None)
tmp11 = tl.load(in_ptr0 + (192 + r0), None)
tmp12 = tl.load(in_ptr1 + (192 + r0), None)
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp0 * tmp0
tmp18 = tmp3 * tmp3
tmp19 = tmp17 + tmp18
tmp20 = tmp7 * tmp7
tmp21 = tmp19 + tmp20
tmp22 = tmp11 * tmp11
tmp23 = tmp21 + tmp22
tmp24 = tmp23 / tmp15
tmp25 = 1e-12
tmp26 = tmp24 + tmp25
tmp27 = tmp1 * tmp1
tmp28 = tmp4 * tmp4
tmp29 = tmp27 + tmp28
tmp30 = tmp8 * tmp8
tmp31 = tmp29 + tmp30
tmp32 = tmp12 * tmp12
tmp33 = tmp31 + tmp32
tmp34 = tmp33 / tmp15
tmp35 = tmp34 + tmp25
tmp36 = tmp26 * tmp35
tmp37 = libdevice.sqrt(tmp36)
tmp38 = tmp16 / tmp37
tmp39 = tl.broadcast_to(tmp38, [XBLOCK, RBLOCK])
tmp41 = tl.sum(tmp39, 1)[:, None]
tmp42 = 64.0
tmp43 = tmp41 / tmp42
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([XBLOCK, 1], 0, tl.int32), tmp43, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_sub_0[grid(256)](arg1_1, buf0, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_sub_0[grid(256)](arg0_1, buf1, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del arg0_1
buf3 = empty_strided_cuda((), (), torch.float32)
buf4 = buf3
del buf3
triton_per_fused_add_div_mean_mul_pow_sqrt_1[grid(1)](buf4, buf0,
buf1, 1, 64, XBLOCK=1, num_warps=2, num_stages=1)
del buf0
del buf1
return buf4,
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class AvgCorrNew(nn.Module):
def __init__(self, eps=1e-12):
self.eps = eps
super().__init__()
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
dattientran/attorch
|
AvgCorr
| false | 12,391 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
Elu1
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ir/ciredr2wczt7o3ccswmy77vz2tgudsn3os5vwcuutede7rcvrozg.py
# Topologically Sorted Source Nodes: [elu, add], Original ATen: [aten.elu, aten.add]
# Source node to ATen node mapping:
# add => add
# elu => expm1, gt, mul, mul_1, mul_2, where
# Graph fragment:
# %gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%arg0_1, 0), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, 1.0), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, 1.0), kwargs = {})
# %expm1 : [num_users=1] = call_function[target=torch.ops.aten.expm1.default](args = (%mul_1,), kwargs = {})
# %mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%expm1, 1.0), kwargs = {})
# %where : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt, %mul, %mul_2), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%where, 1.0), kwargs = {})
# %copy_ : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%arg0_1, %where), kwargs = {})
triton_poi_fused_add_elu_0 = async_compile.triton('triton_poi_fused_add_elu_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_elu_0', 'mutated_arg_names': ['in_ptr0', 'out_ptr2'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_elu_0(in_ptr0, out_ptr0, out_ptr2, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tl.store(out_ptr0 + (x0), tmp8, xmask)
tl.store(out_ptr2 + (x0), tmp7, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [elu, add], Original ATen: [aten.elu, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_elu_0.run(arg0_1, buf0, arg0_1, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn import functional as F
def elu1(x):
return F.elu(x, inplace=True) + 1.0
class Elu1(nn.Module):
"""
Elu activation function shifted by 1 to ensure that the
output stays positive. That is:
Elu1(x) = Elu(x) + 1
"""
def forward(self, x):
return elu1(x)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
from torch.nn import functional as F
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_elu_0(in_ptr0, out_ptr0, out_ptr2, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tl.store(out_ptr0 + x0, tmp8, xmask)
tl.store(out_ptr2 + x0, tmp7, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_elu_0[grid(256)](arg0_1, buf0, arg0_1, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
def elu1(x):
return F.elu(x, inplace=True) + 1.0
class Elu1New(nn.Module):
"""
Elu activation function shifted by 1 to ensure that the
output stays positive. That is:
Elu1(x) = Elu(x) + 1
"""
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
dattientran/attorch
|
Elu1
| false | 12,392 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
XSigmoid
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/of/cofynvv2vkexy5uq2apowq2ccvicws45zrrwqqzypwoqaac7hq2g.py
# Topologically Sorted Source Nodes: [error, mul, neg, exp, add, truediv, sub_1, mean], Original ATen: [aten.sub, aten.mul, aten.neg, aten.exp, aten.add, aten.div, aten.mean]
# Source node to ATen node mapping:
# add => add
# error => sub
# exp => exp
# mean => mean
# mul => mul
# neg => neg
# sub_1 => sub_1
# truediv => div
# Graph fragment:
# %sub : [num_users=3] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg0_1, %arg1_1), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub, 2), kwargs = {})
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%sub,), kwargs = {})
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%neg,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%exp, 1), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%mul, %add), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%div, %sub), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%sub_1,), kwargs = {})
triton_per_fused_add_div_exp_mean_mul_neg_sub_0 = async_compile.triton('triton_per_fused_add_div_exp_mean_mul_neg_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_div_exp_mean_mul_neg_sub_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_div_exp_mean_mul_neg_sub_0(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp1 = tl.load(in_ptr1 + (r0), None)
tmp2 = tmp0 - tmp1
tmp3 = 2.0
tmp4 = tmp2 * tmp3
tmp5 = -tmp2
tmp6 = tl_math.exp(tmp5)
tmp7 = 1.0
tmp8 = tmp6 + tmp7
tmp9 = tmp4 / tmp8
tmp10 = tmp9 - tmp2
tmp11 = tl.broadcast_to(tmp10, [RBLOCK])
tmp13 = triton_helpers.promote_to_tensor(tl.sum(tmp11, 0))
tmp14 = 256.0
tmp15 = tmp13 / tmp14
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp15, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [error, mul, neg, exp, add, truediv, sub_1, mean], Original ATen: [aten.sub, aten.mul, aten.neg, aten.exp, aten.add, aten.div, aten.mean]
stream0 = get_raw_stream(0)
triton_per_fused_add_div_exp_mean_mul_neg_sub_0.run(buf1, arg0_1, arg1_1, 1, 256, grid=grid(1), stream=stream0)
del arg0_1
del arg1_1
return (buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class XSigmoid(torch.nn.Module):
def __init__(self):
super().__init__()
def forward(self, output, target):
_assert_no_grad(target)
error = target - output
return torch.mean(2 * error / (1 + torch.exp(-error)) - error)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_add_div_exp_mean_mul_neg_sub_0(in_out_ptr0, in_ptr0,
in_ptr1, xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp1 = tl.load(in_ptr1 + r0, None)
tmp2 = tmp0 - tmp1
tmp3 = 2.0
tmp4 = tmp2 * tmp3
tmp5 = -tmp2
tmp6 = tl_math.exp(tmp5)
tmp7 = 1.0
tmp8 = tmp6 + tmp7
tmp9 = tmp4 / tmp8
tmp10 = tmp9 - tmp2
tmp11 = tl.broadcast_to(tmp10, [RBLOCK])
tmp13 = triton_helpers.promote_to_tensor(tl.sum(tmp11, 0))
tmp14 = 256.0
tmp15 = tmp13 / tmp14
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp15, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0
del buf0
get_raw_stream(0)
triton_per_fused_add_div_exp_mean_mul_neg_sub_0[grid(1)](buf1,
arg0_1, arg1_1, 1, 256, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
return buf1,
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class XSigmoidNew(torch.nn.Module):
def __init__(self):
super().__init__()
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
dattientran/attorch
|
XSigmoid
| false | 12,393 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
ycbcr_to_rgb_jpeg
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/vx/cvxxlynxah34yu572a3wn3fl35yzhysx2xti2izljcb2iii5cxw4.py
# Topologically Sorted Source Nodes: [add], Original ATen: [aten.add]
# Source node to ATen node mapping:
# add => add
# Graph fragment:
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%primals_2, %primals_1), kwargs = {})
triton_poi_fused_add_0 = async_compile.triton('triton_poi_fused_add_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 192
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 3
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (x2), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (3, ), (1, ))
assert_size_stride(primals_2, (4, 4, 4, 3), (48, 12, 3, 1))
assert_size_stride(primals_3, (3, 3), (1, 3))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 3), (48, 12, 3, 1), torch.float32)
# Topologically Sorted Source Nodes: [add], Original ATen: [aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_0.run(primals_2, primals_1, buf0, 192, grid=grid(192), stream=stream0)
del primals_1
del primals_2
buf1 = empty_strided_cuda((64, 3), (3, 1), torch.float32)
# Topologically Sorted Source Nodes: [result], Original ATen: [aten.mm]
extern_kernels.mm(reinterpret_tensor(buf0, (64, 3), (3, 1), 0), primals_3, out=buf1)
return (reinterpret_tensor(buf1, (4, 3, 4, 4), (48, 1, 12, 3), 0), reinterpret_tensor(buf0, (64, 3), (3, 1), 0), reinterpret_tensor(primals_3, (3, 3), (3, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((3, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 3), (48, 12, 3, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((3, 3), (1, 3), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import numpy as np
import torch.nn as nn
class ycbcr_to_rgb_jpeg(nn.Module):
""" Converts YCbCr image to RGB JPEG
Input:
image(tensor): batch x height x width x 3
Outpput:
result(tensor): batch x 3 x height x width
"""
def __init__(self):
super(ycbcr_to_rgb_jpeg, self).__init__()
matrix = np.array([[1.0, 0.0, 1.402], [1, -0.344136, -0.714136], [1,
1.772, 0]], dtype=np.float32).T
self.shift = nn.Parameter(torch.tensor([0, -128.0, -128.0]))
self.matrix = nn.Parameter(torch.from_numpy(matrix))
def forward(self, image):
result = torch.tensordot(image + self.shift, self.matrix, dims=1)
result.view(image.shape)
return result.permute(0, 3, 1, 2)
def get_inputs():
return [torch.rand([4, 4, 4, 3])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import numpy as np
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_add_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 192
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 3
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + x2, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (3,), (1,))
assert_size_stride(primals_2, (4, 4, 4, 3), (48, 12, 3, 1))
assert_size_stride(primals_3, (3, 3), (1, 3))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 3), (48, 12, 3, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_0[grid(192)](primals_2, primals_1, buf0, 192,
XBLOCK=128, num_warps=4, num_stages=1)
del primals_1
del primals_2
buf1 = empty_strided_cuda((64, 3), (3, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf0, (64, 3), (3, 1), 0),
primals_3, out=buf1)
return reinterpret_tensor(buf1, (4, 3, 4, 4), (48, 1, 12, 3), 0
), reinterpret_tensor(buf0, (64, 3), (3, 1), 0), reinterpret_tensor(
primals_3, (3, 3), (3, 1), 0)
class ycbcr_to_rgb_jpegNew(nn.Module):
""" Converts YCbCr image to RGB JPEG
Input:
image(tensor): batch x height x width x 3
Outpput:
result(tensor): batch x 3 x height x width
"""
def __init__(self):
super(ycbcr_to_rgb_jpegNew, self).__init__()
matrix = np.array([[1.0, 0.0, 1.402], [1, -0.344136, -0.714136], [1,
1.772, 0]], dtype=np.float32).T
self.shift = nn.Parameter(torch.tensor([0, -128.0, -128.0]))
self.matrix = nn.Parameter(torch.from_numpy(matrix))
def forward(self, input_0):
primals_1 = self.shift
primals_3 = self.matrix
primals_2 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
foxtrotmike/DiffJPEG
|
ycbcr_to_rgb_jpeg
| false | 12,394 |
[
"MIT"
] | 0 |
7dbc44b1e921f20a213a7206a8578d6a1c8131b4
|
https://github.com/foxtrotmike/DiffJPEG/tree/7dbc44b1e921f20a213a7206a8578d6a1c8131b4
|
Corr
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/pm/cpmnwgjehgeioaapsw7n3blodric4lxkefozxm5qgof54pgi43ed.py
# Topologically Sorted Source Nodes: [delta_out], Original ATen: [aten.sub]
# Source node to ATen node mapping:
# delta_out => sub
# Graph fragment:
# %sub : [num_users=2] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg1_1, %expand), kwargs = {})
triton_poi_fused_sub_0 = async_compile.triton('triton_poi_fused_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_sub_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_sub_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tl.store(out_ptr0 + (x2), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/i2/ci2d56ny7vwf2y5hgujwey5irj2hjtjz2tmn7nhrojoe4igdlrju.py
# Topologically Sorted Source Nodes: [mul, mean_4, pow_1, var_out, add, pow_2, var_target, add_1, mul_1, sqrt, corrs], Original ATen: [aten.mul, aten.mean, aten.pow, aten.add, aten.sqrt, aten.div]
# Source node to ATen node mapping:
# add => add
# add_1 => add_1
# corrs => div
# mean_4 => mean_4
# mul => mul
# mul_1 => mul_1
# pow_1 => pow_1
# pow_2 => pow_2
# sqrt => sqrt
# var_out => mean_2
# var_target => mean_3
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub, %sub_1), kwargs = {})
# %mean_4 : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%mul, [0], True), kwargs = {})
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub, 2), kwargs = {})
# %mean_2 : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%pow_1, [0], True), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mean_2, 1e-12), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub_1, 2), kwargs = {})
# %mean_3 : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%pow_2, [0], True), kwargs = {})
# %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mean_3, 1e-12), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add, %add_1), kwargs = {})
# %sqrt : [num_users=1] = call_function[target=torch.ops.aten.sqrt.default](args = (%mul_1,), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%mean_4, %sqrt), kwargs = {})
triton_poi_fused_add_div_mean_mul_pow_sqrt_1 = async_compile.triton('triton_poi_fused_add_div_mean_mul_pow_sqrt_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_mean_mul_pow_sqrt_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 8, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_mean_mul_pow_sqrt_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask)
tmp3 = tl.load(in_ptr0 + (64 + x0), xmask)
tmp4 = tl.load(in_ptr1 + (64 + x0), xmask)
tmp7 = tl.load(in_ptr0 + (128 + x0), xmask)
tmp8 = tl.load(in_ptr1 + (128 + x0), xmask)
tmp11 = tl.load(in_ptr0 + (192 + x0), xmask)
tmp12 = tl.load(in_ptr1 + (192 + x0), xmask)
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp0 * tmp0
tmp18 = tmp3 * tmp3
tmp19 = tmp17 + tmp18
tmp20 = tmp7 * tmp7
tmp21 = tmp19 + tmp20
tmp22 = tmp11 * tmp11
tmp23 = tmp21 + tmp22
tmp24 = tmp23 / tmp15
tmp25 = 1e-12
tmp26 = tmp24 + tmp25
tmp27 = tmp1 * tmp1
tmp28 = tmp4 * tmp4
tmp29 = tmp27 + tmp28
tmp30 = tmp8 * tmp8
tmp31 = tmp29 + tmp30
tmp32 = tmp12 * tmp12
tmp33 = tmp31 + tmp32
tmp34 = tmp33 / tmp15
tmp35 = tmp34 + tmp25
tmp36 = tmp26 * tmp35
tmp37 = libdevice.sqrt(tmp36)
tmp38 = tmp16 / tmp37
tl.store(out_ptr0 + (x0), tmp38, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [delta_out], Original ATen: [aten.sub]
stream0 = get_raw_stream(0)
triton_poi_fused_sub_0.run(arg1_1, buf0, 256, grid=grid(256), stream=stream0)
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [delta_target], Original ATen: [aten.sub]
triton_poi_fused_sub_0.run(arg0_1, buf1, 256, grid=grid(256), stream=stream0)
del arg0_1
buf2 = empty_strided_cuda((1, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, mean_4, pow_1, var_out, add, pow_2, var_target, add_1, mul_1, sqrt, corrs], Original ATen: [aten.mul, aten.mean, aten.pow, aten.add, aten.sqrt, aten.div]
triton_poi_fused_add_div_mean_mul_pow_sqrt_1.run(buf0, buf1, buf2, 64, grid=grid(64), stream=stream0)
del buf0
del buf1
return (buf2, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class Corr(nn.Module):
def __init__(self, eps=1e-12):
self.eps = eps
super().__init__()
def forward(self, output, target):
_assert_no_grad(target)
delta_out = output - output.mean(0, keepdim=True).expand_as(output)
delta_target = target - target.mean(0, keepdim=True).expand_as(target)
var_out = delta_out.pow(2).mean(0, keepdim=True)
var_target = delta_target.pow(2).mean(0, keepdim=True)
corrs = (delta_out * delta_target).mean(0, keepdim=True) / ((
var_out + self.eps) * (var_target + self.eps)).sqrt()
return corrs
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_sub_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tl.store(out_ptr0 + x2, tmp10, xmask)
@triton.jit
def triton_poi_fused_add_div_mean_mul_pow_sqrt_1(in_ptr0, in_ptr1, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask)
tmp3 = tl.load(in_ptr0 + (64 + x0), xmask)
tmp4 = tl.load(in_ptr1 + (64 + x0), xmask)
tmp7 = tl.load(in_ptr0 + (128 + x0), xmask)
tmp8 = tl.load(in_ptr1 + (128 + x0), xmask)
tmp11 = tl.load(in_ptr0 + (192 + x0), xmask)
tmp12 = tl.load(in_ptr1 + (192 + x0), xmask)
tmp2 = tmp0 * tmp1
tmp5 = tmp3 * tmp4
tmp6 = tmp2 + tmp5
tmp9 = tmp7 * tmp8
tmp10 = tmp6 + tmp9
tmp13 = tmp11 * tmp12
tmp14 = tmp10 + tmp13
tmp15 = 4.0
tmp16 = tmp14 / tmp15
tmp17 = tmp0 * tmp0
tmp18 = tmp3 * tmp3
tmp19 = tmp17 + tmp18
tmp20 = tmp7 * tmp7
tmp21 = tmp19 + tmp20
tmp22 = tmp11 * tmp11
tmp23 = tmp21 + tmp22
tmp24 = tmp23 / tmp15
tmp25 = 1e-12
tmp26 = tmp24 + tmp25
tmp27 = tmp1 * tmp1
tmp28 = tmp4 * tmp4
tmp29 = tmp27 + tmp28
tmp30 = tmp8 * tmp8
tmp31 = tmp29 + tmp30
tmp32 = tmp12 * tmp12
tmp33 = tmp31 + tmp32
tmp34 = tmp33 / tmp15
tmp35 = tmp34 + tmp25
tmp36 = tmp26 * tmp35
tmp37 = libdevice.sqrt(tmp36)
tmp38 = tmp16 / tmp37
tl.store(out_ptr0 + x0, tmp38, xmask)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_sub_0[grid(256)](arg1_1, buf0, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_sub_0[grid(256)](arg0_1, buf1, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del arg0_1
buf2 = empty_strided_cuda((1, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_add_div_mean_mul_pow_sqrt_1[grid(64)](buf0, buf1,
buf2, 64, XBLOCK=64, num_warps=1, num_stages=1)
del buf0
del buf1
return buf2,
def _assert_no_grad(tensor):
assert not tensor.requires_grad
class CorrNew(nn.Module):
def __init__(self, eps=1e-12):
self.eps = eps
super().__init__()
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
dattientran/attorch
|
Corr
| false | 12,395 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
chroma_subsampling
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/55/c55xeg7h533g7nm5g2b6wdnokzoxkacwj5gi4cwvuithrfob64eb.py
# Topologically Sorted Source Nodes: [cb], Original ATen: [aten.avg_pool2d]
# Source node to ATen node mapping:
# cb => avg_pool2d
# Graph fragment:
# %avg_pool2d : [num_users=1] = call_function[target=torch.ops.aten.avg_pool2d.default](args = (%unsqueeze, [2, 2], [2, 2], [0, 0], False, False), kwargs = {})
triton_poi_fused_avg_pool2d_0 = async_compile.triton('triton_poi_fused_avg_pool2d_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_avg_pool2d_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_avg_pool2d_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 2
x1 = (xindex // 2)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (1 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (5 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (17 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (21 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/vr/cvrunofn7cfbpeyn5wdzlhq4aw752z3x2xgyyzxa2cne36j5mwnp.py
# Topologically Sorted Source Nodes: [cr], Original ATen: [aten.avg_pool2d]
# Source node to ATen node mapping:
# cr => avg_pool2d_1
# Graph fragment:
# %avg_pool2d_1 : [num_users=1] = call_function[target=torch.ops.aten.avg_pool2d.default](args = (%unsqueeze_1, [2, 2], [2, 2], [0, 0], False, False), kwargs = {})
triton_poi_fused_avg_pool2d_1 = async_compile.triton('triton_poi_fused_avg_pool2d_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_avg_pool2d_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_avg_pool2d_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 2
x1 = (xindex // 2)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (2 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (6 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (18 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (22 + (8*x0) + (32*x1)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 2, 2), (4, 1, 2, 1), torch.float32)
# Topologically Sorted Source Nodes: [cb], Original ATen: [aten.avg_pool2d]
stream0 = get_raw_stream(0)
triton_poi_fused_avg_pool2d_0.run(arg0_1, buf0, 16, grid=grid(16), stream=stream0)
buf1 = empty_strided_cuda((4, 1, 2, 2), (4, 1, 2, 1), torch.float32)
# Topologically Sorted Source Nodes: [cr], Original ATen: [aten.avg_pool2d]
triton_poi_fused_avg_pool2d_1.run(arg0_1, buf1, 16, grid=grid(16), stream=stream0)
return (reinterpret_tensor(arg0_1, (4, 4, 4), (64, 16, 4), 0), reinterpret_tensor(buf0, (4, 2, 2), (4, 2, 1), 0), reinterpret_tensor(buf1, (4, 2, 2), (4, 2, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class chroma_subsampling(nn.Module):
""" Chroma subsampling on CbCv channels
Input:
image(tensor): batch x height x width x 3
Output:
y(tensor): batch x height x width
cb(tensor): batch x height/2 x width/2
cr(tensor): batch x height/2 x width/2
"""
def __init__(self):
super(chroma_subsampling, self).__init__()
def forward(self, image):
image_2 = image.permute(0, 3, 1, 2).clone()
avg_pool = nn.AvgPool2d(kernel_size=2, stride=(2, 2),
count_include_pad=False)
cb = avg_pool(image_2[:, 1, :, :].unsqueeze(1))
cr = avg_pool(image_2[:, 2, :, :].unsqueeze(1))
cb = cb.permute(0, 2, 3, 1)
cr = cr.permute(0, 2, 3, 1)
return image[:, :, :, 0], cb.squeeze(3), cr.squeeze(3)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_avg_pool2d_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 2
x1 = xindex // 2
x2 = xindex
tmp0 = tl.load(in_ptr0 + (1 + 8 * x0 + 32 * x1), xmask, eviction_policy
='evict_last')
tmp1 = tl.load(in_ptr0 + (5 + 8 * x0 + 32 * x1), xmask, eviction_policy
='evict_last')
tmp3 = tl.load(in_ptr0 + (17 + 8 * x0 + 32 * x1), xmask,
eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (21 + 8 * x0 + 32 * x1), xmask,
eviction_policy='evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_avg_pool2d_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 2
x1 = xindex // 2
x2 = xindex
tmp0 = tl.load(in_ptr0 + (2 + 8 * x0 + 32 * x1), xmask, eviction_policy
='evict_last')
tmp1 = tl.load(in_ptr0 + (6 + 8 * x0 + 32 * x1), xmask, eviction_policy
='evict_last')
tmp3 = tl.load(in_ptr0 + (18 + 8 * x0 + 32 * x1), xmask,
eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (22 + 8 * x0 + 32 * x1), xmask,
eviction_policy='evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 2, 2), (4, 1, 2, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_avg_pool2d_0[grid(16)](arg0_1, buf0, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf1 = empty_strided_cuda((4, 1, 2, 2), (4, 1, 2, 1), torch.float32)
triton_poi_fused_avg_pool2d_1[grid(16)](arg0_1, buf1, 16, XBLOCK=16,
num_warps=1, num_stages=1)
return reinterpret_tensor(arg0_1, (4, 4, 4), (64, 16, 4), 0
), reinterpret_tensor(buf0, (4, 2, 2), (4, 2, 1), 0
), reinterpret_tensor(buf1, (4, 2, 2), (4, 2, 1), 0)
class chroma_subsamplingNew(nn.Module):
""" Chroma subsampling on CbCv channels
Input:
image(tensor): batch x height x width x 3
Output:
y(tensor): batch x height x width
cb(tensor): batch x height/2 x width/2
cr(tensor): batch x height/2 x width/2
"""
def __init__(self):
super(chroma_subsamplingNew, self).__init__()
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0], output[1], output[2]
|
foxtrotmike/DiffJPEG
|
chroma_subsampling
| false | 12,396 |
[
"MIT"
] | 0 |
7dbc44b1e921f20a213a7206a8578d6a1c8131b4
|
https://github.com/foxtrotmike/DiffJPEG/tree/7dbc44b1e921f20a213a7206a8578d6a1c8131b4
|
idct_8x8
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/b4/cb4ar3ryyt4mp4d5ipmgfvm4veyllxljd2ymzyfikn6eexxqe2ep.py
# Topologically Sorted Source Nodes: [image], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# image => mul
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %primals_1), kwargs = {})
triton_poi_fused_mul_0 = async_compile.triton('triton_poi_fused_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x2), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/y5/cy5rhdemzf6dqbzpnuqnokcny6y6tjsftb3h6y7lac6lexyo22ba.py
# Topologically Sorted Source Nodes: [mul_1, result], Original ATen: [aten.mul, aten.add]
# Source node to ATen node mapping:
# mul_1 => mul_1
# result => add
# Graph fragment:
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_2, 0.25), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul_1, 128), kwargs = {})
triton_poi_fused_add_mul_1 = async_compile.triton('triton_poi_fused_add_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_mul_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_mul_1(in_out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + (x0), xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tmp3 = 128.0
tmp4 = tmp2 + tmp3
tl.store(in_out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (8, 8), (8, 1))
assert_size_stride(primals_2, (4, 4, 8, 8), (256, 64, 8, 1))
assert_size_stride(primals_3, (8, 8, 8, 8), (512, 64, 8, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 8, 8), (256, 64, 8, 1), torch.float32)
# Topologically Sorted Source Nodes: [image], Original ATen: [aten.mul]
stream0 = get_raw_stream(0)
triton_poi_fused_mul_0.run(primals_2, primals_1, buf0, 1024, grid=grid(1024), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((16, 64), (64, 1), torch.float32)
# Topologically Sorted Source Nodes: [tensordot], Original ATen: [aten.mm]
extern_kernels.mm(reinterpret_tensor(buf0, (16, 64), (64, 1), 0), reinterpret_tensor(primals_3, (64, 64), (64, 1), 0), out=buf1)
buf2 = reinterpret_tensor(buf1, (4, 4, 8, 8), (256, 64, 8, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [mul_1, result], Original ATen: [aten.mul, aten.add]
triton_poi_fused_add_mul_1.run(buf2, 1024, grid=grid(1024), stream=stream0)
return (buf2, primals_2, reinterpret_tensor(buf0, (64, 16), (1, 64), 0), reinterpret_tensor(primals_3, (64, 64), (1, 64), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((8, 8), (8, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 8, 8), (256, 64, 8, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((8, 8, 8, 8), (512, 64, 8, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import itertools
import torch
import numpy as np
import torch.nn as nn
class idct_8x8(nn.Module):
""" Inverse discrete Cosine Transformation
Input:
dcp(tensor): batch x height x width
Output:
image(tensor): batch x height x width
"""
def __init__(self):
super(idct_8x8, self).__init__()
alpha = np.array([1.0 / np.sqrt(2)] + [1] * 7)
self.alpha = nn.Parameter(torch.from_numpy(np.outer(alpha, alpha)).
float())
tensor = np.zeros((8, 8, 8, 8), dtype=np.float32)
for x, y, u, v in itertools.product(range(8), repeat=4):
tensor[x, y, u, v] = np.cos((2 * u + 1) * x * np.pi / 16) * np.cos(
(2 * v + 1) * y * np.pi / 16)
self.tensor = nn.Parameter(torch.from_numpy(tensor).float())
def forward(self, image):
image = image * self.alpha
result = 0.25 * torch.tensordot(image, self.tensor, dims=2) + 128
result.view(image.shape)
return result
def get_inputs():
return [torch.rand([4, 4, 8, 8])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import itertools
import numpy as np
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_mul_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x2, tmp2, xmask)
@triton.jit
def triton_poi_fused_add_mul_1(in_out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + x0, xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tmp3 = 128.0
tmp4 = tmp2 + tmp3
tl.store(in_out_ptr0 + x0, tmp4, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (8, 8), (8, 1))
assert_size_stride(primals_2, (4, 4, 8, 8), (256, 64, 8, 1))
assert_size_stride(primals_3, (8, 8, 8, 8), (512, 64, 8, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 8, 8), (256, 64, 8, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_mul_0[grid(1024)](primals_2, primals_1, buf0, 1024,
XBLOCK=256, num_warps=4, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((16, 64), (64, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf0, (16, 64), (64, 1), 0),
reinterpret_tensor(primals_3, (64, 64), (64, 1), 0), out=buf1)
buf2 = reinterpret_tensor(buf1, (4, 4, 8, 8), (256, 64, 8, 1), 0)
del buf1
triton_poi_fused_add_mul_1[grid(1024)](buf2, 1024, XBLOCK=128,
num_warps=4, num_stages=1)
return buf2, primals_2, reinterpret_tensor(buf0, (64, 16), (1, 64), 0
), reinterpret_tensor(primals_3, (64, 64), (1, 64), 0)
class idct_8x8New(nn.Module):
""" Inverse discrete Cosine Transformation
Input:
dcp(tensor): batch x height x width
Output:
image(tensor): batch x height x width
"""
def __init__(self):
super(idct_8x8New, self).__init__()
alpha = np.array([1.0 / np.sqrt(2)] + [1] * 7)
self.alpha = nn.Parameter(torch.from_numpy(np.outer(alpha, alpha)).
float())
tensor = np.zeros((8, 8, 8, 8), dtype=np.float32)
for x, y, u, v in itertools.product(range(8), repeat=4):
tensor[x, y, u, v] = np.cos((2 * u + 1) * x * np.pi / 16) * np.cos(
(2 * v + 1) * y * np.pi / 16)
self.tensor = nn.Parameter(torch.from_numpy(tensor).float())
def forward(self, input_0):
primals_1 = self.alpha
primals_3 = self.tensor
primals_2 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
foxtrotmike/DiffJPEG
|
idct_8x8
| false | 12,397 |
[
"MIT"
] | 0 |
7dbc44b1e921f20a213a7206a8578d6a1c8131b4
|
https://github.com/foxtrotmike/DiffJPEG/tree/7dbc44b1e921f20a213a7206a8578d6a1c8131b4
|
WidthXHeightXFeatureLinear
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ov/covytjt4qc2tfdmezks6pnavutef3fddru5qgd2fc3ef35jt766n.py
# Topologically Sorted Source Nodes: [truediv], Original ATen: [aten.div]
# Source node to ATen node mapping:
# truediv => div
# Graph fragment:
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%primals_2, %expand), kwargs = {})
triton_poi_fused_div_0 = async_compile.triton('triton_poi_fused_div_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp9 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 * tmp1
tmp4 = tmp3 * tmp3
tmp5 = tmp2 + tmp4
tmp7 = tmp6 * tmp6
tmp8 = tmp5 + tmp7
tmp10 = tmp9 * tmp9
tmp11 = tmp8 + tmp10
tmp12 = 1e-06
tmp13 = tmp11 + tmp12
tmp14 = libdevice.sqrt(tmp13)
tmp15 = tmp0 / tmp14
tl.store(out_ptr0 + (x2), tmp15, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ps/cpsfesmymo4w2d4yyjxtwxvjgu7xzxsnsoupbcv5qnvn5ntiunhx.py
# Topologically Sorted Source Nodes: [weight_2], Original ATen: [aten.mul]
# Source node to ATen node mapping:
# weight_2 => mul_1
# Graph fragment:
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%expand_4, %expand_5), kwargs = {})
triton_poi_fused_mul_1 = async_compile.triton('triton_poi_fused_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4) % 4
x3 = (xindex // 64)
x0 = xindex % 4
x6 = (xindex // 16)
x7 = xindex
tmp0 = tl.load(in_ptr0 + (x1 + (4*x3)), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr1 + (x0 + (4*x3)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr2 + (x6), xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp4 = tmp2 * tmp3
tl.store(out_ptr0 + (x7), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 1, 4, 1, 1), (4, 4, 1, 1, 1))
assert_size_stride(primals_3, (4, 1, 1, 4, 1), (4, 4, 4, 1, 1))
assert_size_stride(primals_4, (4, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_5, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 1, 1), (4, 16, 1, 16, 16), torch.float32)
# Topologically Sorted Source Nodes: [truediv], Original ATen: [aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_div_0.run(primals_2, buf0, 16, grid=grid(16), stream=stream0)
buf1 = empty_strided_cuda((4, 1, 1, 4, 1), (4, 16, 16, 1, 16), torch.float32)
# Topologically Sorted Source Nodes: [truediv_1], Original ATen: [aten.div]
triton_poi_fused_div_0.run(primals_3, buf1, 16, grid=grid(16), stream=stream0)
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [weight_2], Original ATen: [aten.mul]
triton_poi_fused_mul_1.run(buf0, buf1, primals_4, buf2, 256, grid=grid(256), stream=stream0)
del buf0
buf3 = reinterpret_tensor(buf1, (4, 4), (4, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.addmm(reinterpret_tensor(primals_5, (4, 4), (0, 1), 0), reinterpret_tensor(primals_1, (4, 64), (64, 1), 0), reinterpret_tensor(buf2, (64, 4), (1, 64), 0), alpha=1, beta=1, out=buf3)
del buf2
del primals_5
return (buf3, primals_2, primals_3, primals_4, reinterpret_tensor(primals_1, (4, 64), (64, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 1, 4, 1, 1), (4, 4, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 1, 1, 4, 1), (4, 4, 4, 1, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4, 1, 1), (4, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn import Parameter
def positive(weight, cache=None):
weight.data *= weight.data.ge(0).float()
return cache
class WidthXHeightXFeatureLinear(nn.Module):
"""
Factorized fully connected layer. Weights are a sum of outer products between three vectors over width,
height and spatial.
"""
def __init__(self, in_shape, outdims, components=1, bias=True,
normalize=True, positive=False, width=None, height=None, eps=1e-06):
super().__init__()
self.in_shape = in_shape
self.eps = eps
c, w, h = self.in_shape
self.outdims = outdims
self.normalize = normalize
self.positive = positive
self.components = components
self.width = Parameter(torch.Tensor(self.outdims, 1, w, 1, components)
) if width is None else width
self.height = Parameter(torch.Tensor(self.outdims, 1, 1, h, components)
) if height is None else height
self.features = Parameter(torch.Tensor(self.outdims, c, 1, 1))
assert self.width.size(4) == self.height.size(4
), 'The number of components in width and height do not agree'
self.components = self.width.size(4)
if bias:
bias = Parameter(torch.Tensor(self.outdims))
self.register_parameter('bias', bias)
else:
self.register_parameter('bias', None)
self.initialize()
def initialize(self, init_noise=0.001):
self.width.data.normal_(0, init_noise)
self.height.data.normal_(0, init_noise)
self.features.data.normal_(0, init_noise)
if self.bias is not None:
self.bias.data.fill_(0)
@property
def normalized_width(self):
if self.positive:
positive(self.width)
if self.normalize:
return self.width / (self.width.pow(2).sum(2, keepdim=True) +
self.eps).sqrt().expand_as(self.width)
else:
return self.width
@property
def normalized_height(self):
_c, _w, _h = self.in_shape
if self.positive:
positive(self.height)
if self.normalize:
return self.height / (self.height.pow(2).sum(3, keepdim=True) +
self.eps).sqrt().expand_as(self.height)
else:
return self.height
@property
def spatial(self):
_c, w, h = self.in_shape
n, comp = self.outdims, self.components
weight = self.normalized_width.expand(n, 1, w, h, comp
) * self.normalized_height.expand(n, 1, w, h, comp)
weight = weight.sum(4, keepdim=True).view(n, 1, w, h)
return weight
@property
def weight(self):
c, w, h = self.in_shape
n, _comp = self.outdims, self.components
weight = self.spatial.expand(n, c, w, h) * self.features.expand(n,
c, w, h)
weight = weight.view(self.outdims, -1)
return weight
@property
def basis(self):
c, w, h = self.in_shape
return self.weight.view(-1, c, w, h).data.cpu().numpy()
def forward(self, x):
N = x.size(0)
y = x.view(N, -1) @ self.weight.t()
if self.bias is not None:
y = y + self.bias.expand_as(y)
return y
def __repr__(self):
return ('spatial positive ' if self.positive else '') + ('normalized '
if self.normalize else ''
) + self.__class__.__name__ + ' (' + '{} x {} x {}'.format(*
self.in_shape) + ' -> ' + str(self.outdims
) + ') spatial rank {}'.format(self.components)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_shape': [4, 4, 4], 'outdims': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
from torch.nn import Parameter
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_div_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp9 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp2 = tmp1 * tmp1
tmp4 = tmp3 * tmp3
tmp5 = tmp2 + tmp4
tmp7 = tmp6 * tmp6
tmp8 = tmp5 + tmp7
tmp10 = tmp9 * tmp9
tmp11 = tmp8 + tmp10
tmp12 = 1e-06
tmp13 = tmp11 + tmp12
tmp14 = libdevice.sqrt(tmp13)
tmp15 = tmp0 / tmp14
tl.store(out_ptr0 + x2, tmp15, xmask)
@triton.jit
def triton_poi_fused_mul_1(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4 % 4
x3 = xindex // 64
x0 = xindex % 4
x6 = xindex // 16
x7 = xindex
tmp0 = tl.load(in_ptr0 + (x1 + 4 * x3), xmask, eviction_policy='evict_last'
)
tmp1 = tl.load(in_ptr1 + (x0 + 4 * x3), xmask, eviction_policy='evict_last'
)
tmp3 = tl.load(in_ptr2 + x6, xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp4 = tmp2 * tmp3
tl.store(out_ptr0 + x7, tmp4, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_2, (4, 1, 4, 1, 1), (4, 4, 1, 1, 1))
assert_size_stride(primals_3, (4, 1, 1, 4, 1), (4, 4, 4, 1, 1))
assert_size_stride(primals_4, (4, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_5, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 1, 1), (4, 16, 1, 16, 16),
torch.float32)
get_raw_stream(0)
triton_poi_fused_div_0[grid(16)](primals_2, buf0, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf1 = empty_strided_cuda((4, 1, 1, 4, 1), (4, 16, 16, 1, 16),
torch.float32)
triton_poi_fused_div_0[grid(16)](primals_3, buf1, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf2 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_mul_1[grid(256)](buf0, buf1, primals_4, buf2, 256,
XBLOCK=128, num_warps=4, num_stages=1)
del buf0
buf3 = reinterpret_tensor(buf1, (4, 4), (4, 1), 0)
del buf1
extern_kernels.addmm(reinterpret_tensor(primals_5, (4, 4), (0, 1),
0), reinterpret_tensor(primals_1, (4, 64), (64, 1), 0),
reinterpret_tensor(buf2, (64, 4), (1, 64), 0), alpha=1, beta=1,
out=buf3)
del buf2
del primals_5
return buf3, primals_2, primals_3, primals_4, reinterpret_tensor(primals_1,
(4, 64), (64, 1), 0)
def positive(weight, cache=None):
weight.data *= weight.data.ge(0).float()
return cache
class WidthXHeightXFeatureLinearNew(nn.Module):
"""
Factorized fully connected layer. Weights are a sum of outer products between three vectors over width,
height and spatial.
"""
def __init__(self, in_shape, outdims, components=1, bias=True,
normalize=True, positive=False, width=None, height=None, eps=1e-06):
super().__init__()
self.in_shape = in_shape
self.eps = eps
c, w, h = self.in_shape
self.outdims = outdims
self.normalize = normalize
self.positive = positive
self.components = components
self.width = Parameter(torch.Tensor(self.outdims, 1, w, 1, components)
) if width is None else width
self.height = Parameter(torch.Tensor(self.outdims, 1, 1, h, components)
) if height is None else height
self.features = Parameter(torch.Tensor(self.outdims, c, 1, 1))
assert self.width.size(4) == self.height.size(4
), 'The number of components in width and height do not agree'
self.components = self.width.size(4)
if bias:
bias = Parameter(torch.Tensor(self.outdims))
self.register_parameter('bias', bias)
else:
self.register_parameter('bias', None)
self.initialize()
def initialize(self, init_noise=0.001):
self.width.data.normal_(0, init_noise)
self.height.data.normal_(0, init_noise)
self.features.data.normal_(0, init_noise)
if self.bias is not None:
self.bias.data.fill_(0)
@property
def normalized_width(self):
if self.positive:
positive(self.width)
if self.normalize:
return self.width / (self.width.pow(2).sum(2, keepdim=True) +
self.eps).sqrt().expand_as(self.width)
else:
return self.width
@property
def normalized_height(self):
_c, _w, _h = self.in_shape
if self.positive:
positive(self.height)
if self.normalize:
return self.height / (self.height.pow(2).sum(3, keepdim=True) +
self.eps).sqrt().expand_as(self.height)
else:
return self.height
@property
def spatial(self):
_c, w, h = self.in_shape
n, comp = self.outdims, self.components
weight = self.normalized_width.expand(n, 1, w, h, comp
) * self.normalized_height.expand(n, 1, w, h, comp)
weight = weight.sum(4, keepdim=True).view(n, 1, w, h)
return weight
@property
def weight(self):
c, w, h = self.in_shape
n, _comp = self.outdims, self.components
weight = self.spatial.expand(n, c, w, h) * self.features.expand(n,
c, w, h)
weight = weight.view(self.outdims, -1)
return weight
@property
def basis(self):
c, w, h = self.in_shape
return self.weight.view(-1, c, w, h).data.cpu().numpy()
def __repr__(self):
return ('spatial positive ' if self.positive else '') + ('normalized '
if self.normalize else ''
) + self.__class__.__name__ + ' (' + '{} x {} x {}'.format(*
self.in_shape) + ' -> ' + str(self.outdims
) + ') spatial rank {}'.format(self.components)
def forward(self, input_0):
primals_2 = self.width
primals_3 = self.height
primals_4 = self.features
primals_5 = self.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4, primals_5])
return output[0]
|
dattientran/attorch
|
WidthXHeightXFeatureLinear
| false | 12,398 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
SpatialTransformerXPooled3d
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/7e/c7epu6wqqthyzjiqh2rjoko2wjfngavr2n5s7n5u7qcbhdom27ph.py
# Topologically Sorted Source Nodes: [clamp], Original ATen: [aten.clamp]
# Source node to ATen node mapping:
# clamp => clamp_max, clamp_min
# Graph fragment:
# %clamp_min : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%primals_1, -1), kwargs = {})
# %clamp_max : [num_users=3] = call_function[target=torch.ops.aten.clamp_max.default](args = (%clamp_min, 1), kwargs = {})
triton_poi_fused_clamp_0 = async_compile.triton('triton_poi_fused_clamp_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clamp_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clamp_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 80
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = -1.0
tmp2 = triton_helpers.maximum(tmp0, tmp1)
tmp3 = 1.0
tmp4 = triton_helpers.minimum(tmp2, tmp3)
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/l4/cl4ehc2pkc5zdczwv6muo6fv2f2wt3nw7b4pjkbkbkuwt5fawxpw.py
# Topologically Sorted Source Nodes: [contiguous_1], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# contiguous_1 => clone
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_1 = async_compile.triton('triton_poi_fused_clone_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = (xindex // 16) % 4
x2 = (xindex // 64) % 4
x3 = (xindex // 256)
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (16*x2) + (64*x1) + (256*x3)), xmask)
tl.store(out_ptr0 + (x4), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/fx/cfx5dbppafovirj4qxi4bvhx2h2sl4mdxnfbevky463weqfqpyxq.py
# Topologically Sorted Source Nodes: [mean], Original ATen: [aten.mean]
# Source node to ATen node mapping:
# mean => mean
# Graph fragment:
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%cudnn_grid_sampler, [3], True), kwargs = {})
triton_per_fused_mean_2 = async_compile.triton('triton_per_fused_mean_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[256, 16],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_mean_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_mean_2(in_ptr0, out_ptr1, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 256
rnumel = 10
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = rindex < rnumel
r1 = rindex
x0 = xindex
x2 = xindex % 16
x3 = (xindex // 16)
tmp0 = tl.load(in_ptr0 + (r1 + (10*x0)), rmask & xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(rmask & xmask, tmp1, 0)
tmp4 = tl.sum(tmp3, 1)[:, None]
tmp5 = 10.0
tmp6 = tmp4 / tmp5
tl.store(out_ptr1 + (x2 + (32*x3)), tmp6, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6g/c6gcqfkqk4z6xxi72qlctglv77f2rr4dspwv4beszms2httz4yrg.py
# Topologically Sorted Source Nodes: [z_1], Original ATen: [aten.avg_pool2d]
# Source node to ATen node mapping:
# z_1 => avg_pool2d
# Graph fragment:
# %avg_pool2d : [num_users=2] = call_function[target=torch.ops.aten.avg_pool2d.default](args = (%view_1, [4, 4], [4, 4], [0, 0], False, False), kwargs = {})
triton_poi_fused_avg_pool2d_3 = async_compile.triton('triton_poi_fused_avg_pool2d_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_avg_pool2d_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 16, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_avg_pool2d_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (16*x0), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + (16*x0)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (2 + (16*x0)), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (3 + (16*x0)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr0 + (4 + (16*x0)), xmask, eviction_policy='evict_last')
tmp9 = tl.load(in_ptr0 + (5 + (16*x0)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (6 + (16*x0)), xmask, eviction_policy='evict_last')
tmp13 = tl.load(in_ptr0 + (7 + (16*x0)), xmask, eviction_policy='evict_last')
tmp15 = tl.load(in_ptr0 + (8 + (16*x0)), xmask, eviction_policy='evict_last')
tmp17 = tl.load(in_ptr0 + (9 + (16*x0)), xmask, eviction_policy='evict_last')
tmp19 = tl.load(in_ptr0 + (10 + (16*x0)), xmask, eviction_policy='evict_last')
tmp21 = tl.load(in_ptr0 + (11 + (16*x0)), xmask, eviction_policy='evict_last')
tmp23 = tl.load(in_ptr0 + (12 + (16*x0)), xmask, eviction_policy='evict_last')
tmp25 = tl.load(in_ptr0 + (13 + (16*x0)), xmask, eviction_policy='evict_last')
tmp27 = tl.load(in_ptr0 + (14 + (16*x0)), xmask, eviction_policy='evict_last')
tmp29 = tl.load(in_ptr0 + (15 + (16*x0)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp8 = tmp7 + tmp6
tmp10 = tmp9 + tmp8
tmp12 = tmp11 + tmp10
tmp14 = tmp13 + tmp12
tmp16 = tmp15 + tmp14
tmp18 = tmp17 + tmp16
tmp20 = tmp19 + tmp18
tmp22 = tmp21 + tmp20
tmp24 = tmp23 + tmp22
tmp26 = tmp25 + tmp24
tmp28 = tmp27 + tmp26
tmp30 = tmp29 + tmp28
tmp31 = 0.0625
tmp32 = tmp30 * tmp31
tl.store(out_ptr0 + (x0), tmp32, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/z3/cz3r2grxrxdglbhdsscuaklo2hir5scxptlce76brrdddnxxk4uo.py
# Topologically Sorted Source Nodes: [mul, sum_1, y_2], Original ATen: [aten.mul, aten.sum, aten.add]
# Source node to ATen node mapping:
# mul => mul
# sum_1 => sum_1
# y_2 => add
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze, %view), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul, [1]), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%view_2, %primals_4), kwargs = {})
triton_per_fused_add_mul_sum_4 = async_compile.triton('triton_per_fused_add_mul_sum_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[64, 8],
reduction_hint=ReductionHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_mul_sum_4', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_mul_sum_4(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 64
rnumel = 8
RBLOCK: tl.constexpr = 8
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r2 = rindex
x0 = xindex % 4
x1 = (xindex // 4)
x3 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (4*r2) + (32*x1)), xmask, other=0.0)
tmp1 = tl.load(in_ptr1 + (x0 + (4*r2)), xmask, eviction_policy='evict_last', other=0.0)
tmp7 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp3 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5 = tl.where(xmask, tmp3, 0)
tmp6 = tl.sum(tmp5, 1)[:, None]
tmp8 = tmp6 + tmp7
tl.debug_barrier()
tl.store(in_out_ptr0 + (x3), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 4, 10, 2), (80, 20, 2, 1))
assert_size_stride(primals_2, (4, 4, 4, 4, 4), (256, 64, 16, 4, 1))
assert_size_stride(primals_3, (1, 8, 1, 4), (32, 4, 4, 1))
assert_size_stride(primals_4, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 4, 10, 2), (80, 20, 2, 1), torch.float32)
# Topologically Sorted Source Nodes: [clamp], Original ATen: [aten.clamp]
stream0 = get_raw_stream(0)
triton_poi_fused_clamp_0.run(primals_1, buf0, 80, grid=grid(80), stream=stream0)
buf1 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [contiguous_1], Original ATen: [aten.clone]
triton_poi_fused_clone_1.run(primals_2, buf1, 1024, grid=grid(1024), stream=stream0)
del primals_2
# Topologically Sorted Source Nodes: [grid_sample], Original ATen: [aten.cudnn_grid_sampler]
buf2 = torch.ops.aten.cudnn_grid_sampler.default(reinterpret_tensor(buf1, (16, 4, 4, 4), (64, 16, 4, 1), 0), reinterpret_tensor(buf0, (16, 4, 10, 2), (0, 20, 2, 1), 0))
buf3 = buf2
buf11 = empty_strided_cuda((16, 8, 4, 1), (32, 4, 1, 1), torch.float32)
buf9 = reinterpret_tensor(buf11, (16, 4, 4, 1), (32, 4, 1, 1), 0) # alias
# Topologically Sorted Source Nodes: [mean], Original ATen: [aten.mean]
triton_per_fused_mean_2.run(buf3, buf9, 256, 10, grid=grid(256), stream=stream0)
del buf3
buf5 = empty_strided_cuda((16, 4, 1, 1), (4, 1, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [z_1], Original ATen: [aten.avg_pool2d]
triton_poi_fused_avg_pool2d_3.run(buf1, buf5, 64, grid=grid(64), stream=stream0)
# Topologically Sorted Source Nodes: [grid_sample_1], Original ATen: [aten.cudnn_grid_sampler]
buf6 = torch.ops.aten.cudnn_grid_sampler.default(buf5, reinterpret_tensor(buf0, (16, 4, 10, 2), (0, 20, 2, 1), 0))
buf7 = buf6
buf10 = reinterpret_tensor(buf11, (16, 4, 4, 1), (32, 4, 1, 1), 16) # alias
# Topologically Sorted Source Nodes: [mean_1], Original ATen: [aten.mean]
triton_per_fused_mean_2.run(buf7, buf10, 256, 10, grid=grid(256), stream=stream0)
del buf7
buf12 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
buf13 = reinterpret_tensor(buf12, (4, 4, 4), (16, 4, 1), 0); del buf12 # reuse
# Topologically Sorted Source Nodes: [mul, sum_1, y_2], Original ATen: [aten.mul, aten.sum, aten.add]
triton_per_fused_add_mul_sum_4.run(buf13, buf11, primals_3, primals_4, 64, 8, grid=grid(64), stream=stream0)
del primals_4
# Topologically Sorted Source Nodes: [], Original ATen: []
buf14 = torch.ops.aten.set_.source_Tensor(primals_1, buf0)
assert_size_stride(buf14, (1, 4, 10, 2), (80, 20, 2, 1))
del buf2
del buf6
del primals_1
return (buf13, primals_3, buf0, reinterpret_tensor(buf1, (16, 4, 4, 4), (64, 16, 4, 1), 0), buf5, buf11, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 4, 10, 2), (80, 20, 2, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4, 4), (256, 64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((1, 8, 1, 4), (32, 4, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn import Parameter
from torch.nn import functional as F
def positive(weight, cache=None):
weight.data *= weight.data.ge(0).float()
return cache
class SpatialTransformerXPooled3d(nn.Module):
def __init__(self, in_shape, outdims, pool_steps=1, positive=False,
bias=True, init_range=0.2, grid_points=10, kernel_size=4, stride=4,
grid=None, stop_grad=False):
super().__init__()
self._pool_steps = pool_steps
self.in_shape = in_shape
c, _t, _w, _h = in_shape
self.outdims = outdims
self.positive = positive
self._grid_points = grid_points
if grid is None:
self.grid = Parameter(torch.Tensor(1, outdims, grid_points, 2))
else:
self.grid = grid
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps + 1
), 1, outdims))
if bias:
bias = Parameter(torch.Tensor(outdims))
self.register_parameter('bias', bias)
else:
self.register_parameter('bias', None)
self.avg = nn.AvgPool2d(kernel_size, stride=stride,
count_include_pad=False)
self.init_range = init_range
self.initialize()
self.stop_grad = stop_grad
@property
def pool_steps(self):
return self._pool_steps
@pool_steps.setter
def pool_steps(self, value):
assert value >= 0 and int(value
) - value == 0, 'new pool steps must be a non-negative integer'
if value != self._pool_steps:
None
c, _t, _w, _h = self.in_shape
outdims = self.outdims
self._pool_steps = int(value)
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps +
1), 1, outdims))
self.features.data.fill_(1 / self.in_shape[0])
def initialize(self, init_noise=0.001, grid=True):
self.features.data.fill_(1 / self.in_shape[0])
if self.bias is not None:
self.bias.data.fill_(0)
if grid:
self.grid.data.uniform_(-self.init_range, self.init_range)
def feature_l1(self, average=True, subs_idx=None):
subs_idx = subs_idx if subs_idx is not None else slice(None)
if average:
return self.features[..., subs_idx].abs().mean()
else:
return self.features[..., subs_idx].abs().sum()
def dgrid_l2(self, average=True, subs_idx=None):
subs_idx = subs_idx if subs_idx is not None else slice(None)
if average:
return (self.grid[:, subs_idx, :-1, :] - self.grid[:, subs_idx,
1:, :]).pow(2).mean()
else:
return (self.grid[:, subs_idx, :-1, :] - self.grid[:, subs_idx,
1:, :]).pow(2).sum()
def forward(self, x, shift=None, subs_idx=None):
if self.stop_grad:
x = x.detach()
if self.positive:
positive(self.features)
self.grid.data = torch.clamp(self.grid.data, -1, 1)
N, c, t, w, h = x.size()
m = self._pool_steps + 1
if subs_idx is not None:
feat = self.features[..., subs_idx].contiguous()
outdims = feat.size(-1)
feat = feat.view(1, m * c, outdims)
grid = self.grid[:, subs_idx, ...]
else:
grid = self.grid
feat = self.features.view(1, m * c, self.outdims)
outdims = self.outdims
if shift is None:
grid = grid.expand(N * t, outdims, self._grid_points, 2)
else:
grid = grid.expand(N, outdims, self._grid_points, 2)
grid = torch.stack([(grid + shift[:, i, :][:, None, None, :]) for
i in range(t)], 1)
grid = grid.contiguous().view(-1, outdims, self._grid_points, 2)
z = x.contiguous().transpose(2, 1).contiguous().view(-1, c, w, h)
pools = [F.grid_sample(z, grid, align_corners=True).mean(dim=3,
keepdim=True)]
for i in range(self._pool_steps):
z = self.avg(z)
pools.append(F.grid_sample(z, grid, align_corners=True).mean(
dim=3, keepdim=True))
y = torch.cat(pools, dim=1)
y = (y.squeeze(-1) * feat).sum(1).view(N, t, outdims)
if self.bias is not None:
if subs_idx is None:
y = y + self.bias
else:
y = y + self.bias[subs_idx]
return y
def __repr__(self):
c, _, w, h = self.in_shape
r = self.__class__.__name__ + ' (' + '{} x {} x {}'.format(c, w, h
) + ' -> ' + str(self.outdims) + ')'
if self.bias is not None:
r += ' with bias'
if self.stop_grad:
r += ', stop_grad=True'
r += '\n'
for ch in self.children():
r += ' -> ' + ch.__repr__() + '\n'
return r
def get_inputs():
return [torch.rand([4, 4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_shape': [4, 4, 4, 4], 'outdims': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch import nn
from torch.nn import Parameter
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_clamp_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 80
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = -1.0
tmp2 = triton_helpers.maximum(tmp0, tmp1)
tmp3 = 1.0
tmp4 = triton_helpers.minimum(tmp2, tmp3)
tl.store(out_ptr0 + x0, tmp4, xmask)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = xindex // 16 % 4
x2 = xindex // 64 % 4
x3 = xindex // 256
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 16 * x2 + 64 * x1 + 256 * x3), xmask)
tl.store(out_ptr0 + x4, tmp0, xmask)
@triton.jit
def triton_per_fused_mean_2(in_ptr0, out_ptr1, xnumel, rnumel, XBLOCK: tl.
constexpr):
xnumel = 256
rnumel = 10
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
rmask = rindex < rnumel
r1 = rindex
x0 = xindex
x2 = xindex % 16
x3 = xindex // 16
tmp0 = tl.load(in_ptr0 + (r1 + 10 * x0), rmask & xmask, other=0.0)
tmp1 = tl.broadcast_to(tmp0, [XBLOCK, RBLOCK])
tmp3 = tl.where(rmask & xmask, tmp1, 0)
tmp4 = tl.sum(tmp3, 1)[:, None]
tmp5 = 10.0
tmp6 = tmp4 / tmp5
tl.store(out_ptr1 + (x2 + 32 * x3), tmp6, xmask)
@triton.jit
def triton_poi_fused_avg_pool2d_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 16 * x0, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + 16 * x0), xmask, eviction_policy='evict_last'
)
tmp3 = tl.load(in_ptr0 + (2 + 16 * x0), xmask, eviction_policy='evict_last'
)
tmp5 = tl.load(in_ptr0 + (3 + 16 * x0), xmask, eviction_policy='evict_last'
)
tmp7 = tl.load(in_ptr0 + (4 + 16 * x0), xmask, eviction_policy='evict_last'
)
tmp9 = tl.load(in_ptr0 + (5 + 16 * x0), xmask, eviction_policy='evict_last'
)
tmp11 = tl.load(in_ptr0 + (6 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp13 = tl.load(in_ptr0 + (7 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp15 = tl.load(in_ptr0 + (8 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp17 = tl.load(in_ptr0 + (9 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp19 = tl.load(in_ptr0 + (10 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp21 = tl.load(in_ptr0 + (11 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp23 = tl.load(in_ptr0 + (12 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp25 = tl.load(in_ptr0 + (13 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp27 = tl.load(in_ptr0 + (14 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp29 = tl.load(in_ptr0 + (15 + 16 * x0), xmask, eviction_policy=
'evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp8 = tmp7 + tmp6
tmp10 = tmp9 + tmp8
tmp12 = tmp11 + tmp10
tmp14 = tmp13 + tmp12
tmp16 = tmp15 + tmp14
tmp18 = tmp17 + tmp16
tmp20 = tmp19 + tmp18
tmp22 = tmp21 + tmp20
tmp24 = tmp23 + tmp22
tmp26 = tmp25 + tmp24
tmp28 = tmp27 + tmp26
tmp30 = tmp29 + tmp28
tmp31 = 0.0625
tmp32 = tmp30 * tmp31
tl.store(out_ptr0 + x0, tmp32, xmask)
@triton.jit
def triton_per_fused_add_mul_sum_4(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2,
xnumel, rnumel, XBLOCK: tl.constexpr):
xnumel = 64
RBLOCK: tl.constexpr = 8
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r2 = rindex
x0 = xindex % 4
x1 = xindex // 4
x3 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 4 * r2 + 32 * x1), xmask, other=0.0)
tmp1 = tl.load(in_ptr1 + (x0 + 4 * r2), xmask, eviction_policy=
'evict_last', other=0.0)
tmp7 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp3 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5 = tl.where(xmask, tmp3, 0)
tmp6 = tl.sum(tmp5, 1)[:, None]
tmp8 = tmp6 + tmp7
tl.debug_barrier()
tl.store(in_out_ptr0 + x3, tmp8, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 4, 10, 2), (80, 20, 2, 1))
assert_size_stride(primals_2, (4, 4, 4, 4, 4), (256, 64, 16, 4, 1))
assert_size_stride(primals_3, (1, 8, 1, 4), (32, 4, 4, 1))
assert_size_stride(primals_4, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 4, 10, 2), (80, 20, 2, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_clamp_0[grid(80)](primals_1, buf0, 80, XBLOCK=128,
num_warps=4, num_stages=1)
buf1 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1),
torch.float32)
triton_poi_fused_clone_1[grid(1024)](primals_2, buf1, 1024, XBLOCK=
256, num_warps=4, num_stages=1)
del primals_2
buf2 = torch.ops.aten.cudnn_grid_sampler.default(reinterpret_tensor
(buf1, (16, 4, 4, 4), (64, 16, 4, 1), 0), reinterpret_tensor(
buf0, (16, 4, 10, 2), (0, 20, 2, 1), 0))
buf3 = buf2
buf11 = empty_strided_cuda((16, 8, 4, 1), (32, 4, 1, 1), torch.float32)
buf9 = reinterpret_tensor(buf11, (16, 4, 4, 1), (32, 4, 1, 1), 0)
triton_per_fused_mean_2[grid(256)](buf3, buf9, 256, 10, XBLOCK=8,
num_warps=2, num_stages=1)
del buf3
buf5 = empty_strided_cuda((16, 4, 1, 1), (4, 1, 1, 1), torch.float32)
triton_poi_fused_avg_pool2d_3[grid(64)](buf1, buf5, 64, XBLOCK=64,
num_warps=1, num_stages=1)
buf6 = torch.ops.aten.cudnn_grid_sampler.default(buf5,
reinterpret_tensor(buf0, (16, 4, 10, 2), (0, 20, 2, 1), 0))
buf7 = buf6
buf10 = reinterpret_tensor(buf11, (16, 4, 4, 1), (32, 4, 1, 1), 16)
triton_per_fused_mean_2[grid(256)](buf7, buf10, 256, 10, XBLOCK=8,
num_warps=2, num_stages=1)
del buf7
buf12 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
buf13 = reinterpret_tensor(buf12, (4, 4, 4), (16, 4, 1), 0)
del buf12
triton_per_fused_add_mul_sum_4[grid(64)](buf13, buf11, primals_3,
primals_4, 64, 8, XBLOCK=1, num_warps=2, num_stages=1)
del primals_4
buf14 = torch.ops.aten.set_.source_Tensor(primals_1, buf0)
assert_size_stride(buf14, (1, 4, 10, 2), (80, 20, 2, 1))
del buf2
del buf6
del primals_1
return buf13, primals_3, buf0, reinterpret_tensor(buf1, (16, 4, 4, 4),
(64, 16, 4, 1), 0), buf5, buf11
def positive(weight, cache=None):
weight.data *= weight.data.ge(0).float()
return cache
class SpatialTransformerXPooled3dNew(nn.Module):
def __init__(self, in_shape, outdims, pool_steps=1, positive=False,
bias=True, init_range=0.2, grid_points=10, kernel_size=4, stride=4,
grid=None, stop_grad=False):
super().__init__()
self._pool_steps = pool_steps
self.in_shape = in_shape
c, _t, _w, _h = in_shape
self.outdims = outdims
self.positive = positive
self._grid_points = grid_points
if grid is None:
self.grid = Parameter(torch.Tensor(1, outdims, grid_points, 2))
else:
self.grid = grid
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps + 1
), 1, outdims))
if bias:
bias = Parameter(torch.Tensor(outdims))
self.register_parameter('bias', bias)
else:
self.register_parameter('bias', None)
self.avg = nn.AvgPool2d(kernel_size, stride=stride,
count_include_pad=False)
self.init_range = init_range
self.initialize()
self.stop_grad = stop_grad
@property
def pool_steps(self):
return self._pool_steps
@pool_steps.setter
def pool_steps(self, value):
assert value >= 0 and int(value
) - value == 0, 'new pool steps must be a non-negative integer'
if value != self._pool_steps:
None
c, _t, _w, _h = self.in_shape
outdims = self.outdims
self._pool_steps = int(value)
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps +
1), 1, outdims))
self.features.data.fill_(1 / self.in_shape[0])
def initialize(self, init_noise=0.001, grid=True):
self.features.data.fill_(1 / self.in_shape[0])
if self.bias is not None:
self.bias.data.fill_(0)
if grid:
self.grid.data.uniform_(-self.init_range, self.init_range)
def feature_l1(self, average=True, subs_idx=None):
subs_idx = subs_idx if subs_idx is not None else slice(None)
if average:
return self.features[..., subs_idx].abs().mean()
else:
return self.features[..., subs_idx].abs().sum()
def dgrid_l2(self, average=True, subs_idx=None):
subs_idx = subs_idx if subs_idx is not None else slice(None)
if average:
return (self.grid[:, subs_idx, :-1, :] - self.grid[:, subs_idx,
1:, :]).pow(2).mean()
else:
return (self.grid[:, subs_idx, :-1, :] - self.grid[:, subs_idx,
1:, :]).pow(2).sum()
def __repr__(self):
c, _, w, h = self.in_shape
r = self.__class__.__name__ + ' (' + '{} x {} x {}'.format(c, w, h
) + ' -> ' + str(self.outdims) + ')'
if self.bias is not None:
r += ' with bias'
if self.stop_grad:
r += ', stop_grad=True'
r += '\n'
for ch in self.children():
r += ' -> ' + ch.__repr__() + '\n'
return r
def forward(self, input_0):
primals_1 = self.grid
primals_3 = self.features
primals_4 = self.bias
primals_2 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
dattientran/attorch
|
SpatialTransformerXPooled3d
| false | 12,399 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
GAT
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/t4/ct4b67pdo2nqkmug5ve6psoz6ptovf44cjwac2selnsbhojvain4.py
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cat => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%view, %repeat_1], 1), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x1 = (xindex // 8)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*((((4*x1) + x0) // 16) % 4)) + ((((4*x1) + x0) % 16) % 4)), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr0 + ((4*(x1 % 4)) + ((-4) + x0)), tmp6 & xmask, eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x2), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/rs/crsikvivgof4u6qcelh3gov7oade5uaprup6quh2r4qjgsssen7k.py
# Topologically Sorted Source Nodes: [e], Original ATen: [aten.leaky_relu]
# Source node to ATen node mapping:
# e => gt
# Graph fragment:
# %gt : [num_users=2] = call_function[target=torch.ops.aten.gt.Scalar](args = (%squeeze, 0), kwargs = {})
triton_poi_fused_leaky_relu_1 = async_compile.triton('triton_poi_fused_leaky_relu_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i1', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_leaky_relu_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_leaky_relu_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tl.store(out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2k/c2k6idl77paa5lpvj6v4mi3dzsgbzy45voclv5jrtwxnvtfb6k4v.py
# Topologically Sorted Source Nodes: [e, zero_vec, attention, attention_1, e_1, attention_3, attention_4, e_2, attention_6, attention_7, e_3, attention_9, attention_10], Original ATen: [aten.leaky_relu, aten.mul, aten.where, aten._softmax]
# Source node to ATen node mapping:
# attention => where_1
# attention_1 => amax, exp, sub, sum_1
# attention_10 => amax_3, exp_3, sub_3, sum_4
# attention_3 => where_4
# attention_4 => amax_1, exp_1, sub_1, sum_2
# attention_6 => where_7
# attention_7 => amax_2, exp_2, sub_2, sum_3
# attention_9 => where_10
# e => mul, where
# e_1 => mul_5, where_3
# e_2 => mul_10, where_6
# e_3 => mul_15, where_9
# zero_vec => full_default
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze, 4), kwargs = {})
# %where : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt, %squeeze, %mul), kwargs = {})
# %full_default : [num_users=5] = call_function[target=torch.ops.aten.full.default](args = ([4, 4], -8999999815811072.0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %where_1 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where, %full_default), kwargs = {})
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%where_1, [1], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_1, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [1], True), kwargs = {})
# %mul_5 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_1, 4), kwargs = {})
# %where_3 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_3, %squeeze_1, %mul_5), kwargs = {})
# %where_4 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_3, %full_default), kwargs = {})
# %amax_1 : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%where_4, [1], True), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_4, %amax_1), kwargs = {})
# %exp_1 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_1,), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_1, [1], True), kwargs = {})
# %mul_10 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_2, 4), kwargs = {})
# %where_6 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_6, %squeeze_2, %mul_10), kwargs = {})
# %where_7 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_6, %full_default), kwargs = {})
# %amax_2 : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%where_7, [1], True), kwargs = {})
# %sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_7, %amax_2), kwargs = {})
# %exp_2 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_2,), kwargs = {})
# %sum_3 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_2, [1], True), kwargs = {})
# %mul_15 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_3, 4), kwargs = {})
# %where_9 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_9, %squeeze_3, %mul_15), kwargs = {})
# %where_10 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_9, %full_default), kwargs = {})
# %amax_3 : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%where_10, [1], True), kwargs = {})
# %sub_3 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_10, %amax_3), kwargs = {})
# %exp_3 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_3,), kwargs = {})
# %sum_4 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_3, [1], True), kwargs = {})
triton_poi_fused__softmax_leaky_relu_mul_where_2 = async_compile.triton('triton_poi_fused__softmax_leaky_relu_mul_where_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4],
filename=__file__,
triton_meta={'signature': {0: '*i1', 1: '*i1', 2: '*fp32', 3: '*i1', 4: '*fp32', 5: '*i1', 6: '*fp32', 7: '*i1', 8: '*fp32', 9: '*fp32', 10: '*fp32', 11: '*fp32', 12: '*fp32', 13: '*fp32', 14: '*fp32', 15: '*fp32', 16: '*fp32', 17: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_leaky_relu_mul_where_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 36, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_2(in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, in_ptr6, in_ptr7, in_ptr8, out_ptr0, out_ptr1, out_ptr2, out_ptr3, out_ptr4, out_ptr5, out_ptr6, out_ptr7, xnumel, XBLOCK : tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (4*x0), xmask, eviction_policy='evict_last').to(tl.int1)
tmp1 = tl.load(in_ptr1 + (4*x0), xmask, eviction_policy='evict_last').to(tl.int1)
tmp2 = tl.load(in_ptr2 + (4*x0), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (1 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp9 = tl.load(in_ptr1 + (1 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp10 = tl.load(in_ptr2 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp15 = tl.load(in_ptr0 + (2 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp16 = tl.load(in_ptr1 + (2 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp17 = tl.load(in_ptr2 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp22 = tl.load(in_ptr0 + (3 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp23 = tl.load(in_ptr1 + (3 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp24 = tl.load(in_ptr2 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp40 = tl.load(in_ptr3 + (4*x0), xmask, eviction_policy='evict_last').to(tl.int1)
tmp41 = tl.load(in_ptr4 + (4*x0), xmask, eviction_policy='evict_last')
tmp45 = tl.load(in_ptr3 + (1 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp46 = tl.load(in_ptr4 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp51 = tl.load(in_ptr3 + (2 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp52 = tl.load(in_ptr4 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp57 = tl.load(in_ptr3 + (3 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp58 = tl.load(in_ptr4 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp74 = tl.load(in_ptr5 + (4*x0), xmask, eviction_policy='evict_last').to(tl.int1)
tmp75 = tl.load(in_ptr6 + (4*x0), xmask, eviction_policy='evict_last')
tmp79 = tl.load(in_ptr5 + (1 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp80 = tl.load(in_ptr6 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp85 = tl.load(in_ptr5 + (2 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp86 = tl.load(in_ptr6 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp91 = tl.load(in_ptr5 + (3 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp92 = tl.load(in_ptr6 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp108 = tl.load(in_ptr7 + (4*x0), xmask, eviction_policy='evict_last').to(tl.int1)
tmp109 = tl.load(in_ptr8 + (4*x0), xmask, eviction_policy='evict_last')
tmp113 = tl.load(in_ptr7 + (1 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp114 = tl.load(in_ptr8 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp119 = tl.load(in_ptr7 + (2 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp120 = tl.load(in_ptr8 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp125 = tl.load(in_ptr7 + (3 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp126 = tl.load(in_ptr8 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp11 = tmp10 * tmp3
tmp12 = tl.where(tmp9, tmp10, tmp11)
tmp13 = tl.where(tmp8, tmp12, tmp6)
tmp14 = triton_helpers.maximum(tmp7, tmp13)
tmp18 = tmp17 * tmp3
tmp19 = tl.where(tmp16, tmp17, tmp18)
tmp20 = tl.where(tmp15, tmp19, tmp6)
tmp21 = triton_helpers.maximum(tmp14, tmp20)
tmp25 = tmp24 * tmp3
tmp26 = tl.where(tmp23, tmp24, tmp25)
tmp27 = tl.where(tmp22, tmp26, tmp6)
tmp28 = triton_helpers.maximum(tmp21, tmp27)
tmp29 = tmp7 - tmp28
tmp30 = tl_math.exp(tmp29)
tmp31 = tmp13 - tmp28
tmp32 = tl_math.exp(tmp31)
tmp33 = tmp30 + tmp32
tmp34 = tmp20 - tmp28
tmp35 = tl_math.exp(tmp34)
tmp36 = tmp33 + tmp35
tmp37 = tmp27 - tmp28
tmp38 = tl_math.exp(tmp37)
tmp39 = tmp36 + tmp38
tmp42 = tmp41 * tmp3
tmp43 = tl.where(tmp40, tmp41, tmp42)
tmp44 = tl.where(tmp0, tmp43, tmp6)
tmp47 = tmp46 * tmp3
tmp48 = tl.where(tmp45, tmp46, tmp47)
tmp49 = tl.where(tmp8, tmp48, tmp6)
tmp50 = triton_helpers.maximum(tmp44, tmp49)
tmp53 = tmp52 * tmp3
tmp54 = tl.where(tmp51, tmp52, tmp53)
tmp55 = tl.where(tmp15, tmp54, tmp6)
tmp56 = triton_helpers.maximum(tmp50, tmp55)
tmp59 = tmp58 * tmp3
tmp60 = tl.where(tmp57, tmp58, tmp59)
tmp61 = tl.where(tmp22, tmp60, tmp6)
tmp62 = triton_helpers.maximum(tmp56, tmp61)
tmp63 = tmp44 - tmp62
tmp64 = tl_math.exp(tmp63)
tmp65 = tmp49 - tmp62
tmp66 = tl_math.exp(tmp65)
tmp67 = tmp64 + tmp66
tmp68 = tmp55 - tmp62
tmp69 = tl_math.exp(tmp68)
tmp70 = tmp67 + tmp69
tmp71 = tmp61 - tmp62
tmp72 = tl_math.exp(tmp71)
tmp73 = tmp70 + tmp72
tmp76 = tmp75 * tmp3
tmp77 = tl.where(tmp74, tmp75, tmp76)
tmp78 = tl.where(tmp0, tmp77, tmp6)
tmp81 = tmp80 * tmp3
tmp82 = tl.where(tmp79, tmp80, tmp81)
tmp83 = tl.where(tmp8, tmp82, tmp6)
tmp84 = triton_helpers.maximum(tmp78, tmp83)
tmp87 = tmp86 * tmp3
tmp88 = tl.where(tmp85, tmp86, tmp87)
tmp89 = tl.where(tmp15, tmp88, tmp6)
tmp90 = triton_helpers.maximum(tmp84, tmp89)
tmp93 = tmp92 * tmp3
tmp94 = tl.where(tmp91, tmp92, tmp93)
tmp95 = tl.where(tmp22, tmp94, tmp6)
tmp96 = triton_helpers.maximum(tmp90, tmp95)
tmp97 = tmp78 - tmp96
tmp98 = tl_math.exp(tmp97)
tmp99 = tmp83 - tmp96
tmp100 = tl_math.exp(tmp99)
tmp101 = tmp98 + tmp100
tmp102 = tmp89 - tmp96
tmp103 = tl_math.exp(tmp102)
tmp104 = tmp101 + tmp103
tmp105 = tmp95 - tmp96
tmp106 = tl_math.exp(tmp105)
tmp107 = tmp104 + tmp106
tmp110 = tmp109 * tmp3
tmp111 = tl.where(tmp108, tmp109, tmp110)
tmp112 = tl.where(tmp0, tmp111, tmp6)
tmp115 = tmp114 * tmp3
tmp116 = tl.where(tmp113, tmp114, tmp115)
tmp117 = tl.where(tmp8, tmp116, tmp6)
tmp118 = triton_helpers.maximum(tmp112, tmp117)
tmp121 = tmp120 * tmp3
tmp122 = tl.where(tmp119, tmp120, tmp121)
tmp123 = tl.where(tmp15, tmp122, tmp6)
tmp124 = triton_helpers.maximum(tmp118, tmp123)
tmp127 = tmp126 * tmp3
tmp128 = tl.where(tmp125, tmp126, tmp127)
tmp129 = tl.where(tmp22, tmp128, tmp6)
tmp130 = triton_helpers.maximum(tmp124, tmp129)
tmp131 = tmp112 - tmp130
tmp132 = tl_math.exp(tmp131)
tmp133 = tmp117 - tmp130
tmp134 = tl_math.exp(tmp133)
tmp135 = tmp132 + tmp134
tmp136 = tmp123 - tmp130
tmp137 = tl_math.exp(tmp136)
tmp138 = tmp135 + tmp137
tmp139 = tmp129 - tmp130
tmp140 = tl_math.exp(tmp139)
tmp141 = tmp138 + tmp140
tl.store(out_ptr0 + (x0), tmp28, xmask)
tl.store(out_ptr1 + (x0), tmp39, xmask)
tl.store(out_ptr2 + (x0), tmp62, xmask)
tl.store(out_ptr3 + (x0), tmp73, xmask)
tl.store(out_ptr4 + (x0), tmp96, xmask)
tl.store(out_ptr5 + (x0), tmp107, xmask)
tl.store(out_ptr6 + (x0), tmp130, xmask)
tl.store(out_ptr7 + (x0), tmp141, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/hw/chwxxijpaa43oudupbua5ibrakqm6zclctq55ppd6yvy4nqixzjb.py
# Topologically Sorted Source Nodes: [e, zero_vec, attention, attention_1, e_1, attention_3, attention_4, e_2, attention_6, attention_7, e_3, attention_9, attention_10], Original ATen: [aten.leaky_relu, aten.mul, aten.where, aten._softmax]
# Source node to ATen node mapping:
# attention => where_1
# attention_1 => div, exp, sub
# attention_10 => div_3, exp_3, sub_3
# attention_3 => where_4
# attention_4 => div_1, exp_1, sub_1
# attention_6 => where_7
# attention_7 => div_2, exp_2, sub_2
# attention_9 => where_10
# e => mul, where
# e_1 => mul_5, where_3
# e_2 => mul_10, where_6
# e_3 => mul_15, where_9
# zero_vec => full_default
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze, 4), kwargs = {})
# %where : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt, %squeeze, %mul), kwargs = {})
# %full_default : [num_users=5] = call_function[target=torch.ops.aten.full.default](args = ([4, 4], -8999999815811072.0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %where_1 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where, %full_default), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_1, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
# %div : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp, %sum_1), kwargs = {})
# %mul_5 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_1, 4), kwargs = {})
# %where_3 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_3, %squeeze_1, %mul_5), kwargs = {})
# %where_4 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_3, %full_default), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_4, %amax_1), kwargs = {})
# %exp_1 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_1,), kwargs = {})
# %div_1 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp_1, %sum_2), kwargs = {})
# %mul_10 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_2, 4), kwargs = {})
# %where_6 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_6, %squeeze_2, %mul_10), kwargs = {})
# %where_7 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_6, %full_default), kwargs = {})
# %sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_7, %amax_2), kwargs = {})
# %exp_2 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_2,), kwargs = {})
# %div_2 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp_2, %sum_3), kwargs = {})
# %mul_15 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_3, 4), kwargs = {})
# %where_9 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_9, %squeeze_3, %mul_15), kwargs = {})
# %where_10 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_9, %full_default), kwargs = {})
# %sub_3 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_10, %amax_3), kwargs = {})
# %exp_3 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_3,), kwargs = {})
# %div_3 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp_3, %sum_4), kwargs = {})
triton_poi_fused__softmax_leaky_relu_mul_where_3 = async_compile.triton('triton_poi_fused__softmax_leaky_relu_mul_where_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*i1', 5: '*i1', 6: '*fp32', 7: '*fp32', 8: '*i1', 9: '*fp32', 10: '*fp32', 11: '*i1', 12: '*fp32', 13: '*fp32', 14: '*i1', 15: '*fp32', 16: '*fp32', 17: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_leaky_relu_mul_where_3', 'mutated_arg_names': ['in_out_ptr0', 'in_out_ptr1', 'in_out_ptr2', 'in_out_ptr3'], 'no_x_dim': False, 'num_load': 17, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_3(in_out_ptr0, in_out_ptr1, in_out_ptr2, in_out_ptr3, in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, in_ptr6, in_ptr7, in_ptr8, in_ptr9, in_ptr10, in_ptr11, in_ptr12, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask).to(tl.int1)
tmp1 = tl.load(in_ptr1 + (x2), xmask).to(tl.int1)
tmp2 = tl.load(in_out_ptr0 + (x2), xmask)
tmp8 = tl.load(in_ptr2 + (x1), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr3 + (x1), xmask, eviction_policy='evict_last')
tmp13 = tl.load(in_ptr4 + (x2), xmask).to(tl.int1)
tmp14 = tl.load(in_out_ptr1 + (x2), xmask)
tmp18 = tl.load(in_ptr5 + (x1), xmask, eviction_policy='evict_last')
tmp21 = tl.load(in_ptr6 + (x1), xmask, eviction_policy='evict_last')
tmp23 = tl.load(in_ptr7 + (x2), xmask).to(tl.int1)
tmp24 = tl.load(in_out_ptr2 + (x2), xmask)
tmp28 = tl.load(in_ptr8 + (x1), xmask, eviction_policy='evict_last')
tmp31 = tl.load(in_ptr9 + (x1), xmask, eviction_policy='evict_last')
tmp33 = tl.load(in_ptr10 + (x2), xmask).to(tl.int1)
tmp34 = tl.load(in_out_ptr3 + (x2), xmask)
tmp38 = tl.load(in_ptr11 + (x1), xmask, eviction_policy='evict_last')
tmp41 = tl.load(in_ptr12 + (x1), xmask, eviction_policy='evict_last')
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp9 = tmp7 - tmp8
tmp10 = tl_math.exp(tmp9)
tmp12 = tmp10 / tmp11
tmp15 = tmp14 * tmp3
tmp16 = tl.where(tmp13, tmp14, tmp15)
tmp17 = tl.where(tmp0, tmp16, tmp6)
tmp19 = tmp17 - tmp18
tmp20 = tl_math.exp(tmp19)
tmp22 = tmp20 / tmp21
tmp25 = tmp24 * tmp3
tmp26 = tl.where(tmp23, tmp24, tmp25)
tmp27 = tl.where(tmp0, tmp26, tmp6)
tmp29 = tmp27 - tmp28
tmp30 = tl_math.exp(tmp29)
tmp32 = tmp30 / tmp31
tmp35 = tmp34 * tmp3
tmp36 = tl.where(tmp33, tmp34, tmp35)
tmp37 = tl.where(tmp0, tmp36, tmp6)
tmp39 = tmp37 - tmp38
tmp40 = tl_math.exp(tmp39)
tmp42 = tmp40 / tmp41
tl.store(in_out_ptr0 + (x2), tmp12, xmask)
tl.store(in_out_ptr1 + (x2), tmp22, xmask)
tl.store(in_out_ptr2 + (x2), tmp32, xmask)
tl.store(in_out_ptr3 + (x2), tmp42, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6f/c6fg755hkzgmiizoydcu7wlmcvduiztugqjkietqkvpoph4vrtad.py
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# x_1 => cat_4
# Graph fragment:
# %cat_4 : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%where_2, %where_5, %where_8, %where_11], 1), kwargs = {})
triton_poi_fused_cat_4 = async_compile.triton('triton_poi_fused_cat_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = (xindex // 16)
x2 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + ((4*x1) + x0), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = 0.0
tmp7 = tmp5 > tmp6
tmp8 = 1.0
tmp9 = tmp5 * tmp8
tmp10 = libdevice.expm1(tmp9)
tmp11 = tmp10 * tmp8
tmp12 = tl.where(tmp7, tmp9, tmp11)
tmp13 = tl.full(tmp12.shape, 0.0, tmp12.dtype)
tmp14 = tl.where(tmp4, tmp12, tmp13)
tmp15 = tmp0 >= tmp3
tmp16 = tl.full([1], 8, tl.int64)
tmp17 = tmp0 < tmp16
tmp18 = tmp15 & tmp17
tmp19 = tl.load(in_ptr1 + ((4*x1) + ((-4) + x0)), tmp18 & xmask, eviction_policy='evict_last', other=0.0)
tmp20 = tmp19 > tmp6
tmp21 = tmp19 * tmp8
tmp22 = libdevice.expm1(tmp21)
tmp23 = tmp22 * tmp8
tmp24 = tl.where(tmp20, tmp21, tmp23)
tmp25 = tl.full(tmp24.shape, 0.0, tmp24.dtype)
tmp26 = tl.where(tmp18, tmp24, tmp25)
tmp27 = tmp0 >= tmp16
tmp28 = tl.full([1], 12, tl.int64)
tmp29 = tmp0 < tmp28
tmp30 = tmp27 & tmp29
tmp31 = tl.load(in_ptr2 + ((4*x1) + ((-8) + x0)), tmp30 & xmask, eviction_policy='evict_last', other=0.0)
tmp32 = tmp31 > tmp6
tmp33 = tmp31 * tmp8
tmp34 = libdevice.expm1(tmp33)
tmp35 = tmp34 * tmp8
tmp36 = tl.where(tmp32, tmp33, tmp35)
tmp37 = tl.full(tmp36.shape, 0.0, tmp36.dtype)
tmp38 = tl.where(tmp30, tmp36, tmp37)
tmp39 = tmp0 >= tmp28
tmp40 = tl.full([1], 16, tl.int64)
tmp41 = tmp0 < tmp40
tmp42 = tl.load(in_ptr3 + ((4*x1) + ((-12) + x0)), tmp39 & xmask, eviction_policy='evict_last', other=0.0)
tmp43 = tmp42 > tmp6
tmp44 = tmp42 * tmp8
tmp45 = libdevice.expm1(tmp44)
tmp46 = tmp45 * tmp8
tmp47 = tl.where(tmp43, tmp44, tmp46)
tmp48 = tl.full(tmp47.shape, 0.0, tmp47.dtype)
tmp49 = tl.where(tmp39, tmp47, tmp48)
tmp50 = tl.where(tmp30, tmp38, tmp49)
tmp51 = tl.where(tmp18, tmp26, tmp50)
tmp52 = tl.where(tmp4, tmp14, tmp51)
tl.store(out_ptr0 + (x2), tmp52, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/6q/c6qm3bsuqnbgmfnlkkyiezkvruidr5w4kgvxblmhqrkljef5u2ab.py
# Topologically Sorted Source Nodes: [zero_vec, e_4, attention_12, attention_13], Original ATen: [aten.mul, aten.leaky_relu, aten.where, aten._softmax]
# Source node to ATen node mapping:
# attention_12 => where_13
# attention_13 => amax_4, exp_4, sub_4, sum_5
# e_4 => mul_20, where_12
# zero_vec => full_default
# Graph fragment:
# %full_default : [num_users=5] = call_function[target=torch.ops.aten.full.default](args = ([4, 4], -8999999815811072.0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %mul_20 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_4, 4), kwargs = {})
# %where_12 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_12, %squeeze_4, %mul_20), kwargs = {})
# %where_13 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_12, %full_default), kwargs = {})
# %amax_4 : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%where_13, [1], True), kwargs = {})
# %sub_4 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_13, %amax_4), kwargs = {})
# %exp_4 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_4,), kwargs = {})
# %sum_5 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_4, [1], True), kwargs = {})
triton_poi_fused__softmax_leaky_relu_mul_where_5 = async_compile.triton('triton_poi_fused__softmax_leaky_relu_mul_where_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4],
filename=__file__,
triton_meta={'signature': {0: '*i1', 1: '*i1', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_leaky_relu_mul_where_5', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 12, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_5(in_ptr0, in_ptr1, in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (4*x0), xmask, eviction_policy='evict_last').to(tl.int1)
tmp1 = tl.load(in_ptr1 + (4*x0), xmask, eviction_policy='evict_last').to(tl.int1)
tmp2 = tl.load(in_ptr2 + (4*x0), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (1 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp9 = tl.load(in_ptr1 + (1 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp10 = tl.load(in_ptr2 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp15 = tl.load(in_ptr0 + (2 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp16 = tl.load(in_ptr1 + (2 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp17 = tl.load(in_ptr2 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp22 = tl.load(in_ptr0 + (3 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp23 = tl.load(in_ptr1 + (3 + (4*x0)), xmask, eviction_policy='evict_last').to(tl.int1)
tmp24 = tl.load(in_ptr2 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp11 = tmp10 * tmp3
tmp12 = tl.where(tmp9, tmp10, tmp11)
tmp13 = tl.where(tmp8, tmp12, tmp6)
tmp14 = triton_helpers.maximum(tmp7, tmp13)
tmp18 = tmp17 * tmp3
tmp19 = tl.where(tmp16, tmp17, tmp18)
tmp20 = tl.where(tmp15, tmp19, tmp6)
tmp21 = triton_helpers.maximum(tmp14, tmp20)
tmp25 = tmp24 * tmp3
tmp26 = tl.where(tmp23, tmp24, tmp25)
tmp27 = tl.where(tmp22, tmp26, tmp6)
tmp28 = triton_helpers.maximum(tmp21, tmp27)
tmp29 = tmp7 - tmp28
tmp30 = tl_math.exp(tmp29)
tmp31 = tmp13 - tmp28
tmp32 = tl_math.exp(tmp31)
tmp33 = tmp30 + tmp32
tmp34 = tmp20 - tmp28
tmp35 = tl_math.exp(tmp34)
tmp36 = tmp33 + tmp35
tmp37 = tmp27 - tmp28
tmp38 = tl_math.exp(tmp37)
tmp39 = tmp36 + tmp38
tl.store(out_ptr0 + (x0), tmp28, xmask)
tl.store(out_ptr1 + (x0), tmp39, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/4f/c4fpasprzjcely6grqxmcycpm24taaowwba6rqvchhupdivgc3gx.py
# Topologically Sorted Source Nodes: [zero_vec, e_4, attention_12, attention_13], Original ATen: [aten.mul, aten.leaky_relu, aten.where, aten._softmax]
# Source node to ATen node mapping:
# attention_12 => where_13
# attention_13 => div_4, exp_4, sub_4
# e_4 => mul_20, where_12
# zero_vec => full_default
# Graph fragment:
# %full_default : [num_users=5] = call_function[target=torch.ops.aten.full.default](args = ([4, 4], -8999999815811072.0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %mul_20 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze_4, 4), kwargs = {})
# %where_12 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_12, %squeeze_4, %mul_20), kwargs = {})
# %where_13 : [num_users=2] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %where_12, %full_default), kwargs = {})
# %sub_4 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%where_13, %amax_4), kwargs = {})
# %exp_4 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub_4,), kwargs = {})
# %div_4 : [num_users=2] = call_function[target=torch.ops.aten.div.Tensor](args = (%exp_4, %sum_5), kwargs = {})
triton_poi_fused__softmax_leaky_relu_mul_where_6 = async_compile.triton('triton_poi_fused__softmax_leaky_relu_mul_where_6', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i1', 2: '*i1', 3: '*fp32', 4: '*fp32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_leaky_relu_mul_where_6', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_6(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask).to(tl.int1)
tmp1 = tl.load(in_ptr1 + (x2), xmask).to(tl.int1)
tmp2 = tl.load(in_out_ptr0 + (x2), xmask)
tmp8 = tl.load(in_ptr2 + (x1), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr3 + (x1), xmask, eviction_policy='evict_last')
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp9 = tmp7 - tmp8
tmp10 = tl_math.exp(tmp9)
tmp12 = tmp10 / tmp11
tl.store(in_out_ptr0 + (x2), tmp12, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/k3/ck37q54iqhyqxtcggc5k6r5dr76gmt6kxkyslk6wocaqdn3zqz2z.py
# Topologically Sorted Source Nodes: [relu], Original ATen: [aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%mm_14,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_relu_threshold_backward_7 = async_compile.triton('triton_poi_fused_relu_threshold_backward_7', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i1', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_threshold_backward_7', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_threshold_backward_7(in_out_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + (x0), xmask)
tmp1 = tl.full([1], 0, tl.int32)
tmp2 = triton_helpers.maximum(tmp1, tmp0)
tmp3 = 0.0
tmp4 = tmp2 <= tmp3
tl.store(in_out_ptr0 + (x0), tmp2, xmask)
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (8, 1), (1, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, 4), (4, 1))
assert_size_stride(primals_6, (8, 1), (1, 1))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (8, 1), (1, 1))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (8, 1), (1, 1))
assert_size_stride(primals_11, (16, 4), (4, 1))
assert_size_stride(primals_12, (8, 1), (1, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h], Original ATen: [aten.mm]
extern_kernels.mm(primals_1, primals_2, out=buf0)
del primals_2
buf1 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(buf0, buf1, 128, grid=grid(128), stream=stream0)
buf2 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [matmul], Original ATen: [aten.mm]
extern_kernels.mm(buf1, primals_3, out=buf2)
buf3 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
# Topologically Sorted Source Nodes: [e], Original ATen: [aten.leaky_relu]
triton_poi_fused_leaky_relu_1.run(buf2, buf3, 16, grid=grid(16), stream=stream0)
buf4 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
# Topologically Sorted Source Nodes: [gt], Original ATen: [aten.gt]
triton_poi_fused_leaky_relu_1.run(primals_4, buf4, 16, grid=grid(16), stream=stream0)
del primals_4
buf9 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_1], Original ATen: [aten.mm]
extern_kernels.mm(primals_1, primals_5, out=buf9)
del primals_5
buf10 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat_1], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf9, buf10, 128, grid=grid(128), stream=stream0)
buf11 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [matmul_2], Original ATen: [aten.mm]
extern_kernels.mm(buf10, primals_6, out=buf11)
buf12 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
# Topologically Sorted Source Nodes: [e_1], Original ATen: [aten.leaky_relu]
triton_poi_fused_leaky_relu_1.run(buf11, buf12, 16, grid=grid(16), stream=stream0)
buf17 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_2], Original ATen: [aten.mm]
extern_kernels.mm(primals_1, primals_7, out=buf17)
del primals_7
buf18 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat_2], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf17, buf18, 128, grid=grid(128), stream=stream0)
buf19 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [matmul_4], Original ATen: [aten.mm]
extern_kernels.mm(buf18, primals_8, out=buf19)
buf20 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
# Topologically Sorted Source Nodes: [e_2], Original ATen: [aten.leaky_relu]
triton_poi_fused_leaky_relu_1.run(buf19, buf20, 16, grid=grid(16), stream=stream0)
buf25 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_3], Original ATen: [aten.mm]
extern_kernels.mm(primals_1, primals_9, out=buf25)
del primals_9
buf26 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat_3], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf25, buf26, 128, grid=grid(128), stream=stream0)
buf27 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [matmul_6], Original ATen: [aten.mm]
extern_kernels.mm(buf26, primals_10, out=buf27)
buf28 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
# Topologically Sorted Source Nodes: [e_3], Original ATen: [aten.leaky_relu]
triton_poi_fused_leaky_relu_1.run(buf27, buf28, 16, grid=grid(16), stream=stream0)
buf5 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf6 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf13 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf14 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf21 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf22 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf29 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf30 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
# Topologically Sorted Source Nodes: [e, zero_vec, attention, attention_1, e_1, attention_3, attention_4, e_2, attention_6, attention_7, e_3, attention_9, attention_10], Original ATen: [aten.leaky_relu, aten.mul, aten.where, aten._softmax]
triton_poi_fused__softmax_leaky_relu_mul_where_2.run(buf4, buf3, buf2, buf12, buf11, buf20, buf19, buf28, buf27, buf5, buf6, buf13, buf14, buf21, buf22, buf29, buf30, 4, grid=grid(4), stream=stream0)
buf7 = reinterpret_tensor(buf2, (4, 4), (4, 1), 0); del buf2 # reuse
buf15 = reinterpret_tensor(buf11, (4, 4), (4, 1), 0); del buf11 # reuse
buf23 = reinterpret_tensor(buf19, (4, 4), (4, 1), 0); del buf19 # reuse
buf31 = reinterpret_tensor(buf27, (4, 4), (4, 1), 0); del buf27 # reuse
# Topologically Sorted Source Nodes: [e, zero_vec, attention, attention_1, e_1, attention_3, attention_4, e_2, attention_6, attention_7, e_3, attention_9, attention_10], Original ATen: [aten.leaky_relu, aten.mul, aten.where, aten._softmax]
triton_poi_fused__softmax_leaky_relu_mul_where_3.run(buf7, buf15, buf23, buf31, buf4, buf3, buf5, buf6, buf12, buf13, buf14, buf20, buf21, buf22, buf28, buf29, buf30, 16, grid=grid(16), stream=stream0)
del buf13
del buf14
del buf21
del buf22
del buf29
del buf30
buf8 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_prime], Original ATen: [aten.mm]
extern_kernels.mm(buf7, buf0, out=buf8)
buf16 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_prime_1], Original ATen: [aten.mm]
extern_kernels.mm(buf15, buf9, out=buf16)
buf24 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_prime_2], Original ATen: [aten.mm]
extern_kernels.mm(buf23, buf17, out=buf24)
buf32 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_prime_3], Original ATen: [aten.mm]
extern_kernels.mm(buf31, buf25, out=buf32)
buf33 = empty_strided_cuda((4, 16), (16, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.cat]
triton_poi_fused_cat_4.run(buf8, buf16, buf24, buf32, buf33, 64, grid=grid(64), stream=stream0)
buf34 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_4], Original ATen: [aten.mm]
extern_kernels.mm(buf33, primals_11, out=buf34)
buf35 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat_5], Original ATen: [aten.cat]
triton_poi_fused_cat_0.run(buf34, buf35, 128, grid=grid(128), stream=stream0)
buf36 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
# Topologically Sorted Source Nodes: [matmul_8], Original ATen: [aten.mm]
extern_kernels.mm(buf35, primals_12, out=buf36)
buf37 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
# Topologically Sorted Source Nodes: [e_4], Original ATen: [aten.leaky_relu]
triton_poi_fused_leaky_relu_1.run(buf36, buf37, 16, grid=grid(16), stream=stream0)
buf38 = buf6; del buf6 # reuse
buf39 = buf5; del buf5 # reuse
# Topologically Sorted Source Nodes: [zero_vec, e_4, attention_12, attention_13], Original ATen: [aten.mul, aten.leaky_relu, aten.where, aten._softmax]
triton_poi_fused__softmax_leaky_relu_mul_where_5.run(buf4, buf37, buf36, buf38, buf39, 4, grid=grid(4), stream=stream0)
buf40 = reinterpret_tensor(buf36, (4, 4), (4, 1), 0); del buf36 # reuse
# Topologically Sorted Source Nodes: [zero_vec, e_4, attention_12, attention_13], Original ATen: [aten.mul, aten.leaky_relu, aten.where, aten._softmax]
triton_poi_fused__softmax_leaky_relu_mul_where_6.run(buf40, buf4, buf37, buf38, buf39, 16, grid=grid(16), stream=stream0)
del buf38
del buf39
buf41 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [h_prime_4], Original ATen: [aten.mm]
extern_kernels.mm(buf40, buf34, out=buf41)
buf42 = buf41; del buf41 # reuse
buf43 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu], Original ATen: [aten.relu, aten.threshold_backward]
triton_poi_fused_relu_threshold_backward_7.run(buf42, buf43, 16, grid=grid(16), stream=stream0)
return (buf42, buf3, buf4, buf7, buf8, buf12, buf15, buf16, buf20, buf23, buf24, buf28, buf31, buf32, buf37, buf40, buf43, reinterpret_tensor(buf34, (4, 4), (1, 4), 0), reinterpret_tensor(buf35, (8, 16), (1, 8), 0), reinterpret_tensor(primals_12, (1, 8), (1, 1), 0), reinterpret_tensor(buf33, (16, 4), (1, 16), 0), reinterpret_tensor(primals_11, (4, 16), (1, 4), 0), reinterpret_tensor(buf25, (4, 4), (1, 4), 0), reinterpret_tensor(buf26, (8, 16), (1, 8), 0), reinterpret_tensor(primals_10, (1, 8), (1, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0), reinterpret_tensor(buf17, (4, 4), (1, 4), 0), reinterpret_tensor(buf18, (8, 16), (1, 8), 0), reinterpret_tensor(primals_8, (1, 8), (1, 1), 0), reinterpret_tensor(buf9, (4, 4), (1, 4), 0), reinterpret_tensor(buf10, (8, 16), (1, 8), 0), reinterpret_tensor(primals_6, (1, 8), (1, 1), 0), reinterpret_tensor(buf0, (4, 4), (1, 4), 0), reinterpret_tensor(buf1, (8, 16), (1, 8), 0), reinterpret_tensor(primals_3, (1, 8), (1, 1), 0), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((8, 1), (1, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((8, 1), (1, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((8, 1), (1, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((8, 1), (1, 1), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((16, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_12 = rand_strided((8, 1), (1, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
class GraphAttentionLayer(nn.Module):
"""
Simple GAT layer, similar to https://arxiv.org/abs/1710.10903
"""
def __init__(self, in_features, out_features, dropout, alpha, concat=True):
super(GraphAttentionLayer, self).__init__()
self.dropout = dropout
self.in_features = in_features
self.out_features = out_features
self.alpha = alpha
self.concat = concat
self.W = nn.Parameter(torch.zeros(size=(in_features, out_features)))
nn.init.xavier_uniform_(self.W.data, gain=1.414)
self.a = nn.Parameter(torch.zeros(size=(2 * out_features, 1)))
nn.init.xavier_uniform_(self.a.data, gain=1.414)
self.leakyrelu = nn.LeakyReLU(self.alpha)
def forward(self, input, adj):
h = torch.mm(input, self.W)
N = h.size()[0]
a_input = torch.cat([h.repeat(1, N).view(N * N, -1), h.repeat(N, 1)
], dim=1).view(N, -1, 2 * self.out_features)
e = self.leakyrelu(torch.matmul(a_input, self.a).squeeze(2))
zero_vec = -9000000000000000.0 * torch.ones_like(e)
attention = torch.where(adj > 0, e, zero_vec)
attention = F.softmax(attention, dim=1)
attention = F.dropout(attention, self.dropout, training=self.training)
h_prime = torch.matmul(attention, h)
if self.concat:
return F.elu(h_prime)
else:
return h_prime
def __repr__(self):
return self.__class__.__name__ + ' (' + str(self.in_features
) + ' -> ' + str(self.out_features) + ')'
class GAT(nn.Module):
def __init__(self, nfeat, nhid, nclass, dropout, alpha, nheads):
"""Dense version of GAT."""
super(GAT, self).__init__()
self.dropout = dropout
self.attentions = [GraphAttentionLayer(nfeat, nhid, dropout=dropout,
alpha=alpha, concat=True) for _ in range(nheads)]
for i, attention in enumerate(self.attentions):
self.add_module('attention_{}'.format(i), attention)
self.out_att = GraphAttentionLayer(nhid * nheads, nclass, dropout=
dropout, alpha=alpha, concat=False)
def forward(self, x, adj):
x = F.dropout(x, self.dropout, training=self.training)
x = torch.cat([att(x, adj) for att in self.attentions], dim=1)
x = F.dropout(x, self.dropout, training=self.training)
return F.relu(self.out_att(x, adj))
def get_inputs():
return [torch.rand([4, 4]), torch.rand([4, 4])]
def get_init_inputs():
return [[], {'nfeat': 4, 'nhid': 4, 'nclass': 4, 'dropout': 0.5,
'alpha': 4, 'nheads': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
import torch.nn.functional as F
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 8
x1 = xindex // 8
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * ((4 * x1 + x0) // 16 % 4) + (4 * x1 + x0) %
16 % 4), tmp4 & xmask, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr0 + (4 * (x1 % 4) + (-4 + x0)), tmp6 & xmask,
eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x2, tmp10, xmask)
@triton.jit
def triton_poi_fused_leaky_relu_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tl.store(out_ptr0 + x0, tmp2, xmask)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_2(in_ptr0, in_ptr1,
in_ptr2, in_ptr3, in_ptr4, in_ptr5, in_ptr6, in_ptr7, in_ptr8, out_ptr0,
out_ptr1, out_ptr2, out_ptr3, out_ptr4, out_ptr5, out_ptr6, out_ptr7,
xnumel, XBLOCK: tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 4 * x0, xmask, eviction_policy='evict_last').to(tl
.int1)
tmp1 = tl.load(in_ptr1 + 4 * x0, xmask, eviction_policy='evict_last').to(tl
.int1)
tmp2 = tl.load(in_ptr2 + 4 * x0, xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp9 = tl.load(in_ptr1 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp10 = tl.load(in_ptr2 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp15 = tl.load(in_ptr0 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp16 = tl.load(in_ptr1 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp17 = tl.load(in_ptr2 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp22 = tl.load(in_ptr0 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp23 = tl.load(in_ptr1 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp24 = tl.load(in_ptr2 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp40 = tl.load(in_ptr3 + 4 * x0, xmask, eviction_policy='evict_last').to(
tl.int1)
tmp41 = tl.load(in_ptr4 + 4 * x0, xmask, eviction_policy='evict_last')
tmp45 = tl.load(in_ptr3 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp46 = tl.load(in_ptr4 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp51 = tl.load(in_ptr3 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp52 = tl.load(in_ptr4 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp57 = tl.load(in_ptr3 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp58 = tl.load(in_ptr4 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp74 = tl.load(in_ptr5 + 4 * x0, xmask, eviction_policy='evict_last').to(
tl.int1)
tmp75 = tl.load(in_ptr6 + 4 * x0, xmask, eviction_policy='evict_last')
tmp79 = tl.load(in_ptr5 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp80 = tl.load(in_ptr6 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp85 = tl.load(in_ptr5 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp86 = tl.load(in_ptr6 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp91 = tl.load(in_ptr5 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp92 = tl.load(in_ptr6 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp108 = tl.load(in_ptr7 + 4 * x0, xmask, eviction_policy='evict_last').to(
tl.int1)
tmp109 = tl.load(in_ptr8 + 4 * x0, xmask, eviction_policy='evict_last')
tmp113 = tl.load(in_ptr7 + (1 + 4 * x0), xmask, eviction_policy=
'evict_last').to(tl.int1)
tmp114 = tl.load(in_ptr8 + (1 + 4 * x0), xmask, eviction_policy=
'evict_last')
tmp119 = tl.load(in_ptr7 + (2 + 4 * x0), xmask, eviction_policy=
'evict_last').to(tl.int1)
tmp120 = tl.load(in_ptr8 + (2 + 4 * x0), xmask, eviction_policy=
'evict_last')
tmp125 = tl.load(in_ptr7 + (3 + 4 * x0), xmask, eviction_policy=
'evict_last').to(tl.int1)
tmp126 = tl.load(in_ptr8 + (3 + 4 * x0), xmask, eviction_policy=
'evict_last')
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp11 = tmp10 * tmp3
tmp12 = tl.where(tmp9, tmp10, tmp11)
tmp13 = tl.where(tmp8, tmp12, tmp6)
tmp14 = triton_helpers.maximum(tmp7, tmp13)
tmp18 = tmp17 * tmp3
tmp19 = tl.where(tmp16, tmp17, tmp18)
tmp20 = tl.where(tmp15, tmp19, tmp6)
tmp21 = triton_helpers.maximum(tmp14, tmp20)
tmp25 = tmp24 * tmp3
tmp26 = tl.where(tmp23, tmp24, tmp25)
tmp27 = tl.where(tmp22, tmp26, tmp6)
tmp28 = triton_helpers.maximum(tmp21, tmp27)
tmp29 = tmp7 - tmp28
tmp30 = tl_math.exp(tmp29)
tmp31 = tmp13 - tmp28
tmp32 = tl_math.exp(tmp31)
tmp33 = tmp30 + tmp32
tmp34 = tmp20 - tmp28
tmp35 = tl_math.exp(tmp34)
tmp36 = tmp33 + tmp35
tmp37 = tmp27 - tmp28
tmp38 = tl_math.exp(tmp37)
tmp39 = tmp36 + tmp38
tmp42 = tmp41 * tmp3
tmp43 = tl.where(tmp40, tmp41, tmp42)
tmp44 = tl.where(tmp0, tmp43, tmp6)
tmp47 = tmp46 * tmp3
tmp48 = tl.where(tmp45, tmp46, tmp47)
tmp49 = tl.where(tmp8, tmp48, tmp6)
tmp50 = triton_helpers.maximum(tmp44, tmp49)
tmp53 = tmp52 * tmp3
tmp54 = tl.where(tmp51, tmp52, tmp53)
tmp55 = tl.where(tmp15, tmp54, tmp6)
tmp56 = triton_helpers.maximum(tmp50, tmp55)
tmp59 = tmp58 * tmp3
tmp60 = tl.where(tmp57, tmp58, tmp59)
tmp61 = tl.where(tmp22, tmp60, tmp6)
tmp62 = triton_helpers.maximum(tmp56, tmp61)
tmp63 = tmp44 - tmp62
tmp64 = tl_math.exp(tmp63)
tmp65 = tmp49 - tmp62
tmp66 = tl_math.exp(tmp65)
tmp67 = tmp64 + tmp66
tmp68 = tmp55 - tmp62
tmp69 = tl_math.exp(tmp68)
tmp70 = tmp67 + tmp69
tmp71 = tmp61 - tmp62
tmp72 = tl_math.exp(tmp71)
tmp73 = tmp70 + tmp72
tmp76 = tmp75 * tmp3
tmp77 = tl.where(tmp74, tmp75, tmp76)
tmp78 = tl.where(tmp0, tmp77, tmp6)
tmp81 = tmp80 * tmp3
tmp82 = tl.where(tmp79, tmp80, tmp81)
tmp83 = tl.where(tmp8, tmp82, tmp6)
tmp84 = triton_helpers.maximum(tmp78, tmp83)
tmp87 = tmp86 * tmp3
tmp88 = tl.where(tmp85, tmp86, tmp87)
tmp89 = tl.where(tmp15, tmp88, tmp6)
tmp90 = triton_helpers.maximum(tmp84, tmp89)
tmp93 = tmp92 * tmp3
tmp94 = tl.where(tmp91, tmp92, tmp93)
tmp95 = tl.where(tmp22, tmp94, tmp6)
tmp96 = triton_helpers.maximum(tmp90, tmp95)
tmp97 = tmp78 - tmp96
tmp98 = tl_math.exp(tmp97)
tmp99 = tmp83 - tmp96
tmp100 = tl_math.exp(tmp99)
tmp101 = tmp98 + tmp100
tmp102 = tmp89 - tmp96
tmp103 = tl_math.exp(tmp102)
tmp104 = tmp101 + tmp103
tmp105 = tmp95 - tmp96
tmp106 = tl_math.exp(tmp105)
tmp107 = tmp104 + tmp106
tmp110 = tmp109 * tmp3
tmp111 = tl.where(tmp108, tmp109, tmp110)
tmp112 = tl.where(tmp0, tmp111, tmp6)
tmp115 = tmp114 * tmp3
tmp116 = tl.where(tmp113, tmp114, tmp115)
tmp117 = tl.where(tmp8, tmp116, tmp6)
tmp118 = triton_helpers.maximum(tmp112, tmp117)
tmp121 = tmp120 * tmp3
tmp122 = tl.where(tmp119, tmp120, tmp121)
tmp123 = tl.where(tmp15, tmp122, tmp6)
tmp124 = triton_helpers.maximum(tmp118, tmp123)
tmp127 = tmp126 * tmp3
tmp128 = tl.where(tmp125, tmp126, tmp127)
tmp129 = tl.where(tmp22, tmp128, tmp6)
tmp130 = triton_helpers.maximum(tmp124, tmp129)
tmp131 = tmp112 - tmp130
tmp132 = tl_math.exp(tmp131)
tmp133 = tmp117 - tmp130
tmp134 = tl_math.exp(tmp133)
tmp135 = tmp132 + tmp134
tmp136 = tmp123 - tmp130
tmp137 = tl_math.exp(tmp136)
tmp138 = tmp135 + tmp137
tmp139 = tmp129 - tmp130
tmp140 = tl_math.exp(tmp139)
tmp141 = tmp138 + tmp140
tl.store(out_ptr0 + x0, tmp28, xmask)
tl.store(out_ptr1 + x0, tmp39, xmask)
tl.store(out_ptr2 + x0, tmp62, xmask)
tl.store(out_ptr3 + x0, tmp73, xmask)
tl.store(out_ptr4 + x0, tmp96, xmask)
tl.store(out_ptr5 + x0, tmp107, xmask)
tl.store(out_ptr6 + x0, tmp130, xmask)
tl.store(out_ptr7 + x0, tmp141, xmask)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_3(in_out_ptr0,
in_out_ptr1, in_out_ptr2, in_out_ptr3, in_ptr0, in_ptr1, in_ptr2,
in_ptr3, in_ptr4, in_ptr5, in_ptr6, in_ptr7, in_ptr8, in_ptr9, in_ptr10,
in_ptr11, in_ptr12, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask).to(tl.int1)
tmp1 = tl.load(in_ptr1 + x2, xmask).to(tl.int1)
tmp2 = tl.load(in_out_ptr0 + x2, xmask)
tmp8 = tl.load(in_ptr2 + x1, xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr3 + x1, xmask, eviction_policy='evict_last')
tmp13 = tl.load(in_ptr4 + x2, xmask).to(tl.int1)
tmp14 = tl.load(in_out_ptr1 + x2, xmask)
tmp18 = tl.load(in_ptr5 + x1, xmask, eviction_policy='evict_last')
tmp21 = tl.load(in_ptr6 + x1, xmask, eviction_policy='evict_last')
tmp23 = tl.load(in_ptr7 + x2, xmask).to(tl.int1)
tmp24 = tl.load(in_out_ptr2 + x2, xmask)
tmp28 = tl.load(in_ptr8 + x1, xmask, eviction_policy='evict_last')
tmp31 = tl.load(in_ptr9 + x1, xmask, eviction_policy='evict_last')
tmp33 = tl.load(in_ptr10 + x2, xmask).to(tl.int1)
tmp34 = tl.load(in_out_ptr3 + x2, xmask)
tmp38 = tl.load(in_ptr11 + x1, xmask, eviction_policy='evict_last')
tmp41 = tl.load(in_ptr12 + x1, xmask, eviction_policy='evict_last')
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp9 = tmp7 - tmp8
tmp10 = tl_math.exp(tmp9)
tmp12 = tmp10 / tmp11
tmp15 = tmp14 * tmp3
tmp16 = tl.where(tmp13, tmp14, tmp15)
tmp17 = tl.where(tmp0, tmp16, tmp6)
tmp19 = tmp17 - tmp18
tmp20 = tl_math.exp(tmp19)
tmp22 = tmp20 / tmp21
tmp25 = tmp24 * tmp3
tmp26 = tl.where(tmp23, tmp24, tmp25)
tmp27 = tl.where(tmp0, tmp26, tmp6)
tmp29 = tmp27 - tmp28
tmp30 = tl_math.exp(tmp29)
tmp32 = tmp30 / tmp31
tmp35 = tmp34 * tmp3
tmp36 = tl.where(tmp33, tmp34, tmp35)
tmp37 = tl.where(tmp0, tmp36, tmp6)
tmp39 = tmp37 - tmp38
tmp40 = tl_math.exp(tmp39)
tmp42 = tmp40 / tmp41
tl.store(in_out_ptr0 + x2, tmp12, xmask)
tl.store(in_out_ptr1 + x2, tmp22, xmask)
tl.store(in_out_ptr2 + x2, tmp32, xmask)
tl.store(in_out_ptr3 + x2, tmp42, xmask)
@triton.jit
def triton_poi_fused_cat_4(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = xindex // 16
x2 = xindex
tmp0 = x0
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (4 * x1 + x0), tmp4 & xmask, eviction_policy=
'evict_last', other=0.0)
tmp6 = 0.0
tmp7 = tmp5 > tmp6
tmp8 = 1.0
tmp9 = tmp5 * tmp8
tmp10 = libdevice.expm1(tmp9)
tmp11 = tmp10 * tmp8
tmp12 = tl.where(tmp7, tmp9, tmp11)
tmp13 = tl.full(tmp12.shape, 0.0, tmp12.dtype)
tmp14 = tl.where(tmp4, tmp12, tmp13)
tmp15 = tmp0 >= tmp3
tmp16 = tl.full([1], 8, tl.int64)
tmp17 = tmp0 < tmp16
tmp18 = tmp15 & tmp17
tmp19 = tl.load(in_ptr1 + (4 * x1 + (-4 + x0)), tmp18 & xmask,
eviction_policy='evict_last', other=0.0)
tmp20 = tmp19 > tmp6
tmp21 = tmp19 * tmp8
tmp22 = libdevice.expm1(tmp21)
tmp23 = tmp22 * tmp8
tmp24 = tl.where(tmp20, tmp21, tmp23)
tmp25 = tl.full(tmp24.shape, 0.0, tmp24.dtype)
tmp26 = tl.where(tmp18, tmp24, tmp25)
tmp27 = tmp0 >= tmp16
tmp28 = tl.full([1], 12, tl.int64)
tmp29 = tmp0 < tmp28
tmp30 = tmp27 & tmp29
tmp31 = tl.load(in_ptr2 + (4 * x1 + (-8 + x0)), tmp30 & xmask,
eviction_policy='evict_last', other=0.0)
tmp32 = tmp31 > tmp6
tmp33 = tmp31 * tmp8
tmp34 = libdevice.expm1(tmp33)
tmp35 = tmp34 * tmp8
tmp36 = tl.where(tmp32, tmp33, tmp35)
tmp37 = tl.full(tmp36.shape, 0.0, tmp36.dtype)
tmp38 = tl.where(tmp30, tmp36, tmp37)
tmp39 = tmp0 >= tmp28
tl.full([1], 16, tl.int64)
tmp42 = tl.load(in_ptr3 + (4 * x1 + (-12 + x0)), tmp39 & xmask,
eviction_policy='evict_last', other=0.0)
tmp43 = tmp42 > tmp6
tmp44 = tmp42 * tmp8
tmp45 = libdevice.expm1(tmp44)
tmp46 = tmp45 * tmp8
tmp47 = tl.where(tmp43, tmp44, tmp46)
tmp48 = tl.full(tmp47.shape, 0.0, tmp47.dtype)
tmp49 = tl.where(tmp39, tmp47, tmp48)
tmp50 = tl.where(tmp30, tmp38, tmp49)
tmp51 = tl.where(tmp18, tmp26, tmp50)
tmp52 = tl.where(tmp4, tmp14, tmp51)
tl.store(out_ptr0 + x2, tmp52, xmask)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_5(in_ptr0, in_ptr1,
in_ptr2, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 4 * x0, xmask, eviction_policy='evict_last').to(tl
.int1)
tmp1 = tl.load(in_ptr1 + 4 * x0, xmask, eviction_policy='evict_last').to(tl
.int1)
tmp2 = tl.load(in_ptr2 + 4 * x0, xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp9 = tl.load(in_ptr1 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp10 = tl.load(in_ptr2 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp15 = tl.load(in_ptr0 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp16 = tl.load(in_ptr1 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp17 = tl.load(in_ptr2 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp22 = tl.load(in_ptr0 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp23 = tl.load(in_ptr1 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
).to(tl.int1)
tmp24 = tl.load(in_ptr2 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp11 = tmp10 * tmp3
tmp12 = tl.where(tmp9, tmp10, tmp11)
tmp13 = tl.where(tmp8, tmp12, tmp6)
tmp14 = triton_helpers.maximum(tmp7, tmp13)
tmp18 = tmp17 * tmp3
tmp19 = tl.where(tmp16, tmp17, tmp18)
tmp20 = tl.where(tmp15, tmp19, tmp6)
tmp21 = triton_helpers.maximum(tmp14, tmp20)
tmp25 = tmp24 * tmp3
tmp26 = tl.where(tmp23, tmp24, tmp25)
tmp27 = tl.where(tmp22, tmp26, tmp6)
tmp28 = triton_helpers.maximum(tmp21, tmp27)
tmp29 = tmp7 - tmp28
tmp30 = tl_math.exp(tmp29)
tmp31 = tmp13 - tmp28
tmp32 = tl_math.exp(tmp31)
tmp33 = tmp30 + tmp32
tmp34 = tmp20 - tmp28
tmp35 = tl_math.exp(tmp34)
tmp36 = tmp33 + tmp35
tmp37 = tmp27 - tmp28
tmp38 = tl_math.exp(tmp37)
tmp39 = tmp36 + tmp38
tl.store(out_ptr0 + x0, tmp28, xmask)
tl.store(out_ptr1 + x0, tmp39, xmask)
@triton.jit
def triton_poi_fused__softmax_leaky_relu_mul_where_6(in_out_ptr0, in_ptr0,
in_ptr1, in_ptr2, in_ptr3, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask).to(tl.int1)
tmp1 = tl.load(in_ptr1 + x2, xmask).to(tl.int1)
tmp2 = tl.load(in_out_ptr0 + x2, xmask)
tmp8 = tl.load(in_ptr2 + x1, xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr3 + x1, xmask, eviction_policy='evict_last')
tmp3 = 4.0
tmp4 = tmp2 * tmp3
tmp5 = tl.where(tmp1, tmp2, tmp4)
tmp6 = -8999999815811072.0
tmp7 = tl.where(tmp0, tmp5, tmp6)
tmp9 = tmp7 - tmp8
tmp10 = tl_math.exp(tmp9)
tmp12 = tmp10 / tmp11
tl.store(in_out_ptr0 + x2, tmp12, xmask)
@triton.jit
def triton_poi_fused_relu_threshold_backward_7(in_out_ptr0, out_ptr0,
xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + x0, xmask)
tmp1 = tl.full([1], 0, tl.int32)
tmp2 = triton_helpers.maximum(tmp1, tmp0)
tmp3 = 0.0
tmp4 = tmp2 <= tmp3
tl.store(in_out_ptr0 + x0, tmp2, xmask)
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11, primals_12
) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (4, 4), (4, 1))
assert_size_stride(primals_3, (8, 1), (1, 1))
assert_size_stride(primals_4, (4, 4), (4, 1))
assert_size_stride(primals_5, (4, 4), (4, 1))
assert_size_stride(primals_6, (8, 1), (1, 1))
assert_size_stride(primals_7, (4, 4), (4, 1))
assert_size_stride(primals_8, (8, 1), (1, 1))
assert_size_stride(primals_9, (4, 4), (4, 1))
assert_size_stride(primals_10, (8, 1), (1, 1))
assert_size_stride(primals_11, (16, 4), (4, 1))
assert_size_stride(primals_12, (8, 1), (1, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(primals_1, primals_2, out=buf0)
del primals_2
buf1 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(128)](buf0, buf1, 128, XBLOCK=128,
num_warps=4, num_stages=1)
buf2 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
extern_kernels.mm(buf1, primals_3, out=buf2)
buf3 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
triton_poi_fused_leaky_relu_1[grid(16)](buf2, buf3, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf4 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
triton_poi_fused_leaky_relu_1[grid(16)](primals_4, buf4, 16, XBLOCK
=16, num_warps=1, num_stages=1)
del primals_4
buf9 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(primals_1, primals_5, out=buf9)
del primals_5
buf10 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
triton_poi_fused_cat_0[grid(128)](buf9, buf10, 128, XBLOCK=128,
num_warps=4, num_stages=1)
buf11 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
extern_kernels.mm(buf10, primals_6, out=buf11)
buf12 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
triton_poi_fused_leaky_relu_1[grid(16)](buf11, buf12, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf17 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(primals_1, primals_7, out=buf17)
del primals_7
buf18 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
triton_poi_fused_cat_0[grid(128)](buf17, buf18, 128, XBLOCK=128,
num_warps=4, num_stages=1)
buf19 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
extern_kernels.mm(buf18, primals_8, out=buf19)
buf20 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
triton_poi_fused_leaky_relu_1[grid(16)](buf19, buf20, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf25 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(primals_1, primals_9, out=buf25)
del primals_9
buf26 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
triton_poi_fused_cat_0[grid(128)](buf25, buf26, 128, XBLOCK=128,
num_warps=4, num_stages=1)
buf27 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
extern_kernels.mm(buf26, primals_10, out=buf27)
buf28 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
triton_poi_fused_leaky_relu_1[grid(16)](buf27, buf28, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf5 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf6 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf13 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf14 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf21 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf22 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf29 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
buf30 = empty_strided_cuda((4, 1), (1, 4), torch.float32)
triton_poi_fused__softmax_leaky_relu_mul_where_2[grid(4)](buf4,
buf3, buf2, buf12, buf11, buf20, buf19, buf28, buf27, buf5,
buf6, buf13, buf14, buf21, buf22, buf29, buf30, 4, XBLOCK=4,
num_warps=1, num_stages=1)
buf7 = reinterpret_tensor(buf2, (4, 4), (4, 1), 0)
del buf2
buf15 = reinterpret_tensor(buf11, (4, 4), (4, 1), 0)
del buf11
buf23 = reinterpret_tensor(buf19, (4, 4), (4, 1), 0)
del buf19
buf31 = reinterpret_tensor(buf27, (4, 4), (4, 1), 0)
del buf27
triton_poi_fused__softmax_leaky_relu_mul_where_3[grid(16)](buf7,
buf15, buf23, buf31, buf4, buf3, buf5, buf6, buf12, buf13,
buf14, buf20, buf21, buf22, buf28, buf29, buf30, 16, XBLOCK=16,
num_warps=1, num_stages=1)
del buf13
del buf14
del buf21
del buf22
del buf29
del buf30
buf8 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(buf7, buf0, out=buf8)
buf16 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(buf15, buf9, out=buf16)
buf24 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(buf23, buf17, out=buf24)
buf32 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(buf31, buf25, out=buf32)
buf33 = empty_strided_cuda((4, 16), (16, 1), torch.float32)
triton_poi_fused_cat_4[grid(64)](buf8, buf16, buf24, buf32, buf33,
64, XBLOCK=64, num_warps=1, num_stages=1)
buf34 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(buf33, primals_11, out=buf34)
buf35 = empty_strided_cuda((16, 8), (8, 1), torch.float32)
triton_poi_fused_cat_0[grid(128)](buf34, buf35, 128, XBLOCK=128,
num_warps=4, num_stages=1)
buf36 = empty_strided_cuda((16, 1), (1, 1), torch.float32)
extern_kernels.mm(buf35, primals_12, out=buf36)
buf37 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
triton_poi_fused_leaky_relu_1[grid(16)](buf36, buf37, 16, XBLOCK=16,
num_warps=1, num_stages=1)
buf38 = buf6
del buf6
buf39 = buf5
del buf5
triton_poi_fused__softmax_leaky_relu_mul_where_5[grid(4)](buf4,
buf37, buf36, buf38, buf39, 4, XBLOCK=4, num_warps=1, num_stages=1)
buf40 = reinterpret_tensor(buf36, (4, 4), (4, 1), 0)
del buf36
triton_poi_fused__softmax_leaky_relu_mul_where_6[grid(16)](buf40,
buf4, buf37, buf38, buf39, 16, XBLOCK=16, num_warps=1, num_stages=1
)
del buf38
del buf39
buf41 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(buf40, buf34, out=buf41)
buf42 = buf41
del buf41
buf43 = empty_strided_cuda((4, 4), (4, 1), torch.bool)
triton_poi_fused_relu_threshold_backward_7[grid(16)](buf42, buf43,
16, XBLOCK=16, num_warps=1, num_stages=1)
return (buf42, buf3, buf4, buf7, buf8, buf12, buf15, buf16, buf20,
buf23, buf24, buf28, buf31, buf32, buf37, buf40, buf43,
reinterpret_tensor(buf34, (4, 4), (1, 4), 0), reinterpret_tensor(
buf35, (8, 16), (1, 8), 0), reinterpret_tensor(primals_12, (1, 8),
(1, 1), 0), reinterpret_tensor(buf33, (16, 4), (1, 16), 0),
reinterpret_tensor(primals_11, (4, 16), (1, 4), 0),
reinterpret_tensor(buf25, (4, 4), (1, 4), 0), reinterpret_tensor(
buf26, (8, 16), (1, 8), 0), reinterpret_tensor(primals_10, (1, 8),
(1, 1), 0), reinterpret_tensor(primals_1, (4, 4), (1, 4), 0),
reinterpret_tensor(buf17, (4, 4), (1, 4), 0), reinterpret_tensor(
buf18, (8, 16), (1, 8), 0), reinterpret_tensor(primals_8, (1, 8), (
1, 1), 0), reinterpret_tensor(buf9, (4, 4), (1, 4), 0),
reinterpret_tensor(buf10, (8, 16), (1, 8), 0), reinterpret_tensor(
primals_6, (1, 8), (1, 1), 0), reinterpret_tensor(buf0, (4, 4), (1,
4), 0), reinterpret_tensor(buf1, (8, 16), (1, 8), 0),
reinterpret_tensor(primals_3, (1, 8), (1, 1), 0))
class GraphAttentionLayer(nn.Module):
"""
Simple GAT layer, similar to https://arxiv.org/abs/1710.10903
"""
def __init__(self, in_features, out_features, dropout, alpha, concat=True):
super(GraphAttentionLayer, self).__init__()
self.dropout = dropout
self.in_features = in_features
self.out_features = out_features
self.alpha = alpha
self.concat = concat
self.W = nn.Parameter(torch.zeros(size=(in_features, out_features)))
nn.init.xavier_uniform_(self.W.data, gain=1.414)
self.a = nn.Parameter(torch.zeros(size=(2 * out_features, 1)))
nn.init.xavier_uniform_(self.a.data, gain=1.414)
self.leakyrelu = nn.LeakyReLU(self.alpha)
def forward(self, input, adj):
h = torch.mm(input, self.W)
N = h.size()[0]
a_input = torch.cat([h.repeat(1, N).view(N * N, -1), h.repeat(N, 1)
], dim=1).view(N, -1, 2 * self.out_features)
e = self.leakyrelu(torch.matmul(a_input, self.a).squeeze(2))
zero_vec = -9000000000000000.0 * torch.ones_like(e)
attention = torch.where(adj > 0, e, zero_vec)
attention = F.softmax(attention, dim=1)
attention = F.dropout(attention, self.dropout, training=self.training)
h_prime = torch.matmul(attention, h)
if self.concat:
return F.elu(h_prime)
else:
return h_prime
def __repr__(self):
return self.__class__.__name__ + ' (' + str(self.in_features
) + ' -> ' + str(self.out_features) + ')'
class GATNew(nn.Module):
def __init__(self, nfeat, nhid, nclass, dropout, alpha, nheads):
"""Dense version of GAT."""
super(GATNew, self).__init__()
self.dropout = dropout
self.attentions = [GraphAttentionLayer(nfeat, nhid, dropout=dropout,
alpha=alpha, concat=True) for _ in range(nheads)]
for i, attention in enumerate(self.attentions):
self.add_module('attention_{}'.format(i), attention)
self.out_att = GraphAttentionLayer(nhid * nheads, nclass, dropout=
dropout, alpha=alpha, concat=False)
def forward(self, input_0, input_1):
primals_1 = self.attention_0.W
primals_3 = self.attention_0.a
primals_2 = self.attention_1.W
primals_6 = self.attention_1.a
primals_4 = self.attention_2.W
primals_8 = self.attention_2.a
primals_5 = self.attention_3.W
primals_10 = self.attention_3.a
primals_11 = self.out_att.W
primals_12 = self.out_att.a
primals_7 = input_0
primals_9 = input_1
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11, primals_12])
return output[0]
|
dzb1998/pyGAT
|
GAT
| false | 12,400 |
[
"MIT"
] | 0 |
b794c65683bd12d3211e62b97732a905a24b9940
|
https://github.com/dzb1998/pyGAT/tree/b794c65683bd12d3211e62b97732a905a24b9940
|
FullAttention
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ay/caylcn737p2wwjm32cacv462xdgdut6ho32ptwxfu34t3i2tr75z.py
# Topologically Sorted Source Nodes: [QK], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# QK => clone
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_2,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_0 = async_compile.triton('triton_poi_fused_clone_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = (xindex // 4) % 4
x2 = (xindex // 16) % 4
x3 = (xindex // 64)
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (4*x2) + (16*x1) + (64*x3)), xmask)
tl.store(out_ptr0 + (x4), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ri/cricgdtr5c24l63g746gjtdd45qor3pkzmi7qmyygyd24ejrijb7.py
# Topologically Sorted Source Nodes: [QK], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# QK => clone_1
# Graph fragment:
# %clone_1 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_3,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_1 = async_compile.triton('triton_poi_fused_clone_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64, 4], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 64
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = (yindex // 16)
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (16*x2) + (64*y1)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + (4*y3)), tmp0, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ya/cyacjmbos4hqjbfqrids3ws7umxa2yjx6uocj6nk4q5qb3ifqsdu.py
# Topologically Sorted Source Nodes: [mul, A], Original ATen: [aten.mul, aten._softmax]
# Source node to ATen node mapping:
# A => amax, clone_2, exp, sub
# mul => mul
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_3, 0.5), kwargs = {})
# %clone_2 : [num_users=2] = call_function[target=torch.ops.aten.clone.default](args = (%mul,), kwargs = {memory_format: torch.contiguous_format})
# %amax : [num_users=1] = call_function[target=torch.ops.aten.amax.default](args = (%clone_2, [2], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%clone_2, %amax), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%sub,), kwargs = {})
triton_poi_fused__softmax_mul_2 = async_compile.triton('triton_poi_fused__softmax_mul_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused__softmax_mul_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused__softmax_mul_2(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp3 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp1 = 0.5
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = tl_math.exp(tmp14)
tl.store(out_ptr0 + (x2), tmp15, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/vb/cvb3oguu73uopye5dab6pedxgpbqv7hv76dtm7n5cat2xumuixgo.py
# Topologically Sorted Source Nodes: [queried_values], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# queried_values => clone_3
# Graph fragment:
# %clone_3 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_7,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_3 = async_compile.triton('triton_poi_fused_clone_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = (xindex // 4)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (4*x1), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + (4*x1)), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + (4*x1)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + (4*x1)), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1, arg2_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg2_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [QK], Original ATen: [aten.clone]
stream0 = get_raw_stream(0)
triton_poi_fused_clone_0.run(arg1_1, buf0, 256, grid=grid(256), stream=stream0)
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [QK], Original ATen: [aten.clone]
triton_poi_fused_clone_1.run(arg0_1, buf1, 64, 4, grid=grid(64, 4), stream=stream0)
del arg0_1
buf2 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [QK], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf0, (16, 4, 4), (16, 4, 1), 0), reinterpret_tensor(buf1, (16, 4, 4), (16, 4, 1), 0), out=buf2)
buf3 = reinterpret_tensor(buf1, (4, 4, 4, 4), (64, 4, 1, 16), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [mul, A], Original ATen: [aten.mul, aten._softmax]
triton_poi_fused__softmax_mul_2.run(buf2, buf3, 256, grid=grid(256), stream=stream0)
buf4 = reinterpret_tensor(buf2, (4, 4, 4, 4, 1), (64, 16, 4, 1, 1), 0); del buf2 # reuse
# Topologically Sorted Source Nodes: [queried_values], Original ATen: [aten.clone]
triton_poi_fused_clone_3.run(buf3, buf4, 256, grid=grid(256), stream=stream0)
buf5 = reinterpret_tensor(buf3, (4, 4, 4, 4, 1), (64, 16, 4, 1, 1), 0); del buf3 # reuse
# Topologically Sorted Source Nodes: [queried_values], Original ATen: [aten.clone]
triton_poi_fused_clone_0.run(arg2_1, buf5, 256, grid=grid(256), stream=stream0)
del arg2_1
buf6 = reinterpret_tensor(buf0, (16, 4, 4), (16, 4, 1), 0); del buf0 # reuse
# Topologically Sorted Source Nodes: [queried_values], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf4, (16, 4, 4), (16, 4, 1), 0), reinterpret_tensor(buf5, (16, 4, 4), (16, 4, 1), 0), out=buf6)
del buf4
buf7 = reinterpret_tensor(buf5, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf5 # reuse
# Topologically Sorted Source Nodes: [contiguous], Original ATen: [aten.clone]
triton_poi_fused_clone_0.run(buf6, buf7, 256, grid=grid(256), stream=stream0)
del buf6
return (buf7, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg2_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1, arg2_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from torch.nn import Module
import torch
from torch.nn import Dropout
class FullAttention(Module):
def __init__(self, use_dropout=False, attention_dropout=0.1):
super().__init__()
self.use_dropout = use_dropout
self.dropout = Dropout(attention_dropout)
def forward(self, queries, keys, values, q_mask=None, kv_mask=None):
""" Multi-head scaled dot-product attention, a.k.a full attention.
Args:
queries: [N, L, H, D]
keys: [N, S, H, D]
values: [N, S, H, D]
q_mask: [N, L]
kv_mask: [N, S]
Returns:
queried_values: (N, L, H, D)
"""
QK = torch.einsum('nlhd,nshd->nlsh', queries, keys)
if kv_mask is not None:
QK.masked_fill_(~(q_mask[:, :, None, None] * kv_mask[:, None, :,
None]), float('-inf'))
softmax_temp = 1.0 / queries.size(3) ** 0.5
A = torch.softmax(softmax_temp * QK, dim=2)
if self.use_dropout:
A = self.dropout(A)
queried_values = torch.einsum('nlsh,nshd->nlhd', A, values)
return queried_values.contiguous()
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4]), torch.rand(
[4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
from torch.nn import Module
from torch.nn import Dropout
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_clone_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = xindex // 4 % 4
x2 = xindex // 16 % 4
x3 = xindex // 64
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 4 * x2 + 16 * x1 + 64 * x3), xmask)
tl.store(out_ptr0 + x4, tmp0, xmask)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 64
xnumel = 4
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y0 = yindex % 16
y1 = yindex // 16
y3 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 16 * x2 + 64 * y1), xmask & ymask,
eviction_policy='evict_last')
tl.store(out_ptr0 + (x2 + 4 * y3), tmp0, xmask & ymask)
@triton.jit
def triton_poi_fused__softmax_mul_2(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp3 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp8 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp11 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last'
)
tmp1 = 0.5
tmp2 = tmp0 * tmp1
tmp4 = tmp3 * tmp1
tmp6 = tmp5 * tmp1
tmp7 = triton_helpers.maximum(tmp4, tmp6)
tmp9 = tmp8 * tmp1
tmp10 = triton_helpers.maximum(tmp7, tmp9)
tmp12 = tmp11 * tmp1
tmp13 = triton_helpers.maximum(tmp10, tmp12)
tmp14 = tmp2 - tmp13
tmp15 = tl_math.exp(tmp14)
tl.store(out_ptr0 + x2, tmp15, xmask)
@triton.jit
def triton_poi_fused_clone_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x1 = xindex // 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + 4 * x1, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (1 + 4 * x1), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (2 + 4 * x1), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (3 + 4 * x1), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = tmp0 / tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
def call(args):
arg0_1, arg1_1, arg2_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg2_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch
.float32)
get_raw_stream(0)
triton_poi_fused_clone_0[grid(256)](arg1_1, buf0, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch
.float32)
triton_poi_fused_clone_1[grid(64, 4)](arg0_1, buf1, 64, 4, XBLOCK=4,
YBLOCK=32, num_warps=4, num_stages=1)
del arg0_1
buf2 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf0, (16, 4, 4), (16, 4, 1),
0), reinterpret_tensor(buf1, (16, 4, 4), (16, 4, 1), 0), out=buf2)
buf3 = reinterpret_tensor(buf1, (4, 4, 4, 4), (64, 4, 1, 16), 0)
del buf1
triton_poi_fused__softmax_mul_2[grid(256)](buf2, buf3, 256, XBLOCK=
256, num_warps=4, num_stages=1)
buf4 = reinterpret_tensor(buf2, (4, 4, 4, 4, 1), (64, 16, 4, 1, 1), 0)
del buf2
triton_poi_fused_clone_3[grid(256)](buf3, buf4, 256, XBLOCK=256,
num_warps=4, num_stages=1)
buf5 = reinterpret_tensor(buf3, (4, 4, 4, 4, 1), (64, 16, 4, 1, 1), 0)
del buf3
triton_poi_fused_clone_0[grid(256)](arg2_1, buf5, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del arg2_1
buf6 = reinterpret_tensor(buf0, (16, 4, 4), (16, 4, 1), 0)
del buf0
extern_kernels.bmm(reinterpret_tensor(buf4, (16, 4, 4), (16, 4, 1),
0), reinterpret_tensor(buf5, (16, 4, 4), (16, 4, 1), 0), out=buf6)
del buf4
buf7 = reinterpret_tensor(buf5, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf5
triton_poi_fused_clone_0[grid(256)](buf6, buf7, 256, XBLOCK=128,
num_warps=4, num_stages=1)
del buf6
return buf7,
class FullAttentionNew(Module):
def __init__(self, use_dropout=False, attention_dropout=0.1):
super().__init__()
self.use_dropout = use_dropout
self.dropout = Dropout(attention_dropout)
def forward(self, input_0, input_1, input_2):
arg0_1 = input_0
arg1_1 = input_1
arg2_1 = input_2
output = call([arg0_1, arg1_1, arg2_1])
return output[0]
|
francescodisalvo05/LoFTR
|
FullAttention
| false | 12,401 |
[
"Apache-2.0"
] | 0 |
66372ebbe1ea97d57fe6cb8b5acf5cd92a87ef8d
|
https://github.com/francescodisalvo05/LoFTR/tree/66372ebbe1ea97d57fe6cb8b5acf5cd92a87ef8d
|
SpatialTransformerPooled2d
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/nu/cnuwsai2e6ger7eapl25fpdvtr5txtxw3k77jhqjlxa723cyqadi.py
# Topologically Sorted Source Nodes: [clamp], Original ATen: [aten.clamp]
# Source node to ATen node mapping:
# clamp => clamp_max, clamp_min
# Graph fragment:
# %clamp_min : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%primals_1, -1), kwargs = {})
# %clamp_max : [num_users=3] = call_function[target=torch.ops.aten.clamp_max.default](args = (%clamp_min, 1), kwargs = {})
triton_poi_fused_clamp_0 = async_compile.triton('triton_poi_fused_clamp_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clamp_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clamp_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 8
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = -1.0
tmp2 = triton_helpers.maximum(tmp0, tmp1)
tmp3 = 1.0
tmp4 = triton_helpers.minimum(tmp2, tmp3)
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/4f/c4frkkhm2pmx5e7fgt7wuzef2gfzodf25s5iiclgfvq2z3k7rmgq.py
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.avg_pool2d]
# Source node to ATen node mapping:
# x => avg_pool2d
# Graph fragment:
# %avg_pool2d : [num_users=2] = call_function[target=torch.ops.aten.avg_pool2d.default](args = (%primals_2, [2, 2], [2, 2], [0, 0], False, False), kwargs = {})
triton_poi_fused_avg_pool2d_1 = async_compile.triton('triton_poi_fused_avg_pool2d_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_avg_pool2d_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_avg_pool2d_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 2
x1 = (xindex // 2)
x2 = xindex
tmp0 = tl.load(in_ptr0 + ((2*x0) + (8*x1)), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + (2*x0) + (8*x1)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (4 + (2*x0) + (8*x1)), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (5 + (2*x0) + (8*x1)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + (x2), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kf/ckfghrkocduvmovgzgimf5n4tizoinlqpdnmgn4qbkr27p4qmmji.py
# Topologically Sorted Source Nodes: [y], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# y => cat
# Graph fragment:
# %cat : [num_users=2] = call_function[target=torch.ops.aten.cat.default](args = ([%cudnn_grid_sampler, %cudnn_grid_sampler_1], 1), kwargs = {})
triton_poi_fused_cat_2 = async_compile.triton('triton_poi_fused_cat_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4) % 8
x0 = xindex % 4
x2 = (xindex // 32)
x3 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (4*x1) + (16*x2)), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + (x0 + (4*((-4) + x1)) + (16*x2)), tmp6 & xmask, other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x3), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/k5/ck52tyqilmtir6psuojqrywf3qyj2eukeapqioueittl6drbmtsz.py
# Topologically Sorted Source Nodes: [mul, sum_1, y_2], Original ATen: [aten.mul, aten.sum, aten.add]
# Source node to ATen node mapping:
# mul => mul
# sum_1 => sum_1
# y_2 => add
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%squeeze, %view), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul, [1]), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sum_1, %primals_4), kwargs = {})
triton_per_fused_add_mul_sum_3 = async_compile.triton('triton_per_fused_add_mul_sum_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[16, 8],
reduction_hint=ReductionHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_mul_sum_3', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_mul_sum_3(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 16
rnumel = 8
RBLOCK: tl.constexpr = 8
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r2 = rindex
x0 = xindex % 4
x1 = (xindex // 4)
x3 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (4*r2) + (32*x1)), xmask, other=0.0)
tmp1 = tl.load(in_ptr1 + (x0 + (4*r2)), xmask, eviction_policy='evict_last', other=0.0)
tmp7 = tl.load(in_ptr2 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp3 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5 = tl.where(xmask, tmp3, 0)
tmp6 = tl.sum(tmp5, 1)[:, None]
tmp8 = tmp6 + tmp7
tl.debug_barrier()
tl.store(in_out_ptr0 + (x3), tmp8, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 4, 1, 2), (8, 2, 2, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (1, 8, 1, 4), (32, 4, 4, 1))
assert_size_stride(primals_4, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 4, 1, 2), (8, 2, 2, 1), torch.float32)
# Topologically Sorted Source Nodes: [clamp], Original ATen: [aten.clamp]
stream0 = get_raw_stream(0)
triton_poi_fused_clamp_0.run(primals_1, buf0, 8, grid=grid(8), stream=stream0)
# Topologically Sorted Source Nodes: [grid_sample], Original ATen: [aten.cudnn_grid_sampler]
buf1 = torch.ops.aten.cudnn_grid_sampler.default(primals_2, reinterpret_tensor(buf0, (4, 4, 1, 2), (0, 2, 0, 1), 0))
buf2 = buf1
buf3 = empty_strided_cuda((4, 4, 2, 2), (16, 4, 2, 1), torch.float32)
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.avg_pool2d]
triton_poi_fused_avg_pool2d_1.run(primals_2, buf3, 64, grid=grid(64), stream=stream0)
# Topologically Sorted Source Nodes: [grid_sample_1], Original ATen: [aten.cudnn_grid_sampler]
buf4 = torch.ops.aten.cudnn_grid_sampler.default(buf3, reinterpret_tensor(buf0, (4, 4, 1, 2), (0, 2, 0, 1), 0))
buf5 = buf4
buf6 = empty_strided_cuda((4, 8, 4, 1), (32, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [y], Original ATen: [aten.cat]
triton_poi_fused_cat_2.run(buf2, buf5, buf6, 128, grid=grid(128), stream=stream0)
del buf2
del buf5
buf7 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
buf8 = buf7; del buf7 # reuse
# Topologically Sorted Source Nodes: [mul, sum_1, y_2], Original ATen: [aten.mul, aten.sum, aten.add]
triton_per_fused_add_mul_sum_3.run(buf8, buf6, primals_3, primals_4, 16, 8, grid=grid(16), stream=stream0)
del primals_4
# Topologically Sorted Source Nodes: [], Original ATen: []
buf9 = torch.ops.aten.set_.source_Tensor(primals_1, buf0)
assert_size_stride(buf9, (1, 4, 1, 2), (8, 2, 2, 1))
del buf1
del buf4
del primals_1
return (buf8, primals_2, primals_3, buf0, buf3, buf6, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((1, 4, 1, 2), (8, 2, 2, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((1, 8, 1, 4), (32, 4, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn import Parameter
from torch.nn import functional as F
def positive(weight, cache=None):
weight.data *= weight.data.ge(0).float()
return cache
class SpatialTransformerPooled2d(nn.Module):
def __init__(self, in_shape, outdims, pool_steps=1, positive=False,
bias=True, pool_kern=2, init_range=0.1):
super().__init__()
self._pool_steps = pool_steps
self.in_shape = in_shape
c, _w, _h = in_shape
self.outdims = outdims
self.positive = positive
self.grid = Parameter(torch.Tensor(1, outdims, 1, 2))
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps + 1
), 1, outdims))
if bias:
bias = Parameter(torch.Tensor(outdims))
self.register_parameter('bias', bias)
else:
self.register_parameter('bias', None)
self.pool_kern = pool_kern
self.avg = nn.AvgPool2d((pool_kern, pool_kern), stride=pool_kern,
count_include_pad=False)
self.init_range = init_range
self.initialize()
@property
def pool_steps(self):
return self._pool_steps
@pool_steps.setter
def pool_steps(self, value):
assert value >= 0 and int(value
) - value == 0, 'new pool steps must be a non-negative integer'
if value != self._pool_steps:
None
c, _w, _h = self.in_shape
self._pool_steps = int(value)
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps +
1), 1, self.outdims))
self.features.data.fill_(1 / self.in_shape[0])
def initialize(self):
self.grid.data.uniform_(-self.init_range, self.init_range)
self.features.data.fill_(1 / self.in_shape[0])
if self.bias is not None:
self.bias.data.fill_(0)
def feature_l1(self, average=True):
if average:
return self.features.abs().mean()
else:
return self.features.abs().sum()
def group_sparsity(self, group_size):
f = self.features.size(1)
n = f // group_size
ret = 0
for chunk in range(0, f, group_size):
ret = ret + (self.features[:, chunk:chunk + group_size, ...].
pow(2).mean(1) + 1e-12).sqrt().mean() / n
return ret
def forward(self, x, shift=None, out_idx=None):
if self.positive:
positive(self.features)
self.grid.data = torch.clamp(self.grid.data, -1, 1)
N, c, _w, _h = x.size()
m = self.pool_steps + 1
feat = self.features.view(1, m * c, self.outdims)
if out_idx is None:
grid = self.grid
bias = self.bias
outdims = self.outdims
else:
feat = feat[:, :, out_idx]
grid = self.grid[:, out_idx]
if self.bias is not None:
bias = self.bias[out_idx]
outdims = len(out_idx)
if shift is None:
grid = grid.expand(N, outdims, 1, 2)
else:
grid = grid.expand(N, outdims, 1, 2) + shift[:, None, None, :]
pools = [F.grid_sample(x, grid, align_corners=True)]
for _ in range(self.pool_steps):
x = self.avg(x)
pools.append(F.grid_sample(x, grid, align_corners=True))
y = torch.cat(pools, dim=1)
y = (y.squeeze(-1) * feat).sum(1).view(N, outdims)
if self.bias is not None:
y = y + bias
return y
def __repr__(self):
c, w, h = self.in_shape
r = self.__class__.__name__ + ' (' + '{} x {} x {}'.format(c, w, h
) + ' -> ' + str(self.outdims) + ')'
if self.bias is not None:
r += ' with bias'
r += ' and pooling for {} steps\n'.format(self.pool_steps)
for ch in self.children():
r += ' -> ' + ch.__repr__() + '\n'
return r
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'in_shape': [4, 4, 4], 'outdims': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch import nn
from torch.nn import Parameter
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_clamp_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 8
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = -1.0
tmp2 = triton_helpers.maximum(tmp0, tmp1)
tmp3 = 1.0
tmp4 = triton_helpers.minimum(tmp2, tmp3)
tl.store(out_ptr0 + x0, tmp4, xmask)
@triton.jit
def triton_poi_fused_avg_pool2d_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 2
x1 = xindex // 2
x2 = xindex
tmp0 = tl.load(in_ptr0 + (2 * x0 + 8 * x1), xmask, eviction_policy=
'evict_last')
tmp1 = tl.load(in_ptr0 + (1 + 2 * x0 + 8 * x1), xmask, eviction_policy=
'evict_last')
tmp3 = tl.load(in_ptr0 + (4 + 2 * x0 + 8 * x1), xmask, eviction_policy=
'evict_last')
tmp5 = tl.load(in_ptr0 + (5 + 2 * x0 + 8 * x1), xmask, eviction_policy=
'evict_last')
tmp2 = tmp1 + tmp0
tmp4 = tmp3 + tmp2
tmp6 = tmp5 + tmp4
tmp7 = 0.25
tmp8 = tmp6 * tmp7
tl.store(out_ptr0 + x2, tmp8, xmask)
@triton.jit
def triton_poi_fused_cat_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4 % 8
x0 = xindex % 4
x2 = xindex // 32
x3 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 4 * x1 + 16 * x2), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (x0 + 4 * (-4 + x1) + 16 * x2), tmp6 & xmask,
other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x3, tmp10, xmask)
@triton.jit
def triton_per_fused_add_mul_sum_3(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2,
xnumel, rnumel, XBLOCK: tl.constexpr):
xnumel = 16
RBLOCK: tl.constexpr = 8
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r2 = rindex
x0 = xindex % 4
x1 = xindex // 4
x3 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 4 * r2 + 32 * x1), xmask, other=0.0)
tmp1 = tl.load(in_ptr1 + (x0 + 4 * r2), xmask, eviction_policy=
'evict_last', other=0.0)
tmp7 = tl.load(in_ptr2 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp3 = tl.broadcast_to(tmp2, [XBLOCK, RBLOCK])
tmp5 = tl.where(xmask, tmp3, 0)
tmp6 = tl.sum(tmp5, 1)[:, None]
tmp8 = tmp6 + tmp7
tl.debug_barrier()
tl.store(in_out_ptr0 + x3, tmp8, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4 = args
args.clear()
assert_size_stride(primals_1, (1, 4, 1, 2), (8, 2, 2, 1))
assert_size_stride(primals_2, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_3, (1, 8, 1, 4), (32, 4, 4, 1))
assert_size_stride(primals_4, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((1, 4, 1, 2), (8, 2, 2, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_clamp_0[grid(8)](primals_1, buf0, 8, XBLOCK=8,
num_warps=1, num_stages=1)
buf1 = torch.ops.aten.cudnn_grid_sampler.default(primals_2,
reinterpret_tensor(buf0, (4, 4, 1, 2), (0, 2, 0, 1), 0))
buf2 = buf1
buf3 = empty_strided_cuda((4, 4, 2, 2), (16, 4, 2, 1), torch.float32)
triton_poi_fused_avg_pool2d_1[grid(64)](primals_2, buf3, 64, XBLOCK
=64, num_warps=1, num_stages=1)
buf4 = torch.ops.aten.cudnn_grid_sampler.default(buf3,
reinterpret_tensor(buf0, (4, 4, 1, 2), (0, 2, 0, 1), 0))
buf5 = buf4
buf6 = empty_strided_cuda((4, 8, 4, 1), (32, 4, 1, 1), torch.float32)
triton_poi_fused_cat_2[grid(128)](buf2, buf5, buf6, 128, XBLOCK=128,
num_warps=4, num_stages=1)
del buf2
del buf5
buf7 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
buf8 = buf7
del buf7
triton_per_fused_add_mul_sum_3[grid(16)](buf8, buf6, primals_3,
primals_4, 16, 8, XBLOCK=1, num_warps=2, num_stages=1)
del primals_4
buf9 = torch.ops.aten.set_.source_Tensor(primals_1, buf0)
assert_size_stride(buf9, (1, 4, 1, 2), (8, 2, 2, 1))
del buf1
del buf4
del primals_1
return buf8, primals_2, primals_3, buf0, buf3, buf6
def positive(weight, cache=None):
weight.data *= weight.data.ge(0).float()
return cache
class SpatialTransformerPooled2dNew(nn.Module):
def __init__(self, in_shape, outdims, pool_steps=1, positive=False,
bias=True, pool_kern=2, init_range=0.1):
super().__init__()
self._pool_steps = pool_steps
self.in_shape = in_shape
c, _w, _h = in_shape
self.outdims = outdims
self.positive = positive
self.grid = Parameter(torch.Tensor(1, outdims, 1, 2))
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps + 1
), 1, outdims))
if bias:
bias = Parameter(torch.Tensor(outdims))
self.register_parameter('bias', bias)
else:
self.register_parameter('bias', None)
self.pool_kern = pool_kern
self.avg = nn.AvgPool2d((pool_kern, pool_kern), stride=pool_kern,
count_include_pad=False)
self.init_range = init_range
self.initialize()
@property
def pool_steps(self):
return self._pool_steps
@pool_steps.setter
def pool_steps(self, value):
assert value >= 0 and int(value
) - value == 0, 'new pool steps must be a non-negative integer'
if value != self._pool_steps:
None
c, _w, _h = self.in_shape
self._pool_steps = int(value)
self.features = Parameter(torch.Tensor(1, c * (self._pool_steps +
1), 1, self.outdims))
self.features.data.fill_(1 / self.in_shape[0])
def initialize(self):
self.grid.data.uniform_(-self.init_range, self.init_range)
self.features.data.fill_(1 / self.in_shape[0])
if self.bias is not None:
self.bias.data.fill_(0)
def feature_l1(self, average=True):
if average:
return self.features.abs().mean()
else:
return self.features.abs().sum()
def group_sparsity(self, group_size):
f = self.features.size(1)
n = f // group_size
ret = 0
for chunk in range(0, f, group_size):
ret = ret + (self.features[:, chunk:chunk + group_size, ...].
pow(2).mean(1) + 1e-12).sqrt().mean() / n
return ret
def __repr__(self):
c, w, h = self.in_shape
r = self.__class__.__name__ + ' (' + '{} x {} x {}'.format(c, w, h
) + ' -> ' + str(self.outdims) + ')'
if self.bias is not None:
r += ' with bias'
r += ' and pooling for {} steps\n'.format(self.pool_steps)
for ch in self.children():
r += ' -> ' + ch.__repr__() + '\n'
return r
def forward(self, input_0):
primals_1 = self.grid
primals_3 = self.features
primals_4 = self.bias
primals_2 = input_0
output = call([primals_1, primals_2, primals_3, primals_4])
return output[0]
|
dattientran/attorch
|
SpatialTransformerPooled2d
| false | 12,402 |
[
"MIT"
] | 0 |
469b225846c6d8a7d833ebac19d040c7a407a0ff
|
https://github.com/dattientran/attorch/tree/469b225846c6d8a7d833ebac19d040c7a407a0ff
|
DeepHeadModule
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/td/ctdybbibnws4d7ukbk3fpn35zkgapxylowdhzwx7vgsllncbdrxa.py
# Topologically Sorted Source Nodes: [conv2d, relu], Original ATen: [aten.convolution, aten.relu]
# Source node to ATen node mapping:
# conv2d => convolution
# relu => relu
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution,), kwargs = {})
triton_poi_fused_convolution_relu_0 = async_compile.triton('triton_poi_fused_convolution_relu_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 16) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x3), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/32/c32v7egt4mupqssam3gmac2qgv3ujprjybthsgweflmot256qqw7.py
# Topologically Sorted Source Nodes: [conv2d_3], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# conv2d_3 => convolution_3
# Graph fragment:
# %convolution_3 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%relu_2, %primals_8, %primals_9, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
triton_poi_fused_convolution_1 = async_compile.triton('triton_poi_fused_convolution_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 16) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_5, (4, ), (1, ))
assert_size_stride(primals_6, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_7, (4, ), (1, ))
assert_size_stride(primals_8, (4, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_9, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 4, 4), (64, 16, 4, 1))
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [conv2d, relu], Original ATen: [aten.convolution, aten.relu]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_relu_0.run(buf1, primals_2, 256, grid=grid(256), stream=stream0)
del primals_2
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
buf2 = extern_kernels.convolution(buf1, primals_4, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 4, 4, 4), (64, 16, 4, 1))
buf3 = buf2; del buf2 # reuse
# Topologically Sorted Source Nodes: [conv2d_1, relu_1], Original ATen: [aten.convolution, aten.relu]
triton_poi_fused_convolution_relu_0.run(buf3, primals_5, 256, grid=grid(256), stream=stream0)
del primals_5
# Topologically Sorted Source Nodes: [conv2d_2], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(buf3, primals_6, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 4, 4), (64, 16, 4, 1))
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [conv2d_2, relu_2], Original ATen: [aten.convolution, aten.relu]
triton_poi_fused_convolution_relu_0.run(buf5, primals_7, 256, grid=grid(256), stream=stream0)
del primals_7
# Topologically Sorted Source Nodes: [conv2d_3], Original ATen: [aten.convolution]
buf6 = extern_kernels.convolution(buf5, primals_8, stride=(1, 1), padding=(0, 0), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 4, 4, 4), (64, 16, 4, 1))
buf7 = buf6; del buf6 # reuse
# Topologically Sorted Source Nodes: [conv2d_3], Original ATen: [aten.convolution]
triton_poi_fused_convolution_1.run(buf7, primals_9, 256, grid=grid(256), stream=stream0)
del primals_9
return (buf7, primals_1, primals_3, primals_4, primals_6, primals_8, buf1, buf3, buf5, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 3, 3), (36, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4, 3, 3), (36, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4, 3, 3), (36, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, 4, 1, 1), (4, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
from math import sqrt as sqrt
from itertools import product as product
class DeepHeadModule(nn.Module):
def __init__(self, input_channels, output_channels):
super(DeepHeadModule, self).__init__()
self._input_channels = input_channels
self._output_channels = output_channels
self._mid_channels = min(self._input_channels, 256)
self.conv1 = nn.Conv2d(self._input_channels, self._mid_channels,
kernel_size=3, dilation=1, stride=1, padding=1)
self.conv2 = nn.Conv2d(self._mid_channels, self._mid_channels,
kernel_size=3, dilation=1, stride=1, padding=1)
self.conv3 = nn.Conv2d(self._mid_channels, self._mid_channels,
kernel_size=3, dilation=1, stride=1, padding=1)
self.conv4 = nn.Conv2d(self._mid_channels, self._output_channels,
kernel_size=1, dilation=1, stride=1, padding=0)
def forward(self, x):
return self.conv4(F.relu(self.conv3(F.relu(self.conv2(F.relu(self.
conv1(x), inplace=True)), inplace=True)), inplace=True))
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'input_channels': 4, 'output_channels': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
from math import sqrt as sqrt
from itertools import product as product
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
@triton.jit
def triton_poi_fused_convolution_relu_0(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 16 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x3, tmp4, xmask)
@triton.jit
def triton_poi_fused_convolution_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 16 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9) = args
args.clear()
assert_size_stride(primals_1, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_5, (4,), (1,))
assert_size_stride(primals_6, (4, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_7, (4,), (1,))
assert_size_stride(primals_8, (4, 4, 1, 1), (4, 1, 1, 1))
assert_size_stride(primals_9, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,
1), padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 4, 4), (64, 16, 4, 1))
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_convolution_relu_0[grid(256)](buf1, primals_2, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del primals_2
buf2 = extern_kernels.convolution(buf1, primals_4, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 4, 4, 4), (64, 16, 4, 1))
buf3 = buf2
del buf2
triton_poi_fused_convolution_relu_0[grid(256)](buf3, primals_5, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del primals_5
buf4 = extern_kernels.convolution(buf3, primals_6, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 4, 4, 4), (64, 16, 4, 1))
buf5 = buf4
del buf4
triton_poi_fused_convolution_relu_0[grid(256)](buf5, primals_7, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del primals_7
buf6 = extern_kernels.convolution(buf5, primals_8, stride=(1, 1),
padding=(0, 0), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 4, 4, 4), (64, 16, 4, 1))
buf7 = buf6
del buf6
triton_poi_fused_convolution_1[grid(256)](buf7, primals_9, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del primals_9
return (buf7, primals_1, primals_3, primals_4, primals_6, primals_8,
buf1, buf3, buf5)
class DeepHeadModuleNew(nn.Module):
def __init__(self, input_channels, output_channels):
super(DeepHeadModuleNew, self).__init__()
self._input_channels = input_channels
self._output_channels = output_channels
self._mid_channels = min(self._input_channels, 256)
self.conv1 = nn.Conv2d(self._input_channels, self._mid_channels,
kernel_size=3, dilation=1, stride=1, padding=1)
self.conv2 = nn.Conv2d(self._mid_channels, self._mid_channels,
kernel_size=3, dilation=1, stride=1, padding=1)
self.conv3 = nn.Conv2d(self._mid_channels, self._mid_channels,
kernel_size=3, dilation=1, stride=1, padding=1)
self.conv4 = nn.Conv2d(self._mid_channels, self._output_channels,
kernel_size=1, dilation=1, stride=1, padding=0)
def forward(self, input_0):
primals_1 = self.conv1.weight
primals_2 = self.conv1.bias
primals_4 = self.conv2.weight
primals_5 = self.conv2.bias
primals_6 = self.conv3.weight
primals_7 = self.conv3.bias
primals_8 = self.conv4.weight
primals_9 = self.conv4.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9])
return output[0]
|
fuankarion/FaceDetection-DSFD
|
DeepHeadModule
| false | 12,403 |
[
"Apache-2.0"
] | 0 |
f1e464ec5c9d95c2fe73edf44e4d414a464839b1
|
https://github.com/fuankarion/FaceDetection-DSFD/tree/f1e464ec5c9d95c2fe73edf44e4d414a464839b1
|
LeCunTanh
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/y5/cy57u6wvqibomvk4ejb62hdxm5umhgocmkfyideehrduq5mkrygt.py
# Topologically Sorted Source Nodes: [mul, tanh, mul_1], Original ATen: [aten.mul, aten.tanh]
# Source node to ATen node mapping:
# mul => mul
# mul_1 => mul_1
# tanh => tanh
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, 0.6666666666666666), kwargs = {})
# %tanh : [num_users=1] = call_function[target=torch.ops.aten.tanh.default](args = (%mul,), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%tanh, 1.7159), kwargs = {})
triton_poi_fused_mul_tanh_0 = async_compile.triton('triton_poi_fused_mul_tanh_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_tanh_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_tanh_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = 0.6666666666666666
tmp2 = tmp0 * tmp1
tmp3 = libdevice.tanh(tmp2)
tmp4 = 1.7159
tmp5 = tmp3 * tmp4
tl.store(out_ptr0 + (x0), tmp5, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, tanh, mul_1], Original ATen: [aten.mul, aten.tanh]
stream0 = get_raw_stream(0)
triton_poi_fused_mul_tanh_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class LeCunTanh(nn.Module):
def forward(self, x):
return 1.7159 * torch.tanh(2.0 / 3 * x)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_mul_tanh_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = 0.6666666666666666
tmp2 = tmp0 * tmp1
tmp3 = libdevice.tanh(tmp2)
tmp4 = 1.7159
tmp5 = tmp3 * tmp4
tl.store(out_ptr0 + x0, tmp5, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_mul_tanh_0[grid(256)](arg0_1, buf0, 256, XBLOCK=
256, num_warps=4, num_stages=1)
del arg0_1
return buf0,
class LeCunTanhNew(nn.Module):
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
fmhoward/pysurvival
|
LeCunTanh
| false | 12,404 |
[
"Apache-2.0"
] | 0 |
3fea55f09477e9f0844845e09d6ea60434436e2e
|
https://github.com/fmhoward/pysurvival/tree/3fea55f09477e9f0844845e09d6ea60434436e2e
|
CausalConv1d
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/gk/cgkw7mtnlzc2x4r4dlaxhd27twdsy4zp5nunyiqlzxzxlut26kgd.py
# Topologically Sorted Source Nodes: [conv1d], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# conv1d => convolution
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1], [2], [2], False, [0], 1), kwargs = {})
triton_poi_fused_convolution_0 = async_compile.triton('triton_poi_fused_convolution_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 96
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 6) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 2), (8, 2, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv1d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,), padding=(2,), dilation=(2,), transposed=False, output_padding=(0,), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 6), (24, 6, 1))
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [conv1d], Original ATen: [aten.convolution]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_0.run(buf1, primals_2, 96, grid=grid(96), stream=stream0)
del primals_2
return (reinterpret_tensor(buf1, (4, 4, 4), (24, 6, 1), 0), primals_1, primals_3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 2), (8, 2, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
class CausalConv1d(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size=2, dilation=2):
super(CausalConv1d, self).__init__()
self.padding = dilation
self.causal_conv = nn.Conv1d(in_channels, out_channels, kernel_size,
padding=self.padding, dilation=dilation)
def forward(self, minibatch):
return self.causal_conv(minibatch)[:, :, :-self.padding]
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'in_channels': 4, 'out_channels': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 96
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 6 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
def call(args):
primals_1, primals_2, primals_3 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 2), (8, 2, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,),
padding=(2,), dilation=(2,), transposed=False, output_padding=(
0,), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 6), (24, 6, 1))
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_convolution_0[grid(96)](buf1, primals_2, 96,
XBLOCK=128, num_warps=4, num_stages=1)
del primals_2
return reinterpret_tensor(buf1, (4, 4, 4), (24, 6, 1), 0
), primals_1, primals_3
class CausalConv1dNew(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size=2, dilation=2):
super(CausalConv1dNew, self).__init__()
self.padding = dilation
self.causal_conv = nn.Conv1d(in_channels, out_channels, kernel_size,
padding=self.padding, dilation=dilation)
def forward(self, input_0):
primals_1 = self.causal_conv.weight
primals_2 = self.causal_conv.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3])
return output[0]
|
gaetangate/FewRel
|
CausalConv1d
| false | 12,405 |
[
"MIT"
] | 0 |
150199d1060571315b1f370b3b3352d7a7c72dd5
|
https://github.com/gaetangate/FewRel/tree/150199d1060571315b1f370b3b3352d7a7c72dd5
|
LinearAttention
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/og/cogo4pfcy56l7ulvsx6und2vmih3qojwpvkiazyzh7hsptmp4vos.py
# Topologically Sorted Source Nodes: [elu_1, K, sum_1], Original ATen: [aten.elu, aten.add, aten.sum]
# Source node to ATen node mapping:
# K => add_1
# elu_1 => expm1_1, gt_1, mul_3, mul_4, mul_5, where_1
# sum_1 => sum_1
# Graph fragment:
# %gt_1 : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%arg1_1, 0), kwargs = {})
# %mul_3 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg1_1, 1.0), kwargs = {})
# %mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg1_1, 1.0), kwargs = {})
# %expm1_1 : [num_users=1] = call_function[target=torch.ops.aten.expm1.default](args = (%mul_4,), kwargs = {})
# %mul_5 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%expm1_1, 1.0), kwargs = {})
# %where_1 : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt_1, %mul_3, %mul_5), kwargs = {})
# %add_1 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%where_1, 1), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%add_1, [1]), kwargs = {})
triton_poi_fused_add_elu_sum_0 = async_compile.triton('triton_poi_fused_add_elu_sum_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_elu_sum_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_elu_sum_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = (xindex // 16)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (64*x1)), xmask)
tmp9 = tl.load(in_ptr0 + (16 + x0 + (64*x1)), xmask)
tmp17 = tl.load(in_ptr0 + (32 + x0 + (64*x1)), xmask)
tmp25 = tl.load(in_ptr0 + (48 + x0 + (64*x1)), xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tmp10 = tmp9 > tmp1
tmp11 = tmp9 * tmp3
tmp12 = libdevice.expm1(tmp11)
tmp13 = tmp12 * tmp3
tmp14 = tl.where(tmp10, tmp11, tmp13)
tmp15 = tmp14 + tmp3
tmp16 = tmp8 + tmp15
tmp18 = tmp17 > tmp1
tmp19 = tmp17 * tmp3
tmp20 = libdevice.expm1(tmp19)
tmp21 = tmp20 * tmp3
tmp22 = tl.where(tmp18, tmp19, tmp21)
tmp23 = tmp22 + tmp3
tmp24 = tmp16 + tmp23
tmp26 = tmp25 > tmp1
tmp27 = tmp25 * tmp3
tmp28 = libdevice.expm1(tmp27)
tmp29 = tmp28 * tmp3
tmp30 = tl.where(tmp26, tmp27, tmp29)
tmp31 = tmp30 + tmp3
tmp32 = tmp24 + tmp31
tl.store(out_ptr0 + (x2), tmp32, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ui/cuicp7lcen5tosk4olu3x3oaa7yz6cvucka5zwp4hr2qbvjrvvxh.py
# Topologically Sorted Source Nodes: [einsum_1, einsum_2], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# einsum_1 => clone_2
# einsum_2 => clone_3
# Graph fragment:
# %clone_2 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_7,), kwargs = {memory_format: torch.contiguous_format})
# %clone_3 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_13,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_1 = async_compile.triton('triton_poi_fused_clone_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x0 = xindex % 4
x1 = (xindex // 4) % 4
x2 = (xindex // 16) % 4
x3 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x4), xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tl.store(out_ptr0 + (x0 + (4*x2) + (16*x1) + (64*x3)), tmp8, xmask)
tl.store(out_ptr1 + (x0 + (4*x2) + (16*x1) + (64*x3)), tmp8, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ds/cdscs66baluy2qjp2egcmspwpssho6a6gwzskgkoo4xnrapvrluw.py
# Topologically Sorted Source Nodes: [KV], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# KV => clone
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_2,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_2 = async_compile.triton('triton_poi_fused_clone_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 16], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 4
y1 = (yindex // 4)
tmp0 = tl.load(in_ptr0 + (x2 + (16*y3)), xmask & ymask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tl.store(out_ptr0 + (y0 + (4*x2) + (64*y1)), tmp8, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/mf/cmffnm5icoqilxjyfgljttokrgpbt5d2oz7xydx6dxdlysmdthuj.py
# Topologically Sorted Source Nodes: [KV], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# KV => clone_1
# Graph fragment:
# %clone_1 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_3,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_3 = async_compile.triton('triton_poi_fused_clone_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = (xindex // 4) % 4
x2 = (xindex // 16) % 4
x3 = (xindex // 64)
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (4*x2) + (16*x1) + (64*x3)), xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + (x4), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kd/ckd25s7vujyv5hxcsgrxyfy25zilz32mxdkf3yx2dqm3ovvchilg.py
# Topologically Sorted Source Nodes: [queried_values, contiguous], Original ATen: [aten.mul, aten.clone]
# Source node to ATen node mapping:
# contiguous => clone_4
# queried_values => mul_8
# Graph fragment:
# %mul_8 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_11, 4), kwargs = {})
# %clone_4 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%mul_8,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_mul_4 = async_compile.triton('triton_poi_fused_clone_mul_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_mul_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_mul_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = (xindex // 4)
x5 = xindex
x0 = xindex % 4
x1 = (xindex // 4) % 4
x2 = (xindex // 16) % 4
x3 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x4), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (x5), xmask)
tmp1 = 1e-06
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 1, tl.int32)
tmp4 = tmp3 / tmp2
tmp5 = 1.0
tmp6 = tmp4 * tmp5
tmp8 = tmp6 * tmp7
tmp9 = 4.0
tmp10 = tmp8 * tmp9
tl.store(out_ptr0 + (x0 + (4*x2) + (16*x1) + (64*x3)), tmp10, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1, arg2_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg2_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [elu_1, K, sum_1], Original ATen: [aten.elu, aten.add, aten.sum]
stream0 = get_raw_stream(0)
triton_poi_fused_add_elu_sum_0.run(arg1_1, buf0, 64, grid=grid(64), stream=stream0)
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
buf6 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [einsum_1, einsum_2], Original ATen: [aten.clone]
triton_poi_fused_clone_1.run(arg0_1, buf1, buf6, 256, grid=grid(256), stream=stream0)
del arg0_1
buf2 = empty_strided_cuda((16, 4, 1), (4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [einsum_1], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf1, (16, 4, 4), (16, 4, 1), 0), reinterpret_tensor(buf0, (16, 4, 1), (4, 1, 1), 0), out=buf2)
del buf0
buf3 = reinterpret_tensor(buf1, (4, 4, 4, 4, 1), (64, 16, 4, 1, 1), 0); del buf1 # reuse
# Topologically Sorted Source Nodes: [KV], Original ATen: [aten.clone]
triton_poi_fused_clone_2.run(arg1_1, buf3, 16, 16, grid=grid(16, 16), stream=stream0)
del arg1_1
buf4 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch.float32)
# Topologically Sorted Source Nodes: [KV], Original ATen: [aten.clone]
triton_poi_fused_clone_3.run(arg2_1, buf4, 256, grid=grid(256), stream=stream0)
del arg2_1
buf5 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [KV], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 4), (16, 4, 1), 0), reinterpret_tensor(buf4, (16, 4, 4), (16, 4, 1), 0), out=buf5)
del buf3
buf7 = reinterpret_tensor(buf4, (16, 4, 4), (16, 4, 1), 0); del buf4 # reuse
# Topologically Sorted Source Nodes: [einsum_2], Original ATen: [aten.bmm]
extern_kernels.bmm(reinterpret_tensor(buf6, (16, 4, 4), (16, 4, 1), 0), buf5, out=buf7)
del buf5
buf8 = reinterpret_tensor(buf6, (4, 4, 4, 4), (64, 16, 4, 1), 0); del buf6 # reuse
# Topologically Sorted Source Nodes: [queried_values, contiguous], Original ATen: [aten.mul, aten.clone]
triton_poi_fused_clone_mul_4.run(buf2, buf7, buf8, 256, grid=grid(256), stream=stream0)
del buf2
del buf7
return (buf8, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg2_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1, arg2_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from torch.nn import Module
import torch
def elu_feature_map(x):
return torch.nn.functional.elu(x) + 1
class LinearAttention(Module):
def __init__(self, eps=1e-06):
super().__init__()
self.feature_map = elu_feature_map
self.eps = eps
def forward(self, queries, keys, values, q_mask=None, kv_mask=None):
""" Multi-Head linear attention proposed in "Transformers are RNNs"
Args:
queries: [N, L, H, D]
keys: [N, S, H, D]
values: [N, S, H, D]
q_mask: [N, L]
kv_mask: [N, S]
Returns:
queried_values: (N, L, H, D)
"""
Q = self.feature_map(queries)
K = self.feature_map(keys)
if q_mask is not None:
Q = Q * q_mask[:, :, None, None]
if kv_mask is not None:
K = K * kv_mask[:, :, None, None]
values = values * kv_mask[:, :, None, None]
v_length = values.size(1)
values = values / v_length
KV = torch.einsum('nshd,nshv->nhdv', K, values)
Z = 1 / (torch.einsum('nlhd,nhd->nlh', Q, K.sum(dim=1)) + self.eps)
queried_values = torch.einsum('nlhd,nhdv,nlh->nlhv', Q, KV, Z
) * v_length
return queried_values.contiguous()
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4]), torch.rand(
[4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch.nn import Module
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_add_elu_sum_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 16
x1 = xindex // 16
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 64 * x1), xmask)
tmp9 = tl.load(in_ptr0 + (16 + x0 + 64 * x1), xmask)
tmp17 = tl.load(in_ptr0 + (32 + x0 + 64 * x1), xmask)
tmp25 = tl.load(in_ptr0 + (48 + x0 + 64 * x1), xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tmp10 = tmp9 > tmp1
tmp11 = tmp9 * tmp3
tmp12 = libdevice.expm1(tmp11)
tmp13 = tmp12 * tmp3
tmp14 = tl.where(tmp10, tmp11, tmp13)
tmp15 = tmp14 + tmp3
tmp16 = tmp8 + tmp15
tmp18 = tmp17 > tmp1
tmp19 = tmp17 * tmp3
tmp20 = libdevice.expm1(tmp19)
tmp21 = tmp20 * tmp3
tmp22 = tl.where(tmp18, tmp19, tmp21)
tmp23 = tmp22 + tmp3
tmp24 = tmp16 + tmp23
tmp26 = tmp25 > tmp1
tmp27 = tmp25 * tmp3
tmp28 = libdevice.expm1(tmp27)
tmp29 = tmp28 * tmp3
tmp30 = tl.where(tmp26, tmp27, tmp29)
tmp31 = tmp30 + tmp3
tmp32 = tmp24 + tmp31
tl.store(out_ptr0 + x2, tmp32, xmask)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, out_ptr1, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x0 = xindex % 4
x1 = xindex // 4 % 4
x2 = xindex // 16 % 4
x3 = xindex // 64
tmp0 = tl.load(in_ptr0 + x4, xmask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tl.store(out_ptr0 + (x0 + 4 * x2 + 16 * x1 + 64 * x3), tmp8, xmask)
tl.store(out_ptr1 + (x0 + 4 * x2 + 16 * x1 + 64 * x3), tmp8, xmask)
@triton.jit
def triton_poi_fused_clone_2(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x2 = xindex
y3 = yindex
y0 = yindex % 4
y1 = yindex // 4
tmp0 = tl.load(in_ptr0 + (x2 + 16 * y3), xmask & ymask)
tmp1 = 0.0
tmp2 = tmp0 > tmp1
tmp3 = 1.0
tmp4 = tmp0 * tmp3
tmp5 = libdevice.expm1(tmp4)
tmp6 = tmp5 * tmp3
tmp7 = tl.where(tmp2, tmp4, tmp6)
tmp8 = tmp7 + tmp3
tl.store(out_ptr0 + (y0 + 4 * x2 + 64 * y1), tmp8, xmask & ymask)
@triton.jit
def triton_poi_fused_clone_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = xindex // 4 % 4
x2 = xindex // 16 % 4
x3 = xindex // 64
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 4 * x2 + 16 * x1 + 64 * x3), xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tl.store(out_ptr0 + x4, tmp2, xmask)
@triton.jit
def triton_poi_fused_clone_mul_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex // 4
x5 = xindex
x0 = xindex % 4
x1 = xindex // 4 % 4
x2 = xindex // 16 % 4
x3 = xindex // 64
tmp0 = tl.load(in_ptr0 + x4, xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + x5, xmask)
tmp1 = 1e-06
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 1, tl.int32)
tmp4 = tmp3 / tmp2
tmp5 = 1.0
tmp6 = tmp4 * tmp5
tmp8 = tmp6 * tmp7
tmp9 = 4.0
tmp10 = tmp8 * tmp9
tl.store(out_ptr0 + (x0 + 4 * x2 + 16 * x1 + 64 * x3), tmp10, xmask)
def call(args):
arg0_1, arg1_1, arg2_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg2_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_elu_sum_0[grid(64)](arg1_1, buf0, 64, XBLOCK=
64, num_warps=1, num_stages=1)
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
buf6 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch
.float32)
triton_poi_fused_clone_1[grid(256)](arg0_1, buf1, buf6, 256, XBLOCK
=128, num_warps=4, num_stages=1)
del arg0_1
buf2 = empty_strided_cuda((16, 4, 1), (4, 1, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf1, (16, 4, 4), (16, 4, 1),
0), reinterpret_tensor(buf0, (16, 4, 1), (4, 1, 1), 0), out=buf2)
del buf0
buf3 = reinterpret_tensor(buf1, (4, 4, 4, 4, 1), (64, 16, 4, 1, 1), 0)
del buf1
triton_poi_fused_clone_2[grid(16, 16)](arg1_1, buf3, 16, 16, XBLOCK
=16, YBLOCK=16, num_warps=4, num_stages=1)
del arg1_1
buf4 = empty_strided_cuda((4, 4, 4, 4, 1), (64, 16, 4, 1, 1), torch
.float32)
triton_poi_fused_clone_3[grid(256)](arg2_1, buf4, 256, XBLOCK=256,
num_warps=4, num_stages=1)
del arg2_1
buf5 = empty_strided_cuda((16, 4, 4), (16, 4, 1), torch.float32)
extern_kernels.bmm(reinterpret_tensor(buf3, (16, 4, 4), (16, 4, 1),
0), reinterpret_tensor(buf4, (16, 4, 4), (16, 4, 1), 0), out=buf5)
del buf3
buf7 = reinterpret_tensor(buf4, (16, 4, 4), (16, 4, 1), 0)
del buf4
extern_kernels.bmm(reinterpret_tensor(buf6, (16, 4, 4), (16, 4, 1),
0), buf5, out=buf7)
del buf5
buf8 = reinterpret_tensor(buf6, (4, 4, 4, 4), (64, 16, 4, 1), 0)
del buf6
triton_poi_fused_clone_mul_4[grid(256)](buf2, buf7, buf8, 256,
XBLOCK=128, num_warps=4, num_stages=1)
del buf2
del buf7
return buf8,
def elu_feature_map(x):
return torch.nn.functional.elu(x) + 1
class LinearAttentionNew(Module):
def __init__(self, eps=1e-06):
super().__init__()
self.feature_map = elu_feature_map
self.eps = eps
def forward(self, input_0, input_1, input_2):
arg0_1 = input_0
arg1_1 = input_1
arg2_1 = input_2
output = call([arg0_1, arg1_1, arg2_1])
return output[0]
|
francescodisalvo05/LoFTR
|
LinearAttention
| false | 12,406 |
[
"Apache-2.0"
] | 0 |
66372ebbe1ea97d57fe6cb8b5acf5cd92a87ef8d
|
https://github.com/francescodisalvo05/LoFTR/tree/66372ebbe1ea97d57fe6cb8b5acf5cd92a87ef8d
|
BPRLoss
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/7v/c7vgqnqd6tfrddpue57u7sh2gdpwg4q2fzz35hfbkaxbzaahdbrb.py
# Topologically Sorted Source Nodes: [sub, sigmoid, add, log, loss, mean], Original ATen: [aten.sub, aten.sigmoid, aten.add, aten.log, aten.neg, aten.mean]
# Source node to ATen node mapping:
# add => add
# log => log
# loss => neg
# mean => mean
# sigmoid => sigmoid
# sub => sub
# Graph fragment:
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg0_1, %arg1_1), kwargs = {})
# %sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%sub,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sigmoid, 1e-10), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%add,), kwargs = {})
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%log,), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%neg,), kwargs = {})
triton_per_fused_add_log_mean_neg_sigmoid_sub_0 = async_compile.triton('triton_per_fused_add_log_mean_neg_sigmoid_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_log_mean_neg_sigmoid_sub_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_log_mean_neg_sigmoid_sub_0(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp1 = tl.load(in_ptr1 + (r0), None)
tmp2 = tmp0 - tmp1
tmp3 = tl.sigmoid(tmp2)
tmp4 = 1e-10
tmp5 = tmp3 + tmp4
tmp6 = tl_math.log(tmp5)
tmp7 = -tmp6
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = 256.0
tmp12 = tmp10 / tmp11
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp12, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [sub, sigmoid, add, log, loss, mean], Original ATen: [aten.sub, aten.sigmoid, aten.add, aten.log, aten.neg, aten.mean]
stream0 = get_raw_stream(0)
triton_per_fused_add_log_mean_neg_sigmoid_sub_0.run(buf1, arg0_1, arg1_1, 1, 256, grid=grid(1), stream=stream0)
del arg0_1
del arg1_1
return (buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn.modules.loss import *
from torch.nn.modules import *
from torch.optim import *
from torch.optim.lr_scheduler import *
import torch.backends
class PairwiseLoss(nn.Module):
"""Base class for pairwise loss functions.
Pairwise approached looks at a pair of documents at a time in the loss function.
Given a pair of documents the algorithm try to come up with the optimal ordering
For that pair and compare it with the ground truth. The goal for the ranker is to
minimize the number of inversions in ranker.
Input space: pairs of documents (d1, d2)
Output space: preferences (yes/no) for a given doc. pair
"""
@staticmethod
def _assert_equal_size(positive_score: 'torch.Tensor', negative_score:
'torch.Tensor') ->None:
if positive_score.size() != negative_score.size():
raise ValueError(
f'Shape mismatch: {positive_score.size()}, {negative_score.size()}'
)
def forward(self, positive_score: 'torch.Tensor', negative_score:
'torch.Tensor'):
raise NotImplementedError()
class BPRLoss(PairwiseLoss):
"""Bayesian Personalised Ranking loss function.
It has been proposed in `BPRLoss\\: Bayesian Personalized Ranking from Implicit Feedback`_.
.. _BPRLoss\\: Bayesian Personalized Ranking from Implicit Feedback:
https://arxiv.org/pdf/1205.2618.pdf
Args:
gamma (float): Small value to avoid division by zero. Default: ``1e-10``.
Example:
.. code-block:: python
import torch
from catalyst.contrib.nn.criterion import recsys
pos_score = torch.randn(3, requires_grad=True)
neg_score = torch.randn(3, requires_grad=True)
output = recsys.BPRLoss()(pos_score, neg_score)
output.backward()
"""
def __init__(self, gamma=1e-10) ->None:
super().__init__()
self.gamma = gamma
def forward(self, positive_score: 'torch.Tensor', negative_score:
'torch.Tensor') ->torch.Tensor:
"""Forward propagation method for the BPR loss.
Args:
positive_score: Tensor containing predictions for known positive items.
negative_score: Tensor containing predictions for sampled negative items.
Returns:
computed loss
"""
self._assert_equal_size(positive_score, negative_score)
loss = -torch.log(self.gamma + torch.sigmoid(positive_score -
negative_score))
return loss.mean()
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import math as tl_math
from torch import nn
from torch.nn.modules.loss import *
from torch.nn.modules import *
from torch.optim import *
from torch.optim.lr_scheduler import *
import torch.backends
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_add_log_mean_neg_sigmoid_sub_0(in_out_ptr0, in_ptr0,
in_ptr1, xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp1 = tl.load(in_ptr1 + r0, None)
tmp2 = tmp0 - tmp1
tmp3 = tl.sigmoid(tmp2)
tmp4 = 1e-10
tmp5 = tmp3 + tmp4
tmp6 = tl_math.log(tmp5)
tmp7 = -tmp6
tmp8 = tl.broadcast_to(tmp7, [RBLOCK])
tmp10 = triton_helpers.promote_to_tensor(tl.sum(tmp8, 0))
tmp11 = 256.0
tmp12 = tmp10 / tmp11
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp12, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0
del buf0
get_raw_stream(0)
triton_per_fused_add_log_mean_neg_sigmoid_sub_0[grid(1)](buf1,
arg0_1, arg1_1, 1, 256, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
return buf1,
class PairwiseLoss(nn.Module):
"""Base class for pairwise loss functions.
Pairwise approached looks at a pair of documents at a time in the loss function.
Given a pair of documents the algorithm try to come up with the optimal ordering
For that pair and compare it with the ground truth. The goal for the ranker is to
minimize the number of inversions in ranker.
Input space: pairs of documents (d1, d2)
Output space: preferences (yes/no) for a given doc. pair
"""
@staticmethod
def _assert_equal_size(positive_score: 'torch.Tensor', negative_score:
'torch.Tensor') ->None:
if positive_score.size() != negative_score.size():
raise ValueError(
f'Shape mismatch: {positive_score.size()}, {negative_score.size()}'
)
def forward(self, positive_score: 'torch.Tensor', negative_score:
'torch.Tensor'):
raise NotImplementedError()
class BPRLossNew(PairwiseLoss):
"""Bayesian Personalised Ranking loss function.
It has been proposed in `BPRLoss\\: Bayesian Personalized Ranking from Implicit Feedback`_.
.. _BPRLoss\\: Bayesian Personalized Ranking from Implicit Feedback:
https://arxiv.org/pdf/1205.2618.pdf
Args:
gamma (float): Small value to avoid division by zero. Default: ``1e-10``.
Example:
.. code-block:: python
import torch
from catalyst.contrib.nn.criterion import recsys
pos_score = torch.randn(3, requires_grad=True)
neg_score = torch.randn(3, requires_grad=True)
output = recsys.BPRLoss()(pos_score, neg_score)
output.backward()
"""
def __init__(self, gamma=1e-10) ->None:
super().__init__()
self.gamma = gamma
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
gavrin-s/catalyst
|
BPRLoss
| false | 12,407 |
[
"Apache-2.0"
] | 0 |
81087d8348b359e501d899f7a8350e0bedfc2b7d
|
https://github.com/gavrin-s/catalyst/tree/81087d8348b359e501d899f7a8350e0bedfc2b7d
|
HuEtAl
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ei/ceiznlk6kg5oe2livq5fqchenhep4cjvqjwxh5pt3sqi5ybf5ncz.py
# Topologically Sorted Source Nodes: [x_2, max_pool1d, x_3], Original ATen: [aten.convolution, aten.max_pool2d_with_indices, aten.tanh]
# Source node to ATen node mapping:
# max_pool1d => getitem_1
# x_2 => convolution
# x_3 => tanh
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%unsqueeze, %primals_2, %primals_3, [1], [0], [1], False, [0], 1), kwargs = {})
# %getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%_low_memory_max_pool2d_with_offsets, 1), kwargs = {})
# %tanh : [num_users=2] = call_function[target=torch.ops.aten.tanh.default](args = (%squeeze_2,), kwargs = {})
triton_poi_fused_convolution_max_pool2d_with_indices_tanh_0 = async_compile.triton('triton_poi_fused_convolution_max_pool2d_with_indices_tanh_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[512],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i8', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_max_pool2d_with_indices_tanh_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_max_pool2d_with_indices_tanh_0(in_out_ptr0, in_ptr0, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 320
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 4) % 20
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int8)
tmp4 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
tl.store(out_ptr0 + (x3), tmp3, xmask)
tl.store(out_ptr1 + (x3), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/yy/cyy2iyhelbruzgtpkuv4lif6qt3iicbck2w7hamal2vbsoea75do.py
# Topologically Sorted Source Nodes: [x_5], Original ATen: [aten.tanh]
# Source node to ATen node mapping:
# x_5 => tanh_1
# Graph fragment:
# %add_tensor : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mm_default, %primals_5), kwargs = {})
# %tanh_1 : [num_users=2] = call_function[target=torch.ops.aten.tanh.default](args = (%add_tensor,), kwargs = {})
triton_poi_fused_tanh_1 = async_compile.triton('triton_poi_fused_tanh_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[512],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_tanh_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_tanh_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 400
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 100
tmp0 = tl.load(in_out_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + (x2), tmp3, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7 = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (20, 1, 1), (1, 1, 1))
assert_size_stride(primals_3, (20, ), (1, ))
assert_size_stride(primals_4, (100, 80), (80, 1))
assert_size_stride(primals_5, (100, ), (1, ))
assert_size_stride(primals_6, (4, 100), (100, 1))
assert_size_stride(primals_7, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [x_2], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(reinterpret_tensor(primals_1, (4, 1, 4), (4, 4, 1), 0), primals_2, stride=(1,), padding=(0,), dilation=(1,), transposed=False, output_padding=(0,), groups=1, bias=None)
assert_size_stride(buf0, (4, 20, 4), (80, 4, 1))
buf1 = buf0; del buf0 # reuse
buf2 = empty_strided_cuda((4, 20, 1, 4), (80, 4, 4, 1), torch.int8)
buf3 = empty_strided_cuda((4, 20, 4), (80, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_2, max_pool1d, x_3], Original ATen: [aten.convolution, aten.max_pool2d_with_indices, aten.tanh]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_max_pool2d_with_indices_tanh_0.run(buf1, primals_3, buf2, buf3, 320, grid=grid(320), stream=stream0)
del primals_3
buf4 = empty_strided_cuda((4, 100), (100, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(buf3, (4, 80), (80, 1), 0), reinterpret_tensor(primals_4, (80, 100), (1, 80), 0), out=buf4)
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [x_5], Original ATen: [aten.tanh]
triton_poi_fused_tanh_1.run(buf5, primals_5, 400, grid=grid(400), stream=stream0)
del primals_5
buf6 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_6], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_7, buf5, reinterpret_tensor(primals_6, (100, 4), (1, 100), 0), alpha=1, beta=1, out=buf6)
del primals_7
return (buf6, primals_2, reinterpret_tensor(primals_1, (4, 1, 4), (4, 4, 1), 0), reinterpret_tensor(buf1, (4, 20, 1, 4), (80, 4, 4, 1), 0), buf2, buf3, buf5, primals_6, primals_4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((20, 1, 1), (1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((20, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((100, 80), (80, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((100, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 100), (100, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import math
import torch
import torch.utils
import torch.utils.data
import torch.nn as nn
from torch.nn import init
class HuEtAl(nn.Module):
"""
Deep Convolutional Neural Networks for Hyperspectral Image Classification
Wei Hu, Yangyu Huang, Li Wei, Fan Zhang and Hengchao Li
Journal of Sensors, Volume 2015 (2015)
https://www.hindawi.com/journals/js/2015/258619/
"""
@staticmethod
def weight_init(m):
if isinstance(m, nn.Linear) or isinstance(m, nn.Conv1d):
init.uniform_(m.weight, -0.05, 0.05)
init.zeros_(m.bias)
def _get_final_flattened_size(self):
with torch.no_grad():
x = torch.zeros(1, 1, self.input_channels)
x = self.pool(self.conv(x))
return x.numel()
def __init__(self, input_channels, n_classes, kernel_size=None,
pool_size=None):
super(HuEtAl, self).__init__()
if kernel_size is None:
kernel_size = math.ceil(input_channels / 9)
if pool_size is None:
pool_size = math.ceil(kernel_size / 5)
self.input_channels = input_channels
self.conv = nn.Conv1d(1, 20, kernel_size)
self.pool = nn.MaxPool1d(pool_size)
self.features_size = self._get_final_flattened_size()
self.fc1 = nn.Linear(self.features_size, 100)
self.fc2 = nn.Linear(100, n_classes)
self.apply(self.weight_init)
def forward(self, x):
x = x.squeeze(dim=-1).squeeze(dim=-1)
x = x.unsqueeze(1)
x = self.conv(x)
x = torch.tanh(self.pool(x))
x = x.view(-1, self.features_size)
x = torch.tanh(self.fc1(x))
x = self.fc2(x)
return x
def get_inputs():
return [torch.rand([4, 4])]
def get_init_inputs():
return [[], {'input_channels': 4, 'n_classes': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import math
import torch.utils
import torch.utils.data
import torch.nn as nn
from torch.nn import init
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_convolution_max_pool2d_with_indices_tanh_0(in_out_ptr0,
in_ptr0, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 320
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 4 % 20
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int8)
tmp4 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x3, tmp2, xmask)
tl.store(out_ptr0 + x3, tmp3, xmask)
tl.store(out_ptr1 + x3, tmp4, xmask)
@triton.jit
def triton_poi_fused_tanh_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl.constexpr
):
xnumel = 400
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 100
tmp0 = tl.load(in_out_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = libdevice.tanh(tmp2)
tl.store(in_out_ptr0 + x2, tmp3, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7) = args
args.clear()
assert_size_stride(primals_1, (4, 4), (4, 1))
assert_size_stride(primals_2, (20, 1, 1), (1, 1, 1))
assert_size_stride(primals_3, (20,), (1,))
assert_size_stride(primals_4, (100, 80), (80, 1))
assert_size_stride(primals_5, (100,), (1,))
assert_size_stride(primals_6, (4, 100), (100, 1))
assert_size_stride(primals_7, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(reinterpret_tensor(primals_1, (4,
1, 4), (4, 4, 1), 0), primals_2, stride=(1,), padding=(0,),
dilation=(1,), transposed=False, output_padding=(0,), groups=1,
bias=None)
assert_size_stride(buf0, (4, 20, 4), (80, 4, 1))
buf1 = buf0
del buf0
buf2 = empty_strided_cuda((4, 20, 1, 4), (80, 4, 4, 1), torch.int8)
buf3 = empty_strided_cuda((4, 20, 4), (80, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_convolution_max_pool2d_with_indices_tanh_0[grid(320)](
buf1, primals_3, buf2, buf3, 320, XBLOCK=256, num_warps=4,
num_stages=1)
del primals_3
buf4 = empty_strided_cuda((4, 100), (100, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf3, (4, 80), (80, 1), 0),
reinterpret_tensor(primals_4, (80, 100), (1, 80), 0), out=buf4)
buf5 = buf4
del buf4
triton_poi_fused_tanh_1[grid(400)](buf5, primals_5, 400, XBLOCK=128,
num_warps=4, num_stages=1)
del primals_5
buf6 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_7, buf5, reinterpret_tensor(primals_6,
(100, 4), (1, 100), 0), alpha=1, beta=1, out=buf6)
del primals_7
return buf6, primals_2, reinterpret_tensor(primals_1, (4, 1, 4), (4, 4,
1), 0), reinterpret_tensor(buf1, (4, 20, 1, 4), (80, 4, 4, 1), 0
), buf2, buf3, buf5, primals_6, primals_4
class HuEtAlNew(nn.Module):
"""
Deep Convolutional Neural Networks for Hyperspectral Image Classification
Wei Hu, Yangyu Huang, Li Wei, Fan Zhang and Hengchao Li
Journal of Sensors, Volume 2015 (2015)
https://www.hindawi.com/journals/js/2015/258619/
"""
@staticmethod
def weight_init(m):
if isinstance(m, nn.Linear) or isinstance(m, nn.Conv1d):
init.uniform_(m.weight, -0.05, 0.05)
init.zeros_(m.bias)
def _get_final_flattened_size(self):
with torch.no_grad():
x = torch.zeros(1, 1, self.input_channels)
x = self.pool(self.conv(x))
return x.numel()
def __init__(self, input_channels, n_classes, kernel_size=None,
pool_size=None):
super(HuEtAlNew, self).__init__()
if kernel_size is None:
kernel_size = math.ceil(input_channels / 9)
if pool_size is None:
pool_size = math.ceil(kernel_size / 5)
self.input_channels = input_channels
self.conv = nn.Conv1d(1, 20, kernel_size)
self.pool = nn.MaxPool1d(pool_size)
self.features_size = self._get_final_flattened_size()
self.fc1 = nn.Linear(self.features_size, 100)
self.fc2 = nn.Linear(100, n_classes)
self.apply(self.weight_init)
def forward(self, input_0):
primals_2 = self.conv.weight
primals_3 = self.conv.bias
primals_4 = self.fc1.weight
primals_5 = self.fc1.bias
primals_6 = self.fc2.weight
primals_7 = self.fc2.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7])
return output[0]
|
dikers/DeepHyper
|
HuEtAl
| false | 12,408 |
[
"Apache-2.0"
] | 0 |
827a8f3077e18b71cf448a2e56e49670428b1bfd
|
https://github.com/dikers/DeepHyper/tree/827a8f3077e18b71cf448a2e56e49670428b1bfd
|
Lookahead
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/xf/cxfjorpj7nzakemkplfiue4lxn3wssubujqanweih6rojbuoolb7.py
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.constant_pad_nd]
# Source node to ATen node mapping:
# x_1 => constant_pad_nd
# Graph fragment:
# %constant_pad_nd : [num_users=2] = call_function[target=torch.ops.aten.constant_pad_nd.default](args = (%permute_1, [0, 3], 0.0), kwargs = {})
triton_poi_fused_constant_pad_nd_0 = async_compile.triton('triton_poi_fused_constant_pad_nd_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16, 8], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_constant_pad_nd_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_constant_pad_nd_0(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 16
xnumel = 7
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x1 = xindex
y0 = yindex
tmp0 = x1
tmp1 = tl.full([1, 1], 4, tl.int64)
tmp2 = tmp0 < tmp1
tmp3 = tl.load(in_ptr0 + (y0 + (16*x1)), tmp2 & xmask & ymask, eviction_policy='evict_last', other=0.0)
tl.store(out_ptr0 + (x1 + (7*y0)), tmp3, xmask & ymask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/r7/cr7evfztwmdzfk7dfomwxcc4g457gqiibpdthzqr2y7vyuw4ptpg.py
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.clone]
# Source node to ATen node mapping:
# x_3 => clone
# Graph fragment:
# %clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%permute_3,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_clone_1 = async_compile.triton('triton_poi_fused_clone_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4, 16], tile_hint=TileHint.SQUARE,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clone_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK : tl.constexpr, XBLOCK : tl.constexpr):
ynumel = 4
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x1 = xindex
y0 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + (4*x1)), xmask & ymask, eviction_policy='evict_last')
tl.store(out_ptr0 + (x1 + (16*y0)), tmp0, xmask & ymask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 1, 4), (4, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 7), (28, 7, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.constant_pad_nd]
stream0 = get_raw_stream(0)
triton_poi_fused_constant_pad_nd_0.run(primals_1, buf0, 16, 7, grid=grid(16, 7), stream=stream0)
del primals_1
# Topologically Sorted Source Nodes: [x_2], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(buf0, primals_2, stride=(1,), padding=(0,), dilation=(1,), transposed=False, output_padding=(0,), groups=4, bias=None)
assert_size_stride(buf1, (4, 4, 4), (16, 4, 1))
buf2 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.clone]
triton_poi_fused_clone_1.run(buf1, buf2, 4, 16, grid=grid(4, 16), stream=stream0)
del buf1
return (buf2, primals_2, buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, 1, 4), (4, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.utils.data.distributed
import torch.nn as nn
import torch.nn.functional as F
class Lookahead(nn.Module):
def __init__(self, n_features, context):
super(Lookahead, self).__init__()
assert context > 0
self.context = context
self.n_features = n_features
self.pad = 0, self.context - 1
self.conv = nn.Conv1d(self.n_features, self.n_features, kernel_size
=self.context, stride=1, groups=self.n_features, padding=0,
bias=None)
def forward(self, x):
x = x.transpose(0, 1).transpose(1, 2)
x = F.pad(x, pad=self.pad, value=0)
x = self.conv(x)
x = x.transpose(1, 2).transpose(0, 1).contiguous()
return x
def __repr__(self):
return self.__class__.__name__ + '(' + 'n_features=' + str(self.
n_features) + ', context=' + str(self.context) + ')'
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'n_features': 4, 'context': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.utils.data.distributed
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_constant_pad_nd_0(in_ptr0, out_ptr0, ynumel, xnumel,
YBLOCK: tl.constexpr, XBLOCK: tl.constexpr):
ynumel = 16
xnumel = 7
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x1 = xindex
y0 = yindex
tmp0 = x1
tmp1 = tl.full([1, 1], 4, tl.int64)
tmp2 = tmp0 < tmp1
tmp3 = tl.load(in_ptr0 + (y0 + 16 * x1), tmp2 & xmask & ymask,
eviction_policy='evict_last', other=0.0)
tl.store(out_ptr0 + (x1 + 7 * y0), tmp3, xmask & ymask)
@triton.jit
def triton_poi_fused_clone_1(in_ptr0, out_ptr0, ynumel, xnumel, YBLOCK: tl.
constexpr, XBLOCK: tl.constexpr):
ynumel = 4
xnumel = 16
yoffset = tl.program_id(1) * YBLOCK
yindex = yoffset + tl.arange(0, YBLOCK)[None, :]
ymask = yindex < ynumel
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = xindex < xnumel
x1 = xindex
y0 = yindex
tmp0 = tl.load(in_ptr0 + (y0 + 4 * x1), xmask & ymask, eviction_policy=
'evict_last')
tl.store(out_ptr0 + (x1 + 16 * y0), tmp0, xmask & ymask)
def call(args):
primals_1, primals_2 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (4, 1, 4), (4, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 7), (28, 7, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_constant_pad_nd_0[grid(16, 7)](primals_1, buf0, 16,
7, XBLOCK=8, YBLOCK=16, num_warps=4, num_stages=1)
del primals_1
buf1 = extern_kernels.convolution(buf0, primals_2, stride=(1,),
padding=(0,), dilation=(1,), transposed=False, output_padding=(
0,), groups=4, bias=None)
assert_size_stride(buf1, (4, 4, 4), (16, 4, 1))
buf2 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_clone_1[grid(4, 16)](buf1, buf2, 4, 16, XBLOCK=16,
YBLOCK=4, num_warps=1, num_stages=1)
del buf1
return buf2, primals_2, buf0
class LookaheadNew(nn.Module):
def __init__(self, n_features, context):
super(LookaheadNew, self).__init__()
assert context > 0
self.context = context
self.n_features = n_features
self.pad = 0, self.context - 1
self.conv = nn.Conv1d(self.n_features, self.n_features, kernel_size
=self.context, stride=1, groups=self.n_features, padding=0,
bias=None)
def __repr__(self):
return self.__class__.__name__ + '(' + 'n_features=' + str(self.
n_features) + ', context=' + str(self.context) + ')'
def forward(self, input_0):
primals_2 = self.conv.weight
primals_1 = input_0
output = call([primals_1, primals_2])
return output[0]
|
gbaril/End-to-end-E2E-Named-Entity-Recognition-from-English-Speech
|
Lookahead
| false | 12,409 |
[
"Apache-2.0"
] | 0 |
9760a4ec3ba1c55bb4740c12c4542f13dd028695
|
https://github.com/gbaril/End-to-end-E2E-Named-Entity-Recognition-from-English-Speech/tree/9760a4ec3ba1c55bb4740c12c4542f13dd028695
|
DenseBlock
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/gk/cgkw7mtnlzc2x4r4dlaxhd27twdsy4zp5nunyiqlzxzxlut26kgd.py
# Topologically Sorted Source Nodes: [conv1d], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# conv1d => convolution
# Graph fragment:
# %convolution : [num_users=2] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1], [2], [2], False, [0], 1), kwargs = {})
triton_poi_fused_convolution_0 = async_compile.triton('triton_poi_fused_convolution_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 96
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 6) % 4
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/n4/cn4nrao6ng2dt25fyig6ikjf3xz5iqxb22ofakq7pljb2sg3u6rd.py
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# out => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%primals_3, %mul], 1), kwargs = {})
triton_poi_fused_cat_1 = async_compile.triton('triton_poi_fused_cat_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[128],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 3, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_1(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = (xindex // 4) % 8
x0 = xindex % 4
x2 = (xindex // 32)
x3 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (4*x1) + (16*x2)), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + (x0 + (6*((-4) + x1)) + (24*x2)), tmp6 & xmask, other=0.0)
tmp10 = libdevice.tanh(tmp9)
tmp11 = tl.load(in_ptr2 + (x0 + (6*((-4) + x1)) + (24*x2)), tmp6 & xmask, other=0.0)
tmp12 = tl.sigmoid(tmp11)
tmp13 = tmp10 * tmp12
tmp14 = tl.full(tmp13.shape, 0.0, tmp13.dtype)
tmp15 = tl.where(tmp6, tmp13, tmp14)
tmp16 = tl.where(tmp4, tmp5, tmp15)
tl.store(out_ptr0 + (x3), tmp16, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 2), (8, 2, 1))
assert_size_stride(primals_2, (4, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4, 2), (8, 2, 1))
assert_size_stride(primals_5, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv1d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,), padding=(2,), dilation=(2,), transposed=False, output_padding=(0,), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 6), (24, 6, 1))
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [conv1d], Original ATen: [aten.convolution]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_0.run(buf1, primals_2, 96, grid=grid(96), stream=stream0)
del primals_2
# Topologically Sorted Source Nodes: [conv1d_1], Original ATen: [aten.convolution]
buf2 = extern_kernels.convolution(primals_3, primals_4, stride=(1,), padding=(2,), dilation=(2,), transposed=False, output_padding=(0,), groups=1, bias=None)
assert_size_stride(buf2, (4, 4, 6), (24, 6, 1))
buf3 = buf2; del buf2 # reuse
# Topologically Sorted Source Nodes: [conv1d_1], Original ATen: [aten.convolution]
triton_poi_fused_convolution_0.run(buf3, primals_5, 96, grid=grid(96), stream=stream0)
del primals_5
buf4 = empty_strided_cuda((4, 8, 4), (32, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [out], Original ATen: [aten.cat]
triton_poi_fused_cat_1.run(primals_3, buf1, buf3, buf4, 128, grid=grid(128), stream=stream0)
return (buf4, primals_1, primals_3, primals_4, buf1, buf3, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 2), (8, 2, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 4, 2), (8, 2, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn import functional as F
class CausalConv1d(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size=2, dilation=2):
super(CausalConv1d, self).__init__()
self.padding = dilation
self.causal_conv = nn.Conv1d(in_channels, out_channels, kernel_size,
padding=self.padding, dilation=dilation)
def forward(self, minibatch):
return self.causal_conv(minibatch)[:, :, :-self.padding]
class DenseBlock(nn.Module):
def __init__(self, in_channels, filters, dilation=2):
super(DenseBlock, self).__init__()
self.causal_conv1 = CausalConv1d(in_channels, filters, dilation=
dilation)
self.causal_conv2 = CausalConv1d(in_channels, filters, dilation=
dilation)
def forward(self, minibatch):
tanh = F.tanh(self.causal_conv1(minibatch))
sig = F.sigmoid(self.causal_conv2(minibatch))
out = torch.cat([minibatch, tanh * sig], dim=1)
return out
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'in_channels': 4, 'filters': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_convolution_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 96
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 6 % 4
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
@triton.jit
def triton_poi_fused_cat_1(in_ptr0, in_ptr1, in_ptr2, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 128
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x1 = xindex // 4 % 8
x0 = xindex % 4
x2 = xindex // 32
x3 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 4 * x1 + 16 * x2), tmp4 & xmask, other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (x0 + 6 * (-4 + x1) + 24 * x2), tmp6 & xmask,
other=0.0)
tmp10 = libdevice.tanh(tmp9)
tmp11 = tl.load(in_ptr2 + (x0 + 6 * (-4 + x1) + 24 * x2), tmp6 & xmask,
other=0.0)
tmp12 = tl.sigmoid(tmp11)
tmp13 = tmp10 * tmp12
tmp14 = tl.full(tmp13.shape, 0.0, tmp13.dtype)
tmp15 = tl.where(tmp6, tmp13, tmp14)
tmp16 = tl.where(tmp4, tmp5, tmp15)
tl.store(out_ptr0 + x3, tmp16, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 2), (8, 2, 1))
assert_size_stride(primals_2, (4,), (1,))
assert_size_stride(primals_3, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_4, (4, 4, 2), (8, 2, 1))
assert_size_stride(primals_5, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,),
padding=(2,), dilation=(2,), transposed=False, output_padding=(
0,), groups=1, bias=None)
assert_size_stride(buf0, (4, 4, 6), (24, 6, 1))
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_convolution_0[grid(96)](buf1, primals_2, 96,
XBLOCK=128, num_warps=4, num_stages=1)
del primals_2
buf2 = extern_kernels.convolution(primals_3, primals_4, stride=(1,),
padding=(2,), dilation=(2,), transposed=False, output_padding=(
0,), groups=1, bias=None)
assert_size_stride(buf2, (4, 4, 6), (24, 6, 1))
buf3 = buf2
del buf2
triton_poi_fused_convolution_0[grid(96)](buf3, primals_5, 96,
XBLOCK=128, num_warps=4, num_stages=1)
del primals_5
buf4 = empty_strided_cuda((4, 8, 4), (32, 4, 1), torch.float32)
triton_poi_fused_cat_1[grid(128)](primals_3, buf1, buf3, buf4, 128,
XBLOCK=128, num_warps=4, num_stages=1)
return buf4, primals_1, primals_3, primals_4, buf1, buf3
class CausalConv1d(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size=2, dilation=2):
super(CausalConv1d, self).__init__()
self.padding = dilation
self.causal_conv = nn.Conv1d(in_channels, out_channels, kernel_size,
padding=self.padding, dilation=dilation)
def forward(self, minibatch):
return self.causal_conv(minibatch)[:, :, :-self.padding]
class DenseBlockNew(nn.Module):
def __init__(self, in_channels, filters, dilation=2):
super(DenseBlockNew, self).__init__()
self.causal_conv1 = CausalConv1d(in_channels, filters, dilation=
dilation)
self.causal_conv2 = CausalConv1d(in_channels, filters, dilation=
dilation)
def forward(self, input_0):
primals_1 = self.causal_conv1.causal_conv.weight
primals_2 = self.causal_conv1.causal_conv.bias
primals_4 = self.causal_conv2.causal_conv.weight
primals_5 = self.causal_conv2.causal_conv.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4, primals_5])
return output[0]
|
gaetangate/FewRel
|
DenseBlock
| false | 12,410 |
[
"MIT"
] | 0 |
150199d1060571315b1f370b3b3352d7a7c72dd5
|
https://github.com/gaetangate/FewRel/tree/150199d1060571315b1f370b3b3352d7a7c72dd5
|
MinibatchStd
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/2z/c2ztcux5cffvqjv632chxjfgjvsh5kiy2j3qznrskvufcfchphso.py
# Topologically Sorted Source Nodes: [mean, y_2, x], Original ATen: [aten.mean, aten.sub, aten.cat]
# Source node to ATen node mapping:
# mean => mean
# x => cat
# y_2 => sub
# Graph fragment:
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%view, [0], True), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view, %mean), kwargs = {})
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%view_1, %expand], 1), kwargs = {})
# %copy_ : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%arg0_1, %view_1), kwargs = {})
triton_poi_fused_cat_mean_sub_0 = async_compile.triton('triton_poi_fused_cat_mean_sub_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_mean_sub_0', 'mutated_arg_names': ['in_ptr0', 'out_ptr2'], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_mean_sub_0(in_ptr0, out_ptr0, out_ptr1, out_ptr2, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
x1 = (xindex // 64)
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tl.store(out_ptr0 + (x2), tmp10, xmask)
tl.store(out_ptr1 + (x0 + (80*x1)), tmp10, xmask)
tl.store(out_ptr2 + (x2), tmp10, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/52/c52fvha3se2kccjivsimg2dchvpu2igv5pa544e2dzcwbkplcnbv.py
# Topologically Sorted Source Nodes: [y_5, x], Original ATen: [aten.mean, aten.cat]
# Source node to ATen node mapping:
# x => cat
# y_5 => mean_2
# Graph fragment:
# %mean_2 : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%view_3, [-1]), kwargs = {})
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%view_1, %expand], 1), kwargs = {})
triton_per_fused_cat_mean_1 = async_compile.triton('triton_per_fused_cat_mean_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 64],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {2: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 3), equal_to_1=(2,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_cat_mean_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 4, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_cat_mean_1(in_ptr0, out_ptr1, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 1
rnumel = 64
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
r1 = rindex % 16
r2 = (rindex // 16)
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp2 = tl.load(in_ptr0 + (64 + r0), None)
tmp5 = tl.load(in_ptr0 + (128 + r0), None)
tmp8 = tl.load(in_ptr0 + (192 + r0), None)
tmp1 = tmp0 * tmp0
tmp3 = tmp2 * tmp2
tmp4 = tmp1 + tmp3
tmp6 = tmp5 * tmp5
tmp7 = tmp4 + tmp6
tmp9 = tmp8 * tmp8
tmp10 = tmp7 + tmp9
tmp11 = 4.0
tmp12 = tmp10 / tmp11
tmp13 = 1e-08
tmp14 = tmp12 + tmp13
tmp15 = libdevice.sqrt(tmp14)
tmp16 = tl.broadcast_to(tmp15, [XBLOCK, RBLOCK])
tmp18 = tl.sum(tmp16, 1)[:, None]
tmp19 = 64.0
tmp20 = tmp18 / tmp19
tl.store(out_ptr1 + (tl.broadcast_to(r1 + (80*r2), [XBLOCK, RBLOCK])), tmp20, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 4, 4), (64, 256, 16, 4, 1), torch.float32)
buf4 = empty_strided_cuda((4, 5, 4, 4), (80, 16, 4, 1), torch.float32)
buf2 = reinterpret_tensor(buf4, (4, 4, 4, 4), (80, 16, 4, 1), 0) # alias
# Topologically Sorted Source Nodes: [mean, y_2, x], Original ATen: [aten.mean, aten.sub, aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_mean_sub_0.run(arg0_1, buf0, buf2, arg0_1, 256, grid=grid(256), stream=stream0)
del arg0_1
buf3 = reinterpret_tensor(buf4, (4, 1, 4, 4), (80, 16, 4, 1), 64) # alias
# Topologically Sorted Source Nodes: [y_5, x], Original ATen: [aten.mean, aten.cat]
triton_per_fused_cat_mean_1.run(buf0, buf3, 1, 64, grid=grid(1), stream=stream0)
del buf0
return (buf4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.utils.tensorboard
class MinibatchStd(nn.Module):
"""
Adds the aveage std of each data point over a
slice of the minibatch to that slice as a new
feature map. This gives an output with one extra
channel.
Arguments:
group_size (int): Number of entries in each slice
of the batch. If <= 0, the entire batch is used.
Default value is 4.
eps (float): Epsilon value added for numerical stability.
Default value is 1e-8.
"""
def __init__(self, group_size=4, eps=1e-08, *args, **kwargs):
super(MinibatchStd, self).__init__()
if group_size is None or group_size <= 0:
group_size = 0
assert group_size != 1, 'Can not use 1 as minibatch std group size.'
self.group_size = group_size
self.eps = eps
def forward(self, input, **kwargs):
"""
Add a new feature map to the input containing the average
standard deviation for each slice.
Arguments:
input (torch.Tensor)
Returns:
output (torch.Tensor)
"""
group_size = self.group_size or input.size(0)
assert input.size(0
) >= group_size, 'Can not use a smaller batch size ' + '({}) than the specified '.format(
input.size(0)) + 'group size ({}) '.format(group_size
) + 'of this minibatch std layer.'
assert input.size(0
) % group_size == 0, 'Can not use a batch of a size ' + '({}) that is not '.format(
input.size(0)) + 'evenly divisible by the group size ({})'.format(
group_size)
x = input
y = input.view(group_size, -1, *input.size()[1:])
y = y.float()
y -= y.mean(dim=0, keepdim=True)
y = torch.mean(y ** 2, dim=0)
y = torch.sqrt(y + self.eps)
y = torch.mean(y.view(y.size(0), -1), dim=-1)
y = y.view(-1, *([1] * (input.dim() - 1)))
y = y
y = y.repeat(group_size, *([1] * (y.dim() - 1)))
y = y.expand(y.size(0), 1, *x.size()[2:])
x = torch.cat([x, y], dim=1)
return x
def extra_repr(self):
return 'group_size={}'.format(self.group_size or '-1')
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
import torch.utils.tensorboard
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_mean_sub_0(in_ptr0, out_ptr0, out_ptr1, out_ptr2,
xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 64
x1 = xindex // 64
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (64 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (128 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (192 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tl.store(out_ptr0 + x2, tmp10, xmask)
tl.store(out_ptr1 + (x0 + 80 * x1), tmp10, xmask)
tl.store(out_ptr2 + x2, tmp10, xmask)
@triton.jit
def triton_per_fused_cat_mean_1(in_ptr0, out_ptr1, xnumel, rnumel, XBLOCK:
tl.constexpr):
RBLOCK: tl.constexpr = 64
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
r1 = rindex % 16
r2 = rindex // 16
tmp0 = tl.load(in_ptr0 + r0, None)
tmp2 = tl.load(in_ptr0 + (64 + r0), None)
tmp5 = tl.load(in_ptr0 + (128 + r0), None)
tmp8 = tl.load(in_ptr0 + (192 + r0), None)
tmp1 = tmp0 * tmp0
tmp3 = tmp2 * tmp2
tmp4 = tmp1 + tmp3
tmp6 = tmp5 * tmp5
tmp7 = tmp4 + tmp6
tmp9 = tmp8 * tmp8
tmp10 = tmp7 + tmp9
tmp11 = 4.0
tmp12 = tmp10 / tmp11
tmp13 = 1e-08
tmp14 = tmp12 + tmp13
tmp15 = libdevice.sqrt(tmp14)
tmp16 = tl.broadcast_to(tmp15, [XBLOCK, RBLOCK])
tmp18 = tl.sum(tmp16, 1)[:, None]
tmp19 = 64.0
tmp20 = tmp18 / tmp19
tl.store(out_ptr1 + tl.broadcast_to(r1 + 80 * r2, [XBLOCK, RBLOCK]),
tmp20, None)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 1, 4, 4, 4), (64, 256, 16, 4, 1),
torch.float32)
buf4 = empty_strided_cuda((4, 5, 4, 4), (80, 16, 4, 1), torch.float32)
buf2 = reinterpret_tensor(buf4, (4, 4, 4, 4), (80, 16, 4, 1), 0)
get_raw_stream(0)
triton_poi_fused_cat_mean_sub_0[grid(256)](arg0_1, buf0, buf2,
arg0_1, 256, XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
buf3 = reinterpret_tensor(buf4, (4, 1, 4, 4), (80, 16, 4, 1), 64)
triton_per_fused_cat_mean_1[grid(1)](buf0, buf3, 1, 64, XBLOCK=1,
num_warps=2, num_stages=1)
del buf0
return buf4,
class MinibatchStdNew(nn.Module):
"""
Adds the aveage std of each data point over a
slice of the minibatch to that slice as a new
feature map. This gives an output with one extra
channel.
Arguments:
group_size (int): Number of entries in each slice
of the batch. If <= 0, the entire batch is used.
Default value is 4.
eps (float): Epsilon value added for numerical stability.
Default value is 1e-8.
"""
def __init__(self, group_size=4, eps=1e-08, *args, **kwargs):
super(MinibatchStdNew, self).__init__()
if group_size is None or group_size <= 0:
group_size = 0
assert group_size != 1, 'Can not use 1 as minibatch std group size.'
self.group_size = group_size
self.eps = eps
def extra_repr(self):
return 'group_size={}'.format(self.group_size or '-1')
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
allenbao64/jamm-bandit
|
MinibatchStd
| false | 12,411 |
[
"Apache-2.0"
] | 0 |
06c9d8815ce907a68b0bc4ecf8bee4a2465c6a9e
|
https://github.com/allenbao64/jamm-bandit/tree/06c9d8815ce907a68b0bc4ecf8bee4a2465c6a9e
|
SimpleSliceModel
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ng/cnggtmai2hzxc7e5creviqseyyf7qiy5pfpdjlp2pomqsserjuzj.py
# Topologically Sorted Source Nodes: [add], Original ATen: [aten.add]
# Source node to ATen node mapping:
# add => add
# Graph fragment:
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arg0_1, %arg0_1), kwargs = {})
triton_poi_fused_add_0 = async_compile.triton('triton_poi_fused_add_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tmp0 + tmp0
tl.store(out_ptr0 + (x0), tmp1, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [add], Original ATen: [aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_0.run(arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
return (reinterpret_tensor(buf0, (3, 4, 4), (16, 4, 1), 80), )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.onnx
import torch.nn
class SimpleSliceModel(torch.nn.Module):
def __init__(self):
super(SimpleSliceModel, self).__init__()
def forward(self, tensor):
other = (tensor + tensor)[1:]
return other[0][1:]
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.onnx
import torch.nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_add_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tmp0 + tmp0
tl.store(out_ptr0 + x0, tmp1, xmask)
def call(args):
arg0_1, = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_0[grid(256)](arg0_1, buf0, 256, XBLOCK=256,
num_warps=4, num_stages=1)
del arg0_1
return reinterpret_tensor(buf0, (3, 4, 4), (16, 4, 1), 80),
class SimpleSliceModelNew(torch.nn.Module):
def __init__(self):
super(SimpleSliceModelNew, self).__init__()
def forward(self, input_0):
arg0_1 = input_0
output = call([arg0_1])
return output[0]
|
geoffberry/glow
|
SimpleSliceModel
| false | 12,412 |
[
"Apache-2.0"
] | 0 |
24b2827c830eb58af56a0704e899968026832e9c
|
https://github.com/geoffberry/glow/tree/24b2827c830eb58af56a0704e899968026832e9c
|
HamidaEtAl
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/pa/cpaohuvckqq7lbfpx3nel7h3eb65rfn7acxfhuytgkck5h34niux.py
# Topologically Sorted Source Nodes: [conv3d, x], Original ATen: [aten.convolution, aten.relu]
# Source node to ATen node mapping:
# conv3d => convolution
# x => relu
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1, 1], [0, 0, 0], [1, 1, 1], False, [0, 0, 0], 1), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution,), kwargs = {})
triton_poi_fused_convolution_relu_0 = async_compile.triton('triton_poi_fused_convolution_relu_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[33554432],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 19066240
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 238328) % 20
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x3), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kh/ckha7g6ws7mumpiujhk6y5h5vexe6vprahxnkpdnnvhkrzvcg2tg.py
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# x_1 => convolution_1
# Graph fragment:
# %convolution_1 : [num_users=2] = call_function[target=torch.ops.aten.convolution.default](args = (%relu, %primals_4, %primals_5, [2, 1, 1], [1, 0, 0], [1, 1, 1], False, [0, 0, 0], 1), kwargs = {})
triton_poi_fused_convolution_1 = async_compile.triton('triton_poi_fused_convolution_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16777216],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 9533120
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 119164) % 20
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/4k/c4kggtolinytq3dk6a7oyawlonqv5raxonjuhpvgnheskglysy5l.py
# Topologically Sorted Source Nodes: [conv3d_2, x_2], Original ATen: [aten.convolution, aten.relu]
# Source node to ATen node mapping:
# conv3d_2 => convolution_2
# x_2 => relu_1
# Graph fragment:
# %convolution_2 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%convolution_1, %primals_6, %primals_7, [1, 1, 1], [1, 0, 0], [1, 1, 1], False, [0, 0, 0], 1), kwargs = {})
# %relu_1 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_2,), kwargs = {})
triton_poi_fused_convolution_relu_2 = async_compile.triton('triton_poi_fused_convolution_relu_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16777216],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_2', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_2(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 15624000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 111600) % 35
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x3), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/le/cles5mw2qfnkle3lp4ufjq3a36ijdmf5zq37npaxku2jpwl7rzyn.py
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.convolution]
# Source node to ATen node mapping:
# x_3 => convolution_3
# Graph fragment:
# %convolution_3 : [num_users=2] = call_function[target=torch.ops.aten.convolution.default](args = (%relu_1, %primals_8, %primals_9, [2, 1, 1], [1, 0, 0], [1, 1, 1], False, [0, 0, 0], 1), kwargs = {})
triton_poi_fused_convolution_3 = async_compile.triton('triton_poi_fused_convolution_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8388608],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_3', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_3(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 8064000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 57600) % 35
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + (x3), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/t7/ct753eoeg253nmb27xd2tv746tepqemrsiyltthh35gvgq53s6gr.py
# Topologically Sorted Source Nodes: [conv3d_4, x_4], Original ATen: [aten.convolution, aten.relu]
# Source node to ATen node mapping:
# conv3d_4 => convolution_4
# x_4 => relu_2
# Graph fragment:
# %convolution_4 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%convolution_3, %primals_10, %primals_11, [1, 1, 1], [1, 0, 0], [1, 1, 1], False, [0, 0, 0], 1), kwargs = {})
# %relu_2 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_4,), kwargs = {})
triton_poi_fused_convolution_relu_4 = async_compile.triton('triton_poi_fused_convolution_relu_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8388608],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_4', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_4(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 8064000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = (xindex // 57600) % 35
tmp0 = tl.load(in_out_ptr0 + (x3), xmask)
tmp1 = tl.load(in_ptr0 + (x1), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x3), tmp4, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/su/csuhyxi7wqxpurmhxz7ksl2i6rpl34lf3oh42houo2itblaef5tz.py
# Topologically Sorted Source Nodes: [conv3d_5, x_5], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# conv3d_5 => convolution_5
# x_5 => relu_3
# Graph fragment:
# %convolution_5 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%relu_2, %primals_12, %primals_13, [2, 1, 1], [1, 0, 0], [1, 1, 1], False, [0, 0, 0], 1), kwargs = {})
# %relu_3 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_5,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu_3, 0), kwargs = {})
triton_poi_fused_convolution_relu_threshold_backward_5 = async_compile.triton('triton_poi_fused_convolution_relu_threshold_backward_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8388608],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*i1', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_threshold_backward_5', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_5(in_ptr0, in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 4536000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x5 = xindex
x2 = (xindex // 32400) % 35
x0 = xindex % 3600
x4 = (xindex // 3600)
tmp0 = tl.load(in_ptr0 + (x5), xmask)
tmp1 = tl.load(in_ptr1 + (x2), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x0 + (3616*x4)), tmp4, xmask)
tl.store(out_ptr1 + (x0 + (3712*x4)), tmp6, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/eq/ceqvitldybse7r552dojqrid5w3gvjjq6vkf7ltz6z43fttzbx2c.py
# Topologically Sorted Source Nodes: [conv3d_5, x_5, x_6], Original ATen: [aten.convolution, aten.relu, aten.view]
# Source node to ATen node mapping:
# conv3d_5 => convolution_5
# x_5 => relu_3
# x_6 => view
# Graph fragment:
# %convolution_5 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%relu_2, %primals_12, %primals_13, [2, 1, 1], [1, 0, 0], [1, 1, 1], False, [0, 0, 0], 1), kwargs = {})
# %relu_3 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_5,), kwargs = {})
# %view : [num_users=2] = call_function[target=torch.ops.aten.reshape.default](args = (%relu_3, [-1, 35]), kwargs = {})
triton_poi_fused_convolution_relu_view_6 = async_compile.triton('triton_poi_fused_convolution_relu_view_6', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8388608],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_view_6', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_view_6(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4536000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + ((3616*(x0 // 3600)) + (x0 % 3600)), xmask)
tl.store(out_ptr0 + (x0), tmp0, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12, primals_13, primals_14, primals_15 = args
args.clear()
assert_size_stride(primals_1, (20, 1, 3, 3, 3), (27, 27, 9, 3, 1))
assert_size_stride(primals_2, (20, ), (1, ))
assert_size_stride(primals_3, (4, 1, 64, 64, 64), (262144, 262144, 4096, 64, 1))
assert_size_stride(primals_4, (20, 20, 3, 1, 1), (60, 3, 1, 1, 1))
assert_size_stride(primals_5, (20, ), (1, ))
assert_size_stride(primals_6, (35, 20, 3, 3, 3), (540, 27, 9, 3, 1))
assert_size_stride(primals_7, (35, ), (1, ))
assert_size_stride(primals_8, (35, 35, 3, 1, 1), (105, 3, 1, 1, 1))
assert_size_stride(primals_9, (35, ), (1, ))
assert_size_stride(primals_10, (35, 35, 3, 1, 1), (105, 3, 1, 1, 1))
assert_size_stride(primals_11, (35, ), (1, ))
assert_size_stride(primals_12, (35, 35, 2, 1, 1), (70, 2, 1, 1, 1))
assert_size_stride(primals_13, (35, ), (1, ))
assert_size_stride(primals_14, (4, 35), (35, 1))
assert_size_stride(primals_15, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv3d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1, 1, 1), padding=(0, 0, 0), dilation=(1, 1, 1), transposed=False, output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 20, 62, 62, 62), (4766560, 238328, 3844, 62, 1))
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [conv3d, x], Original ATen: [aten.convolution, aten.relu]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_relu_0.run(buf1, primals_2, 19066240, grid=grid(19066240), stream=stream0)
del primals_2
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.convolution]
buf2 = extern_kernels.convolution(buf1, primals_4, stride=(2, 1, 1), padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False, output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 20, 31, 62, 62), (2383280, 119164, 3844, 62, 1))
buf3 = buf2; del buf2 # reuse
# Topologically Sorted Source Nodes: [x_1], Original ATen: [aten.convolution]
triton_poi_fused_convolution_1.run(buf3, primals_5, 9533120, grid=grid(9533120), stream=stream0)
del primals_5
# Topologically Sorted Source Nodes: [conv3d_2], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(buf3, primals_6, stride=(1, 1, 1), padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False, output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 35, 31, 60, 60), (3906000, 111600, 3600, 60, 1))
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [conv3d_2, x_2], Original ATen: [aten.convolution, aten.relu]
triton_poi_fused_convolution_relu_2.run(buf5, primals_7, 15624000, grid=grid(15624000), stream=stream0)
del primals_7
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.convolution]
buf6 = extern_kernels.convolution(buf5, primals_8, stride=(2, 1, 1), padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False, output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 35, 16, 60, 60), (2016000, 57600, 3600, 60, 1))
buf7 = buf6; del buf6 # reuse
# Topologically Sorted Source Nodes: [x_3], Original ATen: [aten.convolution]
triton_poi_fused_convolution_3.run(buf7, primals_9, 8064000, grid=grid(8064000), stream=stream0)
del primals_9
# Topologically Sorted Source Nodes: [conv3d_4], Original ATen: [aten.convolution]
buf8 = extern_kernels.convolution(buf7, primals_10, stride=(1, 1, 1), padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False, output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf8, (4, 35, 16, 60, 60), (2016000, 57600, 3600, 60, 1))
buf9 = buf8; del buf8 # reuse
# Topologically Sorted Source Nodes: [conv3d_4, x_4], Original ATen: [aten.convolution, aten.relu]
triton_poi_fused_convolution_relu_4.run(buf9, primals_11, 8064000, grid=grid(8064000), stream=stream0)
del primals_11
# Topologically Sorted Source Nodes: [conv3d_5], Original ATen: [aten.convolution]
buf10 = extern_kernels.convolution(buf9, primals_12, stride=(2, 1, 1), padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False, output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf10, (4, 35, 9, 60, 60), (1134000, 32400, 3600, 60, 1))
buf11 = empty_strided_cuda((4, 35, 9, 60, 60), (1139040, 32544, 3616, 60, 1), torch.float32)
buf14 = empty_strided_cuda((4, 35, 9, 60, 60), (1169280, 33408, 3712, 60, 1), torch.bool)
# Topologically Sorted Source Nodes: [conv3d_5, x_5], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
triton_poi_fused_convolution_relu_threshold_backward_5.run(buf10, primals_13, buf11, buf14, 4536000, grid=grid(4536000), stream=stream0)
del primals_13
buf12 = reinterpret_tensor(buf10, (129600, 35), (35, 1), 0); del buf10 # reuse
# Topologically Sorted Source Nodes: [conv3d_5, x_5, x_6], Original ATen: [aten.convolution, aten.relu, aten.view]
triton_poi_fused_convolution_relu_view_6.run(buf11, buf12, 4536000, grid=grid(4536000), stream=stream0)
del buf11
buf13 = empty_strided_cuda((129600, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x_7], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_15, buf12, reinterpret_tensor(primals_14, (35, 4), (1, 35), 0), alpha=1, beta=1, out=buf13)
del primals_15
return (buf13, primals_1, primals_3, primals_4, primals_6, primals_8, primals_10, primals_12, buf1, buf3, buf5, buf7, buf9, buf12, primals_14, buf14, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((20, 1, 3, 3, 3), (27, 27, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((20, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 1, 64, 64, 64), (262144, 262144, 4096, 64, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((20, 20, 3, 1, 1), (60, 3, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((20, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((35, 20, 3, 3, 3), (540, 27, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((35, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((35, 35, 3, 1, 1), (105, 3, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((35, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((35, 35, 3, 1, 1), (105, 3, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((35, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_12 = rand_strided((35, 35, 2, 1, 1), (70, 2, 1, 1, 1), device='cuda:0', dtype=torch.float32)
primals_13 = rand_strided((35, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_14 = rand_strided((4, 35), (35, 1), device='cuda:0', dtype=torch.float32)
primals_15 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11, primals_12, primals_13, primals_14, primals_15])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.utils
import torch.utils.data
import torch.nn as nn
import torch.nn.functional as F
from torch.nn import init
class HamidaEtAl(nn.Module):
"""
3-D Deep Learning Approach for Remote Sensing Image Classification
Amina Ben Hamida, Alexandre Benoit, Patrick Lambert, Chokri Ben Amar
IEEE TGRS, 2018
https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8344565
"""
@staticmethod
def weight_init(m):
if isinstance(m, nn.Linear) or isinstance(m, nn.Conv3d):
init.kaiming_normal_(m.weight)
init.zeros_(m.bias)
def __init__(self, input_channels, n_classes, patch_size=5, dilation=1):
super(HamidaEtAl, self).__init__()
self.patch_size = patch_size
self.input_channels = input_channels
dilation = dilation, 1, 1
if patch_size == 3:
self.conv1 = nn.Conv3d(1, 20, (3, 3, 3), stride=(1, 1, 1),
dilation=dilation, padding=1)
else:
self.conv1 = nn.Conv3d(1, 20, (3, 3, 3), stride=(1, 1, 1),
dilation=dilation, padding=0)
self.pool1 = nn.Conv3d(20, 20, (3, 1, 1), dilation=dilation, stride
=(2, 1, 1), padding=(1, 0, 0))
self.conv2 = nn.Conv3d(20, 35, (3, 3, 3), dilation=dilation, stride
=(1, 1, 1), padding=(1, 0, 0))
self.pool2 = nn.Conv3d(35, 35, (3, 1, 1), dilation=dilation, stride
=(2, 1, 1), padding=(1, 0, 0))
self.conv3 = nn.Conv3d(35, 35, (3, 1, 1), dilation=dilation, stride
=(1, 1, 1), padding=(1, 0, 0))
self.conv4 = nn.Conv3d(35, 35, (2, 1, 1), dilation=dilation, stride
=(2, 1, 1), padding=(1, 0, 0))
self.features_size = self._get_final_flattened_size()
self.fc = nn.Linear(self.features_size, n_classes)
self.apply(self.weight_init)
def _get_final_flattened_size(self):
with torch.no_grad():
x = torch.zeros((1, 1, self.input_channels, self.patch_size,
self.patch_size))
x = self.pool1(self.conv1(x))
x = self.pool2(self.conv2(x))
x = self.conv3(x)
x = self.conv4(x)
_, t, c, w, h = x.size()
return t * c * w * h
def forward(self, x):
x = F.relu(self.conv1(x))
x = self.pool1(x)
x = F.relu(self.conv2(x))
x = self.pool2(x)
x = F.relu(self.conv3(x))
x = F.relu(self.conv4(x))
x = x.view(-1, self.features_size)
x = self.fc(x)
return x
def get_inputs():
return [torch.rand([4, 1, 64, 64, 64])]
def get_init_inputs():
return [[], {'input_channels': 4, 'n_classes': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.utils
import torch.utils.data
import torch.nn as nn
from torch.nn import init
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_convolution_relu_0(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 19066240
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 238328 % 20
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x3, tmp4, xmask)
@triton.jit
def triton_poi_fused_convolution_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 9533120
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 119164 % 20
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
@triton.jit
def triton_poi_fused_convolution_relu_2(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 15624000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 111600 % 35
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x3, tmp4, xmask)
@triton.jit
def triton_poi_fused_convolution_3(in_out_ptr0, in_ptr0, xnumel, XBLOCK: tl
.constexpr):
xnumel = 8064000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 57600 % 35
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tl.store(in_out_ptr0 + x3, tmp2, xmask)
@triton.jit
def triton_poi_fused_convolution_relu_4(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 8064000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex
x1 = xindex // 57600 % 35
tmp0 = tl.load(in_out_ptr0 + x3, xmask)
tmp1 = tl.load(in_ptr0 + x1, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x3, tmp4, xmask)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_5(in_ptr0, in_ptr1,
out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 4536000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x5 = xindex
x2 = xindex // 32400 % 35
x0 = xindex % 3600
x4 = xindex // 3600
tmp0 = tl.load(in_ptr0 + x5, xmask)
tmp1 = tl.load(in_ptr1 + x2, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x0 + 3616 * x4), tmp4, xmask)
tl.store(out_ptr1 + (x0 + 3712 * x4), tmp6, xmask)
@triton.jit
def triton_poi_fused_convolution_relu_view_6(in_ptr0, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 4536000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (3616 * (x0 // 3600) + x0 % 3600), xmask)
tl.store(out_ptr0 + x0, tmp0, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11, primals_12,
primals_13, primals_14, primals_15) = args
args.clear()
assert_size_stride(primals_1, (20, 1, 3, 3, 3), (27, 27, 9, 3, 1))
assert_size_stride(primals_2, (20,), (1,))
assert_size_stride(primals_3, (4, 1, 64, 64, 64), (262144, 262144, 4096,
64, 1))
assert_size_stride(primals_4, (20, 20, 3, 1, 1), (60, 3, 1, 1, 1))
assert_size_stride(primals_5, (20,), (1,))
assert_size_stride(primals_6, (35, 20, 3, 3, 3), (540, 27, 9, 3, 1))
assert_size_stride(primals_7, (35,), (1,))
assert_size_stride(primals_8, (35, 35, 3, 1, 1), (105, 3, 1, 1, 1))
assert_size_stride(primals_9, (35,), (1,))
assert_size_stride(primals_10, (35, 35, 3, 1, 1), (105, 3, 1, 1, 1))
assert_size_stride(primals_11, (35,), (1,))
assert_size_stride(primals_12, (35, 35, 2, 1, 1), (70, 2, 1, 1, 1))
assert_size_stride(primals_13, (35,), (1,))
assert_size_stride(primals_14, (4, 35), (35, 1))
assert_size_stride(primals_15, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,
1, 1), padding=(0, 0, 0), dilation=(1, 1, 1), transposed=False,
output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 20, 62, 62, 62), (4766560, 238328,
3844, 62, 1))
buf1 = buf0
del buf0
get_raw_stream(0)
triton_poi_fused_convolution_relu_0[grid(19066240)](buf1, primals_2,
19066240, XBLOCK=512, num_warps=8, num_stages=1)
del primals_2
buf2 = extern_kernels.convolution(buf1, primals_4, stride=(2, 1, 1),
padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False,
output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf2, (4, 20, 31, 62, 62), (2383280, 119164,
3844, 62, 1))
buf3 = buf2
del buf2
triton_poi_fused_convolution_1[grid(9533120)](buf3, primals_5,
9533120, XBLOCK=512, num_warps=8, num_stages=1)
del primals_5
buf4 = extern_kernels.convolution(buf3, primals_6, stride=(1, 1, 1),
padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False,
output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 35, 31, 60, 60), (3906000, 111600,
3600, 60, 1))
buf5 = buf4
del buf4
triton_poi_fused_convolution_relu_2[grid(15624000)](buf5, primals_7,
15624000, XBLOCK=512, num_warps=8, num_stages=1)
del primals_7
buf6 = extern_kernels.convolution(buf5, primals_8, stride=(2, 1, 1),
padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False,
output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 35, 16, 60, 60), (2016000, 57600, 3600,
60, 1))
buf7 = buf6
del buf6
triton_poi_fused_convolution_3[grid(8064000)](buf7, primals_9,
8064000, XBLOCK=1024, num_warps=4, num_stages=1)
del primals_9
buf8 = extern_kernels.convolution(buf7, primals_10, stride=(1, 1, 1
), padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False,
output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf8, (4, 35, 16, 60, 60), (2016000, 57600, 3600,
60, 1))
buf9 = buf8
del buf8
triton_poi_fused_convolution_relu_4[grid(8064000)](buf9, primals_11,
8064000, XBLOCK=1024, num_warps=4, num_stages=1)
del primals_11
buf10 = extern_kernels.convolution(buf9, primals_12, stride=(2, 1,
1), padding=(1, 0, 0), dilation=(1, 1, 1), transposed=False,
output_padding=(0, 0, 0), groups=1, bias=None)
assert_size_stride(buf10, (4, 35, 9, 60, 60), (1134000, 32400, 3600,
60, 1))
buf11 = empty_strided_cuda((4, 35, 9, 60, 60), (1139040, 32544,
3616, 60, 1), torch.float32)
buf14 = empty_strided_cuda((4, 35, 9, 60, 60), (1169280, 33408,
3712, 60, 1), torch.bool)
triton_poi_fused_convolution_relu_threshold_backward_5[grid(4536000)](
buf10, primals_13, buf11, buf14, 4536000, XBLOCK=1024,
num_warps=4, num_stages=1)
del primals_13
buf12 = reinterpret_tensor(buf10, (129600, 35), (35, 1), 0)
del buf10
triton_poi_fused_convolution_relu_view_6[grid(4536000)](buf11,
buf12, 4536000, XBLOCK=512, num_warps=8, num_stages=1)
del buf11
buf13 = empty_strided_cuda((129600, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_15, buf12, reinterpret_tensor(
primals_14, (35, 4), (1, 35), 0), alpha=1, beta=1, out=buf13)
del primals_15
return (buf13, primals_1, primals_3, primals_4, primals_6, primals_8,
primals_10, primals_12, buf1, buf3, buf5, buf7, buf9, buf12,
primals_14, buf14)
class HamidaEtAlNew(nn.Module):
"""
3-D Deep Learning Approach for Remote Sensing Image Classification
Amina Ben Hamida, Alexandre Benoit, Patrick Lambert, Chokri Ben Amar
IEEE TGRS, 2018
https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8344565
"""
@staticmethod
def weight_init(m):
if isinstance(m, nn.Linear) or isinstance(m, nn.Conv3d):
init.kaiming_normal_(m.weight)
init.zeros_(m.bias)
def __init__(self, input_channels, n_classes, patch_size=5, dilation=1):
super(HamidaEtAlNew, self).__init__()
self.patch_size = patch_size
self.input_channels = input_channels
dilation = dilation, 1, 1
if patch_size == 3:
self.conv1 = nn.Conv3d(1, 20, (3, 3, 3), stride=(1, 1, 1),
dilation=dilation, padding=1)
else:
self.conv1 = nn.Conv3d(1, 20, (3, 3, 3), stride=(1, 1, 1),
dilation=dilation, padding=0)
self.pool1 = nn.Conv3d(20, 20, (3, 1, 1), dilation=dilation, stride
=(2, 1, 1), padding=(1, 0, 0))
self.conv2 = nn.Conv3d(20, 35, (3, 3, 3), dilation=dilation, stride
=(1, 1, 1), padding=(1, 0, 0))
self.pool2 = nn.Conv3d(35, 35, (3, 1, 1), dilation=dilation, stride
=(2, 1, 1), padding=(1, 0, 0))
self.conv3 = nn.Conv3d(35, 35, (3, 1, 1), dilation=dilation, stride
=(1, 1, 1), padding=(1, 0, 0))
self.conv4 = nn.Conv3d(35, 35, (2, 1, 1), dilation=dilation, stride
=(2, 1, 1), padding=(1, 0, 0))
self.features_size = self._get_final_flattened_size()
self.fc = nn.Linear(self.features_size, n_classes)
self.apply(self.weight_init)
def _get_final_flattened_size(self):
with torch.no_grad():
x = torch.zeros((1, 1, self.input_channels, self.patch_size,
self.patch_size))
x = self.pool1(self.conv1(x))
x = self.pool2(self.conv2(x))
x = self.conv3(x)
x = self.conv4(x)
_, t, c, w, h = x.size()
return t * c * w * h
def forward(self, input_0):
primals_1 = self.conv1.weight
primals_2 = self.conv1.bias
primals_4 = self.pool1.weight
primals_5 = self.pool1.bias
primals_6 = self.conv2.weight
primals_7 = self.conv2.bias
primals_8 = self.pool2.weight
primals_9 = self.pool2.bias
primals_10 = self.conv3.weight
primals_11 = self.conv3.bias
primals_12 = self.conv4.weight
primals_13 = self.conv4.bias
primals_14 = self.fc.weight
primals_15 = self.fc.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11, primals_12, primals_13, primals_14,
primals_15])
return output[0]
|
dikers/DeepHyper
|
HamidaEtAl
| false | 12,413 |
[
"Apache-2.0"
] | 0 |
827a8f3077e18b71cf448a2e56e49670428b1bfd
|
https://github.com/dikers/DeepHyper/tree/827a8f3077e18b71cf448a2e56e49670428b1bfd
|
BarlowTwinsLoss
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/v3/cv3kb3xe3cnhyoaqqw6etvgavblpvrqt4klo34heveuf3mtxqne3.py
# Topologically Sorted Source Nodes: [mean, sub, var, add, pow_1, z_left], Original ATen: [aten.mean, aten.sub, aten.var, aten.add, aten.pow, aten.div]
# Source node to ATen node mapping:
# add => add
# mean => mean
# pow_1 => pow_1
# sub => sub
# var => var
# z_left => div
# Graph fragment:
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.dim](args = (%arg0_1, [0]), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg0_1, %mean), kwargs = {})
# %var : [num_users=1] = call_function[target=torch.ops.aten.var.correction](args = (%arg0_1, [0]), kwargs = {correction: 1})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%var, 1e-12), kwargs = {})
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%add, 0.5), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub, %pow_1), kwargs = {})
triton_poi_fused_add_div_mean_pow_sub_var_0 = async_compile.triton('triton_poi_fused_add_div_mean_pow_sub_var_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_div_mean_pow_sub_var_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 5, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_div_mean_pow_sub_var_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + (x2), xmask)
tmp1 = tl.load(in_ptr0 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (4 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (8 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (12 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tmp11 = tmp1 - tmp9
tmp12 = tmp11 * tmp11
tmp13 = tmp2 - tmp9
tmp14 = tmp13 * tmp13
tmp15 = tmp12 + tmp14
tmp16 = tmp4 - tmp9
tmp17 = tmp16 * tmp16
tmp18 = tmp15 + tmp17
tmp19 = tmp6 - tmp9
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp22 = 3.0
tmp23 = tmp21 / tmp22
tmp24 = 1e-12
tmp25 = tmp23 + tmp24
tmp26 = libdevice.sqrt(tmp25)
tmp27 = tmp10 / tmp26
tl.store(out_ptr0 + (x2), tmp27, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/o4/co43gtmemxr2zlvmvdzxkr37cbfe5mt7o5wuu2udutuyo3fgaovi.py
# Topologically Sorted Source Nodes: [cross_correlation], Original ATen: [aten.div]
# Source node to ATen node mapping:
# cross_correlation => div_2
# Graph fragment:
# %div_2 : [num_users=4] = call_function[target=torch.ops.aten.div.Tensor](args = (%mm, 4), kwargs = {})
triton_poi_fused_div_1 = async_compile.triton('triton_poi_fused_div_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_1(in_out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + (x0), xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tl.store(in_out_ptr0 + (x0), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/4e/c4ems2krpjome7vpom5xb4iaexvrawrectf3qe6lidip6cfk7bmp.py
# Topologically Sorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
# Graph fragment:
# %clone_1 : [num_users=4] = call_function[target=torch.ops.aten.clone](args = (%div_2,), kwargs = {})
triton_poi_fused_2 = async_compile.triton('triton_poi_fused_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_2(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tl.store(out_ptr0 + (x0), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/sh/cshzmbe23ekozat7r23uh7zegdlglw2y6touqiciencckpippagu.py
# Topologically Sorted Source Nodes: [add_], Original ATen: [aten.add]
# Source node to ATen node mapping:
# add_ => add_2
# Graph fragment:
# %add_2 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%diagonal, -1), kwargs = {})
# %copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%diagonal_default, %add_2), kwargs = {})
triton_poi_fused_add_3 = async_compile.triton('triton_poi_fused_add_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_3', 'mutated_arg_names': ['out_ptr0'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_3(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (5*x0), xmask, eviction_policy='evict_last')
tmp1 = -1.0
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + (5*x0), tmp2, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kg/ckg6rqmgdhxkvvs7ji4lw4uorxeuvjsl6bkaharfjchuhrhvcdor.py
# Topologically Sorted Source Nodes: [pow_, on_diag_loss], Original ATen: [aten.pow, aten.sum]
# Source node to ATen node mapping:
# on_diag_loss => sum_1
# pow_ => pow_3
# Graph fragment:
# %pow_3 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%diagonal_1, 2), kwargs = {})
# %copy__default_1 : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%diagonal_default_1, %pow_3), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%diagonal_2,), kwargs = {})
triton_per_fused_pow_sum_4 = async_compile.triton('triton_per_fused_pow_sum_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 4],
reduction_hint=ReductionHint.DEFAULT,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_pow_sum_4', 'mutated_arg_names': ['in_ptr0', 'out_ptr1'], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_pow_sum_4(in_ptr0, out_ptr1, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 1
rnumel = 4
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (5*r0), None, eviction_policy='evict_last')
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.sum(tmp2, 1)[:, None]
tl.store(out_ptr1 + (tl.broadcast_to(5*r0, [XBLOCK, RBLOCK])), tmp1, None)
tl.store(out_ptr2 + (tl.full([XBLOCK, 1], 0, tl.int32)), tmp4, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/js/cjs2r5os6wqtnw2uh2savbghsqjbkrldpi6wk4iohyalrp4xd6rd.py
# Topologically Sorted Source Nodes: [off_diag], Original ATen: [aten.fill]
# Source node to ATen node mapping:
# off_diag => full_default
# Graph fragment:
# %full_default : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([4], 0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %copy__default_2 : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%as_strided_default, %full_default), kwargs = {})
triton_poi_fused_fill_5 = async_compile.triton('triton_poi_fused_fill_5', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[4],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0,), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_fill_5', 'mutated_arg_names': ['out_ptr0'], 'no_x_dim': False, 'num_load': 0, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_fill_5(out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = 0.0
tl.store(out_ptr0 + (5*x0), tmp0, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/kv/ckvw46euabmkgc4wu5t7wl3fk2hcqjptcgi5kpil77sckabkmsmm.py
# Topologically Sorted Source Nodes: [pow__1, off_diag_loss, mul, loss], Original ATen: [aten.pow, aten.sum, aten.mul, aten.add]
# Source node to ATen node mapping:
# loss => add_3
# mul => mul
# off_diag_loss => sum_2
# pow__1 => pow_4
# Graph fragment:
# %pow_4 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%div_2, 2), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%pow_4,), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sum_2, 1.0), kwargs = {})
# %add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sum_1, %mul), kwargs = {})
triton_per_fused_add_mul_pow_sum_6 = async_compile.triton('triton_per_fused_add_mul_pow_sum_6', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 16],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_mul_pow_sum_6', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_mul_pow_sum_6(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel, XBLOCK : tl.constexpr):
xnumel = 1
rnumel = 16
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
xmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
roffset = 0
rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp5 = tl.load(in_ptr1 + (0))
tmp6 = tl.broadcast_to(tmp5, [XBLOCK, 1])
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.sum(tmp2, 1)[:, None]
tmp7 = 1.0
tmp8 = tmp4 * tmp7
tmp9 = tmp6 + tmp8
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([XBLOCK, 1], 0, tl.int32)), tmp9, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4), (4, 1))
assert_size_stride(arg1_1, (4, 4), (4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mean, sub, var, add, pow_1, z_left], Original ATen: [aten.mean, aten.sub, aten.var, aten.add, aten.pow, aten.div]
stream0 = get_raw_stream(0)
triton_poi_fused_add_div_mean_pow_sub_var_0.run(arg0_1, buf0, 16, grid=grid(16), stream=stream0)
del arg0_1
buf1 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mean_1, sub_1, var_1, add_1, pow_2, z_right], Original ATen: [aten.mean, aten.sub, aten.var, aten.add, aten.pow, aten.div]
triton_poi_fused_add_div_mean_pow_sub_var_0.run(arg1_1, buf1, 16, grid=grid(16), stream=stream0)
del arg1_1
buf2 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mean_1, sub_1, var_1, add_1, pow_2, z_right, matmul], Original ATen: [aten.mean, aten.sub, aten.var, aten.add, aten.pow, aten.div, aten.mm]
extern_kernels.mm(reinterpret_tensor(buf0, (4, 4), (1, 4), 0), buf1, out=buf2)
del buf0
buf3 = buf2; del buf2 # reuse
# Topologically Sorted Source Nodes: [cross_correlation], Original ATen: [aten.div]
triton_poi_fused_div_1.run(buf3, 16, grid=grid(16), stream=stream0)
buf4 = buf1; del buf1 # reuse
# Topologically Sorted Source Nodes: [], Original ATen: []
triton_poi_fused_2.run(buf3, buf4, 16, grid=grid(16), stream=stream0)
# Topologically Sorted Source Nodes: [add_], Original ATen: [aten.add]
triton_poi_fused_add_3.run(buf3, buf4, 4, grid=grid(4), stream=stream0)
buf8 = empty_strided_cuda((), (), torch.float32)
# Topologically Sorted Source Nodes: [pow_, on_diag_loss], Original ATen: [aten.pow, aten.sum]
triton_per_fused_pow_sum_4.run(buf4, buf4, buf8, 1, 4, grid=grid(1), stream=stream0)
# Topologically Sorted Source Nodes: [off_diag], Original ATen: [aten.fill]
triton_poi_fused_fill_5.run(buf3, 4, grid=grid(4), stream=stream0)
del buf4
buf11 = empty_strided_cuda((), (), torch.float32)
buf12 = buf11; del buf11 # reuse
# Topologically Sorted Source Nodes: [pow__1, off_diag_loss, mul, loss], Original ATen: [aten.pow, aten.sum, aten.mul, aten.add]
triton_per_fused_add_mul_pow_sum_6.run(buf12, buf3, buf8, 1, 16, grid=grid(1), stream=stream0)
del buf3
del buf8
return (buf12, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
from torch import nn
from torch.nn.modules.loss import *
from torch.nn.modules import *
from torch.optim import *
from torch.optim.lr_scheduler import *
import torch.backends
class BarlowTwinsLoss(nn.Module):
"""The Contrastive embedding loss.
It has been proposed in `Barlow Twins:
Self-Supervised Learning via Redundancy Reduction`_.
Example:
.. code-block:: python
import torch
from torch.nn import functional as F
from catalyst.contrib.nn import BarlowTwinsLoss
embeddings_left = F.normalize(torch.rand(256, 64, requires_grad=True))
embeddings_right = F.normalize(torch.rand(256, 64, requires_grad=True))
criterion = BarlowTwinsLoss(offdiag_lambda = 1)
criterion(embeddings_left, embeddings_right)
.. _`Barlow Twins: Self-Supervised Learning via Redundancy Reduction`:
https://arxiv.org/abs/2103.03230
"""
def __init__(self, offdiag_lambda=1.0, eps=1e-12):
"""
Args:
offdiag_lambda: trade-off parameter
eps: shift for the varience (var + eps)
"""
super().__init__()
self.offdiag_lambda = offdiag_lambda
self.eps = eps
def forward(self, embeddings_left: 'torch.Tensor', embeddings_right:
'torch.Tensor') ->torch.Tensor:
"""Forward propagation method for the contrastive loss.
Args:
embeddings_left: left objects embeddings [batch_size, features_dim]
embeddings_right: right objects embeddings [batch_size, features_dim]
Raises:
ValueError: if the batch size is 1
ValueError: if embeddings_left and embeddings_right shapes are different
ValueError: if embeddings shapes are not in a form (batch_size, features_dim)
Returns:
torch.Tensor: loss
"""
shape_left, shape_right = embeddings_left.shape, embeddings_right.shape
if len(shape_left) != 2:
raise ValueError(
f'Left shape should be (batch_size, feature_dim), but got - {shape_left}!'
)
elif len(shape_right) != 2:
raise ValueError(
f'Right shape should be (batch_size, feature_dim), but got - {shape_right}!'
)
if shape_left[0] == 1:
raise ValueError(
f'Batch size should be >= 2, but got - {shape_left[0]}!')
if shape_left != shape_right:
raise ValueError(
f'Shapes should be equall, but got - {shape_left} and {shape_right}!'
)
z_left = (embeddings_left - embeddings_left.mean(dim=0)) / (
embeddings_left.var(dim=0) + self.eps).pow(1 / 2)
z_right = (embeddings_right - embeddings_right.mean(dim=0)) / (
embeddings_right.var(dim=0) + self.eps).pow(1 / 2)
batch_size = z_left.shape[0]
cross_correlation = torch.matmul(z_left.T, z_right) / batch_size
on_diag = torch.diagonal(cross_correlation)
off_diag = cross_correlation.clone().fill_diagonal_(0)
on_diag_loss = on_diag.add_(-1).pow_(2).sum()
off_diag_loss = off_diag.pow_(2).sum()
loss = on_diag_loss + self.offdiag_lambda * off_diag_loss
return loss
def get_inputs():
return [torch.rand([4, 4]), torch.rand([4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice
from torch import nn
from torch.nn.modules.loss import *
from torch.nn.modules import *
from torch.optim import *
from torch.optim.lr_scheduler import *
import torch.backends
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_add_div_mean_pow_sub_var_0(in_ptr0, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex
x0 = xindex % 4
tmp0 = tl.load(in_ptr0 + x2, xmask)
tmp1 = tl.load(in_ptr0 + x0, xmask, eviction_policy='evict_last')
tmp2 = tl.load(in_ptr0 + (4 + x0), xmask, eviction_policy='evict_last')
tmp4 = tl.load(in_ptr0 + (8 + x0), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (12 + x0), xmask, eviction_policy='evict_last')
tmp3 = tmp1 + tmp2
tmp5 = tmp3 + tmp4
tmp7 = tmp5 + tmp6
tmp8 = 4.0
tmp9 = tmp7 / tmp8
tmp10 = tmp0 - tmp9
tmp11 = tmp1 - tmp9
tmp12 = tmp11 * tmp11
tmp13 = tmp2 - tmp9
tmp14 = tmp13 * tmp13
tmp15 = tmp12 + tmp14
tmp16 = tmp4 - tmp9
tmp17 = tmp16 * tmp16
tmp18 = tmp15 + tmp17
tmp19 = tmp6 - tmp9
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp22 = 3.0
tmp23 = tmp21 / tmp22
tmp24 = 1e-12
tmp25 = tmp23 + tmp24
tmp26 = libdevice.sqrt(tmp25)
tmp27 = tmp10 / tmp26
tl.store(out_ptr0 + x2, tmp27, xmask)
@triton.jit
def triton_poi_fused_div_1(in_out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_out_ptr0 + x0, xmask)
tmp1 = 0.25
tmp2 = tmp0 * tmp1
tl.store(in_out_ptr0 + x0, tmp2, xmask)
@triton.jit
def triton_poi_fused_2(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 16
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tl.store(out_ptr0 + x0, tmp0, xmask)
@triton.jit
def triton_poi_fused_add_3(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 5 * x0, xmask, eviction_policy='evict_last')
tmp1 = -1.0
tmp2 = tmp0 + tmp1
tl.store(out_ptr0 + 5 * x0, tmp2, xmask)
@triton.jit
def triton_per_fused_pow_sum_4(in_ptr0, out_ptr1, out_ptr2, xnumel, rnumel,
XBLOCK: tl.constexpr):
RBLOCK: tl.constexpr = 4
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + 5 * r0, None, eviction_policy='evict_last')
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.sum(tmp2, 1)[:, None]
tl.store(out_ptr1 + tl.broadcast_to(5 * r0, [XBLOCK, RBLOCK]), tmp1, None)
tl.store(out_ptr2 + tl.full([XBLOCK, 1], 0, tl.int32), tmp4, None)
@triton.jit
def triton_poi_fused_fill_5(out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 4
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = 0.0
tl.store(out_ptr0 + 5 * x0, tmp0, xmask)
@triton.jit
def triton_per_fused_add_mul_pow_sum_6(in_out_ptr0, in_ptr0, in_ptr1,
xnumel, rnumel, XBLOCK: tl.constexpr):
RBLOCK: tl.constexpr = 16
xoffset = tl.program_id(0) * XBLOCK
xoffset + tl.arange(0, XBLOCK)[:, None]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[None, :]
tl.full([XBLOCK, RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp5 = tl.load(in_ptr1 + 0)
tmp6 = tl.broadcast_to(tmp5, [XBLOCK, 1])
tmp1 = tmp0 * tmp0
tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
tmp4 = tl.sum(tmp2, 1)[:, None]
tmp7 = 1.0
tmp8 = tmp4 * tmp7
tmp9 = tmp6 + tmp8
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([XBLOCK, 1], 0, tl.int32), tmp9, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4), (4, 1))
assert_size_stride(arg1_1, (4, 4), (4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_div_mean_pow_sub_var_0[grid(16)](arg0_1, buf0,
16, XBLOCK=16, num_warps=1, num_stages=1)
del arg0_1
buf1 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
triton_poi_fused_add_div_mean_pow_sub_var_0[grid(16)](arg1_1, buf1,
16, XBLOCK=16, num_warps=1, num_stages=1)
del arg1_1
buf2 = empty_strided_cuda((4, 4), (4, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(buf0, (4, 4), (1, 4), 0), buf1,
out=buf2)
del buf0
buf3 = buf2
del buf2
triton_poi_fused_div_1[grid(16)](buf3, 16, XBLOCK=16, num_warps=1,
num_stages=1)
buf4 = buf1
del buf1
triton_poi_fused_2[grid(16)](buf3, buf4, 16, XBLOCK=16, num_warps=1,
num_stages=1)
triton_poi_fused_add_3[grid(4)](buf3, buf4, 4, XBLOCK=4, num_warps=
1, num_stages=1)
buf8 = empty_strided_cuda((), (), torch.float32)
triton_per_fused_pow_sum_4[grid(1)](buf4, buf4, buf8, 1, 4, XBLOCK=
1, num_warps=2, num_stages=1)
triton_poi_fused_fill_5[grid(4)](buf3, 4, XBLOCK=4, num_warps=1,
num_stages=1)
del buf4
buf11 = empty_strided_cuda((), (), torch.float32)
buf12 = buf11
del buf11
triton_per_fused_add_mul_pow_sum_6[grid(1)](buf12, buf3, buf8, 1,
16, XBLOCK=1, num_warps=2, num_stages=1)
del buf3
del buf8
return buf12,
class BarlowTwinsLossNew(nn.Module):
"""The Contrastive embedding loss.
It has been proposed in `Barlow Twins:
Self-Supervised Learning via Redundancy Reduction`_.
Example:
.. code-block:: python
import torch
from torch.nn import functional as F
from catalyst.contrib.nn import BarlowTwinsLoss
embeddings_left = F.normalize(torch.rand(256, 64, requires_grad=True))
embeddings_right = F.normalize(torch.rand(256, 64, requires_grad=True))
criterion = BarlowTwinsLoss(offdiag_lambda = 1)
criterion(embeddings_left, embeddings_right)
.. _`Barlow Twins: Self-Supervised Learning via Redundancy Reduction`:
https://arxiv.org/abs/2103.03230
"""
def __init__(self, offdiag_lambda=1.0, eps=1e-12):
"""
Args:
offdiag_lambda: trade-off parameter
eps: shift for the varience (var + eps)
"""
super().__init__()
self.offdiag_lambda = offdiag_lambda
self.eps = eps
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
gavrin-s/catalyst
|
BarlowTwinsLoss
| false | 12,414 |
[
"Apache-2.0"
] | 0 |
81087d8348b359e501d899f7a8350e0bedfc2b7d
|
https://github.com/gavrin-s/catalyst/tree/81087d8348b359e501d899f7a8350e0bedfc2b7d
|
Bar
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/3u/c3u4235hw5fuoyi7zvmqpjvnxqx5asdkyjlt4thtl6ajywk3ptav.py
# Topologically Sorted Source Nodes: [mul, add], Original ATen: [aten.mul, aten.add]
# Source node to ATen node mapping:
# add => add
# mul => mul
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, %arg1_1), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%mul, 4), kwargs = {})
triton_poi_fused_add_mul_0 = async_compile.triton('triton_poi_fused_add_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_mul_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask)
tmp2 = tmp0 * tmp1
tmp3 = 4.0
tmp4 = tmp2 + tmp3
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, add], Original ATen: [aten.mul, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_mul_0.run(arg0_1, arg1_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
del arg1_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.onnx
import torch.nn
class Bar(torch.nn.Module):
def __init__(self, x):
super(Bar, self).__init__()
self.x = x
def forward(self, a, b):
return a * b + self.x
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'x': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.onnx
import torch.nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_mul_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask)
tmp2 = tmp0 * tmp1
tmp3 = 4.0
tmp4 = tmp2 + tmp3
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_mul_0[grid(256)](arg0_1, arg1_1, buf0, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
del arg1_1
return buf0,
class BarNew(torch.nn.Module):
def __init__(self, x):
super(BarNew, self).__init__()
self.x = x
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
geoffberry/glow
|
Bar
| false | 12,415 |
[
"Apache-2.0"
] | 0 |
24b2827c830eb58af56a0704e899968026832e9c
|
https://github.com/geoffberry/glow/tree/24b2827c830eb58af56a0704e899968026832e9c
|
SimpleStackModel
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ju/cjuhkexaq3cgnr573mmdexu24x2qnvpxj6i5p4x2fwqe7eobr4vz.py
# Topologically Sorted Source Nodes: [stack_2], Original ATen: [aten.stack]
# Source node to ATen node mapping:
# stack_2 => clone_2
# Graph fragment:
# %clone_2 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%expand_1,), kwargs = {memory_format: torch.contiguous_format})
triton_poi_fused_stack_0 = async_compile.triton('triton_poi_fused_stack_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[2048],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_stack_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_stack_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 2048
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x1 = (xindex // 64) % 4
x3 = (xindex // 1024)
x0 = xindex % 64
x4 = xindex
tmp0 = x1 + (4*x3)
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (64*(x1 + (4*x3)))), tmp4, eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tmp7 = tl.full([1], 8, tl.int64)
tmp8 = tmp0 < tmp7
tmp9 = tl.load(in_ptr1 + (x0 + (64*((-4) + x1 + (4*x3)))), tmp6, eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + (x4), tmp10, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((2, 2, 2, 4, 4, 4, 4), (1024, 512, 256, 64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [stack_2], Original ATen: [aten.stack]
stream0 = get_raw_stream(0)
triton_poi_fused_stack_0.run(arg0_1, arg1_1, buf0, 2048, grid=grid(2048), stream=stream0)
del arg0_1
del arg1_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.onnx
import torch.nn
class SimpleStackModel(torch.nn.Module):
def __init__(self):
super(SimpleStackModel, self).__init__()
def forward(self, a, b):
c = torch.stack((a, b), 0)
d = torch.stack((c, c), 1)
return torch.stack((d, d), 2)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.onnx
import torch.nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_stack_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK: tl
.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x1 = xindex // 64 % 4
x3 = xindex // 1024
x0 = xindex % 64
x4 = xindex
tmp0 = x1 + 4 * x3
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 4, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 64 * (x1 + 4 * x3)), tmp4,
eviction_policy='evict_last', other=0.0)
tmp6 = tmp0 >= tmp3
tl.full([1], 8, tl.int64)
tmp9 = tl.load(in_ptr1 + (x0 + 64 * (-4 + x1 + 4 * x3)), tmp6,
eviction_policy='evict_last', other=0.0)
tmp10 = tl.where(tmp4, tmp5, tmp9)
tl.store(out_ptr0 + x4, tmp10, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((2, 2, 2, 4, 4, 4, 4), (1024, 512, 256,
64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_stack_0[grid(2048)](arg0_1, arg1_1, buf0, 2048,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
del arg1_1
return buf0,
class SimpleStackModelNew(torch.nn.Module):
def __init__(self):
super(SimpleStackModelNew, self).__init__()
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
geoffberry/glow
|
SimpleStackModel
| false | 12,416 |
[
"Apache-2.0"
] | 0 |
24b2827c830eb58af56a0704e899968026832e9c
|
https://github.com/geoffberry/glow/tree/24b2827c830eb58af56a0704e899968026832e9c
|
Baz
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/3p/c3p5dg3iyqj27rwgxtwdonyxpi46oayqucilggoubv6fbzcw5uhp.py
# Topologically Sorted Source Nodes: [mul, add], Original ATen: [aten.mul, aten.add]
# Source node to ATen node mapping:
# add => add
# mul => mul
# Graph fragment:
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg0_1, 4), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arg1_1, %mul), kwargs = {})
triton_poi_fused_add_mul_0 = async_compile.triton('triton_poi_fused_add_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_add_mul_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask)
tmp2 = 4.0
tmp3 = tmp1 * tmp2
tmp4 = tmp0 + tmp3
tl.store(out_ptr0 + (x0), tmp4, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mul, add], Original ATen: [aten.mul, aten.add]
stream0 = get_raw_stream(0)
triton_poi_fused_add_mul_0.run(arg1_1, arg0_1, buf0, 256, grid=grid(256), stream=stream0)
del arg0_1
del arg1_1
return (buf0, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.onnx
import torch.nn
class Baz(torch.nn.Module):
def __init__(self, x):
super(Baz, self).__init__()
self.x = x
def forward(self, a, b):
return a + b * self.x
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'x': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch.onnx
import torch.nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_add_mul_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask)
tmp2 = 4.0
tmp3 = tmp1 * tmp2
tmp4 = tmp0 + tmp3
tl.store(out_ptr0 + x0, tmp4, xmask)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_add_mul_0[grid(256)](arg1_1, arg0_1, buf0, 256,
XBLOCK=256, num_warps=4, num_stages=1)
del arg0_1
del arg1_1
return buf0,
class BazNew(torch.nn.Module):
def __init__(self, x):
super(BazNew, self).__init__()
self.x = x
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
geoffberry/glow
|
Baz
| false | 12,417 |
[
"Apache-2.0"
] | 0 |
24b2827c830eb58af56a0704e899968026832e9c
|
https://github.com/geoffberry/glow/tree/24b2827c830eb58af56a0704e899968026832e9c
|
Conditional_Contrastive_loss_plus
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/ik/ciknrrp6dre74dpeiauduto7u6b23dfn67nxh7k442j4zhxjlrep.py
# Topologically Sorted Source Nodes: [v, v_1], Original ATen: [aten.linalg_vector_norm, aten.clamp_min, aten.div, aten.mul]
# Source node to ATen node mapping:
# v => clamp_min, clamp_min_1, div, div_1, mul, pow_1, pow_2, pow_3, pow_4, sum_1, sum_2
# v_1 => clamp_min_2, clamp_min_3, div_2, div_3, mul_1, pow_5, pow_6, pow_7, pow_8, sum_4, sum_5
# Graph fragment:
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%expand_1, 2), kwargs = {})
# %sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_1, [-1], True), kwargs = {})
# %pow_2 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_1, 0.5), kwargs = {})
# %clamp_min : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%pow_2, 1e-08), kwargs = {})
# %div_1 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%expand_1, %clamp_min), kwargs = {})
# %pow_3 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%expand, 2), kwargs = {})
# %sum_2 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_3, [-1], True), kwargs = {})
# %pow_4 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_2, 0.5), kwargs = {})
# %clamp_min_1 : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%pow_4, 1e-08), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%expand, %clamp_min_1), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%div_1, %div), kwargs = {})
# %pow_5 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%expand_3, 2), kwargs = {})
# %sum_4 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_5, [-1], True), kwargs = {})
# %pow_6 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_4, 0.5), kwargs = {})
# %clamp_min_2 : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%pow_6, 1e-08), kwargs = {})
# %div_3 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%expand_3, %clamp_min_2), kwargs = {})
# %pow_7 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%expand_2, 2), kwargs = {})
# %sum_5 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%pow_7, [-1], True), kwargs = {})
# %pow_8 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sum_5, 0.5), kwargs = {})
# %clamp_min_3 : [num_users=1] = call_function[target=torch.ops.aten.clamp_min.default](args = (%pow_8, 1e-08), kwargs = {})
# %div_2 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%expand_2, %clamp_min_3), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%div_3, %div_2), kwargs = {})
triton_poi_fused_clamp_min_div_linalg_vector_norm_mul_0 = async_compile.triton('triton_poi_fused_clamp_min_div_linalg_vector_norm_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[1024],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_clamp_min_div_linalg_vector_norm_mul_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 15, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_clamp_min_div_linalg_vector_norm_mul_0(in_ptr0, in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = (xindex // 256)
x4 = xindex % 64
x1 = (xindex // 4) % 16
x5 = xindex % 256
x6 = (xindex // 4) % 64
x7 = xindex
tmp0 = tl.load(in_ptr0 + (x4 + (64*x3)), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + ((4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (1 + (4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp6 = tl.load(in_ptr0 + (2 + (4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp9 = tl.load(in_ptr0 + (3 + (4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp16 = tl.load(in_ptr1 + (x5), xmask, eviction_policy='evict_last')
tmp17 = tl.load(in_ptr1 + (4*x6), xmask, eviction_policy='evict_last')
tmp19 = tl.load(in_ptr1 + (1 + (4*x6)), xmask, eviction_policy='evict_last')
tmp22 = tl.load(in_ptr1 + (2 + (4*x6)), xmask, eviction_policy='evict_last')
tmp25 = tl.load(in_ptr1 + (3 + (4*x6)), xmask, eviction_policy='evict_last')
tmp32 = tl.load(in_ptr1 + (x4 + (64*x3)), xmask, eviction_policy='evict_last')
tmp33 = tl.load(in_ptr1 + ((4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp35 = tl.load(in_ptr1 + (1 + (4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp38 = tl.load(in_ptr1 + (2 + (4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp41 = tl.load(in_ptr1 + (3 + (4*x1) + (64*x3)), xmask, eviction_policy='evict_last')
tmp2 = tmp1 * tmp1
tmp4 = tmp3 * tmp3
tmp5 = tmp2 + tmp4
tmp7 = tmp6 * tmp6
tmp8 = tmp5 + tmp7
tmp10 = tmp9 * tmp9
tmp11 = tmp8 + tmp10
tmp12 = libdevice.sqrt(tmp11)
tmp13 = 1e-08
tmp14 = triton_helpers.maximum(tmp12, tmp13)
tmp15 = tmp0 / tmp14
tmp18 = tmp17 * tmp17
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp23 = tmp22 * tmp22
tmp24 = tmp21 + tmp23
tmp26 = tmp25 * tmp25
tmp27 = tmp24 + tmp26
tmp28 = libdevice.sqrt(tmp27)
tmp29 = triton_helpers.maximum(tmp28, tmp13)
tmp30 = tmp16 / tmp29
tmp31 = tmp15 * tmp30
tmp34 = tmp33 * tmp33
tmp36 = tmp35 * tmp35
tmp37 = tmp34 + tmp36
tmp39 = tmp38 * tmp38
tmp40 = tmp37 + tmp39
tmp42 = tmp41 * tmp41
tmp43 = tmp40 + tmp42
tmp44 = libdevice.sqrt(tmp43)
tmp45 = triton_helpers.maximum(tmp44, tmp13)
tmp46 = tmp32 / tmp45
tmp47 = tmp46 * tmp30
tl.store(out_ptr0 + (x7), tmp31, xmask)
tl.store(out_ptr1 + (x7), tmp47, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ka/ckal3vcwdqfggevtu5rxezv764uurq7s5yy6lkgcwtiwsild5vck.py
# Topologically Sorted Source Nodes: [v, sub, truediv, p2i_similarity_zone, v_1, sub_1, truediv_1, i2i_similarity_zone], Original ATen: [aten.sum, aten.sub, aten.div, aten.exp]
# Source node to ATen node mapping:
# i2i_similarity_zone => exp_1
# p2i_similarity_zone => exp
# sub => sub
# sub_1 => sub_1
# truediv => div_4
# truediv_1 => div_5
# v => sum_3
# v_1 => sum_6
# Graph fragment:
# %sum_3 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul, [-1]), kwargs = {})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sum_3, %arg2_1), kwargs = {})
# %div_4 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub, %arg3_1), kwargs = {})
# %exp : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%div_4,), kwargs = {})
# %sum_6 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul_1, [-1]), kwargs = {})
# %sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sum_6, %arg2_1), kwargs = {})
# %div_5 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sub_1, %arg3_1), kwargs = {})
# %exp_1 : [num_users=2] = call_function[target=torch.ops.aten.exp.default](args = (%div_5,), kwargs = {})
triton_poi_fused_div_exp_sub_sum_1 = async_compile.triton('triton_poi_fused_div_exp_sub_sum_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_exp_sub_sum_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 10, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_exp_sub_sum_1(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (4*x0), xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + (x0), xmask)
tmp9 = tl.load(in_ptr2 + (x0), xmask)
tmp12 = tl.load(in_ptr3 + (4*x0), xmask, eviction_policy='evict_last')
tmp13 = tl.load(in_ptr3 + (1 + (4*x0)), xmask, eviction_policy='evict_last')
tmp15 = tl.load(in_ptr3 + (2 + (4*x0)), xmask, eviction_policy='evict_last')
tmp17 = tl.load(in_ptr3 + (3 + (4*x0)), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp8 = tmp6 - tmp7
tmp10 = tmp8 / tmp9
tmp11 = tl_math.exp(tmp10)
tmp14 = tmp12 + tmp13
tmp16 = tmp14 + tmp15
tmp18 = tmp16 + tmp17
tmp19 = tmp18 - tmp7
tmp20 = tmp19 / tmp9
tmp21 = tl_math.exp(tmp20)
tl.store(out_ptr0 + (x0), tmp11, xmask)
tl.store(out_ptr1 + (x0), tmp21, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2x/c2xfmuwypi6t35j64lr3jvkazb5jja23doiausp2ko7z6utscfdh.py
# Topologically Sorted Source Nodes: [mask_4_remove_negatives, p2i_positives, p2i_numerator, p2i_denomerator, truediv_2, i2i_positives, i2i_numerator, i2i_denomerator, truediv_3], Original ATen: [aten.index, aten.mul, aten.sum, aten.div]
# Source node to ATen node mapping:
# i2i_denomerator => sum_10
# i2i_numerator => sum_8
# i2i_positives => mul_3
# mask_4_remove_negatives => index
# p2i_denomerator => sum_9
# p2i_numerator => sum_7
# p2i_positives => mul_2
# truediv_2 => div_6
# truediv_3 => div_7
# Graph fragment:
# %index : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%arg4_1, [%arg5_1]), kwargs = {})
# %mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%exp, %index), kwargs = {})
# %sum_7 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul_2, [1]), kwargs = {})
# %sum_9 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [1]), kwargs = {})
# %div_6 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sum_7, %sum_9), kwargs = {})
# %mul_3 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%exp_1, %index), kwargs = {})
# %sum_8 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul_3, [1]), kwargs = {})
# %sum_10 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_1, [1]), kwargs = {})
# %div_7 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sum_8, %sum_10), kwargs = {})
triton_poi_fused_div_index_mul_sum_2 = async_compile.triton('triton_poi_fused_div_index_mul_sum_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*i64', 2: '*i64', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_div_index_mul_sum_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 9, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_div_index_mul_sum_2(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = (xindex // 16)
x3 = xindex % 16
x0 = xindex % 4
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x3 + (64*x2)), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (16 + x3 + (64*x2)), xmask)
tmp13 = tl.load(in_ptr0 + (32 + x3 + (64*x2)), xmask)
tmp16 = tl.load(in_ptr0 + (48 + x3 + (64*x2)), xmask)
tmp23 = tl.load(in_ptr3 + (x3 + (64*x2)), xmask)
tmp25 = tl.load(in_ptr3 + (16 + x3 + (64*x2)), xmask)
tmp28 = tl.load(in_ptr3 + (32 + x3 + (64*x2)), xmask)
tmp31 = tl.load(in_ptr3 + (48 + x3 + (64*x2)), xmask)
tmp2 = tl.full([XBLOCK], 4, tl.int32)
tmp3 = tmp1 + tmp2
tmp4 = tmp1 < 0
tmp5 = tl.where(tmp4, tmp3, tmp1)
tl.device_assert(((0 <= tmp5) & (tmp5 < 4)) | ~(xmask), "index out of bounds: 0 <= tmp5 < 4")
tmp7 = tl.load(in_ptr2 + (tmp5), xmask, eviction_policy='evict_last')
tmp8 = tmp7.to(tl.float32)
tmp9 = tmp0 * tmp8
tmp11 = tmp10 * tmp8
tmp12 = tmp9 + tmp11
tmp14 = tmp13 * tmp8
tmp15 = tmp12 + tmp14
tmp17 = tmp16 * tmp8
tmp18 = tmp15 + tmp17
tmp19 = tmp0 + tmp10
tmp20 = tmp19 + tmp13
tmp21 = tmp20 + tmp16
tmp22 = tmp18 / tmp21
tmp24 = tmp23 * tmp8
tmp26 = tmp25 * tmp8
tmp27 = tmp24 + tmp26
tmp29 = tmp28 * tmp8
tmp30 = tmp27 + tmp29
tmp32 = tmp31 * tmp8
tmp33 = tmp30 + tmp32
tmp34 = tmp23 + tmp25
tmp35 = tmp34 + tmp28
tmp36 = tmp35 + tmp31
tmp37 = tmp33 / tmp36
tl.store(out_ptr0 + (x4), tmp22, xmask)
tl.store(out_ptr1 + (x4), tmp37, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/wp/cwpoyahk3rtja5g2hrfgdkkwv7legjidgnwrdp4ql35n6dk6tuel.py
# Topologically Sorted Source Nodes: [mask_4_remove_negatives, p2i_positives, p2i_numerator, p2i_denomerator, truediv_2, mul_2, log, mean, p2i_contra_loss, i2i_positives, i2i_numerator, i2i_denomerator, truediv_3, mul_3, log_1, mean_1, i2i_contra_loss, add], Original ATen: [aten.index, aten.mul, aten.sum, aten.div, aten.log, aten.mean, aten.neg, aten.add]
# Source node to ATen node mapping:
# add => add
# i2i_contra_loss => neg_1
# i2i_denomerator => sum_10
# i2i_numerator => sum_8
# i2i_positives => mul_3
# log => log
# log_1 => log_1
# mask_4_remove_negatives => index
# mean => mean
# mean_1 => mean_1
# mul_2 => mul_4
# mul_3 => mul_5
# p2i_contra_loss => neg
# p2i_denomerator => sum_9
# p2i_numerator => sum_7
# p2i_positives => mul_2
# truediv_2 => div_6
# truediv_3 => div_7
# Graph fragment:
# %index : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%arg4_1, [%arg5_1]), kwargs = {})
# %mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%exp, %index), kwargs = {})
# %sum_7 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul_2, [1]), kwargs = {})
# %sum_9 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp, [1]), kwargs = {})
# %div_6 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sum_7, %sum_9), kwargs = {})
# %mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg3_1, %div_6), kwargs = {})
# %log : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%mul_4,), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%log,), kwargs = {})
# %neg : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%mean,), kwargs = {})
# %mul_3 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%exp_1, %index), kwargs = {})
# %sum_8 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%mul_3, [1]), kwargs = {})
# %sum_10 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%exp_1, [1]), kwargs = {})
# %div_7 : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%sum_8, %sum_10), kwargs = {})
# %mul_5 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%arg3_1, %div_7), kwargs = {})
# %log_1 : [num_users=1] = call_function[target=torch.ops.aten.log.default](args = (%mul_5,), kwargs = {})
# %mean_1 : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%log_1,), kwargs = {})
# %neg_1 : [num_users=1] = call_function[target=torch.ops.aten.neg.default](args = (%mean_1,), kwargs = {})
# %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%neg, %neg_1), kwargs = {})
triton_per_fused_add_div_index_log_mean_mul_neg_sum_3 = async_compile.triton('triton_per_fused_add_div_index_log_mean_mul_neg_sum_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: 'i32', 5: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {4: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 5), equal_to_1=(4,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_add_div_index_log_mean_mul_neg_sum_3', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 3, 'num_reduction': 2, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_add_div_index_log_mean_mul_neg_sum_3(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r2 = rindex
r0 = rindex % 64
tmp0 = tl.load(in_ptr0 + (r2), None)
tmp1 = tl.load(in_ptr1 + (r0), None, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr2 + (r0), None, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp3 = tl_math.log(tmp2)
tmp4 = tl.broadcast_to(tmp3, [RBLOCK])
tmp6 = triton_helpers.promote_to_tensor(tl.sum(tmp4, 0))
tmp8 = tmp0 * tmp7
tmp9 = tl_math.log(tmp8)
tmp10 = tl.broadcast_to(tmp9, [RBLOCK])
tmp12 = triton_helpers.promote_to_tensor(tl.sum(tmp10, 0))
tmp13 = 256.0
tmp14 = tmp6 / tmp13
tmp15 = -tmp14
tmp16 = tmp12 / tmp13
tmp17 = -tmp16
tmp18 = tmp15 + tmp17
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp18, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1, arg2_1, arg3_1, arg4_1, arg5_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg2_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg3_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg4_1, (4, ), (1, ))
assert_size_stride(arg5_1, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1), torch.float32)
buf4 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [v, v_1], Original ATen: [aten.linalg_vector_norm, aten.clamp_min, aten.div, aten.mul]
stream0 = get_raw_stream(0)
triton_poi_fused_clamp_min_div_linalg_vector_norm_mul_0.run(arg0_1, arg1_1, buf0, buf4, 1024, grid=grid(1024), stream=stream0)
del arg0_1
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
buf5 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [v, sub, truediv, p2i_similarity_zone, v_1, sub_1, truediv_1, i2i_similarity_zone], Original ATen: [aten.sum, aten.sub, aten.div, aten.exp]
triton_poi_fused_div_exp_sub_sum_1.run(buf0, arg2_1, arg3_1, buf4, buf1, buf5, 256, grid=grid(256), stream=stream0)
del arg2_1
del buf0
del buf4
buf2 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf6 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [mask_4_remove_negatives, p2i_positives, p2i_numerator, p2i_denomerator, truediv_2, i2i_positives, i2i_numerator, i2i_denomerator, truediv_3], Original ATen: [aten.index, aten.mul, aten.sum, aten.div]
triton_poi_fused_div_index_mul_sum_2.run(buf1, arg5_1, arg4_1, buf5, buf2, buf6, 64, grid=grid(64), stream=stream0)
del arg4_1
del arg5_1
del buf1
del buf5
buf3 = empty_strided_cuda((), (), torch.float32)
buf8 = buf3; del buf3 # reuse
# Topologically Sorted Source Nodes: [mask_4_remove_negatives, p2i_positives, p2i_numerator, p2i_denomerator, truediv_2, mul_2, log, mean, p2i_contra_loss, i2i_positives, i2i_numerator, i2i_denomerator, truediv_3, mul_3, log_1, mean_1, i2i_contra_loss, add], Original ATen: [aten.index, aten.mul, aten.sum, aten.div, aten.log, aten.mean, aten.neg, aten.add]
triton_per_fused_add_div_index_log_mean_mul_neg_sum_3.run(buf8, arg3_1, buf2, buf6, 1, 256, grid=grid(1), stream=stream0)
del arg3_1
del buf2
del buf6
return (buf8, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg2_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg3_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg4_1 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.int64)
arg5_1 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.int64)
fn = lambda: call([arg0_1, arg1_1, arg2_1, arg3_1, arg4_1, arg5_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import numpy as np
class Conditional_Contrastive_loss_plus(torch.nn.Module):
def __init__(self, device, batch_size, pos_collected_numerator):
super(Conditional_Contrastive_loss_plus, self).__init__()
self.device = device
self.batch_size = batch_size
self.pos_collected_numerator = pos_collected_numerator
self.calculate_similarity_matrix = self._calculate_similarity_matrix()
self.cosine_similarity = torch.nn.CosineSimilarity(dim=-1)
def _calculate_similarity_matrix(self):
return self._cosine_simililarity_matrix
def remove_diag(self, M):
h, w = M.shape
assert h == w, 'h and w should be same'
mask = np.ones((h, w)) - np.eye(h)
mask = torch.from_numpy(mask)
mask = mask.type(torch.bool)
return M[mask].view(h, -1)
def _cosine_simililarity_matrix(self, x, y):
v = self.cosine_similarity(x.unsqueeze(1), y.unsqueeze(0))
return v
def forward(self, inst_embed, proxy, negative_mask, labels, temperature,
margin):
p2i_similarity_matrix = self.calculate_similarity_matrix(proxy,
inst_embed)
i2i_similarity_matrix = self.calculate_similarity_matrix(inst_embed,
inst_embed)
p2i_similarity_zone = torch.exp((p2i_similarity_matrix - margin) /
temperature)
i2i_similarity_zone = torch.exp((i2i_similarity_matrix - margin) /
temperature)
mask_4_remove_negatives = negative_mask[labels]
p2i_positives = p2i_similarity_zone * mask_4_remove_negatives
i2i_positives = i2i_similarity_zone * mask_4_remove_negatives
p2i_numerator = p2i_positives.sum(dim=1)
i2i_numerator = i2i_positives.sum(dim=1)
p2i_denomerator = p2i_similarity_zone.sum(dim=1)
i2i_denomerator = i2i_similarity_zone.sum(dim=1)
p2i_contra_loss = -torch.log(temperature * (p2i_numerator /
p2i_denomerator)).mean()
i2i_contra_loss = -torch.log(temperature * (i2i_numerator /
i2i_denomerator)).mean()
return p2i_contra_loss + i2i_contra_loss
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4]), torch.ones(
[4], dtype=torch.int64), torch.ones([4], dtype=torch.int64), torch.
rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'device': 0, 'batch_size': 4, 'pos_collected_numerator': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import numpy as np
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_clamp_min_div_linalg_vector_norm_mul_0(in_ptr0,
in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 1024
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x3 = xindex // 256
x4 = xindex % 64
x1 = xindex // 4 % 16
x5 = xindex % 256
x6 = xindex // 4 % 64
x7 = xindex
tmp0 = tl.load(in_ptr0 + (x4 + 64 * x3), xmask, eviction_policy=
'evict_last')
tmp1 = tl.load(in_ptr0 + (4 * x1 + 64 * x3), xmask, eviction_policy=
'evict_last')
tmp3 = tl.load(in_ptr0 + (1 + 4 * x1 + 64 * x3), xmask, eviction_policy
='evict_last')
tmp6 = tl.load(in_ptr0 + (2 + 4 * x1 + 64 * x3), xmask, eviction_policy
='evict_last')
tmp9 = tl.load(in_ptr0 + (3 + 4 * x1 + 64 * x3), xmask, eviction_policy
='evict_last')
tmp16 = tl.load(in_ptr1 + x5, xmask, eviction_policy='evict_last')
tmp17 = tl.load(in_ptr1 + 4 * x6, xmask, eviction_policy='evict_last')
tmp19 = tl.load(in_ptr1 + (1 + 4 * x6), xmask, eviction_policy='evict_last'
)
tmp22 = tl.load(in_ptr1 + (2 + 4 * x6), xmask, eviction_policy='evict_last'
)
tmp25 = tl.load(in_ptr1 + (3 + 4 * x6), xmask, eviction_policy='evict_last'
)
tmp32 = tl.load(in_ptr1 + (x4 + 64 * x3), xmask, eviction_policy=
'evict_last')
tmp33 = tl.load(in_ptr1 + (4 * x1 + 64 * x3), xmask, eviction_policy=
'evict_last')
tmp35 = tl.load(in_ptr1 + (1 + 4 * x1 + 64 * x3), xmask,
eviction_policy='evict_last')
tmp38 = tl.load(in_ptr1 + (2 + 4 * x1 + 64 * x3), xmask,
eviction_policy='evict_last')
tmp41 = tl.load(in_ptr1 + (3 + 4 * x1 + 64 * x3), xmask,
eviction_policy='evict_last')
tmp2 = tmp1 * tmp1
tmp4 = tmp3 * tmp3
tmp5 = tmp2 + tmp4
tmp7 = tmp6 * tmp6
tmp8 = tmp5 + tmp7
tmp10 = tmp9 * tmp9
tmp11 = tmp8 + tmp10
tmp12 = libdevice.sqrt(tmp11)
tmp13 = 1e-08
tmp14 = triton_helpers.maximum(tmp12, tmp13)
tmp15 = tmp0 / tmp14
tmp18 = tmp17 * tmp17
tmp20 = tmp19 * tmp19
tmp21 = tmp18 + tmp20
tmp23 = tmp22 * tmp22
tmp24 = tmp21 + tmp23
tmp26 = tmp25 * tmp25
tmp27 = tmp24 + tmp26
tmp28 = libdevice.sqrt(tmp27)
tmp29 = triton_helpers.maximum(tmp28, tmp13)
tmp30 = tmp16 / tmp29
tmp31 = tmp15 * tmp30
tmp34 = tmp33 * tmp33
tmp36 = tmp35 * tmp35
tmp37 = tmp34 + tmp36
tmp39 = tmp38 * tmp38
tmp40 = tmp37 + tmp39
tmp42 = tmp41 * tmp41
tmp43 = tmp40 + tmp42
tmp44 = libdevice.sqrt(tmp43)
tmp45 = triton_helpers.maximum(tmp44, tmp13)
tmp46 = tmp32 / tmp45
tmp47 = tmp46 * tmp30
tl.store(out_ptr0 + x7, tmp31, xmask)
tl.store(out_ptr1 + x7, tmp47, xmask)
@triton.jit
def triton_poi_fused_div_exp_sub_sum_1(in_ptr0, in_ptr1, in_ptr2, in_ptr3,
out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + 4 * x0, xmask, eviction_policy='evict_last')
tmp1 = tl.load(in_ptr0 + (1 + 4 * x0), xmask, eviction_policy='evict_last')
tmp3 = tl.load(in_ptr0 + (2 + 4 * x0), xmask, eviction_policy='evict_last')
tmp5 = tl.load(in_ptr0 + (3 + 4 * x0), xmask, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr1 + x0, xmask)
tmp9 = tl.load(in_ptr2 + x0, xmask)
tmp12 = tl.load(in_ptr3 + 4 * x0, xmask, eviction_policy='evict_last')
tmp13 = tl.load(in_ptr3 + (1 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp15 = tl.load(in_ptr3 + (2 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp17 = tl.load(in_ptr3 + (3 + 4 * x0), xmask, eviction_policy='evict_last'
)
tmp2 = tmp0 + tmp1
tmp4 = tmp2 + tmp3
tmp6 = tmp4 + tmp5
tmp8 = tmp6 - tmp7
tmp10 = tmp8 / tmp9
tmp11 = tl_math.exp(tmp10)
tmp14 = tmp12 + tmp13
tmp16 = tmp14 + tmp15
tmp18 = tmp16 + tmp17
tmp19 = tmp18 - tmp7
tmp20 = tmp19 / tmp9
tmp21 = tl_math.exp(tmp20)
tl.store(out_ptr0 + x0, tmp11, xmask)
tl.store(out_ptr1 + x0, tmp21, xmask)
@triton.jit
def triton_poi_fused_div_index_mul_sum_2(in_ptr0, in_ptr1, in_ptr2, in_ptr3,
out_ptr0, out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x2 = xindex // 16
x3 = xindex % 16
x0 = xindex % 4
x4 = xindex
tmp0 = tl.load(in_ptr0 + (x3 + 64 * x2), xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask, eviction_policy='evict_last')
tmp10 = tl.load(in_ptr0 + (16 + x3 + 64 * x2), xmask)
tmp13 = tl.load(in_ptr0 + (32 + x3 + 64 * x2), xmask)
tmp16 = tl.load(in_ptr0 + (48 + x3 + 64 * x2), xmask)
tmp23 = tl.load(in_ptr3 + (x3 + 64 * x2), xmask)
tmp25 = tl.load(in_ptr3 + (16 + x3 + 64 * x2), xmask)
tmp28 = tl.load(in_ptr3 + (32 + x3 + 64 * x2), xmask)
tmp31 = tl.load(in_ptr3 + (48 + x3 + 64 * x2), xmask)
tmp2 = tl.full([XBLOCK], 4, tl.int32)
tmp3 = tmp1 + tmp2
tmp4 = tmp1 < 0
tmp5 = tl.where(tmp4, tmp3, tmp1)
tl.device_assert((0 <= tmp5) & (tmp5 < 4) | ~xmask,
'index out of bounds: 0 <= tmp5 < 4')
tmp7 = tl.load(in_ptr2 + tmp5, xmask, eviction_policy='evict_last')
tmp8 = tmp7.to(tl.float32)
tmp9 = tmp0 * tmp8
tmp11 = tmp10 * tmp8
tmp12 = tmp9 + tmp11
tmp14 = tmp13 * tmp8
tmp15 = tmp12 + tmp14
tmp17 = tmp16 * tmp8
tmp18 = tmp15 + tmp17
tmp19 = tmp0 + tmp10
tmp20 = tmp19 + tmp13
tmp21 = tmp20 + tmp16
tmp22 = tmp18 / tmp21
tmp24 = tmp23 * tmp8
tmp26 = tmp25 * tmp8
tmp27 = tmp24 + tmp26
tmp29 = tmp28 * tmp8
tmp30 = tmp27 + tmp29
tmp32 = tmp31 * tmp8
tmp33 = tmp30 + tmp32
tmp34 = tmp23 + tmp25
tmp35 = tmp34 + tmp28
tmp36 = tmp35 + tmp31
tmp37 = tmp33 / tmp36
tl.store(out_ptr0 + x4, tmp22, xmask)
tl.store(out_ptr1 + x4, tmp37, xmask)
@triton.jit
def triton_per_fused_add_div_index_log_mean_mul_neg_sum_3(in_out_ptr0,
in_ptr0, in_ptr1, in_ptr2, xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r2 = rindex
r0 = rindex % 64
tmp0 = tl.load(in_ptr0 + r2, None)
tmp1 = tl.load(in_ptr1 + r0, None, eviction_policy='evict_last')
tmp7 = tl.load(in_ptr2 + r0, None, eviction_policy='evict_last')
tmp2 = tmp0 * tmp1
tmp3 = tl_math.log(tmp2)
tmp4 = tl.broadcast_to(tmp3, [RBLOCK])
tmp6 = triton_helpers.promote_to_tensor(tl.sum(tmp4, 0))
tmp8 = tmp0 * tmp7
tmp9 = tl_math.log(tmp8)
tmp10 = tl.broadcast_to(tmp9, [RBLOCK])
tmp12 = triton_helpers.promote_to_tensor(tl.sum(tmp10, 0))
tmp13 = 256.0
tmp14 = tmp6 / tmp13
tmp15 = -tmp14
tmp16 = tmp12 / tmp13
tmp17 = -tmp16
tmp18 = tmp15 + tmp17
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp18, None)
def call(args):
arg0_1, arg1_1, arg2_1, arg3_1, arg4_1, arg5_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg2_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg3_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg4_1, (4,), (1,))
assert_size_stride(arg5_1, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1),
torch.float32)
buf4 = empty_strided_cuda((4, 4, 4, 4, 4), (256, 64, 16, 4, 1),
torch.float32)
get_raw_stream(0)
triton_poi_fused_clamp_min_div_linalg_vector_norm_mul_0[grid(1024)](
arg0_1, arg1_1, buf0, buf4, 1024, XBLOCK=128, num_warps=4,
num_stages=1)
del arg0_1
del arg1_1
buf1 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
buf5 = empty_strided_cuda((4, 4, 4, 4), (64, 16, 4, 1), torch.float32)
triton_poi_fused_div_exp_sub_sum_1[grid(256)](buf0, arg2_1, arg3_1,
buf4, buf1, buf5, 256, XBLOCK=128, num_warps=4, num_stages=1)
del arg2_1
del buf0
del buf4
buf2 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
buf6 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_div_index_mul_sum_2[grid(64)](buf1, arg5_1, arg4_1,
buf5, buf2, buf6, 64, XBLOCK=64, num_warps=1, num_stages=1)
del arg4_1
del arg5_1
del buf1
del buf5
buf3 = empty_strided_cuda((), (), torch.float32)
buf8 = buf3
del buf3
triton_per_fused_add_div_index_log_mean_mul_neg_sum_3[grid(1)](buf8,
arg3_1, buf2, buf6, 1, 256, num_warps=2, num_stages=1)
del arg3_1
del buf2
del buf6
return buf8,
class Conditional_Contrastive_loss_plusNew(torch.nn.Module):
def __init__(self, device, batch_size, pos_collected_numerator):
super(Conditional_Contrastive_loss_plusNew, self).__init__()
self.device = device
self.batch_size = batch_size
self.pos_collected_numerator = pos_collected_numerator
self.calculate_similarity_matrix = self._calculate_similarity_matrix()
self.cosine_similarity = torch.nn.CosineSimilarity(dim=-1)
def _calculate_similarity_matrix(self):
return self._cosine_simililarity_matrix
def remove_diag(self, M):
h, w = M.shape
assert h == w, 'h and w should be same'
mask = np.ones((h, w)) - np.eye(h)
mask = torch.from_numpy(mask)
mask = mask.type(torch.bool)
return M[mask].view(h, -1)
def _cosine_simililarity_matrix(self, x, y):
v = self.cosine_similarity(x.unsqueeze(1), y.unsqueeze(0))
return v
def forward(self, input_0, input_1, input_2, input_3, input_4, input_5):
arg0_1 = input_0
arg1_1 = input_1
arg4_1 = input_2
arg5_1 = input_3
arg2_1 = input_4
arg3_1 = input_5
output = call([arg0_1, arg1_1, arg2_1, arg3_1, arg4_1, arg5_1])
return output[0]
|
fywang0327/PyTorch-ECGAN
|
Conditional_Contrastive_loss_plus
| false | 12,418 |
[
"MIT"
] | 0 |
7c7c8c28c609b1bd2d3aecaeec4bffeb4c9cda6c
|
https://github.com/fywang0327/PyTorch-ECGAN/tree/7c7c8c28c609b1bd2d3aecaeec4bffeb4c9cda6c
|
MLP
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/cc/ccc3rfqzulhvn4nmibqhwjsomjzanxwwc34nr7d7enltyb7czw67.py
# Topologically Sorted Source Nodes: [relu], Original ATen: [aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_1,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_relu_threshold_backward_0 = async_compile.triton('triton_poi_fused_relu_threshold_backward_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32768],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*i1', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_threshold_backward_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_ptr0, in_ptr1, out_ptr0, out_ptr1, xnumel, XBLOCK : tl.constexpr):
xnumel = 32000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x0 = xindex % 500
x2 = (xindex // 2000)
x3 = xindex % 2000
tmp0 = tl.load(in_ptr0 + (x4), xmask)
tmp1 = tl.load(in_ptr1 + (x0), xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x3 + (2016*x2)), tmp4, xmask)
tl.store(out_ptr1 + (x3 + (2048*x2)), tmp6, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/od/cod4chcodxhfkpwgvalpo3wjcotr3qr3momwpd2myrvazuq2fy67.py
# Topologically Sorted Source Nodes: [relu, linear_1], Original ATen: [aten.relu, aten.view]
# Source node to ATen node mapping:
# linear_1 => view_2
# relu => relu
# Graph fragment:
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%view_1,), kwargs = {})
# %view_2 : [num_users=2] = call_function[target=torch.ops.aten.reshape.default](args = (%relu, [64, 500]), kwargs = {})
triton_poi_fused_relu_view_1 = async_compile.triton('triton_poi_fused_relu_view_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32768],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_relu_view_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_relu_view_1(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 32000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 500
x1 = (xindex // 500)
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + (500*(x1 % 4)) + (2016*(x1 // 4))), xmask)
tl.store(out_ptr0 + (x2), tmp0, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (500, 4), (4, 1))
assert_size_stride(primals_2, (500, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 500), (500, 1))
assert_size_stride(primals_5, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 500), (500, 1), torch.float32)
# Topologically Sorted Source Nodes: [], Original ATen: []
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), reinterpret_tensor(primals_1, (4, 500), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((4, 4, 4, 500), (8064, 2016, 500, 1), torch.float32)
buf4 = empty_strided_cuda((4, 4, 4, 500), (8192, 2048, 500, 1), torch.bool)
# Topologically Sorted Source Nodes: [relu], Original ATen: [aten.relu, aten.threshold_backward]
stream0 = get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0.run(buf0, primals_2, buf1, buf4, 32000, grid=grid(32000), stream=stream0)
del primals_2
buf2 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [relu, linear_1], Original ATen: [aten.relu, aten.view]
triton_poi_fused_relu_view_1.run(buf1, buf2, 32000, grid=grid(32000), stream=stream0)
del buf1
buf3 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [linear_1], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_5, buf2, reinterpret_tensor(primals_4, (500, 4), (1, 500), 0), alpha=1, beta=1, out=buf3)
del primals_5
return (reinterpret_tensor(buf3, (4, 4, 4, 4), (64, 16, 4, 1), 0), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0), buf2, primals_4, buf4, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((500, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((500, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((4, 500), (500, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
class MLP(nn.Module):
def __init__(self, input_size, output_size, hidden_size=500,
weight_decay=0.0):
super(MLP, self).__init__()
self.i2h = nn.Linear(in_features=input_size, out_features=hidden_size)
self.Dropout = nn.Dropout(p=0.5)
self.h2o = nn.Linear(in_features=hidden_size, out_features=output_size)
self.loss = nn.functional.cross_entropy
self.weight_decay = weight_decay
def forward(self, x):
return self.h2o(torch.relu(self.Dropout(self.i2h(x))))
def predict(self, x):
return self.forward(x).argmax(dim=1)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'input_size': 4, 'output_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_relu_threshold_backward_0(in_ptr0, in_ptr1, out_ptr0,
out_ptr1, xnumel, XBLOCK: tl.constexpr):
xnumel = 32000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x4 = xindex
x0 = xindex % 500
x2 = xindex // 2000
x3 = xindex % 2000
tmp0 = tl.load(in_ptr0 + x4, xmask)
tmp1 = tl.load(in_ptr1 + x0, xmask, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x3 + 2016 * x2), tmp4, xmask)
tl.store(out_ptr1 + (x3 + 2048 * x2), tmp6, xmask)
@triton.jit
def triton_poi_fused_relu_view_1(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.
constexpr):
xnumel = 32000
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 500
x1 = xindex // 500
x2 = xindex
tmp0 = tl.load(in_ptr0 + (x0 + 500 * (x1 % 4) + 2016 * (x1 // 4)), xmask)
tl.store(out_ptr0 + x2, tmp0, xmask)
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5 = args
args.clear()
assert_size_stride(primals_1, (500, 4), (4, 1))
assert_size_stride(primals_2, (500,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (4, 500), (500, 1))
assert_size_stride(primals_5, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((64, 500), (500, 1), torch.float32)
extern_kernels.mm(reinterpret_tensor(primals_3, (64, 4), (4, 1), 0),
reinterpret_tensor(primals_1, (4, 500), (1, 4), 0), out=buf0)
del primals_1
buf1 = empty_strided_cuda((4, 4, 4, 500), (8064, 2016, 500, 1),
torch.float32)
buf4 = empty_strided_cuda((4, 4, 4, 500), (8192, 2048, 500, 1),
torch.bool)
get_raw_stream(0)
triton_poi_fused_relu_threshold_backward_0[grid(32000)](buf0,
primals_2, buf1, buf4, 32000, XBLOCK=256, num_warps=4, num_stages=1
)
del primals_2
buf2 = buf0
del buf0
triton_poi_fused_relu_view_1[grid(32000)](buf1, buf2, 32000, XBLOCK
=256, num_warps=4, num_stages=1)
del buf1
buf3 = empty_strided_cuda((64, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_5, buf2, reinterpret_tensor(primals_4,
(500, 4), (1, 500), 0), alpha=1, beta=1, out=buf3)
del primals_5
return reinterpret_tensor(buf3, (4, 4, 4, 4), (64, 16, 4, 1), 0
), reinterpret_tensor(primals_3, (64, 4), (4, 1), 0
), buf2, primals_4, buf4
class MLPNew(nn.Module):
def __init__(self, input_size, output_size, hidden_size=500,
weight_decay=0.0):
super(MLPNew, self).__init__()
self.i2h = nn.Linear(in_features=input_size, out_features=hidden_size)
self.Dropout = nn.Dropout(p=0.5)
self.h2o = nn.Linear(in_features=hidden_size, out_features=output_size)
self.loss = nn.functional.cross_entropy
self.weight_decay = weight_decay
def predict(self, x):
return self.forward(x).argmax(dim=1)
def forward(self, input_0):
primals_1 = self.i2h.weight
primals_2 = self.i2h.bias
primals_4 = self.h2o.weight
primals_5 = self.h2o.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4, primals_5])
return output[0]
|
gchrupala/lyz
|
MLP
| false | 12,419 |
[
"MIT"
] | 0 |
e1e99899af65f6c4cb1fd77485f6fa61ba3500f5
|
https://github.com/gchrupala/lyz/tree/e1e99899af65f6c4cb1fd77485f6fa61ba3500f5
|
RWKV_ChannelMix
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/rn/crnf7npolybp3ziafhkvky7ug44jdqf6s6i7qo4iucr5nuinu5z6.py
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# x => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%constant_pad_nd, %slice_6], -1), kwargs = {})
triton_poi_fused_cat_0 = async_compile.triton('triton_poi_fused_cat_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_0', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = (xindex // 4) % 4
x3 = (xindex // 4)
x4 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 2, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = (-1) + x1
tmp6 = tmp5 >= tmp1
tmp7 = tl.full([1], 4, tl.int64)
tmp8 = tmp5 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tmp9 & tmp4
tmp11 = tl.load(in_ptr0 + ((-4) + (4*x3) + x0), tmp10 & xmask, eviction_policy='evict_last', other=0.0)
tmp12 = tl.full(tmp11.shape, 0.0, tmp11.dtype)
tmp13 = tl.where(tmp4, tmp11, tmp12)
tmp14 = tmp0 >= tmp3
tmp15 = tmp0 < tmp7
tmp16 = tl.load(in_ptr0 + (2 + (4*x3) + ((-2) + x0)), tmp14 & xmask, eviction_policy='evict_last', other=0.0)
tmp17 = tl.where(tmp4, tmp13, tmp16)
tl.store(out_ptr0 + (x4), tmp17, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/2h/c2hegytq7hmgogj6drh6cdha4zg2zuv4vozh5vbekubt4zlpteyb.py
# Topologically Sorted Source Nodes: [mish, mul], Original ATen: [aten.mish, aten.mul]
# Source node to ATen node mapping:
# mish => exp, gt, log1p, mul, tanh, where
# mul => mul_1
# Graph fragment:
# %exp : [num_users=1] = call_function[target=torch.ops.aten.exp.default](args = (%view_1,), kwargs = {})
# %log1p : [num_users=1] = call_function[target=torch.ops.aten.log1p.default](args = (%exp,), kwargs = {})
# %gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%view_1, 20), kwargs = {})
# %where : [num_users=1] = call_function[target=torch.ops.aten.where.self](args = (%gt, %view_1, %log1p), kwargs = {})
# %tanh : [num_users=1] = call_function[target=torch.ops.aten.tanh.default](args = (%where,), kwargs = {})
# %mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%view_1, %tanh), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%mul, %view_3), kwargs = {})
triton_poi_fused_mish_mul_1 = async_compile.triton('triton_poi_fused_mish_mul_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[256],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mish_mul_1', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mish_mul_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 160
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp8 = tl.load(in_ptr1 + (x0), xmask)
tmp1 = 20.0
tmp2 = tmp0 > tmp1
tmp3 = tl_math.exp(tmp0)
tmp4 = libdevice.log1p(tmp3)
tmp5 = tl.where(tmp2, tmp0, tmp4)
tmp6 = libdevice.tanh(tmp5)
tmp7 = tmp0 * tmp6
tmp9 = tmp7 * tmp8
tl.store(out_ptr0 + (x0), tmp9, xmask)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/oj/cojhz6pvfpzarbxfv5iqqf34ppyloqixpicpqrh62t4lizd7jny2.py
# Topologically Sorted Source Nodes: [sigmoid, rwkv], Original ATen: [aten.sigmoid, aten.mul]
# Source node to ATen node mapping:
# rwkv => mul_2
# sigmoid => sigmoid
# Graph fragment:
# %sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%view_5,), kwargs = {})
# %mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sigmoid, %view_7), kwargs = {})
triton_poi_fused_mul_sigmoid_2 = async_compile.triton('triton_poi_fused_mul_sigmoid_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[64],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_mul_sigmoid_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_mul_sigmoid_2(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + (x0), xmask)
tmp2 = tl.load(in_ptr1 + (x0), xmask)
tmp1 = tl.sigmoid(tmp0)
tmp3 = tmp1 * tmp2
tl.store(out_ptr0 + (x0), tmp3, xmask)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9 = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (10, 4), (4, 1))
assert_size_stride(primals_3, (10, ), (1, ))
assert_size_stride(primals_4, (10, 4), (4, 1))
assert_size_stride(primals_5, (10, ), (1, ))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4, ), (1, ))
assert_size_stride(primals_8, (4, 10), (10, 1))
assert_size_stride(primals_9, (4, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [x], Original ATen: [aten.cat]
stream0 = get_raw_stream(0)
triton_poi_fused_cat_0.run(primals_1, buf0, 64, grid=grid(64), stream=stream0)
del primals_1
buf1 = empty_strided_cuda((16, 10), (10, 1), torch.float32)
# Topologically Sorted Source Nodes: [k], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_3, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), reinterpret_tensor(primals_2, (4, 10), (1, 4), 0), alpha=1, beta=1, out=buf1)
del primals_2
del primals_3
buf2 = empty_strided_cuda((16, 10), (10, 1), torch.float32)
# Topologically Sorted Source Nodes: [v], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_5, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), reinterpret_tensor(primals_4, (4, 10), (1, 4), 0), alpha=1, beta=1, out=buf2)
del primals_4
del primals_5
buf3 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [r], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_7, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0), alpha=1, beta=1, out=buf3)
del primals_6
del primals_7
buf4 = empty_strided_cuda((4, 4, 10), (40, 10, 1), torch.float32)
# Topologically Sorted Source Nodes: [mish, mul], Original ATen: [aten.mish, aten.mul]
triton_poi_fused_mish_mul_1.run(buf1, buf2, buf4, 160, grid=grid(160), stream=stream0)
buf5 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
# Topologically Sorted Source Nodes: [wkv], Original ATen: [aten.addmm]
extern_kernels.addmm(primals_9, reinterpret_tensor(buf4, (16, 10), (10, 1), 0), reinterpret_tensor(primals_8, (10, 4), (1, 10), 0), alpha=1, beta=1, out=buf5)
del primals_9
buf6 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [sigmoid, rwkv], Original ATen: [aten.sigmoid, aten.mul]
triton_poi_fused_mul_sigmoid_2.run(buf3, buf5, buf6, 64, grid=grid(64), stream=stream0)
return (buf6, reinterpret_tensor(buf0, (16, 4), (4, 1), 0), buf1, buf2, buf3, reinterpret_tensor(buf4, (16, 10), (10, 1), 0), buf5, primals_8, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((4, 4, 4), (16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((10, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((10, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((10, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((10, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((4, 4), (4, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((4, 10), (10, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((4, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
from _paritybench_helpers import _mock_config
import torch
import torch.nn as nn
from torch.nn import functional as F
class RWKV_ChannelMix(nn.Module):
def __init__(self, config, layer_id):
super().__init__()
self.layer_id = layer_id
self.time_shift = nn.ZeroPad2d((0, 0, 1, -1))
hidden_sz = 5 * config.n_ffn // 2
self.key = nn.Linear(config.n_embd, hidden_sz)
self.value = nn.Linear(config.n_embd, hidden_sz)
self.weight = nn.Linear(hidden_sz, config.n_embd)
self.receptance = nn.Linear(config.n_embd, config.n_embd)
self.receptance.scale_init = 0
self.weight.scale_init = 0
def forward(self, x):
_B, _T, C = x.size()
x = torch.cat([self.time_shift(x[:, :, :C // 2]), x[:, :, C // 2:]],
dim=-1)
k = self.key(x)
v = self.value(x)
r = self.receptance(x)
wkv = self.weight(F.mish(k) * v)
rwkv = torch.sigmoid(r) * wkv
return rwkv
def get_inputs():
return [torch.rand([4, 4, 4])]
def get_init_inputs():
return [[], {'config': _mock_config(n_ffn=4, n_embd=4), 'layer_id': 1}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
import torch.nn as nn
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
@triton.jit
def triton_poi_fused_cat_0(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex % 4
x1 = xindex // 4 % 4
x3 = xindex // 4
x4 = xindex
tmp0 = x0
tmp1 = tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 2, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = -1 + x1
tmp6 = tmp5 >= tmp1
tmp7 = tl.full([1], 4, tl.int64)
tmp8 = tmp5 < tmp7
tmp9 = tmp6 & tmp8
tmp10 = tmp9 & tmp4
tmp11 = tl.load(in_ptr0 + (-4 + 4 * x3 + x0), tmp10 & xmask,
eviction_policy='evict_last', other=0.0)
tmp12 = tl.full(tmp11.shape, 0.0, tmp11.dtype)
tmp13 = tl.where(tmp4, tmp11, tmp12)
tmp14 = tmp0 >= tmp3
tmp16 = tl.load(in_ptr0 + (2 + 4 * x3 + (-2 + x0)), tmp14 & xmask,
eviction_policy='evict_last', other=0.0)
tmp17 = tl.where(tmp4, tmp13, tmp16)
tl.store(out_ptr0 + x4, tmp17, xmask)
@triton.jit
def triton_poi_fused_mish_mul_1(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK:
tl.constexpr):
xnumel = 160
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp8 = tl.load(in_ptr1 + x0, xmask)
tmp1 = 20.0
tmp2 = tmp0 > tmp1
tmp3 = tl_math.exp(tmp0)
tmp4 = libdevice.log1p(tmp3)
tmp5 = tl.where(tmp2, tmp0, tmp4)
tmp6 = libdevice.tanh(tmp5)
tmp7 = tmp0 * tmp6
tmp9 = tmp7 * tmp8
tl.store(out_ptr0 + x0, tmp9, xmask)
@triton.jit
def triton_poi_fused_mul_sigmoid_2(in_ptr0, in_ptr1, out_ptr0, xnumel,
XBLOCK: tl.constexpr):
xnumel = 64
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = xindex < xnumel
x0 = xindex
tmp0 = tl.load(in_ptr0 + x0, xmask)
tmp2 = tl.load(in_ptr1 + x0, xmask)
tmp1 = tl.sigmoid(tmp0)
tmp3 = tmp1 * tmp2
tl.store(out_ptr0 + x0, tmp3, xmask)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9) = args
args.clear()
assert_size_stride(primals_1, (4, 4, 4), (16, 4, 1))
assert_size_stride(primals_2, (10, 4), (4, 1))
assert_size_stride(primals_3, (10,), (1,))
assert_size_stride(primals_4, (10, 4), (4, 1))
assert_size_stride(primals_5, (10,), (1,))
assert_size_stride(primals_6, (4, 4), (4, 1))
assert_size_stride(primals_7, (4,), (1,))
assert_size_stride(primals_8, (4, 10), (10, 1))
assert_size_stride(primals_9, (4,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
get_raw_stream(0)
triton_poi_fused_cat_0[grid(64)](primals_1, buf0, 64, XBLOCK=64,
num_warps=1, num_stages=1)
del primals_1
buf1 = empty_strided_cuda((16, 10), (10, 1), torch.float32)
extern_kernels.addmm(primals_3, reinterpret_tensor(buf0, (16, 4), (
4, 1), 0), reinterpret_tensor(primals_2, (4, 10), (1, 4), 0),
alpha=1, beta=1, out=buf1)
del primals_2
del primals_3
buf2 = empty_strided_cuda((16, 10), (10, 1), torch.float32)
extern_kernels.addmm(primals_5, reinterpret_tensor(buf0, (16, 4), (
4, 1), 0), reinterpret_tensor(primals_4, (4, 10), (1, 4), 0),
alpha=1, beta=1, out=buf2)
del primals_4
del primals_5
buf3 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_7, reinterpret_tensor(buf0, (16, 4), (
4, 1), 0), reinterpret_tensor(primals_6, (4, 4), (1, 4), 0),
alpha=1, beta=1, out=buf3)
del primals_6
del primals_7
buf4 = empty_strided_cuda((4, 4, 10), (40, 10, 1), torch.float32)
triton_poi_fused_mish_mul_1[grid(160)](buf1, buf2, buf4, 160,
XBLOCK=128, num_warps=4, num_stages=1)
buf5 = empty_strided_cuda((16, 4), (4, 1), torch.float32)
extern_kernels.addmm(primals_9, reinterpret_tensor(buf4, (16, 10),
(10, 1), 0), reinterpret_tensor(primals_8, (10, 4), (1, 10), 0),
alpha=1, beta=1, out=buf5)
del primals_9
buf6 = empty_strided_cuda((4, 4, 4), (16, 4, 1), torch.float32)
triton_poi_fused_mul_sigmoid_2[grid(64)](buf3, buf5, buf6, 64,
XBLOCK=64, num_warps=1, num_stages=1)
return buf6, reinterpret_tensor(buf0, (16, 4), (4, 1), 0
), buf1, buf2, buf3, reinterpret_tensor(buf4, (16, 10), (10, 1), 0
), buf5, primals_8
class RWKV_ChannelMixNew(nn.Module):
def __init__(self, config, layer_id):
super().__init__()
self.layer_id = layer_id
self.time_shift = nn.ZeroPad2d((0, 0, 1, -1))
hidden_sz = 5 * config.n_ffn // 2
self.key = nn.Linear(config.n_embd, hidden_sz)
self.value = nn.Linear(config.n_embd, hidden_sz)
self.weight = nn.Linear(hidden_sz, config.n_embd)
self.receptance = nn.Linear(config.n_embd, config.n_embd)
self.receptance.scale_init = 0
self.weight.scale_init = 0
def forward(self, input_0):
primals_2 = self.key.weight
primals_3 = self.key.bias
primals_4 = self.value.weight
primals_5 = self.value.bias
primals_8 = self.weight.weight
primals_7 = self.weight.bias
primals_6 = self.receptance.weight
primals_9 = self.receptance.bias
primals_1 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9])
return output[0]
|
gdtool/AI-Writer
|
RWKV_ChannelMix
| false | 12,420 |
[
"BSD-3-Clause"
] | 0 |
25582175376a1feb09aab9079f7e32bba30d0519
|
https://github.com/gdtool/AI-Writer/tree/25582175376a1feb09aab9079f7e32bba30d0519
|
PSNRLoss
|
# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/bz/cbz4jfqnivmqo3kagpgcw2e6ojdrfmyv6p6tqz7c2t57rwabulea.py
# Topologically Sorted Source Nodes: [mul, mse_val, truediv, log10, mul_1], Original ATen: [aten.mul, aten.mse_loss, aten.div, aten.log10]
# Source node to ATen node mapping:
# log10 => log10
# mse_val => mean, pow_1, sub
# mul => full_default_1
# mul_1 => mul_1
# truediv => div
# Graph fragment:
# %full_default_1 : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([], 16.0), kwargs = {dtype: torch.float32, layout: torch.strided, device: cuda:0, pin_memory: False})
# %sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%arg0_1, %arg1_1), kwargs = {})
# %pow_1 : [num_users=1] = call_function[target=torch.ops.aten.pow.Tensor_Scalar](args = (%sub, 2), kwargs = {})
# %mean : [num_users=1] = call_function[target=torch.ops.aten.mean.default](args = (%pow_1,), kwargs = {})
# %div : [num_users=1] = call_function[target=torch.ops.aten.div.Tensor](args = (%full_default_1, %mean), kwargs = {})
# %log10 : [num_users=1] = call_function[target=torch.ops.aten.log10.default](args = (%div,), kwargs = {})
# %mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%log10, 10), kwargs = {})
triton_per_fused_div_log10_mse_loss_mul_0 = async_compile.triton('triton_per_fused_div_log10_mse_loss_mul_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.persistent_reduction(
size_hints=[1, 256],
reduction_hint=ReductionHint.INNER,
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: 'i32', 4: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {3: 1}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 4), equal_to_1=(3,))]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_div_log10_mse_loss_mul_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': True, 'num_load': 2, 'num_reduction': 1, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False}
)
@triton.jit
def triton_per_fused_div_log10_mse_loss_mul_0(in_out_ptr0, in_ptr0, in_ptr1, xnumel, rnumel):
xnumel = 1
XBLOCK: tl.constexpr = 1
rnumel = 256
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
xindex = tl.full([1], xoffset, tl.int32)
xmask = tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
roffset = 0
rmask = tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + (r0), None)
tmp1 = tl.load(in_ptr1 + (r0), None)
tmp2 = tmp0 - tmp1
tmp3 = tmp2 * tmp2
tmp4 = tl.broadcast_to(tmp3, [RBLOCK])
tmp6 = triton_helpers.promote_to_tensor(tl.sum(tmp4, 0))
tmp7 = 256.0
tmp8 = tmp6 / tmp7
tmp9 = 16.0
tmp10 = tmp9 / tmp8
tmp11 = libdevice.log10(tmp10)
tmp12 = 10.0
tmp13 = tmp11 * tmp12
tl.debug_barrier()
tl.store(in_out_ptr0 + (tl.full([1], 0, tl.int32)), tmp13, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0; del buf0 # reuse
# Topologically Sorted Source Nodes: [mul, mse_val, truediv, log10, mul_1], Original ATen: [aten.mul, aten.mse_loss, aten.div, aten.log10]
stream0 = get_raw_stream(0)
triton_per_fused_div_log10_mse_loss_mul_0.run(buf1, arg0_1, arg1_1, 1, 256, grid=grid(1), stream=stream0)
del arg0_1
del arg1_1
return (buf1, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
arg0_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
arg1_1 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
fn = lambda: call([arg0_1, arg1_1])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
from torch.nn.functional import mse_loss
def psnr_loss(input: 'torch.Tensor', target: 'torch.Tensor', max_val: 'float'
) ->torch.Tensor:
"""Function that computes PSNR
See :class:`~kornia.losses.PSNRLoss` for details.
"""
if not torch.is_tensor(input) or not torch.is_tensor(target):
raise TypeError(
f'Expected 2 torch tensors but got {type(input)} and {type(target)}'
)
if input.shape != target.shape:
raise TypeError(
f'Expected tensors of equal shapes, but got {input.shape} and {target.shape}'
)
mse_val = mse_loss(input, target, reduction='mean')
max_val_tensor: 'torch.Tensor' = torch.tensor(max_val).to(input.device).to(
input.dtype)
return 10 * torch.log10(max_val_tensor * max_val_tensor / mse_val)
class PSNRLoss(nn.Module):
"""Creates a criterion that calculates the PSNR between 2 images. Given an m x n image, the PSNR is:
.. math::
\\text{PSNR} = 10 \\log_{10} \\bigg(\\frac{\\text{MAX}_I^2}{MSE(I,T)}\\bigg)
where
.. math::
\\text{MSE}(I,T) = \\frac{1}{mn}\\sum_{i=0}^{m-1}\\sum_{j=0}^{n-1} [I(i,j) - T(i,j)]^2
and :math:`\\text{MAX}_I` is the maximum possible input value
(e.g for floating point images :math:`\\text{MAX}_I=1`).
Arguments:
max_val (float): Maximum value of input
Shape:
- input: :math:`(*)`
- approximation: :math:`(*)` same shape as input
- output: :math:`()` a scalar
Examples:
>>> psnr_loss(torch.ones(1), 1.2*torch.ones(1), 2) # 10 * log(4/((1.2-1)**2)) / log(10)
tensor(20.0000)
Reference:
https://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio#Definition
"""
def __init__(self, max_val: 'float') ->None:
super(PSNRLoss, self).__init__()
self.max_val: 'float' = max_val
def forward(self, input: 'torch.Tensor', target: 'torch.Tensor'
) ->torch.Tensor:
return psnr_loss(input, target, self.max_val)
def get_inputs():
return [torch.rand([4, 4, 4, 4]), torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'max_val': 4}]
|
import torch
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
from torch._inductor.runtime.triton_helpers import libdevice
import torch.nn as nn
from torch.nn.functional import mse_loss
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_per_fused_div_log10_mse_loss_mul_0(in_out_ptr0, in_ptr0, in_ptr1,
xnumel, rnumel):
XBLOCK: tl.constexpr = 1
RBLOCK: tl.constexpr = 256
xoffset = tl.program_id(0) * XBLOCK
tl.full([1], xoffset, tl.int32)
tl.full([RBLOCK], True, tl.int1)
rindex = tl.arange(0, RBLOCK)[:]
tl.full([RBLOCK], True, tl.int1)
r0 = rindex
tmp0 = tl.load(in_ptr0 + r0, None)
tmp1 = tl.load(in_ptr1 + r0, None)
tmp2 = tmp0 - tmp1
tmp3 = tmp2 * tmp2
tmp4 = tl.broadcast_to(tmp3, [RBLOCK])
tmp6 = triton_helpers.promote_to_tensor(tl.sum(tmp4, 0))
tmp7 = 256.0
tmp8 = tmp6 / tmp7
tmp9 = 16.0
tmp10 = tmp9 / tmp8
tmp11 = libdevice.log10(tmp10)
tmp12 = 10.0
tmp13 = tmp11 * tmp12
tl.debug_barrier()
tl.store(in_out_ptr0 + tl.full([1], 0, tl.int32), tmp13, None)
def call(args):
arg0_1, arg1_1 = args
args.clear()
assert_size_stride(arg0_1, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(arg1_1, (4, 4, 4, 4), (64, 16, 4, 1))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = empty_strided_cuda((), (), torch.float32)
buf1 = buf0
del buf0
get_raw_stream(0)
triton_per_fused_div_log10_mse_loss_mul_0[grid(1)](buf1, arg0_1,
arg1_1, 1, 256, num_warps=2, num_stages=1)
del arg0_1
del arg1_1
return buf1,
def psnr_loss(input: 'torch.Tensor', target: 'torch.Tensor', max_val: 'float'
) ->torch.Tensor:
"""Function that computes PSNR
See :class:`~kornia.losses.PSNRLoss` for details.
"""
if not torch.is_tensor(input) or not torch.is_tensor(target):
raise TypeError(
f'Expected 2 torch tensors but got {type(input)} and {type(target)}'
)
if input.shape != target.shape:
raise TypeError(
f'Expected tensors of equal shapes, but got {input.shape} and {target.shape}'
)
mse_val = mse_loss(input, target, reduction='mean')
max_val_tensor: 'torch.Tensor' = torch.tensor(max_val).to(input.device).to(
input.dtype)
return 10 * torch.log10(max_val_tensor * max_val_tensor / mse_val)
class PSNRLossNew(nn.Module):
"""Creates a criterion that calculates the PSNR between 2 images. Given an m x n image, the PSNR is:
.. math::
\\text{PSNR} = 10 \\log_{10} \\bigg(\\frac{\\text{MAX}_I^2}{MSE(I,T)}\\bigg)
where
.. math::
\\text{MSE}(I,T) = \\frac{1}{mn}\\sum_{i=0}^{m-1}\\sum_{j=0}^{n-1} [I(i,j) - T(i,j)]^2
and :math:`\\text{MAX}_I` is the maximum possible input value
(e.g for floating point images :math:`\\text{MAX}_I=1`).
Arguments:
max_val (float): Maximum value of input
Shape:
- input: :math:`(*)`
- approximation: :math:`(*)` same shape as input
- output: :math:`()` a scalar
Examples:
>>> psnr_loss(torch.ones(1), 1.2*torch.ones(1), 2) # 10 * log(4/((1.2-1)**2)) / log(10)
tensor(20.0000)
Reference:
https://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio#Definition
"""
def __init__(self, max_val: 'float') ->None:
super(PSNRLossNew, self).__init__()
self.max_val: 'float' = max_val
def forward(self, input_0, input_1):
arg0_1 = input_0
arg1_1 = input_1
output = call([arg0_1, arg1_1])
return output[0]
|
gf0507033/kornia
|
PSNRLoss
| false | 12,421 |
[
"ECL-2.0",
"Apache-2.0"
] | 0 |
2624f40a62d3639e6d946f3ca41fd1ce4b9de82d
|
https://github.com/gf0507033/kornia/tree/2624f40a62d3639e6d946f3ca41fd1ce4b9de82d
|
FEM
|
# AOT ID: ['0_forward']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid, grid_combo_kernels, start_graph, end_graph
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
# kernel path: runs/run_shard_9/inductor_cache/au/caug6nesiygukdpkrndsclkfho3dygoeotjtbnihl4wlyyiuddug.py
# Topologically Sorted Source Nodes: [conv2d_1, x1_2], Original ATen: [aten.convolution, aten.relu]
# Source node to ATen node mapping:
# conv2d_1 => convolution_1
# x1_2 => relu_1
# Graph fragment:
# %convolution_1 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_4, %primals_5, [1, 1], [2, 2], [2, 2], False, [0, 0], 1), kwargs = {})
# %relu_1 : [num_users=3] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_1,), kwargs = {})
triton_poi_fused_convolution_relu_0 = async_compile.triton('triton_poi_fused_convolution_relu_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_0', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_0(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16384
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = (xindex // 16) % 256
tmp0 = tl.load(in_out_ptr0 + (x3), None)
tmp1 = tl.load(in_ptr0 + (x1), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x3), tmp4, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/bi/cbi6mgxjq7myatdrsprtn45ft32zhfbh47z2ecvegeshlkxcnkmb.py
# Topologically Sorted Source Nodes: [conv2d_3, x2_2], Original ATen: [aten.convolution, aten.relu]
# Source node to ATen node mapping:
# conv2d_3 => convolution_3
# x2_2 => relu_3
# Graph fragment:
# %convolution_3 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%relu_1, %primals_8, %primals_9, [1, 1], [2, 2], [2, 2], False, [0, 0], 1), kwargs = {})
# %relu_3 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_3,), kwargs = {})
triton_poi_fused_convolution_relu_1 = async_compile.triton('triton_poi_fused_convolution_relu_1', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8192],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_1', 'mutated_arg_names': ['in_out_ptr0'], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_1(in_out_ptr0, in_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 8192
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = (xindex // 16) % 128
tmp0 = tl.load(in_out_ptr0 + (x3), None)
tmp1 = tl.load(in_ptr0 + (x1), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + (x3), tmp4, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/px/cpxz7rvk5ntgwsbrfsyn45wdiqnk4w4qfy7ybjomcymdd2uhhvzp.py
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
# Source node to ATen node mapping:
# cat => cat
# Graph fragment:
# %cat : [num_users=1] = call_function[target=torch.ops.aten.cat.default](args = ([%relu, %relu_2, %relu_4], 1), kwargs = {})
triton_poi_fused_cat_2 = async_compile.triton('triton_poi_fused_cat_2', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[32768],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*fp32', 3: '*fp32', 4: '*fp32', 5: '*fp32', 6: '*fp32', 7: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_cat_2', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 6, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_cat_2(in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 32768
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x1 = (xindex // 16) % 512
x0 = xindex % 16
x2 = (xindex // 8192)
x3 = xindex
tmp0 = x1
tmp1 = tl.full([1], 0, tl.int64)
tmp2 = tmp0 >= tmp1
tmp3 = tl.full([1], 256, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + (16*x1) + (4096*x2)), tmp4, other=0.0)
tmp6 = tl.load(in_ptr1 + (x1), tmp4, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full([1], 0, tl.int32)
tmp9 = triton_helpers.maximum(tmp8, tmp7)
tmp10 = tl.full(tmp9.shape, 0.0, tmp9.dtype)
tmp11 = tl.where(tmp4, tmp9, tmp10)
tmp12 = tmp0 >= tmp3
tmp13 = tl.full([1], 384, tl.int64)
tmp14 = tmp0 < tmp13
tmp15 = tmp12 & tmp14
tmp16 = tl.load(in_ptr2 + (x0 + (16*((-256) + x1)) + (2048*x2)), tmp15, other=0.0)
tmp17 = tl.load(in_ptr3 + ((-256) + x1), tmp15, eviction_policy='evict_last', other=0.0)
tmp18 = tmp16 + tmp17
tmp19 = triton_helpers.maximum(tmp8, tmp18)
tmp20 = tl.full(tmp19.shape, 0.0, tmp19.dtype)
tmp21 = tl.where(tmp15, tmp19, tmp20)
tmp22 = tmp0 >= tmp13
tmp23 = tl.full([1], 512, tl.int64)
tmp24 = tmp0 < tmp23
tmp25 = tl.load(in_ptr4 + (x0 + (16*((-384) + x1)) + (2048*x2)), tmp22, other=0.0)
tmp26 = tl.load(in_ptr5 + ((-384) + x1), tmp22, eviction_policy='evict_last', other=0.0)
tmp27 = tmp25 + tmp26
tmp28 = triton_helpers.maximum(tmp8, tmp27)
tmp29 = tl.full(tmp28.shape, 0.0, tmp28.dtype)
tmp30 = tl.where(tmp22, tmp28, tmp29)
tmp31 = tl.where(tmp15, tmp21, tmp30)
tmp32 = tl.where(tmp4, tmp11, tmp31)
tl.store(out_ptr0 + (x3), tmp32, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/ol/col4zcgg4cd6wrq4wve5wzb33mam2v7odywfgv546lznybwt4ers.py
# Topologically Sorted Source Nodes: [conv2d_4, x3_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# conv2d_4 => convolution_4
# x3_1 => relu_4
# Graph fragment:
# %convolution_4 : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%relu_3, %primals_10, %primals_11, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
# %relu_4 : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution_4,), kwargs = {})
# %le : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu_4, 0), kwargs = {})
triton_poi_fused_convolution_relu_threshold_backward_3 = async_compile.triton('triton_poi_fused_convolution_relu_threshold_backward_3', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[8192],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i1', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_threshold_backward_3', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_3(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 8192
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = (xindex // 16) % 128
tmp0 = tl.load(in_ptr0 + (x3), None)
tmp1 = tl.load(in_ptr1 + (x1), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x3), tmp6, None)
''', device_str='cuda')
# kernel path: runs/run_shard_9/inductor_cache/cj/ccj6nfiz77cbo4qn6bpoaocjigblzoawpjmrfrmgil5nsm6gp4z2.py
# Topologically Sorted Source Nodes: [conv2d, x1_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
# Source node to ATen node mapping:
# conv2d => convolution
# x1_1 => relu
# Graph fragment:
# %convolution : [num_users=1] = call_function[target=torch.ops.aten.convolution.default](args = (%primals_3, %primals_1, %primals_2, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
# %relu : [num_users=2] = call_function[target=torch.ops.aten.relu.default](args = (%convolution,), kwargs = {})
# %le_4 : [num_users=1] = call_function[target=torch.ops.aten.le.Scalar](args = (%relu, 0), kwargs = {})
triton_poi_fused_convolution_relu_threshold_backward_4 = async_compile.triton('triton_poi_fused_convolution_relu_threshold_backward_4', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor
from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, instance_descriptor, DeviceProperties
@triton_heuristics.pointwise(
size_hints=[16384],
filename=__file__,
triton_meta={'signature': {0: '*fp32', 1: '*fp32', 2: '*i1', 3: 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=80, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=108), 'constants': {}, 'configs': [AttrsDescriptor(divisible_by_16=(0, 1, 2, 3), equal_to_1=())]},
inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_convolution_relu_threshold_backward_4', 'mutated_arg_names': [], 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': 'A9C866B4A14FD3277824029365D703C2427B2E685E54EC9B3EF4ADC8D1EEAC1D', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_4(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
xnumel = 16384
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
xmask = tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = (xindex // 16) % 256
tmp0 = tl.load(in_ptr0 + (x3), None)
tmp1 = tl.load(in_ptr1 + (x1), None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + (x3), tmp6, None)
''', device_str='cuda')
async_compile.wait(globals())
del async_compile
def call(args):
primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11 = args
args.clear()
assert_size_stride(primals_1, (256, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_2, (256, ), (1, ))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (256, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_5, (256, ), (1, ))
assert_size_stride(primals_6, (128, 256, 3, 3), (2304, 9, 3, 1))
assert_size_stride(primals_7, (128, ), (1, ))
assert_size_stride(primals_8, (128, 256, 3, 3), (2304, 9, 3, 1))
assert_size_stride(primals_9, (128, ), (1, ))
assert_size_stride(primals_10, (128, 128, 3, 3), (1152, 9, 3, 1))
assert_size_stride(primals_11, (128, ), (1, ))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
# Topologically Sorted Source Nodes: [conv2d], Original ATen: [aten.convolution]
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 256, 4, 4), (4096, 16, 4, 1))
# Topologically Sorted Source Nodes: [conv2d_1], Original ATen: [aten.convolution]
buf1 = extern_kernels.convolution(primals_3, primals_4, stride=(1, 1), padding=(2, 2), dilation=(2, 2), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 256, 4, 4), (4096, 16, 4, 1))
buf2 = buf1; del buf1 # reuse
# Topologically Sorted Source Nodes: [conv2d_1, x1_2], Original ATen: [aten.convolution, aten.relu]
stream0 = get_raw_stream(0)
triton_poi_fused_convolution_relu_0.run(buf2, primals_5, 16384, grid=grid(16384), stream=stream0)
del primals_5
# Topologically Sorted Source Nodes: [conv2d_2], Original ATen: [aten.convolution]
buf3 = extern_kernels.convolution(buf2, primals_6, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 128, 4, 4), (2048, 16, 4, 1))
# Topologically Sorted Source Nodes: [conv2d_3], Original ATen: [aten.convolution]
buf4 = extern_kernels.convolution(buf2, primals_8, stride=(1, 1), padding=(2, 2), dilation=(2, 2), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 128, 4, 4), (2048, 16, 4, 1))
buf5 = buf4; del buf4 # reuse
# Topologically Sorted Source Nodes: [conv2d_3, x2_2], Original ATen: [aten.convolution, aten.relu]
triton_poi_fused_convolution_relu_1.run(buf5, primals_9, 8192, grid=grid(8192), stream=stream0)
del primals_9
# Topologically Sorted Source Nodes: [conv2d_4], Original ATen: [aten.convolution]
buf6 = extern_kernels.convolution(buf5, primals_10, stride=(1, 1), padding=(1, 1), dilation=(1, 1), transposed=False, output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 128, 4, 4), (2048, 16, 4, 1))
buf7 = empty_strided_cuda((4, 512, 4, 4), (8192, 16, 4, 1), torch.float32)
# Topologically Sorted Source Nodes: [cat], Original ATen: [aten.cat]
triton_poi_fused_cat_2.run(buf0, primals_2, buf3, primals_7, buf6, primals_11, buf7, 32768, grid=grid(32768), stream=stream0)
buf8 = empty_strided_cuda((4, 128, 4, 4), (2048, 16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [conv2d_4, x3_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
triton_poi_fused_convolution_relu_threshold_backward_3.run(buf6, primals_11, buf8, 8192, grid=grid(8192), stream=stream0)
del buf6
del primals_11
buf9 = empty_strided_cuda((4, 128, 4, 4), (2048, 16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [conv2d_2, x2_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
triton_poi_fused_convolution_relu_threshold_backward_3.run(buf3, primals_7, buf9, 8192, grid=grid(8192), stream=stream0)
del buf3
del primals_7
buf10 = empty_strided_cuda((4, 256, 4, 4), (4096, 16, 4, 1), torch.bool)
# Topologically Sorted Source Nodes: [conv2d, x1_1], Original ATen: [aten.convolution, aten.relu, aten.threshold_backward]
triton_poi_fused_convolution_relu_threshold_backward_4.run(buf0, primals_2, buf10, 16384, grid=grid(16384), stream=stream0)
del buf0
del primals_2
return (buf7, primals_1, primals_3, primals_4, primals_6, primals_8, primals_10, buf2, buf5, buf8, buf9, buf10, )
def benchmark_compiled_module(times=10, repeat=10):
from torch._dynamo.testing import rand_strided
from torch._inductor.utils import print_performance
primals_1 = rand_strided((256, 4, 3, 3), (36, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_2 = rand_strided((256, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_3 = rand_strided((4, 4, 4, 4), (64, 16, 4, 1), device='cuda:0', dtype=torch.float32)
primals_4 = rand_strided((256, 4, 3, 3), (36, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_5 = rand_strided((256, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_6 = rand_strided((128, 256, 3, 3), (2304, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_7 = rand_strided((128, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_8 = rand_strided((128, 256, 3, 3), (2304, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_9 = rand_strided((128, ), (1, ), device='cuda:0', dtype=torch.float32)
primals_10 = rand_strided((128, 128, 3, 3), (1152, 9, 3, 1), device='cuda:0', dtype=torch.float32)
primals_11 = rand_strided((128, ), (1, ), device='cuda:0', dtype=torch.float32)
fn = lambda: call([primals_1, primals_2, primals_3, primals_4, primals_5, primals_6, primals_7, primals_8, primals_9, primals_10, primals_11])
return print_performance(fn, times=times, repeat=repeat)
if __name__ == "__main__":
from torch._inductor.wrapper_benchmark import compiled_module_main
compiled_module_main('None', benchmark_compiled_module)
|
import torch
import torch.nn as nn
import torch.nn.functional as F
from math import sqrt as sqrt
from itertools import product as product
class FEM(nn.Module):
def __init__(self, channel_size):
super(FEM, self).__init__()
self.cs = channel_size
self.cpm1 = nn.Conv2d(self.cs, 256, kernel_size=3, dilation=1,
stride=1, padding=1)
self.cpm2 = nn.Conv2d(self.cs, 256, kernel_size=3, dilation=2,
stride=1, padding=2)
self.cpm3 = nn.Conv2d(256, 128, kernel_size=3, dilation=1, stride=1,
padding=1)
self.cpm4 = nn.Conv2d(256, 128, kernel_size=3, dilation=2, stride=1,
padding=2)
self.cpm5 = nn.Conv2d(128, 128, kernel_size=3, dilation=1, stride=1,
padding=1)
def forward(self, x):
x1_1 = F.relu(self.cpm1(x), inplace=True)
x1_2 = F.relu(self.cpm2(x), inplace=True)
x2_1 = F.relu(self.cpm3(x1_2), inplace=True)
x2_2 = F.relu(self.cpm4(x1_2), inplace=True)
x3_1 = F.relu(self.cpm5(x2_2), inplace=True)
return torch.cat([x1_1, x2_1, x3_1], 1)
def get_inputs():
return [torch.rand([4, 4, 4, 4])]
def get_init_inputs():
return [[], {'channel_size': 4}]
|
import torch
from torch._inductor.select_algorithm import extern_kernels
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import grid
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._inductor.runtime import triton_helpers
import torch.nn as nn
from math import sqrt as sqrt
from itertools import product as product
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
@triton.jit
def triton_poi_fused_convolution_relu_0(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = xindex // 16 % 256
tmp0 = tl.load(in_out_ptr0 + x3, None)
tmp1 = tl.load(in_ptr0 + x1, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x3, tmp4, None)
@triton.jit
def triton_poi_fused_convolution_relu_1(in_out_ptr0, in_ptr0, xnumel,
XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = xindex // 16 % 128
tmp0 = tl.load(in_out_ptr0 + x3, None)
tmp1 = tl.load(in_ptr0 + x1, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tl.store(in_out_ptr0 + x3, tmp4, None)
@triton.jit
def triton_poi_fused_cat_2(in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4,
in_ptr5, out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x1 = xindex // 16 % 512
x0 = xindex % 16
x2 = xindex // 8192
x3 = xindex
tmp0 = x1
tl.full([1], 0, tl.int64)
tmp3 = tl.full([1], 256, tl.int64)
tmp4 = tmp0 < tmp3
tmp5 = tl.load(in_ptr0 + (x0 + 16 * x1 + 4096 * x2), tmp4, other=0.0)
tmp6 = tl.load(in_ptr1 + x1, tmp4, eviction_policy='evict_last', other=0.0)
tmp7 = tmp5 + tmp6
tmp8 = tl.full([1], 0, tl.int32)
tmp9 = triton_helpers.maximum(tmp8, tmp7)
tmp10 = tl.full(tmp9.shape, 0.0, tmp9.dtype)
tmp11 = tl.where(tmp4, tmp9, tmp10)
tmp12 = tmp0 >= tmp3
tmp13 = tl.full([1], 384, tl.int64)
tmp14 = tmp0 < tmp13
tmp15 = tmp12 & tmp14
tmp16 = tl.load(in_ptr2 + (x0 + 16 * (-256 + x1) + 2048 * x2), tmp15,
other=0.0)
tmp17 = tl.load(in_ptr3 + (-256 + x1), tmp15, eviction_policy=
'evict_last', other=0.0)
tmp18 = tmp16 + tmp17
tmp19 = triton_helpers.maximum(tmp8, tmp18)
tmp20 = tl.full(tmp19.shape, 0.0, tmp19.dtype)
tmp21 = tl.where(tmp15, tmp19, tmp20)
tmp22 = tmp0 >= tmp13
tl.full([1], 512, tl.int64)
tmp25 = tl.load(in_ptr4 + (x0 + 16 * (-384 + x1) + 2048 * x2), tmp22,
other=0.0)
tmp26 = tl.load(in_ptr5 + (-384 + x1), tmp22, eviction_policy=
'evict_last', other=0.0)
tmp27 = tmp25 + tmp26
tmp28 = triton_helpers.maximum(tmp8, tmp27)
tmp29 = tl.full(tmp28.shape, 0.0, tmp28.dtype)
tmp30 = tl.where(tmp22, tmp28, tmp29)
tmp31 = tl.where(tmp15, tmp21, tmp30)
tmp32 = tl.where(tmp4, tmp11, tmp31)
tl.store(out_ptr0 + x3, tmp32, None)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_3(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = xindex // 16 % 128
tmp0 = tl.load(in_ptr0 + x3, None)
tmp1 = tl.load(in_ptr1 + x1, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + x3, tmp6, None)
@triton.jit
def triton_poi_fused_convolution_relu_threshold_backward_4(in_ptr0, in_ptr1,
out_ptr0, xnumel, XBLOCK: tl.constexpr):
xoffset = tl.program_id(0) * XBLOCK
xindex = xoffset + tl.arange(0, XBLOCK)[:]
tl.full([XBLOCK], True, tl.int1)
x3 = xindex
x1 = xindex // 16 % 256
tmp0 = tl.load(in_ptr0 + x3, None)
tmp1 = tl.load(in_ptr1 + x1, None, eviction_policy='evict_last')
tmp2 = tmp0 + tmp1
tmp3 = tl.full([1], 0, tl.int32)
tmp4 = triton_helpers.maximum(tmp3, tmp2)
tmp5 = 0.0
tmp6 = tmp4 <= tmp5
tl.store(out_ptr0 + x3, tmp6, None)
def call(args):
(primals_1, primals_2, primals_3, primals_4, primals_5, primals_6,
primals_7, primals_8, primals_9, primals_10, primals_11) = args
args.clear()
assert_size_stride(primals_1, (256, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_2, (256,), (1,))
assert_size_stride(primals_3, (4, 4, 4, 4), (64, 16, 4, 1))
assert_size_stride(primals_4, (256, 4, 3, 3), (36, 9, 3, 1))
assert_size_stride(primals_5, (256,), (1,))
assert_size_stride(primals_6, (128, 256, 3, 3), (2304, 9, 3, 1))
assert_size_stride(primals_7, (128,), (1,))
assert_size_stride(primals_8, (128, 256, 3, 3), (2304, 9, 3, 1))
assert_size_stride(primals_9, (128,), (1,))
assert_size_stride(primals_10, (128, 128, 3, 3), (1152, 9, 3, 1))
assert_size_stride(primals_11, (128,), (1,))
with torch.cuda._DeviceGuard(0):
torch.cuda.set_device(0)
buf0 = extern_kernels.convolution(primals_3, primals_1, stride=(1,
1), padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf0, (4, 256, 4, 4), (4096, 16, 4, 1))
buf1 = extern_kernels.convolution(primals_3, primals_4, stride=(1,
1), padding=(2, 2), dilation=(2, 2), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf1, (4, 256, 4, 4), (4096, 16, 4, 1))
buf2 = buf1
del buf1
get_raw_stream(0)
triton_poi_fused_convolution_relu_0[grid(16384)](buf2, primals_5,
16384, XBLOCK=256, num_warps=4, num_stages=1)
del primals_5
buf3 = extern_kernels.convolution(buf2, primals_6, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf3, (4, 128, 4, 4), (2048, 16, 4, 1))
buf4 = extern_kernels.convolution(buf2, primals_8, stride=(1, 1),
padding=(2, 2), dilation=(2, 2), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf4, (4, 128, 4, 4), (2048, 16, 4, 1))
buf5 = buf4
del buf4
triton_poi_fused_convolution_relu_1[grid(8192)](buf5, primals_9,
8192, XBLOCK=256, num_warps=4, num_stages=1)
del primals_9
buf6 = extern_kernels.convolution(buf5, primals_10, stride=(1, 1),
padding=(1, 1), dilation=(1, 1), transposed=False,
output_padding=(0, 0), groups=1, bias=None)
assert_size_stride(buf6, (4, 128, 4, 4), (2048, 16, 4, 1))
buf7 = empty_strided_cuda((4, 512, 4, 4), (8192, 16, 4, 1), torch.
float32)
triton_poi_fused_cat_2[grid(32768)](buf0, primals_2, buf3,
primals_7, buf6, primals_11, buf7, 32768, XBLOCK=256, num_warps
=4, num_stages=1)
buf8 = empty_strided_cuda((4, 128, 4, 4), (2048, 16, 4, 1), torch.bool)
triton_poi_fused_convolution_relu_threshold_backward_3[grid(8192)](buf6
, primals_11, buf8, 8192, XBLOCK=128, num_warps=4, num_stages=1)
del buf6
del primals_11
buf9 = empty_strided_cuda((4, 128, 4, 4), (2048, 16, 4, 1), torch.bool)
triton_poi_fused_convolution_relu_threshold_backward_3[grid(8192)](buf3
, primals_7, buf9, 8192, XBLOCK=128, num_warps=4, num_stages=1)
del buf3
del primals_7
buf10 = empty_strided_cuda((4, 256, 4, 4), (4096, 16, 4, 1), torch.bool
)
triton_poi_fused_convolution_relu_threshold_backward_4[grid(16384)](
buf0, primals_2, buf10, 16384, XBLOCK=128, num_warps=4,
num_stages=1)
del buf0
del primals_2
return (buf7, primals_1, primals_3, primals_4, primals_6, primals_8,
primals_10, buf2, buf5, buf8, buf9, buf10)
class FEMNew(nn.Module):
def __init__(self, channel_size):
super(FEMNew, self).__init__()
self.cs = channel_size
self.cpm1 = nn.Conv2d(self.cs, 256, kernel_size=3, dilation=1,
stride=1, padding=1)
self.cpm2 = nn.Conv2d(self.cs, 256, kernel_size=3, dilation=2,
stride=1, padding=2)
self.cpm3 = nn.Conv2d(256, 128, kernel_size=3, dilation=1, stride=1,
padding=1)
self.cpm4 = nn.Conv2d(256, 128, kernel_size=3, dilation=2, stride=1,
padding=2)
self.cpm5 = nn.Conv2d(128, 128, kernel_size=3, dilation=1, stride=1,
padding=1)
def forward(self, input_0):
primals_1 = self.cpm1.weight
primals_2 = self.cpm1.bias
primals_4 = self.cpm2.weight
primals_5 = self.cpm2.bias
primals_6 = self.cpm3.weight
primals_7 = self.cpm3.bias
primals_8 = self.cpm4.weight
primals_9 = self.cpm4.bias
primals_10 = self.cpm5.weight
primals_11 = self.cpm5.bias
primals_3 = input_0
output = call([primals_1, primals_2, primals_3, primals_4,
primals_5, primals_6, primals_7, primals_8, primals_9,
primals_10, primals_11])
return output[0]
|
fuankarion/FaceDetection-DSFD
|
FEM
| false | 12,422 |
[
"Apache-2.0"
] | 0 |
f1e464ec5c9d95c2fe73edf44e4d414a464839b1
|
https://github.com/fuankarion/FaceDetection-DSFD/tree/f1e464ec5c9d95c2fe73edf44e4d414a464839b1
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.