drbh
commited on
Commit
·
0e97a7c
1
Parent(s):
13afbbe
feat: bump build
Browse files- build/torch26-cxx11-cu118-x86_64-linux/megablocks/{_megablocks_63599de.abi3.so → _megablocks_13afbbe_dirty.abi3.so} +2 -2
- build/torch26-cxx11-cu118-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch26-cxx11-cu118-x86_64-linux/megablocks/layers.py +195 -20
- build/torch26-cxx11-cu118-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/torch26-cxx11-cu124-x86_64-linux/megablocks/{_megablocks_63599de.abi3.so → _megablocks_13afbbe_dirty.abi3.so} +2 -2
- build/torch26-cxx11-cu124-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch26-cxx11-cu124-x86_64-linux/megablocks/layers.py +195 -20
- build/torch26-cxx11-cu124-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/{torch27-cxx11-cu126-x86_64-linux/megablocks/_megablocks_63599de.abi3.so → torch26-cxx11-cu126-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so} +1 -1
- build/torch26-cxx11-cu126-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch26-cxx11-cu126-x86_64-linux/megablocks/layers.py +195 -20
- build/torch26-cxx11-cu126-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/{torch26-cxx11-cu126-x86_64-linux/megablocks/_megablocks_63599de.abi3.so → torch26-cxx98-cu118-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so} +2 -2
- build/torch26-cxx98-cu118-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch26-cxx98-cu118-x86_64-linux/megablocks/layers.py +195 -20
- build/torch26-cxx98-cu118-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/{torch26-cxx98-cu118-x86_64-linux/megablocks/_megablocks_63599de.abi3.so → torch26-cxx98-cu124-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so} +2 -2
- build/torch26-cxx98-cu124-x86_64-linux/megablocks/_megablocks_63599de.abi3.so +0 -3
- build/torch26-cxx98-cu124-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch26-cxx98-cu124-x86_64-linux/megablocks/layers.py +195 -20
- build/torch26-cxx98-cu124-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/torch26-cxx98-cu126-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so +3 -0
- build/torch26-cxx98-cu126-x86_64-linux/megablocks/_megablocks_63599de.abi3.so +0 -3
- build/torch26-cxx98-cu126-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch26-cxx98-cu126-x86_64-linux/megablocks/layers.py +195 -20
- build/torch26-cxx98-cu126-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/torch27-cxx11-cu118-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so +3 -0
- build/torch27-cxx11-cu118-x86_64-linux/megablocks/_megablocks_63599de.abi3.so +0 -3
- build/torch27-cxx11-cu118-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch27-cxx11-cu118-x86_64-linux/megablocks/layers.py +195 -20
- build/torch27-cxx11-cu118-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/torch27-cxx11-cu126-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so +3 -0
- build/torch27-cxx11-cu126-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch27-cxx11-cu126-x86_64-linux/megablocks/layers.py +195 -20
- build/torch27-cxx11-cu126-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
- build/torch27-cxx11-cu128-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so +3 -0
- build/torch27-cxx11-cu128-x86_64-linux/megablocks/_megablocks_63599de.abi3.so +0 -3
- build/torch27-cxx11-cu128-x86_64-linux/megablocks/_ops.py +3 -3
- build/torch27-cxx11-cu128-x86_64-linux/megablocks/layers.py +195 -20
- build/torch27-cxx11-cu128-x86_64-linux/megablocks/ops/all_to_all_benchmark.py +121 -21
build/torch26-cxx11-cu118-x86_64-linux/megablocks/{_megablocks_63599de.abi3.so → _megablocks_13afbbe_dirty.abi3.so}
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5683ac8b3e98fc8b8ab19f964b0dbfb9a980b6135220b0a0c1b50180665ce341
|
3 |
+
size 10517608
|
build/torch26-cxx11-cu118-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch26-cxx11-cu118-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch26-cxx11-cu118-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/torch26-cxx11-cu124-x86_64-linux/megablocks/{_megablocks_63599de.abi3.so → _megablocks_13afbbe_dirty.abi3.so}
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b55d6ee3d41404603fdb75ad9a2949aa92e0224f7056fdbeb4c66934035ebd4b
|
3 |
+
size 11869424
|
build/torch26-cxx11-cu124-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch26-cxx11-cu124-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch26-cxx11-cu124-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/{torch27-cxx11-cu126-x86_64-linux/megablocks/_megablocks_63599de.abi3.so → torch26-cxx11-cu126-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so}
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 11931080
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:516c5026180d4a8d013c500ed284a60ecbed4bc6c9dc084b838913f40327d1a6
|
3 |
size 11931080
|
build/torch26-cxx11-cu126-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch26-cxx11-cu126-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch26-cxx11-cu126-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/{torch26-cxx11-cu126-x86_64-linux/megablocks/_megablocks_63599de.abi3.so → torch26-cxx98-cu118-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so}
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a5c8c1b700d297741dd86e8c388e03913a30769ceb51b7c12a01245fbdf30128
|
3 |
+
size 10510072
|
build/torch26-cxx98-cu118-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch26-cxx98-cu118-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch26-cxx98-cu118-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/{torch26-cxx98-cu118-x86_64-linux/megablocks/_megablocks_63599de.abi3.so → torch26-cxx98-cu124-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so}
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d915db521f8d37fb887ed8334db60165e5923f8dce817d69f6441c5ba2d210d6
|
3 |
+
size 11857952
|
build/torch26-cxx98-cu124-x86_64-linux/megablocks/_megablocks_63599de.abi3.so
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:f8bfaaeb2a5e226a80403463d15f2c762ac8cb70ca7a44d2156aadfac63ab0d1
|
3 |
-
size 11857920
|
|
|
|
|
|
|
|
build/torch26-cxx98-cu124-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch26-cxx98-cu124-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch26-cxx98-cu124-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/torch26-cxx98-cu126-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:94a9a3bb426adceab66b39fe9d179b73e4524167aeb63bed5a67cd7734d31b24
|
3 |
+
size 11923704
|
build/torch26-cxx98-cu126-x86_64-linux/megablocks/_megablocks_63599de.abi3.so
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:637a8c7ef51b1d35911546ef7456854f1ee7cc3278565d2e144e16f733487148
|
3 |
-
size 11923672
|
|
|
|
|
|
|
|
build/torch26-cxx98-cu126-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch26-cxx98-cu126-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch26-cxx98-cu126-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/torch27-cxx11-cu118-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:aa9d1964e47ec6ff3c4ec77947f6a2a19868b03cec3618daf0555e011f69924d
|
3 |
+
size 10517848
|
build/torch27-cxx11-cu118-x86_64-linux/megablocks/_megablocks_63599de.abi3.so
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:002a58b415ed9e0f6418b103368c4f57f17fa86a851a02f594a33b097b33da09
|
3 |
-
size 10517816
|
|
|
|
|
|
|
|
build/torch27-cxx11-cu118-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch27-cxx11-cu118-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch27-cxx11-cu118-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/torch27-cxx11-cu126-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b204da58db0f8be45dda62abd98b74a8e60f1f983bfc6a128c74ff66f67cf502
|
3 |
+
size 11931112
|
build/torch27-cxx11-cu126-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch27-cxx11-cu126-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch27-cxx11-cu126-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|
build/torch27-cxx11-cu128-x86_64-linux/megablocks/_megablocks_13afbbe_dirty.abi3.so
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f861a8bffedbbf14341d39355f3f43a7c24fee2b99bb9ea7b3a2b9ad21c7ee28
|
3 |
+
size 17892656
|
build/torch27-cxx11-cu128-x86_64-linux/megablocks/_megablocks_63599de.abi3.so
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:dadccc59929c2fdbdf3b153f564d223013924c7b617d1eb2b3ecdc04470a4a60
|
3 |
-
size 17892624
|
|
|
|
|
|
|
|
build/torch27-cxx11-cu128-x86_64-linux/megablocks/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _megablocks_13afbbe_dirty
|
3 |
+
ops = torch.ops._megablocks_13afbbe_dirty
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_megablocks_13afbbe_dirty::{op_name}"
|
build/torch27-cxx11-cu128-x86_64-linux/megablocks/layers.py
CHANGED
@@ -121,7 +121,15 @@ def scale_grad(
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
-
def mlp_forward(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
# Scale weights
|
126 |
w1 = scale_grad(w1, gradient_scale)
|
127 |
w2 = scale_grad(w2, gradient_scale)
|
@@ -144,8 +152,6 @@ def mlp_forward(x, w1, w2, w1_bias, w2_bias, gradient_scale=None, alpha: float =
|
|
144 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
145 |
|
146 |
|
147 |
-
## START: Load Balancing Loss (unused at the moment)
|
148 |
-
|
149 |
# Global variable to store load balancing loss
|
150 |
_LOAD_BALANCING_LOSS = []
|
151 |
|
@@ -234,9 +240,6 @@ def batched_load_balancing_loss(args):
|
|
234 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
235 |
|
236 |
|
237 |
-
## END Load Balancing Loss
|
238 |
-
|
239 |
-
|
240 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
241 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
242 |
def expert_capacity(
|
@@ -410,7 +413,6 @@ def forward_once(
|
|
410 |
return x, tokens_per_expert
|
411 |
|
412 |
|
413 |
-
# TODO: replace with functional logic once aligned with ref
|
414 |
def parallel_forward_once(
|
415 |
x: torch.Tensor,
|
416 |
expert_weights: torch.Tensor,
|
@@ -429,15 +431,180 @@ def parallel_forward_once(
|
|
429 |
moe_expert_model_parallelism: bool = True,
|
430 |
hidden_size: int = 1152,
|
431 |
):
|
432 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
-
|
436 |
-
|
437 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
438 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
439 |
def forward(
|
440 |
-
# self,
|
441 |
x: torch.Tensor,
|
442 |
router_weight: torch.Tensor,
|
443 |
moe_top_k: int,
|
@@ -446,7 +613,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
446 |
moe_normalize_expert_weights: int = None,
|
447 |
uniform_expert_assignment: bool = False,
|
448 |
training: bool = False,
|
449 |
-
#
|
450 |
w1: torch.Tensor = None,
|
451 |
w2: torch.Tensor = None,
|
452 |
w1_bias: torch.Tensor = None,
|
@@ -522,7 +688,6 @@ class MyReplacementLayer(torch.nn.Module):
|
|
522 |
return x, expert_weights, router_scores
|
523 |
|
524 |
|
525 |
-
|
526 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
527 |
|
528 |
def forward(
|
@@ -536,11 +701,21 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
536 |
w2 = self.experts.down_proj.data
|
537 |
w1_bias = self.experts.gate_up_proj_bias.data
|
538 |
w2_bias = self.experts.down_proj_bias.data
|
539 |
-
expert_parallel_group = None
|
540 |
|
541 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
hidden_size = self.experts.hidden_size
|
543 |
-
|
544 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
545 |
x=x,
|
546 |
router_weight=router_weight,
|
@@ -559,8 +734,8 @@ class MegaBlocksMoeMLP(torch.nn.Module):
|
|
559 |
sort_end_bit=sort_end_bit,
|
560 |
expert_parallel_group=expert_parallel_group,
|
561 |
moe_capacity_factor=1.0,
|
562 |
-
moe_expert_model_parallelism=
|
563 |
-
forward_fn=
|
564 |
hidden_size=hidden_size,
|
565 |
)
|
566 |
-
return output, expert_weights_out
|
|
|
121 |
|
122 |
|
123 |
# Forward pass for the MLP layer
|
124 |
+
def mlp_forward(
|
125 |
+
x: torch.Tensor,
|
126 |
+
w1: torch.Tensor,
|
127 |
+
w2: torch.Tensor,
|
128 |
+
w1_bias: torch.Tensor,
|
129 |
+
w2_bias: torch.Tensor,
|
130 |
+
gradient_scale: Optional[float] = None,
|
131 |
+
alpha: float = 1.702,
|
132 |
+
):
|
133 |
# Scale weights
|
134 |
w1 = scale_grad(w1, gradient_scale)
|
135 |
w2 = scale_grad(w2, gradient_scale)
|
|
|
152 |
return torch.bmm(x, w2) + w2_bias[..., None, :]
|
153 |
|
154 |
|
|
|
|
|
155 |
# Global variable to store load balancing loss
|
156 |
_LOAD_BALANCING_LOSS = []
|
157 |
|
|
|
240 |
return scale * torch.dot(tokens_per_expert, expert_scores)
|
241 |
|
242 |
|
|
|
|
|
|
|
243 |
# Calculate the expert capacity based on tokens, top_k, number of experts,
|
244 |
# expert parallel group, capacity factor, and whether expert model parallelism is used.
|
245 |
def expert_capacity(
|
|
|
413 |
return x, tokens_per_expert
|
414 |
|
415 |
|
|
|
416 |
def parallel_forward_once(
|
417 |
x: torch.Tensor,
|
418 |
expert_weights: torch.Tensor,
|
|
|
431 |
moe_expert_model_parallelism: bool = True,
|
432 |
hidden_size: int = 1152,
|
433 |
):
|
434 |
+
# Flatten inputs
|
435 |
+
expert_weights = expert_weights.flatten()
|
436 |
+
top_experts = top_experts.flatten()
|
437 |
+
|
438 |
+
with torch.no_grad():
|
439 |
+
# Step 1: Local permutation setup
|
440 |
+
indices, bin_ids, bins, tokens_per_expert = indices_and_bins(
|
441 |
+
top_experts, sort_end_bit, num_experts
|
442 |
+
)
|
443 |
|
444 |
+
# Calculate sharding parameters
|
445 |
+
world_size = dist.get_world_size(expert_parallel_group)
|
446 |
+
hidden_sharding_deg = hidden_sharding_degree(
|
447 |
+
world_size, num_experts, hidden_size
|
448 |
+
)
|
449 |
+
experts_per_rank_val = experts_per_rank(num_experts, world_size)
|
450 |
|
451 |
+
# Replicate token counts for hidden sharding
|
452 |
+
repeated_tokens_per_expert = ops.repeat(
|
453 |
+
tokens_per_expert, (hidden_sharding_deg,)
|
454 |
+
)
|
455 |
+
|
456 |
+
# Exchange token counts across devices
|
457 |
+
parallel_tokens_per_expert = torch.empty_like(repeated_tokens_per_expert)
|
458 |
+
# print("world_size:", world_size)
|
459 |
+
# print("experts_per_rank_val:", experts_per_rank_val)
|
460 |
+
|
461 |
+
# Ensure CUB knows which device to use
|
462 |
+
tpe_handle = dist.all_to_all_single(
|
463 |
+
parallel_tokens_per_expert,
|
464 |
+
repeated_tokens_per_expert,
|
465 |
+
group=expert_parallel_group,
|
466 |
+
async_op=True,
|
467 |
+
)
|
468 |
+
|
469 |
+
# Step 2: Local permutation - group tokens by target device
|
470 |
+
x = x.view(-1, x.shape[-1]) # [sl * bs, hs]
|
471 |
+
x = ops.gather(x, indices, bin_ids, bins, top_k)
|
472 |
+
|
473 |
+
# Step 3: Compute communication counts and exchange tokens
|
474 |
+
with torch.no_grad():
|
475 |
+
tpe_handle.wait()
|
476 |
+
|
477 |
+
# Reshape for per-device calculations
|
478 |
+
repeated_tokens_per_expert = repeated_tokens_per_expert.view(
|
479 |
+
world_size, experts_per_rank_val
|
480 |
+
)
|
481 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.view(
|
482 |
+
world_size, experts_per_rank_val
|
483 |
+
)
|
484 |
+
|
485 |
+
# Calculate send/recv counts
|
486 |
+
send_counts = repeated_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
487 |
+
# recv_counts = parallel_tokens_per_expert.cpu().sum(dim=-1).tolist()
|
488 |
+
parallel_tokens_per_expert_cpu = parallel_tokens_per_expert.cpu()
|
489 |
+
recv_counts = parallel_tokens_per_expert_cpu.sum(dim=-1).tolist()
|
490 |
+
tokens_received = sum(recv_counts)
|
491 |
+
|
492 |
+
# Replicate for hidden sharding
|
493 |
+
x = ops.repeat(x, (hidden_sharding_deg, 1))
|
494 |
+
|
495 |
+
# Cross-device token exchange
|
496 |
+
parallel_x, parallel_x_handle = ops.all_to_all(
|
497 |
+
x,
|
498 |
+
recv_counts,
|
499 |
+
send_counts,
|
500 |
+
expert_parallel_group,
|
501 |
+
async_op=True
|
502 |
+
)
|
503 |
|
504 |
+
with torch.no_grad():
|
505 |
+
# Step 4: Setup for local expert computation
|
506 |
+
replicate_bins = ops.inclusive_cumsum(
|
507 |
+
parallel_tokens_per_expert.flatten(),
|
508 |
+
0
|
509 |
+
)
|
510 |
+
replicate_bins = (
|
511 |
+
replicate_bins.view(1) if not len(replicate_bins.size()) else replicate_bins
|
512 |
+
)
|
513 |
+
|
514 |
+
# Create expert indices for received tokens
|
515 |
+
parallel_top_expert = torch.remainder(
|
516 |
+
torch.arange(
|
517 |
+
num_experts * hidden_sharding_deg,
|
518 |
+
dtype=torch.int32,
|
519 |
+
device=indices.device,
|
520 |
+
),
|
521 |
+
experts_per_rank_val,
|
522 |
+
)
|
523 |
+
parallel_top_expert = ops.replicate(
|
524 |
+
parallel_top_expert.unsqueeze(dim=0),
|
525 |
+
replicate_bins,
|
526 |
+
tokens_received,
|
527 |
+
).flatten()
|
528 |
+
|
529 |
+
# Sort tokens by expert assignment
|
530 |
+
parallel_bin_ids, parallel_indices = ops.sort(
|
531 |
+
parallel_top_expert,
|
532 |
+
sort_end_bit,
|
533 |
+
)
|
534 |
+
|
535 |
+
# Calculate bins for local experts
|
536 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert.sum(
|
537 |
+
dim=0, dtype=torch.int
|
538 |
+
)
|
539 |
+
parallel_bins = ops.inclusive_cumsum(
|
540 |
+
parallel_tokens_per_expert,
|
541 |
+
0
|
542 |
+
)
|
543 |
+
parallel_bins = (
|
544 |
+
parallel_bins.view(1) if not len(parallel_bins.size()) else parallel_bins
|
545 |
+
)
|
546 |
+
|
547 |
+
# Calculate expert capacity
|
548 |
+
expert_capacity = expert_capacity_fn(
|
549 |
+
tokens_received,
|
550 |
+
top_k,
|
551 |
+
experts_per_rank_val,
|
552 |
+
expert_parallel_group,
|
553 |
+
moe_capacity_factor,
|
554 |
+
moe_expert_model_parallelism,
|
555 |
+
)
|
556 |
+
if expert_capacity == 0:
|
557 |
+
expert_capacity = torch.max(parallel_tokens_per_expert).item()
|
558 |
+
|
559 |
+
# Locally permute the tokens and perform the expert computation.
|
560 |
+
# Block to make sure that the cross-device permutation is complete.
|
561 |
+
# if self.args.mlp_impl == 'grouped':
|
562 |
+
|
563 |
+
# TODO: dont always assume grouped MLP
|
564 |
+
if True:
|
565 |
+
# GroupedMLP requires counts on CPU. We can use the tensor already
|
566 |
+
# moved to CPU for the prior all_to_all, which avoids an extra
|
567 |
+
# device synchronization.
|
568 |
+
parallel_tokens_per_expert = parallel_tokens_per_expert_cpu.sum(
|
569 |
+
dim=0,
|
570 |
+
dtype=torch.int,
|
571 |
+
)
|
572 |
+
|
573 |
+
# Step 5: Expert computation
|
574 |
+
parallel_x_handle.wait()
|
575 |
+
|
576 |
+
parallel_x = permute_and_compute(
|
577 |
+
parallel_x,
|
578 |
+
parallel_tokens_per_expert,
|
579 |
+
parallel_indices,
|
580 |
+
parallel_bin_ids,
|
581 |
+
None, # expert_weights
|
582 |
+
parallel_bins,
|
583 |
+
expert_capacity,
|
584 |
+
top_k=1,
|
585 |
+
w1=w1,
|
586 |
+
w2=w2,
|
587 |
+
w1_bias=w1_bias,
|
588 |
+
w2_bias=w2_bias,
|
589 |
+
gradient_scale=gradient_scale,
|
590 |
+
alpha=alpha,
|
591 |
+
)
|
592 |
+
|
593 |
+
# Step 6: Reverse communication - send results back
|
594 |
+
x, _ = ops.all_to_all(parallel_x, send_counts, recv_counts, expert_parallel_group)
|
595 |
+
|
596 |
+
# Step 7: Reduce across hidden sharding dimension
|
597 |
+
shape = (hidden_sharding_deg, -1, hidden_size)
|
598 |
+
x = x.view(shape).sum(dim=0)
|
599 |
+
|
600 |
+
# Step 8: Final local unpermutation
|
601 |
+
x = ops.scatter(x, indices, bin_ids, expert_weights, bins, top_k)
|
602 |
+
|
603 |
+
return x, tokens_per_expert.flatten()
|
604 |
+
|
605 |
+
|
606 |
+
class MyReplacementLayer(torch.nn.Module):
|
607 |
def forward(
|
|
|
608 |
x: torch.Tensor,
|
609 |
router_weight: torch.Tensor,
|
610 |
moe_top_k: int,
|
|
|
613 |
moe_normalize_expert_weights: int = None,
|
614 |
uniform_expert_assignment: bool = False,
|
615 |
training: bool = False,
|
|
|
616 |
w1: torch.Tensor = None,
|
617 |
w2: torch.Tensor = None,
|
618 |
w1_bias: torch.Tensor = None,
|
|
|
688 |
return x, expert_weights, router_scores
|
689 |
|
690 |
|
|
|
691 |
class MegaBlocksMoeMLP(torch.nn.Module):
|
692 |
|
693 |
def forward(
|
|
|
701 |
w2 = self.experts.down_proj.data
|
702 |
w1_bias = self.experts.gate_up_proj_bias.data
|
703 |
w2_bias = self.experts.down_proj_bias.data
|
|
|
704 |
|
705 |
+
# check if the expert_parallel_group attribute is set
|
706 |
+
if hasattr(self, "expert_parallel_group"):
|
707 |
+
expert_parallel_group = self.expert_parallel_group
|
708 |
+
moe_expert_model_parallelism = True
|
709 |
+
forward_fn = parallel_forward_once
|
710 |
+
else:
|
711 |
+
expert_parallel_group = None
|
712 |
+
moe_expert_model_parallelism = False
|
713 |
+
forward_fn = forward_once
|
714 |
+
|
715 |
+
sort_end_bit = max(
|
716 |
+
int(torch.ceil(torch.log2(torch.tensor(moe_num_experts)))), 1
|
717 |
+
)
|
718 |
hidden_size = self.experts.hidden_size
|
|
|
719 |
output, expert_weights_out, router_scores = MyReplacementLayer.forward(
|
720 |
x=x,
|
721 |
router_weight=router_weight,
|
|
|
734 |
sort_end_bit=sort_end_bit,
|
735 |
expert_parallel_group=expert_parallel_group,
|
736 |
moe_capacity_factor=1.0,
|
737 |
+
moe_expert_model_parallelism=moe_expert_model_parallelism,
|
738 |
+
forward_fn=forward_fn,
|
739 |
hidden_size=hidden_size,
|
740 |
)
|
741 |
+
return output, expert_weights_out
|
build/torch27-cxx11-cu128-x86_64-linux/megablocks/ops/all_to_all_benchmark.py
CHANGED
@@ -7,28 +7,126 @@ import torch.distributed as dist
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
-
from .. import benchmark_util
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
_ALL_TO_ALL_BENCHMARK = (
|
14 |
(8, 1024),
|
15 |
-
(16, 1024),
|
16 |
-
(32, 1024),
|
17 |
-
(64, 1024),
|
18 |
-
(128, 1024),
|
19 |
-
(256, 1024),
|
20 |
-
(512, 1024),
|
21 |
-
(1024, 1024),
|
22 |
-
(2 * 1024, 1024),
|
23 |
-
(4 * 1024, 1024),
|
24 |
-
(8 * 1024, 1024),
|
25 |
-
(16 * 1024, 1024),
|
26 |
-
(32 * 1024, 1024),
|
27 |
-
(64 * 1024, 1024),
|
28 |
-
(128 * 1024, 1024),
|
29 |
-
(256 * 1024, 1024),
|
30 |
-
(512 * 1024, 1024),
|
31 |
-
(1024 * 1024, 1024),
|
32 |
)
|
33 |
|
34 |
|
@@ -47,10 +145,12 @@ def benchmark_all_to_all(group, sl, hs):
|
|
47 |
def benchmark():
|
48 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
49 |
|
50 |
-
time, std = benchmark_util.benchmark_function(benchmark)
|
|
|
51 |
|
52 |
if dist.get_rank(group) == 0:
|
53 |
-
|
|
|
54 |
|
55 |
|
56 |
if __name__ == '__main__':
|
|
|
7 |
# from megablocks import benchmark_util
|
8 |
# from megablocks.layers.all_to_all import all_to_all
|
9 |
|
10 |
+
# from .. import benchmark_util
|
11 |
+
|
12 |
+
# Copyright 2024 Databricks
|
13 |
+
# SPDX-License-Identifier: Apache-2.0
|
14 |
+
|
15 |
+
import numpy as np
|
16 |
+
import torch
|
17 |
+
|
18 |
+
|
19 |
+
def log_benchmark(name, arguments, time, std):
|
20 |
+
print("=" * 60)
|
21 |
+
print(f"{name} Benchmark")
|
22 |
+
print("Benchmark Parameters:")
|
23 |
+
for key, value in arguments.items():
|
24 |
+
print(f"{key} = {value}")
|
25 |
+
print("Results:")
|
26 |
+
print("mean time = {:.3f}ms, std time = {:.3f}ms".format(time, std))
|
27 |
+
print("=" * 60)
|
28 |
+
|
29 |
+
|
30 |
+
def benchmark_function(fn, iterations=100, warmup=10):
|
31 |
+
print(f"Benchmarking {fn.__name__} with {iterations} iterations and {warmup} warmup iterations")
|
32 |
+
# Warmup iterations.
|
33 |
+
for _ in range(warmup):
|
34 |
+
fn()
|
35 |
+
|
36 |
+
times = []
|
37 |
+
print(f"Running {iterations} iterations...")
|
38 |
+
for i in range(iterations):
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
fn()
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
times.append(start.elapsed_time(end))
|
48 |
+
return np.mean(times), np.std(times)
|
49 |
+
|
50 |
+
|
51 |
+
# from .._layers.all_to_all import all_to_all
|
52 |
+
|
53 |
+
# Copyright 2024 Databricks
|
54 |
+
# SPDX-License-Identifier: Apache-2.0
|
55 |
+
|
56 |
+
import torch
|
57 |
+
import torch.distributed as dist
|
58 |
+
|
59 |
+
|
60 |
+
class AllToAllOp(torch.autograd.Function):
|
61 |
+
|
62 |
+
@staticmethod
|
63 |
+
def forward(ctx, x, output_split_sizes, input_split_sizes, group, async_op):
|
64 |
+
out = torch.empty(
|
65 |
+
(sum(output_split_sizes),) + x.shape[1:], device=x.device, dtype=x.dtype
|
66 |
+
)
|
67 |
+
|
68 |
+
ctx.input_shape = x.shape
|
69 |
+
ctx.output_split_sizes = output_split_sizes
|
70 |
+
ctx.input_split_sizes = input_split_sizes
|
71 |
+
ctx.group = group
|
72 |
+
handle = dist.all_to_all_single(
|
73 |
+
out,
|
74 |
+
x,
|
75 |
+
output_split_sizes=output_split_sizes,
|
76 |
+
input_split_sizes=input_split_sizes,
|
77 |
+
group=group,
|
78 |
+
async_op=async_op,
|
79 |
+
)
|
80 |
+
return out, handle
|
81 |
+
|
82 |
+
@staticmethod
|
83 |
+
def backward(ctx, grad, _):
|
84 |
+
if ctx.needs_input_grad[0]:
|
85 |
+
out = torch.empty(
|
86 |
+
ctx.input_shape,
|
87 |
+
device=grad.device,
|
88 |
+
dtype=grad.dtype,
|
89 |
+
)
|
90 |
+
dist.all_to_all_single(
|
91 |
+
out,
|
92 |
+
grad,
|
93 |
+
output_split_sizes=ctx.input_split_sizes,
|
94 |
+
input_split_sizes=ctx.output_split_sizes,
|
95 |
+
group=ctx.group,
|
96 |
+
)
|
97 |
+
return out, None, None, None, None
|
98 |
+
return None, None, None, None, None
|
99 |
+
|
100 |
+
|
101 |
+
def all_to_all(x, output_split_sizes, input_split_sizes, group, async_op=False):
|
102 |
+
return AllToAllOp.apply(
|
103 |
+
x,
|
104 |
+
output_split_sizes,
|
105 |
+
input_split_sizes,
|
106 |
+
group,
|
107 |
+
async_op,
|
108 |
+
)
|
109 |
+
|
110 |
|
111 |
_ALL_TO_ALL_BENCHMARK = (
|
112 |
(8, 1024),
|
113 |
+
# (16, 1024),
|
114 |
+
# (32, 1024),
|
115 |
+
# (64, 1024),
|
116 |
+
# (128, 1024),
|
117 |
+
# (256, 1024),
|
118 |
+
# (512, 1024),
|
119 |
+
# (1024, 1024),
|
120 |
+
# (2 * 1024, 1024),
|
121 |
+
# (4 * 1024, 1024),
|
122 |
+
# (8 * 1024, 1024),
|
123 |
+
# (16 * 1024, 1024),
|
124 |
+
# (32 * 1024, 1024),
|
125 |
+
# (64 * 1024, 1024),
|
126 |
+
# (128 * 1024, 1024),
|
127 |
+
# (256 * 1024, 1024),
|
128 |
+
# (512 * 1024, 1024),
|
129 |
+
# (1024 * 1024, 1024),
|
130 |
)
|
131 |
|
132 |
|
|
|
145 |
def benchmark():
|
146 |
return all_to_all(x, send_recv_sizes, send_recv_sizes, group)
|
147 |
|
148 |
+
# time, std = benchmark_util.benchmark_function(benchmark)
|
149 |
+
time, std = benchmark_function(benchmark)
|
150 |
|
151 |
if dist.get_rank(group) == 0:
|
152 |
+
log_benchmark('All-To-All', details, time, std)
|
153 |
+
# benchmark_util.log_benchmark('All-To-All', details, time, std)
|
154 |
|
155 |
|
156 |
if __name__ == '__main__':
|