code
stringlengths 82
54.1k
| code_codestyle
int64 0
699
| style_context
stringlengths 111
35.6k
| style_context_codestyle
int64 0
699
| label
int64 0
1
|
|---|---|---|---|---|
from collections.abc import Sequence
from queue import Queue
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : int=None , SCREAMING_SNAKE_CASE__ : Tuple=None ):
'''simple docstring'''
__a : Tuple = start
__a : Dict = end
__a : List[str] = val
__a : List[Any] = (start + end) // 2
__a : Optional[Any] = left
__a : List[str] = right
def __repr__( self : Dict ):
'''simple docstring'''
return f'''SegmentTreeNode(start={self.start}, end={self.end}, val={self.val})'''
class _UpperCamelCase:
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : Sequence , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Tuple = collection
__a : Dict = function
if self.collection:
__a : int = self._build_tree(0 , len(SCREAMING_SNAKE_CASE__ ) - 1 )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self._update_tree(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
return self._query_range(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if start == end:
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.collection[start] )
__a : Tuple = (start + end) // 2
__a : Optional[int] = self._build_tree(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = self._build_tree(mid + 1 , SCREAMING_SNAKE_CASE__ )
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.fn(left.val , right.val ) , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if node.start == i and node.end == i:
__a : Optional[Any] = val
return
if i <= node.mid:
self._update_tree(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
self._update_tree(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = self.fn(node.left.val , node.right.val )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
if node.start == i and node.end == j:
return node.val
if i <= node.mid:
if j <= node.mid:
# range in left child tree
return self._query_range(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
# range in left child tree and right child tree
return self.fn(
self._query_range(node.left , SCREAMING_SNAKE_CASE__ , node.mid ) , self._query_range(node.right , node.mid + 1 , SCREAMING_SNAKE_CASE__ ) , )
else:
# range in right child tree
return self._query_range(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
if self.root is not None:
__a : Tuple = Queue()
queue.put(self.root )
while not queue.empty():
__a : Tuple = queue.get()
yield node
if node.left is not None:
queue.put(node.left )
if node.right is not None:
queue.put(node.right )
if __name__ == "__main__":
import operator
for fn in [operator.add, max, min]:
print('''*''' * 50)
SCREAMING_SNAKE_CASE__ = SegmentTree([2, 1, 5, 3, 4], fn)
for node in arr.traverse():
print(node)
print()
arr.update(1, 5)
for node in arr.traverse():
print(node)
print()
print(arr.query_range(3, 4)) # 7
print(arr.query_range(2, 2)) # 5
print(arr.query_range(1, 3)) # 13
print()
| 47 |
from ...utils import is_note_seq_available, is_transformers_available, is_torch_available
from ...utils import OptionalDependencyNotAvailable
try:
if not (is_transformers_available() and is_torch_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_torch_and_transformers_objects import * # noqa F403
else:
from .notes_encoder import SpectrogramNotesEncoder
from .continous_encoder import SpectrogramContEncoder
from .pipeline_spectrogram_diffusion import (
SpectrogramContEncoder,
SpectrogramDiffusionPipeline,
TaFilmDecoder,
)
try:
if not (is_transformers_available() and is_torch_available() and is_note_seq_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_transformers_and_torch_and_note_seq_objects import * # noqa F403
else:
from .midi_utils import MidiProcessor
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple ):
# Return True if there is node that has not iterated.
__a : int = [False] * len(lowerCamelCase_ )
__a : Optional[int] = []
queue.append(lowerCamelCase_ )
__a : List[str] = True
while queue:
__a : Optional[Any] = queue.pop(0 )
for ind in range(len(graph[u] ) ):
if visited[ind] is False and graph[u][ind] > 0:
queue.append(lowerCamelCase_ )
__a : Dict = True
__a : int = u
return visited[t]
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple , lowerCamelCase_ : Any ):
# This array is filled by BFS and to store path
__a : List[str] = [-1] * (len(lowerCamelCase_ ))
__a : int = 0
while bfs(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ ):
__a : List[Any] = float('Inf' )
__a : List[Any] = sink
while s != source:
# Find the minimum value in select path
__a : List[str] = min(lowerCamelCase_ , graph[parent[s]][s] )
__a : List[str] = parent[s]
max_flow += path_flow
__a : int = sink
while v != source:
__a : Optional[Any] = parent[v]
graph[u][v] -= path_flow
graph[v][u] += path_flow
__a : Optional[Any] = parent[v]
return max_flow
SCREAMING_SNAKE_CASE__ = [
[0, 16, 13, 0, 0, 0],
[0, 0, 10, 12, 0, 0],
[0, 4, 0, 0, 14, 0],
[0, 0, 9, 0, 0, 20],
[0, 0, 0, 7, 0, 4],
[0, 0, 0, 0, 0, 0],
]
SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ = 0, 5
print(ford_fulkerson(graph, source, sink))
| 47 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_bridgetower''': [
'''BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''BridgeTowerConfig''',
'''BridgeTowerTextConfig''',
'''BridgeTowerVisionConfig''',
],
'''processing_bridgetower''': ['''BridgeTowerProcessor'''],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''BridgeTowerImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''BridgeTowerForContrastiveLearning''',
'''BridgeTowerForImageAndTextRetrieval''',
'''BridgeTowerForMaskedLM''',
'''BridgeTowerModel''',
'''BridgeTowerPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_bridgetower import (
BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP,
BridgeTowerConfig,
BridgeTowerTextConfig,
BridgeTowerVisionConfig,
)
from .processing_bridgetower import BridgeTowerProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_bridgetower import BridgeTowerImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_bridgetower import (
BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST,
BridgeTowerForContrastiveLearning,
BridgeTowerForImageAndTextRetrieval,
BridgeTowerForMaskedLM,
BridgeTowerModel,
BridgeTowerPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure)
| 47 | 1 |
import numpy
# List of input, output pairs
SCREAMING_SNAKE_CASE__ = (
((5, 2, 3), 15),
((6, 5, 9), 25),
((11, 12, 13), 41),
((1, 1, 1), 8),
((11, 12, 13), 41),
)
SCREAMING_SNAKE_CASE__ = (((515, 22, 13), 555), ((61, 35, 49), 150))
SCREAMING_SNAKE_CASE__ = [2, 4, 1, 5]
SCREAMING_SNAKE_CASE__ = len(train_data)
SCREAMING_SNAKE_CASE__ = 0.0_0_9
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Dict="train" ):
return calculate_hypothesis_value(lowerCamelCase_ , lowerCamelCase_ ) - output(
lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ):
__a : List[Any] = 0
for i in range(len(lowerCamelCase_ ) - 1 ):
hyp_val += data_input_tuple[i] * parameter_vector[i + 1]
hyp_val += parameter_vector[0]
return hyp_val
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] ):
if data_set == "train":
return train_data[example_no][1]
elif data_set == "test":
return test_data[example_no][1]
return None
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : Optional[Any] ):
if data_set == "train":
return _hypothesis_value(train_data[example_no][0] )
elif data_set == "test":
return _hypothesis_value(test_data[example_no][0] )
return None
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Optional[int]=m ):
__a : Optional[Any] = 0
for i in range(lowerCamelCase_ ):
if index == -1:
summation_value += _error(lowerCamelCase_ )
else:
summation_value += _error(lowerCamelCase_ ) * train_data[i][0][index]
return summation_value
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
__a : Union[str, Any] = summation_of_cost_derivative(lowerCamelCase_ , lowerCamelCase_ ) / m
return cost_derivative_value
def UpperCAmelCase__ ( ):
global parameter_vector
# Tune these values to set a tolerance value for predicted output
__a : List[str] = 0.000002
__a : str = 0
__a : List[str] = 0
while True:
j += 1
__a : Dict = [0, 0, 0, 0]
for i in range(0 , len(lowerCamelCase_ ) ):
__a : List[Any] = get_cost_derivative(i - 1 )
__a : Dict = (
parameter_vector[i] - LEARNING_RATE * cost_derivative
)
if numpy.allclose(
lowerCamelCase_ , lowerCamelCase_ , atol=lowerCamelCase_ , rtol=lowerCamelCase_ , ):
break
__a : Dict = temp_parameter_vector
print(('Number of iterations:', j) )
def UpperCAmelCase__ ( ):
for i in range(len(lowerCamelCase_ ) ):
print(('Actual output value:', output(lowerCamelCase_ , 'test' )) )
print(('Hypothesis output:', calculate_hypothesis_value(lowerCamelCase_ , 'test' )) )
if __name__ == "__main__":
run_gradient_descent()
print('''\nTesting gradient descent for a linear hypothesis function.\n''')
test_gradient_descent()
| 47 |
from string import ascii_lowercase, ascii_uppercase
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
if not sentence:
return ""
__a : Union[str, Any] = dict(zip(lowerCamelCase_ , lowerCamelCase_ ) )
return lower_to_upper.get(sentence[0] , sentence[0] ) + sentence[1:]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 47 | 1 |
import unittest
import numpy as np
import torch
from torch import nn
from transformers import (
CLIPImageProcessor,
CLIPTextConfig,
CLIPTextModelWithProjection,
CLIPTokenizer,
CLIPVisionConfig,
CLIPVisionModelWithProjection,
)
from diffusers import KandinskyVaaPriorPipeline, PriorTransformer, UnCLIPScheduler
from diffusers.utils import torch_device
from diffusers.utils.testing_utils import enable_full_determinism, skip_mps
from ..test_pipelines_common import PipelineTesterMixin
enable_full_determinism()
class _UpperCamelCase( __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : List[str] = KandinskyVaaPriorPipeline
__SCREAMING_SNAKE_CASE : int = ['''prompt''']
__SCREAMING_SNAKE_CASE : Tuple = ['''prompt''', '''negative_prompt''']
__SCREAMING_SNAKE_CASE : Any = [
'''num_images_per_prompt''',
'''generator''',
'''num_inference_steps''',
'''latents''',
'''negative_prompt''',
'''guidance_scale''',
'''output_type''',
'''return_dict''',
]
__SCREAMING_SNAKE_CASE : Tuple = False
@property
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
return 3_2
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return 3_2
@property
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return self.time_input_dim
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.time_input_dim * 4
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return 1_0_0
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : Optional[int] = CLIPTokenizer.from_pretrained('hf-internal-testing/tiny-random-clip' )
return tokenizer
@property
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
torch.manual_seed(0 )
__a : Tuple = CLIPTextConfig(
bos_token_id=0 , eos_token_id=2 , hidden_size=self.text_embedder_hidden_size , projection_dim=self.text_embedder_hidden_size , intermediate_size=3_7 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1_0_0_0 , )
return CLIPTextModelWithProjection(SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
torch.manual_seed(0 )
__a : Union[str, Any] = {
'num_attention_heads': 2,
'attention_head_dim': 1_2,
'embedding_dim': self.text_embedder_hidden_size,
'num_layers': 1,
}
__a : Optional[Any] = PriorTransformer(**SCREAMING_SNAKE_CASE__ )
# clip_std and clip_mean is initialized to be 0 so PriorTransformer.post_process_latents will always return 0 - set clip_std to be 1 so it won't return 0
__a : Tuple = nn.Parameter(torch.ones(model.clip_std.shape ) )
return model
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
torch.manual_seed(0 )
__a : Dict = CLIPVisionConfig(
hidden_size=self.text_embedder_hidden_size , image_size=2_2_4 , projection_dim=self.text_embedder_hidden_size , intermediate_size=3_7 , num_attention_heads=4 , num_channels=3 , num_hidden_layers=5 , patch_size=1_4 , )
__a : int = CLIPVisionModelWithProjection(SCREAMING_SNAKE_CASE__ )
return model
@property
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : Optional[Any] = CLIPImageProcessor(
crop_size=2_2_4 , do_center_crop=SCREAMING_SNAKE_CASE__ , do_normalize=SCREAMING_SNAKE_CASE__ , do_resize=SCREAMING_SNAKE_CASE__ , image_mean=[0.48_145_466, 0.4_578_275, 0.40_821_073] , image_std=[0.26_862_954, 0.26_130_258, 0.27_577_711] , resample=3 , size=2_2_4 , )
return image_processor
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Union[str, Any] = self.dummy_prior
__a : int = self.dummy_image_encoder
__a : Optional[Any] = self.dummy_text_encoder
__a : Tuple = self.dummy_tokenizer
__a : Dict = self.dummy_image_processor
__a : Tuple = UnCLIPScheduler(
variance_type='fixed_small_log' , prediction_type='sample' , num_train_timesteps=1_0_0_0 , clip_sample=SCREAMING_SNAKE_CASE__ , clip_sample_range=10.0 , )
__a : Any = {
'prior': prior,
'image_encoder': image_encoder,
'text_encoder': text_encoder,
'tokenizer': tokenizer,
'scheduler': scheduler,
'image_processor': image_processor,
}
return components
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Any=0 ):
'''simple docstring'''
if str(SCREAMING_SNAKE_CASE__ ).startswith('mps' ):
__a : List[str] = torch.manual_seed(SCREAMING_SNAKE_CASE__ )
else:
__a : str = torch.Generator(device=SCREAMING_SNAKE_CASE__ ).manual_seed(SCREAMING_SNAKE_CASE__ )
__a : Tuple = {
'prompt': 'horse',
'generator': generator,
'guidance_scale': 4.0,
'num_inference_steps': 2,
'output_type': 'np',
}
return inputs
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : Union[str, Any] = 'cpu'
__a : str = self.get_dummy_components()
__a : Optional[Any] = self.pipeline_class(**SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = pipe.to(SCREAMING_SNAKE_CASE__ )
pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = pipe(**self.get_dummy_inputs(SCREAMING_SNAKE_CASE__ ) )
__a : List[Any] = output.image_embeds
__a : Optional[int] = pipe(
**self.get_dummy_inputs(SCREAMING_SNAKE_CASE__ ) , return_dict=SCREAMING_SNAKE_CASE__ , )[0]
__a : List[Any] = image[0, -1_0:]
__a : Union[str, Any] = image_from_tuple[0, -1_0:]
assert image.shape == (1, 3_2)
__a : Optional[int] = np.array(
[-0.0_532, 1.7_120, 0.3_656, -1.0_852, -0.8_946, -1.1_756, 0.4_348, 0.2_482, 0.5_146, -0.1_156] )
assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2
assert np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2
@skip_mps
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : str = torch_device == 'cpu'
__a : Tuple = True
__a : Optional[Any] = False
self._test_inference_batch_single_identical(
test_max_difference=SCREAMING_SNAKE_CASE__ , relax_max_difference=SCREAMING_SNAKE_CASE__ , test_mean_pixel_difference=SCREAMING_SNAKE_CASE__ , )
@skip_mps
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Optional[int] = torch_device == 'cpu'
__a : Any = False
self._test_attention_slicing_forward_pass(
test_max_difference=SCREAMING_SNAKE_CASE__ , test_mean_pixel_difference=SCREAMING_SNAKE_CASE__ , )
| 47 |
import functools
import operator
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''asapp/sew-d-tiny-100k''': '''https://huggingface.co/asapp/sew-d-tiny-100k/resolve/main/config.json''',
# See all SEW-D models at https://huggingface.co/models?filter=sew-d
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = '''sew-d'''
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : Dict=3_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=7_6_8 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : str=3_0_7_2 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : List[str]=2_5_6 , SCREAMING_SNAKE_CASE__ : Optional[Any]=True , SCREAMING_SNAKE_CASE__ : Tuple=True , SCREAMING_SNAKE_CASE__ : List[str]=("p2c", "c2p") , SCREAMING_SNAKE_CASE__ : str="layer_norm" , SCREAMING_SNAKE_CASE__ : Tuple="gelu_python" , SCREAMING_SNAKE_CASE__ : Tuple=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.0 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.02 , SCREAMING_SNAKE_CASE__ : int=1e-7 , SCREAMING_SNAKE_CASE__ : Any=1e-5 , SCREAMING_SNAKE_CASE__ : Optional[int]="group" , SCREAMING_SNAKE_CASE__ : Optional[Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=(6_4, 1_2_8, 1_2_8, 1_2_8, 1_2_8, 2_5_6, 2_5_6, 2_5_6, 2_5_6, 5_1_2, 5_1_2, 5_1_2, 5_1_2) , SCREAMING_SNAKE_CASE__ : List[Any]=(5, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : str=(1_0, 3, 1, 3, 1, 3, 1, 3, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : Optional[int]=False , SCREAMING_SNAKE_CASE__ : Optional[int]=1_2_8 , SCREAMING_SNAKE_CASE__ : Tuple=1_6 , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[Any]=0.05 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , SCREAMING_SNAKE_CASE__ : int=0.0 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : Optional[int]=0 , SCREAMING_SNAKE_CASE__ : Optional[int]="mean" , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : List[str]=False , SCREAMING_SNAKE_CASE__ : str=2_5_6 , SCREAMING_SNAKE_CASE__ : str=0 , SCREAMING_SNAKE_CASE__ : List[Any]=1 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ , pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = hidden_size
__a : Optional[Any] = feat_extract_norm
__a : List[str] = feat_extract_activation
__a : Dict = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = list(SCREAMING_SNAKE_CASE__ )
__a : List[str] = list(SCREAMING_SNAKE_CASE__ )
__a : int = conv_bias
__a : Tuple = num_conv_pos_embeddings
__a : List[str] = num_conv_pos_embedding_groups
__a : Optional[Any] = len(self.conv_dim )
__a : Union[str, Any] = num_hidden_layers
__a : Optional[Any] = intermediate_size
__a : Union[str, Any] = squeeze_factor
__a : List[Any] = max_position_embeddings
__a : Tuple = position_buckets
__a : Optional[int] = share_att_key
__a : List[str] = relative_attention
__a : Any = norm_rel_ebd
__a : Any = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = hidden_act
__a : str = num_attention_heads
__a : Union[str, Any] = hidden_dropout
__a : Optional[int] = attention_dropout
__a : List[str] = activation_dropout
__a : int = feat_proj_dropout
__a : int = final_dropout
__a : Dict = layer_norm_eps
__a : Tuple = feature_layer_norm_eps
__a : str = initializer_range
__a : Tuple = vocab_size
if (
(len(self.conv_stride ) != self.num_feat_extract_layers)
or (len(self.conv_kernel ) != self.num_feat_extract_layers)
or (len(self.conv_dim ) != self.num_feat_extract_layers)
):
raise ValueError(
'Configuration for convolutional layers is incorrect.'
'It is required that `len(config.conv_dim)` == `len(config.conv_stride)` == `len(config.conv_kernel)`,'
f'''but is `len(config.conv_dim) = {len(self.conv_dim )}`, `len(config.conv_stride)'''
f'''= {len(self.conv_stride )}`, `len(config.conv_kernel) = {len(self.conv_kernel )}`.''' )
# fine-tuning config parameters for SpecAugment: https://arxiv.org/abs/1904.08779
__a : Tuple = apply_spec_augment
__a : Optional[Any] = mask_time_prob
__a : Any = mask_time_length
__a : List[str] = mask_time_min_masks
__a : List[str] = mask_feature_prob
__a : Tuple = mask_feature_length
__a : Any = mask_feature_min_masks
# ctc loss
__a : Optional[int] = ctc_loss_reduction
__a : List[Any] = ctc_zero_infinity
# sequence classification
__a : Dict = use_weighted_layer_sum
__a : Optional[Any] = classifier_proj_size
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return functools.reduce(operator.mul , self.conv_stride , 1 )
| 47 | 1 |
from copy import deepcopy
import torch
import torch.nn.functional as F
from torch.optim import AdamW
from torch.optim.lr_scheduler import LambdaLR
from torch.utils.data import DataLoader
from accelerate.accelerator import Accelerator
from accelerate.state import GradientState
from accelerate.test_utils import RegressionDataset, RegressionModel
from accelerate.utils import DistributedType, is_torch_version, set_seed
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[str] ):
for param, grad_param in zip(model_a.parameters() , model_b.parameters() ):
if not param.requires_grad:
continue
if not did_step:
# Grads should not be in sync
assert (
torch.allclose(param.grad , grad_param.grad ) is False
), f'''Gradients in sync when they should not be at iteration {iteration}:\nmodel_a grad ({param.grad}) == model_b grad ({grad_param.grad})'''
else:
# Grads should be in sync
assert (
torch.allclose(param.grad , grad_param.grad ) is True
), f'''Gradients not in sync when they should be at iteration {iteration}:\nmodel_a grad ({param.grad}) != model_b grad ({grad_param.grad})'''
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : Tuple , lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : List[Any] , lowerCamelCase_ : Optional[Any]=True ):
model.train()
__a : str = model(lowerCamelCase_ )
__a : List[str] = F.mse_loss(lowerCamelCase_ , target.to(output.device ) )
if not do_backward:
loss /= accelerator.gradient_accumulation_steps
loss.backward()
else:
accelerator.backward(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : str=False ):
set_seed(4_2 )
__a : Any = RegressionModel()
__a : List[str] = deepcopy(lowerCamelCase_ )
__a : Any = RegressionDataset(length=8_0 )
__a : Tuple = DataLoader(lowerCamelCase_ , batch_size=1_6 )
model.to(accelerator.device )
if sched:
__a : Tuple = AdamW(params=model.parameters() , lr=1e-3 )
__a : Tuple = AdamW(params=ddp_model.parameters() , lr=1e-3 )
__a : Optional[int] = LambdaLR(lowerCamelCase_ , lr_lambda=lambda lowerCamelCase_ : epoch**0.65 )
__a : List[Any] = LambdaLR(lowerCamelCase_ , lr_lambda=lambda lowerCamelCase_ : epoch**0.65 )
# Make a copy of `model`
if sched:
__a , __a , __a , __a : List[str] = accelerator.prepare(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
else:
__a , __a : Optional[Any] = accelerator.prepare(lowerCamelCase_ , lowerCamelCase_ )
if sched:
return (model, opt, sched, dataloader, ddp_model, ddp_opt, ddp_sched)
return model, ddp_model, dataloader
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ):
# Test when on a single CPU or GPU that the context manager does nothing
__a , __a , __a : Any = get_training_setup(lowerCamelCase_ )
# Use a single batch
__a , __a : int = next(iter(lowerCamelCase_ ) ).values()
for iteration in range(3 ):
# Gather the distributed inputs and targs for the base model
__a , __a : Dict = accelerator.gather((ddp_input, ddp_target) )
__a , __a : List[str] = input.to(accelerator.device ), target.to(accelerator.device )
# Perform our initial ground truth step in non "DDP"
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Do "gradient accumulation" (noop)
if iteration % 2 == 0:
# Accumulate grads locally
with accelerator.no_sync(lowerCamelCase_ ):
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
else:
# Sync grads
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Since `no_sync` is a noop, `ddp_model` and `model` grads should always be in sync
check_model_parameters(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
for param, ddp_param in zip(model.parameters() , ddp_model.parameters() ):
if not param.requires_grad:
continue
assert torch.allclose(
param.grad , ddp_param.grad ), f'''Gradients not in sync when they should be:\nModel grad ({param.grad}) != DDP grad ({ddp_param.grad})'''
# Shuffle ddp_input on each iteration
torch.manual_seed(1_3_3_7 + iteration )
__a : Union[str, Any] = ddp_input[torch.randperm(len(lowerCamelCase_ ) )]
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] ):
# Test on distributed setup that context manager behaves properly
__a , __a , __a : Any = get_training_setup(lowerCamelCase_ )
# Use a single batch
__a , __a : Optional[int] = next(iter(lowerCamelCase_ ) ).values()
for iteration in range(3 ):
# Gather the distributed inputs and targs for the base model
__a , __a : List[str] = accelerator.gather((ddp_input, ddp_target) )
__a , __a : List[str] = input.to(accelerator.device ), target.to(accelerator.device )
# Perform our initial ground truth step in non "DDP"
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Do "gradient accumulation" (noop)
if iteration % 2 == 0:
# Accumulate grads locally
with accelerator.no_sync(lowerCamelCase_ ):
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
else:
# Sync grads
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# DDP model and model should only be in sync when not (iteration % 2 == 0)
for param, ddp_param in zip(model.parameters() , ddp_model.parameters() ):
if not param.requires_grad:
continue
if iteration % 2 == 0:
# Grads should not be in sync
assert (
torch.allclose(param.grad , ddp_param.grad ) is False
), f'''Gradients in sync when they should not be:\nModel grad ({param.grad}) == DDP grad ({ddp_param.grad})'''
else:
# Grads should be in sync
assert (
torch.allclose(param.grad , ddp_param.grad ) is True
), f'''Gradients not in sync when they should be:\nModel grad ({param.grad}) != DDP grad ({ddp_param.grad})'''
# Shuffle ddp_input on each iteration
torch.manual_seed(1_3_3_7 + iteration )
__a : int = ddp_input[torch.randperm(len(lowerCamelCase_ ) )]
def UpperCAmelCase__ ( lowerCamelCase_ : List[str]=False , lowerCamelCase_ : Union[str, Any]=False ):
__a : Optional[int] = Accelerator(
split_batches=lowerCamelCase_ , dispatch_batches=lowerCamelCase_ , gradient_accumulation_steps=2 )
# Test that context manager behaves properly
__a , __a , __a : Optional[int] = get_training_setup(lowerCamelCase_ )
for iteration, batch in enumerate(lowerCamelCase_ ):
__a , __a : Dict = batch.values()
# Gather the distributed inputs and targs for the base model
__a , __a : Optional[Any] = accelerator.gather((ddp_input, ddp_target) )
__a , __a : int = input.to(accelerator.device ), target.to(accelerator.device )
# Perform our initial ground truth step in non "DDP"
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Do "gradient accumulation" (noop)
with accelerator.accumulate(lowerCamelCase_ ):
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# DDP model and model should only be in sync when not (iteration % 2 == 0)
for param, ddp_param in zip(model.parameters() , ddp_model.parameters() ):
if not param.requires_grad:
continue
if ((iteration + 1) % 2 == 0) or (iteration == len(lowerCamelCase_ ) - 1):
# Grads should be in sync
assert (
torch.allclose(param.grad , ddp_param.grad ) is True
), f'''Gradients not in sync when they should be at iteration {iteration}:\nModel grad ({param.grad}) != DDP grad ({ddp_param.grad})'''
else:
# Grads should not be in sync
assert (
torch.allclose(param.grad , ddp_param.grad ) is False
), f'''Gradients in sync when they should not be at iteration {iteration}:\nModel grad ({param.grad}) == DDP grad ({ddp_param.grad})'''
# Shuffle ddp_input on each iteration
torch.manual_seed(1_3_3_7 + iteration )
__a : Union[str, Any] = ddp_input[torch.randperm(len(lowerCamelCase_ ) )]
GradientState._reset_state()
def UpperCAmelCase__ ( lowerCamelCase_ : str=False , lowerCamelCase_ : Optional[int]=False ):
__a : Any = Accelerator(
split_batches=lowerCamelCase_ , dispatch_batches=lowerCamelCase_ , gradient_accumulation_steps=2 )
# Test that context manager behaves properly
__a , __a , __a , __a , __a , __a , __a : int = get_training_setup(lowerCamelCase_ , lowerCamelCase_ )
for iteration, batch in enumerate(lowerCamelCase_ ):
__a , __a : Optional[int] = batch.values()
# Gather the distributed inputs and targs for the base model
__a , __a : Optional[Any] = accelerator.gather((ddp_input, ddp_target) )
__a , __a : int = input.to(accelerator.device ), target.to(accelerator.device )
# Perform our initial ground truth step in non "DDP"
model.train()
ddp_model.train()
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
opt.step()
if ((iteration + 1) % 2 == 0) or ((iteration + 1) == len(lowerCamelCase_ )):
if split_batches:
sched.step()
else:
for _ in range(accelerator.num_processes ):
sched.step()
opt.zero_grad()
# Perform gradient accumulation under wrapper
with accelerator.accumulate(lowerCamelCase_ ):
step_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
ddp_opt.step()
ddp_sched.step()
ddp_opt.zero_grad()
# Learning rates should be the same
assert (
opt.param_groups[0]["lr"] == ddp_opt.param_groups[0]["lr"]
), f'''Learning rates found in each optimizer did not align\nopt: {opt.param_groups[0]["lr"]}\nDDP opt: {ddp_opt.param_groups[0]["lr"]}\n'''
__a : Tuple = (((iteration + 1) % 2) == 0) or ((iteration + 1) == len(lowerCamelCase_ ))
if accelerator.num_processes > 1:
check_model_parameters(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Shuffle ddp_input on each iteration
torch.manual_seed(1_3_3_7 + iteration )
GradientState._reset_state()
def UpperCAmelCase__ ( ):
__a : List[Any] = Accelerator()
__a : List[Any] = RegressionDataset(length=8_0 )
__a : str = DataLoader(lowerCamelCase_ , batch_size=1_6 )
__a : List[str] = RegressionDataset(length=9_6 )
__a : Optional[int] = DataLoader(lowerCamelCase_ , batch_size=1_6 )
__a , __a : Any = accelerator.prepare(lowerCamelCase_ , lowerCamelCase_ )
assert accelerator.gradient_state.active_dataloader is None
for iteration, _ in enumerate(lowerCamelCase_ ):
assert id(accelerator.gradient_state.active_dataloader ) == id(lowerCamelCase_ )
if iteration < len(lowerCamelCase_ ) - 1:
assert not accelerator.gradient_state.end_of_dataloader
if iteration == 1:
for batch_num, _ in enumerate(lowerCamelCase_ ):
assert id(accelerator.gradient_state.active_dataloader ) == id(lowerCamelCase_ )
if batch_num < len(lowerCamelCase_ ) - 1:
assert not accelerator.gradient_state.end_of_dataloader
else:
assert accelerator.gradient_state.end_of_dataloader
else:
assert accelerator.gradient_state.end_of_dataloader
assert accelerator.gradient_state.active_dataloader is None
def UpperCAmelCase__ ( ):
__a : int = Accelerator()
__a : Optional[Any] = accelerator.state
if state.local_process_index == 0:
print('**Test `accumulate` gradient accumulation with dataloader break**' )
test_dataloader_break()
if state.distributed_type == DistributedType.NO:
if state.local_process_index == 0:
print('**Test NOOP `no_sync` context manager**' )
test_noop_sync(lowerCamelCase_ )
if state.distributed_type in (DistributedType.MULTI_GPU, DistributedType.MULTI_CPU):
if state.local_process_index == 0:
print('**Test Distributed `no_sync` context manager**' )
test_distributed_sync(lowerCamelCase_ )
if state.distributed_type == DistributedType.MULTI_GPU:
for split_batch in [True, False]:
for dispatch_batches in [True, False]:
if state.local_process_index == 0:
print(
'**Test `accumulate` gradient accumulation, ' , f'''`split_batches={split_batch}` and `dispatch_batches={dispatch_batches}`**''' , )
test_gradient_accumulation(lowerCamelCase_ , lowerCamelCase_ )
# Currently will break on torch 2.0 +, need to investigate why
if is_torch_version('<' , '2.0' ) or state.distributed_type == DistributedType.NO:
if state.local_process_index == 0:
print(
'**Test `accumulate` gradient accumulation with optimizer and scheduler, ' , '`split_batches=False`, `dispatch_batches=False`**' , )
test_gradient_accumulation_with_opt_and_scheduler()
if state.distributed_type == DistributedType.MULTI_GPU:
for split_batch in [True, False]:
for dispatch_batches in [True, False]:
if not split_batch and not dispatch_batches:
continue
if state.local_process_index == 0:
print(
'**Test `accumulate` gradient accumulation with optimizer and scheduler, ' , f'''`split_batches={split_batch}` and `dispatch_batches={dispatch_batches}`**''' , )
test_gradient_accumulation_with_opt_and_scheduler(lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple ):
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main()
| 47 |
from __future__ import annotations
from sys import maxsize
from typing import Generic, TypeVar
SCREAMING_SNAKE_CASE__ = TypeVar('''T''')
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (position - 1) // 2
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 1
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 2
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[str] ):
'''simple docstring'''
__a : list[tuple[T, int]] = []
__a : dict[T, int] = {}
__a : int = 0
def __len__( self : Any ):
'''simple docstring'''
return self.elements
def __repr__( self : Any ):
'''simple docstring'''
return str(self.heap )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.elements == 0
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.heap.append((elem, weight) )
__a : List[Any] = self.elements
self.elements += 1
self._bubble_up(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self.elements > 1:
self._swap_nodes(0 , self.elements - 1 )
__a , __a : Union[str, Any] = self.heap.pop()
del self.position_map[elem]
self.elements -= 1
if self.elements > 0:
__a , __a : Dict = self.heap[0]
self._bubble_down(SCREAMING_SNAKE_CASE__ )
return elem
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
__a : str = (elem, weight)
if position > 0:
__a : Tuple = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = self.heap[parent_position]
if parent_weight > weight:
self._bubble_up(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
if curr_pos == 0:
return None
__a : List[str] = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : str = self.heap[curr_pos]
__a , __a : Optional[int] = self.heap[parent_position]
if parent_weight > weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_up(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : int = self.position_map[elem]
__a , __a : Optional[Any] = self.heap[curr_pos]
__a : Tuple = get_child_left_position(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = get_child_right_position(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements and child_right_position < self.elements:
__a , __a : str = self.heap[child_left_position]
__a , __a : List[str] = self.heap[child_right_position]
if child_right_weight < child_left_weight and child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements:
__a , __a : Any = self.heap[child_left_position]
if child_left_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
return None
if child_right_position < self.elements:
__a , __a : Union[str, Any] = self.heap[child_right_position]
if child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Optional[Any] = self.heap[nodea_pos][0]
__a : str = self.heap[nodea_pos][0]
__a , __a : int = (
self.heap[nodea_pos],
self.heap[nodea_pos],
)
__a : str = nodea_pos
__a : Optional[int] = nodea_pos
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[Any] ):
'''simple docstring'''
__a : dict[T, dict[T, int]] = {}
__a : int = 0
def __repr__( self : Tuple ):
'''simple docstring'''
return str(self.connections )
def __len__( self : Dict ):
'''simple docstring'''
return self.nodes
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
if node not in self.connections:
__a : Tuple = {}
self.nodes += 1
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.add_node(SCREAMING_SNAKE_CASE__ )
self.add_node(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = weight
__a : Any = weight
def UpperCAmelCase__ ( lowerCamelCase_ : GraphUndirectedWeighted[T] , ):
__a : dict[T, int] = {node: maxsize for node in graph.connections}
__a : dict[T, T | None] = {node: None for node in graph.connections}
__a : MinPriorityQueue[T] = MinPriorityQueue()
for node, weight in dist.items():
priority_queue.push(lowerCamelCase_ , lowerCamelCase_ )
if priority_queue.is_empty():
return dist, parent
# initialization
__a : Optional[int] = priority_queue.extract_min()
__a : int = 0
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : str = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Optional[int] = node
# running prim's algorithm
while not priority_queue.is_empty():
__a : Any = priority_queue.extract_min()
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : Tuple = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Dict = node
return dist, parent
| 47 | 1 |
# Usage:
# ./gen-card-facebook-wmt19.py
import os
from pathlib import Path
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : List[str] , lowerCamelCase_ : Dict ):
__a : List[str] = {
'en': 'Machine learning is great, isn\'t it?',
'ru': 'Машинное обучение - это здорово, не так ли?',
'de': 'Maschinelles Lernen ist großartig, oder?',
}
# BLUE scores as follows:
# "pair": [fairseq, transformers]
__a : List[Any] = {
'ru-en': ['[41.3](http://matrix.statmt.org/matrix/output/1907?run_id=6937)', '39.20'],
'en-ru': ['[36.4](http://matrix.statmt.org/matrix/output/1914?run_id=6724)', '33.47'],
'en-de': ['[43.1](http://matrix.statmt.org/matrix/output/1909?run_id=6862)', '42.83'],
'de-en': ['[42.3](http://matrix.statmt.org/matrix/output/1902?run_id=6750)', '41.35'],
}
__a : int = f'''{src_lang}-{tgt_lang}'''
__a : int = f'''
---
language:
- {src_lang}
- {tgt_lang}
thumbnail:
tags:
- translation
- wmt19
- facebook
license: apache-2.0
datasets:
- wmt19
metrics:
- bleu
---
# FSMT
## Model description
This is a ported version of [fairseq wmt19 transformer](https://github.com/pytorch/fairseq/blob/master/examples/wmt19/README.md) for {src_lang}-{tgt_lang}.
For more details, please see, [Facebook FAIR\'s WMT19 News Translation Task Submission](https://arxiv.org/abs/1907.06616).
The abbreviation FSMT stands for FairSeqMachineTranslation
All four models are available:
* [wmt19-en-ru](https://huggingface.co/facebook/wmt19-en-ru)
* [wmt19-ru-en](https://huggingface.co/facebook/wmt19-ru-en)
* [wmt19-en-de](https://huggingface.co/facebook/wmt19-en-de)
* [wmt19-de-en](https://huggingface.co/facebook/wmt19-de-en)
## Intended uses & limitations
#### How to use
```python
from transformers import FSMTForConditionalGeneration, FSMTTokenizer
mname = "facebook/wmt19-{src_lang}-{tgt_lang}"
tokenizer = FSMTTokenizer.from_pretrained(mname)
model = FSMTForConditionalGeneration.from_pretrained(mname)
input = "{texts[src_lang]}"
input_ids = tokenizer.encode(input, return_tensors="pt")
outputs = model.generate(input_ids)
decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(decoded) # {texts[tgt_lang]}
```
#### Limitations and bias
- The original (and this ported model) doesn\'t seem to handle well inputs with repeated sub-phrases, [content gets truncated](https://discuss.huggingface.co/t/issues-with-translating-inputs-containing-repeated-phrases/981)
## Training data
Pretrained weights were left identical to the original model released by fairseq. For more details, please, see the [paper](https://arxiv.org/abs/1907.06616).
## Eval results
pair | fairseq | transformers
-------|---------|----------
{pair} | {scores[pair][0]} | {scores[pair][1]}
The score is slightly below the score reported by `fairseq`, since `transformers`` currently doesn\'t support:
- model ensemble, therefore the best performing checkpoint was ported (``model4.pt``).
- re-ranking
The score was calculated using this code:
```bash
git clone https://github.com/huggingface/transformers
cd transformers
export PAIR={pair}
export DATA_DIR=data/$PAIR
export SAVE_DIR=data/$PAIR
export BS=8
export NUM_BEAMS=15
mkdir -p $DATA_DIR
sacrebleu -t wmt19 -l $PAIR --echo src > $DATA_DIR/val.source
sacrebleu -t wmt19 -l $PAIR --echo ref > $DATA_DIR/val.target
echo $PAIR
PYTHONPATH="src:examples/seq2seq" python examples/seq2seq/run_eval.py facebook/wmt19-$PAIR $DATA_DIR/val.source $SAVE_DIR/test_translations.txt --reference_path $DATA_DIR/val.target --score_path $SAVE_DIR/test_bleu.json --bs $BS --task translation --num_beams $NUM_BEAMS
```
note: fairseq reports using a beam of 50, so you should get a slightly higher score if re-run with `--num_beams 50`.
## Data Sources
- [training, etc.](http://www.statmt.org/wmt19/)
- [test set](http://matrix.statmt.org/test_sets/newstest2019.tgz?1556572561)
### BibTeX entry and citation info
```bibtex
@inproceedings{{...,
year={{2020}},
title={{Facebook FAIR\'s WMT19 News Translation Task Submission}},
author={{Ng, Nathan and Yee, Kyra and Baevski, Alexei and Ott, Myle and Auli, Michael and Edunov, Sergey}},
booktitle={{Proc. of WMT}},
}}
```
## TODO
- port model ensemble (fairseq uses 4 model checkpoints)
'''
os.makedirs(lowerCamelCase_ , exist_ok=lowerCamelCase_ )
__a : Optional[int] = os.path.join(lowerCamelCase_ , 'README.md' )
print(f'''Generating {path}''' )
with open(lowerCamelCase_ , 'w' , encoding='utf-8' ) as f:
f.write(lowerCamelCase_ )
# make sure we are under the root of the project
SCREAMING_SNAKE_CASE__ = Path(__file__).resolve().parent.parent.parent
SCREAMING_SNAKE_CASE__ = repo_dir / '''model_cards'''
for model_name in ["wmt19-ru-en", "wmt19-en-ru", "wmt19-en-de", "wmt19-de-en"]:
SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ = model_name.split('''-''')
SCREAMING_SNAKE_CASE__ = model_cards_dir / '''facebook''' / model_name
write_model_card(model_card_dir, src_lang=src_lang, tgt_lang=tgt_lang)
| 47 |
from collections.abc import Sequence
from queue import Queue
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : int=None , SCREAMING_SNAKE_CASE__ : Tuple=None ):
'''simple docstring'''
__a : Tuple = start
__a : Dict = end
__a : List[str] = val
__a : List[Any] = (start + end) // 2
__a : Optional[Any] = left
__a : List[str] = right
def __repr__( self : Dict ):
'''simple docstring'''
return f'''SegmentTreeNode(start={self.start}, end={self.end}, val={self.val})'''
class _UpperCamelCase:
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : Sequence , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Tuple = collection
__a : Dict = function
if self.collection:
__a : int = self._build_tree(0 , len(SCREAMING_SNAKE_CASE__ ) - 1 )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self._update_tree(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
return self._query_range(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if start == end:
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.collection[start] )
__a : Tuple = (start + end) // 2
__a : Optional[int] = self._build_tree(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = self._build_tree(mid + 1 , SCREAMING_SNAKE_CASE__ )
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.fn(left.val , right.val ) , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if node.start == i and node.end == i:
__a : Optional[Any] = val
return
if i <= node.mid:
self._update_tree(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
self._update_tree(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = self.fn(node.left.val , node.right.val )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
if node.start == i and node.end == j:
return node.val
if i <= node.mid:
if j <= node.mid:
# range in left child tree
return self._query_range(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
# range in left child tree and right child tree
return self.fn(
self._query_range(node.left , SCREAMING_SNAKE_CASE__ , node.mid ) , self._query_range(node.right , node.mid + 1 , SCREAMING_SNAKE_CASE__ ) , )
else:
# range in right child tree
return self._query_range(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
if self.root is not None:
__a : Tuple = Queue()
queue.put(self.root )
while not queue.empty():
__a : Tuple = queue.get()
yield node
if node.left is not None:
queue.put(node.left )
if node.right is not None:
queue.put(node.right )
if __name__ == "__main__":
import operator
for fn in [operator.add, max, min]:
print('''*''' * 50)
SCREAMING_SNAKE_CASE__ = SegmentTree([2, 1, 5, 3, 4], fn)
for node in arr.traverse():
print(node)
print()
arr.update(1, 5)
for node in arr.traverse():
print(node)
print()
print(arr.query_range(3, 4)) # 7
print(arr.query_range(2, 2)) # 5
print(arr.query_range(1, 3)) # 13
print()
| 47 | 1 |
import unittest
from parameterized import parameterized
from transformers import AutoTokenizer, GPTNeoXConfig, is_torch_available, set_seed
from transformers.testing_utils import require_torch, slow, torch_device
from ...generation.test_utils import GenerationTesterMixin
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import (
GPTNeoXForCausalLM,
GPTNeoXForQuestionAnswering,
GPTNeoXForSequenceClassification,
GPTNeoXForTokenClassification,
GPTNeoXModel,
)
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Tuple=1_3 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=7 , SCREAMING_SNAKE_CASE__ : List[Any]=True , SCREAMING_SNAKE_CASE__ : int=True , SCREAMING_SNAKE_CASE__ : Optional[Any]=True , SCREAMING_SNAKE_CASE__ : List[str]=True , SCREAMING_SNAKE_CASE__ : Tuple=9_9 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=6_4 , SCREAMING_SNAKE_CASE__ : str=5 , SCREAMING_SNAKE_CASE__ : Optional[Any]=4 , SCREAMING_SNAKE_CASE__ : Optional[int]=3_7 , SCREAMING_SNAKE_CASE__ : str="gelu" , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0.1 , SCREAMING_SNAKE_CASE__ : int=0.1 , SCREAMING_SNAKE_CASE__ : Tuple=5_1_2 , SCREAMING_SNAKE_CASE__ : List[Any]=1_6 , SCREAMING_SNAKE_CASE__ : int=2 , SCREAMING_SNAKE_CASE__ : Dict=0.02 , SCREAMING_SNAKE_CASE__ : Optional[int]=3 , SCREAMING_SNAKE_CASE__ : str=4 , SCREAMING_SNAKE_CASE__ : Dict=None , ):
'''simple docstring'''
__a : Tuple = parent
__a : str = batch_size
__a : int = seq_length
__a : Union[str, Any] = is_training
__a : str = use_input_mask
__a : Dict = use_token_type_ids
__a : int = use_labels
__a : Tuple = vocab_size
__a : Union[str, Any] = hidden_size
__a : Optional[Any] = num_hidden_layers
__a : List[str] = num_attention_heads
__a : List[str] = intermediate_size
__a : Dict = hidden_act
__a : str = hidden_dropout_prob
__a : str = attention_probs_dropout_prob
__a : List[str] = max_position_embeddings
__a : List[Any] = type_vocab_size
__a : List[Any] = type_sequence_label_size
__a : str = initializer_range
__a : Optional[Any] = num_labels
__a : List[str] = num_choices
__a : Tuple = scope
__a : int = vocab_size - 1
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
__a : Optional[Any] = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a : Optional[Any] = None
if self.use_input_mask:
__a : Tuple = random_attention_mask([self.batch_size, self.seq_length] )
__a : Tuple = None
if self.use_labels:
__a : Tuple = ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a : Dict = self.get_config()
return config, input_ids, input_mask, token_labels
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
return GPTNeoXConfig(
vocab_size=self.vocab_size , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , max_position_embeddings=self.max_position_embeddings , type_vocab_size=self.type_vocab_size , is_decoder=SCREAMING_SNAKE_CASE__ , initializer_range=self.initializer_range , pad_token_id=self.pad_token_id , )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a , __a , __a , __a : Optional[int] = self.prepare_config_and_inputs()
__a : str = True
return config, input_ids, input_mask, token_labels
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Tuple = GPTNeoXModel(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Dict = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ )
__a : Tuple = model(SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : int = True
__a : str = GPTNeoXModel(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Optional[Any] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a : str = GPTNeoXForCausalLM(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Union[str, Any] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Any = self.num_labels
__a : int = GPTNeoXForQuestionAnswering(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : List[str] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.start_logits.shape , (self.batch_size, self.seq_length) )
self.parent.assertEqual(result.end_logits.shape , (self.batch_size, self.seq_length) )
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : Dict = self.num_labels
__a : List[str] = GPTNeoXForSequenceClassification(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a : List[str] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : Dict = self.num_labels
__a : Dict = GPTNeoXForTokenClassification(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : str = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Optional[Any] = True
__a : str = GPTNeoXForCausalLM(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
# first forward pass
__a : Optional[Any] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , use_cache=SCREAMING_SNAKE_CASE__ )
__a : Tuple = outputs.past_key_values
# create hypothetical multiple next token and extent to next_input_ids
__a : Optional[Any] = ids_tensor((self.batch_size, 3) , config.vocab_size )
__a : List[Any] = ids_tensor((self.batch_size, 3) , vocab_size=2 )
# append to next input_ids and
__a : str = torch.cat([input_ids, next_tokens] , dim=-1 )
__a : Dict = torch.cat([input_mask, next_mask] , dim=-1 )
__a : int = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , output_hidden_states=SCREAMING_SNAKE_CASE__ )
__a : Dict = output_from_no_past['hidden_states'][0]
__a : List[str] = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , past_key_values=SCREAMING_SNAKE_CASE__ , output_hidden_states=SCREAMING_SNAKE_CASE__ , )['hidden_states'][0]
# select random slice
__a : Any = ids_tensor((1,) , output_from_past.shape[-1] ).item()
__a : Union[str, Any] = output_from_no_past[:, -3:, random_slice_idx].detach()
__a : List[str] = output_from_past[:, :, random_slice_idx].detach()
self.parent.assertTrue(output_from_past_slice.shape[1] == next_tokens.shape[1] )
# test that outputs are equal for slice
self.parent.assertTrue(torch.allclose(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , atol=1e-3 ) )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Union[str, Any] = self.prepare_config_and_inputs()
__a , __a , __a , __a : str = config_and_inputs
__a : Tuple = {'input_ids': input_ids, 'attention_mask': input_mask}
return config, inputs_dict
@require_torch
class _UpperCamelCase( __lowerCamelCase , __lowerCamelCase , __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : Optional[int] = (
(
GPTNeoXModel,
GPTNeoXForCausalLM,
GPTNeoXForQuestionAnswering,
GPTNeoXForSequenceClassification,
GPTNeoXForTokenClassification,
)
if is_torch_available()
else ()
)
__SCREAMING_SNAKE_CASE : str = (GPTNeoXForCausalLM,) if is_torch_available() else ()
__SCREAMING_SNAKE_CASE : Optional[int] = (
{
'''feature-extraction''': GPTNeoXModel,
'''question-answering''': GPTNeoXForQuestionAnswering,
'''text-classification''': GPTNeoXForSequenceClassification,
'''text-generation''': GPTNeoXForCausalLM,
'''token-classification''': GPTNeoXForTokenClassification,
'''zero-shot''': GPTNeoXForSequenceClassification,
}
if is_torch_available()
else {}
)
__SCREAMING_SNAKE_CASE : Union[str, Any] = False
__SCREAMING_SNAKE_CASE : Optional[int] = False
__SCREAMING_SNAKE_CASE : str = False
__SCREAMING_SNAKE_CASE : Any = False
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
__a : str = GPTNeoXModelTester(self )
__a : Union[str, Any] = ConfigTester(self , config_class=SCREAMING_SNAKE_CASE__ , hidden_size=6_4 , num_attention_heads=8 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
self.config_tester.run_common_tests()
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a , __a , __a , __a : str = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a , __a , __a , __a : Dict = self.model_tester.prepare_config_and_inputs_for_decoder()
self.model_tester.create_and_check_model_as_decoder(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a , __a , __a , __a : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_decoder()
__a : int = None
self.model_tester.create_and_check_model_as_decoder(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a , __a , __a , __a : Union[str, Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_decoder_model_past_large_inputs(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : List[str] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_causal_lm(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : Any = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_question_answering(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : Tuple = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_sequence_classification(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Any = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_token_classification(*SCREAMING_SNAKE_CASE__ )
@unittest.skip(reason='Feed forward chunking is not implemented' )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
pass
@parameterized.expand([('linear',), ('dynamic',)] )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
__a , __a : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
__a : int = ids_tensor([1, 1_0] , config.vocab_size )
__a : Optional[Any] = ids_tensor([1, int(config.max_position_embeddings * 1.5 )] , config.vocab_size )
set_seed(4_2 ) # Fixed seed at init time so the two models get the same random weights
__a : List[str] = GPTNeoXModel(SCREAMING_SNAKE_CASE__ )
original_model.to(SCREAMING_SNAKE_CASE__ )
original_model.eval()
__a : Union[str, Any] = original_model(SCREAMING_SNAKE_CASE__ ).last_hidden_state
__a : int = original_model(SCREAMING_SNAKE_CASE__ ).last_hidden_state
set_seed(4_2 ) # Fixed seed at init time so the two models get the same random weights
__a : str = {'type': scaling_type, 'factor': 10.0}
__a : Any = GPTNeoXModel(SCREAMING_SNAKE_CASE__ )
scaled_model.to(SCREAMING_SNAKE_CASE__ )
scaled_model.eval()
__a : Optional[int] = scaled_model(SCREAMING_SNAKE_CASE__ ).last_hidden_state
__a : Union[str, Any] = scaled_model(SCREAMING_SNAKE_CASE__ ).last_hidden_state
# Dynamic scaling does not change the RoPE embeddings until it receives an input longer than the original
# maximum sequence length, so the outputs for the short input should match.
if scaling_type == "dynamic":
self.assertTrue(torch.allclose(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , atol=1e-5 ) )
else:
self.assertFalse(torch.allclose(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , atol=1e-5 ) )
# The output should be different for long inputs
self.assertFalse(torch.allclose(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , atol=1e-5 ) )
@require_torch
class _UpperCamelCase( unittest.TestCase ):
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : str = AutoTokenizer.from_pretrained('EleutherAI/pythia-410m-deduped' )
for checkpointing in [True, False]:
__a : str = GPTNeoXForCausalLM.from_pretrained('EleutherAI/pythia-410m-deduped' )
if checkpointing:
model.gradient_checkpointing_enable()
else:
model.gradient_checkpointing_disable()
model.to(SCREAMING_SNAKE_CASE__ )
__a : Dict = tokenizer('My favorite food is' , return_tensors='pt' ).to(SCREAMING_SNAKE_CASE__ )
# The hub repo. is updated on 2023-04-04, resulting in poor outputs.
# See: https://github.com/huggingface/transformers/pull/24193
__a : Optional[int] = 'My favorite food is a good old-fashioned, old-fashioned, old-fashioned.\n\nI\'m not sure'
__a : Tuple = model.generate(**SCREAMING_SNAKE_CASE__ , do_sample=SCREAMING_SNAKE_CASE__ , max_new_tokens=2_0 )
__a : int = tokenizer.batch_decode(SCREAMING_SNAKE_CASE__ )[0]
self.assertEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
| 47 |
import os
import posixpath
import uuid
from dataclasses import dataclass
from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, Union
import numpy as np
import pyarrow as pa
import datasets
from datasets.arrow_writer import ArrowWriter, ParquetWriter
from datasets.config import MAX_SHARD_SIZE
from datasets.filesystems import (
is_remote_filesystem,
rename,
)
from datasets.iterable_dataset import _BaseExamplesIterable
from datasets.utils.py_utils import convert_file_size_to_int
SCREAMING_SNAKE_CASE__ = datasets.utils.logging.get_logger(__name__)
if TYPE_CHECKING:
import pyspark
@dataclass
class _UpperCamelCase( datasets.BuilderConfig ):
__SCREAMING_SNAKE_CASE : Optional[datasets.Features] = None
def UpperCAmelCase__ ( lowerCamelCase_ : "pyspark.sql.DataFrame" , lowerCamelCase_ : List[int] , ):
import pyspark
def generate_fn():
__a : List[Any] = df.select('*' , pyspark.sql.functions.spark_partition_id().alias('part_id' ) )
for partition_id in partition_order:
__a : Optional[int] = df_with_partition_id.select('*' ).where(f'''part_id = {partition_id}''' ).drop('part_id' )
__a : Optional[Any] = partition_df.collect()
__a : Union[str, Any] = 0
for row in rows:
yield f'''{partition_id}_{row_id}''', row.asDict()
row_id += 1
return generate_fn
class _UpperCamelCase( _BaseExamplesIterable ):
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : Dict=None , ):
'''simple docstring'''
__a : List[str] = df
__a : Tuple = partition_order or range(self.df.rdd.getNumPartitions() )
__a : List[Any] = _generate_iterable_examples(self.df , self.partition_order )
def __iter__( self : Tuple ):
'''simple docstring'''
yield from self.generate_examples_fn()
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : np.random.Generator ):
'''simple docstring'''
__a : Union[str, Any] = list(range(self.df.rdd.getNumPartitions() ) )
generator.shuffle(SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Union[str, Any] = self.split_shard_indices_by_worker(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return len(self.partition_order )
class _UpperCamelCase( datasets.DatasetBuilder ):
__SCREAMING_SNAKE_CASE : List[str] = SparkConfig
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : str = None , SCREAMING_SNAKE_CASE__ : str = None , **SCREAMING_SNAKE_CASE__ : Optional[int] , ):
'''simple docstring'''
import pyspark
__a : int = pyspark.sql.SparkSession.builder.getOrCreate()
__a : Optional[int] = df
__a : List[Any] = working_dir
super().__init__(
cache_dir=SCREAMING_SNAKE_CASE__ , config_name=str(self.df.semanticHash() ) , **SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
def create_cache_and_write_probe(SCREAMING_SNAKE_CASE__ : List[str] ):
# makedirs with exist_ok will recursively create the directory. It will not throw an error if directories
# already exist.
os.makedirs(self._cache_dir , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(self._cache_dir , 'fs_test' + uuid.uuida().hex )
# Opening the file in append mode will create a new file unless it already exists, in which case it will not
# change the file contents.
open(SCREAMING_SNAKE_CASE__ , 'a' )
return [probe_file]
if self._spark.conf.get('spark.master' , '' ).startswith('local' ):
return
# If the cluster is multi-node, make sure that the user provided a cache_dir and that it is on an NFS
# accessible to the driver.
# TODO: Stream batches to the driver using ArrowCollectSerializer instead of throwing an error.
if self._cache_dir:
__a : List[Any] = (
self._spark.sparkContext.parallelize(range(1 ) , 1 ).mapPartitions(SCREAMING_SNAKE_CASE__ ).collect()
)
if os.path.isfile(probe[0] ):
return
raise ValueError(
'When using Dataset.from_spark on a multi-node cluster, the driver and all workers should be able to access cache_dir' )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
return datasets.DatasetInfo(features=self.config.features )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : datasets.download.download_manager.DownloadManager ):
'''simple docstring'''
return [datasets.SplitGenerator(name=datasets.Split.TRAIN )]
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
import pyspark
def get_arrow_batch_size(SCREAMING_SNAKE_CASE__ : int ):
for batch in it:
yield pa.RecordBatch.from_pydict({'batch_bytes': [batch.nbytes]} )
__a : List[str] = self.df.count()
__a : Dict = df_num_rows if df_num_rows <= 1_0_0 else 1_0_0
# Approximate the size of each row (in Arrow format) by averaging over a max-100-row sample.
__a : List[str] = (
self.df.limit(SCREAMING_SNAKE_CASE__ )
.repartition(1 )
.mapInArrow(SCREAMING_SNAKE_CASE__ , 'batch_bytes: long' )
.agg(pyspark.sql.functions.sum('batch_bytes' ).alias('sample_bytes' ) )
.collect()[0]
.sample_bytes
/ sample_num_rows
)
__a : Dict = approx_bytes_per_row * df_num_rows
if approx_total_size > max_shard_size:
# Make sure there is at least one row per partition.
__a : Union[str, Any] = min(SCREAMING_SNAKE_CASE__ , int(approx_total_size / max_shard_size ) )
__a : int = self.df.repartition(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : int , ):
'''simple docstring'''
import pyspark
__a : Any = ParquetWriter if file_format == 'parquet' else ArrowWriter
__a : Union[str, Any] = os.path.join(self._working_dir , os.path.basename(SCREAMING_SNAKE_CASE__ ) ) if self._working_dir else fpath
__a : Optional[int] = file_format == 'parquet'
# Define these so that we don't reference self in write_arrow, which will result in a pickling error due to
# pickling the SparkContext.
__a : List[str] = self.config.features
__a : int = self._writer_batch_size
__a : Union[str, Any] = self._fs.storage_options
def write_arrow(SCREAMING_SNAKE_CASE__ : Optional[int] ):
# Within the same SparkContext, no two task attempts will share the same attempt ID.
__a : Any = pyspark.TaskContext().taskAttemptId()
__a : str = next(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
if first_batch is None:
# Some partitions might not receive any data.
return pa.RecordBatch.from_arrays(
[[task_id], [0], [0]] , names=['task_id', 'num_examples', 'num_bytes'] , )
__a : Any = 0
__a : List[str] = writer_class(
features=SCREAMING_SNAKE_CASE__ , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Optional[Any] = pa.Table.from_batches([first_batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
for batch in it:
if max_shard_size is not None and writer._num_bytes >= max_shard_size:
__a , __a : Optional[int] = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
shard_id += 1
__a : Optional[Any] = writer_class(
features=writer._features , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Union[str, Any] = pa.Table.from_batches([batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
if writer._num_bytes > 0:
__a , __a : str = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
if working_fpath != fpath:
for file in os.listdir(os.path.dirname(SCREAMING_SNAKE_CASE__ ) ):
__a : Any = os.path.join(os.path.dirname(SCREAMING_SNAKE_CASE__ ) , os.path.basename(SCREAMING_SNAKE_CASE__ ) )
shutil.move(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Dict = (
self.df.mapInArrow(SCREAMING_SNAKE_CASE__ , 'task_id: long, num_examples: long, num_bytes: long' )
.groupBy('task_id' )
.agg(
pyspark.sql.functions.sum('num_examples' ).alias('total_num_examples' ) , pyspark.sql.functions.sum('num_bytes' ).alias('total_num_bytes' ) , pyspark.sql.functions.count('num_bytes' ).alias('num_shards' ) , pyspark.sql.functions.collect_list('num_examples' ).alias('shard_lengths' ) , )
.collect()
)
for row in stats:
yield row.task_id, (row.total_num_examples, row.total_num_bytes, row.num_shards, row.shard_lengths)
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , SCREAMING_SNAKE_CASE__ : str = "arrow" , SCREAMING_SNAKE_CASE__ : Optional[Union[str, int]] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
self._validate_cache_dir()
__a : List[str] = convert_file_size_to_int(max_shard_size or MAX_SHARD_SIZE )
self._repartition_df_if_needed(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = not is_remote_filesystem(self._fs )
__a : Optional[Any] = os.path.join if is_local else posixpath.join
__a : Any = '-TTTTT-SSSSS-of-NNNNN'
__a : Union[str, Any] = f'''{self.name}-{split_generator.name}{SUFFIX}.{file_format}'''
__a : Any = path_join(self._output_dir , SCREAMING_SNAKE_CASE__ )
__a : Any = 0
__a : Dict = 0
__a : int = 0
__a : List[str] = []
__a : Optional[int] = []
for task_id, content in self._prepare_split_single(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : Optional[int] = content
if num_bytes > 0:
total_num_examples += num_examples
total_num_bytes += num_bytes
total_shards += num_shards
task_id_and_num_shards.append((task_id, num_shards) )
all_shard_lengths.extend(SCREAMING_SNAKE_CASE__ )
__a : List[str] = total_num_examples
__a : Optional[int] = total_num_bytes
# should rename everything at the end
logger.debug(f'''Renaming {total_shards} shards.''' )
if total_shards > 1:
__a : Any = all_shard_lengths
# Define fs outside of _rename_shard so that we don't reference self in the function, which will result in a
# pickling error due to pickling the SparkContext.
__a : Dict = self._fs
# use the -SSSSS-of-NNNNN pattern
def _rename_shard(
SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , ):
rename(
SCREAMING_SNAKE_CASE__ , fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace('TTTTT-SSSSS' , f'''{global_shard_id:05d}''' ).replace('NNNNN' , f'''{total_shards:05d}''' ) , )
__a : Union[str, Any] = []
__a : List[str] = 0
for i in range(len(SCREAMING_SNAKE_CASE__ ) ):
__a , __a : Union[str, Any] = task_id_and_num_shards[i]
for shard_id in range(SCREAMING_SNAKE_CASE__ ):
args.append([task_id, shard_id, global_shard_id] )
global_shard_id += 1
self._spark.sparkContext.parallelize(SCREAMING_SNAKE_CASE__ , len(SCREAMING_SNAKE_CASE__ ) ).map(lambda SCREAMING_SNAKE_CASE__ : _rename_shard(*SCREAMING_SNAKE_CASE__ ) ).collect()
else:
# don't use any pattern
__a : List[Any] = 0
__a : Any = task_id_and_num_shards[0][0]
self._rename(
fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace(SCREAMING_SNAKE_CASE__ , '' ) , )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , ):
'''simple docstring'''
return SparkExamplesIterable(self.df )
| 47 | 1 |
import argparse
import evaluate
import torch
from datasets import load_dataset
from torch.optim import AdamW
from torch.utils.data import DataLoader
from transformers import AutoModelForSequenceClassification, AutoTokenizer, get_linear_schedule_with_warmup, set_seed
from accelerate import Accelerator, DistributedType
########################################################################
# This is a fully working simple example to use Accelerate
#
# This example trains a Bert base model on GLUE MRPC
# in any of the following settings (with the same script):
# - single CPU or single GPU
# - multi GPUS (using PyTorch distributed mode)
# - (multi) TPUs
# - fp16 (mixed-precision) or fp32 (normal precision)
#
# To run it in each of these various modes, follow the instructions
# in the readme for examples:
# https://github.com/huggingface/accelerate/tree/main/examples
#
########################################################################
SCREAMING_SNAKE_CASE__ = 16
SCREAMING_SNAKE_CASE__ = 32
def UpperCAmelCase__ ( lowerCamelCase_ : Accelerator , lowerCamelCase_ : int = 1_6 ):
__a : Tuple = AutoTokenizer.from_pretrained('bert-base-cased' )
__a : int = load_dataset('glue' , 'mrpc' )
def tokenize_function(lowerCamelCase_ : Union[str, Any] ):
# max_length=None => use the model max length (it's actually the default)
__a : int = tokenizer(examples['sentence1'] , examples['sentence2'] , truncation=lowerCamelCase_ , max_length=lowerCamelCase_ )
return outputs
# Apply the method we just defined to all the examples in all the splits of the dataset
# starting with the main process first:
with accelerator.main_process_first():
__a : Optional[Any] = datasets.map(
lowerCamelCase_ , batched=lowerCamelCase_ , remove_columns=['idx', 'sentence1', 'sentence2'] , )
# We also rename the 'label' column to 'labels' which is the expected name for labels by the models of the
# transformers library
__a : Optional[int] = tokenized_datasets.rename_column('label' , 'labels' )
def collate_fn(lowerCamelCase_ : Dict ):
# On TPU it's best to pad everything to the same length or training will be very slow.
__a : Optional[int] = 1_2_8 if accelerator.distributed_type == DistributedType.TPU else None
# When using mixed precision we want round multiples of 8/16
if accelerator.mixed_precision == "fp8":
__a : List[Any] = 1_6
elif accelerator.mixed_precision != "no":
__a : Union[str, Any] = 8
else:
__a : List[str] = None
return tokenizer.pad(
lowerCamelCase_ , padding='longest' , max_length=lowerCamelCase_ , pad_to_multiple_of=lowerCamelCase_ , return_tensors='pt' , )
# Instantiate dataloaders.
__a : List[str] = DataLoader(
tokenized_datasets['train'] , shuffle=lowerCamelCase_ , collate_fn=lowerCamelCase_ , batch_size=lowerCamelCase_ , drop_last=lowerCamelCase_ )
__a : List[str] = DataLoader(
tokenized_datasets['validation'] , shuffle=lowerCamelCase_ , collate_fn=lowerCamelCase_ , batch_size=lowerCamelCase_ , drop_last=(accelerator.mixed_precision == 'fp8') , )
return train_dataloader, eval_dataloader
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : List[str] ):
# Initialize accelerator
__a : Tuple = Accelerator(cpu=args.cpu , mixed_precision=args.mixed_precision )
# Sample hyper-parameters for learning rate, batch size, seed and a few other HPs
__a : List[str] = config['lr']
__a : Tuple = int(config['num_epochs'] )
__a : Tuple = int(config['seed'] )
__a : Union[str, Any] = int(config['batch_size'] )
__a : str = evaluate.load('glue' , 'mrpc' )
# If the batch size is too big we use gradient accumulation
__a : int = 1
if batch_size > MAX_GPU_BATCH_SIZE and accelerator.distributed_type != DistributedType.TPU:
__a : str = batch_size // MAX_GPU_BATCH_SIZE
__a : Optional[int] = MAX_GPU_BATCH_SIZE
set_seed(lowerCamelCase_ )
__a , __a : List[str] = get_dataloaders(lowerCamelCase_ , lowerCamelCase_ )
# Instantiate the model (we build the model here so that the seed also control new weights initialization)
__a : Dict = AutoModelForSequenceClassification.from_pretrained('bert-base-cased' , return_dict=lowerCamelCase_ )
# We could avoid this line since the accelerator is set with `device_placement=True` (default value).
# Note that if you are placing tensors on devices manually, this line absolutely needs to be before the optimizer
# creation otherwise training will not work on TPU (`accelerate` will kindly throw an error to make us aware of that).
__a : int = model.to(accelerator.device )
# Instantiate optimizer
__a : Any = AdamW(params=model.parameters() , lr=lowerCamelCase_ )
# Instantiate scheduler
__a : Optional[int] = get_linear_schedule_with_warmup(
optimizer=lowerCamelCase_ , num_warmup_steps=1_0_0 , num_training_steps=(len(lowerCamelCase_ ) * num_epochs) // gradient_accumulation_steps , )
# Prepare everything
# There is no specific order to remember, we just need to unpack the objects in the same order we gave them to the
# prepare method.
__a , __a , __a , __a , __a : Dict = accelerator.prepare(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Now we train the model
for epoch in range(lowerCamelCase_ ):
model.train()
for step, batch in enumerate(lowerCamelCase_ ):
# We could avoid this line since we set the accelerator with `device_placement=True`.
batch.to(accelerator.device )
__a : int = model(**lowerCamelCase_ )
__a : Optional[Any] = outputs.loss
__a : Optional[int] = loss / gradient_accumulation_steps
accelerator.backward(lowerCamelCase_ )
if step % gradient_accumulation_steps == 0:
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
model.eval()
for step, batch in enumerate(lowerCamelCase_ ):
# We could avoid this line since we set the accelerator with `device_placement=True`.
batch.to(accelerator.device )
with torch.no_grad():
__a : List[str] = model(**lowerCamelCase_ )
__a : Any = outputs.logits.argmax(dim=-1 )
__a , __a : int = accelerator.gather_for_metrics((predictions, batch['labels']) )
metric.add_batch(
predictions=lowerCamelCase_ , references=lowerCamelCase_ , )
__a : Optional[Any] = metric.compute()
# Use accelerator.print to print only on the main process.
accelerator.print(f'''epoch {epoch}:''' , lowerCamelCase_ )
def UpperCAmelCase__ ( ):
__a : Optional[int] = argparse.ArgumentParser(description='Simple example of training script.' )
parser.add_argument(
'--mixed_precision' , type=lowerCamelCase_ , default=lowerCamelCase_ , choices=['no', 'fp16', 'bf16', 'fp8'] , help='Whether to use mixed precision. Choose'
'between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >= 1.10.'
'and an Nvidia Ampere GPU.' , )
parser.add_argument('--cpu' , action='store_true' , help='If passed, will train on the CPU.' )
__a : List[Any] = parser.parse_args()
__a : Optional[int] = {'lr': 2e-5, 'num_epochs': 3, 'seed': 4_2, 'batch_size': 1_6}
training_function(lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 |
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, TensorDataset
from tqdm import tqdm
from transformers import GPTaLMHeadModel
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : int ):
# save results
if os.path.exists(lowerCamelCase_ ):
if os.path.exists(os.path.join(lowerCamelCase_ , 'config.json' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'config.json' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'config.json' ) )
if os.path.exists(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) )
else:
os.makedirs(lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Any=False ):
__a : Dict = 2
if unlogit:
__a : Optional[Any] = torch.pow(lowerCamelCase_ , lowerCamelCase_ )
__a : Any = p * torch.log(lowerCamelCase_ )
__a : Union[str, Any] = 0
return -plogp.sum(dim=-1 )
def UpperCAmelCase__ ( lowerCamelCase_ : Any ):
logger.info('lv, h >\t' + '\t'.join(f'''{x + 1}''' for x in range(len(lowerCamelCase_ ) ) ) )
for row in range(len(lowerCamelCase_ ) ):
if tensor.dtype != torch.long:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:.5f}''' for x in tensor[row].cpu().data ) )
else:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:d}''' for x in tensor[row].cpu().data ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Any , lowerCamelCase_ : int , lowerCamelCase_ : int=True , lowerCamelCase_ : Optional[Any]=True , lowerCamelCase_ : List[Any]=None , lowerCamelCase_ : List[Any]=False ):
__a , __a : Optional[int] = model.config.num_hidden_layers, model.config.num_attention_heads
__a : str = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
__a : int = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
if head_mask is None:
__a : Union[str, Any] = torch.ones(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
head_mask.requires_grad_(requires_grad=lowerCamelCase_ )
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
__a : Any = None
__a : Optional[int] = 0.0
__a : Optional[Any] = 0.0
for step, inputs in enumerate(tqdm(lowerCamelCase_ , desc='Iteration' , disable=args.local_rank not in [-1, 0] ) ):
__a : Dict = tuple(t.to(args.device ) for t in inputs )
((__a) , ) : Dict = inputs
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
__a : List[Any] = model(lowerCamelCase_ , labels=lowerCamelCase_ , head_mask=lowerCamelCase_ )
# (loss), lm_logits, presents, (all hidden_states), (attentions)
__a , __a , __a : int = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
total_loss += loss.detach().cpu().numpy()
if compute_entropy:
for layer, attn in enumerate(lowerCamelCase_ ):
__a : List[str] = entropy(attn.detach() , lowerCamelCase_ )
attn_entropy[layer] += masked_entropy.sum(-1 ).sum(0 ).sum(0 ).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
tot_tokens += torch.ones_like(lowerCamelCase_ ).float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
__a : Optional[Any] = 2
__a : Union[str, Any] = torch.pow(torch.pow(lowerCamelCase_ , lowerCamelCase_ ).sum(-1 ) , 1 / exponent )
head_importance /= norm_by_layer.unsqueeze(-1 ) + 1e-20
if not args.dont_normalize_global_importance:
__a : List[str] = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print matrices
if compute_entropy:
logger.info('Attention entropies' )
print_ad_tensor(lowerCamelCase_ )
if compute_importance:
logger.info('Head importance scores' )
print_ad_tensor(lowerCamelCase_ )
logger.info('Head ranked by importance scores' )
__a : Optional[Any] = torch.zeros(head_importance.numel() , dtype=torch.long , device=args.device )
__a : str = torch.arange(
head_importance.numel() , device=args.device )
__a : Tuple = head_ranks.view_as(lowerCamelCase_ )
print_ad_tensor(lowerCamelCase_ )
return attn_entropy, head_importance, total_loss
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Tuple , lowerCamelCase_ : int ):
__a , __a , __a : Optional[int] = compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ )
__a : Tuple = 1 / loss # instead of downsteam score use the LM loss
logger.info('Pruning: original score: %f, threshold: %f' , lowerCamelCase_ , original_score * args.masking_threshold )
__a : Tuple = torch.ones_like(lowerCamelCase_ )
__a : int = max(1 , int(new_head_mask.numel() * args.masking_amount ) )
__a : Tuple = original_score
while current_score >= original_score * args.masking_threshold:
__a : Optional[Any] = new_head_mask.clone().detach() # save current head mask
# heads from least important to most - keep only not-masked heads
__a : List[str] = float('Inf' )
__a : List[Any] = head_importance.view(-1 ).sort()[1]
if len(lowerCamelCase_ ) <= num_to_mask:
print('BREAK BY num_to_mask' )
break
# mask heads
__a : Any = current_heads_to_mask[:num_to_mask]
logger.info('Heads to mask: %s' , str(current_heads_to_mask.tolist() ) )
__a : int = new_head_mask.view(-1 )
__a : Tuple = 0.0
__a : int = new_head_mask.view_as(lowerCamelCase_ )
__a : Optional[int] = new_head_mask.clone().detach()
print_ad_tensor(lowerCamelCase_ )
# Compute metric and head importance again
__a , __a , __a : int = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[Any] = 1 / loss
logger.info(
'Masking: current score: %f, remaining heads %d (%.1f percents)' , lowerCamelCase_ , new_head_mask.sum() , new_head_mask.sum() / new_head_mask.numel() * 1_0_0 , )
logger.info('Final head mask' )
print_ad_tensor(lowerCamelCase_ )
np.save(os.path.join(args.output_dir , 'head_mask.npy' ) , head_mask.detach().cpu().numpy() )
return head_mask
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Union[str, Any] ):
__a : List[Any] = datetime.now()
__a , __a , __a : List[str] = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[str] = 1 / loss
__a : List[Any] = datetime.now() - before_time
__a : List[str] = sum(p.numel() for p in model.parameters() )
__a : Dict = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(lowerCamelCase_ ) )
}
for k, v in heads_to_prune.items():
if isinstance(lowerCamelCase_ , lowerCamelCase_ ):
__a : Tuple = [
v,
]
assert sum(len(lowerCamelCase_ ) for h in heads_to_prune.values() ) == (1 - head_mask.long()).sum().item()
model.prune_heads(lowerCamelCase_ )
__a : Optional[Any] = sum(p.numel() for p in model.parameters() )
__a : Tuple = datetime.now()
__a , __a , __a : Tuple = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ , actually_pruned=lowerCamelCase_ , )
__a : Optional[Any] = 1 / loss
__a : List[Any] = datetime.now() - before_time
logger.info(
'Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)' , lowerCamelCase_ , lowerCamelCase_ , pruned_num_params / original_num_params * 1_0_0 , )
logger.info('Pruning: score with masking: %f score with pruning: %f' , lowerCamelCase_ , lowerCamelCase_ )
logger.info('Pruning: speed ratio (original timing / new timing): %f percents' , original_time / new_time * 1_0_0 )
save_model(lowerCamelCase_ , args.output_dir )
def UpperCAmelCase__ ( ):
__a : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'--data_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The input data dir. Should contain the .tsv files (or other data files) for the task.' , )
parser.add_argument(
'--model_name_or_path' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='Path to pretrained model or model identifier from huggingface.co/models' , )
parser.add_argument(
'--output_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
# Other parameters
parser.add_argument(
'--config_name' , default='' , type=lowerCamelCase_ , help='Pretrained config name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--tokenizer_name' , default='' , type=lowerCamelCase_ , help='Pretrained tokenizer name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--cache_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , help='Where do you want to store the pre-trained models downloaded from s3' , )
parser.add_argument(
'--data_subset' , type=lowerCamelCase_ , default=-1 , help='If > 0: limit the data to a subset of data_subset instances.' )
parser.add_argument(
'--overwrite_output_dir' , action='store_true' , help='Whether to overwrite data in output directory' )
parser.add_argument(
'--overwrite_cache' , action='store_true' , help='Overwrite the cached training and evaluation sets' )
parser.add_argument(
'--dont_normalize_importance_by_layer' , action='store_true' , help='Don\'t normalize importance score by layers' )
parser.add_argument(
'--dont_normalize_global_importance' , action='store_true' , help='Don\'t normalize all importance scores between 0 and 1' , )
parser.add_argument(
'--try_masking' , action='store_true' , help='Whether to try to mask head until a threshold of accuracy.' )
parser.add_argument(
'--masking_threshold' , default=0.9 , type=lowerCamelCase_ , help='masking threshold in term of metrics (stop masking when metric < threshold * original metric value).' , )
parser.add_argument(
'--masking_amount' , default=0.1 , type=lowerCamelCase_ , help='Amount to heads to masking at each masking step.' )
parser.add_argument('--metric_name' , default='acc' , type=lowerCamelCase_ , help='Metric to use for head masking.' )
parser.add_argument(
'--max_seq_length' , default=1_2_8 , type=lowerCamelCase_ , help=(
'The maximum total input sequence length after WordPiece tokenization. \n'
'Sequences longer than this will be truncated, sequences shorter padded.'
) , )
parser.add_argument('--batch_size' , default=1 , type=lowerCamelCase_ , help='Batch size.' )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 )
parser.add_argument('--local_rank' , type=lowerCamelCase_ , default=-1 , help='local_rank for distributed training on gpus' )
parser.add_argument('--no_cuda' , action='store_true' , help='Whether not to use CUDA when available' )
parser.add_argument('--server_ip' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
parser.add_argument('--server_port' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
__a : Optional[Any] = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print('Waiting for debugger attach' )
ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=lowerCamelCase_ )
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
__a : List[str] = torch.device('cuda' if torch.cuda.is_available() and not args.no_cuda else 'cpu' )
__a : Tuple = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank )
__a : Union[str, Any] = torch.device('cuda' , args.local_rank )
__a : Any = 1
torch.distributed.init_process_group(backend='nccl' ) # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN )
logger.info('device: {} n_gpu: {}, distributed: {}'.format(args.device , args.n_gpu , bool(args.local_rank != -1 ) ) )
__a : Optional[Any] = GPTaLMHeadModel.from_pretrained(args.model_name_or_path )
# Distributed and parallel training
model.to(args.device )
if args.local_rank != -1:
__a : List[Any] = nn.parallel.DistributedDataParallel(
lowerCamelCase_ , device_ids=[args.local_rank] , output_device=args.local_rank , find_unused_parameters=lowerCamelCase_ )
elif args.n_gpu > 1:
__a : Union[str, Any] = nn.DataParallel(lowerCamelCase_ )
# Print/save training arguments
os.makedirs(args.output_dir , exist_ok=lowerCamelCase_ )
torch.save(lowerCamelCase_ , os.path.join(args.output_dir , 'run_args.bin' ) )
logger.info('Training/evaluation parameters %s' , lowerCamelCase_ )
# Prepare dataset
__a : Tuple = np.concatenate(
[
np.loadtxt(args.data_dir , dtype=np.intaa ),
] )
__a : str = (torch.from_numpy(lowerCamelCase_ ),)
__a : List[str] = TensorDataset(*lowerCamelCase_ )
__a : Optional[Any] = RandomSampler(lowerCamelCase_ )
__a : Union[str, Any] = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.batch_size )
# Compute head entropy and importance score
compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
__a : Union[str, Any] = mask_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
prune_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 | 1 |
import os
import tempfile
import unittest
from transformers import is_torch_available
from transformers.testing_utils import require_torch
if is_torch_available():
import torch
from torch import nn
from transformers import (
Adafactor,
AdamW,
get_constant_schedule,
get_constant_schedule_with_warmup,
get_cosine_schedule_with_warmup,
get_cosine_with_hard_restarts_schedule_with_warmup,
get_inverse_sqrt_schedule,
get_linear_schedule_with_warmup,
get_polynomial_decay_schedule_with_warmup,
)
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : List[str]=1_0 ):
__a : Optional[Any] = []
for _ in range(lowerCamelCase_ ):
lrs.append(scheduler.get_lr()[0] )
scheduler.step()
return lrs
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : Tuple=1_0 ):
__a : Union[str, Any] = []
for step in range(lowerCamelCase_ ):
lrs.append(scheduler.get_lr()[0] )
scheduler.step()
if step == num_steps // 2:
with tempfile.TemporaryDirectory() as tmpdirname:
__a : List[Any] = os.path.join(lowerCamelCase_ , 'schedule.bin' )
torch.save(scheduler.state_dict() , lowerCamelCase_ )
__a : Tuple = torch.load(lowerCamelCase_ )
scheduler.load_state_dict(lowerCamelCase_ )
return lrs
@require_torch
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : Union[str, Any] = torch.tensor([0.1, -0.2, -0.1] , requires_grad=SCREAMING_SNAKE_CASE__ )
__a : List[str] = torch.tensor([0.4, 0.2, -0.5] )
__a : List[Any] = nn.MSELoss()
# No warmup, constant schedule, no gradient clipping
__a : List[Any] = AdamW(params=[w] , lr=2e-1 , weight_decay=0.0 )
for _ in range(1_0_0 ):
__a : int = criterion(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
loss.backward()
optimizer.step()
w.grad.detach_() # No zero_grad() function on simple tensors. we do it ourselves.
w.grad.zero_()
self.assertListAlmostEqual(w.tolist() , [0.4, 0.2, -0.5] , tol=1e-2 )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : List[str] = torch.tensor([0.1, -0.2, -0.1] , requires_grad=SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = torch.tensor([0.4, 0.2, -0.5] )
__a : Union[str, Any] = nn.MSELoss()
# No warmup, constant schedule, no gradient clipping
__a : int = Adafactor(
params=[w] , lr=1e-2 , eps=(1e-30, 1e-3) , clip_threshold=1.0 , decay_rate=-0.8 , betaa=SCREAMING_SNAKE_CASE__ , weight_decay=0.0 , relative_step=SCREAMING_SNAKE_CASE__ , scale_parameter=SCREAMING_SNAKE_CASE__ , warmup_init=SCREAMING_SNAKE_CASE__ , )
for _ in range(1_0_0_0 ):
__a : str = criterion(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
loss.backward()
optimizer.step()
w.grad.detach_() # No zero_grad() function on simple tensors. we do it ourselves.
w.grad.zero_()
self.assertListAlmostEqual(w.tolist() , [0.4, 0.2, -0.5] , tol=1e-2 )
@require_torch
class _UpperCamelCase( unittest.TestCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = nn.Linear(50 , 50 ) if is_torch_available() else None
__SCREAMING_SNAKE_CASE : str = AdamW(m.parameters() , lr=10.0 ) if is_torch_available() else None
__SCREAMING_SNAKE_CASE : str = 10
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Dict=None ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ , msg=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : str = {'num_warmup_steps': 2, 'num_training_steps': 1_0}
# schedulers doct format
# function: (sched_args_dict, expected_learning_rates)
__a : Any = {
get_constant_schedule: ({}, [10.0] * self.num_steps),
get_constant_schedule_with_warmup: (
{'num_warmup_steps': 4},
[0.0, 2.5, 5.0, 7.5, 10.0, 10.0, 10.0, 10.0, 10.0, 10.0],
),
get_linear_schedule_with_warmup: (
{**common_kwargs},
[0.0, 5.0, 10.0, 8.75, 7.5, 6.25, 5.0, 3.75, 2.5, 1.25],
),
get_cosine_schedule_with_warmup: (
{**common_kwargs},
[0.0, 5.0, 10.0, 9.61, 8.53, 6.91, 5.0, 3.08, 1.46, 0.38],
),
get_cosine_with_hard_restarts_schedule_with_warmup: (
{**common_kwargs, 'num_cycles': 2},
[0.0, 5.0, 10.0, 8.53, 5.0, 1.46, 10.0, 8.53, 5.0, 1.46],
),
get_polynomial_decay_schedule_with_warmup: (
{**common_kwargs, 'power': 2.0, 'lr_end': 1e-7},
[0.0, 5.0, 10.0, 7.656, 5.625, 3.906, 2.5, 1.406, 0.625, 0.156],
),
get_inverse_sqrt_schedule: (
{'num_warmup_steps': 2},
[0.0, 5.0, 10.0, 8.165, 7.071, 6.325, 5.774, 5.345, 5.0, 4.714],
),
}
for scheduler_func, data in scheds.items():
__a , __a : str = data
__a : Tuple = scheduler_func(self.optimizer , **SCREAMING_SNAKE_CASE__ )
self.assertEqual(len([scheduler.get_lr()[0]] ) , 1 )
__a : List[Any] = unwrap_schedule(SCREAMING_SNAKE_CASE__ , self.num_steps )
self.assertListAlmostEqual(
SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , tol=1e-2 , msg=f'''failed for {scheduler_func} in normal scheduler''' , )
__a : Dict = scheduler_func(self.optimizer , **SCREAMING_SNAKE_CASE__ )
if scheduler_func.__name__ != "get_constant_schedule":
LambdaScheduleWrapper.wrap_scheduler(SCREAMING_SNAKE_CASE__ ) # wrap to test picklability of the schedule
__a : Optional[int] = unwrap_and_save_reload_schedule(SCREAMING_SNAKE_CASE__ , self.num_steps )
self.assertListEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , msg=f'''failed for {scheduler_func} in save and reload''' )
class _UpperCamelCase:
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : str = fn
def __call__( self : int , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
return self.fn(*SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@classmethod
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : List[Any] = list(map(self , scheduler.lr_lambdas ) )
| 47 |
import argparse
import os
import gluonnlp as nlp
import mxnet as mx
import numpy as np
import torch
from gluonnlp.base import get_home_dir
from gluonnlp.model.bert import BERTEncoder
from gluonnlp.model.utils import _load_vocab
from gluonnlp.vocab import Vocab
from packaging import version
from torch import nn
from transformers import BertConfig, BertForMaskedLM, BertModel, RobertaTokenizer
from transformers.models.bert.modeling_bert import (
BertIntermediate,
BertLayer,
BertOutput,
BertSelfAttention,
BertSelfOutput,
)
from transformers.utils import logging
if version.parse(nlp.__version__) != version.parse('''0.8.3'''):
raise Exception('''requires gluonnlp == 0.8.3''')
if version.parse(mx.__version__) != version.parse('''1.5.0'''):
raise Exception('''requires mxnet == 1.5.0''')
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''The Nymphenburg Palace is a beautiful palace in Munich!'''
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
__a : List[Any] = {
'attention_cell': 'multi_head',
'num_layers': 4,
'units': 1_0_2_4,
'hidden_size': 7_6_8,
'max_length': 5_1_2,
'num_heads': 8,
'scaled': True,
'dropout': 0.1,
'use_residual': True,
'embed_size': 1_0_2_4,
'embed_dropout': 0.1,
'word_embed': None,
'layer_norm_eps': 1e-5,
'token_type_vocab_size': 2,
}
__a : Optional[int] = bort_4_8_768_1024_hparams
# Let's construct the original Bort model here
# Taken from official BERT implementation, see:
# https://github.com/alexa/bort/blob/master/bort/bort.py
__a : List[str] = BERTEncoder(
attention_cell=predefined_args['attention_cell'] , num_layers=predefined_args['num_layers'] , units=predefined_args['units'] , hidden_size=predefined_args['hidden_size'] , max_length=predefined_args['max_length'] , num_heads=predefined_args['num_heads'] , scaled=predefined_args['scaled'] , dropout=predefined_args['dropout'] , output_attention=lowerCamelCase_ , output_all_encodings=lowerCamelCase_ , use_residual=predefined_args['use_residual'] , activation=predefined_args.get('activation' , 'gelu' ) , layer_norm_eps=predefined_args.get('layer_norm_eps' , lowerCamelCase_ ) , )
# Vocab information needs to be fetched first
# It's the same as RoBERTa, so RobertaTokenizer can be used later
__a : int = 'openwebtext_ccnews_stories_books_cased'
# Specify download folder to Gluonnlp's vocab
__a : Optional[Any] = os.path.join(get_home_dir() , 'models' )
__a : Optional[Any] = _load_vocab(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , cls=lowerCamelCase_ )
__a : Any = nlp.model.BERTModel(
lowerCamelCase_ , len(lowerCamelCase_ ) , units=predefined_args['units'] , embed_size=predefined_args['embed_size'] , embed_dropout=predefined_args['embed_dropout'] , word_embed=predefined_args['word_embed'] , use_pooler=lowerCamelCase_ , use_token_type_embed=lowerCamelCase_ , token_type_vocab_size=predefined_args['token_type_vocab_size'] , use_classifier=lowerCamelCase_ , use_decoder=lowerCamelCase_ , )
original_bort.load_parameters(lowerCamelCase_ , cast_dtype=lowerCamelCase_ , ignore_extra=lowerCamelCase_ )
__a : Dict = original_bort._collect_params_with_prefix()
# Build our config 🤗
__a : Optional[Any] = {
'architectures': ['BertForMaskedLM'],
'attention_probs_dropout_prob': predefined_args['dropout'],
'hidden_act': 'gelu',
'hidden_dropout_prob': predefined_args['dropout'],
'hidden_size': predefined_args['embed_size'],
'initializer_range': 0.02,
'intermediate_size': predefined_args['hidden_size'],
'layer_norm_eps': predefined_args['layer_norm_eps'],
'max_position_embeddings': predefined_args['max_length'],
'model_type': 'bort',
'num_attention_heads': predefined_args['num_heads'],
'num_hidden_layers': predefined_args['num_layers'],
'pad_token_id': 1, # 2 = BERT, 1 = RoBERTa
'type_vocab_size': 1, # 2 = BERT, 1 = RoBERTa
'vocab_size': len(lowerCamelCase_ ),
}
__a : str = BertConfig.from_dict(lowerCamelCase_ )
__a : Optional[int] = BertForMaskedLM(lowerCamelCase_ )
hf_bort_model.eval()
# Parameter mapping table (Gluonnlp to Transformers)
# * denotes layer index
#
# | Gluon Parameter | Transformers Parameter
# | -------------------------------------------------------------- | ----------------------
# | `encoder.layer_norm.beta` | `bert.embeddings.LayerNorm.bias`
# | `encoder.layer_norm.gamma` | `bert.embeddings.LayerNorm.weight`
# | `encoder.position_weight` | `bert.embeddings.position_embeddings.weight`
# | `word_embed.0.weight` | `bert.embeddings.word_embeddings.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_key.bias` | `bert.encoder.layer.*.attention.self.key.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_key.weight` | `bert.encoder.layer.*.attention.self.key.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_query.bias` | `bert.encoder.layer.*.attention.self.query.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_query.weight` | `bert.encoder.layer.*.attention.self.query.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_value.bias` | `bert.encoder.layer.*.attention.self.value.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_value.weight` | `bert.encoder.layer.*.attention.self.value.weight`
# | `encoder.transformer_cells.*.ffn.ffn_2.bias` | `bert.encoder.layer.*.attention.output.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_2.weight` | `bert.encoder.layer.*.attention.output.dense.weight`
# | `encoder.transformer_cells.*.layer_norm.beta` | `bert.encoder.layer.*.attention.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.layer_norm.gamma` | `bert.encoder.layer.*.attention.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.ffn.ffn_1.bias` | `bert.encoder.layer.*.intermediate.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_1.weight` | `bert.encoder.layer.*.intermediate.dense.weight`
# | `encoder.transformer_cells.*.ffn.layer_norm.beta` | `bert.encoder.layer.*.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.ffn.layer_norm.gamma` | `bert.encoder.layer.*.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.proj.bias` | `bert.encoder.layer.*.output.dense.bias`
# | `encoder.transformer_cells.*.proj.weight` | `bert.encoder.layer.*.output.dense.weight`
# Helper function to convert MXNET Arrays to PyTorch
def to_torch(lowerCamelCase_ : Optional[Any] ) -> nn.Parameter:
return nn.Parameter(torch.FloatTensor(mx_array.data().asnumpy() ) )
# Check param shapes and map new HF param back
def check_and_map_params(lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[str] ):
__a : Optional[int] = hf_param.shape
__a : int = to_torch(params[gluon_param] )
__a : int = gluon_param.shape
assert (
shape_hf == shape_gluon
), f'''The gluon parameter {gluon_param} has shape {shape_gluon}, but expects shape {shape_hf} for Transformers'''
return gluon_param
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.word_embeddings.weight , 'word_embed.0.weight' )
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.position_embeddings.weight , 'encoder.position_weight' )
__a : Tuple = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.bias , 'encoder.layer_norm.beta' )
__a : Union[str, Any] = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.weight , 'encoder.layer_norm.gamma' )
# Inspired by RoBERTa conversion script, we just zero them out (Bort does not use them)
__a : Union[str, Any] = torch.zeros_like(
hf_bort_model.bert.embeddings.token_type_embeddings.weight.data )
for i in range(hf_bort_config.num_hidden_layers ):
__a : BertLayer = hf_bort_model.bert.encoder.layer[i]
# self attention
__a : BertSelfAttention = layer.attention.self
__a : Optional[int] = check_and_map_params(
self_attn.key.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.bias''' )
__a : str = check_and_map_params(
self_attn.key.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.weight''' )
__a : List[str] = check_and_map_params(
self_attn.query.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.bias''' )
__a : str = check_and_map_params(
self_attn.query.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.weight''' )
__a : Dict = check_and_map_params(
self_attn.value.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.bias''' )
__a : str = check_and_map_params(
self_attn.value.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.weight''' )
# self attention output
__a : BertSelfOutput = layer.attention.output
__a : Tuple = check_and_map_params(
self_output.dense.bias , f'''encoder.transformer_cells.{i}.proj.bias''' )
__a : Dict = check_and_map_params(
self_output.dense.weight , f'''encoder.transformer_cells.{i}.proj.weight''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.layer_norm.beta''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.layer_norm.gamma''' )
# intermediate
__a : BertIntermediate = layer.intermediate
__a : List[str] = check_and_map_params(
intermediate.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_1.bias''' )
__a : Optional[Any] = check_and_map_params(
intermediate.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_1.weight''' )
# output
__a : BertOutput = layer.output
__a : str = check_and_map_params(
bert_output.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_2.bias''' )
__a : List[Any] = check_and_map_params(
bert_output.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_2.weight''' )
__a : str = check_and_map_params(
bert_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.ffn.layer_norm.beta''' )
__a : List[str] = check_and_map_params(
bert_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.ffn.layer_norm.gamma''' )
# Save space and energy 🎄
hf_bort_model.half()
# Compare output of both models
__a : Union[str, Any] = RobertaTokenizer.from_pretrained('roberta-base' )
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ )['input_ids']
# Get gluon output
__a : Optional[int] = mx.nd.array([input_ids] )
__a : Tuple = original_bort(inputs=lowerCamelCase_ , token_types=[] )
# Get Transformer output (save and reload model again)
hf_bort_model.save_pretrained(lowerCamelCase_ )
__a : Optional[Any] = BertModel.from_pretrained(lowerCamelCase_ )
hf_bort_model.eval()
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ , return_tensors='pt' )
__a : int = hf_bort_model(**lowerCamelCase_ )[0]
__a : Dict = output_gluon[0].asnumpy()
__a : str = output_hf[0].detach().numpy()
__a : List[Any] = np.max(np.abs(hf_layer - gluon_layer ) ).item()
__a : str = np.allclose(lowerCamelCase_ , lowerCamelCase_ , atol=1e-3 )
if success:
print('✔️ Both model do output the same tensors' )
else:
print('❌ Both model do **NOT** output the same tensors' )
print('Absolute difference is:' , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--bort_checkpoint_path''', default=None, type=str, required=True, help='''Path the official Bort params file.'''
)
parser.add_argument(
'''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.'''
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_bort_checkpoint_to_pytorch(args.bort_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_tokenizers_available, is_torch_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_nezha''': ['''NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''NezhaConfig'''],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''NezhaForNextSentencePrediction''',
'''NezhaForMaskedLM''',
'''NezhaForPreTraining''',
'''NezhaForMultipleChoice''',
'''NezhaForQuestionAnswering''',
'''NezhaForSequenceClassification''',
'''NezhaForTokenClassification''',
'''NezhaModel''',
'''NezhaPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_nezha import NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP, NezhaConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_nezha import (
NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST,
NezhaForMaskedLM,
NezhaForMultipleChoice,
NezhaForNextSentencePrediction,
NezhaForPreTraining,
NezhaForQuestionAnswering,
NezhaForSequenceClassification,
NezhaForTokenClassification,
NezhaModel,
NezhaPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[str] ):
__a : Any = ''
for i in table:
res += inp[i - 1]
return res
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
return data[1:] + data[0]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] ):
__a : Optional[int] = ''
for i in range(len(lowerCamelCase_ ) ):
if a[i] == b[i]:
res += "0"
else:
res += "1"
return res
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : str ):
__a : List[str] = int('0b' + data[0] + data[-1] , 2 )
__a : List[str] = int('0b' + data[1:3] , 2 )
return bin(s[row][col] )[2:]
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : List[str] , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] ):
__a : List[Any] = message[:4]
__a : str = message[4:]
__a : Any = apply_table(lowerCamelCase_ , lowerCamelCase_ )
__a : int = xor(lowerCamelCase_ , lowerCamelCase_ )
__a : Dict = apply_sbox(lowerCamelCase_ , temp[:4] ) # noqa: E741
__a : Tuple = apply_sbox(lowerCamelCase_ , temp[4:] )
__a : List[Any] = '0' * (2 - len(lowerCamelCase_ )) + l # noqa: E741
__a : List[str] = '0' * (2 - len(lowerCamelCase_ )) + r
__a : List[Any] = apply_table(l + r , lowerCamelCase_ )
__a : Dict = xor(lowerCamelCase_ , lowerCamelCase_ )
return temp + right
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter 10 bit key: ''')
SCREAMING_SNAKE_CASE__ = input('''Enter 8 bit message: ''')
SCREAMING_SNAKE_CASE__ = [6, 3, 7, 4, 8, 5, 10, 9]
SCREAMING_SNAKE_CASE__ = [3, 5, 2, 7, 4, 10, 1, 9, 8, 6]
SCREAMING_SNAKE_CASE__ = [2, 4, 3, 1]
SCREAMING_SNAKE_CASE__ = [2, 6, 3, 1, 4, 8, 5, 7]
SCREAMING_SNAKE_CASE__ = [4, 1, 3, 5, 7, 2, 8, 6]
SCREAMING_SNAKE_CASE__ = [4, 1, 2, 3, 2, 3, 4, 1]
SCREAMING_SNAKE_CASE__ = [[1, 0, 3, 2], [3, 2, 1, 0], [0, 2, 1, 3], [3, 1, 3, 2]]
SCREAMING_SNAKE_CASE__ = [[0, 1, 2, 3], [2, 0, 1, 3], [3, 0, 1, 0], [2, 1, 0, 3]]
# key generation
SCREAMING_SNAKE_CASE__ = apply_table(key, paa_table)
SCREAMING_SNAKE_CASE__ = temp[:5]
SCREAMING_SNAKE_CASE__ = temp[5:]
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
# encryption
SCREAMING_SNAKE_CASE__ = apply_table(message, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Cipher text is:''', CT)
# decryption
SCREAMING_SNAKE_CASE__ = apply_table(CT, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Plain text after decypting is:''', PT)
| 47 | 1 |
from __future__ import annotations
import unittest
from transformers import LEDConfig, is_tf_available
from transformers.testing_utils import require_tf, slow
from ...test_configuration_common import ConfigTester
from ...test_modeling_tf_common import TFModelTesterMixin, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_tf_available():
import tensorflow as tf
from transformers import TFLEDForConditionalGeneration, TFLEDModel
@require_tf
class _UpperCamelCase:
__SCREAMING_SNAKE_CASE : str = LEDConfig
__SCREAMING_SNAKE_CASE : Optional[Any] = {}
__SCREAMING_SNAKE_CASE : Optional[int] = '''gelu'''
def __init__( self : str , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Tuple=1_3 , SCREAMING_SNAKE_CASE__ : Tuple=7 , SCREAMING_SNAKE_CASE__ : Optional[Any]=True , SCREAMING_SNAKE_CASE__ : List[str]=False , SCREAMING_SNAKE_CASE__ : Optional[int]=9_9 , SCREAMING_SNAKE_CASE__ : Any=3_2 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : List[Any]=4 , SCREAMING_SNAKE_CASE__ : Tuple=3_7 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : Tuple=0.1 , SCREAMING_SNAKE_CASE__ : List[Any]=2_0 , SCREAMING_SNAKE_CASE__ : Tuple=2 , SCREAMING_SNAKE_CASE__ : Any=1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=0 , SCREAMING_SNAKE_CASE__ : Optional[Any]=4 , ):
'''simple docstring'''
__a : Optional[Any] = parent
__a : Tuple = batch_size
__a : Optional[Any] = seq_length
__a : List[Any] = is_training
__a : List[Any] = use_labels
__a : Dict = vocab_size
__a : Tuple = hidden_size
__a : Union[str, Any] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : Any = intermediate_size
__a : Union[str, Any] = hidden_dropout_prob
__a : Optional[Any] = attention_probs_dropout_prob
__a : List[str] = max_position_embeddings
__a : int = eos_token_id
__a : List[Any] = pad_token_id
__a : Any = bos_token_id
__a : Dict = attention_window
# `ModelTesterMixin.test_attention_outputs` is expecting attention tensors to be of size
# [num_attention_heads, encoder_seq_length, encoder_key_length], but TFLongformerSelfAttention
# returns attention of shape [num_attention_heads, encoder_seq_length, self.attention_window + 1]
# because its local attention only attends to `self.attention_window` and one before and one after
__a : Optional[Any] = self.attention_window + 2
# because of padding `encoder_seq_length`, is different from `seq_length`. Relevant for
# the `test_attention_outputs` and `test_hidden_states_output` tests
__a : List[str] = (
self.seq_length + (self.attention_window - self.seq_length % self.attention_window) % self.attention_window
)
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Dict = ids_tensor([self.batch_size, self.seq_length - 1] , self.vocab_size )
__a : str = tf.expand_dims(tf.constant([self.eos_token_id] * self.batch_size ) , 1 )
__a : Optional[Any] = tf.concat([input_ids, eos_tensor] , axis=1 )
__a : Union[str, Any] = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a : Optional[Any] = self.config_cls(
vocab_size=self.vocab_size , d_model=self.hidden_size , encoder_layers=self.num_hidden_layers , decoder_layers=self.num_hidden_layers , encoder_attention_heads=self.num_attention_heads , decoder_attention_heads=self.num_attention_heads , encoder_ffn_dim=self.intermediate_size , decoder_ffn_dim=self.intermediate_size , dropout=self.hidden_dropout_prob , attention_dropout=self.attention_probs_dropout_prob , max_position_embeddings=self.max_position_embeddings , eos_token_ids=[2] , bos_token_id=self.bos_token_id , pad_token_id=self.pad_token_id , decoder_start_token_id=self.pad_token_id , attention_window=self.attention_window , **self.config_updates , )
__a : int = prepare_led_inputs_dict(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = tf.concat(
[tf.zeros_like(SCREAMING_SNAKE_CASE__ )[:, :-1], tf.ones_like(SCREAMING_SNAKE_CASE__ )[:, -1:]] , axis=-1 , )
__a : str = global_attention_mask
return config, inputs_dict
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
__a : Any = TFLEDModel(config=SCREAMING_SNAKE_CASE__ ).get_decoder()
__a : Tuple = inputs_dict['input_ids']
__a : str = input_ids[:1, :]
__a : Optional[int] = inputs_dict['attention_mask'][:1, :]
__a : Optional[Any] = 1
# first forward pass
__a : Dict = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , use_cache=SCREAMING_SNAKE_CASE__ )
__a , __a : str = outputs.to_tuple()
# create hypothetical next token and extent to next_input_ids
__a : List[Any] = ids_tensor((self.batch_size, 3) , config.vocab_size )
__a : Dict = tf.cast(ids_tensor((self.batch_size, 3) , 2 ) , tf.inta )
# append to next input_ids and
__a : List[Any] = tf.concat([input_ids, next_tokens] , axis=-1 )
__a : Optional[Any] = tf.concat([attention_mask, next_attn_mask] , axis=-1 )
__a : str = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ )[0]
__a : Optional[Any] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , past_key_values=SCREAMING_SNAKE_CASE__ )[0]
self.parent.assertEqual(next_tokens.shape[1] , output_from_past.shape[1] )
# select random slice
__a : Optional[Any] = int(ids_tensor((1,) , output_from_past.shape[-1] ) )
__a : List[str] = output_from_no_past[:, -3:, random_slice_idx]
__a : Dict = output_from_past[:, :, random_slice_idx]
# test that outputs are equal for slice
tf.debugging.assert_near(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , rtol=1e-3 )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : str , lowerCamelCase_ : Dict , lowerCamelCase_ : str=None , lowerCamelCase_ : Tuple=None , lowerCamelCase_ : str=None , lowerCamelCase_ : Optional[int]=None , ):
if attention_mask is None:
__a : Tuple = tf.cast(tf.math.not_equal(lowerCamelCase_ , config.pad_token_id ) , tf.inta )
if decoder_attention_mask is None:
__a : Union[str, Any] = tf.concat(
[
tf.ones(decoder_input_ids[:, :1].shape , dtype=tf.inta ),
tf.cast(tf.math.not_equal(decoder_input_ids[:, 1:] , config.pad_token_id ) , tf.inta ),
] , axis=-1 , )
if head_mask is None:
__a : str = tf.ones((config.encoder_layers, config.encoder_attention_heads) )
if decoder_head_mask is None:
__a : Optional[Any] = tf.ones((config.decoder_layers, config.decoder_attention_heads) )
return {
"input_ids": input_ids,
"attention_mask": attention_mask,
"decoder_input_ids": decoder_input_ids,
"decoder_attention_mask": decoder_attention_mask,
"head_mask": head_mask,
"decoder_head_mask": decoder_head_mask,
}
@require_tf
class _UpperCamelCase( __lowerCamelCase , __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : Optional[int] = (TFLEDForConditionalGeneration, TFLEDModel) if is_tf_available() else ()
__SCREAMING_SNAKE_CASE : Optional[Any] = (TFLEDForConditionalGeneration,) if is_tf_available() else ()
__SCREAMING_SNAKE_CASE : Optional[Any] = (
{
'''conversational''': TFLEDForConditionalGeneration,
'''feature-extraction''': TFLEDModel,
'''summarization''': TFLEDForConditionalGeneration,
'''text2text-generation''': TFLEDForConditionalGeneration,
'''translation''': TFLEDForConditionalGeneration,
}
if is_tf_available()
else {}
)
__SCREAMING_SNAKE_CASE : Optional[int] = True
__SCREAMING_SNAKE_CASE : List[Any] = False
__SCREAMING_SNAKE_CASE : Dict = False
__SCREAMING_SNAKE_CASE : Tuple = False
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Tuple = TFLEDModelTester(self )
__a : Any = ConfigTester(self , config_class=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
self.config_tester.run_common_tests()
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Dict = self.model_tester.prepare_config_and_inputs_for_common()
self.model_tester.check_decoder_model_past_large_inputs(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a , __a : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
__a : str = tf.zeros_like(inputs_dict['attention_mask'] )
__a : Dict = 2
__a : List[Any] = tf.where(
tf.range(self.model_tester.seq_length )[None, :] < num_global_attn_indices , 1 , inputs_dict['global_attention_mask'] , )
__a : Tuple = True
__a : Optional[Any] = self.model_tester.seq_length
__a : Optional[int] = self.model_tester.encoder_seq_length
def check_decoder_attentions_output(SCREAMING_SNAKE_CASE__ : Any ):
__a : List[Any] = outputs.decoder_attentions
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , self.model_tester.num_hidden_layers )
self.assertListEqual(
list(decoder_attentions[0].shape[-3:] ) , [self.model_tester.num_attention_heads, seq_length, seq_length] , )
def check_encoder_attentions_output(SCREAMING_SNAKE_CASE__ : Tuple ):
__a : Optional[int] = [t.numpy() for t in outputs.encoder_attentions]
__a : List[str] = [t.numpy() for t in outputs.encoder_global_attentions]
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , self.model_tester.num_hidden_layers )
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , self.model_tester.num_hidden_layers )
self.assertListEqual(
list(attentions[0].shape[-3:] ) , [self.model_tester.num_attention_heads, seq_length, seq_length] , )
self.assertListEqual(
list(global_attentions[0].shape[-3:] ) , [self.model_tester.num_attention_heads, encoder_seq_length, num_global_attn_indices] , )
for model_class in self.all_model_classes:
__a : Union[str, Any] = True
__a : Union[str, Any] = False
__a : Optional[int] = False
__a : Dict = model_class(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = model(self._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a : Optional[int] = len(SCREAMING_SNAKE_CASE__ )
self.assertEqual(config.output_hidden_states , SCREAMING_SNAKE_CASE__ )
check_encoder_attentions_output(SCREAMING_SNAKE_CASE__ )
if self.is_encoder_decoder:
__a : List[Any] = model_class(SCREAMING_SNAKE_CASE__ )
__a : Dict = model(self._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
self.assertEqual(config.output_hidden_states , SCREAMING_SNAKE_CASE__ )
check_decoder_attentions_output(SCREAMING_SNAKE_CASE__ )
# Check that output attentions can also be changed via the config
del inputs_dict["output_attentions"]
__a : Optional[Any] = True
__a : List[str] = model_class(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = model(self._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
self.assertEqual(config.output_hidden_states , SCREAMING_SNAKE_CASE__ )
check_encoder_attentions_output(SCREAMING_SNAKE_CASE__ )
# Check attention is always last and order is fine
__a : Dict = True
__a : List[str] = True
__a : int = model_class(SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = model(self._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
self.assertEqual(out_len + (2 if self.is_encoder_decoder else 1) , len(SCREAMING_SNAKE_CASE__ ) )
self.assertEqual(model.config.output_hidden_states , SCREAMING_SNAKE_CASE__ )
check_encoder_attentions_output(SCREAMING_SNAKE_CASE__ )
@unittest.skip('LED keeps using potentially symbolic tensors in conditionals and breaks tracing.' )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
pass
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
pass
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] ):
return tf.constant(lowerCamelCase_ , dtype=tf.intaa )
SCREAMING_SNAKE_CASE__ = 1E-4
@slow
@require_tf
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Union[str, Any] = TFLEDForConditionalGeneration.from_pretrained('allenai/led-base-16384' ).led
# change to intended input here
__a : List[Any] = _long_tensor([5_1_2 * [0, 3_1_4_1_4, 2_3_2, 3_2_8, 7_4_0, 1_1_4_0, 1_2_6_9_5, 6_9]] )
__a : Optional[int] = _long_tensor([1_2_8 * [0, 3_1_4_1_4, 2_3_2, 3_2_8, 7_4_0, 1_1_4_0, 1_2_6_9_5, 6_9]] )
__a : Dict = prepare_led_inputs_dict(model.config , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = model(**SCREAMING_SNAKE_CASE__ )[0]
__a : Tuple = (1, 1_0_2_4, 7_6_8)
self.assertEqual(output.shape , SCREAMING_SNAKE_CASE__ )
# change to expected output here
__a : str = tf.convert_to_tensor(
[[2.3_050, 2.8_279, 0.6_531], [-1.8_457, -0.1_455, -3.5_661], [-1.0_186, 0.4_586, -2.2_043]] , )
tf.debugging.assert_near(output[:, :3, :3] , SCREAMING_SNAKE_CASE__ , atol=1e-3 )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : int = TFLEDForConditionalGeneration.from_pretrained('allenai/led-base-16384' )
# change to intended input here
__a : Optional[int] = _long_tensor([5_1_2 * [0, 3_1_4_1_4, 2_3_2, 3_2_8, 7_4_0, 1_1_4_0, 1_2_6_9_5, 6_9]] )
__a : Any = _long_tensor([1_2_8 * [0, 3_1_4_1_4, 2_3_2, 3_2_8, 7_4_0, 1_1_4_0, 1_2_6_9_5, 6_9]] )
__a : Dict = prepare_led_inputs_dict(model.config , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = model(**SCREAMING_SNAKE_CASE__ )[0]
__a : Optional[Any] = (1, 1_0_2_4, model.config.vocab_size)
self.assertEqual(output.shape , SCREAMING_SNAKE_CASE__ )
# change to expected output here
__a : Union[str, Any] = tf.convert_to_tensor(
[[33.6_507, 6.4_572, 16.8_089], [5.8_739, -2.4_238, 11.2_902], [-3.2_139, -4.3_149, 4.2_783]] , )
tf.debugging.assert_near(output[:, :3, :3] , SCREAMING_SNAKE_CASE__ , atol=1e-3 , rtol=1e-3 )
| 47 |
import unittest
from transformers import is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from transformers import GradientAccumulator, create_optimizer
@require_tf
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : List[Any] = GradientAccumulator()
accumulator([tf.constant([1.0, 2.0] )] )
accumulator([tf.constant([-2.0, 1.0] )] )
accumulator([tf.constant([-1.0, 2.0] )] )
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
accumulator([tf.constant([1.0, 1.0] ), tf.constant([2.0, 2.0] )] )
self.assertEqual(accumulator.step , 3 )
self.assertEqual(len(accumulator.gradients ) , 1 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [-2.0, 5.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [0.0, 0.0] , tol=1e-2 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : int = None
ops.enable_eager_execution_internal()
__a : Optional[Any] = tf.config.list_physical_devices('CPU' )
if len(SCREAMING_SNAKE_CASE__ ) == 1:
tf.config.set_logical_device_configuration(
physical_devices[0] , [tf.config.LogicalDeviceConfiguration(), tf.config.LogicalDeviceConfiguration()] )
__a : int = tf.config.list_logical_devices(device_type='CPU' )
__a : str = tf.distribute.MirroredStrategy(devices=devices[:2] )
with strategy.scope():
__a : List[str] = GradientAccumulator()
__a : Tuple = tf.Variable([4.0, 3.0] )
__a , __a : int = create_optimizer(5e-5 , 1_0 , 5 )
__a : List[Any] = tf.Variable([0.0, 0.0] , trainable=SCREAMING_SNAKE_CASE__ )
def accumulate_on_replica(SCREAMING_SNAKE_CASE__ : Optional[Any] ):
accumulator([gradient] )
def apply_on_replica():
optimizer.apply_gradients(list(zip(accumulator.gradients , [variable] ) ) )
@tf.function
def accumulate(SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
with strategy.scope():
__a : Optional[Any] = strategy.experimental_local_results(SCREAMING_SNAKE_CASE__ )
local_variables[0].assign(SCREAMING_SNAKE_CASE__ )
local_variables[1].assign(SCREAMING_SNAKE_CASE__ )
strategy.run(SCREAMING_SNAKE_CASE__ , args=(gradient_placeholder,) )
@tf.function
def apply_grad():
with strategy.scope():
strategy.run(SCREAMING_SNAKE_CASE__ )
def _check_local_values(SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int ):
__a : Union[str, Any] = strategy.experimental_local_results(accumulator._gradients[0] )
self.assertListAlmostEqual(values[0].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
self.assertListAlmostEqual(values[1].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
accumulate([1.0, 2.0] , [-1.0, 1.0] )
accumulate([3.0, -1.0] , [-1.0, -1.0] )
accumulate([-2.0, 2.0] , [3.0, -2.0] )
self.assertEqual(accumulator.step , 3 )
_check_local_values([2.0, 3.0] , [1.0, -2.0] )
apply_grad()
self.assertListAlmostEqual(variable.value() , [4.0, 3.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
_check_local_values([0.0, 0.0] , [0.0, 0.0] )
| 47 | 1 |
from argparse import ArgumentParser
from .add_new_model import AddNewModelCommand
from .add_new_model_like import AddNewModelLikeCommand
from .convert import ConvertCommand
from .download import DownloadCommand
from .env import EnvironmentCommand
from .lfs import LfsCommands
from .pt_to_tf import PTtoTFCommand
from .run import RunCommand
from .serving import ServeCommand
from .user import UserCommands
def UpperCAmelCase__ ( ):
__a : int = ArgumentParser('Transformers CLI tool' , usage='transformers-cli <command> [<args>]' )
__a : int = parser.add_subparsers(help='transformers-cli command helpers' )
# Register commands
ConvertCommand.register_subcommand(lowerCamelCase_ )
DownloadCommand.register_subcommand(lowerCamelCase_ )
EnvironmentCommand.register_subcommand(lowerCamelCase_ )
RunCommand.register_subcommand(lowerCamelCase_ )
ServeCommand.register_subcommand(lowerCamelCase_ )
UserCommands.register_subcommand(lowerCamelCase_ )
AddNewModelCommand.register_subcommand(lowerCamelCase_ )
AddNewModelLikeCommand.register_subcommand(lowerCamelCase_ )
LfsCommands.register_subcommand(lowerCamelCase_ )
PTtoTFCommand.register_subcommand(lowerCamelCase_ )
# Let's go
__a : Union[str, Any] = parser.parse_args()
if not hasattr(lowerCamelCase_ , 'func' ):
parser.print_help()
exit(1 )
# Run
__a : Dict = args.func(lowerCamelCase_ )
service.run()
if __name__ == "__main__":
main()
| 47 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''roberta-base''': '''https://huggingface.co/roberta-base/resolve/main/config.json''',
'''roberta-large''': '''https://huggingface.co/roberta-large/resolve/main/config.json''',
'''roberta-large-mnli''': '''https://huggingface.co/roberta-large-mnli/resolve/main/config.json''',
'''distilroberta-base''': '''https://huggingface.co/distilroberta-base/resolve/main/config.json''',
'''roberta-base-openai-detector''': '''https://huggingface.co/roberta-base-openai-detector/resolve/main/config.json''',
'''roberta-large-openai-detector''': '''https://huggingface.co/roberta-large-openai-detector/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = '''roberta'''
def __init__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=5_0_2_6_5 , SCREAMING_SNAKE_CASE__ : Optional[int]=7_6_8 , SCREAMING_SNAKE_CASE__ : str=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Any="gelu" , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.1 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=2 , SCREAMING_SNAKE_CASE__ : Any=0.02 , SCREAMING_SNAKE_CASE__ : List[str]=1e-12 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0 , SCREAMING_SNAKE_CASE__ : List[str]=2 , SCREAMING_SNAKE_CASE__ : Tuple="absolute" , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[str]=None , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = vocab_size
__a : Tuple = hidden_size
__a : List[str] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : str = hidden_act
__a : Optional[Any] = intermediate_size
__a : Dict = hidden_dropout_prob
__a : List[str] = attention_probs_dropout_prob
__a : Optional[Any] = max_position_embeddings
__a : Dict = type_vocab_size
__a : str = initializer_range
__a : List[str] = layer_norm_eps
__a : Optional[int] = position_embedding_type
__a : Union[str, Any] = use_cache
__a : str = classifier_dropout
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
if self.task == "multiple-choice":
__a : List[str] = {0: 'batch', 1: 'choice', 2: 'sequence'}
else:
__a : Dict = {0: 'batch', 1: 'sequence'}
return OrderedDict(
[
('input_ids', dynamic_axis),
('attention_mask', dynamic_axis),
] )
| 47 | 1 |
import argparse
import os
import transformers
from .convert_slow_tokenizer import SLOW_TO_FAST_CONVERTERS
from .utils import logging
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {name: getattr(transformers, name + '''Fast''') for name in SLOW_TO_FAST_CONVERTERS}
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : Dict , lowerCamelCase_ : str , lowerCamelCase_ : List[Any] ):
if tokenizer_name is not None and tokenizer_name not in TOKENIZER_CLASSES:
raise ValueError(f'''Unrecognized tokenizer name, should be one of {list(TOKENIZER_CLASSES.keys() )}.''' )
if tokenizer_name is None:
__a : Dict = TOKENIZER_CLASSES
else:
__a : Optional[Any] = {tokenizer_name: getattr(lowerCamelCase_ , tokenizer_name + 'Fast' )}
logger.info(f'''Loading tokenizer classes: {tokenizer_names}''' )
for tokenizer_name in tokenizer_names:
__a : Optional[int] = TOKENIZER_CLASSES[tokenizer_name]
__a : List[str] = True
if checkpoint_name is None:
__a : int = list(tokenizer_class.max_model_input_sizes.keys() )
else:
__a : Any = [checkpoint_name]
logger.info(f'''For tokenizer {tokenizer_class.__class__.__name__} loading checkpoints: {checkpoint_names}''' )
for checkpoint in checkpoint_names:
logger.info(f'''Loading {tokenizer_class.__class__.__name__} {checkpoint}''' )
# Load tokenizer
__a : int = tokenizer_class.from_pretrained(lowerCamelCase_ , force_download=lowerCamelCase_ )
# Save fast tokenizer
logger.info(f'''Save fast tokenizer to {dump_path} with prefix {checkpoint} add_prefix {add_prefix}''' )
# For organization names we create sub-directories
if "/" in checkpoint:
__a , __a : Any = checkpoint.split('/' )
__a : List[str] = os.path.join(lowerCamelCase_ , lowerCamelCase_ )
elif add_prefix:
__a : Optional[int] = checkpoint
__a : Optional[int] = dump_path
else:
__a : Tuple = None
__a : Tuple = dump_path
logger.info(f'''=> {dump_path_full} with prefix {checkpoint_prefix_name}, add_prefix {add_prefix}''' )
if checkpoint in list(tokenizer.pretrained_vocab_files_map.values() )[0]:
__a : Optional[Any] = list(tokenizer.pretrained_vocab_files_map.values() )[0][checkpoint]
__a : str = file_path.split(lowerCamelCase_ )[-1][0]
if next_char == "/":
__a : Optional[Any] = os.path.join(lowerCamelCase_ , lowerCamelCase_ )
__a : int = None
logger.info(f'''=> {dump_path_full} with prefix {checkpoint_prefix_name}, add_prefix {add_prefix}''' )
__a : Dict = tokenizer.save_pretrained(
lowerCamelCase_ , legacy_format=lowerCamelCase_ , filename_prefix=lowerCamelCase_ )
logger.info(f'''=> File names {file_names}''' )
for file_name in file_names:
if not file_name.endswith('tokenizer.json' ):
os.remove(lowerCamelCase_ )
logger.info(f'''=> removing {file_name}''' )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--dump_path''', default=None, type=str, required=True, help='''Path to output generated fast tokenizer files.'''
)
parser.add_argument(
'''--tokenizer_name''',
default=None,
type=str,
help=(
F"Optional tokenizer type selected in the list of {list(TOKENIZER_CLASSES.keys())}. If not given, will "
'''download and convert all the checkpoints from AWS.'''
),
)
parser.add_argument(
'''--checkpoint_name''',
default=None,
type=str,
help='''Optional checkpoint name. If not given, will download and convert the canonical checkpoints from AWS.''',
)
parser.add_argument(
'''--force_download''',
action='''store_true''',
help='''Re-download checkpoints.''',
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_slow_checkpoint_to_fast(args.tokenizer_name, args.checkpoint_name, args.dump_path, args.force_download)
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''▁'''
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''sentencepiece.bpe.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''facebook/xglm-564M''': '''https://huggingface.co/facebook/xglm-564M/resolve/main/sentencepiece.bpe.model''',
}
}
SCREAMING_SNAKE_CASE__ = {
'''facebook/xglm-564M''': 2048,
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : Union[str, Any] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : str = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : Any = ['''input_ids''', '''attention_mask''']
def __init__( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : List[str]="</s>" , SCREAMING_SNAKE_CASE__ : Optional[Any]="</s>" , SCREAMING_SNAKE_CASE__ : Union[str, Any]="<s>" , SCREAMING_SNAKE_CASE__ : str="<unk>" , SCREAMING_SNAKE_CASE__ : Dict="<pad>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : List[str] , ):
'''simple docstring'''
__a : Union[str, Any] = {} if sp_model_kwargs is None else sp_model_kwargs
# Compatibility with the original tokenizer
__a : Any = 7
__a : Union[str, Any] = [f'''<madeupword{i}>''' for i in range(self.num_madeup_words )]
__a : Union[str, Any] = kwargs.get('additional_special_tokens' , [] )
kwargs["additional_special_tokens"] += [
word for word in madeup_words if word not in kwargs["additional_special_tokens"]
]
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , cls_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : List[str] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(str(SCREAMING_SNAKE_CASE__ ) )
__a : str = vocab_file
# Original fairseq vocab and spm vocab must be "aligned":
# Vocab | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
# -------- | ------- | ------- | ------ | ------- | --- | --- | --- | ----- | ----- | ----
# fairseq | '<s>' | '<pad>' | '</s>' | '<unk>' | ',' | '.' | '▁' | 's' | '▁de' | '-'
# spm | '<unk>' | '<s>' | '</s>' | ',' | '.' | '▁' | 's' | '▁de' | '-' | '▁a'
# The first "real" token "," has position 4 in the original fairseq vocab and position 3 in the spm vocab
__a : Any = 1
# Mimic fairseq token-to-id alignment for the first 4 token
__a : str = {'<s>': 0, '<pad>': 1, '</s>': 2, '<unk>': 3}
__a : List[str] = len(self.sp_model )
__a : Optional[int] = {f'''<madeupword{i}>''': sp_size + i + self.fairseq_offset for i in range(self.num_madeup_words )}
self.fairseq_tokens_to_ids.update(SCREAMING_SNAKE_CASE__ )
__a : Dict = {v: k for k, v in self.fairseq_tokens_to_ids.items()}
def __getstate__( self : List[str] ):
'''simple docstring'''
__a : Tuple = self.__dict__.copy()
__a : List[str] = None
__a : Optional[int] = self.sp_model.serialized_model_proto()
return state
def __setstate__( self : List[str] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : int = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : Dict = {}
__a : Tuple = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.LoadFromSerializedProto(self.sp_model_proto )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
if token_ids_a is None:
return [self.sep_token_id] + token_ids_a
__a : Optional[Any] = [self.sep_token_id]
return sep + token_ids_a + sep + sep + token_ids_a
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : bool = False ):
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=SCREAMING_SNAKE_CASE__ , token_ids_a=SCREAMING_SNAKE_CASE__ , already_has_special_tokens=SCREAMING_SNAKE_CASE__ )
if token_ids_a is None:
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ )) + [1, 1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
__a : Optional[int] = [self.sep_token_id]
if token_ids_a is None:
return len(sep + token_ids_a ) * [0]
return len(sep + token_ids_a + sep + sep + token_ids_a ) * [0]
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return len(self.sp_model ) + self.fairseq_offset + self.num_madeup_words
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : str = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if token in self.fairseq_tokens_to_ids:
return self.fairseq_tokens_to_ids[token]
__a : List[str] = self.sp_model.PieceToId(SCREAMING_SNAKE_CASE__ )
# Need to return unknown token if the SP model returned 0
return spm_id + self.fairseq_offset if spm_id else self.unk_token_id
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if index in self.fairseq_ids_to_tokens:
return self.fairseq_ids_to_tokens[index]
return self.sp_model.IdToPiece(index - self.fairseq_offset )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
__a : Optional[int] = ''.join(SCREAMING_SNAKE_CASE__ ).replace(SCREAMING_SNAKE_CASE__ , ' ' ).strip()
return out_string
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Any = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[Any] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : list[int] , lowerCamelCase_ : int ):
def count_of_possible_combinations(lowerCamelCase_ : int ) -> int:
if target < 0:
return 0
if target == 0:
return 1
return sum(count_of_possible_combinations(target - item ) for item in array )
return count_of_possible_combinations(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : list[int] , lowerCamelCase_ : int ):
def count_of_possible_combinations_with_dp_array(
lowerCamelCase_ : int , lowerCamelCase_ : list[int] ) -> int:
if target < 0:
return 0
if target == 0:
return 1
if dp_array[target] != -1:
return dp_array[target]
__a : Dict = sum(
count_of_possible_combinations_with_dp_array(target - item , lowerCamelCase_ )
for item in array )
__a : Any = answer
return answer
__a : Dict = [-1] * (target + 1)
return count_of_possible_combinations_with_dp_array(lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : list[int] , lowerCamelCase_ : int ):
__a : Optional[int] = [0] * (target + 1)
__a : Tuple = 1
for i in range(1 , target + 1 ):
for j in range(lowerCamelCase_ ):
if i - array[j] >= 0:
dp_array[i] += dp_array[i - array[j]]
return dp_array[target]
if __name__ == "__main__":
import doctest
doctest.testmod()
SCREAMING_SNAKE_CASE__ = 3
SCREAMING_SNAKE_CASE__ = 5
SCREAMING_SNAKE_CASE__ = [1, 2, 5]
print(combination_sum_iv(n, array, target))
| 47 |
import argparse
from collections import OrderedDict
from pathlib import Path
import torch
from transformers import (
VisualBertConfig,
VisualBertForMultipleChoice,
VisualBertForPreTraining,
VisualBertForQuestionAnswering,
VisualBertForVisualReasoning,
)
from transformers.utils import logging
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = [
('''bert.bert''', '''visual_bert'''),
('''bert.cls''', '''cls'''),
('''bert.classifier''', '''cls'''),
('''token_type_embeddings_visual''', '''visual_token_type_embeddings'''),
('''position_embeddings_visual''', '''visual_position_embeddings'''),
('''projection''', '''visual_projection'''),
]
SCREAMING_SNAKE_CASE__ = [
'''nlvr2_coco_pre_trained.th''',
'''nlvr2_fine_tuned.th''',
'''nlvr2_pre_trained.th''',
'''vcr_coco_pre_train.th''',
'''vcr_fine_tune.th''',
'''vcr_pre_train.th''',
'''vqa_coco_pre_trained.th''',
'''vqa_fine_tuned.th''',
'''vqa_pre_trained.th''',
]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ):
__a : str = torch.load(lowerCamelCase_ , map_location='cpu' )
return sd
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : int , lowerCamelCase_ : Dict=rename_keys_prefix ):
__a : Optional[Any] = OrderedDict()
__a : Any = torch.arange(config.max_position_embeddings ).expand((1, -1) )
# detector_d = OrderedDict()
for key in d:
if "detector" in key:
# detector_d[key.replace('detector.','')] = d[key]
continue
__a : List[Any] = key
for name_pair in rename_keys_prefix:
__a : List[str] = new_key.replace(name_pair[0] , name_pair[1] )
__a : Any = d[key]
if key == "bert.cls.predictions.decoder.weight":
# Old bert code didn't have `decoder.bias`, but was added separately
__a : int = new_d['cls.predictions.bias']
return new_d
@torch.no_grad()
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : Any ):
assert (
checkpoint_path.split('/' )[-1] in ACCEPTABLE_CHECKPOINTS
), f'''The checkpoint provided must be in {ACCEPTABLE_CHECKPOINTS}.'''
# Get Config
if "pre" in checkpoint_path:
__a : Dict = 'pretraining'
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
elif "vqa_advanced" in checkpoint_path:
__a : int = {'visual_embedding_dim': 2_0_4_8}
elif "vqa" in checkpoint_path:
__a : Tuple = {'visual_embedding_dim': 2_0_4_8}
elif "nlvr" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 1_0_2_4}
else:
raise NotImplementedError(f'''No implementation found for `{checkpoint_path}`.''' )
else:
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
__a : Any = 'multichoice'
elif "vqa_advanced" in checkpoint_path:
__a : Any = {'visual_embedding_dim': 2_0_4_8}
__a : List[str] = 'vqa_advanced'
elif "vqa" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 2_0_4_8, 'num_labels': 3_1_2_9}
__a : List[Any] = 'vqa'
elif "nlvr" in checkpoint_path:
__a : Optional[int] = {
'visual_embedding_dim': 1_0_2_4,
'num_labels': 2,
}
__a : Optional[Any] = 'nlvr'
__a : str = VisualBertConfig(**lowerCamelCase_ )
# Load State Dict
__a : str = load_state_dict(lowerCamelCase_ )
__a : str = get_new_dict(lowerCamelCase_ , lowerCamelCase_ )
if model_type == "pretraining":
__a : Optional[Any] = VisualBertForPreTraining(lowerCamelCase_ )
elif model_type == "vqa":
__a : Any = VisualBertForQuestionAnswering(lowerCamelCase_ )
elif model_type == "nlvr":
__a : int = VisualBertForVisualReasoning(lowerCamelCase_ )
elif model_type == "multichoice":
__a : Optional[int] = VisualBertForMultipleChoice(lowerCamelCase_ )
model.load_state_dict(lowerCamelCase_ )
# Save Checkpoints
Path(lowerCamelCase_ ).mkdir(exist_ok=lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument('''orig_checkpoint_path''', type=str, help='''A path to .th on local filesystem.''')
parser.add_argument('''pytorch_dump_folder_path''', type=str, help='''Path to the output PyTorch model.''')
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_visual_bert_checkpoint(args.orig_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
from collections.abc import Generator
def UpperCAmelCase__ ( ):
__a , __a : Tuple = 0, 1
while True:
__a , __a : Dict = b, a + b
yield b
def UpperCAmelCase__ ( lowerCamelCase_ : int = 1_0_0_0 ):
__a : str = 1
__a : Dict = fibonacci_generator()
while len(str(next(lowerCamelCase_ ) ) ) < n:
answer += 1
return answer + 1
if __name__ == "__main__":
print(solution(int(str(input()).strip())))
| 47 |
print((lambda quine: quine % quine)('''print((lambda quine: quine %% quine)(%r))'''))
| 47 | 1 |
import math
def UpperCAmelCase__ ( lowerCamelCase_ : list , lowerCamelCase_ : int = 0 , lowerCamelCase_ : int = 0 ):
__a : List[str] = end or len(lowerCamelCase_ )
for i in range(lowerCamelCase_ , lowerCamelCase_ ):
__a : str = i
__a : int = array[i]
while temp_index != start and temp_index_value < array[temp_index - 1]:
__a : Dict = array[temp_index - 1]
temp_index -= 1
__a : List[Any] = temp_index_value
return array
def UpperCAmelCase__ ( lowerCamelCase_ : list , lowerCamelCase_ : int , lowerCamelCase_ : int ): # Max Heap
__a : Optional[int] = index
__a : int = 2 * index + 1 # Left Node
__a : str = 2 * index + 2 # Right Node
if left_index < heap_size and array[largest] < array[left_index]:
__a : Dict = left_index
if right_index < heap_size and array[largest] < array[right_index]:
__a : Tuple = right_index
if largest != index:
__a , __a : int = array[largest], array[index]
heapify(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : list ):
__a : Dict = len(lowerCamelCase_ )
for i in range(n // 2 , -1 , -1 ):
heapify(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
for i in range(n - 1 , 0 , -1 ):
__a , __a : Any = array[0], array[i]
heapify(lowerCamelCase_ , 0 , lowerCamelCase_ )
return array
def UpperCAmelCase__ ( lowerCamelCase_ : list , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : int ):
if (array[first_index] > array[middle_index]) != (
array[first_index] > array[last_index]
):
return array[first_index]
elif (array[middle_index] > array[first_index]) != (
array[middle_index] > array[last_index]
):
return array[middle_index]
else:
return array[last_index]
def UpperCAmelCase__ ( lowerCamelCase_ : list , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : int ):
__a : int = low
__a : Tuple = high
while True:
while array[i] < pivot:
i += 1
j -= 1
while pivot < array[j]:
j -= 1
if i >= j:
return i
__a , __a : int = array[j], array[i]
i += 1
def UpperCAmelCase__ ( lowerCamelCase_ : list ):
if len(lowerCamelCase_ ) == 0:
return array
__a : List[str] = 2 * math.ceil(math.loga(len(lowerCamelCase_ ) ) )
__a : List[Any] = 1_6
return intro_sort(lowerCamelCase_ , 0 , len(lowerCamelCase_ ) , lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : list , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : int ):
while end - start > size_threshold:
if max_depth == 0:
return heap_sort(lowerCamelCase_ )
max_depth -= 1
__a : Tuple = median_of_a(lowerCamelCase_ , lowerCamelCase_ , start + ((end - start) // 2) + 1 , end - 1 )
__a : Union[str, Any] = partition(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
intro_sort(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
__a : Dict = p
return insertion_sort(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
import doctest
doctest.testmod()
SCREAMING_SNAKE_CASE__ = input('''Enter numbers separated by a comma : ''').strip()
SCREAMING_SNAKE_CASE__ = [float(item) for item in user_input.split(''',''')]
print(sort(unsorted))
| 47 |
import json
import os
import shutil
import tempfile
from unittest import TestCase
from transformers import BartTokenizer, BartTokenizerFast, DPRQuestionEncoderTokenizer, DPRQuestionEncoderTokenizerFast
from transformers.models.bart.configuration_bart import BartConfig
from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES
from transformers.models.dpr.configuration_dpr import DPRConfig
from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
from transformers.testing_utils import require_faiss, require_tokenizers, require_torch, slow
from transformers.utils import is_datasets_available, is_faiss_available, is_torch_available
if is_torch_available() and is_datasets_available() and is_faiss_available():
from transformers.models.rag.configuration_rag import RagConfig
from transformers.models.rag.tokenization_rag import RagTokenizer
@require_faiss
@require_torch
class _UpperCamelCase( __lowerCamelCase ):
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[Any] = tempfile.mkdtemp()
__a : int = 8
# DPR tok
__a : Dict = [
'[UNK]',
'[CLS]',
'[SEP]',
'[PAD]',
'[MASK]',
'want',
'##want',
'##ed',
'wa',
'un',
'runn',
'##ing',
',',
'low',
'lowest',
]
__a : int = os.path.join(self.tmpdirname , 'dpr_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , DPR_VOCAB_FILES_NAMES['vocab_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as vocab_writer:
vocab_writer.write(''.join([x + '\n' for x in vocab_tokens] ) )
# BART tok
__a : str = [
'l',
'o',
'w',
'e',
'r',
's',
't',
'i',
'd',
'n',
'\u0120',
'\u0120l',
'\u0120n',
'\u0120lo',
'\u0120low',
'er',
'\u0120lowest',
'\u0120newer',
'\u0120wider',
'<unk>',
]
__a : Optional[int] = dict(zip(SCREAMING_SNAKE_CASE__ , range(len(SCREAMING_SNAKE_CASE__ ) ) ) )
__a : List[str] = ['#version: 0.2', '\u0120 l', '\u0120l o', '\u0120lo w', 'e r', '']
__a : List[str] = {'unk_token': '<unk>'}
__a : Dict = os.path.join(self.tmpdirname , 'bart_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['vocab_file'] )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(SCREAMING_SNAKE_CASE__ ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'dpr_tokenizer' ) )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'bart_tokenizer' ) )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
shutil.rmtree(self.tmpdirname )
@require_tokenizers
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Tuple = os.path.join(self.tmpdirname , 'rag_tokenizer' )
__a : Optional[Any] = RagConfig(question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() )
__a : Optional[Any] = RagTokenizer(question_encoder=self.get_dpr_tokenizer() , generator=self.get_bart_tokenizer() )
rag_config.save_pretrained(SCREAMING_SNAKE_CASE__ )
rag_tokenizer.save_pretrained(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = RagTokenizer.from_pretrained(SCREAMING_SNAKE_CASE__ , config=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(new_rag_tokenizer.question_encoder , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.question_encoder.get_vocab() , rag_tokenizer.question_encoder.get_vocab() )
self.assertIsInstance(new_rag_tokenizer.generator , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.generator.get_vocab() , rag_tokenizer.generator.get_vocab() )
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Optional[Any] = RagTokenizer.from_pretrained('facebook/rag-token-nq' )
__a : List[Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : Tuple = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Any = RagTokenizer.from_pretrained('facebook/rag-sequence-nq' )
__a : Union[str, Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : str = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
| 47 | 1 |
from string import ascii_lowercase, ascii_uppercase
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
if not sentence:
return ""
__a : Union[str, Any] = dict(zip(lowerCamelCase_ , lowerCamelCase_ ) )
return lower_to_upper.get(sentence[0] , sentence[0] ) + sentence[1:]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''spiece.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''bert_for_seq_generation''': (
'''https://huggingface.co/google/bert_for_seq_generation_L-24_bbc_encoder/resolve/main/spiece.model'''
),
}
}
SCREAMING_SNAKE_CASE__ = {'''bert_for_seq_generation''': 512}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : Any = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : List[int] = []
__SCREAMING_SNAKE_CASE : int = ['''input_ids''', '''attention_mask''']
def __init__( self : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : Tuple="</s>" , SCREAMING_SNAKE_CASE__ : Any="<unk>" , SCREAMING_SNAKE_CASE__ : int="<pad>" , SCREAMING_SNAKE_CASE__ : List[str]="<::::>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Tuple = {} if sp_model_kwargs is None else sp_model_kwargs
# Add extra_ids to the special token list
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : int = vocab_file
__a : Union[str, Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return self.sp_model.get_piece_size()
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Dict = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __getstate__( self : Optional[Any] ):
'''simple docstring'''
__a : Union[str, Any] = self.__dict__.copy()
__a : Any = None
return state
def __setstate__( self : int , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a : str = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : str = {}
__a : List[Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(self.vocab_file )
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
return self.sp_model.piece_to_id(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : int = self.sp_model.IdToPiece(SCREAMING_SNAKE_CASE__ )
return token
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Optional[Any] = []
__a : Optional[int] = ''
for token in tokens:
# make sure that special tokens are not decoded using sentencepiece model
if token in self.all_special_tokens:
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ ) + token
__a : Dict = []
else:
current_sub_tokens.append(SCREAMING_SNAKE_CASE__ )
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ )
return out_string.strip()
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Tuple = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[str] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''vinvino02/glpn-kitti''': '''https://huggingface.co/vinvino02/glpn-kitti/resolve/main/config.json''',
# See all GLPN models at https://huggingface.co/models?filter=glpn
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = '''glpn'''
def __init__( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Dict=3 , SCREAMING_SNAKE_CASE__ : List[Any]=4 , SCREAMING_SNAKE_CASE__ : Optional[Any]=[2, 2, 2, 2] , SCREAMING_SNAKE_CASE__ : List[Any]=[8, 4, 2, 1] , SCREAMING_SNAKE_CASE__ : str=[3_2, 6_4, 1_6_0, 2_5_6] , SCREAMING_SNAKE_CASE__ : Dict=[7, 3, 3, 3] , SCREAMING_SNAKE_CASE__ : Tuple=[4, 2, 2, 2] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=[1, 2, 5, 8] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=[4, 4, 4, 4] , SCREAMING_SNAKE_CASE__ : List[Any]="gelu" , SCREAMING_SNAKE_CASE__ : List[Any]=0.0 , SCREAMING_SNAKE_CASE__ : Any=0.0 , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.02 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : List[Any]=1e-6 , SCREAMING_SNAKE_CASE__ : int=6_4 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_0 , SCREAMING_SNAKE_CASE__ : Any=-1 , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ )
__a : List[str] = num_channels
__a : Dict = num_encoder_blocks
__a : List[str] = depths
__a : Optional[int] = sr_ratios
__a : int = hidden_sizes
__a : str = patch_sizes
__a : Union[str, Any] = strides
__a : str = mlp_ratios
__a : Optional[int] = num_attention_heads
__a : List[Any] = hidden_act
__a : Any = hidden_dropout_prob
__a : List[Any] = attention_probs_dropout_prob
__a : str = initializer_range
__a : int = drop_path_rate
__a : int = layer_norm_eps
__a : Optional[Any] = decoder_hidden_size
__a : Dict = max_depth
__a : Union[str, Any] = head_in_index
| 47 |
from ..utils import DummyObject, requires_backends
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Dict , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Tuple , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Any = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Tuple , *SCREAMING_SNAKE_CASE__ : List[str] , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Dict , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Dict = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[int] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Optional[int] , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
| 47 | 1 |
import warnings
from ...utils import is_sklearn_available, requires_backends
if is_sklearn_available():
from scipy.stats import pearsonr, spearmanr
from sklearn.metrics import fa_score, matthews_corrcoef
SCREAMING_SNAKE_CASE__ = (
'''This metric will be removed from the library soon, metrics should be handled with the 🤗 Evaluate '''
'''library. You can have a look at this example script for pointers: '''
'''https://github.com/huggingface/transformers/blob/main/examples/pytorch/text-classification/run_glue.py'''
)
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : List[str] ):
warnings.warn(lowerCamelCase_ , lowerCamelCase_ )
requires_backends(lowerCamelCase_ , 'sklearn' )
return (preds == labels).mean()
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : Any ):
warnings.warn(lowerCamelCase_ , lowerCamelCase_ )
requires_backends(lowerCamelCase_ , 'sklearn' )
__a : List[str] = simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )
__a : Any = fa_score(y_true=lowerCamelCase_ , y_pred=lowerCamelCase_ )
return {
"acc": acc,
"f1": fa,
"acc_and_f1": (acc + fa) / 2,
}
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : str ):
warnings.warn(lowerCamelCase_ , lowerCamelCase_ )
requires_backends(lowerCamelCase_ , 'sklearn' )
__a : Dict = pearsonr(lowerCamelCase_ , lowerCamelCase_ )[0]
__a : List[str] = spearmanr(lowerCamelCase_ , lowerCamelCase_ )[0]
return {
"pearson": pearson_corr,
"spearmanr": spearman_corr,
"corr": (pearson_corr + spearman_corr) / 2,
}
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Dict , lowerCamelCase_ : Any ):
warnings.warn(lowerCamelCase_ , lowerCamelCase_ )
requires_backends(lowerCamelCase_ , 'sklearn' )
assert len(lowerCamelCase_ ) == len(lowerCamelCase_ ), f'''Predictions and labels have mismatched lengths {len(lowerCamelCase_ )} and {len(lowerCamelCase_ )}'''
if task_name == "cola":
return {"mcc": matthews_corrcoef(lowerCamelCase_ , lowerCamelCase_ )}
elif task_name == "sst-2":
return {"acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
elif task_name == "mrpc":
return acc_and_fa(lowerCamelCase_ , lowerCamelCase_ )
elif task_name == "sts-b":
return pearson_and_spearman(lowerCamelCase_ , lowerCamelCase_ )
elif task_name == "qqp":
return acc_and_fa(lowerCamelCase_ , lowerCamelCase_ )
elif task_name == "mnli":
return {"mnli/acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
elif task_name == "mnli-mm":
return {"mnli-mm/acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
elif task_name == "qnli":
return {"acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
elif task_name == "rte":
return {"acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
elif task_name == "wnli":
return {"acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
elif task_name == "hans":
return {"acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
else:
raise KeyError(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Any , lowerCamelCase_ : List[Any] ):
warnings.warn(lowerCamelCase_ , lowerCamelCase_ )
requires_backends(lowerCamelCase_ , 'sklearn' )
if len(lowerCamelCase_ ) != len(lowerCamelCase_ ):
raise ValueError(f'''Predictions and labels have mismatched lengths {len(lowerCamelCase_ )} and {len(lowerCamelCase_ )}''' )
if task_name == "xnli":
return {"acc": simple_accuracy(lowerCamelCase_ , lowerCamelCase_ )}
else:
raise KeyError(lowerCamelCase_ )
| 47 |
import math
from datetime import datetime, timedelta
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
__a : Union[str, Any] = year % 1_9
__a : int = year % 4
__a : Optional[int] = year % 7
__a : Dict = math.floor(year / 1_0_0 )
__a : Optional[Any] = math.floor((1_3 + 8 * leap_day_inhibits) / 2_5 )
__a : Union[str, Any] = leap_day_inhibits / 4
__a : str = (
1_5 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 3_0
__a : Union[str, Any] = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
__a : List[Any] = (1_9 * metonic_cycle + secular_moon_shift) % 3_0
# PHM -> Paschal Full Moon
__a : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 2_9 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_9 )
elif days_to_add == 2_8 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_8 )
else:
return datetime(lowerCamelCase_ , 3 , 2_2 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
SCREAMING_SNAKE_CASE__ = '''will be''' if year > datetime.now().year else '''was'''
print(F"Easter in {year} {tense} {gauss_easter(year)}")
| 47 | 1 |
import pytest
import requests
from datasets.utils.file_utils import http_head
from .utils import OfflineSimulationMode, RequestWouldHangIndefinitelyError, offline
@pytest.mark.integration
def UpperCAmelCase__ ( ):
with offline(OfflineSimulationMode.CONNECTION_TIMES_OUT ):
with pytest.raises(lowerCamelCase_ ):
requests.request('GET' , 'https://huggingface.co' )
with pytest.raises(requests.exceptions.ConnectTimeout ):
requests.request('GET' , 'https://huggingface.co' , timeout=1.0 )
@pytest.mark.integration
def UpperCAmelCase__ ( ):
with offline(OfflineSimulationMode.CONNECTION_FAILS ):
with pytest.raises(requests.exceptions.ConnectionError ):
requests.request('GET' , 'https://huggingface.co' )
def UpperCAmelCase__ ( ):
with offline(OfflineSimulationMode.HF_DATASETS_OFFLINE_SET_TO_1 ):
with pytest.raises(lowerCamelCase_ ):
http_head('https://huggingface.co' )
| 47 |
from typing import List, Optional, Union
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''huggingface/informer-tourism-monthly''': (
'''https://huggingface.co/huggingface/informer-tourism-monthly/resolve/main/config.json'''
),
# See all Informer models at https://huggingface.co/models?filter=informer
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = '''informer'''
__SCREAMING_SNAKE_CASE : List[Any] = {
'''hidden_size''': '''d_model''',
'''num_attention_heads''': '''encoder_attention_heads''',
'''num_hidden_layers''': '''encoder_layers''',
}
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : str = "student_t" , SCREAMING_SNAKE_CASE__ : str = "nll" , SCREAMING_SNAKE_CASE__ : int = 1 , SCREAMING_SNAKE_CASE__ : List[int] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[str, bool]] = "mean" , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : int = 6_4 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : str = "gelu" , SCREAMING_SNAKE_CASE__ : float = 0.05 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : int = 1_0_0 , SCREAMING_SNAKE_CASE__ : float = 0.02 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=True , SCREAMING_SNAKE_CASE__ : str = "prob" , SCREAMING_SNAKE_CASE__ : int = 5 , SCREAMING_SNAKE_CASE__ : bool = True , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Dict = prediction_length
__a : Tuple = context_length or prediction_length
__a : Tuple = distribution_output
__a : Tuple = loss
__a : str = input_size
__a : Dict = num_time_features
__a : Optional[int] = lags_sequence if lags_sequence is not None else [1, 2, 3, 4, 5, 6, 7]
__a : str = scaling
__a : Tuple = num_dynamic_real_features
__a : int = num_static_real_features
__a : Dict = num_static_categorical_features
# set cardinality
if cardinality and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The cardinality should be a list of the same length as `num_static_categorical_features`' )
__a : Optional[Any] = cardinality
else:
__a : Optional[int] = [0]
# set embedding_dimension
if embedding_dimension and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The embedding dimension should be a list of the same length as `num_static_categorical_features`' )
__a : int = embedding_dimension
else:
__a : List[Any] = [min(5_0 , (cat + 1) // 2 ) for cat in self.cardinality]
__a : int = num_parallel_samples
# Transformer architecture configuration
__a : str = input_size * len(self.lags_sequence ) + self._number_of_features
__a : Optional[int] = d_model
__a : Union[str, Any] = encoder_attention_heads
__a : int = decoder_attention_heads
__a : Any = encoder_ffn_dim
__a : Union[str, Any] = decoder_ffn_dim
__a : List[Any] = encoder_layers
__a : Optional[int] = decoder_layers
__a : int = dropout
__a : Optional[Any] = attention_dropout
__a : Dict = activation_dropout
__a : Union[str, Any] = encoder_layerdrop
__a : Optional[int] = decoder_layerdrop
__a : List[str] = activation_function
__a : str = init_std
__a : Optional[int] = use_cache
# Informer
__a : Union[str, Any] = attention_type
__a : str = sampling_factor
__a : Dict = distil
super().__init__(is_encoder_decoder=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
return (
sum(self.embedding_dimension )
+ self.num_dynamic_real_features
+ self.num_time_features
+ self.num_static_real_features
+ self.input_size * 2 # the log1p(abs(loc)) and log(scale) features
)
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : list ):
if len(lowerCamelCase_ ) <= 1:
return lst
__a : str = 1
while i < len(lowerCamelCase_ ):
if lst[i - 1] <= lst[i]:
i += 1
else:
__a , __a : Any = lst[i], lst[i - 1]
i -= 1
if i == 0:
__a : int = 1
return lst
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter numbers separated by a comma:\n''').strip()
SCREAMING_SNAKE_CASE__ = [int(item) for item in user_input.split(''',''')]
print(gnome_sort(unsorted))
| 47 |
import torch
from diffusers import DDIMParallelScheduler
from .test_schedulers import SchedulerCommonTest
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = (DDIMParallelScheduler,)
__SCREAMING_SNAKE_CASE : Union[str, Any] = (('''eta''', 0.0), ('''num_inference_steps''', 50))
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : List[Any] = {
'num_train_timesteps': 1_0_0_0,
'beta_start': 0.0_001,
'beta_end': 0.02,
'beta_schedule': 'linear',
'clip_sample': True,
}
config.update(**SCREAMING_SNAKE_CASE__ )
return config
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Tuple = self.scheduler_classes[0]
__a : Optional[Any] = self.get_scheduler_config(**SCREAMING_SNAKE_CASE__ )
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : List[str] = 1_0, 0.0
__a : Dict = self.dummy_model()
__a : str = self.dummy_sample_deter
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
for t in scheduler.timesteps:
__a : str = model(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[str] = scheduler.step(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ).prev_sample
return sample
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
for timesteps in [1_0_0, 5_0_0, 1_0_0_0]:
self.check_over_configs(num_train_timesteps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
for steps_offset in [0, 1]:
self.check_over_configs(steps_offset=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config(steps_offset=1 )
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
scheduler.set_timesteps(5 )
assert torch.equal(scheduler.timesteps , torch.LongTensor([8_0_1, 6_0_1, 4_0_1, 2_0_1, 1] ) )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for beta_start, beta_end in zip([0.0_001, 0.001, 0.01, 0.1] , [0.002, 0.02, 0.2, 2] ):
self.check_over_configs(beta_start=SCREAMING_SNAKE_CASE__ , beta_end=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for schedule in ["linear", "squaredcos_cap_v2"]:
self.check_over_configs(beta_schedule=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(prediction_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for clip_sample in [True, False]:
self.check_over_configs(clip_sample=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for timestep_spacing in ["trailing", "leading"]:
self.check_over_configs(timestep_spacing=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for rescale_betas_zero_snr in [True, False]:
self.check_over_configs(rescale_betas_zero_snr=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
self.check_over_configs(thresholding=SCREAMING_SNAKE_CASE__ )
for threshold in [0.5, 1.0, 2.0]:
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(
thresholding=SCREAMING_SNAKE_CASE__ , prediction_type=SCREAMING_SNAKE_CASE__ , sample_max_value=SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for t in [1, 1_0, 4_9]:
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
for t, num_inference_steps in zip([1, 1_0, 5_0] , [1_0, 5_0, 5_0_0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , num_inference_steps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for t, eta in zip([1, 1_0, 4_9] , [0.0, 0.5, 1.0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , eta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : Union[str, Any] = self.get_scheduler_config()
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_2_0 , 4_0_0 ) - 0.14_771 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_8_0 , 9_6_0 ) - 0.32_460 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_8_7 , 4_8_6 ) - 0.00_979 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_9_9 , 9_9_8 ) - 0.02 ) ) < 1e-5
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config()
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : Any = 1_0, 0.0
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = self.dummy_model()
__a : int = self.dummy_sample_deter
__a : List[Any] = self.dummy_sample_deter + 0.1
__a : List[str] = self.dummy_sample_deter - 0.1
__a : Optional[Any] = samplea.shape[0]
__a : Optional[Any] = torch.stack([samplea, samplea, samplea] , dim=0 )
__a : Union[str, Any] = torch.arange(SCREAMING_SNAKE_CASE__ )[0:3, None].repeat(1 , SCREAMING_SNAKE_CASE__ )
__a : int = model(samples.flatten(0 , 1 ) , timesteps.flatten(0 , 1 ) )
__a : int = scheduler.batch_step_no_noise(SCREAMING_SNAKE_CASE__ , timesteps.flatten(0 , 1 ) , samples.flatten(0 , 1 ) , SCREAMING_SNAKE_CASE__ )
__a : Dict = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 1_147.7_904 ) < 1e-2
assert abs(result_mean.item() - 0.4_982 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : List[str] = self.full_loop()
__a : Tuple = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : int = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 172.0_067 ) < 1e-2
assert abs(result_mean.item() - 0.223_967 ) < 1e-3
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Optional[int] = self.full_loop(prediction_type='v_prediction' )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Union[str, Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 52.5_302 ) < 1e-2
assert abs(result_mean.item() - 0.0_684 ) < 1e-3
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Union[str, Any] = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : Optional[int] = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[int] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.8_295 ) < 1e-2
assert abs(result_mean.item() - 0.1_951 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Dict = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Tuple = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.0_784 ) < 1e-2
assert abs(result_mean.item() - 0.1_941 ) < 1e-3
| 47 | 1 |
import os
from distutils.util import strtobool
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple , lowerCamelCase_ : Optional[int] ):
for e in env_keys:
__a : List[str] = int(os.environ.get(lowerCamelCase_ , -1 ) )
if val >= 0:
return val
return default
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : str=False ):
__a : Dict = os.environ.get(lowerCamelCase_ , str(lowerCamelCase_ ) )
return strtobool(lowerCamelCase_ ) == 1 # As its name indicates `strtobool` actually returns an int...
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Tuple="no" ):
__a : str = os.environ.get(lowerCamelCase_ , str(lowerCamelCase_ ) )
return value
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : list[int] ):
# Check if the input is valid
if not len(lowerCamelCase_ ) == len(lowerCamelCase_ ) == 3:
raise ValueError('Please enter a valid equation.' )
if equationa[0] == equationa[1] == equationa[0] == equationa[1] == 0:
raise ValueError('Both a & b of two equations can\'t be zero.' )
# Extract the coefficients
__a , __a , __a : Optional[Any] = equationa
__a , __a , __a : Optional[int] = equationa
# Calculate the determinants of the matrices
__a : str = aa * ba - aa * ba
__a : Tuple = ca * ba - ca * ba
__a : Union[str, Any] = aa * ca - aa * ca
# Check if the system of linear equations has a solution (using Cramer's rule)
if determinant == 0:
if determinant_x == determinant_y == 0:
raise ValueError('Infinite solutions. (Consistent system)' )
else:
raise ValueError('No solution. (Inconsistent system)' )
else:
if determinant_x == determinant_y == 0:
# Trivial solution (Inconsistent system)
return (0.0, 0.0)
else:
__a : Any = determinant_x / determinant
__a : Optional[Any] = determinant_y / determinant
# Non-Trivial Solution (Consistent system)
return (x, y)
| 47 | 1 |
import os
import tempfile
import unittest
from transformers import FlaubertConfig, is_torch_available
from transformers.testing_utils import require_torch, require_torch_gpu, slow, torch_device
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import (
FlaubertForMultipleChoice,
FlaubertForQuestionAnswering,
FlaubertForQuestionAnsweringSimple,
FlaubertForSequenceClassification,
FlaubertForTokenClassification,
FlaubertModel,
FlaubertWithLMHeadModel,
)
from transformers.models.flaubert.modeling_flaubert import FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST
class _UpperCamelCase( __lowerCamelCase ):
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Any=1_3 , SCREAMING_SNAKE_CASE__ : Optional[Any]=7 , SCREAMING_SNAKE_CASE__ : str=True , SCREAMING_SNAKE_CASE__ : str=True , SCREAMING_SNAKE_CASE__ : List[Any]=True , SCREAMING_SNAKE_CASE__ : Dict=True , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : Tuple=False , SCREAMING_SNAKE_CASE__ : List[str]=2 , SCREAMING_SNAKE_CASE__ : List[str]=9_9 , SCREAMING_SNAKE_CASE__ : Any=0 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=3_2 , SCREAMING_SNAKE_CASE__ : Any=5 , SCREAMING_SNAKE_CASE__ : Any=4 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : List[Any]=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=5_1_2 , SCREAMING_SNAKE_CASE__ : Tuple=1_2 , SCREAMING_SNAKE_CASE__ : Dict=2 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.02 , SCREAMING_SNAKE_CASE__ : Optional[int]=3 , SCREAMING_SNAKE_CASE__ : Any=4 , SCREAMING_SNAKE_CASE__ : List[Any]="last" , SCREAMING_SNAKE_CASE__ : Optional[int]=None , SCREAMING_SNAKE_CASE__ : str=None , ):
'''simple docstring'''
__a : Dict = parent
__a : Dict = batch_size
__a : Union[str, Any] = seq_length
__a : Union[str, Any] = is_training
__a : List[Any] = use_input_lengths
__a : Any = use_token_type_ids
__a : Any = use_labels
__a : Optional[Any] = gelu_activation
__a : List[Any] = sinusoidal_embeddings
__a : int = causal
__a : Tuple = asm
__a : List[str] = n_langs
__a : Tuple = vocab_size
__a : Dict = n_special
__a : int = hidden_size
__a : Dict = num_hidden_layers
__a : Dict = num_attention_heads
__a : str = hidden_dropout_prob
__a : Optional[Any] = attention_probs_dropout_prob
__a : List[str] = max_position_embeddings
__a : Tuple = type_vocab_size
__a : str = type_sequence_label_size
__a : Dict = initializer_range
__a : int = num_labels
__a : Optional[int] = num_choices
__a : List[Any] = summary_type
__a : Any = use_proj
__a : Any = scope
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : Any = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a : Dict = random_attention_mask([self.batch_size, self.seq_length] )
__a : List[Any] = None
if self.use_input_lengths:
__a : List[Any] = (
ids_tensor([self.batch_size] , vocab_size=2 ) + self.seq_length - 2
) # small variation of seq_length
__a : str = None
if self.use_token_type_ids:
__a : List[Any] = ids_tensor([self.batch_size, self.seq_length] , self.n_langs )
__a : int = None
__a : str = None
__a : Dict = None
if self.use_labels:
__a : Optional[int] = ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a : int = ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a : List[Any] = ids_tensor([self.batch_size] , 2 ).float()
__a : Union[str, Any] = ids_tensor([self.batch_size] , self.num_choices )
__a : Dict = self.get_config()
return (
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
)
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return FlaubertConfig(
vocab_size=self.vocab_size , n_special=self.n_special , emb_dim=self.hidden_size , n_layers=self.num_hidden_layers , n_heads=self.num_attention_heads , dropout=self.hidden_dropout_prob , attention_dropout=self.attention_probs_dropout_prob , gelu_activation=self.gelu_activation , sinusoidal_embeddings=self.sinusoidal_embeddings , asm=self.asm , causal=self.causal , n_langs=self.n_langs , max_position_embeddings=self.max_position_embeddings , initializer_range=self.initializer_range , summary_type=self.summary_type , use_proj=self.use_proj , )
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : str , ):
'''simple docstring'''
__a : Tuple = FlaubertModel(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : List[Any] = model(SCREAMING_SNAKE_CASE__ , lengths=SCREAMING_SNAKE_CASE__ , langs=SCREAMING_SNAKE_CASE__ )
__a : int = model(SCREAMING_SNAKE_CASE__ , langs=SCREAMING_SNAKE_CASE__ )
__a : int = model(SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Union[str, Any] , ):
'''simple docstring'''
__a : Any = FlaubertWithLMHeadModel(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Optional[int] = model(SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Union[str, Any] = FlaubertForQuestionAnsweringSimple(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : int = model(SCREAMING_SNAKE_CASE__ )
__a : Tuple = model(SCREAMING_SNAKE_CASE__ , start_positions=SCREAMING_SNAKE_CASE__ , end_positions=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.start_logits.shape , (self.batch_size, self.seq_length) )
self.parent.assertEqual(result.end_logits.shape , (self.batch_size, self.seq_length) )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Dict , ):
'''simple docstring'''
__a : Tuple = FlaubertForQuestionAnswering(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : List[Any] = model(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = model(
SCREAMING_SNAKE_CASE__ , start_positions=SCREAMING_SNAKE_CASE__ , end_positions=SCREAMING_SNAKE_CASE__ , cls_index=SCREAMING_SNAKE_CASE__ , is_impossible=SCREAMING_SNAKE_CASE__ , p_mask=SCREAMING_SNAKE_CASE__ , )
__a : List[Any] = model(
SCREAMING_SNAKE_CASE__ , start_positions=SCREAMING_SNAKE_CASE__ , end_positions=SCREAMING_SNAKE_CASE__ , cls_index=SCREAMING_SNAKE_CASE__ , is_impossible=SCREAMING_SNAKE_CASE__ , )
((__a) , ) : Optional[Any] = result_with_labels.to_tuple()
__a : List[Any] = model(SCREAMING_SNAKE_CASE__ , start_positions=SCREAMING_SNAKE_CASE__ , end_positions=SCREAMING_SNAKE_CASE__ )
((__a) , ) : List[Any] = result_with_labels.to_tuple()
self.parent.assertEqual(result_with_labels.loss.shape , () )
self.parent.assertEqual(result.start_top_log_probs.shape , (self.batch_size, model.config.start_n_top) )
self.parent.assertEqual(result.start_top_index.shape , (self.batch_size, model.config.start_n_top) )
self.parent.assertEqual(
result.end_top_log_probs.shape , (self.batch_size, model.config.start_n_top * model.config.end_n_top) )
self.parent.assertEqual(
result.end_top_index.shape , (self.batch_size, model.config.start_n_top * model.config.end_n_top) )
self.parent.assertEqual(result.cls_logits.shape , (self.batch_size,) )
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : int , ):
'''simple docstring'''
__a : Optional[Any] = FlaubertForSequenceClassification(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : List[str] = model(SCREAMING_SNAKE_CASE__ )
__a : int = model(SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : str , ):
'''simple docstring'''
__a : List[Any] = self.num_labels
__a : str = FlaubertForTokenClassification(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Any = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Optional[int] , ):
'''simple docstring'''
__a : Any = self.num_choices
__a : List[str] = FlaubertForMultipleChoice(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Optional[int] = input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a : Optional[int] = token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a : List[str] = input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a : List[Any] = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Dict = self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : str = config_and_inputs
__a : List[str] = {
'input_ids': input_ids,
'token_type_ids': token_type_ids,
'lengths': input_lengths,
'attention_mask': input_mask,
}
return config, inputs_dict
@require_torch
class _UpperCamelCase( __lowerCamelCase , __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : str = (
(
FlaubertModel,
FlaubertWithLMHeadModel,
FlaubertForQuestionAnswering,
FlaubertForQuestionAnsweringSimple,
FlaubertForSequenceClassification,
FlaubertForTokenClassification,
FlaubertForMultipleChoice,
)
if is_torch_available()
else ()
)
__SCREAMING_SNAKE_CASE : str = (
{
'''feature-extraction''': FlaubertModel,
'''fill-mask''': FlaubertWithLMHeadModel,
'''question-answering''': FlaubertForQuestionAnsweringSimple,
'''text-classification''': FlaubertForSequenceClassification,
'''token-classification''': FlaubertForTokenClassification,
'''zero-shot''': FlaubertForSequenceClassification,
}
if is_torch_available()
else {}
)
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
if (
pipeline_test_casse_name == "QAPipelineTests"
and tokenizer_name is not None
and not tokenizer_name.endswith('Fast' )
):
# `QAPipelineTests` fails for a few models when the slower tokenizer are used.
# (The slower tokenizers were never used for pipeline tests before the pipeline testing rework)
# TODO: check (and possibly fix) the `QAPipelineTests` with slower tokenizer
return True
return False
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Tuple=False ):
'''simple docstring'''
__a : List[Any] = super()._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , return_labels=SCREAMING_SNAKE_CASE__ )
if return_labels:
if model_class.__name__ == "FlaubertForQuestionAnswering":
__a : Tuple = torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=SCREAMING_SNAKE_CASE__ )
__a : Tuple = torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=SCREAMING_SNAKE_CASE__ )
return inputs_dict
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : Union[str, Any] = FlaubertModelTester(self )
__a : Dict = ConfigTester(self , config_class=SCREAMING_SNAKE_CASE__ , emb_dim=3_7 )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
self.config_tester.run_common_tests()
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : List[Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_flaubert_model(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Any = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_flaubert_lm_head(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_flaubert_simple_qa(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : str = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_flaubert_qa(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_flaubert_sequence_classif(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_flaubert_token_classif(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : List[Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_flaubert_multiple_choice(*SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
for model_name in FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a : Dict = FlaubertModel.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@slow
@require_torch_gpu
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a , __a : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
# FlauBertForMultipleChoice behaves incorrectly in JIT environments.
if model_class == FlaubertForMultipleChoice:
return
__a : str = True
__a : Tuple = model_class(config=SCREAMING_SNAKE_CASE__ )
__a : Any = self._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = torch.jit.trace(
SCREAMING_SNAKE_CASE__ , (inputs_dict['input_ids'].to('cpu' ), inputs_dict['attention_mask'].to('cpu' )) )
with tempfile.TemporaryDirectory() as tmp:
torch.jit.save(SCREAMING_SNAKE_CASE__ , os.path.join(SCREAMING_SNAKE_CASE__ , 'traced_model.pt' ) )
__a : Tuple = torch.jit.load(os.path.join(SCREAMING_SNAKE_CASE__ , 'traced_model.pt' ) , map_location=SCREAMING_SNAKE_CASE__ )
loaded(inputs_dict['input_ids'].to(SCREAMING_SNAKE_CASE__ ) , inputs_dict['attention_mask'].to(SCREAMING_SNAKE_CASE__ ) )
@require_torch
class _UpperCamelCase( unittest.TestCase ):
@slow
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : List[str] = FlaubertModel.from_pretrained('flaubert/flaubert_base_cased' )
__a : Union[str, Any] = torch.tensor([[0, 3_4_5, 2_3_2, 3_2_8, 7_4_0, 1_4_0, 1_6_9_5, 6_9, 6_0_7_8, 1_5_8_8, 2]] )
with torch.no_grad():
__a : str = model(SCREAMING_SNAKE_CASE__ )[0]
__a : Optional[Any] = torch.Size((1, 1_1, 7_6_8) )
self.assertEqual(output.shape , SCREAMING_SNAKE_CASE__ )
__a : Any = torch.tensor(
[[[-2.6_251, -1.4_298, -0.0_227], [-2.8_510, -1.6_387, 0.2_258], [-2.8_114, -1.1_832, -0.3_066]]] )
self.assertTrue(torch.allclose(output[:, :3, :3] , SCREAMING_SNAKE_CASE__ , atol=1e-4 ) )
| 47 |
from ...utils import is_note_seq_available, is_transformers_available, is_torch_available
from ...utils import OptionalDependencyNotAvailable
try:
if not (is_transformers_available() and is_torch_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_torch_and_transformers_objects import * # noqa F403
else:
from .notes_encoder import SpectrogramNotesEncoder
from .continous_encoder import SpectrogramContEncoder
from .pipeline_spectrogram_diffusion import (
SpectrogramContEncoder,
SpectrogramDiffusionPipeline,
TaFilmDecoder,
)
try:
if not (is_transformers_available() and is_torch_available() and is_note_seq_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_transformers_and_torch_and_note_seq_objects import * # noqa F403
else:
from .midi_utils import MidiProcessor
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : int ):
return "\n".join(
f'''{number} * {i} = {number * i}''' for i in range(1 , number_of_terms + 1 ) )
if __name__ == "__main__":
print(multiplication_table(number=5, number_of_terms=10))
| 47 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_bridgetower''': [
'''BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''BridgeTowerConfig''',
'''BridgeTowerTextConfig''',
'''BridgeTowerVisionConfig''',
],
'''processing_bridgetower''': ['''BridgeTowerProcessor'''],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''BridgeTowerImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''BridgeTowerForContrastiveLearning''',
'''BridgeTowerForImageAndTextRetrieval''',
'''BridgeTowerForMaskedLM''',
'''BridgeTowerModel''',
'''BridgeTowerPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_bridgetower import (
BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP,
BridgeTowerConfig,
BridgeTowerTextConfig,
BridgeTowerVisionConfig,
)
from .processing_bridgetower import BridgeTowerProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_bridgetower import BridgeTowerImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_bridgetower import (
BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST,
BridgeTowerForContrastiveLearning,
BridgeTowerForImageAndTextRetrieval,
BridgeTowerForMaskedLM,
BridgeTowerModel,
BridgeTowerPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure)
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[str] ):
__a : Any = ''
for i in table:
res += inp[i - 1]
return res
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
return data[1:] + data[0]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] ):
__a : Optional[int] = ''
for i in range(len(lowerCamelCase_ ) ):
if a[i] == b[i]:
res += "0"
else:
res += "1"
return res
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : str ):
__a : List[str] = int('0b' + data[0] + data[-1] , 2 )
__a : List[str] = int('0b' + data[1:3] , 2 )
return bin(s[row][col] )[2:]
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : List[str] , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] ):
__a : List[Any] = message[:4]
__a : str = message[4:]
__a : Any = apply_table(lowerCamelCase_ , lowerCamelCase_ )
__a : int = xor(lowerCamelCase_ , lowerCamelCase_ )
__a : Dict = apply_sbox(lowerCamelCase_ , temp[:4] ) # noqa: E741
__a : Tuple = apply_sbox(lowerCamelCase_ , temp[4:] )
__a : List[Any] = '0' * (2 - len(lowerCamelCase_ )) + l # noqa: E741
__a : List[str] = '0' * (2 - len(lowerCamelCase_ )) + r
__a : List[Any] = apply_table(l + r , lowerCamelCase_ )
__a : Dict = xor(lowerCamelCase_ , lowerCamelCase_ )
return temp + right
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter 10 bit key: ''')
SCREAMING_SNAKE_CASE__ = input('''Enter 8 bit message: ''')
SCREAMING_SNAKE_CASE__ = [6, 3, 7, 4, 8, 5, 10, 9]
SCREAMING_SNAKE_CASE__ = [3, 5, 2, 7, 4, 10, 1, 9, 8, 6]
SCREAMING_SNAKE_CASE__ = [2, 4, 3, 1]
SCREAMING_SNAKE_CASE__ = [2, 6, 3, 1, 4, 8, 5, 7]
SCREAMING_SNAKE_CASE__ = [4, 1, 3, 5, 7, 2, 8, 6]
SCREAMING_SNAKE_CASE__ = [4, 1, 2, 3, 2, 3, 4, 1]
SCREAMING_SNAKE_CASE__ = [[1, 0, 3, 2], [3, 2, 1, 0], [0, 2, 1, 3], [3, 1, 3, 2]]
SCREAMING_SNAKE_CASE__ = [[0, 1, 2, 3], [2, 0, 1, 3], [3, 0, 1, 0], [2, 1, 0, 3]]
# key generation
SCREAMING_SNAKE_CASE__ = apply_table(key, paa_table)
SCREAMING_SNAKE_CASE__ = temp[:5]
SCREAMING_SNAKE_CASE__ = temp[5:]
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
# encryption
SCREAMING_SNAKE_CASE__ = apply_table(message, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Cipher text is:''', CT)
# decryption
SCREAMING_SNAKE_CASE__ = apply_table(CT, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Plain text after decypting is:''', PT)
| 47 |
from string import ascii_lowercase, ascii_uppercase
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
if not sentence:
return ""
__a : Union[str, Any] = dict(zip(lowerCamelCase_ , lowerCamelCase_ ) )
return lower_to_upper.get(sentence[0] , sentence[0] ) + sentence[1:]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : float , lowerCamelCase_ : float ):
if mass < 0:
raise ValueError('The mass of a body cannot be negative' )
return 0.5 * mass * abs(lowerCamelCase_ ) * abs(lowerCamelCase_ )
if __name__ == "__main__":
import doctest
doctest.testmod(verbose=True)
| 47 |
import functools
import operator
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''asapp/sew-d-tiny-100k''': '''https://huggingface.co/asapp/sew-d-tiny-100k/resolve/main/config.json''',
# See all SEW-D models at https://huggingface.co/models?filter=sew-d
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = '''sew-d'''
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : Dict=3_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=7_6_8 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : str=3_0_7_2 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : List[str]=2_5_6 , SCREAMING_SNAKE_CASE__ : Optional[Any]=True , SCREAMING_SNAKE_CASE__ : Tuple=True , SCREAMING_SNAKE_CASE__ : List[str]=("p2c", "c2p") , SCREAMING_SNAKE_CASE__ : str="layer_norm" , SCREAMING_SNAKE_CASE__ : Tuple="gelu_python" , SCREAMING_SNAKE_CASE__ : Tuple=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.0 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.02 , SCREAMING_SNAKE_CASE__ : int=1e-7 , SCREAMING_SNAKE_CASE__ : Any=1e-5 , SCREAMING_SNAKE_CASE__ : Optional[int]="group" , SCREAMING_SNAKE_CASE__ : Optional[Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=(6_4, 1_2_8, 1_2_8, 1_2_8, 1_2_8, 2_5_6, 2_5_6, 2_5_6, 2_5_6, 5_1_2, 5_1_2, 5_1_2, 5_1_2) , SCREAMING_SNAKE_CASE__ : List[Any]=(5, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : str=(1_0, 3, 1, 3, 1, 3, 1, 3, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : Optional[int]=False , SCREAMING_SNAKE_CASE__ : Optional[int]=1_2_8 , SCREAMING_SNAKE_CASE__ : Tuple=1_6 , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[Any]=0.05 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , SCREAMING_SNAKE_CASE__ : int=0.0 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : Optional[int]=0 , SCREAMING_SNAKE_CASE__ : Optional[int]="mean" , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : List[str]=False , SCREAMING_SNAKE_CASE__ : str=2_5_6 , SCREAMING_SNAKE_CASE__ : str=0 , SCREAMING_SNAKE_CASE__ : List[Any]=1 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ , pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = hidden_size
__a : Optional[Any] = feat_extract_norm
__a : List[str] = feat_extract_activation
__a : Dict = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = list(SCREAMING_SNAKE_CASE__ )
__a : List[str] = list(SCREAMING_SNAKE_CASE__ )
__a : int = conv_bias
__a : Tuple = num_conv_pos_embeddings
__a : List[str] = num_conv_pos_embedding_groups
__a : Optional[Any] = len(self.conv_dim )
__a : Union[str, Any] = num_hidden_layers
__a : Optional[Any] = intermediate_size
__a : Union[str, Any] = squeeze_factor
__a : List[Any] = max_position_embeddings
__a : Tuple = position_buckets
__a : Optional[int] = share_att_key
__a : List[str] = relative_attention
__a : Any = norm_rel_ebd
__a : Any = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = hidden_act
__a : str = num_attention_heads
__a : Union[str, Any] = hidden_dropout
__a : Optional[int] = attention_dropout
__a : List[str] = activation_dropout
__a : int = feat_proj_dropout
__a : int = final_dropout
__a : Dict = layer_norm_eps
__a : Tuple = feature_layer_norm_eps
__a : str = initializer_range
__a : Tuple = vocab_size
if (
(len(self.conv_stride ) != self.num_feat_extract_layers)
or (len(self.conv_kernel ) != self.num_feat_extract_layers)
or (len(self.conv_dim ) != self.num_feat_extract_layers)
):
raise ValueError(
'Configuration for convolutional layers is incorrect.'
'It is required that `len(config.conv_dim)` == `len(config.conv_stride)` == `len(config.conv_kernel)`,'
f'''but is `len(config.conv_dim) = {len(self.conv_dim )}`, `len(config.conv_stride)'''
f'''= {len(self.conv_stride )}`, `len(config.conv_kernel) = {len(self.conv_kernel )}`.''' )
# fine-tuning config parameters for SpecAugment: https://arxiv.org/abs/1904.08779
__a : Tuple = apply_spec_augment
__a : Optional[Any] = mask_time_prob
__a : Any = mask_time_length
__a : List[str] = mask_time_min_masks
__a : List[str] = mask_feature_prob
__a : Tuple = mask_feature_length
__a : Any = mask_feature_min_masks
# ctc loss
__a : Optional[int] = ctc_loss_reduction
__a : List[Any] = ctc_zero_infinity
# sequence classification
__a : Dict = use_weighted_layer_sum
__a : Optional[Any] = classifier_proj_size
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return functools.reduce(operator.mul , self.conv_stride , 1 )
| 47 | 1 |
from ...utils import is_torch_available, is_transformers_available
if is_transformers_available() and is_torch_available():
from .pipeline_vq_diffusion import LearnedClassifierFreeSamplingEmbeddings, VQDiffusionPipeline
| 47 |
from __future__ import annotations
from sys import maxsize
from typing import Generic, TypeVar
SCREAMING_SNAKE_CASE__ = TypeVar('''T''')
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (position - 1) // 2
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 1
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 2
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[str] ):
'''simple docstring'''
__a : list[tuple[T, int]] = []
__a : dict[T, int] = {}
__a : int = 0
def __len__( self : Any ):
'''simple docstring'''
return self.elements
def __repr__( self : Any ):
'''simple docstring'''
return str(self.heap )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.elements == 0
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.heap.append((elem, weight) )
__a : List[Any] = self.elements
self.elements += 1
self._bubble_up(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self.elements > 1:
self._swap_nodes(0 , self.elements - 1 )
__a , __a : Union[str, Any] = self.heap.pop()
del self.position_map[elem]
self.elements -= 1
if self.elements > 0:
__a , __a : Dict = self.heap[0]
self._bubble_down(SCREAMING_SNAKE_CASE__ )
return elem
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
__a : str = (elem, weight)
if position > 0:
__a : Tuple = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = self.heap[parent_position]
if parent_weight > weight:
self._bubble_up(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
if curr_pos == 0:
return None
__a : List[str] = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : str = self.heap[curr_pos]
__a , __a : Optional[int] = self.heap[parent_position]
if parent_weight > weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_up(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : int = self.position_map[elem]
__a , __a : Optional[Any] = self.heap[curr_pos]
__a : Tuple = get_child_left_position(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = get_child_right_position(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements and child_right_position < self.elements:
__a , __a : str = self.heap[child_left_position]
__a , __a : List[str] = self.heap[child_right_position]
if child_right_weight < child_left_weight and child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements:
__a , __a : Any = self.heap[child_left_position]
if child_left_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
return None
if child_right_position < self.elements:
__a , __a : Union[str, Any] = self.heap[child_right_position]
if child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Optional[Any] = self.heap[nodea_pos][0]
__a : str = self.heap[nodea_pos][0]
__a , __a : int = (
self.heap[nodea_pos],
self.heap[nodea_pos],
)
__a : str = nodea_pos
__a : Optional[int] = nodea_pos
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[Any] ):
'''simple docstring'''
__a : dict[T, dict[T, int]] = {}
__a : int = 0
def __repr__( self : Tuple ):
'''simple docstring'''
return str(self.connections )
def __len__( self : Dict ):
'''simple docstring'''
return self.nodes
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
if node not in self.connections:
__a : Tuple = {}
self.nodes += 1
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.add_node(SCREAMING_SNAKE_CASE__ )
self.add_node(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = weight
__a : Any = weight
def UpperCAmelCase__ ( lowerCamelCase_ : GraphUndirectedWeighted[T] , ):
__a : dict[T, int] = {node: maxsize for node in graph.connections}
__a : dict[T, T | None] = {node: None for node in graph.connections}
__a : MinPriorityQueue[T] = MinPriorityQueue()
for node, weight in dist.items():
priority_queue.push(lowerCamelCase_ , lowerCamelCase_ )
if priority_queue.is_empty():
return dist, parent
# initialization
__a : Optional[int] = priority_queue.extract_min()
__a : int = 0
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : str = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Optional[int] = node
# running prim's algorithm
while not priority_queue.is_empty():
__a : Any = priority_queue.extract_min()
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : Tuple = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Dict = node
return dist, parent
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_torch_available
SCREAMING_SNAKE_CASE__ = {'''configuration_speech_encoder_decoder''': ['''SpeechEncoderDecoderConfig''']}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''SpeechEncoderDecoderModel''']
try:
if not is_flax_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''FlaxSpeechEncoderDecoderModel''']
if TYPE_CHECKING:
from .configuration_speech_encoder_decoder import SpeechEncoderDecoderConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_speech_encoder_decoder import SpeechEncoderDecoderModel
try:
if not is_flax_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_flax_speech_encoder_decoder import FlaxSpeechEncoderDecoderModel
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
from collections.abc import Sequence
from queue import Queue
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : int=None , SCREAMING_SNAKE_CASE__ : Tuple=None ):
'''simple docstring'''
__a : Tuple = start
__a : Dict = end
__a : List[str] = val
__a : List[Any] = (start + end) // 2
__a : Optional[Any] = left
__a : List[str] = right
def __repr__( self : Dict ):
'''simple docstring'''
return f'''SegmentTreeNode(start={self.start}, end={self.end}, val={self.val})'''
class _UpperCamelCase:
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : Sequence , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Tuple = collection
__a : Dict = function
if self.collection:
__a : int = self._build_tree(0 , len(SCREAMING_SNAKE_CASE__ ) - 1 )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self._update_tree(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
return self._query_range(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if start == end:
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.collection[start] )
__a : Tuple = (start + end) // 2
__a : Optional[int] = self._build_tree(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = self._build_tree(mid + 1 , SCREAMING_SNAKE_CASE__ )
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.fn(left.val , right.val ) , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if node.start == i and node.end == i:
__a : Optional[Any] = val
return
if i <= node.mid:
self._update_tree(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
self._update_tree(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = self.fn(node.left.val , node.right.val )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
if node.start == i and node.end == j:
return node.val
if i <= node.mid:
if j <= node.mid:
# range in left child tree
return self._query_range(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
# range in left child tree and right child tree
return self.fn(
self._query_range(node.left , SCREAMING_SNAKE_CASE__ , node.mid ) , self._query_range(node.right , node.mid + 1 , SCREAMING_SNAKE_CASE__ ) , )
else:
# range in right child tree
return self._query_range(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
if self.root is not None:
__a : Tuple = Queue()
queue.put(self.root )
while not queue.empty():
__a : Tuple = queue.get()
yield node
if node.left is not None:
queue.put(node.left )
if node.right is not None:
queue.put(node.right )
if __name__ == "__main__":
import operator
for fn in [operator.add, max, min]:
print('''*''' * 50)
SCREAMING_SNAKE_CASE__ = SegmentTree([2, 1, 5, 3, 4], fn)
for node in arr.traverse():
print(node)
print()
arr.update(1, 5)
for node in arr.traverse():
print(node)
print()
print(arr.query_range(3, 4)) # 7
print(arr.query_range(2, 2)) # 5
print(arr.query_range(1, 3)) # 13
print()
| 47 | 1 |
from __future__ import annotations
import math
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
if 1 < number < 4:
# 2 and 3 are primes
return True
elif number < 2 or number % 2 == 0 or number % 3 == 0:
# Negatives, 0, 1, all even numbers, all multiples of 3 are not primes
return False
# All primes number are in format of 6k +/- 1
for i in range(5 , int(math.sqrt(lowerCamelCase_ ) + 1 ) , 6 ):
if number % i == 0 or number % (i + 2) == 0:
return False
return True
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
__a : Optional[int] = str(lowerCamelCase_ )
__a : Union[str, Any] = [n]
for i in range(1 , len(lowerCamelCase_ ) ):
list_nums.append(int(str_num[i:] ) )
list_nums.append(int(str_num[:-i] ) )
return list_nums
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
if len(str(lowerCamelCase_ ) ) > 3:
if not is_prime(int(str(lowerCamelCase_ )[-3:] ) ) or not is_prime(int(str(lowerCamelCase_ )[:3] ) ):
return False
return True
def UpperCAmelCase__ ( lowerCamelCase_ : int = 1_1 ):
__a : list[int] = []
__a : Union[str, Any] = 1_3
while len(lowerCamelCase_ ) != count:
if validate(lowerCamelCase_ ):
__a : Union[str, Any] = list_truncated_nums(lowerCamelCase_ )
if all(is_prime(lowerCamelCase_ ) for i in list_nums ):
list_truncated_primes.append(lowerCamelCase_ )
num += 2
return list_truncated_primes
def UpperCAmelCase__ ( ):
return sum(compute_truncated_primes(1_1 ) )
if __name__ == "__main__":
print(F"{sum(compute_truncated_primes(11)) = }")
| 47 |
import os
import posixpath
import uuid
from dataclasses import dataclass
from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, Union
import numpy as np
import pyarrow as pa
import datasets
from datasets.arrow_writer import ArrowWriter, ParquetWriter
from datasets.config import MAX_SHARD_SIZE
from datasets.filesystems import (
is_remote_filesystem,
rename,
)
from datasets.iterable_dataset import _BaseExamplesIterable
from datasets.utils.py_utils import convert_file_size_to_int
SCREAMING_SNAKE_CASE__ = datasets.utils.logging.get_logger(__name__)
if TYPE_CHECKING:
import pyspark
@dataclass
class _UpperCamelCase( datasets.BuilderConfig ):
__SCREAMING_SNAKE_CASE : Optional[datasets.Features] = None
def UpperCAmelCase__ ( lowerCamelCase_ : "pyspark.sql.DataFrame" , lowerCamelCase_ : List[int] , ):
import pyspark
def generate_fn():
__a : List[Any] = df.select('*' , pyspark.sql.functions.spark_partition_id().alias('part_id' ) )
for partition_id in partition_order:
__a : Optional[int] = df_with_partition_id.select('*' ).where(f'''part_id = {partition_id}''' ).drop('part_id' )
__a : Optional[Any] = partition_df.collect()
__a : Union[str, Any] = 0
for row in rows:
yield f'''{partition_id}_{row_id}''', row.asDict()
row_id += 1
return generate_fn
class _UpperCamelCase( _BaseExamplesIterable ):
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : Dict=None , ):
'''simple docstring'''
__a : List[str] = df
__a : Tuple = partition_order or range(self.df.rdd.getNumPartitions() )
__a : List[Any] = _generate_iterable_examples(self.df , self.partition_order )
def __iter__( self : Tuple ):
'''simple docstring'''
yield from self.generate_examples_fn()
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : np.random.Generator ):
'''simple docstring'''
__a : Union[str, Any] = list(range(self.df.rdd.getNumPartitions() ) )
generator.shuffle(SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Union[str, Any] = self.split_shard_indices_by_worker(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return len(self.partition_order )
class _UpperCamelCase( datasets.DatasetBuilder ):
__SCREAMING_SNAKE_CASE : List[str] = SparkConfig
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : str = None , SCREAMING_SNAKE_CASE__ : str = None , **SCREAMING_SNAKE_CASE__ : Optional[int] , ):
'''simple docstring'''
import pyspark
__a : int = pyspark.sql.SparkSession.builder.getOrCreate()
__a : Optional[int] = df
__a : List[Any] = working_dir
super().__init__(
cache_dir=SCREAMING_SNAKE_CASE__ , config_name=str(self.df.semanticHash() ) , **SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
def create_cache_and_write_probe(SCREAMING_SNAKE_CASE__ : List[str] ):
# makedirs with exist_ok will recursively create the directory. It will not throw an error if directories
# already exist.
os.makedirs(self._cache_dir , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(self._cache_dir , 'fs_test' + uuid.uuida().hex )
# Opening the file in append mode will create a new file unless it already exists, in which case it will not
# change the file contents.
open(SCREAMING_SNAKE_CASE__ , 'a' )
return [probe_file]
if self._spark.conf.get('spark.master' , '' ).startswith('local' ):
return
# If the cluster is multi-node, make sure that the user provided a cache_dir and that it is on an NFS
# accessible to the driver.
# TODO: Stream batches to the driver using ArrowCollectSerializer instead of throwing an error.
if self._cache_dir:
__a : List[Any] = (
self._spark.sparkContext.parallelize(range(1 ) , 1 ).mapPartitions(SCREAMING_SNAKE_CASE__ ).collect()
)
if os.path.isfile(probe[0] ):
return
raise ValueError(
'When using Dataset.from_spark on a multi-node cluster, the driver and all workers should be able to access cache_dir' )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
return datasets.DatasetInfo(features=self.config.features )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : datasets.download.download_manager.DownloadManager ):
'''simple docstring'''
return [datasets.SplitGenerator(name=datasets.Split.TRAIN )]
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
import pyspark
def get_arrow_batch_size(SCREAMING_SNAKE_CASE__ : int ):
for batch in it:
yield pa.RecordBatch.from_pydict({'batch_bytes': [batch.nbytes]} )
__a : List[str] = self.df.count()
__a : Dict = df_num_rows if df_num_rows <= 1_0_0 else 1_0_0
# Approximate the size of each row (in Arrow format) by averaging over a max-100-row sample.
__a : List[str] = (
self.df.limit(SCREAMING_SNAKE_CASE__ )
.repartition(1 )
.mapInArrow(SCREAMING_SNAKE_CASE__ , 'batch_bytes: long' )
.agg(pyspark.sql.functions.sum('batch_bytes' ).alias('sample_bytes' ) )
.collect()[0]
.sample_bytes
/ sample_num_rows
)
__a : Dict = approx_bytes_per_row * df_num_rows
if approx_total_size > max_shard_size:
# Make sure there is at least one row per partition.
__a : Union[str, Any] = min(SCREAMING_SNAKE_CASE__ , int(approx_total_size / max_shard_size ) )
__a : int = self.df.repartition(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : int , ):
'''simple docstring'''
import pyspark
__a : Any = ParquetWriter if file_format == 'parquet' else ArrowWriter
__a : Union[str, Any] = os.path.join(self._working_dir , os.path.basename(SCREAMING_SNAKE_CASE__ ) ) if self._working_dir else fpath
__a : Optional[int] = file_format == 'parquet'
# Define these so that we don't reference self in write_arrow, which will result in a pickling error due to
# pickling the SparkContext.
__a : List[str] = self.config.features
__a : int = self._writer_batch_size
__a : Union[str, Any] = self._fs.storage_options
def write_arrow(SCREAMING_SNAKE_CASE__ : Optional[int] ):
# Within the same SparkContext, no two task attempts will share the same attempt ID.
__a : Any = pyspark.TaskContext().taskAttemptId()
__a : str = next(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
if first_batch is None:
# Some partitions might not receive any data.
return pa.RecordBatch.from_arrays(
[[task_id], [0], [0]] , names=['task_id', 'num_examples', 'num_bytes'] , )
__a : Any = 0
__a : List[str] = writer_class(
features=SCREAMING_SNAKE_CASE__ , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Optional[Any] = pa.Table.from_batches([first_batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
for batch in it:
if max_shard_size is not None and writer._num_bytes >= max_shard_size:
__a , __a : Optional[int] = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
shard_id += 1
__a : Optional[Any] = writer_class(
features=writer._features , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Union[str, Any] = pa.Table.from_batches([batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
if writer._num_bytes > 0:
__a , __a : str = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
if working_fpath != fpath:
for file in os.listdir(os.path.dirname(SCREAMING_SNAKE_CASE__ ) ):
__a : Any = os.path.join(os.path.dirname(SCREAMING_SNAKE_CASE__ ) , os.path.basename(SCREAMING_SNAKE_CASE__ ) )
shutil.move(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Dict = (
self.df.mapInArrow(SCREAMING_SNAKE_CASE__ , 'task_id: long, num_examples: long, num_bytes: long' )
.groupBy('task_id' )
.agg(
pyspark.sql.functions.sum('num_examples' ).alias('total_num_examples' ) , pyspark.sql.functions.sum('num_bytes' ).alias('total_num_bytes' ) , pyspark.sql.functions.count('num_bytes' ).alias('num_shards' ) , pyspark.sql.functions.collect_list('num_examples' ).alias('shard_lengths' ) , )
.collect()
)
for row in stats:
yield row.task_id, (row.total_num_examples, row.total_num_bytes, row.num_shards, row.shard_lengths)
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , SCREAMING_SNAKE_CASE__ : str = "arrow" , SCREAMING_SNAKE_CASE__ : Optional[Union[str, int]] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
self._validate_cache_dir()
__a : List[str] = convert_file_size_to_int(max_shard_size or MAX_SHARD_SIZE )
self._repartition_df_if_needed(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = not is_remote_filesystem(self._fs )
__a : Optional[Any] = os.path.join if is_local else posixpath.join
__a : Any = '-TTTTT-SSSSS-of-NNNNN'
__a : Union[str, Any] = f'''{self.name}-{split_generator.name}{SUFFIX}.{file_format}'''
__a : Any = path_join(self._output_dir , SCREAMING_SNAKE_CASE__ )
__a : Any = 0
__a : Dict = 0
__a : int = 0
__a : List[str] = []
__a : Optional[int] = []
for task_id, content in self._prepare_split_single(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : Optional[int] = content
if num_bytes > 0:
total_num_examples += num_examples
total_num_bytes += num_bytes
total_shards += num_shards
task_id_and_num_shards.append((task_id, num_shards) )
all_shard_lengths.extend(SCREAMING_SNAKE_CASE__ )
__a : List[str] = total_num_examples
__a : Optional[int] = total_num_bytes
# should rename everything at the end
logger.debug(f'''Renaming {total_shards} shards.''' )
if total_shards > 1:
__a : Any = all_shard_lengths
# Define fs outside of _rename_shard so that we don't reference self in the function, which will result in a
# pickling error due to pickling the SparkContext.
__a : Dict = self._fs
# use the -SSSSS-of-NNNNN pattern
def _rename_shard(
SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , ):
rename(
SCREAMING_SNAKE_CASE__ , fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace('TTTTT-SSSSS' , f'''{global_shard_id:05d}''' ).replace('NNNNN' , f'''{total_shards:05d}''' ) , )
__a : Union[str, Any] = []
__a : List[str] = 0
for i in range(len(SCREAMING_SNAKE_CASE__ ) ):
__a , __a : Union[str, Any] = task_id_and_num_shards[i]
for shard_id in range(SCREAMING_SNAKE_CASE__ ):
args.append([task_id, shard_id, global_shard_id] )
global_shard_id += 1
self._spark.sparkContext.parallelize(SCREAMING_SNAKE_CASE__ , len(SCREAMING_SNAKE_CASE__ ) ).map(lambda SCREAMING_SNAKE_CASE__ : _rename_shard(*SCREAMING_SNAKE_CASE__ ) ).collect()
else:
# don't use any pattern
__a : List[Any] = 0
__a : Any = task_id_and_num_shards[0][0]
self._rename(
fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace(SCREAMING_SNAKE_CASE__ , '' ) , )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , ):
'''simple docstring'''
return SparkExamplesIterable(self.df )
| 47 | 1 |
from __future__ import annotations
import bisect
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : int , lowerCamelCase_ : int = 0 , lowerCamelCase_ : int = -1 ):
if hi < 0:
__a : Optional[Any] = len(lowerCamelCase_ )
while lo < hi:
__a : str = lo + (hi - lo) // 2
if sorted_collection[mid] < item:
__a : List[Any] = mid + 1
else:
__a : Union[str, Any] = mid
return lo
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : int , lowerCamelCase_ : int = 0 , lowerCamelCase_ : int = -1 ):
if hi < 0:
__a : Union[str, Any] = len(lowerCamelCase_ )
while lo < hi:
__a : List[str] = lo + (hi - lo) // 2
if sorted_collection[mid] <= item:
__a : List[Any] = mid + 1
else:
__a : Optional[int] = mid
return lo
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : int , lowerCamelCase_ : int = 0 , lowerCamelCase_ : int = -1 ):
sorted_collection.insert(bisect_left(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ ) , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : int , lowerCamelCase_ : int = 0 , lowerCamelCase_ : int = -1 ):
sorted_collection.insert(bisect_right(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ ) , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : int ):
__a : str = 0
__a : Union[str, Any] = len(lowerCamelCase_ ) - 1
while left <= right:
__a : List[Any] = left + (right - left) // 2
__a : Union[str, Any] = sorted_collection[midpoint]
if current_item == item:
return midpoint
elif item < current_item:
__a : List[str] = midpoint - 1
else:
__a : int = midpoint + 1
return None
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : int ):
__a : List[str] = bisect.bisect_left(lowerCamelCase_ , lowerCamelCase_ )
if index != len(lowerCamelCase_ ) and sorted_collection[index] == item:
return index
return None
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : int ):
if right < left:
return None
__a : Optional[Any] = left + (right - left) // 2
if sorted_collection[midpoint] == item:
return midpoint
elif sorted_collection[midpoint] > item:
return binary_search_by_recursion(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , midpoint - 1 )
else:
return binary_search_by_recursion(lowerCamelCase_ , lowerCamelCase_ , midpoint + 1 , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter numbers separated by comma:\n''').strip()
SCREAMING_SNAKE_CASE__ = sorted(int(item) for item in user_input.split(''','''))
SCREAMING_SNAKE_CASE__ = int(input('''Enter a single number to be found in the list:\n'''))
SCREAMING_SNAKE_CASE__ = binary_search(collection, target)
if result is None:
print(F"{target} was not found in {collection}.")
else:
print(F"{target} was found at position {result} in {collection}.")
| 47 |
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, TensorDataset
from tqdm import tqdm
from transformers import GPTaLMHeadModel
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : int ):
# save results
if os.path.exists(lowerCamelCase_ ):
if os.path.exists(os.path.join(lowerCamelCase_ , 'config.json' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'config.json' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'config.json' ) )
if os.path.exists(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) )
else:
os.makedirs(lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Any=False ):
__a : Dict = 2
if unlogit:
__a : Optional[Any] = torch.pow(lowerCamelCase_ , lowerCamelCase_ )
__a : Any = p * torch.log(lowerCamelCase_ )
__a : Union[str, Any] = 0
return -plogp.sum(dim=-1 )
def UpperCAmelCase__ ( lowerCamelCase_ : Any ):
logger.info('lv, h >\t' + '\t'.join(f'''{x + 1}''' for x in range(len(lowerCamelCase_ ) ) ) )
for row in range(len(lowerCamelCase_ ) ):
if tensor.dtype != torch.long:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:.5f}''' for x in tensor[row].cpu().data ) )
else:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:d}''' for x in tensor[row].cpu().data ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Any , lowerCamelCase_ : int , lowerCamelCase_ : int=True , lowerCamelCase_ : Optional[Any]=True , lowerCamelCase_ : List[Any]=None , lowerCamelCase_ : List[Any]=False ):
__a , __a : Optional[int] = model.config.num_hidden_layers, model.config.num_attention_heads
__a : str = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
__a : int = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
if head_mask is None:
__a : Union[str, Any] = torch.ones(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
head_mask.requires_grad_(requires_grad=lowerCamelCase_ )
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
__a : Any = None
__a : Optional[int] = 0.0
__a : Optional[Any] = 0.0
for step, inputs in enumerate(tqdm(lowerCamelCase_ , desc='Iteration' , disable=args.local_rank not in [-1, 0] ) ):
__a : Dict = tuple(t.to(args.device ) for t in inputs )
((__a) , ) : Dict = inputs
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
__a : List[Any] = model(lowerCamelCase_ , labels=lowerCamelCase_ , head_mask=lowerCamelCase_ )
# (loss), lm_logits, presents, (all hidden_states), (attentions)
__a , __a , __a : int = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
total_loss += loss.detach().cpu().numpy()
if compute_entropy:
for layer, attn in enumerate(lowerCamelCase_ ):
__a : List[str] = entropy(attn.detach() , lowerCamelCase_ )
attn_entropy[layer] += masked_entropy.sum(-1 ).sum(0 ).sum(0 ).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
tot_tokens += torch.ones_like(lowerCamelCase_ ).float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
__a : Optional[Any] = 2
__a : Union[str, Any] = torch.pow(torch.pow(lowerCamelCase_ , lowerCamelCase_ ).sum(-1 ) , 1 / exponent )
head_importance /= norm_by_layer.unsqueeze(-1 ) + 1e-20
if not args.dont_normalize_global_importance:
__a : List[str] = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print matrices
if compute_entropy:
logger.info('Attention entropies' )
print_ad_tensor(lowerCamelCase_ )
if compute_importance:
logger.info('Head importance scores' )
print_ad_tensor(lowerCamelCase_ )
logger.info('Head ranked by importance scores' )
__a : Optional[Any] = torch.zeros(head_importance.numel() , dtype=torch.long , device=args.device )
__a : str = torch.arange(
head_importance.numel() , device=args.device )
__a : Tuple = head_ranks.view_as(lowerCamelCase_ )
print_ad_tensor(lowerCamelCase_ )
return attn_entropy, head_importance, total_loss
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Tuple , lowerCamelCase_ : int ):
__a , __a , __a : Optional[int] = compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ )
__a : Tuple = 1 / loss # instead of downsteam score use the LM loss
logger.info('Pruning: original score: %f, threshold: %f' , lowerCamelCase_ , original_score * args.masking_threshold )
__a : Tuple = torch.ones_like(lowerCamelCase_ )
__a : int = max(1 , int(new_head_mask.numel() * args.masking_amount ) )
__a : Tuple = original_score
while current_score >= original_score * args.masking_threshold:
__a : Optional[Any] = new_head_mask.clone().detach() # save current head mask
# heads from least important to most - keep only not-masked heads
__a : List[str] = float('Inf' )
__a : List[Any] = head_importance.view(-1 ).sort()[1]
if len(lowerCamelCase_ ) <= num_to_mask:
print('BREAK BY num_to_mask' )
break
# mask heads
__a : Any = current_heads_to_mask[:num_to_mask]
logger.info('Heads to mask: %s' , str(current_heads_to_mask.tolist() ) )
__a : int = new_head_mask.view(-1 )
__a : Tuple = 0.0
__a : int = new_head_mask.view_as(lowerCamelCase_ )
__a : Optional[int] = new_head_mask.clone().detach()
print_ad_tensor(lowerCamelCase_ )
# Compute metric and head importance again
__a , __a , __a : int = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[Any] = 1 / loss
logger.info(
'Masking: current score: %f, remaining heads %d (%.1f percents)' , lowerCamelCase_ , new_head_mask.sum() , new_head_mask.sum() / new_head_mask.numel() * 1_0_0 , )
logger.info('Final head mask' )
print_ad_tensor(lowerCamelCase_ )
np.save(os.path.join(args.output_dir , 'head_mask.npy' ) , head_mask.detach().cpu().numpy() )
return head_mask
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Union[str, Any] ):
__a : List[Any] = datetime.now()
__a , __a , __a : List[str] = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[str] = 1 / loss
__a : List[Any] = datetime.now() - before_time
__a : List[str] = sum(p.numel() for p in model.parameters() )
__a : Dict = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(lowerCamelCase_ ) )
}
for k, v in heads_to_prune.items():
if isinstance(lowerCamelCase_ , lowerCamelCase_ ):
__a : Tuple = [
v,
]
assert sum(len(lowerCamelCase_ ) for h in heads_to_prune.values() ) == (1 - head_mask.long()).sum().item()
model.prune_heads(lowerCamelCase_ )
__a : Optional[Any] = sum(p.numel() for p in model.parameters() )
__a : Tuple = datetime.now()
__a , __a , __a : Tuple = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ , actually_pruned=lowerCamelCase_ , )
__a : Optional[Any] = 1 / loss
__a : List[Any] = datetime.now() - before_time
logger.info(
'Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)' , lowerCamelCase_ , lowerCamelCase_ , pruned_num_params / original_num_params * 1_0_0 , )
logger.info('Pruning: score with masking: %f score with pruning: %f' , lowerCamelCase_ , lowerCamelCase_ )
logger.info('Pruning: speed ratio (original timing / new timing): %f percents' , original_time / new_time * 1_0_0 )
save_model(lowerCamelCase_ , args.output_dir )
def UpperCAmelCase__ ( ):
__a : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'--data_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The input data dir. Should contain the .tsv files (or other data files) for the task.' , )
parser.add_argument(
'--model_name_or_path' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='Path to pretrained model or model identifier from huggingface.co/models' , )
parser.add_argument(
'--output_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
# Other parameters
parser.add_argument(
'--config_name' , default='' , type=lowerCamelCase_ , help='Pretrained config name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--tokenizer_name' , default='' , type=lowerCamelCase_ , help='Pretrained tokenizer name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--cache_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , help='Where do you want to store the pre-trained models downloaded from s3' , )
parser.add_argument(
'--data_subset' , type=lowerCamelCase_ , default=-1 , help='If > 0: limit the data to a subset of data_subset instances.' )
parser.add_argument(
'--overwrite_output_dir' , action='store_true' , help='Whether to overwrite data in output directory' )
parser.add_argument(
'--overwrite_cache' , action='store_true' , help='Overwrite the cached training and evaluation sets' )
parser.add_argument(
'--dont_normalize_importance_by_layer' , action='store_true' , help='Don\'t normalize importance score by layers' )
parser.add_argument(
'--dont_normalize_global_importance' , action='store_true' , help='Don\'t normalize all importance scores between 0 and 1' , )
parser.add_argument(
'--try_masking' , action='store_true' , help='Whether to try to mask head until a threshold of accuracy.' )
parser.add_argument(
'--masking_threshold' , default=0.9 , type=lowerCamelCase_ , help='masking threshold in term of metrics (stop masking when metric < threshold * original metric value).' , )
parser.add_argument(
'--masking_amount' , default=0.1 , type=lowerCamelCase_ , help='Amount to heads to masking at each masking step.' )
parser.add_argument('--metric_name' , default='acc' , type=lowerCamelCase_ , help='Metric to use for head masking.' )
parser.add_argument(
'--max_seq_length' , default=1_2_8 , type=lowerCamelCase_ , help=(
'The maximum total input sequence length after WordPiece tokenization. \n'
'Sequences longer than this will be truncated, sequences shorter padded.'
) , )
parser.add_argument('--batch_size' , default=1 , type=lowerCamelCase_ , help='Batch size.' )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 )
parser.add_argument('--local_rank' , type=lowerCamelCase_ , default=-1 , help='local_rank for distributed training on gpus' )
parser.add_argument('--no_cuda' , action='store_true' , help='Whether not to use CUDA when available' )
parser.add_argument('--server_ip' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
parser.add_argument('--server_port' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
__a : Optional[Any] = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print('Waiting for debugger attach' )
ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=lowerCamelCase_ )
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
__a : List[str] = torch.device('cuda' if torch.cuda.is_available() and not args.no_cuda else 'cpu' )
__a : Tuple = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank )
__a : Union[str, Any] = torch.device('cuda' , args.local_rank )
__a : Any = 1
torch.distributed.init_process_group(backend='nccl' ) # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN )
logger.info('device: {} n_gpu: {}, distributed: {}'.format(args.device , args.n_gpu , bool(args.local_rank != -1 ) ) )
__a : Optional[Any] = GPTaLMHeadModel.from_pretrained(args.model_name_or_path )
# Distributed and parallel training
model.to(args.device )
if args.local_rank != -1:
__a : List[Any] = nn.parallel.DistributedDataParallel(
lowerCamelCase_ , device_ids=[args.local_rank] , output_device=args.local_rank , find_unused_parameters=lowerCamelCase_ )
elif args.n_gpu > 1:
__a : Union[str, Any] = nn.DataParallel(lowerCamelCase_ )
# Print/save training arguments
os.makedirs(args.output_dir , exist_ok=lowerCamelCase_ )
torch.save(lowerCamelCase_ , os.path.join(args.output_dir , 'run_args.bin' ) )
logger.info('Training/evaluation parameters %s' , lowerCamelCase_ )
# Prepare dataset
__a : Tuple = np.concatenate(
[
np.loadtxt(args.data_dir , dtype=np.intaa ),
] )
__a : str = (torch.from_numpy(lowerCamelCase_ ),)
__a : List[str] = TensorDataset(*lowerCamelCase_ )
__a : Optional[Any] = RandomSampler(lowerCamelCase_ )
__a : Union[str, Any] = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.batch_size )
# Compute head entropy and importance score
compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
__a : Union[str, Any] = mask_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
prune_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 | 1 |
import math
import tensorflow as tf
from packaging import version
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple ):
__a : Any = tf.convert_to_tensor(lowerCamelCase_ )
__a : Any = 0.5 * (1.0 + tf.math.erf(x / tf.cast(tf.sqrt(2.0 ) , x.dtype ) ))
return x * cdf
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] ):
__a : Optional[Any] = tf.convert_to_tensor(lowerCamelCase_ )
__a : Union[str, Any] = tf.cast(math.pi , x.dtype )
__a : str = tf.cast(0.044715 , x.dtype )
__a : int = 0.5 * (1.0 + tf.tanh(tf.sqrt(2.0 / pi ) * (x + coeff * tf.pow(lowerCamelCase_ , 3 )) ))
return x * cdf
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple ):
__a : Union[str, Any] = tf.convert_to_tensor(lowerCamelCase_ )
return x * tf.tanh(tf.math.softplus(lowerCamelCase_ ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] ):
__a : List[str] = tf.convert_to_tensor(lowerCamelCase_ )
__a : int = tf.cast(0.044715 , x.dtype )
__a : List[Any] = tf.cast(0.7978845608 , x.dtype )
return 0.5 * x * (1.0 + tf.tanh(x * coeffa * (1.0 + coeffa * x * x) ))
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
__a : Dict = tf.convert_to_tensor(lowerCamelCase_ )
__a : int = tf.cast(1.702 , x.dtype )
return x * tf.math.sigmoid(coeff * x )
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] ):
return tf.clip_by_value(_gelu(lowerCamelCase_ ) , -1_0 , 1_0 )
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : List[Any]=-1 ):
__a , __a : Optional[Any] = tf.split(lowerCamelCase_ , 2 , axis=lowerCamelCase_ )
return a * tf.math.sigmoid(lowerCamelCase_ )
if version.parse(tf.version.VERSION) >= version.parse('''2.4'''):
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return tf.keras.activations.gelu(lowerCamelCase_ , approximate=lowerCamelCase_ )
SCREAMING_SNAKE_CASE__ = tf.keras.activations.gelu
SCREAMING_SNAKE_CASE__ = approximate_gelu_wrap
else:
SCREAMING_SNAKE_CASE__ = _gelu
SCREAMING_SNAKE_CASE__ = _gelu_new
SCREAMING_SNAKE_CASE__ = {
'''gelu''': gelu,
'''gelu_10''': gelu_aa,
'''gelu_fast''': gelu_fast,
'''gelu_new''': gelu_new,
'''glu''': glu,
'''mish''': mish,
'''quick_gelu''': quick_gelu,
'''relu''': tf.keras.activations.relu,
'''sigmoid''': tf.keras.activations.sigmoid,
'''silu''': tf.keras.activations.swish,
'''swish''': tf.keras.activations.swish,
'''tanh''': tf.keras.activations.tanh,
}
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
if activation_string in ACTaFN:
return ACTaFN[activation_string]
else:
raise KeyError(f'''function {activation_string} not found in ACT2FN mapping {list(ACTaFN.keys() )}''' )
| 47 |
import argparse
import os
import gluonnlp as nlp
import mxnet as mx
import numpy as np
import torch
from gluonnlp.base import get_home_dir
from gluonnlp.model.bert import BERTEncoder
from gluonnlp.model.utils import _load_vocab
from gluonnlp.vocab import Vocab
from packaging import version
from torch import nn
from transformers import BertConfig, BertForMaskedLM, BertModel, RobertaTokenizer
from transformers.models.bert.modeling_bert import (
BertIntermediate,
BertLayer,
BertOutput,
BertSelfAttention,
BertSelfOutput,
)
from transformers.utils import logging
if version.parse(nlp.__version__) != version.parse('''0.8.3'''):
raise Exception('''requires gluonnlp == 0.8.3''')
if version.parse(mx.__version__) != version.parse('''1.5.0'''):
raise Exception('''requires mxnet == 1.5.0''')
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''The Nymphenburg Palace is a beautiful palace in Munich!'''
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
__a : List[Any] = {
'attention_cell': 'multi_head',
'num_layers': 4,
'units': 1_0_2_4,
'hidden_size': 7_6_8,
'max_length': 5_1_2,
'num_heads': 8,
'scaled': True,
'dropout': 0.1,
'use_residual': True,
'embed_size': 1_0_2_4,
'embed_dropout': 0.1,
'word_embed': None,
'layer_norm_eps': 1e-5,
'token_type_vocab_size': 2,
}
__a : Optional[int] = bort_4_8_768_1024_hparams
# Let's construct the original Bort model here
# Taken from official BERT implementation, see:
# https://github.com/alexa/bort/blob/master/bort/bort.py
__a : List[str] = BERTEncoder(
attention_cell=predefined_args['attention_cell'] , num_layers=predefined_args['num_layers'] , units=predefined_args['units'] , hidden_size=predefined_args['hidden_size'] , max_length=predefined_args['max_length'] , num_heads=predefined_args['num_heads'] , scaled=predefined_args['scaled'] , dropout=predefined_args['dropout'] , output_attention=lowerCamelCase_ , output_all_encodings=lowerCamelCase_ , use_residual=predefined_args['use_residual'] , activation=predefined_args.get('activation' , 'gelu' ) , layer_norm_eps=predefined_args.get('layer_norm_eps' , lowerCamelCase_ ) , )
# Vocab information needs to be fetched first
# It's the same as RoBERTa, so RobertaTokenizer can be used later
__a : int = 'openwebtext_ccnews_stories_books_cased'
# Specify download folder to Gluonnlp's vocab
__a : Optional[Any] = os.path.join(get_home_dir() , 'models' )
__a : Optional[Any] = _load_vocab(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , cls=lowerCamelCase_ )
__a : Any = nlp.model.BERTModel(
lowerCamelCase_ , len(lowerCamelCase_ ) , units=predefined_args['units'] , embed_size=predefined_args['embed_size'] , embed_dropout=predefined_args['embed_dropout'] , word_embed=predefined_args['word_embed'] , use_pooler=lowerCamelCase_ , use_token_type_embed=lowerCamelCase_ , token_type_vocab_size=predefined_args['token_type_vocab_size'] , use_classifier=lowerCamelCase_ , use_decoder=lowerCamelCase_ , )
original_bort.load_parameters(lowerCamelCase_ , cast_dtype=lowerCamelCase_ , ignore_extra=lowerCamelCase_ )
__a : Dict = original_bort._collect_params_with_prefix()
# Build our config 🤗
__a : Optional[Any] = {
'architectures': ['BertForMaskedLM'],
'attention_probs_dropout_prob': predefined_args['dropout'],
'hidden_act': 'gelu',
'hidden_dropout_prob': predefined_args['dropout'],
'hidden_size': predefined_args['embed_size'],
'initializer_range': 0.02,
'intermediate_size': predefined_args['hidden_size'],
'layer_norm_eps': predefined_args['layer_norm_eps'],
'max_position_embeddings': predefined_args['max_length'],
'model_type': 'bort',
'num_attention_heads': predefined_args['num_heads'],
'num_hidden_layers': predefined_args['num_layers'],
'pad_token_id': 1, # 2 = BERT, 1 = RoBERTa
'type_vocab_size': 1, # 2 = BERT, 1 = RoBERTa
'vocab_size': len(lowerCamelCase_ ),
}
__a : str = BertConfig.from_dict(lowerCamelCase_ )
__a : Optional[int] = BertForMaskedLM(lowerCamelCase_ )
hf_bort_model.eval()
# Parameter mapping table (Gluonnlp to Transformers)
# * denotes layer index
#
# | Gluon Parameter | Transformers Parameter
# | -------------------------------------------------------------- | ----------------------
# | `encoder.layer_norm.beta` | `bert.embeddings.LayerNorm.bias`
# | `encoder.layer_norm.gamma` | `bert.embeddings.LayerNorm.weight`
# | `encoder.position_weight` | `bert.embeddings.position_embeddings.weight`
# | `word_embed.0.weight` | `bert.embeddings.word_embeddings.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_key.bias` | `bert.encoder.layer.*.attention.self.key.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_key.weight` | `bert.encoder.layer.*.attention.self.key.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_query.bias` | `bert.encoder.layer.*.attention.self.query.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_query.weight` | `bert.encoder.layer.*.attention.self.query.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_value.bias` | `bert.encoder.layer.*.attention.self.value.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_value.weight` | `bert.encoder.layer.*.attention.self.value.weight`
# | `encoder.transformer_cells.*.ffn.ffn_2.bias` | `bert.encoder.layer.*.attention.output.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_2.weight` | `bert.encoder.layer.*.attention.output.dense.weight`
# | `encoder.transformer_cells.*.layer_norm.beta` | `bert.encoder.layer.*.attention.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.layer_norm.gamma` | `bert.encoder.layer.*.attention.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.ffn.ffn_1.bias` | `bert.encoder.layer.*.intermediate.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_1.weight` | `bert.encoder.layer.*.intermediate.dense.weight`
# | `encoder.transformer_cells.*.ffn.layer_norm.beta` | `bert.encoder.layer.*.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.ffn.layer_norm.gamma` | `bert.encoder.layer.*.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.proj.bias` | `bert.encoder.layer.*.output.dense.bias`
# | `encoder.transformer_cells.*.proj.weight` | `bert.encoder.layer.*.output.dense.weight`
# Helper function to convert MXNET Arrays to PyTorch
def to_torch(lowerCamelCase_ : Optional[Any] ) -> nn.Parameter:
return nn.Parameter(torch.FloatTensor(mx_array.data().asnumpy() ) )
# Check param shapes and map new HF param back
def check_and_map_params(lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[str] ):
__a : Optional[int] = hf_param.shape
__a : int = to_torch(params[gluon_param] )
__a : int = gluon_param.shape
assert (
shape_hf == shape_gluon
), f'''The gluon parameter {gluon_param} has shape {shape_gluon}, but expects shape {shape_hf} for Transformers'''
return gluon_param
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.word_embeddings.weight , 'word_embed.0.weight' )
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.position_embeddings.weight , 'encoder.position_weight' )
__a : Tuple = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.bias , 'encoder.layer_norm.beta' )
__a : Union[str, Any] = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.weight , 'encoder.layer_norm.gamma' )
# Inspired by RoBERTa conversion script, we just zero them out (Bort does not use them)
__a : Union[str, Any] = torch.zeros_like(
hf_bort_model.bert.embeddings.token_type_embeddings.weight.data )
for i in range(hf_bort_config.num_hidden_layers ):
__a : BertLayer = hf_bort_model.bert.encoder.layer[i]
# self attention
__a : BertSelfAttention = layer.attention.self
__a : Optional[int] = check_and_map_params(
self_attn.key.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.bias''' )
__a : str = check_and_map_params(
self_attn.key.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.weight''' )
__a : List[str] = check_and_map_params(
self_attn.query.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.bias''' )
__a : str = check_and_map_params(
self_attn.query.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.weight''' )
__a : Dict = check_and_map_params(
self_attn.value.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.bias''' )
__a : str = check_and_map_params(
self_attn.value.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.weight''' )
# self attention output
__a : BertSelfOutput = layer.attention.output
__a : Tuple = check_and_map_params(
self_output.dense.bias , f'''encoder.transformer_cells.{i}.proj.bias''' )
__a : Dict = check_and_map_params(
self_output.dense.weight , f'''encoder.transformer_cells.{i}.proj.weight''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.layer_norm.beta''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.layer_norm.gamma''' )
# intermediate
__a : BertIntermediate = layer.intermediate
__a : List[str] = check_and_map_params(
intermediate.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_1.bias''' )
__a : Optional[Any] = check_and_map_params(
intermediate.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_1.weight''' )
# output
__a : BertOutput = layer.output
__a : str = check_and_map_params(
bert_output.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_2.bias''' )
__a : List[Any] = check_and_map_params(
bert_output.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_2.weight''' )
__a : str = check_and_map_params(
bert_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.ffn.layer_norm.beta''' )
__a : List[str] = check_and_map_params(
bert_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.ffn.layer_norm.gamma''' )
# Save space and energy 🎄
hf_bort_model.half()
# Compare output of both models
__a : Union[str, Any] = RobertaTokenizer.from_pretrained('roberta-base' )
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ )['input_ids']
# Get gluon output
__a : Optional[int] = mx.nd.array([input_ids] )
__a : Tuple = original_bort(inputs=lowerCamelCase_ , token_types=[] )
# Get Transformer output (save and reload model again)
hf_bort_model.save_pretrained(lowerCamelCase_ )
__a : Optional[Any] = BertModel.from_pretrained(lowerCamelCase_ )
hf_bort_model.eval()
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ , return_tensors='pt' )
__a : int = hf_bort_model(**lowerCamelCase_ )[0]
__a : Dict = output_gluon[0].asnumpy()
__a : str = output_hf[0].detach().numpy()
__a : List[Any] = np.max(np.abs(hf_layer - gluon_layer ) ).item()
__a : str = np.allclose(lowerCamelCase_ , lowerCamelCase_ , atol=1e-3 )
if success:
print('✔️ Both model do output the same tensors' )
else:
print('❌ Both model do **NOT** output the same tensors' )
print('Absolute difference is:' , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--bort_checkpoint_path''', default=None, type=str, required=True, help='''Path the official Bort params file.'''
)
parser.add_argument(
'''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.'''
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_bort_checkpoint_to_pytorch(args.bort_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
from __future__ import annotations
import random
import unittest
from transformers import TransfoXLConfig, is_tf_available
from transformers.testing_utils import require_tf, slow
from ...test_configuration_common import ConfigTester
from ...test_modeling_tf_common import TFModelTesterMixin, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_tf_available():
import tensorflow as tf
from transformers import (
TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST,
TFTransfoXLForSequenceClassification,
TFTransfoXLLMHeadModel,
TFTransfoXLModel,
)
class _UpperCamelCase:
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : List[Any] , ):
'''simple docstring'''
__a : Optional[int] = parent
__a : int = 1_3
__a : Union[str, Any] = 7
__a : Dict = 3_0
__a : Tuple = self.seq_length + self.mem_len
__a : Optional[int] = 1_5
__a : Tuple = True
__a : Optional[Any] = True
__a : str = 9_9
__a : int = [1_0, 5_0, 8_0]
__a : Union[str, Any] = 3_2
__a : Dict = 3_2
__a : Tuple = 4
__a : Union[str, Any] = 8
__a : Optional[Any] = 1_2_8
__a : str = 2
__a : str = 2
__a : Dict = None
__a : List[str] = 1
__a : Union[str, Any] = 0
__a : Optional[Any] = 3
__a : List[str] = self.vocab_size - 1
__a : int = 0.01
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : str = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a : str = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a : Tuple = None
if self.use_labels:
__a : Tuple = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a : Optional[int] = TransfoXLConfig(
vocab_size=self.vocab_size , mem_len=self.mem_len , clamp_len=self.clamp_len , cutoffs=self.cutoffs , d_model=self.hidden_size , d_embed=self.d_embed , n_head=self.num_attention_heads , d_head=self.d_head , d_inner=self.d_inner , div_val=self.div_val , n_layer=self.num_hidden_layers , eos_token_id=self.eos_token_id , pad_token_id=self.vocab_size - 1 , init_range=self.init_range , num_labels=self.num_labels , )
return (config, input_ids_a, input_ids_a, lm_labels)
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
random.seed(self.seed )
tf.random.set_seed(self.seed )
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
__a : Optional[int] = TFTransfoXLModel(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = model(SCREAMING_SNAKE_CASE__ ).to_tuple()
__a : str = {'input_ids': input_ids_a, 'mems': mems_a}
__a , __a : Optional[Any] = model(SCREAMING_SNAKE_CASE__ ).to_tuple()
self.parent.assertEqual(hidden_states_a.shape , (self.batch_size, self.seq_length, self.hidden_size) )
self.parent.assertEqual(hidden_states_a.shape , (self.batch_size, self.seq_length, self.hidden_size) )
self.parent.assertListEqual(
[mem.shape for mem in mems_a] , [(self.mem_len, self.batch_size, self.hidden_size)] * self.num_hidden_layers , )
self.parent.assertListEqual(
[mem.shape for mem in mems_a] , [(self.mem_len, self.batch_size, self.hidden_size)] * self.num_hidden_layers , )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : str = TFTransfoXLLMHeadModel(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = model(SCREAMING_SNAKE_CASE__ ).to_tuple()
__a : List[str] = {'input_ids': input_ids_a, 'labels': lm_labels}
__a , __a : List[str] = model(SCREAMING_SNAKE_CASE__ ).to_tuple()
__a , __a : int = model([input_ids_a, mems_a] ).to_tuple()
__a : Dict = {'input_ids': input_ids_a, 'mems': mems_a, 'labels': lm_labels}
__a , __a : Tuple = model(SCREAMING_SNAKE_CASE__ ).to_tuple()
self.parent.assertEqual(lm_logits_a.shape , (self.batch_size, self.seq_length, self.vocab_size) )
self.parent.assertListEqual(
[mem.shape for mem in mems_a] , [(self.mem_len, self.batch_size, self.hidden_size)] * self.num_hidden_layers , )
self.parent.assertEqual(lm_logits_a.shape , (self.batch_size, self.seq_length, self.vocab_size) )
self.parent.assertListEqual(
[mem.shape for mem in mems_a] , [(self.mem_len, self.batch_size, self.hidden_size)] * self.num_hidden_layers , )
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Any = TFTransfoXLForSequenceClassification(SCREAMING_SNAKE_CASE__ )
__a : Tuple = model(SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : str = self.prepare_config_and_inputs()
((__a) , (__a) , (__a) , (__a)) : Union[str, Any] = config_and_inputs
__a : str = {'input_ids': input_ids_a}
return config, inputs_dict
@require_tf
class _UpperCamelCase( __lowerCamelCase , __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : List[str] = (
(TFTransfoXLModel, TFTransfoXLLMHeadModel, TFTransfoXLForSequenceClassification) if is_tf_available() else ()
)
__SCREAMING_SNAKE_CASE : int = () if is_tf_available() else ()
__SCREAMING_SNAKE_CASE : Dict = (
{
'''feature-extraction''': TFTransfoXLModel,
'''text-classification''': TFTransfoXLForSequenceClassification,
'''text-generation''': TFTransfoXLLMHeadModel,
'''zero-shot''': TFTransfoXLForSequenceClassification,
}
if is_tf_available()
else {}
)
# TODO: add this test when TFTransfoXLLMHead has a linear output layer implemented
__SCREAMING_SNAKE_CASE : List[Any] = False
__SCREAMING_SNAKE_CASE : str = False
__SCREAMING_SNAKE_CASE : Optional[Any] = False
__SCREAMING_SNAKE_CASE : Any = False
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
if pipeline_test_casse_name == "TextGenerationPipelineTests":
# Get `ValueError: AttributeError: 'NoneType' object has no attribute 'new_ones'` or `AssertionError`.
# `TransfoXLConfig` was never used in pipeline tests: cannot create a simple
# tokenizer.
return True
return False
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : List[str] = TFTransfoXLModelTester(self )
__a : List[str] = ConfigTester(self , config_class=SCREAMING_SNAKE_CASE__ , d_embed=3_7 )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
self.config_tester.run_common_tests()
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
self.model_tester.set_seed()
__a : str = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_transfo_xl_model(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
self.model_tester.set_seed()
__a : List[Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_transfo_xl_lm_head(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_transfo_xl_for_sequence_classification(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a , __a : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
__a : Any = [TFTransfoXLForSequenceClassification]
for model_class in self.all_model_classes:
__a : int = model_class(SCREAMING_SNAKE_CASE__ )
assert isinstance(model.get_input_embeddings() , tf.keras.layers.Layer )
if model_class in list_other_models_with_output_ebd:
__a : Tuple = model.get_output_embeddings()
assert isinstance(SCREAMING_SNAKE_CASE__ , tf.keras.layers.Layer )
__a : Optional[int] = model.get_bias()
assert name is None
else:
__a : Optional[Any] = model.get_output_embeddings()
assert x is None
__a : Union[str, Any] = model.get_bias()
assert name is None
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
pass
@slow
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
for model_name in TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a : Optional[int] = TFTransfoXLModel.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@unittest.skip(reason='This model doesn\'t play well with fit() due to not returning a single loss.' )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
pass
@require_tf
class _UpperCamelCase( unittest.TestCase ):
@unittest.skip('Skip test until #12651 is resolved.' )
@slow
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : str = TFTransfoXLLMHeadModel.from_pretrained('transfo-xl-wt103' )
# fmt: off
__a : Optional[int] = tf.convert_to_tensor([[3_3,1_2_9_7,2,1,1_0_0_9,4,1_1_0_9,1_1_7_3_9,4_7_6_2,3_5_8,5,2_5,2_4_5,2_2,1_7_0_6,1_7,2_0_0_9_8,5,3_2_1_5,2_1,3_7,1_1_1_0,3,1_3,1_0_4_1,4,2_4,6_0_3,4_9_0,2,7_1_4_7_7,2_0_0_9_8,1_0_4_4_4_7,2,2_0_9_6_1,1,2_6_0_4,4,1,3_2_9,3,6_2_2_4,8_3_1,1_6_0_0_2,2,8,6_0_3,7_8_9_6_7,2_9_5_4_6,2_3,8_0_3,2_0,2_5,4_1_6,5,8,2_3_2,4,2_7_7,6,1_8_5_5,4_6_0_1,3,2_9_5_4_6,5_4,8,3_6_0_9,5,5_7_2_1_1,4_9,4,1,2_7_7,1_8,8,1_7_5_5,1_5_6_9_1,3,3_4_1,2_5,4_1_6,6_9_3,4_2_5_7_3,7_1,1_7,4_0_1,9_4,3_1,1_7_9_1_9,2,2_9_5_4_6,7_8_7_3,1_8,1,4_3_5,2_3,1_1_0_1_1,7_5_5,5,5_1_6_7,3,7_9_8_3,9_8,8_4,2,2_9_5_4_6,3_2_6_7,8,3_6_0_9,4,1,4_8_6_5,1_0_7_5,2,6_0_8_7,7_1,6,3_4_6,8,5_8_5_4,3,2_9_5_4_6,8_2_4,1_4_0_0,1_8_6_8,2,1_9,1_6_0,2,3_1_1,8,5_4_9_6,2,2_0_9_2_0,1_7,2_5,1_5_0_9_7,3,2_4,2_4,0]] , dtype=tf.intaa ) # noqa: E231
# fmt: on
# In 1991 , the remains of Russian Tsar Nicholas II and his family
# ( except for Alexei and Maria ) are discovered .
# The voice of Nicholas's young son , Tsarevich Alexei Nikolaevich , narrates the
# remainder of the story . 1883 Western Siberia ,
# a young Grigori Rasputin is asked by his father and a group of men to perform magic .
# Rasputin has a vision and denounces one of the men as a horse thief . Although his
# father initially slaps him for making such an accusation , Rasputin watches as the
# man is chased outside and beaten . Twenty years later , Rasputin sees a vision of
# the Virgin Mary , prompting him to become a priest . Rasputin quickly becomes famous ,
# with people , even a bishop , begging for his blessing . <eod> </s> <eos>
# fmt: off
__a : List[Any] = [3_3,1_2_9_7,2,1,1_0_0_9,4,1_1_0_9,1_1_7_3_9,4_7_6_2,3_5_8,5,2_5,2_4_5,2_2,1_7_0_6,1_7,2_0_0_9_8,5,3_2_1_5,2_1,3_7,1_1_1_0,3,1_3,1_0_4_1,4,2_4,6_0_3,4_9_0,2,7_1_4_7_7,2_0_0_9_8,1_0_4_4_4_7,2,2_0_9_6_1,1,2_6_0_4,4,1,3_2_9,3,6_2_2_4,8_3_1,1_6_0_0_2,2,8,6_0_3,7_8_9_6_7,2_9_5_4_6,2_3,8_0_3,2_0,2_5,4_1_6,5,8,2_3_2,4,2_7_7,6,1_8_5_5,4_6_0_1,3,2_9_5_4_6,5_4,8,3_6_0_9,5,5_7_2_1_1,4_9,4,1,2_7_7,1_8,8,1_7_5_5,1_5_6_9_1,3,3_4_1,2_5,4_1_6,6_9_3,4_2_5_7_3,7_1,1_7,4_0_1,9_4,3_1,1_7_9_1_9,2,2_9_5_4_6,7_8_7_3,1_8,1,4_3_5,2_3,1_1_0_1_1,7_5_5,5,5_1_6_7,3,7_9_8_3,9_8,8_4,2,2_9_5_4_6,3_2_6_7,8,3_6_0_9,4,1,4_8_6_5,1_0_7_5,2,6_0_8_7,7_1,6,3_4_6,8,5_8_5_4,3,2_9_5_4_6,8_2_4,1_4_0_0,1_8_6_8,2,1_9,1_6_0,2,3_1_1,8,5_4_9_6,2,2_0_9_2_0,1_7,2_5,1_5_0_9_7,3,2_4,2_4,0,3_3,1,1_8_5_7,2,1,1_0_0_9,4,1_1_0_9,1_1_7_3_9,4_7_6_2,3_5_8,5,2_5,2_4_5,2_8,1_1_1_0,3,1_3,1_0_4_1,4,2_4,6_0_3,4_9_0,2,7_1_4_7_7,2_0_0_9_8,1_0_4_4_4_7,2,2_0_9_6_1,1,2_6_0_4,4,1,3_2_9,3,0] # noqa: E231
# fmt: on
# In 1991, the remains of Russian Tsar Nicholas II and his family (
# except for Alexei and Maria ) are discovered. The voice of young son,
# Tsarevich Alexei Nikolaevich, narrates the remainder of the story.
# 1883 Western Siberia, a young Grigori Rasputin is asked by his father
# and a group of men to perform magic. Rasputin has a vision and
# denounces one of the men as a horse thief. Although his father initially
# slaps him for making such an accusation, Rasputin watches as the man
# is chased outside and beaten. Twenty years later, Rasputin sees a vision
# of the Virgin Mary, prompting him to become a priest.
# Rasputin quickly becomes famous, with people, even a bishop, begging for
# his blessing. <unk> <unk> <eos> In the 1990s, the remains of Russian Tsar
# Nicholas II and his family were discovered. The voice of <unk> young son,
# Tsarevich Alexei Nikolaevich, narrates the remainder of the story.<eos>
__a : List[str] = model.generate(SCREAMING_SNAKE_CASE__ , max_length=2_0_0 , do_sample=SCREAMING_SNAKE_CASE__ )
self.assertListEqual(output_ids[0].numpy().tolist() , SCREAMING_SNAKE_CASE__ )
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[str] ):
__a : Any = ''
for i in table:
res += inp[i - 1]
return res
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
return data[1:] + data[0]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] ):
__a : Optional[int] = ''
for i in range(len(lowerCamelCase_ ) ):
if a[i] == b[i]:
res += "0"
else:
res += "1"
return res
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : str ):
__a : List[str] = int('0b' + data[0] + data[-1] , 2 )
__a : List[str] = int('0b' + data[1:3] , 2 )
return bin(s[row][col] )[2:]
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : List[str] , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] ):
__a : List[Any] = message[:4]
__a : str = message[4:]
__a : Any = apply_table(lowerCamelCase_ , lowerCamelCase_ )
__a : int = xor(lowerCamelCase_ , lowerCamelCase_ )
__a : Dict = apply_sbox(lowerCamelCase_ , temp[:4] ) # noqa: E741
__a : Tuple = apply_sbox(lowerCamelCase_ , temp[4:] )
__a : List[Any] = '0' * (2 - len(lowerCamelCase_ )) + l # noqa: E741
__a : List[str] = '0' * (2 - len(lowerCamelCase_ )) + r
__a : List[Any] = apply_table(l + r , lowerCamelCase_ )
__a : Dict = xor(lowerCamelCase_ , lowerCamelCase_ )
return temp + right
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter 10 bit key: ''')
SCREAMING_SNAKE_CASE__ = input('''Enter 8 bit message: ''')
SCREAMING_SNAKE_CASE__ = [6, 3, 7, 4, 8, 5, 10, 9]
SCREAMING_SNAKE_CASE__ = [3, 5, 2, 7, 4, 10, 1, 9, 8, 6]
SCREAMING_SNAKE_CASE__ = [2, 4, 3, 1]
SCREAMING_SNAKE_CASE__ = [2, 6, 3, 1, 4, 8, 5, 7]
SCREAMING_SNAKE_CASE__ = [4, 1, 3, 5, 7, 2, 8, 6]
SCREAMING_SNAKE_CASE__ = [4, 1, 2, 3, 2, 3, 4, 1]
SCREAMING_SNAKE_CASE__ = [[1, 0, 3, 2], [3, 2, 1, 0], [0, 2, 1, 3], [3, 1, 3, 2]]
SCREAMING_SNAKE_CASE__ = [[0, 1, 2, 3], [2, 0, 1, 3], [3, 0, 1, 0], [2, 1, 0, 3]]
# key generation
SCREAMING_SNAKE_CASE__ = apply_table(key, paa_table)
SCREAMING_SNAKE_CASE__ = temp[:5]
SCREAMING_SNAKE_CASE__ = temp[5:]
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
# encryption
SCREAMING_SNAKE_CASE__ = apply_table(message, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Cipher text is:''', CT)
# decryption
SCREAMING_SNAKE_CASE__ = apply_table(CT, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Plain text after decypting is:''', PT)
| 47 | 1 |
from random import randint
from tempfile import TemporaryFile
import numpy as np
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : str , lowerCamelCase_ : Union[str, Any] ):
__a : Optional[Any] = 0
if start < end:
__a : List[str] = randint(lowerCamelCase_ , lowerCamelCase_ )
__a : int = a[end]
__a : str = a[pivot]
__a : str = temp
__a , __a : str = _in_place_partition(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
count += _in_place_quick_sort(lowerCamelCase_ , lowerCamelCase_ , p - 1 )
count += _in_place_quick_sort(lowerCamelCase_ , p + 1 , lowerCamelCase_ )
return count
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Tuple , lowerCamelCase_ : str ):
__a : int = 0
__a : str = randint(lowerCamelCase_ , lowerCamelCase_ )
__a : List[str] = a[end]
__a : List[Any] = a[pivot]
__a : Union[str, Any] = temp
__a : Optional[Any] = start - 1
for index in range(lowerCamelCase_ , lowerCamelCase_ ):
count += 1
if a[index] < a[end]: # check if current val is less than pivot value
__a : int = new_pivot_index + 1
__a : List[str] = a[new_pivot_index]
__a : List[Any] = a[index]
__a : List[Any] = temp
__a : int = a[new_pivot_index + 1]
__a : Tuple = a[end]
__a : str = temp
return new_pivot_index + 1, count
SCREAMING_SNAKE_CASE__ = TemporaryFile()
SCREAMING_SNAKE_CASE__ = 100 # 1000 elements are to be sorted
SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ = 0, 1 # mean and standard deviation
SCREAMING_SNAKE_CASE__ = np.random.normal(mu, sigma, p)
np.save(outfile, X)
print('''The array is''')
print(X)
outfile.seek(0) # using the same array
SCREAMING_SNAKE_CASE__ = np.load(outfile)
SCREAMING_SNAKE_CASE__ = len(M) - 1
SCREAMING_SNAKE_CASE__ = _in_place_quick_sort(M, 0, r)
print(
'''No of Comparisons for 100 elements selected from a standard normal distribution'''
'''is :'''
)
print(z)
| 47 |
import unittest
from transformers import is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from transformers import GradientAccumulator, create_optimizer
@require_tf
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : List[Any] = GradientAccumulator()
accumulator([tf.constant([1.0, 2.0] )] )
accumulator([tf.constant([-2.0, 1.0] )] )
accumulator([tf.constant([-1.0, 2.0] )] )
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
accumulator([tf.constant([1.0, 1.0] ), tf.constant([2.0, 2.0] )] )
self.assertEqual(accumulator.step , 3 )
self.assertEqual(len(accumulator.gradients ) , 1 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [-2.0, 5.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [0.0, 0.0] , tol=1e-2 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : int = None
ops.enable_eager_execution_internal()
__a : Optional[Any] = tf.config.list_physical_devices('CPU' )
if len(SCREAMING_SNAKE_CASE__ ) == 1:
tf.config.set_logical_device_configuration(
physical_devices[0] , [tf.config.LogicalDeviceConfiguration(), tf.config.LogicalDeviceConfiguration()] )
__a : int = tf.config.list_logical_devices(device_type='CPU' )
__a : str = tf.distribute.MirroredStrategy(devices=devices[:2] )
with strategy.scope():
__a : List[str] = GradientAccumulator()
__a : Tuple = tf.Variable([4.0, 3.0] )
__a , __a : int = create_optimizer(5e-5 , 1_0 , 5 )
__a : List[Any] = tf.Variable([0.0, 0.0] , trainable=SCREAMING_SNAKE_CASE__ )
def accumulate_on_replica(SCREAMING_SNAKE_CASE__ : Optional[Any] ):
accumulator([gradient] )
def apply_on_replica():
optimizer.apply_gradients(list(zip(accumulator.gradients , [variable] ) ) )
@tf.function
def accumulate(SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
with strategy.scope():
__a : Optional[Any] = strategy.experimental_local_results(SCREAMING_SNAKE_CASE__ )
local_variables[0].assign(SCREAMING_SNAKE_CASE__ )
local_variables[1].assign(SCREAMING_SNAKE_CASE__ )
strategy.run(SCREAMING_SNAKE_CASE__ , args=(gradient_placeholder,) )
@tf.function
def apply_grad():
with strategy.scope():
strategy.run(SCREAMING_SNAKE_CASE__ )
def _check_local_values(SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int ):
__a : Union[str, Any] = strategy.experimental_local_results(accumulator._gradients[0] )
self.assertListAlmostEqual(values[0].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
self.assertListAlmostEqual(values[1].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
accumulate([1.0, 2.0] , [-1.0, 1.0] )
accumulate([3.0, -1.0] , [-1.0, -1.0] )
accumulate([-2.0, 2.0] , [3.0, -2.0] )
self.assertEqual(accumulator.step , 3 )
_check_local_values([2.0, 3.0] , [1.0, -2.0] )
apply_grad()
self.assertListAlmostEqual(variable.value() , [4.0, 3.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
_check_local_values([0.0, 0.0] , [0.0, 0.0] )
| 47 | 1 |
import os
import numpy
import onnx
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : List[Any] ):
__a : Union[str, Any] = a.name
__a : Dict = b.name
__a : Dict = ''
__a : Tuple = ''
__a : Optional[int] = a == b
__a : Dict = name_a
__a : Tuple = name_b
return res
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Tuple ):
for i, input_name in enumerate(node_proto.input ):
if input_name == name:
node_proto.input.insert(lowerCamelCase_ , lowerCamelCase_ )
node_proto.input.pop(i + 1 )
if node_proto.op_type == "If":
_graph_replace_input_with(node_proto.attribute[0].g , lowerCamelCase_ , lowerCamelCase_ )
_graph_replace_input_with(node_proto.attribute[1].g , lowerCamelCase_ , lowerCamelCase_ )
if node_proto.op_type == "Loop":
_graph_replace_input_with(node_proto.attribute[0].g , lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : int , lowerCamelCase_ : Optional[int] ):
for n in graph_proto.node:
_node_replace_input_with(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : List[str] , lowerCamelCase_ : List[str] ):
__a : Any = list(model.graph.initializer )
__a : Optional[Any] = list(model_without_ext.graph.initializer )
for i, ref_i in ind_to_replace:
assert inits_with_data[i].name == inits[i].name
assert inits_with_data[ref_i].name == inits[ref_i].name
assert i > ref_i
__a : Tuple = inits[i].name
__a : str = inits[ref_i].name
model_without_ext.graph.initializer.remove(inits[i] )
# for n in model.graph.node:
_graph_replace_input_with(model_without_ext.graph , lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] ):
__a : int = os.path.dirname(lowerCamelCase_ )
__a : Any = os.path.basename(lowerCamelCase_ )
__a : Any = onnx.load(os.path.join(lowerCamelCase_ , lowerCamelCase_ ) )
__a : str = list(model.graph.initializer )
__a : Union[str, Any] = set()
__a : Union[str, Any] = {}
__a : Optional[int] = []
__a : Dict = 0
for i in range(len(lowerCamelCase_ ) ):
if i in dup_set:
continue
for j in range(i + 1 , len(lowerCamelCase_ ) ):
if j in dup_set:
continue
if _is_equal_tensor_proto(inits[i] , inits[j] ):
dup_set.add(lowerCamelCase_ )
dup_set.add(lowerCamelCase_ )
__a : Union[str, Any] = inits[j].data_type
__a : Tuple = numpy.prod(inits[j].dims )
if dtype == 1:
mem_size *= 4
elif dtype == 6:
mem_size *= 4
elif dtype == 7 or dtype == 1_1:
mem_size *= 8
else:
print('unexpected data type: ' , lowerCamelCase_ )
total_reduced_size += mem_size
__a : Dict = inits[i].name
__a : Any = inits[j].name
if name_i in dup_map:
dup_map[name_i].append(lowerCamelCase_ )
else:
__a : List[Any] = [name_j]
ind_to_replace.append((j, i) )
print('total reduced size: ' , total_reduced_size / 1_0_2_4 / 1_0_2_4 / 1_0_2_4 , 'GB' )
__a : Optional[Any] = sorted(lowerCamelCase_ )
_remove_dup_initializers_from_model(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
__a : Tuple = 'optimized_' + model_file_name
__a : Tuple = os.path.join(lowerCamelCase_ , lowerCamelCase_ )
onnx.save(lowerCamelCase_ , lowerCamelCase_ )
return new_model
| 47 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''roberta-base''': '''https://huggingface.co/roberta-base/resolve/main/config.json''',
'''roberta-large''': '''https://huggingface.co/roberta-large/resolve/main/config.json''',
'''roberta-large-mnli''': '''https://huggingface.co/roberta-large-mnli/resolve/main/config.json''',
'''distilroberta-base''': '''https://huggingface.co/distilroberta-base/resolve/main/config.json''',
'''roberta-base-openai-detector''': '''https://huggingface.co/roberta-base-openai-detector/resolve/main/config.json''',
'''roberta-large-openai-detector''': '''https://huggingface.co/roberta-large-openai-detector/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = '''roberta'''
def __init__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=5_0_2_6_5 , SCREAMING_SNAKE_CASE__ : Optional[int]=7_6_8 , SCREAMING_SNAKE_CASE__ : str=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Any="gelu" , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.1 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=2 , SCREAMING_SNAKE_CASE__ : Any=0.02 , SCREAMING_SNAKE_CASE__ : List[str]=1e-12 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0 , SCREAMING_SNAKE_CASE__ : List[str]=2 , SCREAMING_SNAKE_CASE__ : Tuple="absolute" , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[str]=None , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = vocab_size
__a : Tuple = hidden_size
__a : List[str] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : str = hidden_act
__a : Optional[Any] = intermediate_size
__a : Dict = hidden_dropout_prob
__a : List[str] = attention_probs_dropout_prob
__a : Optional[Any] = max_position_embeddings
__a : Dict = type_vocab_size
__a : str = initializer_range
__a : List[str] = layer_norm_eps
__a : Optional[int] = position_embedding_type
__a : Union[str, Any] = use_cache
__a : str = classifier_dropout
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
if self.task == "multiple-choice":
__a : List[str] = {0: 'batch', 1: 'choice', 2: 'sequence'}
else:
__a : Dict = {0: 'batch', 1: 'sequence'}
return OrderedDict(
[
('input_ids', dynamic_axis),
('attention_mask', dynamic_axis),
] )
| 47 | 1 |
from abc import ABC, abstractmethod
from typing import Optional, Union
from .. import Dataset, DatasetDict, Features, IterableDataset, IterableDatasetDict, NamedSplit
from ..utils.typing import NestedDataStructureLike, PathLike
class _UpperCamelCase( __lowerCamelCase ):
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : Optional[NestedDataStructureLike[PathLike]] = None , SCREAMING_SNAKE_CASE__ : Optional[NamedSplit] = None , SCREAMING_SNAKE_CASE__ : Optional[Features] = None , SCREAMING_SNAKE_CASE__ : str = None , SCREAMING_SNAKE_CASE__ : bool = False , SCREAMING_SNAKE_CASE__ : bool = False , SCREAMING_SNAKE_CASE__ : Optional[int] = None , **SCREAMING_SNAKE_CASE__ : Dict , ):
'''simple docstring'''
__a : Any = path_or_paths
__a : List[Any] = split if split or isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) else 'train'
__a : Any = features
__a : str = cache_dir
__a : List[str] = keep_in_memory
__a : Dict = streaming
__a : str = num_proc
__a : List[str] = kwargs
@abstractmethod
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
pass
class _UpperCamelCase( __lowerCamelCase ):
def __init__( self : int , SCREAMING_SNAKE_CASE__ : Optional[Features] = None , SCREAMING_SNAKE_CASE__ : str = None , SCREAMING_SNAKE_CASE__ : bool = False , SCREAMING_SNAKE_CASE__ : bool = False , SCREAMING_SNAKE_CASE__ : Optional[int] = None , **SCREAMING_SNAKE_CASE__ : Dict , ):
'''simple docstring'''
__a : Tuple = features
__a : Union[str, Any] = cache_dir
__a : Optional[Any] = keep_in_memory
__a : Tuple = streaming
__a : Tuple = num_proc
__a : Optional[int] = kwargs
@abstractmethod
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
pass
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''▁'''
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''sentencepiece.bpe.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''facebook/xglm-564M''': '''https://huggingface.co/facebook/xglm-564M/resolve/main/sentencepiece.bpe.model''',
}
}
SCREAMING_SNAKE_CASE__ = {
'''facebook/xglm-564M''': 2048,
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : Union[str, Any] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : str = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : Any = ['''input_ids''', '''attention_mask''']
def __init__( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : List[str]="</s>" , SCREAMING_SNAKE_CASE__ : Optional[Any]="</s>" , SCREAMING_SNAKE_CASE__ : Union[str, Any]="<s>" , SCREAMING_SNAKE_CASE__ : str="<unk>" , SCREAMING_SNAKE_CASE__ : Dict="<pad>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : List[str] , ):
'''simple docstring'''
__a : Union[str, Any] = {} if sp_model_kwargs is None else sp_model_kwargs
# Compatibility with the original tokenizer
__a : Any = 7
__a : Union[str, Any] = [f'''<madeupword{i}>''' for i in range(self.num_madeup_words )]
__a : Union[str, Any] = kwargs.get('additional_special_tokens' , [] )
kwargs["additional_special_tokens"] += [
word for word in madeup_words if word not in kwargs["additional_special_tokens"]
]
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , cls_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : List[str] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(str(SCREAMING_SNAKE_CASE__ ) )
__a : str = vocab_file
# Original fairseq vocab and spm vocab must be "aligned":
# Vocab | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
# -------- | ------- | ------- | ------ | ------- | --- | --- | --- | ----- | ----- | ----
# fairseq | '<s>' | '<pad>' | '</s>' | '<unk>' | ',' | '.' | '▁' | 's' | '▁de' | '-'
# spm | '<unk>' | '<s>' | '</s>' | ',' | '.' | '▁' | 's' | '▁de' | '-' | '▁a'
# The first "real" token "," has position 4 in the original fairseq vocab and position 3 in the spm vocab
__a : Any = 1
# Mimic fairseq token-to-id alignment for the first 4 token
__a : str = {'<s>': 0, '<pad>': 1, '</s>': 2, '<unk>': 3}
__a : List[str] = len(self.sp_model )
__a : Optional[int] = {f'''<madeupword{i}>''': sp_size + i + self.fairseq_offset for i in range(self.num_madeup_words )}
self.fairseq_tokens_to_ids.update(SCREAMING_SNAKE_CASE__ )
__a : Dict = {v: k for k, v in self.fairseq_tokens_to_ids.items()}
def __getstate__( self : List[str] ):
'''simple docstring'''
__a : Tuple = self.__dict__.copy()
__a : List[str] = None
__a : Optional[int] = self.sp_model.serialized_model_proto()
return state
def __setstate__( self : List[str] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : int = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : Dict = {}
__a : Tuple = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.LoadFromSerializedProto(self.sp_model_proto )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
if token_ids_a is None:
return [self.sep_token_id] + token_ids_a
__a : Optional[Any] = [self.sep_token_id]
return sep + token_ids_a + sep + sep + token_ids_a
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : bool = False ):
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=SCREAMING_SNAKE_CASE__ , token_ids_a=SCREAMING_SNAKE_CASE__ , already_has_special_tokens=SCREAMING_SNAKE_CASE__ )
if token_ids_a is None:
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ )) + [1, 1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
__a : Optional[int] = [self.sep_token_id]
if token_ids_a is None:
return len(sep + token_ids_a ) * [0]
return len(sep + token_ids_a + sep + sep + token_ids_a ) * [0]
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return len(self.sp_model ) + self.fairseq_offset + self.num_madeup_words
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : str = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if token in self.fairseq_tokens_to_ids:
return self.fairseq_tokens_to_ids[token]
__a : List[str] = self.sp_model.PieceToId(SCREAMING_SNAKE_CASE__ )
# Need to return unknown token if the SP model returned 0
return spm_id + self.fairseq_offset if spm_id else self.unk_token_id
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if index in self.fairseq_ids_to_tokens:
return self.fairseq_ids_to_tokens[index]
return self.sp_model.IdToPiece(index - self.fairseq_offset )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
__a : Optional[int] = ''.join(SCREAMING_SNAKE_CASE__ ).replace(SCREAMING_SNAKE_CASE__ , ' ' ).strip()
return out_string
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Any = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[Any] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : int ):
if b == 0:
return 1
if (b % 2) == 0:
return actual_power(lowerCamelCase_ , int(b / 2 ) ) * actual_power(lowerCamelCase_ , int(b / 2 ) )
else:
return a * actual_power(lowerCamelCase_ , int(b / 2 ) ) * actual_power(lowerCamelCase_ , int(b / 2 ) )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : int ):
if b < 0:
return 1 / actual_power(lowerCamelCase_ , lowerCamelCase_ )
return actual_power(lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
print(power(-2, -3))
| 47 |
import argparse
from collections import OrderedDict
from pathlib import Path
import torch
from transformers import (
VisualBertConfig,
VisualBertForMultipleChoice,
VisualBertForPreTraining,
VisualBertForQuestionAnswering,
VisualBertForVisualReasoning,
)
from transformers.utils import logging
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = [
('''bert.bert''', '''visual_bert'''),
('''bert.cls''', '''cls'''),
('''bert.classifier''', '''cls'''),
('''token_type_embeddings_visual''', '''visual_token_type_embeddings'''),
('''position_embeddings_visual''', '''visual_position_embeddings'''),
('''projection''', '''visual_projection'''),
]
SCREAMING_SNAKE_CASE__ = [
'''nlvr2_coco_pre_trained.th''',
'''nlvr2_fine_tuned.th''',
'''nlvr2_pre_trained.th''',
'''vcr_coco_pre_train.th''',
'''vcr_fine_tune.th''',
'''vcr_pre_train.th''',
'''vqa_coco_pre_trained.th''',
'''vqa_fine_tuned.th''',
'''vqa_pre_trained.th''',
]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ):
__a : str = torch.load(lowerCamelCase_ , map_location='cpu' )
return sd
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : int , lowerCamelCase_ : Dict=rename_keys_prefix ):
__a : Optional[Any] = OrderedDict()
__a : Any = torch.arange(config.max_position_embeddings ).expand((1, -1) )
# detector_d = OrderedDict()
for key in d:
if "detector" in key:
# detector_d[key.replace('detector.','')] = d[key]
continue
__a : List[Any] = key
for name_pair in rename_keys_prefix:
__a : List[str] = new_key.replace(name_pair[0] , name_pair[1] )
__a : Any = d[key]
if key == "bert.cls.predictions.decoder.weight":
# Old bert code didn't have `decoder.bias`, but was added separately
__a : int = new_d['cls.predictions.bias']
return new_d
@torch.no_grad()
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : Any ):
assert (
checkpoint_path.split('/' )[-1] in ACCEPTABLE_CHECKPOINTS
), f'''The checkpoint provided must be in {ACCEPTABLE_CHECKPOINTS}.'''
# Get Config
if "pre" in checkpoint_path:
__a : Dict = 'pretraining'
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
elif "vqa_advanced" in checkpoint_path:
__a : int = {'visual_embedding_dim': 2_0_4_8}
elif "vqa" in checkpoint_path:
__a : Tuple = {'visual_embedding_dim': 2_0_4_8}
elif "nlvr" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 1_0_2_4}
else:
raise NotImplementedError(f'''No implementation found for `{checkpoint_path}`.''' )
else:
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
__a : Any = 'multichoice'
elif "vqa_advanced" in checkpoint_path:
__a : Any = {'visual_embedding_dim': 2_0_4_8}
__a : List[str] = 'vqa_advanced'
elif "vqa" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 2_0_4_8, 'num_labels': 3_1_2_9}
__a : List[Any] = 'vqa'
elif "nlvr" in checkpoint_path:
__a : Optional[int] = {
'visual_embedding_dim': 1_0_2_4,
'num_labels': 2,
}
__a : Optional[Any] = 'nlvr'
__a : str = VisualBertConfig(**lowerCamelCase_ )
# Load State Dict
__a : str = load_state_dict(lowerCamelCase_ )
__a : str = get_new_dict(lowerCamelCase_ , lowerCamelCase_ )
if model_type == "pretraining":
__a : Optional[Any] = VisualBertForPreTraining(lowerCamelCase_ )
elif model_type == "vqa":
__a : Any = VisualBertForQuestionAnswering(lowerCamelCase_ )
elif model_type == "nlvr":
__a : int = VisualBertForVisualReasoning(lowerCamelCase_ )
elif model_type == "multichoice":
__a : Optional[int] = VisualBertForMultipleChoice(lowerCamelCase_ )
model.load_state_dict(lowerCamelCase_ )
# Save Checkpoints
Path(lowerCamelCase_ ).mkdir(exist_ok=lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument('''orig_checkpoint_path''', type=str, help='''A path to .th on local filesystem.''')
parser.add_argument('''pytorch_dump_folder_path''', type=str, help='''Path to the output PyTorch model.''')
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_visual_bert_checkpoint(args.orig_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
import math
def UpperCAmelCase__ ( lowerCamelCase_ : int = 1_0_0 ):
__a : Dict = sum(i * i for i in range(1 , n + 1 ) )
__a : List[Any] = int(math.pow(sum(range(1 , n + 1 ) ) , 2 ) )
return square_of_sum - sum_of_squares
if __name__ == "__main__":
print(F"{solution() = }")
| 47 |
print((lambda quine: quine % quine)('''print((lambda quine: quine %% quine)(%r))'''))
| 47 | 1 |
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, TensorDataset
from tqdm import tqdm
from transformers import GPTaLMHeadModel
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : int ):
# save results
if os.path.exists(lowerCamelCase_ ):
if os.path.exists(os.path.join(lowerCamelCase_ , 'config.json' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'config.json' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'config.json' ) )
if os.path.exists(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) )
else:
os.makedirs(lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Any=False ):
__a : Dict = 2
if unlogit:
__a : Optional[Any] = torch.pow(lowerCamelCase_ , lowerCamelCase_ )
__a : Any = p * torch.log(lowerCamelCase_ )
__a : Union[str, Any] = 0
return -plogp.sum(dim=-1 )
def UpperCAmelCase__ ( lowerCamelCase_ : Any ):
logger.info('lv, h >\t' + '\t'.join(f'''{x + 1}''' for x in range(len(lowerCamelCase_ ) ) ) )
for row in range(len(lowerCamelCase_ ) ):
if tensor.dtype != torch.long:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:.5f}''' for x in tensor[row].cpu().data ) )
else:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:d}''' for x in tensor[row].cpu().data ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Any , lowerCamelCase_ : int , lowerCamelCase_ : int=True , lowerCamelCase_ : Optional[Any]=True , lowerCamelCase_ : List[Any]=None , lowerCamelCase_ : List[Any]=False ):
__a , __a : Optional[int] = model.config.num_hidden_layers, model.config.num_attention_heads
__a : str = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
__a : int = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
if head_mask is None:
__a : Union[str, Any] = torch.ones(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
head_mask.requires_grad_(requires_grad=lowerCamelCase_ )
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
__a : Any = None
__a : Optional[int] = 0.0
__a : Optional[Any] = 0.0
for step, inputs in enumerate(tqdm(lowerCamelCase_ , desc='Iteration' , disable=args.local_rank not in [-1, 0] ) ):
__a : Dict = tuple(t.to(args.device ) for t in inputs )
((__a) , ) : Dict = inputs
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
__a : List[Any] = model(lowerCamelCase_ , labels=lowerCamelCase_ , head_mask=lowerCamelCase_ )
# (loss), lm_logits, presents, (all hidden_states), (attentions)
__a , __a , __a : int = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
total_loss += loss.detach().cpu().numpy()
if compute_entropy:
for layer, attn in enumerate(lowerCamelCase_ ):
__a : List[str] = entropy(attn.detach() , lowerCamelCase_ )
attn_entropy[layer] += masked_entropy.sum(-1 ).sum(0 ).sum(0 ).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
tot_tokens += torch.ones_like(lowerCamelCase_ ).float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
__a : Optional[Any] = 2
__a : Union[str, Any] = torch.pow(torch.pow(lowerCamelCase_ , lowerCamelCase_ ).sum(-1 ) , 1 / exponent )
head_importance /= norm_by_layer.unsqueeze(-1 ) + 1e-20
if not args.dont_normalize_global_importance:
__a : List[str] = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print matrices
if compute_entropy:
logger.info('Attention entropies' )
print_ad_tensor(lowerCamelCase_ )
if compute_importance:
logger.info('Head importance scores' )
print_ad_tensor(lowerCamelCase_ )
logger.info('Head ranked by importance scores' )
__a : Optional[Any] = torch.zeros(head_importance.numel() , dtype=torch.long , device=args.device )
__a : str = torch.arange(
head_importance.numel() , device=args.device )
__a : Tuple = head_ranks.view_as(lowerCamelCase_ )
print_ad_tensor(lowerCamelCase_ )
return attn_entropy, head_importance, total_loss
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Tuple , lowerCamelCase_ : int ):
__a , __a , __a : Optional[int] = compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ )
__a : Tuple = 1 / loss # instead of downsteam score use the LM loss
logger.info('Pruning: original score: %f, threshold: %f' , lowerCamelCase_ , original_score * args.masking_threshold )
__a : Tuple = torch.ones_like(lowerCamelCase_ )
__a : int = max(1 , int(new_head_mask.numel() * args.masking_amount ) )
__a : Tuple = original_score
while current_score >= original_score * args.masking_threshold:
__a : Optional[Any] = new_head_mask.clone().detach() # save current head mask
# heads from least important to most - keep only not-masked heads
__a : List[str] = float('Inf' )
__a : List[Any] = head_importance.view(-1 ).sort()[1]
if len(lowerCamelCase_ ) <= num_to_mask:
print('BREAK BY num_to_mask' )
break
# mask heads
__a : Any = current_heads_to_mask[:num_to_mask]
logger.info('Heads to mask: %s' , str(current_heads_to_mask.tolist() ) )
__a : int = new_head_mask.view(-1 )
__a : Tuple = 0.0
__a : int = new_head_mask.view_as(lowerCamelCase_ )
__a : Optional[int] = new_head_mask.clone().detach()
print_ad_tensor(lowerCamelCase_ )
# Compute metric and head importance again
__a , __a , __a : int = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[Any] = 1 / loss
logger.info(
'Masking: current score: %f, remaining heads %d (%.1f percents)' , lowerCamelCase_ , new_head_mask.sum() , new_head_mask.sum() / new_head_mask.numel() * 1_0_0 , )
logger.info('Final head mask' )
print_ad_tensor(lowerCamelCase_ )
np.save(os.path.join(args.output_dir , 'head_mask.npy' ) , head_mask.detach().cpu().numpy() )
return head_mask
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Union[str, Any] ):
__a : List[Any] = datetime.now()
__a , __a , __a : List[str] = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[str] = 1 / loss
__a : List[Any] = datetime.now() - before_time
__a : List[str] = sum(p.numel() for p in model.parameters() )
__a : Dict = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(lowerCamelCase_ ) )
}
for k, v in heads_to_prune.items():
if isinstance(lowerCamelCase_ , lowerCamelCase_ ):
__a : Tuple = [
v,
]
assert sum(len(lowerCamelCase_ ) for h in heads_to_prune.values() ) == (1 - head_mask.long()).sum().item()
model.prune_heads(lowerCamelCase_ )
__a : Optional[Any] = sum(p.numel() for p in model.parameters() )
__a : Tuple = datetime.now()
__a , __a , __a : Tuple = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ , actually_pruned=lowerCamelCase_ , )
__a : Optional[Any] = 1 / loss
__a : List[Any] = datetime.now() - before_time
logger.info(
'Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)' , lowerCamelCase_ , lowerCamelCase_ , pruned_num_params / original_num_params * 1_0_0 , )
logger.info('Pruning: score with masking: %f score with pruning: %f' , lowerCamelCase_ , lowerCamelCase_ )
logger.info('Pruning: speed ratio (original timing / new timing): %f percents' , original_time / new_time * 1_0_0 )
save_model(lowerCamelCase_ , args.output_dir )
def UpperCAmelCase__ ( ):
__a : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'--data_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The input data dir. Should contain the .tsv files (or other data files) for the task.' , )
parser.add_argument(
'--model_name_or_path' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='Path to pretrained model or model identifier from huggingface.co/models' , )
parser.add_argument(
'--output_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
# Other parameters
parser.add_argument(
'--config_name' , default='' , type=lowerCamelCase_ , help='Pretrained config name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--tokenizer_name' , default='' , type=lowerCamelCase_ , help='Pretrained tokenizer name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--cache_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , help='Where do you want to store the pre-trained models downloaded from s3' , )
parser.add_argument(
'--data_subset' , type=lowerCamelCase_ , default=-1 , help='If > 0: limit the data to a subset of data_subset instances.' )
parser.add_argument(
'--overwrite_output_dir' , action='store_true' , help='Whether to overwrite data in output directory' )
parser.add_argument(
'--overwrite_cache' , action='store_true' , help='Overwrite the cached training and evaluation sets' )
parser.add_argument(
'--dont_normalize_importance_by_layer' , action='store_true' , help='Don\'t normalize importance score by layers' )
parser.add_argument(
'--dont_normalize_global_importance' , action='store_true' , help='Don\'t normalize all importance scores between 0 and 1' , )
parser.add_argument(
'--try_masking' , action='store_true' , help='Whether to try to mask head until a threshold of accuracy.' )
parser.add_argument(
'--masking_threshold' , default=0.9 , type=lowerCamelCase_ , help='masking threshold in term of metrics (stop masking when metric < threshold * original metric value).' , )
parser.add_argument(
'--masking_amount' , default=0.1 , type=lowerCamelCase_ , help='Amount to heads to masking at each masking step.' )
parser.add_argument('--metric_name' , default='acc' , type=lowerCamelCase_ , help='Metric to use for head masking.' )
parser.add_argument(
'--max_seq_length' , default=1_2_8 , type=lowerCamelCase_ , help=(
'The maximum total input sequence length after WordPiece tokenization. \n'
'Sequences longer than this will be truncated, sequences shorter padded.'
) , )
parser.add_argument('--batch_size' , default=1 , type=lowerCamelCase_ , help='Batch size.' )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 )
parser.add_argument('--local_rank' , type=lowerCamelCase_ , default=-1 , help='local_rank for distributed training on gpus' )
parser.add_argument('--no_cuda' , action='store_true' , help='Whether not to use CUDA when available' )
parser.add_argument('--server_ip' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
parser.add_argument('--server_port' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
__a : Optional[Any] = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print('Waiting for debugger attach' )
ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=lowerCamelCase_ )
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
__a : List[str] = torch.device('cuda' if torch.cuda.is_available() and not args.no_cuda else 'cpu' )
__a : Tuple = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank )
__a : Union[str, Any] = torch.device('cuda' , args.local_rank )
__a : Any = 1
torch.distributed.init_process_group(backend='nccl' ) # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN )
logger.info('device: {} n_gpu: {}, distributed: {}'.format(args.device , args.n_gpu , bool(args.local_rank != -1 ) ) )
__a : Optional[Any] = GPTaLMHeadModel.from_pretrained(args.model_name_or_path )
# Distributed and parallel training
model.to(args.device )
if args.local_rank != -1:
__a : List[Any] = nn.parallel.DistributedDataParallel(
lowerCamelCase_ , device_ids=[args.local_rank] , output_device=args.local_rank , find_unused_parameters=lowerCamelCase_ )
elif args.n_gpu > 1:
__a : Union[str, Any] = nn.DataParallel(lowerCamelCase_ )
# Print/save training arguments
os.makedirs(args.output_dir , exist_ok=lowerCamelCase_ )
torch.save(lowerCamelCase_ , os.path.join(args.output_dir , 'run_args.bin' ) )
logger.info('Training/evaluation parameters %s' , lowerCamelCase_ )
# Prepare dataset
__a : Tuple = np.concatenate(
[
np.loadtxt(args.data_dir , dtype=np.intaa ),
] )
__a : str = (torch.from_numpy(lowerCamelCase_ ),)
__a : List[str] = TensorDataset(*lowerCamelCase_ )
__a : Optional[Any] = RandomSampler(lowerCamelCase_ )
__a : Union[str, Any] = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.batch_size )
# Compute head entropy and importance score
compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
__a : Union[str, Any] = mask_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
prune_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 |
import json
import os
import shutil
import tempfile
from unittest import TestCase
from transformers import BartTokenizer, BartTokenizerFast, DPRQuestionEncoderTokenizer, DPRQuestionEncoderTokenizerFast
from transformers.models.bart.configuration_bart import BartConfig
from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES
from transformers.models.dpr.configuration_dpr import DPRConfig
from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
from transformers.testing_utils import require_faiss, require_tokenizers, require_torch, slow
from transformers.utils import is_datasets_available, is_faiss_available, is_torch_available
if is_torch_available() and is_datasets_available() and is_faiss_available():
from transformers.models.rag.configuration_rag import RagConfig
from transformers.models.rag.tokenization_rag import RagTokenizer
@require_faiss
@require_torch
class _UpperCamelCase( __lowerCamelCase ):
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[Any] = tempfile.mkdtemp()
__a : int = 8
# DPR tok
__a : Dict = [
'[UNK]',
'[CLS]',
'[SEP]',
'[PAD]',
'[MASK]',
'want',
'##want',
'##ed',
'wa',
'un',
'runn',
'##ing',
',',
'low',
'lowest',
]
__a : int = os.path.join(self.tmpdirname , 'dpr_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , DPR_VOCAB_FILES_NAMES['vocab_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as vocab_writer:
vocab_writer.write(''.join([x + '\n' for x in vocab_tokens] ) )
# BART tok
__a : str = [
'l',
'o',
'w',
'e',
'r',
's',
't',
'i',
'd',
'n',
'\u0120',
'\u0120l',
'\u0120n',
'\u0120lo',
'\u0120low',
'er',
'\u0120lowest',
'\u0120newer',
'\u0120wider',
'<unk>',
]
__a : Optional[int] = dict(zip(SCREAMING_SNAKE_CASE__ , range(len(SCREAMING_SNAKE_CASE__ ) ) ) )
__a : List[str] = ['#version: 0.2', '\u0120 l', '\u0120l o', '\u0120lo w', 'e r', '']
__a : List[str] = {'unk_token': '<unk>'}
__a : Dict = os.path.join(self.tmpdirname , 'bart_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['vocab_file'] )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(SCREAMING_SNAKE_CASE__ ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'dpr_tokenizer' ) )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'bart_tokenizer' ) )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
shutil.rmtree(self.tmpdirname )
@require_tokenizers
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Tuple = os.path.join(self.tmpdirname , 'rag_tokenizer' )
__a : Optional[Any] = RagConfig(question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() )
__a : Optional[Any] = RagTokenizer(question_encoder=self.get_dpr_tokenizer() , generator=self.get_bart_tokenizer() )
rag_config.save_pretrained(SCREAMING_SNAKE_CASE__ )
rag_tokenizer.save_pretrained(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = RagTokenizer.from_pretrained(SCREAMING_SNAKE_CASE__ , config=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(new_rag_tokenizer.question_encoder , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.question_encoder.get_vocab() , rag_tokenizer.question_encoder.get_vocab() )
self.assertIsInstance(new_rag_tokenizer.generator , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.generator.get_vocab() , rag_tokenizer.generator.get_vocab() )
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Optional[Any] = RagTokenizer.from_pretrained('facebook/rag-token-nq' )
__a : List[Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : Tuple = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Any = RagTokenizer.from_pretrained('facebook/rag-sequence-nq' )
__a : Union[str, Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : str = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
| 47 | 1 |
import numpy as np
import torch
import tqdm
from ...models.unet_ad import UNetaDModel
from ...pipelines import DiffusionPipeline
from ...utils import randn_tensor
from ...utils.dummy_pt_objects import DDPMScheduler
class _UpperCamelCase( __lowerCamelCase ):
def __init__( self : str , SCREAMING_SNAKE_CASE__ : UNetaDModel , SCREAMING_SNAKE_CASE__ : UNetaDModel , SCREAMING_SNAKE_CASE__ : DDPMScheduler , SCREAMING_SNAKE_CASE__ : str , ):
'''simple docstring'''
super().__init__()
__a : Tuple = value_function
__a : List[Any] = unet
__a : Optional[int] = scheduler
__a : Optional[Any] = env
__a : str = env.get_dataset()
__a : Tuple = {}
for key in self.data.keys():
try:
__a : str = self.data[key].mean()
except: # noqa: E722
pass
__a : str = {}
for key in self.data.keys():
try:
__a : Any = self.data[key].std()
except: # noqa: E722
pass
__a : Optional[int] = env.observation_space.shape[0]
__a : List[str] = env.action_space.shape[0]
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
return (x_in - self.means[key]) / self.stds[key]
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
return x_in * self.stds[key] + self.means[key]
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
if type(SCREAMING_SNAKE_CASE__ ) is dict:
return {k: self.to_torch(SCREAMING_SNAKE_CASE__ ) for k, v in x_in.items()}
elif torch.is_tensor(SCREAMING_SNAKE_CASE__ ):
return x_in.to(self.unet.device )
return torch.tensor(SCREAMING_SNAKE_CASE__ , device=self.unet.device )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
for key, val in cond.items():
__a : str = val.clone()
return x_in
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
__a : int = x.shape[0]
__a : Tuple = None
for i in tqdm.tqdm(self.scheduler.timesteps ):
# create batch of timesteps to pass into model
__a : Union[str, Any] = torch.full((batch_size,) , SCREAMING_SNAKE_CASE__ , device=self.unet.device , dtype=torch.long )
for _ in range(SCREAMING_SNAKE_CASE__ ):
with torch.enable_grad():
x.requires_grad_()
# permute to match dimension for pre-trained models
__a : Optional[int] = self.value_function(x.permute(0 , 2 , 1 ) , SCREAMING_SNAKE_CASE__ ).sample
__a : Optional[int] = torch.autograd.grad([y.sum()] , [x] )[0]
__a : Union[str, Any] = self.scheduler._get_variance(SCREAMING_SNAKE_CASE__ )
__a : Dict = torch.exp(0.5 * posterior_variance )
__a : Tuple = model_std * grad
__a : Tuple = 0
__a : str = x.detach()
__a : List[str] = x + scale * grad
__a : Optional[Any] = self.reset_xa(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.action_dim )
__a : str = self.unet(x.permute(0 , 2 , 1 ) , SCREAMING_SNAKE_CASE__ ).sample.permute(0 , 2 , 1 )
# TODO: verify deprecation of this kwarg
__a : Union[str, Any] = self.scheduler.step(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , predict_epsilon=SCREAMING_SNAKE_CASE__ )['prev_sample']
# apply conditions to the trajectory (set the initial state)
__a : int = self.reset_xa(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.action_dim )
__a : List[str] = self.to_torch(SCREAMING_SNAKE_CASE__ )
return x, y
def __call__( self : Tuple , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=6_4 , SCREAMING_SNAKE_CASE__ : Dict=3_2 , SCREAMING_SNAKE_CASE__ : int=2 , SCREAMING_SNAKE_CASE__ : Any=0.1 ):
'''simple docstring'''
__a : Any = self.normalize(SCREAMING_SNAKE_CASE__ , 'observations' )
__a : Dict = obs[None].repeat(SCREAMING_SNAKE_CASE__ , axis=0 )
__a : Union[str, Any] = {0: self.to_torch(SCREAMING_SNAKE_CASE__ )}
__a : Union[str, Any] = (batch_size, planning_horizon, self.state_dim + self.action_dim)
# generate initial noise and apply our conditions (to make the trajectories start at current state)
__a : Optional[Any] = randn_tensor(SCREAMING_SNAKE_CASE__ , device=self.unet.device )
__a : str = self.reset_xa(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.action_dim )
__a : Any = self.to_torch(SCREAMING_SNAKE_CASE__ )
# run the diffusion process
__a , __a : Dict = self.run_diffusion(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# sort output trajectories by value
__a : List[str] = y.argsort(0 , descending=SCREAMING_SNAKE_CASE__ ).squeeze()
__a : Optional[Any] = x[sorted_idx]
__a : List[Any] = sorted_values[:, :, : self.action_dim]
__a : str = actions.detach().cpu().numpy()
__a : int = self.de_normalize(SCREAMING_SNAKE_CASE__ , key='actions' )
# select the action with the highest value
if y is not None:
__a : List[str] = 0
else:
# if we didn't run value guiding, select a random action
__a : List[str] = np.random.randint(0 , SCREAMING_SNAKE_CASE__ )
__a : Any = denorm_actions[selected_index, 0]
return denorm_actions
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''spiece.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''bert_for_seq_generation''': (
'''https://huggingface.co/google/bert_for_seq_generation_L-24_bbc_encoder/resolve/main/spiece.model'''
),
}
}
SCREAMING_SNAKE_CASE__ = {'''bert_for_seq_generation''': 512}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : Any = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : List[int] = []
__SCREAMING_SNAKE_CASE : int = ['''input_ids''', '''attention_mask''']
def __init__( self : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : Tuple="</s>" , SCREAMING_SNAKE_CASE__ : Any="<unk>" , SCREAMING_SNAKE_CASE__ : int="<pad>" , SCREAMING_SNAKE_CASE__ : List[str]="<::::>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Tuple = {} if sp_model_kwargs is None else sp_model_kwargs
# Add extra_ids to the special token list
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : int = vocab_file
__a : Union[str, Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return self.sp_model.get_piece_size()
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Dict = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __getstate__( self : Optional[Any] ):
'''simple docstring'''
__a : Union[str, Any] = self.__dict__.copy()
__a : Any = None
return state
def __setstate__( self : int , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a : str = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : str = {}
__a : List[Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(self.vocab_file )
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
return self.sp_model.piece_to_id(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : int = self.sp_model.IdToPiece(SCREAMING_SNAKE_CASE__ )
return token
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Optional[Any] = []
__a : Optional[int] = ''
for token in tokens:
# make sure that special tokens are not decoded using sentencepiece model
if token in self.all_special_tokens:
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ ) + token
__a : Dict = []
else:
current_sub_tokens.append(SCREAMING_SNAKE_CASE__ )
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ )
return out_string.strip()
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Tuple = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[str] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
import copy
import os
from collections import OrderedDict
from typing import TYPE_CHECKING, Any, Dict, Mapping, Optional, Union
if TYPE_CHECKING:
from ...processing_utils import ProcessorMixin
from ...utils import TensorType
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''google/owlvit-base-patch32''': '''https://huggingface.co/google/owlvit-base-patch32/resolve/main/config.json''',
'''google/owlvit-base-patch16''': '''https://huggingface.co/google/owlvit-base-patch16/resolve/main/config.json''',
'''google/owlvit-large-patch14''': '''https://huggingface.co/google/owlvit-large-patch14/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = '''owlvit_text_model'''
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Tuple=4_9_4_0_8 , SCREAMING_SNAKE_CASE__ : Optional[int]=5_1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=2_0_4_8 , SCREAMING_SNAKE_CASE__ : int=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=8 , SCREAMING_SNAKE_CASE__ : int=1_6 , SCREAMING_SNAKE_CASE__ : Dict="quick_gelu" , SCREAMING_SNAKE_CASE__ : Any=1e-5 , SCREAMING_SNAKE_CASE__ : str=0.0 , SCREAMING_SNAKE_CASE__ : Tuple=0.02 , SCREAMING_SNAKE_CASE__ : List[Any]=1.0 , SCREAMING_SNAKE_CASE__ : List[Any]=0 , SCREAMING_SNAKE_CASE__ : List[Any]=4_9_4_0_6 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=4_9_4_0_7 , **SCREAMING_SNAKE_CASE__ : Dict , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : List[Any] = vocab_size
__a : Dict = hidden_size
__a : Optional[Any] = intermediate_size
__a : Any = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : Optional[Any] = max_position_embeddings
__a : Dict = hidden_act
__a : str = layer_norm_eps
__a : List[str] = attention_dropout
__a : Optional[int] = initializer_range
__a : List[Any] = initializer_factor
@classmethod
def __lowerCAmelCase ( cls : int , SCREAMING_SNAKE_CASE__ : Union[str, os.PathLike] , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
cls._set_token_in_kwargs(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = cls.get_config_dict(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
# get the text config dict if we are loading from OwlViTConfig
if config_dict.get('model_type' ) == "owlvit":
__a : Dict = config_dict['text_config']
if "model_type" in config_dict and hasattr(cls , 'model_type' ) and config_dict["model_type"] != cls.model_type:
logger.warning(
f'''You are using a model of type {config_dict["model_type"]} to instantiate a model of type '''
f'''{cls.model_type}. This is not supported for all configurations of models and can yield errors.''' )
return cls.from_dict(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : str = '''owlvit_vision_model'''
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : Union[str, Any]=7_6_8 , SCREAMING_SNAKE_CASE__ : Dict=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Dict=1_2 , SCREAMING_SNAKE_CASE__ : List[str]=1_2 , SCREAMING_SNAKE_CASE__ : Dict=3 , SCREAMING_SNAKE_CASE__ : List[Any]=7_6_8 , SCREAMING_SNAKE_CASE__ : int=3_2 , SCREAMING_SNAKE_CASE__ : Tuple="quick_gelu" , SCREAMING_SNAKE_CASE__ : List[Any]=1e-5 , SCREAMING_SNAKE_CASE__ : List[str]=0.0 , SCREAMING_SNAKE_CASE__ : List[str]=0.02 , SCREAMING_SNAKE_CASE__ : Tuple=1.0 , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ )
__a : str = hidden_size
__a : int = intermediate_size
__a : Dict = num_hidden_layers
__a : str = num_attention_heads
__a : Optional[int] = num_channels
__a : Tuple = image_size
__a : str = patch_size
__a : List[str] = hidden_act
__a : Dict = layer_norm_eps
__a : Union[str, Any] = attention_dropout
__a : int = initializer_range
__a : List[str] = initializer_factor
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , SCREAMING_SNAKE_CASE__ : Union[str, os.PathLike] , **SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
cls._set_token_in_kwargs(SCREAMING_SNAKE_CASE__ )
__a , __a : int = cls.get_config_dict(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
# get the vision config dict if we are loading from OwlViTConfig
if config_dict.get('model_type' ) == "owlvit":
__a : List[str] = config_dict['vision_config']
if "model_type" in config_dict and hasattr(cls , 'model_type' ) and config_dict["model_type"] != cls.model_type:
logger.warning(
f'''You are using a model of type {config_dict["model_type"]} to instantiate a model of type '''
f'''{cls.model_type}. This is not supported for all configurations of models and can yield errors.''' )
return cls.from_dict(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Any = '''owlvit'''
__SCREAMING_SNAKE_CASE : str = True
def __init__( self : str , SCREAMING_SNAKE_CASE__ : List[Any]=None , SCREAMING_SNAKE_CASE__ : List[str]=None , SCREAMING_SNAKE_CASE__ : Any=5_1_2 , SCREAMING_SNAKE_CASE__ : Optional[int]=2.6_592 , SCREAMING_SNAKE_CASE__ : Dict=True , **SCREAMING_SNAKE_CASE__ : str , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ )
if text_config is None:
__a : int = {}
logger.info('text_config is None. Initializing the OwlViTTextConfig with default values.' )
if vision_config is None:
__a : Optional[int] = {}
logger.info('vision_config is None. initializing the OwlViTVisionConfig with default values.' )
__a : int = OwlViTTextConfig(**SCREAMING_SNAKE_CASE__ )
__a : str = OwlViTVisionConfig(**SCREAMING_SNAKE_CASE__ )
__a : str = projection_dim
__a : Union[str, Any] = logit_scale_init_value
__a : Union[str, Any] = return_dict
__a : int = 1.0
@classmethod
def __lowerCAmelCase ( cls : Any , SCREAMING_SNAKE_CASE__ : Union[str, os.PathLike] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
cls._set_token_in_kwargs(SCREAMING_SNAKE_CASE__ )
__a , __a : int = cls.get_config_dict(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
if "model_type" in config_dict and hasattr(cls , 'model_type' ) and config_dict["model_type"] != cls.model_type:
logger.warning(
f'''You are using a model of type {config_dict["model_type"]} to instantiate a model of type '''
f'''{cls.model_type}. This is not supported for all configurations of models and can yield errors.''' )
return cls.from_dict(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@classmethod
def __lowerCAmelCase ( cls : Dict , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Union[str, Any] = {}
__a : List[str] = text_config
__a : Tuple = vision_config
return cls.from_dict(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : List[Any] = copy.deepcopy(self.__dict__ )
__a : Optional[Any] = self.text_config.to_dict()
__a : Dict = self.vision_config.to_dict()
__a : Dict = self.__class__.model_type
return output
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return OrderedDict(
[
('input_ids', {0: 'batch', 1: 'sequence'}),
('pixel_values', {0: 'batch', 1: 'num_channels', 2: 'height', 3: 'width'}),
('attention_mask', {0: 'batch', 1: 'sequence'}),
] )
@property
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return OrderedDict(
[
('logits_per_image', {0: 'batch'}),
('logits_per_text', {0: 'batch'}),
('text_embeds', {0: 'batch'}),
('image_embeds', {0: 'batch'}),
] )
@property
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return 1e-4
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : "ProcessorMixin" , SCREAMING_SNAKE_CASE__ : int = -1 , SCREAMING_SNAKE_CASE__ : int = -1 , SCREAMING_SNAKE_CASE__ : Optional["TensorType"] = None , ):
'''simple docstring'''
__a : Tuple = super().generate_dummy_inputs(
processor.tokenizer , batch_size=SCREAMING_SNAKE_CASE__ , seq_length=SCREAMING_SNAKE_CASE__ , framework=SCREAMING_SNAKE_CASE__ )
__a : Dict = super().generate_dummy_inputs(
processor.image_processor , batch_size=SCREAMING_SNAKE_CASE__ , framework=SCREAMING_SNAKE_CASE__ )
return {**text_input_dict, **image_input_dict}
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return 1_4
| 47 |
from ..utils import DummyObject, requires_backends
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Dict , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Tuple , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Any = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Tuple , *SCREAMING_SNAKE_CASE__ : List[str] , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Dict , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Dict = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[int] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Optional[int] , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
| 47 | 1 |
import json
import os
import shutil
import tempfile
from unittest import TestCase
from transformers import BartTokenizer, BartTokenizerFast, DPRQuestionEncoderTokenizer, DPRQuestionEncoderTokenizerFast
from transformers.models.bart.configuration_bart import BartConfig
from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES
from transformers.models.dpr.configuration_dpr import DPRConfig
from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
from transformers.testing_utils import require_faiss, require_tokenizers, require_torch, slow
from transformers.utils import is_datasets_available, is_faiss_available, is_torch_available
if is_torch_available() and is_datasets_available() and is_faiss_available():
from transformers.models.rag.configuration_rag import RagConfig
from transformers.models.rag.tokenization_rag import RagTokenizer
@require_faiss
@require_torch
class _UpperCamelCase( __lowerCamelCase ):
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[Any] = tempfile.mkdtemp()
__a : int = 8
# DPR tok
__a : Dict = [
'[UNK]',
'[CLS]',
'[SEP]',
'[PAD]',
'[MASK]',
'want',
'##want',
'##ed',
'wa',
'un',
'runn',
'##ing',
',',
'low',
'lowest',
]
__a : int = os.path.join(self.tmpdirname , 'dpr_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , DPR_VOCAB_FILES_NAMES['vocab_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as vocab_writer:
vocab_writer.write(''.join([x + '\n' for x in vocab_tokens] ) )
# BART tok
__a : str = [
'l',
'o',
'w',
'e',
'r',
's',
't',
'i',
'd',
'n',
'\u0120',
'\u0120l',
'\u0120n',
'\u0120lo',
'\u0120low',
'er',
'\u0120lowest',
'\u0120newer',
'\u0120wider',
'<unk>',
]
__a : Optional[int] = dict(zip(SCREAMING_SNAKE_CASE__ , range(len(SCREAMING_SNAKE_CASE__ ) ) ) )
__a : List[str] = ['#version: 0.2', '\u0120 l', '\u0120l o', '\u0120lo w', 'e r', '']
__a : List[str] = {'unk_token': '<unk>'}
__a : Dict = os.path.join(self.tmpdirname , 'bart_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['vocab_file'] )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(SCREAMING_SNAKE_CASE__ ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'dpr_tokenizer' ) )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'bart_tokenizer' ) )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
shutil.rmtree(self.tmpdirname )
@require_tokenizers
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Tuple = os.path.join(self.tmpdirname , 'rag_tokenizer' )
__a : Optional[Any] = RagConfig(question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() )
__a : Optional[Any] = RagTokenizer(question_encoder=self.get_dpr_tokenizer() , generator=self.get_bart_tokenizer() )
rag_config.save_pretrained(SCREAMING_SNAKE_CASE__ )
rag_tokenizer.save_pretrained(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = RagTokenizer.from_pretrained(SCREAMING_SNAKE_CASE__ , config=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(new_rag_tokenizer.question_encoder , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.question_encoder.get_vocab() , rag_tokenizer.question_encoder.get_vocab() )
self.assertIsInstance(new_rag_tokenizer.generator , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.generator.get_vocab() , rag_tokenizer.generator.get_vocab() )
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Optional[Any] = RagTokenizer.from_pretrained('facebook/rag-token-nq' )
__a : List[Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : Tuple = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Any = RagTokenizer.from_pretrained('facebook/rag-sequence-nq' )
__a : Union[str, Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : str = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
| 47 |
import math
from datetime import datetime, timedelta
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
__a : Union[str, Any] = year % 1_9
__a : int = year % 4
__a : Optional[int] = year % 7
__a : Dict = math.floor(year / 1_0_0 )
__a : Optional[Any] = math.floor((1_3 + 8 * leap_day_inhibits) / 2_5 )
__a : Union[str, Any] = leap_day_inhibits / 4
__a : str = (
1_5 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 3_0
__a : Union[str, Any] = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
__a : List[Any] = (1_9 * metonic_cycle + secular_moon_shift) % 3_0
# PHM -> Paschal Full Moon
__a : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 2_9 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_9 )
elif days_to_add == 2_8 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_8 )
else:
return datetime(lowerCamelCase_ , 3 , 2_2 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
SCREAMING_SNAKE_CASE__ = '''will be''' if year > datetime.now().year else '''was'''
print(F"Easter in {year} {tense} {gauss_easter(year)}")
| 47 | 1 |
from typing import List, Optional, Union
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''huggingface/informer-tourism-monthly''': (
'''https://huggingface.co/huggingface/informer-tourism-monthly/resolve/main/config.json'''
),
# See all Informer models at https://huggingface.co/models?filter=informer
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = '''informer'''
__SCREAMING_SNAKE_CASE : List[Any] = {
'''hidden_size''': '''d_model''',
'''num_attention_heads''': '''encoder_attention_heads''',
'''num_hidden_layers''': '''encoder_layers''',
}
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : str = "student_t" , SCREAMING_SNAKE_CASE__ : str = "nll" , SCREAMING_SNAKE_CASE__ : int = 1 , SCREAMING_SNAKE_CASE__ : List[int] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[str, bool]] = "mean" , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : int = 6_4 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : str = "gelu" , SCREAMING_SNAKE_CASE__ : float = 0.05 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : int = 1_0_0 , SCREAMING_SNAKE_CASE__ : float = 0.02 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=True , SCREAMING_SNAKE_CASE__ : str = "prob" , SCREAMING_SNAKE_CASE__ : int = 5 , SCREAMING_SNAKE_CASE__ : bool = True , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Dict = prediction_length
__a : Tuple = context_length or prediction_length
__a : Tuple = distribution_output
__a : Tuple = loss
__a : str = input_size
__a : Dict = num_time_features
__a : Optional[int] = lags_sequence if lags_sequence is not None else [1, 2, 3, 4, 5, 6, 7]
__a : str = scaling
__a : Tuple = num_dynamic_real_features
__a : int = num_static_real_features
__a : Dict = num_static_categorical_features
# set cardinality
if cardinality and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The cardinality should be a list of the same length as `num_static_categorical_features`' )
__a : Optional[Any] = cardinality
else:
__a : Optional[int] = [0]
# set embedding_dimension
if embedding_dimension and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The embedding dimension should be a list of the same length as `num_static_categorical_features`' )
__a : int = embedding_dimension
else:
__a : List[Any] = [min(5_0 , (cat + 1) // 2 ) for cat in self.cardinality]
__a : int = num_parallel_samples
# Transformer architecture configuration
__a : str = input_size * len(self.lags_sequence ) + self._number_of_features
__a : Optional[int] = d_model
__a : Union[str, Any] = encoder_attention_heads
__a : int = decoder_attention_heads
__a : Any = encoder_ffn_dim
__a : Union[str, Any] = decoder_ffn_dim
__a : List[Any] = encoder_layers
__a : Optional[int] = decoder_layers
__a : int = dropout
__a : Optional[Any] = attention_dropout
__a : Dict = activation_dropout
__a : Union[str, Any] = encoder_layerdrop
__a : Optional[int] = decoder_layerdrop
__a : List[str] = activation_function
__a : str = init_std
__a : Optional[int] = use_cache
# Informer
__a : Union[str, Any] = attention_type
__a : str = sampling_factor
__a : Dict = distil
super().__init__(is_encoder_decoder=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
return (
sum(self.embedding_dimension )
+ self.num_dynamic_real_features
+ self.num_time_features
+ self.num_static_real_features
+ self.input_size * 2 # the log1p(abs(loc)) and log(scale) features
)
| 47 |
from typing import List, Optional, Union
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''huggingface/informer-tourism-monthly''': (
'''https://huggingface.co/huggingface/informer-tourism-monthly/resolve/main/config.json'''
),
# See all Informer models at https://huggingface.co/models?filter=informer
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = '''informer'''
__SCREAMING_SNAKE_CASE : List[Any] = {
'''hidden_size''': '''d_model''',
'''num_attention_heads''': '''encoder_attention_heads''',
'''num_hidden_layers''': '''encoder_layers''',
}
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : str = "student_t" , SCREAMING_SNAKE_CASE__ : str = "nll" , SCREAMING_SNAKE_CASE__ : int = 1 , SCREAMING_SNAKE_CASE__ : List[int] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[str, bool]] = "mean" , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : int = 6_4 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : str = "gelu" , SCREAMING_SNAKE_CASE__ : float = 0.05 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : int = 1_0_0 , SCREAMING_SNAKE_CASE__ : float = 0.02 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=True , SCREAMING_SNAKE_CASE__ : str = "prob" , SCREAMING_SNAKE_CASE__ : int = 5 , SCREAMING_SNAKE_CASE__ : bool = True , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Dict = prediction_length
__a : Tuple = context_length or prediction_length
__a : Tuple = distribution_output
__a : Tuple = loss
__a : str = input_size
__a : Dict = num_time_features
__a : Optional[int] = lags_sequence if lags_sequence is not None else [1, 2, 3, 4, 5, 6, 7]
__a : str = scaling
__a : Tuple = num_dynamic_real_features
__a : int = num_static_real_features
__a : Dict = num_static_categorical_features
# set cardinality
if cardinality and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The cardinality should be a list of the same length as `num_static_categorical_features`' )
__a : Optional[Any] = cardinality
else:
__a : Optional[int] = [0]
# set embedding_dimension
if embedding_dimension and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The embedding dimension should be a list of the same length as `num_static_categorical_features`' )
__a : int = embedding_dimension
else:
__a : List[Any] = [min(5_0 , (cat + 1) // 2 ) for cat in self.cardinality]
__a : int = num_parallel_samples
# Transformer architecture configuration
__a : str = input_size * len(self.lags_sequence ) + self._number_of_features
__a : Optional[int] = d_model
__a : Union[str, Any] = encoder_attention_heads
__a : int = decoder_attention_heads
__a : Any = encoder_ffn_dim
__a : Union[str, Any] = decoder_ffn_dim
__a : List[Any] = encoder_layers
__a : Optional[int] = decoder_layers
__a : int = dropout
__a : Optional[Any] = attention_dropout
__a : Dict = activation_dropout
__a : Union[str, Any] = encoder_layerdrop
__a : Optional[int] = decoder_layerdrop
__a : List[str] = activation_function
__a : str = init_std
__a : Optional[int] = use_cache
# Informer
__a : Union[str, Any] = attention_type
__a : str = sampling_factor
__a : Dict = distil
super().__init__(is_encoder_decoder=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
return (
sum(self.embedding_dimension )
+ self.num_dynamic_real_features
+ self.num_time_features
+ self.num_static_real_features
+ self.input_size * 2 # the log1p(abs(loc)) and log(scale) features
)
| 47 | 1 |
import json
from typing import TYPE_CHECKING, List, Optional, Tuple
from tokenizers import pre_tokenizers
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import logging
if TYPE_CHECKING:
from transformers.pipelines.conversational import Conversation
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''vocab.json''', '''merges_file''': '''merges.txt''', '''tokenizer_file''': '''tokenizer.json'''}
SCREAMING_SNAKE_CASE__ = {
'''tokenizer_file''': {
'''EleutherAI/gpt-neox-20b''': '''https://huggingface.co/EleutherAI/gpt-neox-20b/resolve/main/tokenizer.json''',
},
}
SCREAMING_SNAKE_CASE__ = {
'''gpt-neox-20b''': 2048,
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : str = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : List[Any] = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : Any = ['''input_ids''', '''attention_mask''']
def __init__( self : str , SCREAMING_SNAKE_CASE__ : Tuple=None , SCREAMING_SNAKE_CASE__ : Any=None , SCREAMING_SNAKE_CASE__ : List[str]=None , SCREAMING_SNAKE_CASE__ : Dict="<|endoftext|>" , SCREAMING_SNAKE_CASE__ : Tuple="<|endoftext|>" , SCREAMING_SNAKE_CASE__ : Any="<|endoftext|>" , SCREAMING_SNAKE_CASE__ : int=False , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
super().__init__(
SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , tokenizer_file=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , add_prefix_space=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ , )
__a : Optional[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get('add_prefix_space' , SCREAMING_SNAKE_CASE__ ) != add_prefix_space:
__a : Any = getattr(SCREAMING_SNAKE_CASE__ , pre_tok_state.pop('type' ) )
__a : int = add_prefix_space
__a : str = pre_tok_class(**SCREAMING_SNAKE_CASE__ )
__a : int = add_prefix_space
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
__a : Union[str, Any] = self._tokenizer.model.save(SCREAMING_SNAKE_CASE__ , name=SCREAMING_SNAKE_CASE__ )
return tuple(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : "Conversation" ):
'''simple docstring'''
__a : List[str] = []
for is_user, text in conversation.iter_texts():
input_ids.extend(self.encode(SCREAMING_SNAKE_CASE__ , add_special_tokens=SCREAMING_SNAKE_CASE__ ) + [self.eos_token_id] )
if len(SCREAMING_SNAKE_CASE__ ) > self.model_max_length:
__a : Optional[int] = input_ids[-self.model_max_length :]
return input_ids
| 47 |
import torch
from diffusers import DDIMParallelScheduler
from .test_schedulers import SchedulerCommonTest
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = (DDIMParallelScheduler,)
__SCREAMING_SNAKE_CASE : Union[str, Any] = (('''eta''', 0.0), ('''num_inference_steps''', 50))
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : List[Any] = {
'num_train_timesteps': 1_0_0_0,
'beta_start': 0.0_001,
'beta_end': 0.02,
'beta_schedule': 'linear',
'clip_sample': True,
}
config.update(**SCREAMING_SNAKE_CASE__ )
return config
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Tuple = self.scheduler_classes[0]
__a : Optional[Any] = self.get_scheduler_config(**SCREAMING_SNAKE_CASE__ )
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : List[str] = 1_0, 0.0
__a : Dict = self.dummy_model()
__a : str = self.dummy_sample_deter
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
for t in scheduler.timesteps:
__a : str = model(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[str] = scheduler.step(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ).prev_sample
return sample
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
for timesteps in [1_0_0, 5_0_0, 1_0_0_0]:
self.check_over_configs(num_train_timesteps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
for steps_offset in [0, 1]:
self.check_over_configs(steps_offset=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config(steps_offset=1 )
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
scheduler.set_timesteps(5 )
assert torch.equal(scheduler.timesteps , torch.LongTensor([8_0_1, 6_0_1, 4_0_1, 2_0_1, 1] ) )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for beta_start, beta_end in zip([0.0_001, 0.001, 0.01, 0.1] , [0.002, 0.02, 0.2, 2] ):
self.check_over_configs(beta_start=SCREAMING_SNAKE_CASE__ , beta_end=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for schedule in ["linear", "squaredcos_cap_v2"]:
self.check_over_configs(beta_schedule=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(prediction_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for clip_sample in [True, False]:
self.check_over_configs(clip_sample=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for timestep_spacing in ["trailing", "leading"]:
self.check_over_configs(timestep_spacing=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for rescale_betas_zero_snr in [True, False]:
self.check_over_configs(rescale_betas_zero_snr=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
self.check_over_configs(thresholding=SCREAMING_SNAKE_CASE__ )
for threshold in [0.5, 1.0, 2.0]:
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(
thresholding=SCREAMING_SNAKE_CASE__ , prediction_type=SCREAMING_SNAKE_CASE__ , sample_max_value=SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for t in [1, 1_0, 4_9]:
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
for t, num_inference_steps in zip([1, 1_0, 5_0] , [1_0, 5_0, 5_0_0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , num_inference_steps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for t, eta in zip([1, 1_0, 4_9] , [0.0, 0.5, 1.0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , eta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : Union[str, Any] = self.get_scheduler_config()
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_2_0 , 4_0_0 ) - 0.14_771 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_8_0 , 9_6_0 ) - 0.32_460 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_8_7 , 4_8_6 ) - 0.00_979 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_9_9 , 9_9_8 ) - 0.02 ) ) < 1e-5
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config()
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : Any = 1_0, 0.0
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = self.dummy_model()
__a : int = self.dummy_sample_deter
__a : List[Any] = self.dummy_sample_deter + 0.1
__a : List[str] = self.dummy_sample_deter - 0.1
__a : Optional[Any] = samplea.shape[0]
__a : Optional[Any] = torch.stack([samplea, samplea, samplea] , dim=0 )
__a : Union[str, Any] = torch.arange(SCREAMING_SNAKE_CASE__ )[0:3, None].repeat(1 , SCREAMING_SNAKE_CASE__ )
__a : int = model(samples.flatten(0 , 1 ) , timesteps.flatten(0 , 1 ) )
__a : int = scheduler.batch_step_no_noise(SCREAMING_SNAKE_CASE__ , timesteps.flatten(0 , 1 ) , samples.flatten(0 , 1 ) , SCREAMING_SNAKE_CASE__ )
__a : Dict = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 1_147.7_904 ) < 1e-2
assert abs(result_mean.item() - 0.4_982 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : List[str] = self.full_loop()
__a : Tuple = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : int = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 172.0_067 ) < 1e-2
assert abs(result_mean.item() - 0.223_967 ) < 1e-3
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Optional[int] = self.full_loop(prediction_type='v_prediction' )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Union[str, Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 52.5_302 ) < 1e-2
assert abs(result_mean.item() - 0.0_684 ) < 1e-3
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Union[str, Any] = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : Optional[int] = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[int] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.8_295 ) < 1e-2
assert abs(result_mean.item() - 0.1_951 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Dict = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Tuple = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.0_784 ) < 1e-2
assert abs(result_mean.item() - 0.1_941 ) < 1e-3
| 47 | 1 |
import logging
import re
import pytorch_quantization
import pytorch_quantization.nn as quant_nn
import torch
from pytorch_quantization import calib
from pytorch_quantization.tensor_quant import QuantDescriptor
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
SCREAMING_SNAKE_CASE__ = 50 # max width of layer names
SCREAMING_SNAKE_CASE__ = 70 # max width of quantizer names
def UpperCAmelCase__ ( lowerCamelCase_ : Dict ):
__a : int = parser.add_argument_group('quant_trainer arguments' )
group.add_argument('--wprec' , type=lowerCamelCase_ , default=8 , help='weight precision' )
group.add_argument('--aprec' , type=lowerCamelCase_ , default=8 , help='activation precision' )
group.add_argument('--quant-per-tensor' , action='store_true' , help='per tensor weight scaling' )
group.add_argument('--quant-disable' , action='store_true' , help='disable all quantizers' )
group.add_argument('--quant-disable-embeddings' , action='store_true' , help='disable all embeddings quantizers' )
group.add_argument('--quant-disable-keyword' , type=lowerCamelCase_ , nargs='+' , help='disable quantizers by keyword' )
group.add_argument('--quant-disable-layer-module' , type=lowerCamelCase_ , help='disable quantizers by keyword under layer.' )
group.add_argument('--quant-enable-layer-module' , type=lowerCamelCase_ , help='enable quantizers by keyword under layer' )
group.add_argument('--calibrator' , default='max' , help='which quantization range calibrator to use' )
group.add_argument('--percentile' , default=lowerCamelCase_ , type=lowerCamelCase_ , help='percentile for PercentileCalibrator' )
group.add_argument('--fuse-qkv' , action='store_true' , help='use the same scale factor for qkv' )
group.add_argument('--clip-gelu' , metavar='N' , type=lowerCamelCase_ , help='clip gelu output maximum value to N' )
group.add_argument(
'--recalibrate-weights' , action='store_true' , help=(
'recalibrate weight amaxes by taking the max of the weights.'
' amaxes will be computed with the current quantization granularity (axis).'
) , )
def UpperCAmelCase__ ( lowerCamelCase_ : Dict ):
if args.calibrator == "max":
__a : str = 'max'
elif args.calibrator == "percentile":
if args.percentile is None:
raise ValueError('Specify --percentile when using percentile calibrator' )
__a : int = 'histogram'
elif args.calibrator == "mse":
__a : List[str] = 'histogram'
else:
raise ValueError(f'''Invalid calibrator {args.calibrator}''' )
__a : Optional[Any] = QuantDescriptor(num_bits=args.aprec , calib_method=lowerCamelCase_ )
__a : Optional[Any] = QuantDescriptor(num_bits=args.wprec , axis=(None if args.quant_per_tensor else (0,)) )
quant_nn.QuantLinear.set_default_quant_desc_input(lowerCamelCase_ )
quant_nn.QuantLinear.set_default_quant_desc_weight(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Optional[Any]=False , lowerCamelCase_ : Dict=False ):
logger.info('Configuring Model for Quantization' )
logger.info(f'''using quantization package {pytorch_quantization.__file__}''' )
if not calib:
if args.quant_disable_embeddings:
set_quantizer_by_name(lowerCamelCase_ , ['embeddings'] , which='weight' , _disabled=lowerCamelCase_ )
if args.quant_disable:
set_quantizer_by_name(lowerCamelCase_ , [''] , _disabled=lowerCamelCase_ )
if args.quant_disable_keyword:
set_quantizer_by_name(lowerCamelCase_ , args.quant_disable_keyword , _disabled=lowerCamelCase_ )
if args.quant_disable_layer_module:
set_quantizer_by_name(lowerCamelCase_ , [R'layer.\d+.' + args.quant_disable_layer_module] , _disabled=lowerCamelCase_ )
if args.quant_enable_layer_module:
set_quantizer_by_name(lowerCamelCase_ , [R'layer.\d+.' + args.quant_enable_layer_module] , _disabled=lowerCamelCase_ )
if args.recalibrate_weights:
recalibrate_weights(lowerCamelCase_ )
if args.fuse_qkv:
fuse_qkv(lowerCamelCase_ , lowerCamelCase_ )
if args.clip_gelu:
clip_gelu(lowerCamelCase_ , args.clip_gelu )
# if args.local_rank in [-1, 0] and not calib:
print_quant_summary(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
logger.info('Enabling Calibration' )
for name, module in model.named_modules():
if name.endswith('_quantizer' ):
if module._calibrator is not None:
module.disable_quant()
module.enable_calib()
else:
module.disable()
logger.info(f'''{name:80}: {module}''' )
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Dict ):
logger.info('Loading calibrated amax' )
for name, module in model.named_modules():
if name.endswith('_quantizer' ):
if module._calibrator is not None:
if isinstance(module._calibrator , calib.MaxCalibrator ):
module.load_calib_amax()
else:
module.load_calib_amax('percentile' , percentile=args.percentile )
module.enable_quant()
module.disable_calib()
else:
module.enable()
model.cuda()
print_quant_summary(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] ):
def fusea(lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Optional[int] ):
for mod in [qq, qk, qv]:
if not hasattr(lowerCamelCase_ , '_amax' ):
print(' WARNING: NO AMAX BUFFER' )
return
__a : Any = qq._amax.detach().item()
__a : Union[str, Any] = qk._amax.detach().item()
__a : int = qv._amax.detach().item()
__a : List[Any] = max(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
qq._amax.fill_(lowerCamelCase_ )
qk._amax.fill_(lowerCamelCase_ )
qv._amax.fill_(lowerCamelCase_ )
logger.info(f''' q={q:5.2f} k={k:5.2f} v={v:5.2f} -> {amax:5.2f}''' )
for name, mod in model.named_modules():
if name.endswith('.attention.self' ):
logger.info(f'''FUSE_QKV: {name:{name_width}}''' )
fusea(mod.matmul_q_input_quantizer , mod.matmul_k_input_quantizer , mod.matmul_v_input_quantizer )
if args.quant_per_tensor:
fusea(mod.query._weight_quantizer , mod.key._weight_quantizer , mod.value._weight_quantizer )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : str ):
for name, mod in model.named_modules():
if name.endswith('.output.dense' ) and not name.endswith('attention.output.dense' ):
__a : Dict = mod._input_quantizer._amax.data.detach().item()
mod._input_quantizer._amax.data.detach().clamp_(max=lowerCamelCase_ )
__a : List[str] = mod._input_quantizer._amax.data.detach().item()
logger.info(f'''CLIP_GELU: {name:{name_width}} amax: {amax_init:5.2f} -> {amax:5.2f}''' )
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] ):
for name, mod in model.named_modules():
if hasattr(lowerCamelCase_ , '_weight_quantizer' ) and mod._weight_quantizer.axis is not None:
__a : Optional[int] = mod.weight.shape[0]
__a : List[str] = mod._weight_quantizer._amax.detach()
__a : List[Any] = torch.ones(lowerCamelCase_ , dtype=amax.dtype , device=amax.device ) * amax
print(f'''expanding {name} {amax} -> {mod._weight_quantizer._amax}''' )
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple ):
for name, mod in model.named_modules():
if hasattr(lowerCamelCase_ , '_weight_quantizer' ):
if not hasattr(mod.weight_quantizer , '_amax' ):
print('RECALIB: {name:{name_width}} WARNING: NO AMAX BUFFER' )
continue
# determine which axes to reduce across
# e.g. a 4D tensor quantized per axis 0 should reduce over (1,2,3)
__a : Tuple = set() if mod._weight_quantizer.axis is None else set(mod._weight_quantizer.axis )
__a : str = set(range(len(mod.weight.size() ) ) ) - axis_set
__a : Union[str, Any] = pytorch_quantization.utils.reduce_amax(mod.weight , axis=lowerCamelCase_ , keepdims=lowerCamelCase_ ).detach()
logger.info(f'''RECALIB: {name:{name_width}} {mod._weight_quantizer._amax.flatten()} -> {amax.flatten()}''' )
__a : Union[str, Any] = amax
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Optional[Any]=2_5 , lowerCamelCase_ : Dict=1_8_0 , lowerCamelCase_ : int=None ):
if ignore is None:
__a : Any = []
elif not isinstance(lowerCamelCase_ , lowerCamelCase_ ):
__a : int = [ignore]
__a : Optional[int] = 0
for name, mod in model.named_modules():
if not hasattr(lowerCamelCase_ , 'weight' ):
continue
__a : List[str] = max(lowerCamelCase_ , len(lowerCamelCase_ ) )
for name, mod in model.named_modules():
__a : Optional[Any] = getattr(lowerCamelCase_ , '_input_quantizer' , lowerCamelCase_ )
__a : int = getattr(lowerCamelCase_ , '_weight_quantizer' , lowerCamelCase_ )
if not hasattr(lowerCamelCase_ , 'weight' ):
continue
if type(lowerCamelCase_ ) in ignore:
continue
if [True for s in ignore if type(lowerCamelCase_ ) is str and s in name]:
continue
__a : Any = f'''Act:{input_q.extra_repr()}'''
__a : str = f'''Wgt:{weight_q.extra_repr()}'''
__a : Optional[int] = f'''{name:{name_width}} {act_str} {wgt_str}'''
if len(lowerCamelCase_ ) <= line_width:
logger.info(lowerCamelCase_ )
else:
logger.info(f'''{name:{name_width}} {act_str}''' )
logger.info(f'''{" ":{name_width}} {wgt_str}''' )
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] ):
__a : Optional[int] = 0
for name, mod in model.named_modules():
if isinstance(lowerCamelCase_ , pytorch_quantization.nn.TensorQuantizer ):
print(f'''{name:80} {mod}''' )
count += 1
print(f'''{count} TensorQuantizers found in model''' )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[Any] , lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] , lowerCamelCase_ : str ):
__a : List[Any] = getattr(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if quantizer_mod is not None:
assert hasattr(lowerCamelCase_ , lowerCamelCase_ )
setattr(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
else:
logger.warning(f'''{name} has no {quantizer}''' )
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Any="both" , **lowerCamelCase_ : Any ):
__a : Union[str, Any] = f'''Warning: changing {which} quantizers of {name:{qname_width}}'''
for k, v in kwargs.items():
s += f''' {k}={v}'''
if which in ["input", "both"]:
set_quantizer(lowerCamelCase_ , lowerCamelCase_ , '_input_quantizer' , lowerCamelCase_ , lowerCamelCase_ )
if which in ["weight", "both"]:
set_quantizer(lowerCamelCase_ , lowerCamelCase_ , '_weight_quantizer' , lowerCamelCase_ , lowerCamelCase_ )
logger.info(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : Tuple , **lowerCamelCase_ : List[str] ):
for name, mod in model.named_modules():
if hasattr(lowerCamelCase_ , '_input_quantizer' ) or hasattr(lowerCamelCase_ , '_weight_quantizer' ):
for n in names:
if re.search(lowerCamelCase_ , lowerCamelCase_ ):
set_quantizers(lowerCamelCase_ , lowerCamelCase_ , **lowerCamelCase_ )
elif name.endswith('_quantizer' ):
for n in names:
if re.search(lowerCamelCase_ , lowerCamelCase_ ):
__a : Dict = f'''Warning: changing {name:{name_width}}'''
for k, v in kwargs.items():
s += f''' {k}={v}'''
setattr(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
logger.info(lowerCamelCase_ )
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : list[int] ):
# Check if the input is valid
if not len(lowerCamelCase_ ) == len(lowerCamelCase_ ) == 3:
raise ValueError('Please enter a valid equation.' )
if equationa[0] == equationa[1] == equationa[0] == equationa[1] == 0:
raise ValueError('Both a & b of two equations can\'t be zero.' )
# Extract the coefficients
__a , __a , __a : Optional[Any] = equationa
__a , __a , __a : Optional[int] = equationa
# Calculate the determinants of the matrices
__a : str = aa * ba - aa * ba
__a : Tuple = ca * ba - ca * ba
__a : Union[str, Any] = aa * ca - aa * ca
# Check if the system of linear equations has a solution (using Cramer's rule)
if determinant == 0:
if determinant_x == determinant_y == 0:
raise ValueError('Infinite solutions. (Consistent system)' )
else:
raise ValueError('No solution. (Inconsistent system)' )
else:
if determinant_x == determinant_y == 0:
# Trivial solution (Inconsistent system)
return (0.0, 0.0)
else:
__a : Any = determinant_x / determinant
__a : Optional[Any] = determinant_y / determinant
# Non-Trivial Solution (Consistent system)
return (x, y)
| 47 | 1 |
from tempfile import TemporaryDirectory
from unittest import TestCase
from unittest.mock import MagicMock, patch
from transformers import AutoModel, TFAutoModel
from transformers.onnx import FeaturesManager
from transformers.testing_utils import SMALL_MODEL_IDENTIFIER, require_tf, require_torch
@require_torch
@require_tf
class _UpperCamelCase( __lowerCamelCase ):
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Dict = SMALL_MODEL_IDENTIFIER
__a : List[str] = 'pt'
__a : int = 'tf'
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Union[str, Any] = AutoModel.from_pretrained(self.test_model )
model_pt.save_pretrained(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : List[str] = TFAutoModel.from_pretrained(self.test_model , from_pt=SCREAMING_SNAKE_CASE__ )
model_tf.save_pretrained(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Tuple = 'mock_framework'
# Framework provided - return whatever the user provides
__a : Any = FeaturesManager.determine_framework(self.test_model , SCREAMING_SNAKE_CASE__ )
self.assertEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# Local checkpoint and framework provided - return provided framework
# PyTorch checkpoint
with TemporaryDirectory() as local_pt_ckpt:
self._setup_pt_ckpt(SCREAMING_SNAKE_CASE__ )
__a : Dict = FeaturesManager.determine_framework(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# TensorFlow checkpoint
with TemporaryDirectory() as local_tf_ckpt:
self._setup_tf_ckpt(SCREAMING_SNAKE_CASE__ )
__a : int = FeaturesManager.determine_framework(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
with TemporaryDirectory() as local_pt_ckpt:
self._setup_pt_ckpt(SCREAMING_SNAKE_CASE__ )
__a : List[str] = FeaturesManager.determine_framework(SCREAMING_SNAKE_CASE__ )
self.assertEqual(SCREAMING_SNAKE_CASE__ , self.framework_pt )
# TensorFlow checkpoint
with TemporaryDirectory() as local_tf_ckpt:
self._setup_tf_ckpt(SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = FeaturesManager.determine_framework(SCREAMING_SNAKE_CASE__ )
self.assertEqual(SCREAMING_SNAKE_CASE__ , self.framework_tf )
# Invalid local checkpoint
with TemporaryDirectory() as local_invalid_ckpt:
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
__a : int = FeaturesManager.determine_framework(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : List[str] = MagicMock(return_value=SCREAMING_SNAKE_CASE__ )
with patch('transformers.onnx.features.is_tf_available' , SCREAMING_SNAKE_CASE__ ):
__a : Any = FeaturesManager.determine_framework(self.test_model )
self.assertEqual(SCREAMING_SNAKE_CASE__ , self.framework_pt )
# PyTorch not in environment -> use TensorFlow
__a : int = MagicMock(return_value=SCREAMING_SNAKE_CASE__ )
with patch('transformers.onnx.features.is_torch_available' , SCREAMING_SNAKE_CASE__ ):
__a : Dict = FeaturesManager.determine_framework(self.test_model )
self.assertEqual(SCREAMING_SNAKE_CASE__ , self.framework_tf )
# Both in environment -> use PyTorch
__a : Any = MagicMock(return_value=SCREAMING_SNAKE_CASE__ )
__a : Tuple = MagicMock(return_value=SCREAMING_SNAKE_CASE__ )
with patch('transformers.onnx.features.is_tf_available' , SCREAMING_SNAKE_CASE__ ), patch(
'transformers.onnx.features.is_torch_available' , SCREAMING_SNAKE_CASE__ ):
__a : List[str] = FeaturesManager.determine_framework(self.test_model )
self.assertEqual(SCREAMING_SNAKE_CASE__ , self.framework_pt )
# Both not in environment -> raise error
__a : Optional[Any] = MagicMock(return_value=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = MagicMock(return_value=SCREAMING_SNAKE_CASE__ )
with patch('transformers.onnx.features.is_tf_available' , SCREAMING_SNAKE_CASE__ ), patch(
'transformers.onnx.features.is_torch_available' , SCREAMING_SNAKE_CASE__ ):
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
__a : Dict = FeaturesManager.determine_framework(self.test_model )
| 47 |
from ...utils import is_note_seq_available, is_transformers_available, is_torch_available
from ...utils import OptionalDependencyNotAvailable
try:
if not (is_transformers_available() and is_torch_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_torch_and_transformers_objects import * # noqa F403
else:
from .notes_encoder import SpectrogramNotesEncoder
from .continous_encoder import SpectrogramContEncoder
from .pipeline_spectrogram_diffusion import (
SpectrogramContEncoder,
SpectrogramDiffusionPipeline,
TaFilmDecoder,
)
try:
if not (is_transformers_available() and is_torch_available() and is_note_seq_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_transformers_and_torch_and_note_seq_objects import * # noqa F403
else:
from .midi_utils import MidiProcessor
| 47 | 1 |
print((lambda quine: quine % quine)('''print((lambda quine: quine %% quine)(%r))'''))
| 47 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_bridgetower''': [
'''BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''BridgeTowerConfig''',
'''BridgeTowerTextConfig''',
'''BridgeTowerVisionConfig''',
],
'''processing_bridgetower''': ['''BridgeTowerProcessor'''],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''BridgeTowerImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''BridgeTowerForContrastiveLearning''',
'''BridgeTowerForImageAndTextRetrieval''',
'''BridgeTowerForMaskedLM''',
'''BridgeTowerModel''',
'''BridgeTowerPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_bridgetower import (
BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP,
BridgeTowerConfig,
BridgeTowerTextConfig,
BridgeTowerVisionConfig,
)
from .processing_bridgetower import BridgeTowerProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_bridgetower import BridgeTowerImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_bridgetower import (
BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST,
BridgeTowerForContrastiveLearning,
BridgeTowerForImageAndTextRetrieval,
BridgeTowerForMaskedLM,
BridgeTowerModel,
BridgeTowerPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure)
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import (
OptionalDependencyNotAvailable,
_LazyModule,
is_flax_available,
is_tf_available,
is_torch_available,
)
SCREAMING_SNAKE_CASE__ = {
'''configuration_vision_encoder_decoder''': ['''VisionEncoderDecoderConfig''', '''VisionEncoderDecoderOnnxConfig''']
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''VisionEncoderDecoderModel''']
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''TFVisionEncoderDecoderModel''']
try:
if not is_flax_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''FlaxVisionEncoderDecoderModel''']
if TYPE_CHECKING:
from .configuration_vision_encoder_decoder import VisionEncoderDecoderConfig, VisionEncoderDecoderOnnxConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_vision_encoder_decoder import VisionEncoderDecoderModel
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_tf_vision_encoder_decoder import TFVisionEncoderDecoderModel
try:
if not is_flax_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_flax_vision_encoder_decoder import FlaxVisionEncoderDecoderModel
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
from string import ascii_lowercase, ascii_uppercase
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
if not sentence:
return ""
__a : Union[str, Any] = dict(zip(lowerCamelCase_ , lowerCamelCase_ ) )
return lower_to_upper.get(sentence[0] , sentence[0] ) + sentence[1:]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 47 | 1 |
import unittest
import numpy as np
import timeout_decorator # noqa
from transformers import BlenderbotConfig, is_flax_available
from transformers.testing_utils import jax_device, require_flax, slow
from ...generation.test_flax_utils import FlaxGenerationTesterMixin
from ...test_modeling_flax_common import FlaxModelTesterMixin, ids_tensor
if is_flax_available():
import os
# The slow tests are often failing with OOM error on GPU
# This makes JAX allocate exactly what is needed on demand, and deallocate memory that is no longer needed
# but will be slower as stated here https://jax.readthedocs.io/en/latest/gpu_memory_allocation.html
SCREAMING_SNAKE_CASE__ = '''platform'''
import jax
import jax.numpy as jnp
from transformers import BlenderbotTokenizer
from transformers.models.blenderbot.modeling_flax_blenderbot import (
FlaxBlenderbotForConditionalGeneration,
FlaxBlenderbotModel,
shift_tokens_right,
)
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : Any , lowerCamelCase_ : List[str]=None , lowerCamelCase_ : Tuple=None , lowerCamelCase_ : List[str]=None , lowerCamelCase_ : Dict=None , lowerCamelCase_ : Union[str, Any]=None , lowerCamelCase_ : Any=None , ):
if attention_mask is None:
__a : Union[str, Any] = np.where(input_ids != config.pad_token_id , 1 , 0 )
if decoder_attention_mask is None:
__a : Dict = np.where(decoder_input_ids != config.pad_token_id , 1 , 0 )
if head_mask is None:
__a : List[str] = np.ones((config.encoder_layers, config.encoder_attention_heads) )
if decoder_head_mask is None:
__a : Dict = np.ones((config.decoder_layers, config.decoder_attention_heads) )
if cross_attn_head_mask is None:
__a : Optional[Any] = np.ones((config.decoder_layers, config.decoder_attention_heads) )
return {
"input_ids": input_ids,
"decoder_input_ids": decoder_input_ids,
"attention_mask": attention_mask,
"decoder_attention_mask": attention_mask,
}
class _UpperCamelCase:
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Optional[int]=1_3 , SCREAMING_SNAKE_CASE__ : str=7 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=True , SCREAMING_SNAKE_CASE__ : Optional[int]=False , SCREAMING_SNAKE_CASE__ : Union[str, Any]=9_9 , SCREAMING_SNAKE_CASE__ : str=1_6 , SCREAMING_SNAKE_CASE__ : Optional[Any]=2 , SCREAMING_SNAKE_CASE__ : Tuple=4 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=4 , SCREAMING_SNAKE_CASE__ : List[Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.1 , SCREAMING_SNAKE_CASE__ : Tuple=3_2 , SCREAMING_SNAKE_CASE__ : Optional[int]=2 , SCREAMING_SNAKE_CASE__ : List[Any]=1 , SCREAMING_SNAKE_CASE__ : List[Any]=0 , SCREAMING_SNAKE_CASE__ : List[str]=0.02 , ):
'''simple docstring'''
__a : Dict = parent
__a : Optional[int] = batch_size
__a : int = seq_length
__a : Any = is_training
__a : Tuple = use_labels
__a : Dict = vocab_size
__a : List[str] = hidden_size
__a : Optional[Any] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : Optional[Any] = intermediate_size
__a : List[str] = hidden_act
__a : Tuple = hidden_dropout_prob
__a : Dict = attention_probs_dropout_prob
__a : Dict = max_position_embeddings
__a : str = eos_token_id
__a : str = pad_token_id
__a : Optional[Any] = bos_token_id
__a : Union[str, Any] = initializer_range
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Optional[Any] = np.clip(ids_tensor([self.batch_size, self.seq_length - 1] , self.vocab_size ) , 3 , self.vocab_size )
__a : Dict = np.concatenate((input_ids, 2 * np.ones((self.batch_size, 1) , dtype=np.intaa )) , -1 )
__a : Optional[Any] = shift_tokens_right(SCREAMING_SNAKE_CASE__ , 1 , 2 )
__a : str = BlenderbotConfig(
vocab_size=self.vocab_size , d_model=self.hidden_size , encoder_layers=self.num_hidden_layers , decoder_layers=self.num_hidden_layers , encoder_attention_heads=self.num_attention_heads , decoder_attention_heads=self.num_attention_heads , encoder_ffn_dim=self.intermediate_size , decoder_ffn_dim=self.intermediate_size , dropout=self.hidden_dropout_prob , attention_dropout=self.attention_probs_dropout_prob , max_position_embeddings=self.max_position_embeddings , eos_token_id=self.eos_token_id , bos_token_id=self.bos_token_id , pad_token_id=self.pad_token_id , initializer_range=self.initializer_range , use_cache=SCREAMING_SNAKE_CASE__ , )
__a : Optional[int] = prepare_blenderbot_inputs_dict(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return config, inputs_dict
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a , __a : Tuple = self.prepare_config_and_inputs()
return config, inputs_dict
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
__a : Union[str, Any] = 2_0
__a : int = model_class_name(SCREAMING_SNAKE_CASE__ )
__a : str = model.encode(inputs_dict['input_ids'] )
__a , __a : Tuple = (
inputs_dict['decoder_input_ids'],
inputs_dict['decoder_attention_mask'],
)
__a : List[str] = model.init_cache(decoder_input_ids.shape[0] , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = jnp.ones((decoder_input_ids.shape[0], max_decoder_length) , dtype='i4' )
__a : Dict = jnp.broadcast_to(
jnp.arange(decoder_input_ids.shape[-1] - 1 )[None, :] , (decoder_input_ids.shape[0], decoder_input_ids.shape[-1] - 1) , )
__a : Dict = model.decode(
decoder_input_ids[:, :-1] , SCREAMING_SNAKE_CASE__ , decoder_attention_mask=SCREAMING_SNAKE_CASE__ , past_key_values=SCREAMING_SNAKE_CASE__ , decoder_position_ids=SCREAMING_SNAKE_CASE__ , )
__a : Any = jnp.array(decoder_input_ids.shape[0] * [[decoder_input_ids.shape[-1] - 1]] , dtype='i4' )
__a : Dict = model.decode(
decoder_input_ids[:, -1:] , SCREAMING_SNAKE_CASE__ , decoder_attention_mask=SCREAMING_SNAKE_CASE__ , past_key_values=outputs_cache.past_key_values , decoder_position_ids=SCREAMING_SNAKE_CASE__ , )
__a : Union[str, Any] = model.decode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = np.max(np.abs((outputs_cache_next[0][:, -1, :5] - outputs[0][:, -1, :5]) ) )
self.parent.assertTrue(diff < 1e-3 , msg=f'''Max diff is {diff}''' )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Optional[Any] = 2_0
__a : List[str] = model_class_name(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = model.encode(inputs_dict['input_ids'] )
__a , __a : Any = (
inputs_dict['decoder_input_ids'],
inputs_dict['decoder_attention_mask'],
)
__a : Any = jnp.concatenate(
[
decoder_attention_mask,
jnp.zeros((decoder_attention_mask.shape[0], max_decoder_length - decoder_attention_mask.shape[1]) ),
] , axis=-1 , )
__a : Dict = model.init_cache(decoder_input_ids.shape[0] , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = jnp.broadcast_to(
jnp.arange(decoder_input_ids.shape[-1] - 1 )[None, :] , (decoder_input_ids.shape[0], decoder_input_ids.shape[-1] - 1) , )
__a : List[str] = model.decode(
decoder_input_ids[:, :-1] , SCREAMING_SNAKE_CASE__ , decoder_attention_mask=SCREAMING_SNAKE_CASE__ , past_key_values=SCREAMING_SNAKE_CASE__ , decoder_position_ids=SCREAMING_SNAKE_CASE__ , )
__a : Tuple = jnp.array(decoder_input_ids.shape[0] * [[decoder_input_ids.shape[-1] - 1]] , dtype='i4' )
__a : int = model.decode(
decoder_input_ids[:, -1:] , SCREAMING_SNAKE_CASE__ , past_key_values=outputs_cache.past_key_values , decoder_attention_mask=SCREAMING_SNAKE_CASE__ , decoder_position_ids=SCREAMING_SNAKE_CASE__ , )
__a : Dict = model.decode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , decoder_attention_mask=SCREAMING_SNAKE_CASE__ )
__a : List[str] = np.max(np.abs((outputs_cache_next[0][:, -1, :5] - outputs[0][:, -1, :5]) ) )
self.parent.assertTrue(diff < 1e-3 , msg=f'''Max diff is {diff}''' )
@require_flax
class _UpperCamelCase( unittest.TestCase ):
__SCREAMING_SNAKE_CASE : List[Any] = 99
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : Dict = np.array(
[
[7_1, 8_2, 1_8, 3_3, 4_6, 9_1, 2],
[6_8, 3_4, 2_6, 5_8, 3_0, 8_2, 2],
[5, 9_7, 1_7, 3_9, 9_4, 4_0, 2],
[7_6, 8_3, 9_4, 2_5, 7_0, 7_8, 2],
[8_7, 5_9, 4_1, 3_5, 4_8, 6_6, 2],
[5_5, 1_3, 1_6, 5_8, 5, 2, 1], # note padding
[6_4, 2_7, 3_1, 5_1, 1_2, 7_5, 2],
[5_2, 6_4, 8_6, 1_7, 8_3, 3_9, 2],
[4_8, 6_1, 9, 2_4, 7_1, 8_2, 2],
[2_6, 1, 6_0, 4_8, 2_2, 1_3, 2],
[2_1, 5, 6_2, 2_8, 1_4, 7_6, 2],
[4_5, 9_8, 3_7, 8_6, 5_9, 4_8, 2],
[7_0, 7_0, 5_0, 9, 2_8, 0, 2],
] , dtype=np.intaa , )
__a : Any = input_ids.shape[0]
__a : List[Any] = BlenderbotConfig(
vocab_size=self.vocab_size , d_model=2_4 , encoder_layers=2 , decoder_layers=2 , encoder_attention_heads=2 , decoder_attention_heads=2 , encoder_ffn_dim=3_2 , decoder_ffn_dim=3_2 , max_position_embeddings=4_8 , eos_token_id=2 , pad_token_id=1 , bos_token_id=0 , )
return config, input_ids, batch_size
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a , __a , __a : List[Any] = self._get_config_and_data()
__a : str = FlaxBlenderbotForConditionalGeneration(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = lm_model(input_ids=SCREAMING_SNAKE_CASE__ )
__a : Tuple = (batch_size, input_ids.shape[1], config.vocab_size)
self.assertEqual(outputs['logits'].shape , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
__a : List[Any] = BlenderbotConfig(
vocab_size=self.vocab_size , d_model=1_4 , encoder_layers=2 , decoder_layers=2 , encoder_attention_heads=2 , decoder_attention_heads=2 , encoder_ffn_dim=8 , decoder_ffn_dim=8 , max_position_embeddings=4_8 , )
__a : Union[str, Any] = FlaxBlenderbotForConditionalGeneration(SCREAMING_SNAKE_CASE__ )
__a : Tuple = np.array([[7_1, 8_2, 1_8, 3_3, 4_6, 9_1, 2], [6_8, 3_4, 2_6, 5_8, 3_0, 2, 1]] , dtype=np.intaa )
__a : Optional[Any] = np.array([[8_2, 7_1, 8_2, 1_8, 2], [5_8, 6_8, 2, 1, 1]] , dtype=np.intaa )
__a : Optional[int] = lm_model(input_ids=SCREAMING_SNAKE_CASE__ , decoder_input_ids=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = (*summary.shape, config.vocab_size)
self.assertEqual(outputs['logits'].shape , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
__a : int = np.array([[7_1, 8_2, 1_8, 3_3, 2, 1, 1], [6_8, 3_4, 2_6, 5_8, 3_0, 8_2, 2]] , dtype=np.intaa )
__a : Optional[int] = shift_tokens_right(SCREAMING_SNAKE_CASE__ , 1 , 2 )
__a : List[str] = np.equal(SCREAMING_SNAKE_CASE__ , 1 ).astype(np.floataa ).sum()
__a : Any = np.equal(SCREAMING_SNAKE_CASE__ , 1 ).astype(np.floataa ).sum()
self.assertEqual(shifted.shape , input_ids.shape )
self.assertEqual(SCREAMING_SNAKE_CASE__ , n_pad_before - 1 )
self.assertTrue(np.equal(shifted[:, 0] , 2 ).all() )
@require_flax
class _UpperCamelCase( __lowerCamelCase , unittest.TestCase , __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = True
__SCREAMING_SNAKE_CASE : Any = (
(
FlaxBlenderbotModel,
FlaxBlenderbotForConditionalGeneration,
)
if is_flax_available()
else ()
)
__SCREAMING_SNAKE_CASE : str = (FlaxBlenderbotForConditionalGeneration,) if is_flax_available() else ()
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : List[Any] = FlaxBlenderbotModelTester(self )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a , __a : Any = self.model_tester.prepare_config_and_inputs()
for model_class in self.all_model_classes:
self.model_tester.check_use_cache_forward(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a , __a : Tuple = self.model_tester.prepare_config_and_inputs()
for model_class in self.all_model_classes:
self.model_tester.check_use_cache_forward_with_attn_mask(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a , __a : Any = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
with self.subTest(model_class.__name__ ):
__a : List[str] = self._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[Any] = model_class(SCREAMING_SNAKE_CASE__ )
@jax.jit
def encode_jitted(SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Union[str, Any]=None , **SCREAMING_SNAKE_CASE__ : List[Any] ):
return model.encode(input_ids=SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ )
with self.subTest('JIT Enabled' ):
__a : Dict = encode_jitted(**SCREAMING_SNAKE_CASE__ ).to_tuple()
with self.subTest('JIT Disabled' ):
with jax.disable_jit():
__a : Dict = encode_jitted(**SCREAMING_SNAKE_CASE__ ).to_tuple()
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for jitted_output, output in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertEqual(jitted_output.shape , output.shape )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a , __a : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
with self.subTest(model_class.__name__ ):
__a : Tuple = model_class(SCREAMING_SNAKE_CASE__ )
__a : Tuple = model.encode(inputs_dict['input_ids'] , inputs_dict['attention_mask'] )
__a : Optional[int] = {
'decoder_input_ids': inputs_dict['decoder_input_ids'],
'decoder_attention_mask': inputs_dict['decoder_attention_mask'],
'encoder_outputs': encoder_outputs,
}
@jax.jit
def decode_jitted(SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : int ):
return model.decode(
decoder_input_ids=SCREAMING_SNAKE_CASE__ , decoder_attention_mask=SCREAMING_SNAKE_CASE__ , encoder_outputs=SCREAMING_SNAKE_CASE__ , )
with self.subTest('JIT Enabled' ):
__a : List[Any] = decode_jitted(**SCREAMING_SNAKE_CASE__ ).to_tuple()
with self.subTest('JIT Disabled' ):
with jax.disable_jit():
__a : str = decode_jitted(**SCREAMING_SNAKE_CASE__ ).to_tuple()
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for jitted_output, output in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertEqual(jitted_output.shape , output.shape )
@slow
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
for model_class_name in self.all_model_classes:
__a : List[Any] = model_class_name.from_pretrained('facebook/blenderbot-400M-distill' )
# FlaxBlenderbotForSequenceClassification expects eos token in input_ids
__a : List[Any] = np.ones((1, 1) ) * model.config.eos_token_id
__a : Union[str, Any] = model(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@unittest.skipUnless(jax_device != 'cpu' , '3B test too slow on CPU.' )
@slow
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : int = {'num_beams': 1, 'early_stopping': True, 'min_length': 1_5, 'max_length': 2_5}
__a : Dict = {'skip_special_tokens': True, 'clean_up_tokenization_spaces': True}
__a : Dict = FlaxBlenderbotForConditionalGeneration.from_pretrained('facebook/blenderbot-3B' , from_pt=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = BlenderbotTokenizer.from_pretrained('facebook/blenderbot-3B' )
__a : Tuple = ['Sam']
__a : int = tokenizer(SCREAMING_SNAKE_CASE__ , return_tensors='jax' )
__a : Any = model.generate(**SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = 'Sam is a great name. It means "sun" in Gaelic.'
__a : Any = tokenizer.batch_decode(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
assert generated_txt[0].strip() == tgt_text
| 47 |
import functools
import operator
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''asapp/sew-d-tiny-100k''': '''https://huggingface.co/asapp/sew-d-tiny-100k/resolve/main/config.json''',
# See all SEW-D models at https://huggingface.co/models?filter=sew-d
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = '''sew-d'''
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : Dict=3_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=7_6_8 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : str=3_0_7_2 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : List[str]=2_5_6 , SCREAMING_SNAKE_CASE__ : Optional[Any]=True , SCREAMING_SNAKE_CASE__ : Tuple=True , SCREAMING_SNAKE_CASE__ : List[str]=("p2c", "c2p") , SCREAMING_SNAKE_CASE__ : str="layer_norm" , SCREAMING_SNAKE_CASE__ : Tuple="gelu_python" , SCREAMING_SNAKE_CASE__ : Tuple=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.0 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.02 , SCREAMING_SNAKE_CASE__ : int=1e-7 , SCREAMING_SNAKE_CASE__ : Any=1e-5 , SCREAMING_SNAKE_CASE__ : Optional[int]="group" , SCREAMING_SNAKE_CASE__ : Optional[Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=(6_4, 1_2_8, 1_2_8, 1_2_8, 1_2_8, 2_5_6, 2_5_6, 2_5_6, 2_5_6, 5_1_2, 5_1_2, 5_1_2, 5_1_2) , SCREAMING_SNAKE_CASE__ : List[Any]=(5, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : str=(1_0, 3, 1, 3, 1, 3, 1, 3, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : Optional[int]=False , SCREAMING_SNAKE_CASE__ : Optional[int]=1_2_8 , SCREAMING_SNAKE_CASE__ : Tuple=1_6 , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[Any]=0.05 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , SCREAMING_SNAKE_CASE__ : int=0.0 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : Optional[int]=0 , SCREAMING_SNAKE_CASE__ : Optional[int]="mean" , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : List[str]=False , SCREAMING_SNAKE_CASE__ : str=2_5_6 , SCREAMING_SNAKE_CASE__ : str=0 , SCREAMING_SNAKE_CASE__ : List[Any]=1 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ , pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = hidden_size
__a : Optional[Any] = feat_extract_norm
__a : List[str] = feat_extract_activation
__a : Dict = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = list(SCREAMING_SNAKE_CASE__ )
__a : List[str] = list(SCREAMING_SNAKE_CASE__ )
__a : int = conv_bias
__a : Tuple = num_conv_pos_embeddings
__a : List[str] = num_conv_pos_embedding_groups
__a : Optional[Any] = len(self.conv_dim )
__a : Union[str, Any] = num_hidden_layers
__a : Optional[Any] = intermediate_size
__a : Union[str, Any] = squeeze_factor
__a : List[Any] = max_position_embeddings
__a : Tuple = position_buckets
__a : Optional[int] = share_att_key
__a : List[str] = relative_attention
__a : Any = norm_rel_ebd
__a : Any = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = hidden_act
__a : str = num_attention_heads
__a : Union[str, Any] = hidden_dropout
__a : Optional[int] = attention_dropout
__a : List[str] = activation_dropout
__a : int = feat_proj_dropout
__a : int = final_dropout
__a : Dict = layer_norm_eps
__a : Tuple = feature_layer_norm_eps
__a : str = initializer_range
__a : Tuple = vocab_size
if (
(len(self.conv_stride ) != self.num_feat_extract_layers)
or (len(self.conv_kernel ) != self.num_feat_extract_layers)
or (len(self.conv_dim ) != self.num_feat_extract_layers)
):
raise ValueError(
'Configuration for convolutional layers is incorrect.'
'It is required that `len(config.conv_dim)` == `len(config.conv_stride)` == `len(config.conv_kernel)`,'
f'''but is `len(config.conv_dim) = {len(self.conv_dim )}`, `len(config.conv_stride)'''
f'''= {len(self.conv_stride )}`, `len(config.conv_kernel) = {len(self.conv_kernel )}`.''' )
# fine-tuning config parameters for SpecAugment: https://arxiv.org/abs/1904.08779
__a : Tuple = apply_spec_augment
__a : Optional[Any] = mask_time_prob
__a : Any = mask_time_length
__a : List[str] = mask_time_min_masks
__a : List[str] = mask_feature_prob
__a : Tuple = mask_feature_length
__a : Any = mask_feature_min_masks
# ctc loss
__a : Optional[int] = ctc_loss_reduction
__a : List[Any] = ctc_zero_infinity
# sequence classification
__a : Dict = use_weighted_layer_sum
__a : Optional[Any] = classifier_proj_size
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return functools.reduce(operator.mul , self.conv_stride , 1 )
| 47 | 1 |
import argparse
import torch
from transformers import BertConfig, BertForPreTraining, load_tf_weights_in_bert
from transformers.utils import logging
logging.set_verbosity_info()
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Dict ):
# Initialise PyTorch model
__a : Any = BertConfig.from_json_file(lowerCamelCase_ )
print(f'''Building PyTorch model from configuration: {config}''' )
__a : List[str] = BertForPreTraining(lowerCamelCase_ )
# Load weights from tf checkpoint
load_tf_weights_in_bert(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Save pytorch-model
print(f'''Save PyTorch model to {pytorch_dump_path}''' )
torch.save(model.state_dict() , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--tf_checkpoint_path''', default=None, type=str, required=True, help='''Path to the TensorFlow checkpoint path.'''
)
parser.add_argument(
'''--bert_config_file''',
default=None,
type=str,
required=True,
help=(
'''The config json file corresponding to the pre-trained BERT model. \n'''
'''This specifies the model architecture.'''
),
)
parser.add_argument(
'''--pytorch_dump_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.'''
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_tf_checkpoint_to_pytorch(args.tf_checkpoint_path, args.bert_config_file, args.pytorch_dump_path)
| 47 |
from __future__ import annotations
from sys import maxsize
from typing import Generic, TypeVar
SCREAMING_SNAKE_CASE__ = TypeVar('''T''')
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (position - 1) // 2
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 1
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 2
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[str] ):
'''simple docstring'''
__a : list[tuple[T, int]] = []
__a : dict[T, int] = {}
__a : int = 0
def __len__( self : Any ):
'''simple docstring'''
return self.elements
def __repr__( self : Any ):
'''simple docstring'''
return str(self.heap )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.elements == 0
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.heap.append((elem, weight) )
__a : List[Any] = self.elements
self.elements += 1
self._bubble_up(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self.elements > 1:
self._swap_nodes(0 , self.elements - 1 )
__a , __a : Union[str, Any] = self.heap.pop()
del self.position_map[elem]
self.elements -= 1
if self.elements > 0:
__a , __a : Dict = self.heap[0]
self._bubble_down(SCREAMING_SNAKE_CASE__ )
return elem
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
__a : str = (elem, weight)
if position > 0:
__a : Tuple = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = self.heap[parent_position]
if parent_weight > weight:
self._bubble_up(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
if curr_pos == 0:
return None
__a : List[str] = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : str = self.heap[curr_pos]
__a , __a : Optional[int] = self.heap[parent_position]
if parent_weight > weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_up(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : int = self.position_map[elem]
__a , __a : Optional[Any] = self.heap[curr_pos]
__a : Tuple = get_child_left_position(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = get_child_right_position(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements and child_right_position < self.elements:
__a , __a : str = self.heap[child_left_position]
__a , __a : List[str] = self.heap[child_right_position]
if child_right_weight < child_left_weight and child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements:
__a , __a : Any = self.heap[child_left_position]
if child_left_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
return None
if child_right_position < self.elements:
__a , __a : Union[str, Any] = self.heap[child_right_position]
if child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Optional[Any] = self.heap[nodea_pos][0]
__a : str = self.heap[nodea_pos][0]
__a , __a : int = (
self.heap[nodea_pos],
self.heap[nodea_pos],
)
__a : str = nodea_pos
__a : Optional[int] = nodea_pos
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[Any] ):
'''simple docstring'''
__a : dict[T, dict[T, int]] = {}
__a : int = 0
def __repr__( self : Tuple ):
'''simple docstring'''
return str(self.connections )
def __len__( self : Dict ):
'''simple docstring'''
return self.nodes
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
if node not in self.connections:
__a : Tuple = {}
self.nodes += 1
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.add_node(SCREAMING_SNAKE_CASE__ )
self.add_node(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = weight
__a : Any = weight
def UpperCAmelCase__ ( lowerCamelCase_ : GraphUndirectedWeighted[T] , ):
__a : dict[T, int] = {node: maxsize for node in graph.connections}
__a : dict[T, T | None] = {node: None for node in graph.connections}
__a : MinPriorityQueue[T] = MinPriorityQueue()
for node, weight in dist.items():
priority_queue.push(lowerCamelCase_ , lowerCamelCase_ )
if priority_queue.is_empty():
return dist, parent
# initialization
__a : Optional[int] = priority_queue.extract_min()
__a : int = 0
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : str = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Optional[int] = node
# running prim's algorithm
while not priority_queue.is_empty():
__a : Any = priority_queue.extract_min()
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : Tuple = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Dict = node
return dist, parent
| 47 | 1 |
from typing import List, Optional, Tuple, Union
import torch
from ...models import UNetaDModel
from ...schedulers import ScoreSdeVeScheduler
from ...utils import randn_tensor
from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : UNetaDModel
__SCREAMING_SNAKE_CASE : ScoreSdeVeScheduler
def __init__( self : str , SCREAMING_SNAKE_CASE__ : UNetaDModel , SCREAMING_SNAKE_CASE__ : ScoreSdeVeScheduler ):
'''simple docstring'''
super().__init__()
self.register_modules(unet=SCREAMING_SNAKE_CASE__ , scheduler=SCREAMING_SNAKE_CASE__ )
@torch.no_grad()
def __call__( self : Tuple , SCREAMING_SNAKE_CASE__ : int = 1 , SCREAMING_SNAKE_CASE__ : int = 2_0_0_0 , SCREAMING_SNAKE_CASE__ : Optional[Union[torch.Generator, List[torch.Generator]]] = None , SCREAMING_SNAKE_CASE__ : Optional[str] = "pil" , SCREAMING_SNAKE_CASE__ : bool = True , **SCREAMING_SNAKE_CASE__ : List[str] , ):
'''simple docstring'''
__a : str = self.unet.config.sample_size
__a : Any = (batch_size, 3, img_size, img_size)
__a : List[str] = self.unet
__a : str = randn_tensor(SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ ) * self.scheduler.init_noise_sigma
__a : Union[str, Any] = sample.to(self.device )
self.scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
self.scheduler.set_sigmas(SCREAMING_SNAKE_CASE__ )
for i, t in enumerate(self.progress_bar(self.scheduler.timesteps ) ):
__a : Tuple = self.scheduler.sigmas[i] * torch.ones(shape[0] , device=self.device )
# correction step
for _ in range(self.scheduler.config.correct_steps ):
__a : Any = self.unet(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ).sample
__a : List[str] = self.scheduler.step_correct(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ ).prev_sample
# prediction step
__a : Tuple = model(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ).sample
__a : Any = self.scheduler.step_pred(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ )
__a , __a : int = output.prev_sample, output.prev_sample_mean
__a : Tuple = sample_mean.clamp(0 , 1 )
__a : Optional[Any] = sample.cpu().permute(0 , 2 , 3 , 1 ).numpy()
if output_type == "pil":
__a : str = self.numpy_to_pil(SCREAMING_SNAKE_CASE__ )
if not return_dict:
return (sample,)
return ImagePipelineOutput(images=SCREAMING_SNAKE_CASE__ )
| 47 |
from collections.abc import Sequence
from queue import Queue
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : int=None , SCREAMING_SNAKE_CASE__ : Tuple=None ):
'''simple docstring'''
__a : Tuple = start
__a : Dict = end
__a : List[str] = val
__a : List[Any] = (start + end) // 2
__a : Optional[Any] = left
__a : List[str] = right
def __repr__( self : Dict ):
'''simple docstring'''
return f'''SegmentTreeNode(start={self.start}, end={self.end}, val={self.val})'''
class _UpperCamelCase:
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : Sequence , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Tuple = collection
__a : Dict = function
if self.collection:
__a : int = self._build_tree(0 , len(SCREAMING_SNAKE_CASE__ ) - 1 )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self._update_tree(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
return self._query_range(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if start == end:
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.collection[start] )
__a : Tuple = (start + end) // 2
__a : Optional[int] = self._build_tree(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = self._build_tree(mid + 1 , SCREAMING_SNAKE_CASE__ )
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.fn(left.val , right.val ) , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if node.start == i and node.end == i:
__a : Optional[Any] = val
return
if i <= node.mid:
self._update_tree(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
self._update_tree(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = self.fn(node.left.val , node.right.val )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
if node.start == i and node.end == j:
return node.val
if i <= node.mid:
if j <= node.mid:
# range in left child tree
return self._query_range(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
# range in left child tree and right child tree
return self.fn(
self._query_range(node.left , SCREAMING_SNAKE_CASE__ , node.mid ) , self._query_range(node.right , node.mid + 1 , SCREAMING_SNAKE_CASE__ ) , )
else:
# range in right child tree
return self._query_range(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
if self.root is not None:
__a : Tuple = Queue()
queue.put(self.root )
while not queue.empty():
__a : Tuple = queue.get()
yield node
if node.left is not None:
queue.put(node.left )
if node.right is not None:
queue.put(node.right )
if __name__ == "__main__":
import operator
for fn in [operator.add, max, min]:
print('''*''' * 50)
SCREAMING_SNAKE_CASE__ = SegmentTree([2, 1, 5, 3, 4], fn)
for node in arr.traverse():
print(node)
print()
arr.update(1, 5)
for node in arr.traverse():
print(node)
print()
print(arr.query_range(3, 4)) # 7
print(arr.query_range(2, 2)) # 5
print(arr.query_range(1, 3)) # 13
print()
| 47 | 1 |
from __future__ import annotations
from sys import maxsize
from typing import Generic, TypeVar
SCREAMING_SNAKE_CASE__ = TypeVar('''T''')
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (position - 1) // 2
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 1
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 2
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[str] ):
'''simple docstring'''
__a : list[tuple[T, int]] = []
__a : dict[T, int] = {}
__a : int = 0
def __len__( self : Any ):
'''simple docstring'''
return self.elements
def __repr__( self : Any ):
'''simple docstring'''
return str(self.heap )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.elements == 0
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.heap.append((elem, weight) )
__a : List[Any] = self.elements
self.elements += 1
self._bubble_up(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self.elements > 1:
self._swap_nodes(0 , self.elements - 1 )
__a , __a : Union[str, Any] = self.heap.pop()
del self.position_map[elem]
self.elements -= 1
if self.elements > 0:
__a , __a : Dict = self.heap[0]
self._bubble_down(SCREAMING_SNAKE_CASE__ )
return elem
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
__a : str = (elem, weight)
if position > 0:
__a : Tuple = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = self.heap[parent_position]
if parent_weight > weight:
self._bubble_up(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
if curr_pos == 0:
return None
__a : List[str] = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : str = self.heap[curr_pos]
__a , __a : Optional[int] = self.heap[parent_position]
if parent_weight > weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_up(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : int = self.position_map[elem]
__a , __a : Optional[Any] = self.heap[curr_pos]
__a : Tuple = get_child_left_position(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = get_child_right_position(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements and child_right_position < self.elements:
__a , __a : str = self.heap[child_left_position]
__a , __a : List[str] = self.heap[child_right_position]
if child_right_weight < child_left_weight and child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements:
__a , __a : Any = self.heap[child_left_position]
if child_left_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
return None
if child_right_position < self.elements:
__a , __a : Union[str, Any] = self.heap[child_right_position]
if child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Optional[Any] = self.heap[nodea_pos][0]
__a : str = self.heap[nodea_pos][0]
__a , __a : int = (
self.heap[nodea_pos],
self.heap[nodea_pos],
)
__a : str = nodea_pos
__a : Optional[int] = nodea_pos
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[Any] ):
'''simple docstring'''
__a : dict[T, dict[T, int]] = {}
__a : int = 0
def __repr__( self : Tuple ):
'''simple docstring'''
return str(self.connections )
def __len__( self : Dict ):
'''simple docstring'''
return self.nodes
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
if node not in self.connections:
__a : Tuple = {}
self.nodes += 1
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.add_node(SCREAMING_SNAKE_CASE__ )
self.add_node(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = weight
__a : Any = weight
def UpperCAmelCase__ ( lowerCamelCase_ : GraphUndirectedWeighted[T] , ):
__a : dict[T, int] = {node: maxsize for node in graph.connections}
__a : dict[T, T | None] = {node: None for node in graph.connections}
__a : MinPriorityQueue[T] = MinPriorityQueue()
for node, weight in dist.items():
priority_queue.push(lowerCamelCase_ , lowerCamelCase_ )
if priority_queue.is_empty():
return dist, parent
# initialization
__a : Optional[int] = priority_queue.extract_min()
__a : int = 0
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : str = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Optional[int] = node
# running prim's algorithm
while not priority_queue.is_empty():
__a : Any = priority_queue.extract_min()
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : Tuple = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Dict = node
return dist, parent
| 47 |
import os
import posixpath
import uuid
from dataclasses import dataclass
from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, Union
import numpy as np
import pyarrow as pa
import datasets
from datasets.arrow_writer import ArrowWriter, ParquetWriter
from datasets.config import MAX_SHARD_SIZE
from datasets.filesystems import (
is_remote_filesystem,
rename,
)
from datasets.iterable_dataset import _BaseExamplesIterable
from datasets.utils.py_utils import convert_file_size_to_int
SCREAMING_SNAKE_CASE__ = datasets.utils.logging.get_logger(__name__)
if TYPE_CHECKING:
import pyspark
@dataclass
class _UpperCamelCase( datasets.BuilderConfig ):
__SCREAMING_SNAKE_CASE : Optional[datasets.Features] = None
def UpperCAmelCase__ ( lowerCamelCase_ : "pyspark.sql.DataFrame" , lowerCamelCase_ : List[int] , ):
import pyspark
def generate_fn():
__a : List[Any] = df.select('*' , pyspark.sql.functions.spark_partition_id().alias('part_id' ) )
for partition_id in partition_order:
__a : Optional[int] = df_with_partition_id.select('*' ).where(f'''part_id = {partition_id}''' ).drop('part_id' )
__a : Optional[Any] = partition_df.collect()
__a : Union[str, Any] = 0
for row in rows:
yield f'''{partition_id}_{row_id}''', row.asDict()
row_id += 1
return generate_fn
class _UpperCamelCase( _BaseExamplesIterable ):
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : Dict=None , ):
'''simple docstring'''
__a : List[str] = df
__a : Tuple = partition_order or range(self.df.rdd.getNumPartitions() )
__a : List[Any] = _generate_iterable_examples(self.df , self.partition_order )
def __iter__( self : Tuple ):
'''simple docstring'''
yield from self.generate_examples_fn()
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : np.random.Generator ):
'''simple docstring'''
__a : Union[str, Any] = list(range(self.df.rdd.getNumPartitions() ) )
generator.shuffle(SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Union[str, Any] = self.split_shard_indices_by_worker(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return len(self.partition_order )
class _UpperCamelCase( datasets.DatasetBuilder ):
__SCREAMING_SNAKE_CASE : List[str] = SparkConfig
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : str = None , SCREAMING_SNAKE_CASE__ : str = None , **SCREAMING_SNAKE_CASE__ : Optional[int] , ):
'''simple docstring'''
import pyspark
__a : int = pyspark.sql.SparkSession.builder.getOrCreate()
__a : Optional[int] = df
__a : List[Any] = working_dir
super().__init__(
cache_dir=SCREAMING_SNAKE_CASE__ , config_name=str(self.df.semanticHash() ) , **SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
def create_cache_and_write_probe(SCREAMING_SNAKE_CASE__ : List[str] ):
# makedirs with exist_ok will recursively create the directory. It will not throw an error if directories
# already exist.
os.makedirs(self._cache_dir , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(self._cache_dir , 'fs_test' + uuid.uuida().hex )
# Opening the file in append mode will create a new file unless it already exists, in which case it will not
# change the file contents.
open(SCREAMING_SNAKE_CASE__ , 'a' )
return [probe_file]
if self._spark.conf.get('spark.master' , '' ).startswith('local' ):
return
# If the cluster is multi-node, make sure that the user provided a cache_dir and that it is on an NFS
# accessible to the driver.
# TODO: Stream batches to the driver using ArrowCollectSerializer instead of throwing an error.
if self._cache_dir:
__a : List[Any] = (
self._spark.sparkContext.parallelize(range(1 ) , 1 ).mapPartitions(SCREAMING_SNAKE_CASE__ ).collect()
)
if os.path.isfile(probe[0] ):
return
raise ValueError(
'When using Dataset.from_spark on a multi-node cluster, the driver and all workers should be able to access cache_dir' )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
return datasets.DatasetInfo(features=self.config.features )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : datasets.download.download_manager.DownloadManager ):
'''simple docstring'''
return [datasets.SplitGenerator(name=datasets.Split.TRAIN )]
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
import pyspark
def get_arrow_batch_size(SCREAMING_SNAKE_CASE__ : int ):
for batch in it:
yield pa.RecordBatch.from_pydict({'batch_bytes': [batch.nbytes]} )
__a : List[str] = self.df.count()
__a : Dict = df_num_rows if df_num_rows <= 1_0_0 else 1_0_0
# Approximate the size of each row (in Arrow format) by averaging over a max-100-row sample.
__a : List[str] = (
self.df.limit(SCREAMING_SNAKE_CASE__ )
.repartition(1 )
.mapInArrow(SCREAMING_SNAKE_CASE__ , 'batch_bytes: long' )
.agg(pyspark.sql.functions.sum('batch_bytes' ).alias('sample_bytes' ) )
.collect()[0]
.sample_bytes
/ sample_num_rows
)
__a : Dict = approx_bytes_per_row * df_num_rows
if approx_total_size > max_shard_size:
# Make sure there is at least one row per partition.
__a : Union[str, Any] = min(SCREAMING_SNAKE_CASE__ , int(approx_total_size / max_shard_size ) )
__a : int = self.df.repartition(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : int , ):
'''simple docstring'''
import pyspark
__a : Any = ParquetWriter if file_format == 'parquet' else ArrowWriter
__a : Union[str, Any] = os.path.join(self._working_dir , os.path.basename(SCREAMING_SNAKE_CASE__ ) ) if self._working_dir else fpath
__a : Optional[int] = file_format == 'parquet'
# Define these so that we don't reference self in write_arrow, which will result in a pickling error due to
# pickling the SparkContext.
__a : List[str] = self.config.features
__a : int = self._writer_batch_size
__a : Union[str, Any] = self._fs.storage_options
def write_arrow(SCREAMING_SNAKE_CASE__ : Optional[int] ):
# Within the same SparkContext, no two task attempts will share the same attempt ID.
__a : Any = pyspark.TaskContext().taskAttemptId()
__a : str = next(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
if first_batch is None:
# Some partitions might not receive any data.
return pa.RecordBatch.from_arrays(
[[task_id], [0], [0]] , names=['task_id', 'num_examples', 'num_bytes'] , )
__a : Any = 0
__a : List[str] = writer_class(
features=SCREAMING_SNAKE_CASE__ , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Optional[Any] = pa.Table.from_batches([first_batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
for batch in it:
if max_shard_size is not None and writer._num_bytes >= max_shard_size:
__a , __a : Optional[int] = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
shard_id += 1
__a : Optional[Any] = writer_class(
features=writer._features , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Union[str, Any] = pa.Table.from_batches([batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
if writer._num_bytes > 0:
__a , __a : str = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
if working_fpath != fpath:
for file in os.listdir(os.path.dirname(SCREAMING_SNAKE_CASE__ ) ):
__a : Any = os.path.join(os.path.dirname(SCREAMING_SNAKE_CASE__ ) , os.path.basename(SCREAMING_SNAKE_CASE__ ) )
shutil.move(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Dict = (
self.df.mapInArrow(SCREAMING_SNAKE_CASE__ , 'task_id: long, num_examples: long, num_bytes: long' )
.groupBy('task_id' )
.agg(
pyspark.sql.functions.sum('num_examples' ).alias('total_num_examples' ) , pyspark.sql.functions.sum('num_bytes' ).alias('total_num_bytes' ) , pyspark.sql.functions.count('num_bytes' ).alias('num_shards' ) , pyspark.sql.functions.collect_list('num_examples' ).alias('shard_lengths' ) , )
.collect()
)
for row in stats:
yield row.task_id, (row.total_num_examples, row.total_num_bytes, row.num_shards, row.shard_lengths)
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , SCREAMING_SNAKE_CASE__ : str = "arrow" , SCREAMING_SNAKE_CASE__ : Optional[Union[str, int]] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
self._validate_cache_dir()
__a : List[str] = convert_file_size_to_int(max_shard_size or MAX_SHARD_SIZE )
self._repartition_df_if_needed(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = not is_remote_filesystem(self._fs )
__a : Optional[Any] = os.path.join if is_local else posixpath.join
__a : Any = '-TTTTT-SSSSS-of-NNNNN'
__a : Union[str, Any] = f'''{self.name}-{split_generator.name}{SUFFIX}.{file_format}'''
__a : Any = path_join(self._output_dir , SCREAMING_SNAKE_CASE__ )
__a : Any = 0
__a : Dict = 0
__a : int = 0
__a : List[str] = []
__a : Optional[int] = []
for task_id, content in self._prepare_split_single(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : Optional[int] = content
if num_bytes > 0:
total_num_examples += num_examples
total_num_bytes += num_bytes
total_shards += num_shards
task_id_and_num_shards.append((task_id, num_shards) )
all_shard_lengths.extend(SCREAMING_SNAKE_CASE__ )
__a : List[str] = total_num_examples
__a : Optional[int] = total_num_bytes
# should rename everything at the end
logger.debug(f'''Renaming {total_shards} shards.''' )
if total_shards > 1:
__a : Any = all_shard_lengths
# Define fs outside of _rename_shard so that we don't reference self in the function, which will result in a
# pickling error due to pickling the SparkContext.
__a : Dict = self._fs
# use the -SSSSS-of-NNNNN pattern
def _rename_shard(
SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , ):
rename(
SCREAMING_SNAKE_CASE__ , fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace('TTTTT-SSSSS' , f'''{global_shard_id:05d}''' ).replace('NNNNN' , f'''{total_shards:05d}''' ) , )
__a : Union[str, Any] = []
__a : List[str] = 0
for i in range(len(SCREAMING_SNAKE_CASE__ ) ):
__a , __a : Union[str, Any] = task_id_and_num_shards[i]
for shard_id in range(SCREAMING_SNAKE_CASE__ ):
args.append([task_id, shard_id, global_shard_id] )
global_shard_id += 1
self._spark.sparkContext.parallelize(SCREAMING_SNAKE_CASE__ , len(SCREAMING_SNAKE_CASE__ ) ).map(lambda SCREAMING_SNAKE_CASE__ : _rename_shard(*SCREAMING_SNAKE_CASE__ ) ).collect()
else:
# don't use any pattern
__a : List[Any] = 0
__a : Any = task_id_and_num_shards[0][0]
self._rename(
fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace(SCREAMING_SNAKE_CASE__ , '' ) , )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , ):
'''simple docstring'''
return SparkExamplesIterable(self.df )
| 47 | 1 |
import argparse
import gc
import json
import os
import shutil
import warnings
import torch
from transformers import LlamaConfig, LlamaForCausalLM, LlamaTokenizer
try:
from transformers import LlamaTokenizerFast
except ImportError as e:
warnings.warn(e)
warnings.warn(
'''The converted tokenizer will be the `slow` tokenizer. To use the fast, update your `tokenizers` library and re-run the tokenizer conversion'''
)
SCREAMING_SNAKE_CASE__ = None
SCREAMING_SNAKE_CASE__ = {
'''7B''': 1_1008,
'''13B''': 1_3824,
'''30B''': 1_7920,
'''65B''': 2_2016,
'''70B''': 2_8672,
}
SCREAMING_SNAKE_CASE__ = {
'''7B''': 1,
'''7Bf''': 1,
'''13B''': 2,
'''13Bf''': 2,
'''30B''': 4,
'''65B''': 8,
'''70B''': 8,
'''70Bf''': 8,
}
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Dict=1 , lowerCamelCase_ : Union[str, Any]=2_5_6 ):
return multiple_of * ((int(ffn_dim_multiplier * int(8 * n / 3 ) ) + multiple_of - 1) // multiple_of)
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] ):
with open(lowerCamelCase_ , 'r' ) as f:
return json.load(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Dict ):
with open(lowerCamelCase_ , 'w' ) as f:
json.dump(lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[str] , lowerCamelCase_ : Any , lowerCamelCase_ : Optional[int]=True ):
os.makedirs(lowerCamelCase_ , exist_ok=lowerCamelCase_ )
__a : Tuple = os.path.join(lowerCamelCase_ , 'tmp' )
os.makedirs(lowerCamelCase_ , exist_ok=lowerCamelCase_ )
__a : str = read_json(os.path.join(lowerCamelCase_ , 'params.json' ) )
__a : str = NUM_SHARDS[model_size]
__a : List[str] = params['n_layers']
__a : Optional[Any] = params['n_heads']
__a : Optional[Any] = n_heads // num_shards
__a : Optional[int] = params['dim']
__a : Union[str, Any] = dim // n_heads
__a : Dict = 10000.0
__a : Tuple = 1.0 / (base ** (torch.arange(0 , lowerCamelCase_ , 2 ).float() / dims_per_head))
if "n_kv_heads" in params:
__a : Tuple = params['n_kv_heads'] # for GQA / MQA
__a : Union[str, Any] = n_heads_per_shard // num_key_value_heads
__a : str = dim // num_key_value_heads
else: # compatibility with other checkpoints
__a : List[str] = n_heads
__a : List[Any] = n_heads_per_shard
__a : List[str] = dim
# permute for sliced rotary
def permute(lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple=n_heads , lowerCamelCase_ : Any=dim , lowerCamelCase_ : List[str]=dim ):
return w.view(lowerCamelCase_ , dima // n_heads // 2 , 2 , lowerCamelCase_ ).transpose(1 , 2 ).reshape(lowerCamelCase_ , lowerCamelCase_ )
print(f'''Fetching all parameters from the checkpoint at {input_base_path}.''' )
# Load weights
if model_size == "7B":
# Not sharded
# (The sharded implementation would also work, but this is simpler.)
__a : Union[str, Any] = torch.load(os.path.join(lowerCamelCase_ , 'consolidated.00.pth' ) , map_location='cpu' )
else:
# Sharded
__a : Any = [
torch.load(os.path.join(lowerCamelCase_ , f'''consolidated.{i:02d}.pth''' ) , map_location='cpu' )
for i in range(lowerCamelCase_ )
]
__a : Any = 0
__a : List[str] = {'weight_map': {}}
for layer_i in range(lowerCamelCase_ ):
__a : Optional[int] = f'''pytorch_model-{layer_i + 1}-of-{n_layers + 1}.bin'''
if model_size == "7B":
# Unsharded
__a : Dict = {
f'''model.layers.{layer_i}.self_attn.q_proj.weight''': permute(
loaded[f'''layers.{layer_i}.attention.wq.weight'''] ),
f'''model.layers.{layer_i}.self_attn.k_proj.weight''': permute(
loaded[f'''layers.{layer_i}.attention.wk.weight'''] ),
f'''model.layers.{layer_i}.self_attn.v_proj.weight''': loaded[f'''layers.{layer_i}.attention.wv.weight'''],
f'''model.layers.{layer_i}.self_attn.o_proj.weight''': loaded[f'''layers.{layer_i}.attention.wo.weight'''],
f'''model.layers.{layer_i}.mlp.gate_proj.weight''': loaded[f'''layers.{layer_i}.feed_forward.w1.weight'''],
f'''model.layers.{layer_i}.mlp.down_proj.weight''': loaded[f'''layers.{layer_i}.feed_forward.w2.weight'''],
f'''model.layers.{layer_i}.mlp.up_proj.weight''': loaded[f'''layers.{layer_i}.feed_forward.w3.weight'''],
f'''model.layers.{layer_i}.input_layernorm.weight''': loaded[f'''layers.{layer_i}.attention_norm.weight'''],
f'''model.layers.{layer_i}.post_attention_layernorm.weight''': loaded[f'''layers.{layer_i}.ffn_norm.weight'''],
}
else:
# Sharded
# Note that attention.w{q,k,v,o}, feed_fordward.w[1,2,3], attention_norm.weight and ffn_norm.weight share
# the same storage object, saving attention_norm and ffn_norm will save other weights too, which is
# redundant as other weights will be stitched from multiple shards. To avoid that, they are cloned.
__a : Optional[int] = {
f'''model.layers.{layer_i}.input_layernorm.weight''': loaded[0][
f'''layers.{layer_i}.attention_norm.weight'''
].clone(),
f'''model.layers.{layer_i}.post_attention_layernorm.weight''': loaded[0][
f'''layers.{layer_i}.ffn_norm.weight'''
].clone(),
}
__a : List[Any] = permute(
torch.cat(
[
loaded[i][f'''layers.{layer_i}.attention.wq.weight'''].view(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
for i in range(lowerCamelCase_ )
] , dim=0 , ).reshape(lowerCamelCase_ , lowerCamelCase_ ) )
__a : List[str] = permute(
torch.cat(
[
loaded[i][f'''layers.{layer_i}.attention.wk.weight'''].view(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
for i in range(lowerCamelCase_ )
] , dim=0 , ).reshape(lowerCamelCase_ , lowerCamelCase_ ) , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , )
__a : List[Any] = torch.cat(
[
loaded[i][f'''layers.{layer_i}.attention.wv.weight'''].view(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
for i in range(lowerCamelCase_ )
] , dim=0 , ).reshape(lowerCamelCase_ , lowerCamelCase_ )
__a : Tuple = torch.cat(
[loaded[i][f'''layers.{layer_i}.attention.wo.weight'''] for i in range(lowerCamelCase_ )] , dim=1 )
__a : List[str] = torch.cat(
[loaded[i][f'''layers.{layer_i}.feed_forward.w1.weight'''] for i in range(lowerCamelCase_ )] , dim=0 )
__a : List[str] = torch.cat(
[loaded[i][f'''layers.{layer_i}.feed_forward.w2.weight'''] for i in range(lowerCamelCase_ )] , dim=1 )
__a : str = torch.cat(
[loaded[i][f'''layers.{layer_i}.feed_forward.w3.weight'''] for i in range(lowerCamelCase_ )] , dim=0 )
__a : Union[str, Any] = inv_freq
for k, v in state_dict.items():
__a : Any = filename
param_count += v.numel()
torch.save(lowerCamelCase_ , os.path.join(lowerCamelCase_ , lowerCamelCase_ ) )
__a : int = f'''pytorch_model-{n_layers + 1}-of-{n_layers + 1}.bin'''
if model_size == "7B":
# Unsharded
__a : Tuple = {
'model.embed_tokens.weight': loaded['tok_embeddings.weight'],
'model.norm.weight': loaded['norm.weight'],
'lm_head.weight': loaded['output.weight'],
}
else:
__a : Optional[int] = {
'model.norm.weight': loaded[0]['norm.weight'],
'model.embed_tokens.weight': torch.cat(
[loaded[i]['tok_embeddings.weight'] for i in range(lowerCamelCase_ )] , dim=1 ),
'lm_head.weight': torch.cat([loaded[i]['output.weight'] for i in range(lowerCamelCase_ )] , dim=0 ),
}
for k, v in state_dict.items():
__a : List[Any] = filename
param_count += v.numel()
torch.save(lowerCamelCase_ , os.path.join(lowerCamelCase_ , lowerCamelCase_ ) )
# Write configs
__a : Any = {'total_size': param_count * 2}
write_json(lowerCamelCase_ , os.path.join(lowerCamelCase_ , 'pytorch_model.bin.index.json' ) )
__a : Optional[int] = params['ffn_dim_multiplier'] if 'ffn_dim_multiplier' in params else 1
__a : Tuple = params['multiple_of'] if 'multiple_of' in params else 2_5_6
__a : Any = LlamaConfig(
hidden_size=lowerCamelCase_ , intermediate_size=compute_intermediate_size(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ ) , num_attention_heads=params['n_heads'] , num_hidden_layers=params['n_layers'] , rms_norm_eps=params['norm_eps'] , num_key_value_heads=lowerCamelCase_ , )
config.save_pretrained(lowerCamelCase_ )
# Make space so we can load the model properly now.
del state_dict
del loaded
gc.collect()
print('Loading the checkpoint in a Llama model.' )
__a : List[str] = LlamaForCausalLM.from_pretrained(lowerCamelCase_ , torch_dtype=torch.floataa , low_cpu_mem_usage=lowerCamelCase_ )
# Avoid saving this as part of the config.
del model.config._name_or_path
print('Saving in the Transformers format.' )
model.save_pretrained(lowerCamelCase_ , safe_serialization=lowerCamelCase_ )
shutil.rmtree(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Optional[int] ):
# Initialize the tokenizer based on the `spm` model
__a : Dict = LlamaTokenizer if LlamaTokenizerFast is None else LlamaTokenizerFast
print(f'''Saving a {tokenizer_class.__name__} to {tokenizer_path}.''' )
__a : Optional[Any] = tokenizer_class(lowerCamelCase_ )
tokenizer.save_pretrained(lowerCamelCase_ )
def UpperCAmelCase__ ( ):
__a : str = argparse.ArgumentParser()
parser.add_argument(
'--input_dir' , help='Location of LLaMA weights, which contains tokenizer.model and model folders' , )
parser.add_argument(
'--model_size' , choices=['7B', '7Bf', '13B', '13Bf', '30B', '65B', '70B', '70Bf', 'tokenizer_only'] , )
parser.add_argument(
'--output_dir' , help='Location to write HF model and tokenizer' , )
parser.add_argument('--safe_serialization' , type=lowerCamelCase_ , help='Whether or not to save using `safetensors`.' )
__a : List[Any] = parser.parse_args()
if args.model_size != "tokenizer_only":
write_model(
model_path=args.output_dir , input_base_path=os.path.join(args.input_dir , args.model_size ) , model_size=args.model_size , safe_serialization=args.safe_serialization , )
__a : List[str] = os.path.join(args.input_dir , 'tokenizer.model' )
write_tokenizer(args.output_dir , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 |
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, TensorDataset
from tqdm import tqdm
from transformers import GPTaLMHeadModel
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : int ):
# save results
if os.path.exists(lowerCamelCase_ ):
if os.path.exists(os.path.join(lowerCamelCase_ , 'config.json' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'config.json' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'config.json' ) )
if os.path.exists(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) )
else:
os.makedirs(lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Any=False ):
__a : Dict = 2
if unlogit:
__a : Optional[Any] = torch.pow(lowerCamelCase_ , lowerCamelCase_ )
__a : Any = p * torch.log(lowerCamelCase_ )
__a : Union[str, Any] = 0
return -plogp.sum(dim=-1 )
def UpperCAmelCase__ ( lowerCamelCase_ : Any ):
logger.info('lv, h >\t' + '\t'.join(f'''{x + 1}''' for x in range(len(lowerCamelCase_ ) ) ) )
for row in range(len(lowerCamelCase_ ) ):
if tensor.dtype != torch.long:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:.5f}''' for x in tensor[row].cpu().data ) )
else:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:d}''' for x in tensor[row].cpu().data ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Any , lowerCamelCase_ : int , lowerCamelCase_ : int=True , lowerCamelCase_ : Optional[Any]=True , lowerCamelCase_ : List[Any]=None , lowerCamelCase_ : List[Any]=False ):
__a , __a : Optional[int] = model.config.num_hidden_layers, model.config.num_attention_heads
__a : str = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
__a : int = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
if head_mask is None:
__a : Union[str, Any] = torch.ones(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
head_mask.requires_grad_(requires_grad=lowerCamelCase_ )
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
__a : Any = None
__a : Optional[int] = 0.0
__a : Optional[Any] = 0.0
for step, inputs in enumerate(tqdm(lowerCamelCase_ , desc='Iteration' , disable=args.local_rank not in [-1, 0] ) ):
__a : Dict = tuple(t.to(args.device ) for t in inputs )
((__a) , ) : Dict = inputs
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
__a : List[Any] = model(lowerCamelCase_ , labels=lowerCamelCase_ , head_mask=lowerCamelCase_ )
# (loss), lm_logits, presents, (all hidden_states), (attentions)
__a , __a , __a : int = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
total_loss += loss.detach().cpu().numpy()
if compute_entropy:
for layer, attn in enumerate(lowerCamelCase_ ):
__a : List[str] = entropy(attn.detach() , lowerCamelCase_ )
attn_entropy[layer] += masked_entropy.sum(-1 ).sum(0 ).sum(0 ).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
tot_tokens += torch.ones_like(lowerCamelCase_ ).float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
__a : Optional[Any] = 2
__a : Union[str, Any] = torch.pow(torch.pow(lowerCamelCase_ , lowerCamelCase_ ).sum(-1 ) , 1 / exponent )
head_importance /= norm_by_layer.unsqueeze(-1 ) + 1e-20
if not args.dont_normalize_global_importance:
__a : List[str] = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print matrices
if compute_entropy:
logger.info('Attention entropies' )
print_ad_tensor(lowerCamelCase_ )
if compute_importance:
logger.info('Head importance scores' )
print_ad_tensor(lowerCamelCase_ )
logger.info('Head ranked by importance scores' )
__a : Optional[Any] = torch.zeros(head_importance.numel() , dtype=torch.long , device=args.device )
__a : str = torch.arange(
head_importance.numel() , device=args.device )
__a : Tuple = head_ranks.view_as(lowerCamelCase_ )
print_ad_tensor(lowerCamelCase_ )
return attn_entropy, head_importance, total_loss
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Tuple , lowerCamelCase_ : int ):
__a , __a , __a : Optional[int] = compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ )
__a : Tuple = 1 / loss # instead of downsteam score use the LM loss
logger.info('Pruning: original score: %f, threshold: %f' , lowerCamelCase_ , original_score * args.masking_threshold )
__a : Tuple = torch.ones_like(lowerCamelCase_ )
__a : int = max(1 , int(new_head_mask.numel() * args.masking_amount ) )
__a : Tuple = original_score
while current_score >= original_score * args.masking_threshold:
__a : Optional[Any] = new_head_mask.clone().detach() # save current head mask
# heads from least important to most - keep only not-masked heads
__a : List[str] = float('Inf' )
__a : List[Any] = head_importance.view(-1 ).sort()[1]
if len(lowerCamelCase_ ) <= num_to_mask:
print('BREAK BY num_to_mask' )
break
# mask heads
__a : Any = current_heads_to_mask[:num_to_mask]
logger.info('Heads to mask: %s' , str(current_heads_to_mask.tolist() ) )
__a : int = new_head_mask.view(-1 )
__a : Tuple = 0.0
__a : int = new_head_mask.view_as(lowerCamelCase_ )
__a : Optional[int] = new_head_mask.clone().detach()
print_ad_tensor(lowerCamelCase_ )
# Compute metric and head importance again
__a , __a , __a : int = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[Any] = 1 / loss
logger.info(
'Masking: current score: %f, remaining heads %d (%.1f percents)' , lowerCamelCase_ , new_head_mask.sum() , new_head_mask.sum() / new_head_mask.numel() * 1_0_0 , )
logger.info('Final head mask' )
print_ad_tensor(lowerCamelCase_ )
np.save(os.path.join(args.output_dir , 'head_mask.npy' ) , head_mask.detach().cpu().numpy() )
return head_mask
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Union[str, Any] ):
__a : List[Any] = datetime.now()
__a , __a , __a : List[str] = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[str] = 1 / loss
__a : List[Any] = datetime.now() - before_time
__a : List[str] = sum(p.numel() for p in model.parameters() )
__a : Dict = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(lowerCamelCase_ ) )
}
for k, v in heads_to_prune.items():
if isinstance(lowerCamelCase_ , lowerCamelCase_ ):
__a : Tuple = [
v,
]
assert sum(len(lowerCamelCase_ ) for h in heads_to_prune.values() ) == (1 - head_mask.long()).sum().item()
model.prune_heads(lowerCamelCase_ )
__a : Optional[Any] = sum(p.numel() for p in model.parameters() )
__a : Tuple = datetime.now()
__a , __a , __a : Tuple = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ , actually_pruned=lowerCamelCase_ , )
__a : Optional[Any] = 1 / loss
__a : List[Any] = datetime.now() - before_time
logger.info(
'Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)' , lowerCamelCase_ , lowerCamelCase_ , pruned_num_params / original_num_params * 1_0_0 , )
logger.info('Pruning: score with masking: %f score with pruning: %f' , lowerCamelCase_ , lowerCamelCase_ )
logger.info('Pruning: speed ratio (original timing / new timing): %f percents' , original_time / new_time * 1_0_0 )
save_model(lowerCamelCase_ , args.output_dir )
def UpperCAmelCase__ ( ):
__a : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'--data_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The input data dir. Should contain the .tsv files (or other data files) for the task.' , )
parser.add_argument(
'--model_name_or_path' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='Path to pretrained model or model identifier from huggingface.co/models' , )
parser.add_argument(
'--output_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
# Other parameters
parser.add_argument(
'--config_name' , default='' , type=lowerCamelCase_ , help='Pretrained config name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--tokenizer_name' , default='' , type=lowerCamelCase_ , help='Pretrained tokenizer name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--cache_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , help='Where do you want to store the pre-trained models downloaded from s3' , )
parser.add_argument(
'--data_subset' , type=lowerCamelCase_ , default=-1 , help='If > 0: limit the data to a subset of data_subset instances.' )
parser.add_argument(
'--overwrite_output_dir' , action='store_true' , help='Whether to overwrite data in output directory' )
parser.add_argument(
'--overwrite_cache' , action='store_true' , help='Overwrite the cached training and evaluation sets' )
parser.add_argument(
'--dont_normalize_importance_by_layer' , action='store_true' , help='Don\'t normalize importance score by layers' )
parser.add_argument(
'--dont_normalize_global_importance' , action='store_true' , help='Don\'t normalize all importance scores between 0 and 1' , )
parser.add_argument(
'--try_masking' , action='store_true' , help='Whether to try to mask head until a threshold of accuracy.' )
parser.add_argument(
'--masking_threshold' , default=0.9 , type=lowerCamelCase_ , help='masking threshold in term of metrics (stop masking when metric < threshold * original metric value).' , )
parser.add_argument(
'--masking_amount' , default=0.1 , type=lowerCamelCase_ , help='Amount to heads to masking at each masking step.' )
parser.add_argument('--metric_name' , default='acc' , type=lowerCamelCase_ , help='Metric to use for head masking.' )
parser.add_argument(
'--max_seq_length' , default=1_2_8 , type=lowerCamelCase_ , help=(
'The maximum total input sequence length after WordPiece tokenization. \n'
'Sequences longer than this will be truncated, sequences shorter padded.'
) , )
parser.add_argument('--batch_size' , default=1 , type=lowerCamelCase_ , help='Batch size.' )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 )
parser.add_argument('--local_rank' , type=lowerCamelCase_ , default=-1 , help='local_rank for distributed training on gpus' )
parser.add_argument('--no_cuda' , action='store_true' , help='Whether not to use CUDA when available' )
parser.add_argument('--server_ip' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
parser.add_argument('--server_port' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
__a : Optional[Any] = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print('Waiting for debugger attach' )
ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=lowerCamelCase_ )
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
__a : List[str] = torch.device('cuda' if torch.cuda.is_available() and not args.no_cuda else 'cpu' )
__a : Tuple = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank )
__a : Union[str, Any] = torch.device('cuda' , args.local_rank )
__a : Any = 1
torch.distributed.init_process_group(backend='nccl' ) # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN )
logger.info('device: {} n_gpu: {}, distributed: {}'.format(args.device , args.n_gpu , bool(args.local_rank != -1 ) ) )
__a : Optional[Any] = GPTaLMHeadModel.from_pretrained(args.model_name_or_path )
# Distributed and parallel training
model.to(args.device )
if args.local_rank != -1:
__a : List[Any] = nn.parallel.DistributedDataParallel(
lowerCamelCase_ , device_ids=[args.local_rank] , output_device=args.local_rank , find_unused_parameters=lowerCamelCase_ )
elif args.n_gpu > 1:
__a : Union[str, Any] = nn.DataParallel(lowerCamelCase_ )
# Print/save training arguments
os.makedirs(args.output_dir , exist_ok=lowerCamelCase_ )
torch.save(lowerCamelCase_ , os.path.join(args.output_dir , 'run_args.bin' ) )
logger.info('Training/evaluation parameters %s' , lowerCamelCase_ )
# Prepare dataset
__a : Tuple = np.concatenate(
[
np.loadtxt(args.data_dir , dtype=np.intaa ),
] )
__a : str = (torch.from_numpy(lowerCamelCase_ ),)
__a : List[str] = TensorDataset(*lowerCamelCase_ )
__a : Optional[Any] = RandomSampler(lowerCamelCase_ )
__a : Union[str, Any] = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.batch_size )
# Compute head entropy and importance score
compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
__a : Union[str, Any] = mask_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
prune_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 | 1 |
import unittest
from transformers import is_flax_available
from transformers.testing_utils import require_flax, require_sentencepiece, require_tokenizers, require_torch, slow
if is_flax_available():
import optax
from flax.training.common_utils import onehot
from transformers import AutoTokenizer, FlaxMTaForConditionalGeneration
from transformers.models.ta.modeling_flax_ta import shift_tokens_right
@require_torch
@require_sentencepiece
@require_tokenizers
@require_flax
class _UpperCamelCase( unittest.TestCase ):
@slow
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Any = FlaxMTaForConditionalGeneration.from_pretrained('google/mt5-small' )
__a : int = AutoTokenizer.from_pretrained('google/mt5-small' )
__a : Any = tokenizer('Hello there' , return_tensors='np' ).input_ids
__a : Any = tokenizer('Hi I am' , return_tensors='np' ).input_ids
__a : List[str] = shift_tokens_right(SCREAMING_SNAKE_CASE__ , model.config.pad_token_id , model.config.decoder_start_token_id )
__a : Tuple = model(SCREAMING_SNAKE_CASE__ , decoder_input_ids=SCREAMING_SNAKE_CASE__ ).logits
__a : Union[str, Any] = optax.softmax_cross_entropy(SCREAMING_SNAKE_CASE__ , onehot(SCREAMING_SNAKE_CASE__ , logits.shape[-1] ) ).mean()
__a : Tuple = -(labels.shape[-1] * loss.item())
__a : Optional[Any] = -84.9_127
self.assertTrue(abs(mtf_score - EXPECTED_SCORE ) < 1e-4 )
| 47 |
import argparse
import os
import gluonnlp as nlp
import mxnet as mx
import numpy as np
import torch
from gluonnlp.base import get_home_dir
from gluonnlp.model.bert import BERTEncoder
from gluonnlp.model.utils import _load_vocab
from gluonnlp.vocab import Vocab
from packaging import version
from torch import nn
from transformers import BertConfig, BertForMaskedLM, BertModel, RobertaTokenizer
from transformers.models.bert.modeling_bert import (
BertIntermediate,
BertLayer,
BertOutput,
BertSelfAttention,
BertSelfOutput,
)
from transformers.utils import logging
if version.parse(nlp.__version__) != version.parse('''0.8.3'''):
raise Exception('''requires gluonnlp == 0.8.3''')
if version.parse(mx.__version__) != version.parse('''1.5.0'''):
raise Exception('''requires mxnet == 1.5.0''')
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''The Nymphenburg Palace is a beautiful palace in Munich!'''
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
__a : List[Any] = {
'attention_cell': 'multi_head',
'num_layers': 4,
'units': 1_0_2_4,
'hidden_size': 7_6_8,
'max_length': 5_1_2,
'num_heads': 8,
'scaled': True,
'dropout': 0.1,
'use_residual': True,
'embed_size': 1_0_2_4,
'embed_dropout': 0.1,
'word_embed': None,
'layer_norm_eps': 1e-5,
'token_type_vocab_size': 2,
}
__a : Optional[int] = bort_4_8_768_1024_hparams
# Let's construct the original Bort model here
# Taken from official BERT implementation, see:
# https://github.com/alexa/bort/blob/master/bort/bort.py
__a : List[str] = BERTEncoder(
attention_cell=predefined_args['attention_cell'] , num_layers=predefined_args['num_layers'] , units=predefined_args['units'] , hidden_size=predefined_args['hidden_size'] , max_length=predefined_args['max_length'] , num_heads=predefined_args['num_heads'] , scaled=predefined_args['scaled'] , dropout=predefined_args['dropout'] , output_attention=lowerCamelCase_ , output_all_encodings=lowerCamelCase_ , use_residual=predefined_args['use_residual'] , activation=predefined_args.get('activation' , 'gelu' ) , layer_norm_eps=predefined_args.get('layer_norm_eps' , lowerCamelCase_ ) , )
# Vocab information needs to be fetched first
# It's the same as RoBERTa, so RobertaTokenizer can be used later
__a : int = 'openwebtext_ccnews_stories_books_cased'
# Specify download folder to Gluonnlp's vocab
__a : Optional[Any] = os.path.join(get_home_dir() , 'models' )
__a : Optional[Any] = _load_vocab(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , cls=lowerCamelCase_ )
__a : Any = nlp.model.BERTModel(
lowerCamelCase_ , len(lowerCamelCase_ ) , units=predefined_args['units'] , embed_size=predefined_args['embed_size'] , embed_dropout=predefined_args['embed_dropout'] , word_embed=predefined_args['word_embed'] , use_pooler=lowerCamelCase_ , use_token_type_embed=lowerCamelCase_ , token_type_vocab_size=predefined_args['token_type_vocab_size'] , use_classifier=lowerCamelCase_ , use_decoder=lowerCamelCase_ , )
original_bort.load_parameters(lowerCamelCase_ , cast_dtype=lowerCamelCase_ , ignore_extra=lowerCamelCase_ )
__a : Dict = original_bort._collect_params_with_prefix()
# Build our config 🤗
__a : Optional[Any] = {
'architectures': ['BertForMaskedLM'],
'attention_probs_dropout_prob': predefined_args['dropout'],
'hidden_act': 'gelu',
'hidden_dropout_prob': predefined_args['dropout'],
'hidden_size': predefined_args['embed_size'],
'initializer_range': 0.02,
'intermediate_size': predefined_args['hidden_size'],
'layer_norm_eps': predefined_args['layer_norm_eps'],
'max_position_embeddings': predefined_args['max_length'],
'model_type': 'bort',
'num_attention_heads': predefined_args['num_heads'],
'num_hidden_layers': predefined_args['num_layers'],
'pad_token_id': 1, # 2 = BERT, 1 = RoBERTa
'type_vocab_size': 1, # 2 = BERT, 1 = RoBERTa
'vocab_size': len(lowerCamelCase_ ),
}
__a : str = BertConfig.from_dict(lowerCamelCase_ )
__a : Optional[int] = BertForMaskedLM(lowerCamelCase_ )
hf_bort_model.eval()
# Parameter mapping table (Gluonnlp to Transformers)
# * denotes layer index
#
# | Gluon Parameter | Transformers Parameter
# | -------------------------------------------------------------- | ----------------------
# | `encoder.layer_norm.beta` | `bert.embeddings.LayerNorm.bias`
# | `encoder.layer_norm.gamma` | `bert.embeddings.LayerNorm.weight`
# | `encoder.position_weight` | `bert.embeddings.position_embeddings.weight`
# | `word_embed.0.weight` | `bert.embeddings.word_embeddings.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_key.bias` | `bert.encoder.layer.*.attention.self.key.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_key.weight` | `bert.encoder.layer.*.attention.self.key.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_query.bias` | `bert.encoder.layer.*.attention.self.query.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_query.weight` | `bert.encoder.layer.*.attention.self.query.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_value.bias` | `bert.encoder.layer.*.attention.self.value.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_value.weight` | `bert.encoder.layer.*.attention.self.value.weight`
# | `encoder.transformer_cells.*.ffn.ffn_2.bias` | `bert.encoder.layer.*.attention.output.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_2.weight` | `bert.encoder.layer.*.attention.output.dense.weight`
# | `encoder.transformer_cells.*.layer_norm.beta` | `bert.encoder.layer.*.attention.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.layer_norm.gamma` | `bert.encoder.layer.*.attention.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.ffn.ffn_1.bias` | `bert.encoder.layer.*.intermediate.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_1.weight` | `bert.encoder.layer.*.intermediate.dense.weight`
# | `encoder.transformer_cells.*.ffn.layer_norm.beta` | `bert.encoder.layer.*.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.ffn.layer_norm.gamma` | `bert.encoder.layer.*.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.proj.bias` | `bert.encoder.layer.*.output.dense.bias`
# | `encoder.transformer_cells.*.proj.weight` | `bert.encoder.layer.*.output.dense.weight`
# Helper function to convert MXNET Arrays to PyTorch
def to_torch(lowerCamelCase_ : Optional[Any] ) -> nn.Parameter:
return nn.Parameter(torch.FloatTensor(mx_array.data().asnumpy() ) )
# Check param shapes and map new HF param back
def check_and_map_params(lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[str] ):
__a : Optional[int] = hf_param.shape
__a : int = to_torch(params[gluon_param] )
__a : int = gluon_param.shape
assert (
shape_hf == shape_gluon
), f'''The gluon parameter {gluon_param} has shape {shape_gluon}, but expects shape {shape_hf} for Transformers'''
return gluon_param
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.word_embeddings.weight , 'word_embed.0.weight' )
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.position_embeddings.weight , 'encoder.position_weight' )
__a : Tuple = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.bias , 'encoder.layer_norm.beta' )
__a : Union[str, Any] = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.weight , 'encoder.layer_norm.gamma' )
# Inspired by RoBERTa conversion script, we just zero them out (Bort does not use them)
__a : Union[str, Any] = torch.zeros_like(
hf_bort_model.bert.embeddings.token_type_embeddings.weight.data )
for i in range(hf_bort_config.num_hidden_layers ):
__a : BertLayer = hf_bort_model.bert.encoder.layer[i]
# self attention
__a : BertSelfAttention = layer.attention.self
__a : Optional[int] = check_and_map_params(
self_attn.key.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.bias''' )
__a : str = check_and_map_params(
self_attn.key.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.weight''' )
__a : List[str] = check_and_map_params(
self_attn.query.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.bias''' )
__a : str = check_and_map_params(
self_attn.query.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.weight''' )
__a : Dict = check_and_map_params(
self_attn.value.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.bias''' )
__a : str = check_and_map_params(
self_attn.value.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.weight''' )
# self attention output
__a : BertSelfOutput = layer.attention.output
__a : Tuple = check_and_map_params(
self_output.dense.bias , f'''encoder.transformer_cells.{i}.proj.bias''' )
__a : Dict = check_and_map_params(
self_output.dense.weight , f'''encoder.transformer_cells.{i}.proj.weight''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.layer_norm.beta''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.layer_norm.gamma''' )
# intermediate
__a : BertIntermediate = layer.intermediate
__a : List[str] = check_and_map_params(
intermediate.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_1.bias''' )
__a : Optional[Any] = check_and_map_params(
intermediate.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_1.weight''' )
# output
__a : BertOutput = layer.output
__a : str = check_and_map_params(
bert_output.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_2.bias''' )
__a : List[Any] = check_and_map_params(
bert_output.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_2.weight''' )
__a : str = check_and_map_params(
bert_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.ffn.layer_norm.beta''' )
__a : List[str] = check_and_map_params(
bert_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.ffn.layer_norm.gamma''' )
# Save space and energy 🎄
hf_bort_model.half()
# Compare output of both models
__a : Union[str, Any] = RobertaTokenizer.from_pretrained('roberta-base' )
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ )['input_ids']
# Get gluon output
__a : Optional[int] = mx.nd.array([input_ids] )
__a : Tuple = original_bort(inputs=lowerCamelCase_ , token_types=[] )
# Get Transformer output (save and reload model again)
hf_bort_model.save_pretrained(lowerCamelCase_ )
__a : Optional[Any] = BertModel.from_pretrained(lowerCamelCase_ )
hf_bort_model.eval()
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ , return_tensors='pt' )
__a : int = hf_bort_model(**lowerCamelCase_ )[0]
__a : Dict = output_gluon[0].asnumpy()
__a : str = output_hf[0].detach().numpy()
__a : List[Any] = np.max(np.abs(hf_layer - gluon_layer ) ).item()
__a : str = np.allclose(lowerCamelCase_ , lowerCamelCase_ , atol=1e-3 )
if success:
print('✔️ Both model do output the same tensors' )
else:
print('❌ Both model do **NOT** output the same tensors' )
print('Absolute difference is:' , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--bort_checkpoint_path''', default=None, type=str, required=True, help='''Path the official Bort params file.'''
)
parser.add_argument(
'''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.'''
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_bort_checkpoint_to_pytorch(args.bort_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
import html
from ...feature_extraction_utils import BatchFeature, FeatureExtractionMixin
from ...utils import is_bsa_available, logging, requires_backends
if is_bsa_available():
import bsa
from bsa import BeautifulSoup
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
class _UpperCamelCase( __lowerCamelCase ):
def __init__( self : Union[str, Any] , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['bs4'] )
super().__init__(**SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Any = []
__a : Tuple = []
__a : int = element if element.name else element.parent
for parent in child.parents: # type: bs4.element.Tag
__a : Tuple = parent.find_all(child.name , recursive=SCREAMING_SNAKE_CASE__ )
xpath_tags.append(child.name )
xpath_subscripts.append(
0 if 1 == len(SCREAMING_SNAKE_CASE__ ) else next(i for i, s in enumerate(SCREAMING_SNAKE_CASE__ , 1 ) if s is child ) )
__a : Union[str, Any] = parent
xpath_tags.reverse()
xpath_subscripts.reverse()
return xpath_tags, xpath_subscripts
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
__a : Union[str, Any] = BeautifulSoup(SCREAMING_SNAKE_CASE__ , 'html.parser' )
__a : Any = []
__a : int = []
__a : Union[str, Any] = []
for element in html_code.descendants:
if type(SCREAMING_SNAKE_CASE__ ) == bsa.element.NavigableString:
if type(element.parent ) != bsa.element.Tag:
continue
__a : str = html.unescape(SCREAMING_SNAKE_CASE__ ).strip()
if not text_in_this_tag:
continue
all_doc_strings.append(SCREAMING_SNAKE_CASE__ )
__a , __a : Optional[Any] = self.xpath_soup(SCREAMING_SNAKE_CASE__ )
stringaxtag_seq.append(SCREAMING_SNAKE_CASE__ )
stringaxsubs_seq.append(SCREAMING_SNAKE_CASE__ )
if len(SCREAMING_SNAKE_CASE__ ) != len(SCREAMING_SNAKE_CASE__ ):
raise ValueError('Number of doc strings and xtags does not correspond' )
if len(SCREAMING_SNAKE_CASE__ ) != len(SCREAMING_SNAKE_CASE__ ):
raise ValueError('Number of doc strings and xsubs does not correspond' )
return all_doc_strings, stringaxtag_seq, stringaxsubs_seq
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : int = ''
for tagname, subs in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
xpath += f'''/{tagname}'''
if subs != 0:
xpath += f'''[{subs}]'''
return xpath
def __call__( self : str , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Union[str, Any] = False
# Check that strings has a valid type
if isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
__a : Union[str, Any] = True
elif isinstance(SCREAMING_SNAKE_CASE__ , (list, tuple) ):
if len(SCREAMING_SNAKE_CASE__ ) == 0 or isinstance(html_strings[0] , SCREAMING_SNAKE_CASE__ ):
__a : str = True
if not valid_strings:
raise ValueError(
'HTML strings must of type `str`, `List[str]` (batch of examples), '
f'''but is of type {type(SCREAMING_SNAKE_CASE__ )}.''' )
__a : Dict = bool(isinstance(SCREAMING_SNAKE_CASE__ , (list, tuple) ) and (isinstance(html_strings[0] , SCREAMING_SNAKE_CASE__ )) )
if not is_batched:
__a : Any = [html_strings]
# Get nodes + xpaths
__a : str = []
__a : List[str] = []
for html_string in html_strings:
__a , __a , __a : Optional[Any] = self.get_three_from_single(SCREAMING_SNAKE_CASE__ )
nodes.append(SCREAMING_SNAKE_CASE__ )
__a : Tuple = []
for node, tag_list, sub_list in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
__a : Tuple = self.construct_xpath(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
xpath_strings.append(SCREAMING_SNAKE_CASE__ )
xpaths.append(SCREAMING_SNAKE_CASE__ )
# return as Dict
__a : str = {'nodes': nodes, 'xpaths': xpaths}
__a : Tuple = BatchFeature(data=SCREAMING_SNAKE_CASE__ , tensor_type=SCREAMING_SNAKE_CASE__ )
return encoded_inputs
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[str] ):
__a : Any = ''
for i in table:
res += inp[i - 1]
return res
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
return data[1:] + data[0]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] ):
__a : Optional[int] = ''
for i in range(len(lowerCamelCase_ ) ):
if a[i] == b[i]:
res += "0"
else:
res += "1"
return res
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : str ):
__a : List[str] = int('0b' + data[0] + data[-1] , 2 )
__a : List[str] = int('0b' + data[1:3] , 2 )
return bin(s[row][col] )[2:]
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : List[str] , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] ):
__a : List[Any] = message[:4]
__a : str = message[4:]
__a : Any = apply_table(lowerCamelCase_ , lowerCamelCase_ )
__a : int = xor(lowerCamelCase_ , lowerCamelCase_ )
__a : Dict = apply_sbox(lowerCamelCase_ , temp[:4] ) # noqa: E741
__a : Tuple = apply_sbox(lowerCamelCase_ , temp[4:] )
__a : List[Any] = '0' * (2 - len(lowerCamelCase_ )) + l # noqa: E741
__a : List[str] = '0' * (2 - len(lowerCamelCase_ )) + r
__a : List[Any] = apply_table(l + r , lowerCamelCase_ )
__a : Dict = xor(lowerCamelCase_ , lowerCamelCase_ )
return temp + right
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter 10 bit key: ''')
SCREAMING_SNAKE_CASE__ = input('''Enter 8 bit message: ''')
SCREAMING_SNAKE_CASE__ = [6, 3, 7, 4, 8, 5, 10, 9]
SCREAMING_SNAKE_CASE__ = [3, 5, 2, 7, 4, 10, 1, 9, 8, 6]
SCREAMING_SNAKE_CASE__ = [2, 4, 3, 1]
SCREAMING_SNAKE_CASE__ = [2, 6, 3, 1, 4, 8, 5, 7]
SCREAMING_SNAKE_CASE__ = [4, 1, 3, 5, 7, 2, 8, 6]
SCREAMING_SNAKE_CASE__ = [4, 1, 2, 3, 2, 3, 4, 1]
SCREAMING_SNAKE_CASE__ = [[1, 0, 3, 2], [3, 2, 1, 0], [0, 2, 1, 3], [3, 1, 3, 2]]
SCREAMING_SNAKE_CASE__ = [[0, 1, 2, 3], [2, 0, 1, 3], [3, 0, 1, 0], [2, 1, 0, 3]]
# key generation
SCREAMING_SNAKE_CASE__ = apply_table(key, paa_table)
SCREAMING_SNAKE_CASE__ = temp[:5]
SCREAMING_SNAKE_CASE__ = temp[5:]
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
# encryption
SCREAMING_SNAKE_CASE__ = apply_table(message, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Cipher text is:''', CT)
# decryption
SCREAMING_SNAKE_CASE__ = apply_table(CT, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Plain text after decypting is:''', PT)
| 47 | 1 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''camembert-base''': '''https://huggingface.co/camembert-base/resolve/main/config.json''',
'''umberto-commoncrawl-cased-v1''': (
'''https://huggingface.co/Musixmatch/umberto-commoncrawl-cased-v1/resolve/main/config.json'''
),
'''umberto-wikipedia-uncased-v1''': (
'''https://huggingface.co/Musixmatch/umberto-wikipedia-uncased-v1/resolve/main/config.json'''
),
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Tuple = '''camembert'''
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : int=3_0_5_2_2 , SCREAMING_SNAKE_CASE__ : Any=7_6_8 , SCREAMING_SNAKE_CASE__ : str=1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_2 , SCREAMING_SNAKE_CASE__ : List[Any]=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : List[Any]=0.1 , SCREAMING_SNAKE_CASE__ : int=5_1_2 , SCREAMING_SNAKE_CASE__ : Any=2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0.02 , SCREAMING_SNAKE_CASE__ : Dict=1e-12 , SCREAMING_SNAKE_CASE__ : str=1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0 , SCREAMING_SNAKE_CASE__ : Dict=2 , SCREAMING_SNAKE_CASE__ : int="absolute" , SCREAMING_SNAKE_CASE__ : str=True , SCREAMING_SNAKE_CASE__ : List[str]=None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : str = vocab_size
__a : Optional[int] = hidden_size
__a : List[str] = num_hidden_layers
__a : Union[str, Any] = num_attention_heads
__a : int = hidden_act
__a : int = intermediate_size
__a : List[str] = hidden_dropout_prob
__a : Union[str, Any] = attention_probs_dropout_prob
__a : int = max_position_embeddings
__a : List[Any] = type_vocab_size
__a : str = initializer_range
__a : Tuple = layer_norm_eps
__a : Tuple = position_embedding_type
__a : Any = use_cache
__a : List[Any] = classifier_dropout
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
if self.task == "multiple-choice":
__a : Optional[Any] = {0: 'batch', 1: 'choice', 2: 'sequence'}
else:
__a : str = {0: 'batch', 1: 'sequence'}
return OrderedDict(
[
('input_ids', dynamic_axis),
('attention_mask', dynamic_axis),
] )
| 47 |
import unittest
from transformers import is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from transformers import GradientAccumulator, create_optimizer
@require_tf
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : List[Any] = GradientAccumulator()
accumulator([tf.constant([1.0, 2.0] )] )
accumulator([tf.constant([-2.0, 1.0] )] )
accumulator([tf.constant([-1.0, 2.0] )] )
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
accumulator([tf.constant([1.0, 1.0] ), tf.constant([2.0, 2.0] )] )
self.assertEqual(accumulator.step , 3 )
self.assertEqual(len(accumulator.gradients ) , 1 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [-2.0, 5.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [0.0, 0.0] , tol=1e-2 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : int = None
ops.enable_eager_execution_internal()
__a : Optional[Any] = tf.config.list_physical_devices('CPU' )
if len(SCREAMING_SNAKE_CASE__ ) == 1:
tf.config.set_logical_device_configuration(
physical_devices[0] , [tf.config.LogicalDeviceConfiguration(), tf.config.LogicalDeviceConfiguration()] )
__a : int = tf.config.list_logical_devices(device_type='CPU' )
__a : str = tf.distribute.MirroredStrategy(devices=devices[:2] )
with strategy.scope():
__a : List[str] = GradientAccumulator()
__a : Tuple = tf.Variable([4.0, 3.0] )
__a , __a : int = create_optimizer(5e-5 , 1_0 , 5 )
__a : List[Any] = tf.Variable([0.0, 0.0] , trainable=SCREAMING_SNAKE_CASE__ )
def accumulate_on_replica(SCREAMING_SNAKE_CASE__ : Optional[Any] ):
accumulator([gradient] )
def apply_on_replica():
optimizer.apply_gradients(list(zip(accumulator.gradients , [variable] ) ) )
@tf.function
def accumulate(SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
with strategy.scope():
__a : Optional[Any] = strategy.experimental_local_results(SCREAMING_SNAKE_CASE__ )
local_variables[0].assign(SCREAMING_SNAKE_CASE__ )
local_variables[1].assign(SCREAMING_SNAKE_CASE__ )
strategy.run(SCREAMING_SNAKE_CASE__ , args=(gradient_placeholder,) )
@tf.function
def apply_grad():
with strategy.scope():
strategy.run(SCREAMING_SNAKE_CASE__ )
def _check_local_values(SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int ):
__a : Union[str, Any] = strategy.experimental_local_results(accumulator._gradients[0] )
self.assertListAlmostEqual(values[0].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
self.assertListAlmostEqual(values[1].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
accumulate([1.0, 2.0] , [-1.0, 1.0] )
accumulate([3.0, -1.0] , [-1.0, -1.0] )
accumulate([-2.0, 2.0] , [3.0, -2.0] )
self.assertEqual(accumulator.step , 3 )
_check_local_values([2.0, 3.0] , [1.0, -2.0] )
apply_grad()
self.assertListAlmostEqual(variable.value() , [4.0, 3.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
_check_local_values([0.0, 0.0] , [0.0, 0.0] )
| 47 | 1 |
import unittest
from transformers import is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from transformers import GradientAccumulator, create_optimizer
@require_tf
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : List[Any] = GradientAccumulator()
accumulator([tf.constant([1.0, 2.0] )] )
accumulator([tf.constant([-2.0, 1.0] )] )
accumulator([tf.constant([-1.0, 2.0] )] )
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
accumulator([tf.constant([1.0, 1.0] ), tf.constant([2.0, 2.0] )] )
self.assertEqual(accumulator.step , 3 )
self.assertEqual(len(accumulator.gradients ) , 1 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [-2.0, 5.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [0.0, 0.0] , tol=1e-2 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : int = None
ops.enable_eager_execution_internal()
__a : Optional[Any] = tf.config.list_physical_devices('CPU' )
if len(SCREAMING_SNAKE_CASE__ ) == 1:
tf.config.set_logical_device_configuration(
physical_devices[0] , [tf.config.LogicalDeviceConfiguration(), tf.config.LogicalDeviceConfiguration()] )
__a : int = tf.config.list_logical_devices(device_type='CPU' )
__a : str = tf.distribute.MirroredStrategy(devices=devices[:2] )
with strategy.scope():
__a : List[str] = GradientAccumulator()
__a : Tuple = tf.Variable([4.0, 3.0] )
__a , __a : int = create_optimizer(5e-5 , 1_0 , 5 )
__a : List[Any] = tf.Variable([0.0, 0.0] , trainable=SCREAMING_SNAKE_CASE__ )
def accumulate_on_replica(SCREAMING_SNAKE_CASE__ : Optional[Any] ):
accumulator([gradient] )
def apply_on_replica():
optimizer.apply_gradients(list(zip(accumulator.gradients , [variable] ) ) )
@tf.function
def accumulate(SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
with strategy.scope():
__a : Optional[Any] = strategy.experimental_local_results(SCREAMING_SNAKE_CASE__ )
local_variables[0].assign(SCREAMING_SNAKE_CASE__ )
local_variables[1].assign(SCREAMING_SNAKE_CASE__ )
strategy.run(SCREAMING_SNAKE_CASE__ , args=(gradient_placeholder,) )
@tf.function
def apply_grad():
with strategy.scope():
strategy.run(SCREAMING_SNAKE_CASE__ )
def _check_local_values(SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int ):
__a : Union[str, Any] = strategy.experimental_local_results(accumulator._gradients[0] )
self.assertListAlmostEqual(values[0].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
self.assertListAlmostEqual(values[1].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
accumulate([1.0, 2.0] , [-1.0, 1.0] )
accumulate([3.0, -1.0] , [-1.0, -1.0] )
accumulate([-2.0, 2.0] , [3.0, -2.0] )
self.assertEqual(accumulator.step , 3 )
_check_local_values([2.0, 3.0] , [1.0, -2.0] )
apply_grad()
self.assertListAlmostEqual(variable.value() , [4.0, 3.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
_check_local_values([0.0, 0.0] , [0.0, 0.0] )
| 47 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''roberta-base''': '''https://huggingface.co/roberta-base/resolve/main/config.json''',
'''roberta-large''': '''https://huggingface.co/roberta-large/resolve/main/config.json''',
'''roberta-large-mnli''': '''https://huggingface.co/roberta-large-mnli/resolve/main/config.json''',
'''distilroberta-base''': '''https://huggingface.co/distilroberta-base/resolve/main/config.json''',
'''roberta-base-openai-detector''': '''https://huggingface.co/roberta-base-openai-detector/resolve/main/config.json''',
'''roberta-large-openai-detector''': '''https://huggingface.co/roberta-large-openai-detector/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = '''roberta'''
def __init__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=5_0_2_6_5 , SCREAMING_SNAKE_CASE__ : Optional[int]=7_6_8 , SCREAMING_SNAKE_CASE__ : str=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Any="gelu" , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.1 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=2 , SCREAMING_SNAKE_CASE__ : Any=0.02 , SCREAMING_SNAKE_CASE__ : List[str]=1e-12 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0 , SCREAMING_SNAKE_CASE__ : List[str]=2 , SCREAMING_SNAKE_CASE__ : Tuple="absolute" , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[str]=None , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = vocab_size
__a : Tuple = hidden_size
__a : List[str] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : str = hidden_act
__a : Optional[Any] = intermediate_size
__a : Dict = hidden_dropout_prob
__a : List[str] = attention_probs_dropout_prob
__a : Optional[Any] = max_position_embeddings
__a : Dict = type_vocab_size
__a : str = initializer_range
__a : List[str] = layer_norm_eps
__a : Optional[int] = position_embedding_type
__a : Union[str, Any] = use_cache
__a : str = classifier_dropout
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
if self.task == "multiple-choice":
__a : List[str] = {0: 'batch', 1: 'choice', 2: 'sequence'}
else:
__a : Dict = {0: 'batch', 1: 'sequence'}
return OrderedDict(
[
('input_ids', dynamic_axis),
('attention_mask', dynamic_axis),
] )
| 47 | 1 |
import argparse
import logging
import os
from pathlib import Path
from typing import Any, Dict
import pytorch_lightning as pl
from pytorch_lightning.utilities import rank_zero_info
from transformers import (
AdamW,
AutoConfig,
AutoModel,
AutoModelForPreTraining,
AutoModelForQuestionAnswering,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoModelForTokenClassification,
AutoModelWithLMHead,
AutoTokenizer,
PretrainedConfig,
PreTrainedTokenizer,
)
from transformers.optimization import (
Adafactor,
get_cosine_schedule_with_warmup,
get_cosine_with_hard_restarts_schedule_with_warmup,
get_linear_schedule_with_warmup,
get_polynomial_decay_schedule_with_warmup,
)
from transformers.utils.versions import require_version
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
require_version('''pytorch_lightning>=1.0.4''')
SCREAMING_SNAKE_CASE__ = {
'''base''': AutoModel,
'''sequence-classification''': AutoModelForSequenceClassification,
'''question-answering''': AutoModelForQuestionAnswering,
'''pretraining''': AutoModelForPreTraining,
'''token-classification''': AutoModelForTokenClassification,
'''language-modeling''': AutoModelWithLMHead,
'''summarization''': AutoModelForSeqaSeqLM,
'''translation''': AutoModelForSeqaSeqLM,
}
# update this and the import above to support new schedulers from transformers.optimization
SCREAMING_SNAKE_CASE__ = {
'''linear''': get_linear_schedule_with_warmup,
'''cosine''': get_cosine_schedule_with_warmup,
'''cosine_w_restarts''': get_cosine_with_hard_restarts_schedule_with_warmup,
'''polynomial''': get_polynomial_decay_schedule_with_warmup,
# '': get_constant_schedule, # not supported for now
# '': get_constant_schedule_with_warmup, # not supported for now
}
SCREAMING_SNAKE_CASE__ = sorted(arg_to_scheduler.keys())
SCREAMING_SNAKE_CASE__ = '''{''' + ''', '''.join(arg_to_scheduler_choices) + '''}'''
class _UpperCamelCase( pl.LightningModule ):
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : argparse.Namespace , SCREAMING_SNAKE_CASE__ : Tuple=None , SCREAMING_SNAKE_CASE__ : Tuple="base" , SCREAMING_SNAKE_CASE__ : Union[str, Any]=None , SCREAMING_SNAKE_CASE__ : str=None , SCREAMING_SNAKE_CASE__ : Optional[int]=None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
super().__init__()
# TODO: move to self.save_hyperparameters()
# self.save_hyperparameters()
# can also expand arguments into trainer signature for easier reading
self.save_hyperparameters(SCREAMING_SNAKE_CASE__ )
__a : str = 0
__a : Tuple = Path(self.hparams.output_dir )
__a : List[str] = self.hparams.cache_dir if self.hparams.cache_dir else None
if config is None:
__a : int = AutoConfig.from_pretrained(
self.hparams.config_name if self.hparams.config_name else self.hparams.model_name_or_path , **({'num_labels': num_labels} if num_labels is not None else {}) , cache_dir=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ , )
else:
__a : PretrainedConfig = config
__a : Union[str, Any] = ('encoder_layerdrop', 'decoder_layerdrop', 'dropout', 'attention_dropout')
for p in extra_model_params:
if getattr(self.hparams , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
assert hasattr(self.config , SCREAMING_SNAKE_CASE__ ), f'''model config doesn\'t have a `{p}` attribute'''
setattr(self.config , SCREAMING_SNAKE_CASE__ , getattr(self.hparams , SCREAMING_SNAKE_CASE__ ) )
if tokenizer is None:
__a : Dict = AutoTokenizer.from_pretrained(
self.hparams.tokenizer_name if self.hparams.tokenizer_name else self.hparams.model_name_or_path , cache_dir=SCREAMING_SNAKE_CASE__ , )
else:
__a : PreTrainedTokenizer = tokenizer
__a : List[str] = MODEL_MODES[mode]
if model is None:
__a : Dict = self.model_type.from_pretrained(
self.hparams.model_name_or_path , from_tf=bool('.ckpt' in self.hparams.model_name_or_path ) , config=self.config , cache_dir=SCREAMING_SNAKE_CASE__ , )
else:
__a : List[str] = model
def __lowerCAmelCase ( self : Dict , *SCREAMING_SNAKE_CASE__ : str , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
__a : List[str] = self.model_type.from_pretrained(*SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : Union[str, Any] = arg_to_scheduler[self.hparams.lr_scheduler]
__a : str = get_schedule_func(
self.opt , num_warmup_steps=self.hparams.warmup_steps , num_training_steps=self.total_steps() )
__a : int = {'scheduler': scheduler, 'interval': 'step', 'frequency': 1}
return scheduler
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : int = self.model
__a : Any = ['bias', 'LayerNorm.weight']
__a : Tuple = [
{
'params': [
p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay )
], # check this named paramters
'weight_decay': self.hparams.weight_decay,
},
{
'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay )],
'weight_decay': 0.0,
},
]
if self.hparams.adafactor:
__a : Union[str, Any] = Adafactor(
SCREAMING_SNAKE_CASE__ , lr=self.hparams.learning_rate , scale_parameter=SCREAMING_SNAKE_CASE__ , relative_step=SCREAMING_SNAKE_CASE__ )
else:
__a : str = AdamW(
SCREAMING_SNAKE_CASE__ , lr=self.hparams.learning_rate , eps=self.hparams.adam_epsilon )
__a : Any = optimizer
__a : str = self.get_lr_scheduler()
return [optimizer], [scheduler]
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
return self.validation_step(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.validation_end(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : List[Any] = max(1 , self.hparams.gpus ) # TODO: consider num_tpu_cores
__a : int = self.hparams.train_batch_size * self.hparams.accumulate_grad_batches * num_devices
return (self.dataset_size / effective_batch_size) * self.hparams.max_epochs
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
if stage == "test":
__a : Any = len(self.test_dataloader().dataset )
else:
__a : Any = self.get_dataloader('train' , self.hparams.train_batch_size , shuffle=SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = len(self.train_dataloader().dataset )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : bool = False ):
'''simple docstring'''
raise NotImplementedError('You must implement this for your task' )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
return self.train_loader
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return self.get_dataloader('dev' , self.hparams.eval_batch_size , shuffle=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
return self.get_dataloader('test' , self.hparams.eval_batch_size , shuffle=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
return os.path.join(
self.hparams.data_dir , 'cached_{}_{}_{}'.format(
SCREAMING_SNAKE_CASE__ , list(filter(SCREAMING_SNAKE_CASE__ , self.hparams.model_name_or_path.split('/' ) ) ).pop() , str(self.hparams.max_seq_length ) , ) , )
@pl.utilities.rank_zero_only
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Dict[str, Any] ):
'''simple docstring'''
__a : Optional[Any] = self.output_dir.joinpath('best_tfmr' )
__a : int = self.step_count
self.model.save_pretrained(SCREAMING_SNAKE_CASE__ )
self.tokenizer.save_pretrained(SCREAMING_SNAKE_CASE__ )
@staticmethod
def __lowerCAmelCase ( SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
parser.add_argument(
'--model_name_or_path' , default=SCREAMING_SNAKE_CASE__ , type=SCREAMING_SNAKE_CASE__ , required=SCREAMING_SNAKE_CASE__ , help='Path to pretrained model or model identifier from huggingface.co/models' , )
parser.add_argument(
'--config_name' , default='' , type=SCREAMING_SNAKE_CASE__ , help='Pretrained config name or path if not the same as model_name' )
parser.add_argument(
'--tokenizer_name' , default=SCREAMING_SNAKE_CASE__ , type=SCREAMING_SNAKE_CASE__ , help='Pretrained tokenizer name or path if not the same as model_name' , )
parser.add_argument(
'--cache_dir' , default=str(Path(SCREAMING_SNAKE_CASE__ ).parent / 'test_run' / 'cache' ) , type=SCREAMING_SNAKE_CASE__ , help='Where do you want to store the pre-trained models downloaded from huggingface.co' , )
parser.add_argument(
'--encoder_layerdrop' , type=SCREAMING_SNAKE_CASE__ , help='Encoder layer dropout probability (Optional). Goes into model.config' , )
parser.add_argument(
'--decoder_layerdrop' , type=SCREAMING_SNAKE_CASE__ , help='Decoder layer dropout probability (Optional). Goes into model.config' , )
parser.add_argument(
'--dropout' , type=SCREAMING_SNAKE_CASE__ , help='Dropout probability (Optional). Goes into model.config' , )
parser.add_argument(
'--attention_dropout' , type=SCREAMING_SNAKE_CASE__ , help='Attention dropout probability (Optional). Goes into model.config' , )
parser.add_argument('--learning_rate' , default=5e-5 , type=SCREAMING_SNAKE_CASE__ , help='The initial learning rate for Adam.' )
parser.add_argument(
'--lr_scheduler' , default='linear' , choices=SCREAMING_SNAKE_CASE__ , metavar=SCREAMING_SNAKE_CASE__ , type=SCREAMING_SNAKE_CASE__ , help='Learning rate scheduler' , )
parser.add_argument('--weight_decay' , default=0.0 , type=SCREAMING_SNAKE_CASE__ , help='Weight decay if we apply some.' )
parser.add_argument('--adam_epsilon' , default=1e-8 , type=SCREAMING_SNAKE_CASE__ , help='Epsilon for Adam optimizer.' )
parser.add_argument('--warmup_steps' , default=0 , type=SCREAMING_SNAKE_CASE__ , help='Linear warmup over warmup_steps.' )
parser.add_argument('--num_workers' , default=4 , type=SCREAMING_SNAKE_CASE__ , help='kwarg passed to DataLoader' )
parser.add_argument('--num_train_epochs' , dest='max_epochs' , default=3 , type=SCREAMING_SNAKE_CASE__ )
parser.add_argument('--train_batch_size' , default=3_2 , type=SCREAMING_SNAKE_CASE__ )
parser.add_argument('--eval_batch_size' , default=3_2 , type=SCREAMING_SNAKE_CASE__ )
parser.add_argument('--adafactor' , action='store_true' )
class _UpperCamelCase( pl.Callback ):
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
if (
trainer.is_global_zero and trainer.global_rank == 0
): # we initialize the retriever only on master worker with RAY. In new pytorch-lightning accelorators are removed.
pl_module.model.rag.retriever.init_retrieval() # better to use hook functions.
class _UpperCamelCase( pl.Callback ):
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
for name, param in pl_module.model.rag.named_parameters():
if param.grad is None:
print(SCREAMING_SNAKE_CASE__ )
class _UpperCamelCase( pl.Callback ):
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Optional[int] = trainer.lr_schedulers[0]['scheduler']
__a : List[str] = {f'''lr_group_{i}''': lr for i, lr in enumerate(lr_scheduler.get_lr() )}
pl_module.logger.log_metrics(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : pl.Trainer , SCREAMING_SNAKE_CASE__ : pl.LightningModule ):
'''simple docstring'''
rank_zero_info('***** Validation results *****' )
__a : List[str] = trainer.callback_metrics
# Log results
for key in sorted(SCREAMING_SNAKE_CASE__ ):
if key not in ["log", "progress_bar"]:
rank_zero_info('{} = {}\n'.format(SCREAMING_SNAKE_CASE__ , str(metrics[key] ) ) )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : pl.Trainer , SCREAMING_SNAKE_CASE__ : pl.LightningModule ):
'''simple docstring'''
rank_zero_info('***** Test results *****' )
__a : List[str] = trainer.callback_metrics
# Log and save results to file
__a : Optional[int] = os.path.join(pl_module.hparams.output_dir , 'test_results.txt' )
with open(SCREAMING_SNAKE_CASE__ , 'w' ) as writer:
for key in sorted(SCREAMING_SNAKE_CASE__ ):
if key not in ["log", "progress_bar"]:
rank_zero_info('{} = {}\n'.format(SCREAMING_SNAKE_CASE__ , str(metrics[key] ) ) )
writer.write('{} = {}\n'.format(SCREAMING_SNAKE_CASE__ , str(metrics[key] ) ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple , lowerCamelCase_ : Tuple ):
# To allow all pl args uncomment the following line
# parser = pl.Trainer.add_argparse_args(parser)
parser.add_argument(
'--output_dir' , default=str(Path(lowerCamelCase_ ).parent / 'test_run' / 'model_checkpoints' ) , type=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
parser.add_argument(
'--fp16' , action='store_true' , help='Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit' , )
parser.add_argument(
'--fp16_opt_level' , type=lowerCamelCase_ , default='O2' , help=(
'For fp16: Apex AMP optimization level selected in [\'O0\', \'O1\', \'O2\', and \'O3\'].'
'See details at https://nvidia.github.io/apex/amp.html'
) , )
parser.add_argument('--n_tpu_cores' , dest='tpu_cores' , type=lowerCamelCase_ )
parser.add_argument('--max_grad_norm' , dest='gradient_clip_val' , default=1.0 , type=lowerCamelCase_ , help='Max gradient norm' )
parser.add_argument('--do_train' , action='store_true' , help='Whether to run training.' )
parser.add_argument('--do_predict' , action='store_true' , help='Whether to run predictions on the test set.' )
parser.add_argument(
'--gradient_accumulation_steps' , dest='accumulate_grad_batches' , type=lowerCamelCase_ , default=1 , help='Number of updates steps to accumulate before performing a backward/update pass.' , )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 , help='random seed for initialization' )
parser.add_argument(
'--data_dir' , default=str(Path(lowerCamelCase_ ).parent / 'test_run' / 'dummy-train-data' ) , type=lowerCamelCase_ , help='The input data dir. Should contain the training files for the CoNLL-2003 NER task.' , )
def UpperCAmelCase__ ( lowerCamelCase_ : BaseTransformer , lowerCamelCase_ : argparse.Namespace , lowerCamelCase_ : Any=None , lowerCamelCase_ : Dict=True , lowerCamelCase_ : Optional[Any]=[] , lowerCamelCase_ : List[Any]=None , lowerCamelCase_ : List[Any]=None , **lowerCamelCase_ : List[str] , ):
pl.seed_everything(args.seed )
# init model
__a : str = Path(model.hparams.output_dir )
odir.mkdir(exist_ok=lowerCamelCase_ )
# add custom checkpoints
if checkpoint_callback is None:
__a : Optional[Any] = pl.callbacks.ModelCheckpoint(
filepath=args.output_dir , prefix='checkpoint' , monitor='val_loss' , mode='min' , save_top_k=1 )
if early_stopping_callback:
extra_callbacks.append(lowerCamelCase_ )
if logging_callback is None:
__a : int = LoggingCallback()
__a : int = {}
if args.fpaa:
__a : Tuple = 1_6
if args.gpus > 1:
__a : int = 'auto'
__a : Optional[int] = 'ddp'
__a : Any = args.accumulate_grad_batches
__a : Union[str, Any] = None
__a : Dict = 'auto'
__a : Union[str, Any] = pl.Trainer.from_argparse_args(
lowerCamelCase_ , weights_summary=lowerCamelCase_ , callbacks=[logging_callback] + extra_callbacks + [InitCallback()] + [checkpoint_callback] , logger=lowerCamelCase_ , val_check_interval=1 , num_sanity_val_steps=2 , **lowerCamelCase_ , )
if args.do_train:
trainer.fit(lowerCamelCase_ )
else:
print('RAG modeling tests with new set functions successfuly executed!' )
return trainer
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''▁'''
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''sentencepiece.bpe.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''facebook/xglm-564M''': '''https://huggingface.co/facebook/xglm-564M/resolve/main/sentencepiece.bpe.model''',
}
}
SCREAMING_SNAKE_CASE__ = {
'''facebook/xglm-564M''': 2048,
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : Union[str, Any] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : str = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : Any = ['''input_ids''', '''attention_mask''']
def __init__( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : List[str]="</s>" , SCREAMING_SNAKE_CASE__ : Optional[Any]="</s>" , SCREAMING_SNAKE_CASE__ : Union[str, Any]="<s>" , SCREAMING_SNAKE_CASE__ : str="<unk>" , SCREAMING_SNAKE_CASE__ : Dict="<pad>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : List[str] , ):
'''simple docstring'''
__a : Union[str, Any] = {} if sp_model_kwargs is None else sp_model_kwargs
# Compatibility with the original tokenizer
__a : Any = 7
__a : Union[str, Any] = [f'''<madeupword{i}>''' for i in range(self.num_madeup_words )]
__a : Union[str, Any] = kwargs.get('additional_special_tokens' , [] )
kwargs["additional_special_tokens"] += [
word for word in madeup_words if word not in kwargs["additional_special_tokens"]
]
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , cls_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : List[str] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(str(SCREAMING_SNAKE_CASE__ ) )
__a : str = vocab_file
# Original fairseq vocab and spm vocab must be "aligned":
# Vocab | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
# -------- | ------- | ------- | ------ | ------- | --- | --- | --- | ----- | ----- | ----
# fairseq | '<s>' | '<pad>' | '</s>' | '<unk>' | ',' | '.' | '▁' | 's' | '▁de' | '-'
# spm | '<unk>' | '<s>' | '</s>' | ',' | '.' | '▁' | 's' | '▁de' | '-' | '▁a'
# The first "real" token "," has position 4 in the original fairseq vocab and position 3 in the spm vocab
__a : Any = 1
# Mimic fairseq token-to-id alignment for the first 4 token
__a : str = {'<s>': 0, '<pad>': 1, '</s>': 2, '<unk>': 3}
__a : List[str] = len(self.sp_model )
__a : Optional[int] = {f'''<madeupword{i}>''': sp_size + i + self.fairseq_offset for i in range(self.num_madeup_words )}
self.fairseq_tokens_to_ids.update(SCREAMING_SNAKE_CASE__ )
__a : Dict = {v: k for k, v in self.fairseq_tokens_to_ids.items()}
def __getstate__( self : List[str] ):
'''simple docstring'''
__a : Tuple = self.__dict__.copy()
__a : List[str] = None
__a : Optional[int] = self.sp_model.serialized_model_proto()
return state
def __setstate__( self : List[str] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : int = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : Dict = {}
__a : Tuple = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.LoadFromSerializedProto(self.sp_model_proto )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
if token_ids_a is None:
return [self.sep_token_id] + token_ids_a
__a : Optional[Any] = [self.sep_token_id]
return sep + token_ids_a + sep + sep + token_ids_a
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : bool = False ):
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=SCREAMING_SNAKE_CASE__ , token_ids_a=SCREAMING_SNAKE_CASE__ , already_has_special_tokens=SCREAMING_SNAKE_CASE__ )
if token_ids_a is None:
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ )) + [1, 1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
__a : Optional[int] = [self.sep_token_id]
if token_ids_a is None:
return len(sep + token_ids_a ) * [0]
return len(sep + token_ids_a + sep + sep + token_ids_a ) * [0]
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return len(self.sp_model ) + self.fairseq_offset + self.num_madeup_words
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : str = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if token in self.fairseq_tokens_to_ids:
return self.fairseq_tokens_to_ids[token]
__a : List[str] = self.sp_model.PieceToId(SCREAMING_SNAKE_CASE__ )
# Need to return unknown token if the SP model returned 0
return spm_id + self.fairseq_offset if spm_id else self.unk_token_id
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if index in self.fairseq_ids_to_tokens:
return self.fairseq_ids_to_tokens[index]
return self.sp_model.IdToPiece(index - self.fairseq_offset )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
__a : Optional[int] = ''.join(SCREAMING_SNAKE_CASE__ ).replace(SCREAMING_SNAKE_CASE__ , ' ' ).strip()
return out_string
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Any = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[Any] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
from __future__ import annotations
import math
SCREAMING_SNAKE_CASE__ = '''2020.9.26'''
SCREAMING_SNAKE_CASE__ = '''xcodz-dot, cclaus, dhruvmanila'''
def UpperCAmelCase__ ( lowerCamelCase_ : float , lowerCamelCase_ : float , lowerCamelCase_ : float , lowerCamelCase_ : float , lowerCamelCase_ : float ):
if not all(isinstance(lowerCamelCase_ , (float, int) ) for val in locals().values() ):
__a : int = f'''Input values must either be float or int: {list(locals().values() )}'''
raise TypeError(lowerCamelCase_ )
__a : Tuple = ((x * distance) / (z + distance)) * scale
__a : List[str] = ((y * distance) / (z + distance)) * scale
return projected_x, projected_y
def UpperCAmelCase__ ( lowerCamelCase_ : float , lowerCamelCase_ : float , lowerCamelCase_ : float , lowerCamelCase_ : str , lowerCamelCase_ : float ):
if not isinstance(lowerCamelCase_ , lowerCamelCase_ ):
raise TypeError('Axis must be a str' )
__a : Optional[Any] = locals()
del input_variables["axis"]
if not all(isinstance(lowerCamelCase_ , (float, int) ) for val in input_variables.values() ):
__a : Any = (
'Input values except axis must either be float or int: '
f'''{list(input_variables.values() )}'''
)
raise TypeError(lowerCamelCase_ )
__a : str = (angle % 3_6_0) / 4_5_0 * 1_8_0 / math.pi
if axis == "z":
__a : Dict = x * math.cos(lowerCamelCase_ ) - y * math.sin(lowerCamelCase_ )
__a : Optional[Any] = y * math.cos(lowerCamelCase_ ) + x * math.sin(lowerCamelCase_ )
__a : List[str] = z
elif axis == "x":
__a : int = y * math.cos(lowerCamelCase_ ) - z * math.sin(lowerCamelCase_ )
__a : Tuple = z * math.cos(lowerCamelCase_ ) + y * math.sin(lowerCamelCase_ )
__a : str = x
elif axis == "y":
__a : Any = x * math.cos(lowerCamelCase_ ) - z * math.sin(lowerCamelCase_ )
__a : List[str] = z * math.cos(lowerCamelCase_ ) + x * math.sin(lowerCamelCase_ )
__a : List[Any] = y
else:
raise ValueError('not a valid axis, choose one of \'x\', \'y\', \'z\'' )
return new_x, new_y, new_z
if __name__ == "__main__":
import doctest
doctest.testmod()
print(F"{convert_to_ad(1.0, 2.0, 3.0, 1_0.0, 1_0.0) = }")
print(F"{rotate(1.0, 2.0, 3.0, 'y', 9_0.0) = }")
| 47 |
import argparse
from collections import OrderedDict
from pathlib import Path
import torch
from transformers import (
VisualBertConfig,
VisualBertForMultipleChoice,
VisualBertForPreTraining,
VisualBertForQuestionAnswering,
VisualBertForVisualReasoning,
)
from transformers.utils import logging
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = [
('''bert.bert''', '''visual_bert'''),
('''bert.cls''', '''cls'''),
('''bert.classifier''', '''cls'''),
('''token_type_embeddings_visual''', '''visual_token_type_embeddings'''),
('''position_embeddings_visual''', '''visual_position_embeddings'''),
('''projection''', '''visual_projection'''),
]
SCREAMING_SNAKE_CASE__ = [
'''nlvr2_coco_pre_trained.th''',
'''nlvr2_fine_tuned.th''',
'''nlvr2_pre_trained.th''',
'''vcr_coco_pre_train.th''',
'''vcr_fine_tune.th''',
'''vcr_pre_train.th''',
'''vqa_coco_pre_trained.th''',
'''vqa_fine_tuned.th''',
'''vqa_pre_trained.th''',
]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ):
__a : str = torch.load(lowerCamelCase_ , map_location='cpu' )
return sd
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : int , lowerCamelCase_ : Dict=rename_keys_prefix ):
__a : Optional[Any] = OrderedDict()
__a : Any = torch.arange(config.max_position_embeddings ).expand((1, -1) )
# detector_d = OrderedDict()
for key in d:
if "detector" in key:
# detector_d[key.replace('detector.','')] = d[key]
continue
__a : List[Any] = key
for name_pair in rename_keys_prefix:
__a : List[str] = new_key.replace(name_pair[0] , name_pair[1] )
__a : Any = d[key]
if key == "bert.cls.predictions.decoder.weight":
# Old bert code didn't have `decoder.bias`, but was added separately
__a : int = new_d['cls.predictions.bias']
return new_d
@torch.no_grad()
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : Any ):
assert (
checkpoint_path.split('/' )[-1] in ACCEPTABLE_CHECKPOINTS
), f'''The checkpoint provided must be in {ACCEPTABLE_CHECKPOINTS}.'''
# Get Config
if "pre" in checkpoint_path:
__a : Dict = 'pretraining'
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
elif "vqa_advanced" in checkpoint_path:
__a : int = {'visual_embedding_dim': 2_0_4_8}
elif "vqa" in checkpoint_path:
__a : Tuple = {'visual_embedding_dim': 2_0_4_8}
elif "nlvr" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 1_0_2_4}
else:
raise NotImplementedError(f'''No implementation found for `{checkpoint_path}`.''' )
else:
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
__a : Any = 'multichoice'
elif "vqa_advanced" in checkpoint_path:
__a : Any = {'visual_embedding_dim': 2_0_4_8}
__a : List[str] = 'vqa_advanced'
elif "vqa" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 2_0_4_8, 'num_labels': 3_1_2_9}
__a : List[Any] = 'vqa'
elif "nlvr" in checkpoint_path:
__a : Optional[int] = {
'visual_embedding_dim': 1_0_2_4,
'num_labels': 2,
}
__a : Optional[Any] = 'nlvr'
__a : str = VisualBertConfig(**lowerCamelCase_ )
# Load State Dict
__a : str = load_state_dict(lowerCamelCase_ )
__a : str = get_new_dict(lowerCamelCase_ , lowerCamelCase_ )
if model_type == "pretraining":
__a : Optional[Any] = VisualBertForPreTraining(lowerCamelCase_ )
elif model_type == "vqa":
__a : Any = VisualBertForQuestionAnswering(lowerCamelCase_ )
elif model_type == "nlvr":
__a : int = VisualBertForVisualReasoning(lowerCamelCase_ )
elif model_type == "multichoice":
__a : Optional[int] = VisualBertForMultipleChoice(lowerCamelCase_ )
model.load_state_dict(lowerCamelCase_ )
# Save Checkpoints
Path(lowerCamelCase_ ).mkdir(exist_ok=lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument('''orig_checkpoint_path''', type=str, help='''A path to .th on local filesystem.''')
parser.add_argument('''pytorch_dump_folder_path''', type=str, help='''Path to the output PyTorch model.''')
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_visual_bert_checkpoint(args.orig_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Union[str, Any] ):
# "extended trapezoidal rule"
# int(f) = dx/2 * (f1 + 2f2 + ... + fn)
__a : str = (boundary[1] - boundary[0]) / steps
__a : Union[str, Any] = boundary[0]
__a : Any = boundary[1]
__a : List[Any] = make_points(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
__a : Union[str, Any] = 0.0
y += (h / 2.0) * f(lowerCamelCase_ )
for i in x_i:
# print(i)
y += h * f(lowerCamelCase_ )
y += (h / 2.0) * f(lowerCamelCase_ )
return y
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : List[str] , lowerCamelCase_ : Tuple ):
__a : Dict = a + h
while x < (b - h):
yield x
__a : Optional[int] = x + h
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ): # enter your function here
__a : Optional[Any] = (x - 0) * (x - 0)
return y
def UpperCAmelCase__ ( ):
__a : Optional[int] = 0.0 # Lower bound of integration
__a : Union[str, Any] = 1.0 # Upper bound of integration
__a : Tuple = 10.0 # define number of steps or resolution
__a : Optional[Any] = [a, b] # define boundary of integration
__a : Tuple = method_a(lowerCamelCase_ , lowerCamelCase_ )
print(f'''y = {y}''' )
if __name__ == "__main__":
main()
| 47 |
print((lambda quine: quine % quine)('''print((lambda quine: quine %% quine)(%r))'''))
| 47 | 1 |
from __future__ import annotations
from bisect import bisect_left
from functools import total_ordering
from heapq import merge
@total_ordering
class _UpperCamelCase( __lowerCamelCase ):
def __lt__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
return self[-1] < other[-1]
def __eq__( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
return self[-1] == other[-1]
def UpperCAmelCase__ ( lowerCamelCase_ : list ):
__a : list[Stack] = []
# sort into stacks
for element in collection:
__a : Dict = Stack([element] )
__a : Union[str, Any] = bisect_left(lowerCamelCase_ , lowerCamelCase_ )
if i != len(lowerCamelCase_ ):
stacks[i].append(lowerCamelCase_ )
else:
stacks.append(lowerCamelCase_ )
# use a heap-based merge to merge stack efficiently
__a : List[str] = merge(*(reversed(lowerCamelCase_ ) for stack in stacks) )
return collection
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter numbers separated by a comma:\n''').strip()
SCREAMING_SNAKE_CASE__ = [int(item) for item in user_input.split(''',''')]
print(patience_sort(unsorted))
| 47 |
import json
import os
import shutil
import tempfile
from unittest import TestCase
from transformers import BartTokenizer, BartTokenizerFast, DPRQuestionEncoderTokenizer, DPRQuestionEncoderTokenizerFast
from transformers.models.bart.configuration_bart import BartConfig
from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES
from transformers.models.dpr.configuration_dpr import DPRConfig
from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
from transformers.testing_utils import require_faiss, require_tokenizers, require_torch, slow
from transformers.utils import is_datasets_available, is_faiss_available, is_torch_available
if is_torch_available() and is_datasets_available() and is_faiss_available():
from transformers.models.rag.configuration_rag import RagConfig
from transformers.models.rag.tokenization_rag import RagTokenizer
@require_faiss
@require_torch
class _UpperCamelCase( __lowerCamelCase ):
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[Any] = tempfile.mkdtemp()
__a : int = 8
# DPR tok
__a : Dict = [
'[UNK]',
'[CLS]',
'[SEP]',
'[PAD]',
'[MASK]',
'want',
'##want',
'##ed',
'wa',
'un',
'runn',
'##ing',
',',
'low',
'lowest',
]
__a : int = os.path.join(self.tmpdirname , 'dpr_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , DPR_VOCAB_FILES_NAMES['vocab_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as vocab_writer:
vocab_writer.write(''.join([x + '\n' for x in vocab_tokens] ) )
# BART tok
__a : str = [
'l',
'o',
'w',
'e',
'r',
's',
't',
'i',
'd',
'n',
'\u0120',
'\u0120l',
'\u0120n',
'\u0120lo',
'\u0120low',
'er',
'\u0120lowest',
'\u0120newer',
'\u0120wider',
'<unk>',
]
__a : Optional[int] = dict(zip(SCREAMING_SNAKE_CASE__ , range(len(SCREAMING_SNAKE_CASE__ ) ) ) )
__a : List[str] = ['#version: 0.2', '\u0120 l', '\u0120l o', '\u0120lo w', 'e r', '']
__a : List[str] = {'unk_token': '<unk>'}
__a : Dict = os.path.join(self.tmpdirname , 'bart_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['vocab_file'] )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(SCREAMING_SNAKE_CASE__ ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'dpr_tokenizer' ) )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'bart_tokenizer' ) )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
shutil.rmtree(self.tmpdirname )
@require_tokenizers
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Tuple = os.path.join(self.tmpdirname , 'rag_tokenizer' )
__a : Optional[Any] = RagConfig(question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() )
__a : Optional[Any] = RagTokenizer(question_encoder=self.get_dpr_tokenizer() , generator=self.get_bart_tokenizer() )
rag_config.save_pretrained(SCREAMING_SNAKE_CASE__ )
rag_tokenizer.save_pretrained(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = RagTokenizer.from_pretrained(SCREAMING_SNAKE_CASE__ , config=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(new_rag_tokenizer.question_encoder , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.question_encoder.get_vocab() , rag_tokenizer.question_encoder.get_vocab() )
self.assertIsInstance(new_rag_tokenizer.generator , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.generator.get_vocab() , rag_tokenizer.generator.get_vocab() )
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Optional[Any] = RagTokenizer.from_pretrained('facebook/rag-token-nq' )
__a : List[Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : Tuple = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Any = RagTokenizer.from_pretrained('facebook/rag-sequence-nq' )
__a : Union[str, Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : str = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
| 47 | 1 |
import json
import os
import tempfile
import unittest
import unittest.mock as mock
from pathlib import Path
from requests.exceptions import HTTPError
from transformers.utils import (
CONFIG_NAME,
FLAX_WEIGHTS_NAME,
TF2_WEIGHTS_NAME,
TRANSFORMERS_CACHE,
WEIGHTS_NAME,
cached_file,
get_file_from_repo,
has_file,
)
SCREAMING_SNAKE_CASE__ = '''hf-internal-testing/tiny-random-bert'''
SCREAMING_SNAKE_CASE__ = os.path.join(TRANSFORMERS_CACHE, '''models--hf-internal-testing--tiny-random-bert''')
SCREAMING_SNAKE_CASE__ = '''9b8c223d42b2188cb49d29af482996f9d0f3e5a6'''
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Tuple = cached_file(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# Should have downloaded the file in here
self.assertTrue(os.path.isdir(SCREAMING_SNAKE_CASE__ ) )
# Cache should contain at least those three subfolders:
for subfolder in ["blobs", "refs", "snapshots"]:
self.assertTrue(os.path.isdir(os.path.join(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) ) )
with open(os.path.join(SCREAMING_SNAKE_CASE__ , 'refs' , 'main' ) ) as f:
__a : List[str] = f.read()
self.assertEqual(SCREAMING_SNAKE_CASE__ , os.path.join(SCREAMING_SNAKE_CASE__ , 'snapshots' , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
self.assertTrue(os.path.isfile(SCREAMING_SNAKE_CASE__ ) )
# File is cached at the same place the second time.
__a : Any = cached_file(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# Using a specific revision to test the full commit hash.
__a : Tuple = cached_file(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , revision='9b8c223' )
self.assertEqual(SCREAMING_SNAKE_CASE__ , os.path.join(SCREAMING_SNAKE_CASE__ , 'snapshots' , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
with self.assertRaisesRegex(SCREAMING_SNAKE_CASE__ , 'is not a valid model identifier' ):
__a : Union[str, Any] = cached_file('tiny-random-bert' , SCREAMING_SNAKE_CASE__ )
with self.assertRaisesRegex(SCREAMING_SNAKE_CASE__ , 'is not a valid git identifier' ):
__a : List[str] = cached_file(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , revision='aaaa' )
with self.assertRaisesRegex(SCREAMING_SNAKE_CASE__ , 'does not appear to have a file named' ):
__a : Dict = cached_file(SCREAMING_SNAKE_CASE__ , 'conf' )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
with self.assertRaisesRegex(SCREAMING_SNAKE_CASE__ , 'does not appear to have a file named' ):
__a : Any = cached_file(SCREAMING_SNAKE_CASE__ , 'conf' )
with open(os.path.join(SCREAMING_SNAKE_CASE__ , 'refs' , 'main' ) ) as f:
__a : Any = f.read()
self.assertTrue(os.path.isfile(os.path.join(SCREAMING_SNAKE_CASE__ , '.no_exist' , SCREAMING_SNAKE_CASE__ , 'conf' ) ) )
__a : List[Any] = cached_file(SCREAMING_SNAKE_CASE__ , 'conf' , _raise_exceptions_for_missing_entries=SCREAMING_SNAKE_CASE__ )
self.assertIsNone(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = cached_file(SCREAMING_SNAKE_CASE__ , 'conf' , local_files_only=SCREAMING_SNAKE_CASE__ , _raise_exceptions_for_missing_entries=SCREAMING_SNAKE_CASE__ )
self.assertIsNone(SCREAMING_SNAKE_CASE__ )
__a : int = mock.Mock()
__a : List[Any] = 5_0_0
__a : Dict = {}
__a : str = HTTPError
__a : Optional[int] = {}
# Under the mock environment we get a 500 error when trying to reach the tokenizer.
with mock.patch('requests.Session.request' , return_value=SCREAMING_SNAKE_CASE__ ) as mock_head:
__a : Optional[int] = cached_file(SCREAMING_SNAKE_CASE__ , 'conf' , _raise_exceptions_for_connection_errors=SCREAMING_SNAKE_CASE__ )
self.assertIsNone(SCREAMING_SNAKE_CASE__ )
# This check we did call the fake head request
mock_head.assert_called()
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
self.assertTrue(has_file('hf-internal-testing/tiny-bert-pt-only' , SCREAMING_SNAKE_CASE__ ) )
self.assertFalse(has_file('hf-internal-testing/tiny-bert-pt-only' , SCREAMING_SNAKE_CASE__ ) )
self.assertFalse(has_file('hf-internal-testing/tiny-bert-pt-only' , SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
self.assertIsNone(get_file_from_repo('bert-base-cased' , 'ahah.txt' ) )
# The function raises if the repository does not exist.
with self.assertRaisesRegex(SCREAMING_SNAKE_CASE__ , 'is not a valid model identifier' ):
get_file_from_repo('bert-base-case' , SCREAMING_SNAKE_CASE__ )
# The function raises if the revision does not exist.
with self.assertRaisesRegex(SCREAMING_SNAKE_CASE__ , 'is not a valid git identifier' ):
get_file_from_repo('bert-base-cased' , SCREAMING_SNAKE_CASE__ , revision='ahaha' )
__a : Optional[Any] = get_file_from_repo('bert-base-cased' , SCREAMING_SNAKE_CASE__ )
# The name is the cached name which is not very easy to test, so instead we load the content.
__a : Optional[Any] = json.loads(open(SCREAMING_SNAKE_CASE__ , 'r' ).read() )
self.assertEqual(config['hidden_size'] , 7_6_8 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
with tempfile.TemporaryDirectory() as tmp_dir:
__a : List[str] = Path(SCREAMING_SNAKE_CASE__ ) / 'a.txt'
filename.touch()
self.assertEqual(get_file_from_repo(SCREAMING_SNAKE_CASE__ , 'a.txt' ) , str(SCREAMING_SNAKE_CASE__ ) )
self.assertIsNone(get_file_from_repo(SCREAMING_SNAKE_CASE__ , 'b.txt' ) )
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''spiece.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''bert_for_seq_generation''': (
'''https://huggingface.co/google/bert_for_seq_generation_L-24_bbc_encoder/resolve/main/spiece.model'''
),
}
}
SCREAMING_SNAKE_CASE__ = {'''bert_for_seq_generation''': 512}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : Any = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : List[int] = []
__SCREAMING_SNAKE_CASE : int = ['''input_ids''', '''attention_mask''']
def __init__( self : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : Tuple="</s>" , SCREAMING_SNAKE_CASE__ : Any="<unk>" , SCREAMING_SNAKE_CASE__ : int="<pad>" , SCREAMING_SNAKE_CASE__ : List[str]="<::::>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Tuple = {} if sp_model_kwargs is None else sp_model_kwargs
# Add extra_ids to the special token list
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : int = vocab_file
__a : Union[str, Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return self.sp_model.get_piece_size()
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Dict = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __getstate__( self : Optional[Any] ):
'''simple docstring'''
__a : Union[str, Any] = self.__dict__.copy()
__a : Any = None
return state
def __setstate__( self : int , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a : str = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : str = {}
__a : List[Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(self.vocab_file )
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
return self.sp_model.piece_to_id(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : int = self.sp_model.IdToPiece(SCREAMING_SNAKE_CASE__ )
return token
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Optional[Any] = []
__a : Optional[int] = ''
for token in tokens:
# make sure that special tokens are not decoded using sentencepiece model
if token in self.all_special_tokens:
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ ) + token
__a : Dict = []
else:
current_sub_tokens.append(SCREAMING_SNAKE_CASE__ )
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ )
return out_string.strip()
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Tuple = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[str] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
import torch
from diffusers import DDIMParallelScheduler
from .test_schedulers import SchedulerCommonTest
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = (DDIMParallelScheduler,)
__SCREAMING_SNAKE_CASE : Union[str, Any] = (('''eta''', 0.0), ('''num_inference_steps''', 50))
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : List[Any] = {
'num_train_timesteps': 1_0_0_0,
'beta_start': 0.0_001,
'beta_end': 0.02,
'beta_schedule': 'linear',
'clip_sample': True,
}
config.update(**SCREAMING_SNAKE_CASE__ )
return config
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Tuple = self.scheduler_classes[0]
__a : Optional[Any] = self.get_scheduler_config(**SCREAMING_SNAKE_CASE__ )
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : List[str] = 1_0, 0.0
__a : Dict = self.dummy_model()
__a : str = self.dummy_sample_deter
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
for t in scheduler.timesteps:
__a : str = model(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[str] = scheduler.step(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ).prev_sample
return sample
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
for timesteps in [1_0_0, 5_0_0, 1_0_0_0]:
self.check_over_configs(num_train_timesteps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
for steps_offset in [0, 1]:
self.check_over_configs(steps_offset=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config(steps_offset=1 )
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
scheduler.set_timesteps(5 )
assert torch.equal(scheduler.timesteps , torch.LongTensor([8_0_1, 6_0_1, 4_0_1, 2_0_1, 1] ) )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for beta_start, beta_end in zip([0.0_001, 0.001, 0.01, 0.1] , [0.002, 0.02, 0.2, 2] ):
self.check_over_configs(beta_start=SCREAMING_SNAKE_CASE__ , beta_end=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for schedule in ["linear", "squaredcos_cap_v2"]:
self.check_over_configs(beta_schedule=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(prediction_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for clip_sample in [True, False]:
self.check_over_configs(clip_sample=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for timestep_spacing in ["trailing", "leading"]:
self.check_over_configs(timestep_spacing=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for rescale_betas_zero_snr in [True, False]:
self.check_over_configs(rescale_betas_zero_snr=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
self.check_over_configs(thresholding=SCREAMING_SNAKE_CASE__ )
for threshold in [0.5, 1.0, 2.0]:
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(
thresholding=SCREAMING_SNAKE_CASE__ , prediction_type=SCREAMING_SNAKE_CASE__ , sample_max_value=SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for t in [1, 1_0, 4_9]:
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
for t, num_inference_steps in zip([1, 1_0, 5_0] , [1_0, 5_0, 5_0_0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , num_inference_steps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for t, eta in zip([1, 1_0, 4_9] , [0.0, 0.5, 1.0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , eta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : Union[str, Any] = self.get_scheduler_config()
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_2_0 , 4_0_0 ) - 0.14_771 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_8_0 , 9_6_0 ) - 0.32_460 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_8_7 , 4_8_6 ) - 0.00_979 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_9_9 , 9_9_8 ) - 0.02 ) ) < 1e-5
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config()
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : Any = 1_0, 0.0
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = self.dummy_model()
__a : int = self.dummy_sample_deter
__a : List[Any] = self.dummy_sample_deter + 0.1
__a : List[str] = self.dummy_sample_deter - 0.1
__a : Optional[Any] = samplea.shape[0]
__a : Optional[Any] = torch.stack([samplea, samplea, samplea] , dim=0 )
__a : Union[str, Any] = torch.arange(SCREAMING_SNAKE_CASE__ )[0:3, None].repeat(1 , SCREAMING_SNAKE_CASE__ )
__a : int = model(samples.flatten(0 , 1 ) , timesteps.flatten(0 , 1 ) )
__a : int = scheduler.batch_step_no_noise(SCREAMING_SNAKE_CASE__ , timesteps.flatten(0 , 1 ) , samples.flatten(0 , 1 ) , SCREAMING_SNAKE_CASE__ )
__a : Dict = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 1_147.7_904 ) < 1e-2
assert abs(result_mean.item() - 0.4_982 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : List[str] = self.full_loop()
__a : Tuple = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : int = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 172.0_067 ) < 1e-2
assert abs(result_mean.item() - 0.223_967 ) < 1e-3
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Optional[int] = self.full_loop(prediction_type='v_prediction' )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Union[str, Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 52.5_302 ) < 1e-2
assert abs(result_mean.item() - 0.0_684 ) < 1e-3
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Union[str, Any] = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : Optional[int] = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[int] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.8_295 ) < 1e-2
assert abs(result_mean.item() - 0.1_951 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Dict = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Tuple = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.0_784 ) < 1e-2
assert abs(result_mean.item() - 0.1_941 ) < 1e-3
| 47 |
from ..utils import DummyObject, requires_backends
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Dict , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Tuple , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Any = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Tuple , *SCREAMING_SNAKE_CASE__ : List[str] , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Dict , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Dict = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[int] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Optional[int] , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
| 47 | 1 |
import glob
import os
import random
from string import ascii_lowercase, digits
import cva
import numpy as np
# Parrameters
SCREAMING_SNAKE_CASE__ = (720, 1280) # Height, Width
SCREAMING_SNAKE_CASE__ = (0.4, 0.6) # if height or width lower than this scale, drop it.
SCREAMING_SNAKE_CASE__ = 1 / 100
SCREAMING_SNAKE_CASE__ = ''''''
SCREAMING_SNAKE_CASE__ = ''''''
SCREAMING_SNAKE_CASE__ = ''''''
SCREAMING_SNAKE_CASE__ = 250
def UpperCAmelCase__ ( ):
__a , __a : Optional[int] = get_dataset(lowerCamelCase_ , lowerCamelCase_ )
for index in range(lowerCamelCase_ ):
__a : Optional[Any] = random.sample(range(len(lowerCamelCase_ ) ) , 4 )
__a , __a , __a : int = update_image_and_anno(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , filter_scale=lowerCamelCase_ , )
# Get random string code: '7b7ad245cdff75241935e4dd860f3bad'
__a : Optional[Any] = random_chars(3_2 )
__a : Tuple = path.split(os.sep )[-1].rsplit('.' , 1 )[0]
__a : Union[str, Any] = f'''{OUTPUT_DIR}/{file_name}_MOSAIC_{letter_code}'''
cva.imwrite(f'''{file_root}.jpg''' , lowerCamelCase_ , [cva.IMWRITE_JPEG_QUALITY, 8_5] )
print(f'''Succeeded {index+1}/{NUMBER_IMAGES} with {file_name}''' )
__a : List[Any] = []
for anno in new_annos:
__a : Optional[int] = anno[3] - anno[1]
__a : str = anno[4] - anno[2]
__a : Tuple = anno[1] + width / 2
__a : Optional[int] = anno[2] + height / 2
__a : Tuple = f'''{anno[0]} {x_center} {y_center} {width} {height}'''
annos_list.append(lowerCamelCase_ )
with open(f'''{file_root}.txt''' , 'w' ) as outfile:
outfile.write('\n'.join(line for line in annos_list ) )
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
__a : List[Any] = []
__a : int = []
for label_file in glob.glob(os.path.join(lowerCamelCase_ , '*.txt' ) ):
__a : Union[str, Any] = label_file.split(os.sep )[-1].rsplit('.' , 1 )[0]
with open(lowerCamelCase_ ) as in_file:
__a : Tuple = in_file.readlines()
__a : Optional[int] = os.path.join(lowerCamelCase_ , f'''{label_name}.jpg''' )
__a : Dict = []
for obj_list in obj_lists:
__a : List[Any] = obj_list.rstrip('\n' ).split(' ' )
__a : Any = float(obj[1] ) - float(obj[3] ) / 2
__a : Optional[int] = float(obj[2] ) - float(obj[4] ) / 2
__a : int = float(obj[1] ) + float(obj[3] ) / 2
__a : Any = float(obj[2] ) + float(obj[4] ) / 2
boxes.append([int(obj[0] ), xmin, ymin, xmax, ymax] )
if not boxes:
continue
img_paths.append(lowerCamelCase_ )
labels.append(lowerCamelCase_ )
return img_paths, labels
def UpperCAmelCase__ ( lowerCamelCase_ : list , lowerCamelCase_ : list , lowerCamelCase_ : list[int] , lowerCamelCase_ : tuple[int, int] , lowerCamelCase_ : tuple[float, float] , lowerCamelCase_ : float = 0.0 , ):
__a : Optional[int] = np.zeros([output_size[0], output_size[1], 3] , dtype=np.uinta )
__a : Union[str, Any] = scale_range[0] + random.random() * (scale_range[1] - scale_range[0])
__a : List[str] = scale_range[0] + random.random() * (scale_range[1] - scale_range[0])
__a : List[str] = int(scale_x * output_size[1] )
__a : Dict = int(scale_y * output_size[0] )
__a : Optional[int] = []
__a : Optional[Any] = []
for i, index in enumerate(lowerCamelCase_ ):
__a : str = all_img_list[index]
path_list.append(lowerCamelCase_ )
__a : int = all_annos[index]
__a : List[Any] = cva.imread(lowerCamelCase_ )
if i == 0: # top-left
__a : List[Any] = cva.resize(lowerCamelCase_ , (divid_point_x, divid_point_y) )
__a : Dict = img
for bbox in img_annos:
__a : int = bbox[1] * scale_x
__a : Any = bbox[2] * scale_y
__a : int = bbox[3] * scale_x
__a : str = bbox[4] * scale_y
new_anno.append([bbox[0], xmin, ymin, xmax, ymax] )
elif i == 1: # top-right
__a : Tuple = cva.resize(lowerCamelCase_ , (output_size[1] - divid_point_x, divid_point_y) )
__a : Dict = img
for bbox in img_annos:
__a : List[str] = scale_x + bbox[1] * (1 - scale_x)
__a : Optional[Any] = bbox[2] * scale_y
__a : Union[str, Any] = scale_x + bbox[3] * (1 - scale_x)
__a : Dict = bbox[4] * scale_y
new_anno.append([bbox[0], xmin, ymin, xmax, ymax] )
elif i == 2: # bottom-left
__a : List[Any] = cva.resize(lowerCamelCase_ , (divid_point_x, output_size[0] - divid_point_y) )
__a : Optional[Any] = img
for bbox in img_annos:
__a : Optional[int] = bbox[1] * scale_x
__a : Optional[Any] = scale_y + bbox[2] * (1 - scale_y)
__a : Union[str, Any] = bbox[3] * scale_x
__a : List[str] = scale_y + bbox[4] * (1 - scale_y)
new_anno.append([bbox[0], xmin, ymin, xmax, ymax] )
else: # bottom-right
__a : str = cva.resize(
lowerCamelCase_ , (output_size[1] - divid_point_x, output_size[0] - divid_point_y) )
__a : Tuple = img
for bbox in img_annos:
__a : Dict = scale_x + bbox[1] * (1 - scale_x)
__a : str = scale_y + bbox[2] * (1 - scale_y)
__a : Union[str, Any] = scale_x + bbox[3] * (1 - scale_x)
__a : List[str] = scale_y + bbox[4] * (1 - scale_y)
new_anno.append([bbox[0], xmin, ymin, xmax, ymax] )
# Remove bounding box small than scale of filter
if filter_scale > 0:
__a : Optional[Any] = [
anno
for anno in new_anno
if filter_scale < (anno[3] - anno[1]) and filter_scale < (anno[4] - anno[2])
]
return output_img, new_anno, path_list[0]
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
assert number_char > 1, "The number of character should greater than 1"
__a : Dict = ascii_lowercase + digits
return "".join(random.choice(lowerCamelCase_ ) for _ in range(lowerCamelCase_ ) )
if __name__ == "__main__":
main()
print('''DONE ✅''')
| 47 |
import math
from datetime import datetime, timedelta
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
__a : Union[str, Any] = year % 1_9
__a : int = year % 4
__a : Optional[int] = year % 7
__a : Dict = math.floor(year / 1_0_0 )
__a : Optional[Any] = math.floor((1_3 + 8 * leap_day_inhibits) / 2_5 )
__a : Union[str, Any] = leap_day_inhibits / 4
__a : str = (
1_5 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 3_0
__a : Union[str, Any] = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
__a : List[Any] = (1_9 * metonic_cycle + secular_moon_shift) % 3_0
# PHM -> Paschal Full Moon
__a : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 2_9 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_9 )
elif days_to_add == 2_8 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_8 )
else:
return datetime(lowerCamelCase_ , 3 , 2_2 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
SCREAMING_SNAKE_CASE__ = '''will be''' if year > datetime.now().year else '''was'''
print(F"Easter in {year} {tense} {gauss_easter(year)}")
| 47 | 1 |
import unittest
import numpy as np
from transformers.testing_utils import require_torch, require_vision
from transformers.utils import is_torch_available, is_vision_available
from ...test_image_processing_common import ImageProcessingSavingTestMixin, prepare_image_inputs
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
from transformers import LevitImageProcessor
class _UpperCamelCase( unittest.TestCase ):
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : str=7 , SCREAMING_SNAKE_CASE__ : Dict=3 , SCREAMING_SNAKE_CASE__ : List[str]=1_8 , SCREAMING_SNAKE_CASE__ : Optional[int]=3_0 , SCREAMING_SNAKE_CASE__ : Dict=4_0_0 , SCREAMING_SNAKE_CASE__ : List[Any]=True , SCREAMING_SNAKE_CASE__ : List[Any]=None , SCREAMING_SNAKE_CASE__ : Union[str, Any]=True , SCREAMING_SNAKE_CASE__ : Union[str, Any]=None , SCREAMING_SNAKE_CASE__ : str=True , SCREAMING_SNAKE_CASE__ : int=[0.5, 0.5, 0.5] , SCREAMING_SNAKE_CASE__ : str=[0.5, 0.5, 0.5] , ):
'''simple docstring'''
__a : Optional[Any] = size if size is not None else {'shortest_edge': 1_8}
__a : List[str] = crop_size if crop_size is not None else {'height': 1_8, 'width': 1_8}
__a : int = parent
__a : Optional[int] = batch_size
__a : Optional[int] = num_channels
__a : Dict = image_size
__a : List[Any] = min_resolution
__a : Optional[int] = max_resolution
__a : Optional[Any] = do_resize
__a : Union[str, Any] = size
__a : Optional[Any] = do_center_crop
__a : Optional[Any] = crop_size
__a : Union[str, Any] = do_normalize
__a : Optional[int] = image_mean
__a : Tuple = image_std
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return {
"image_mean": self.image_mean,
"image_std": self.image_std,
"do_normalize": self.do_normalize,
"do_resize": self.do_resize,
"do_center_crop": self.do_center_crop,
"size": self.size,
"crop_size": self.crop_size,
}
@require_torch
@require_vision
class _UpperCamelCase( __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : List[Any] = LevitImageProcessor if is_vision_available() else None
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : int = LevitImageProcessingTester(self )
@property
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return self.image_processor_tester.prepare_image_processor_dict()
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Union[str, Any] = self.image_processing_class(**self.image_processor_dict )
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , 'image_mean' ) )
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , 'image_std' ) )
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , 'do_normalize' ) )
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , 'do_resize' ) )
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , 'do_center_crop' ) )
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , 'size' ) )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : Optional[Any] = self.image_processing_class.from_dict(self.image_processor_dict )
self.assertEqual(image_processor.size , {'shortest_edge': 1_8} )
self.assertEqual(image_processor.crop_size , {'height': 1_8, 'width': 1_8} )
__a : int = self.image_processing_class.from_dict(self.image_processor_dict , size=4_2 , crop_size=8_4 )
self.assertEqual(image_processor.size , {'shortest_edge': 4_2} )
self.assertEqual(image_processor.crop_size , {'height': 8_4, 'width': 8_4} )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
pass
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : str = self.image_processing_class(**self.image_processor_dict )
# create random PIL images
__a : List[Any] = prepare_image_inputs(self.image_processor_tester , equal_resolution=SCREAMING_SNAKE_CASE__ )
for image in image_inputs:
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , Image.Image )
# Test not batched input
__a : Any = image_processing(image_inputs[0] , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_processor_tester.num_channels,
self.image_processor_tester.crop_size['height'],
self.image_processor_tester.crop_size['width'],
) , )
# Test batched
__a : Any = image_processing(SCREAMING_SNAKE_CASE__ , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_processor_tester.batch_size,
self.image_processor_tester.num_channels,
self.image_processor_tester.crop_size['height'],
self.image_processor_tester.crop_size['width'],
) , )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Tuple = self.image_processing_class(**self.image_processor_dict )
# create random numpy tensors
__a : Optional[Any] = prepare_image_inputs(self.image_processor_tester , equal_resolution=SCREAMING_SNAKE_CASE__ , numpify=SCREAMING_SNAKE_CASE__ )
for image in image_inputs:
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , np.ndarray )
# Test not batched input
__a : Tuple = image_processing(image_inputs[0] , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_processor_tester.num_channels,
self.image_processor_tester.crop_size['height'],
self.image_processor_tester.crop_size['width'],
) , )
# Test batched
__a : List[str] = image_processing(SCREAMING_SNAKE_CASE__ , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_processor_tester.batch_size,
self.image_processor_tester.num_channels,
self.image_processor_tester.crop_size['height'],
self.image_processor_tester.crop_size['width'],
) , )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[Any] = self.image_processing_class(**self.image_processor_dict )
# create random PyTorch tensors
__a : Tuple = prepare_image_inputs(self.image_processor_tester , equal_resolution=SCREAMING_SNAKE_CASE__ , torchify=SCREAMING_SNAKE_CASE__ )
for image in image_inputs:
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , torch.Tensor )
# Test not batched input
__a : Dict = image_processing(image_inputs[0] , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_processor_tester.num_channels,
self.image_processor_tester.crop_size['height'],
self.image_processor_tester.crop_size['width'],
) , )
# Test batched
__a : Dict = image_processing(SCREAMING_SNAKE_CASE__ , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_processor_tester.batch_size,
self.image_processor_tester.num_channels,
self.image_processor_tester.crop_size['height'],
self.image_processor_tester.crop_size['width'],
) , )
| 47 |
from typing import List, Optional, Union
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''huggingface/informer-tourism-monthly''': (
'''https://huggingface.co/huggingface/informer-tourism-monthly/resolve/main/config.json'''
),
# See all Informer models at https://huggingface.co/models?filter=informer
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = '''informer'''
__SCREAMING_SNAKE_CASE : List[Any] = {
'''hidden_size''': '''d_model''',
'''num_attention_heads''': '''encoder_attention_heads''',
'''num_hidden_layers''': '''encoder_layers''',
}
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : str = "student_t" , SCREAMING_SNAKE_CASE__ : str = "nll" , SCREAMING_SNAKE_CASE__ : int = 1 , SCREAMING_SNAKE_CASE__ : List[int] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[str, bool]] = "mean" , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : int = 6_4 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : str = "gelu" , SCREAMING_SNAKE_CASE__ : float = 0.05 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : int = 1_0_0 , SCREAMING_SNAKE_CASE__ : float = 0.02 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=True , SCREAMING_SNAKE_CASE__ : str = "prob" , SCREAMING_SNAKE_CASE__ : int = 5 , SCREAMING_SNAKE_CASE__ : bool = True , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Dict = prediction_length
__a : Tuple = context_length or prediction_length
__a : Tuple = distribution_output
__a : Tuple = loss
__a : str = input_size
__a : Dict = num_time_features
__a : Optional[int] = lags_sequence if lags_sequence is not None else [1, 2, 3, 4, 5, 6, 7]
__a : str = scaling
__a : Tuple = num_dynamic_real_features
__a : int = num_static_real_features
__a : Dict = num_static_categorical_features
# set cardinality
if cardinality and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The cardinality should be a list of the same length as `num_static_categorical_features`' )
__a : Optional[Any] = cardinality
else:
__a : Optional[int] = [0]
# set embedding_dimension
if embedding_dimension and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The embedding dimension should be a list of the same length as `num_static_categorical_features`' )
__a : int = embedding_dimension
else:
__a : List[Any] = [min(5_0 , (cat + 1) // 2 ) for cat in self.cardinality]
__a : int = num_parallel_samples
# Transformer architecture configuration
__a : str = input_size * len(self.lags_sequence ) + self._number_of_features
__a : Optional[int] = d_model
__a : Union[str, Any] = encoder_attention_heads
__a : int = decoder_attention_heads
__a : Any = encoder_ffn_dim
__a : Union[str, Any] = decoder_ffn_dim
__a : List[Any] = encoder_layers
__a : Optional[int] = decoder_layers
__a : int = dropout
__a : Optional[Any] = attention_dropout
__a : Dict = activation_dropout
__a : Union[str, Any] = encoder_layerdrop
__a : Optional[int] = decoder_layerdrop
__a : List[str] = activation_function
__a : str = init_std
__a : Optional[int] = use_cache
# Informer
__a : Union[str, Any] = attention_type
__a : str = sampling_factor
__a : Dict = distil
super().__init__(is_encoder_decoder=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
return (
sum(self.embedding_dimension )
+ self.num_dynamic_real_features
+ self.num_time_features
+ self.num_static_real_features
+ self.input_size * 2 # the log1p(abs(loc)) and log(scale) features
)
| 47 | 1 |
from typing import TYPE_CHECKING
from ..utils import _LazyModule
SCREAMING_SNAKE_CASE__ = {
'''config''': [
'''EXTERNAL_DATA_FORMAT_SIZE_LIMIT''',
'''OnnxConfig''',
'''OnnxConfigWithPast''',
'''OnnxSeq2SeqConfigWithPast''',
'''PatchingSpec''',
],
'''convert''': ['''export''', '''validate_model_outputs'''],
'''features''': ['''FeaturesManager'''],
'''utils''': ['''ParameterFormat''', '''compute_serialized_parameters_size'''],
}
if TYPE_CHECKING:
from .config import (
EXTERNAL_DATA_FORMAT_SIZE_LIMIT,
OnnxConfig,
OnnxConfigWithPast,
OnnxSeqaSeqConfigWithPast,
PatchingSpec,
)
from .convert import export, validate_model_outputs
from .features import FeaturesManager
from .utils import ParameterFormat, compute_serialized_parameters_size
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
import torch
from diffusers import DDIMParallelScheduler
from .test_schedulers import SchedulerCommonTest
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = (DDIMParallelScheduler,)
__SCREAMING_SNAKE_CASE : Union[str, Any] = (('''eta''', 0.0), ('''num_inference_steps''', 50))
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : List[Any] = {
'num_train_timesteps': 1_0_0_0,
'beta_start': 0.0_001,
'beta_end': 0.02,
'beta_schedule': 'linear',
'clip_sample': True,
}
config.update(**SCREAMING_SNAKE_CASE__ )
return config
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Tuple = self.scheduler_classes[0]
__a : Optional[Any] = self.get_scheduler_config(**SCREAMING_SNAKE_CASE__ )
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : List[str] = 1_0, 0.0
__a : Dict = self.dummy_model()
__a : str = self.dummy_sample_deter
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
for t in scheduler.timesteps:
__a : str = model(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[str] = scheduler.step(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ).prev_sample
return sample
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
for timesteps in [1_0_0, 5_0_0, 1_0_0_0]:
self.check_over_configs(num_train_timesteps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
for steps_offset in [0, 1]:
self.check_over_configs(steps_offset=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config(steps_offset=1 )
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
scheduler.set_timesteps(5 )
assert torch.equal(scheduler.timesteps , torch.LongTensor([8_0_1, 6_0_1, 4_0_1, 2_0_1, 1] ) )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for beta_start, beta_end in zip([0.0_001, 0.001, 0.01, 0.1] , [0.002, 0.02, 0.2, 2] ):
self.check_over_configs(beta_start=SCREAMING_SNAKE_CASE__ , beta_end=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for schedule in ["linear", "squaredcos_cap_v2"]:
self.check_over_configs(beta_schedule=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(prediction_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for clip_sample in [True, False]:
self.check_over_configs(clip_sample=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for timestep_spacing in ["trailing", "leading"]:
self.check_over_configs(timestep_spacing=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for rescale_betas_zero_snr in [True, False]:
self.check_over_configs(rescale_betas_zero_snr=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
self.check_over_configs(thresholding=SCREAMING_SNAKE_CASE__ )
for threshold in [0.5, 1.0, 2.0]:
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(
thresholding=SCREAMING_SNAKE_CASE__ , prediction_type=SCREAMING_SNAKE_CASE__ , sample_max_value=SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for t in [1, 1_0, 4_9]:
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
for t, num_inference_steps in zip([1, 1_0, 5_0] , [1_0, 5_0, 5_0_0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , num_inference_steps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for t, eta in zip([1, 1_0, 4_9] , [0.0, 0.5, 1.0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , eta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : Union[str, Any] = self.get_scheduler_config()
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_2_0 , 4_0_0 ) - 0.14_771 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_8_0 , 9_6_0 ) - 0.32_460 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_8_7 , 4_8_6 ) - 0.00_979 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_9_9 , 9_9_8 ) - 0.02 ) ) < 1e-5
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config()
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : Any = 1_0, 0.0
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = self.dummy_model()
__a : int = self.dummy_sample_deter
__a : List[Any] = self.dummy_sample_deter + 0.1
__a : List[str] = self.dummy_sample_deter - 0.1
__a : Optional[Any] = samplea.shape[0]
__a : Optional[Any] = torch.stack([samplea, samplea, samplea] , dim=0 )
__a : Union[str, Any] = torch.arange(SCREAMING_SNAKE_CASE__ )[0:3, None].repeat(1 , SCREAMING_SNAKE_CASE__ )
__a : int = model(samples.flatten(0 , 1 ) , timesteps.flatten(0 , 1 ) )
__a : int = scheduler.batch_step_no_noise(SCREAMING_SNAKE_CASE__ , timesteps.flatten(0 , 1 ) , samples.flatten(0 , 1 ) , SCREAMING_SNAKE_CASE__ )
__a : Dict = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 1_147.7_904 ) < 1e-2
assert abs(result_mean.item() - 0.4_982 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : List[str] = self.full_loop()
__a : Tuple = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : int = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 172.0_067 ) < 1e-2
assert abs(result_mean.item() - 0.223_967 ) < 1e-3
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Optional[int] = self.full_loop(prediction_type='v_prediction' )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Union[str, Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 52.5_302 ) < 1e-2
assert abs(result_mean.item() - 0.0_684 ) < 1e-3
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Union[str, Any] = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : Optional[int] = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[int] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.8_295 ) < 1e-2
assert abs(result_mean.item() - 0.1_951 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Dict = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Tuple = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.0_784 ) < 1e-2
assert abs(result_mean.item() - 0.1_941 ) < 1e-3
| 47 | 1 |
# Copyright 2023 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
SCREAMING_SNAKE_CASE__ = {'''configuration_timm_backbone''': ['''TimmBackboneConfig''']}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''TimmBackbone''']
if TYPE_CHECKING:
from .configuration_timm_backbone import TimmBackboneConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_timm_backbone import TimmBackbone
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : list[int] ):
# Check if the input is valid
if not len(lowerCamelCase_ ) == len(lowerCamelCase_ ) == 3:
raise ValueError('Please enter a valid equation.' )
if equationa[0] == equationa[1] == equationa[0] == equationa[1] == 0:
raise ValueError('Both a & b of two equations can\'t be zero.' )
# Extract the coefficients
__a , __a , __a : Optional[Any] = equationa
__a , __a , __a : Optional[int] = equationa
# Calculate the determinants of the matrices
__a : str = aa * ba - aa * ba
__a : Tuple = ca * ba - ca * ba
__a : Union[str, Any] = aa * ca - aa * ca
# Check if the system of linear equations has a solution (using Cramer's rule)
if determinant == 0:
if determinant_x == determinant_y == 0:
raise ValueError('Infinite solutions. (Consistent system)' )
else:
raise ValueError('No solution. (Inconsistent system)' )
else:
if determinant_x == determinant_y == 0:
# Trivial solution (Inconsistent system)
return (0.0, 0.0)
else:
__a : Any = determinant_x / determinant
__a : Optional[Any] = determinant_y / determinant
# Non-Trivial Solution (Consistent system)
return (x, y)
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_conditional_detr''': [
'''CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''ConditionalDetrConfig''',
'''ConditionalDetrOnnxConfig''',
]
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''ConditionalDetrFeatureExtractor''']
SCREAMING_SNAKE_CASE__ = ['''ConditionalDetrImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''ConditionalDetrForObjectDetection''',
'''ConditionalDetrForSegmentation''',
'''ConditionalDetrModel''',
'''ConditionalDetrPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_conditional_detr import (
CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP,
ConditionalDetrConfig,
ConditionalDetrOnnxConfig,
)
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .feature_extraction_conditional_detr import ConditionalDetrFeatureExtractor
from .image_processing_conditional_detr import ConditionalDetrImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_conditional_detr import (
CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST,
ConditionalDetrForObjectDetection,
ConditionalDetrForSegmentation,
ConditionalDetrModel,
ConditionalDetrPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
from ...utils import is_note_seq_available, is_transformers_available, is_torch_available
from ...utils import OptionalDependencyNotAvailable
try:
if not (is_transformers_available() and is_torch_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_torch_and_transformers_objects import * # noqa F403
else:
from .notes_encoder import SpectrogramNotesEncoder
from .continous_encoder import SpectrogramContEncoder
from .pipeline_spectrogram_diffusion import (
SpectrogramContEncoder,
SpectrogramDiffusionPipeline,
TaFilmDecoder,
)
try:
if not (is_transformers_available() and is_torch_available() and is_note_seq_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_transformers_and_torch_and_note_seq_objects import * # noqa F403
else:
from .midi_utils import MidiProcessor
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : int ):
while a != 0:
__a , __a : Dict = b % a, a
return b
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : int ):
if gcd(lowerCamelCase_ , lowerCamelCase_ ) != 1:
__a : int = f'''mod inverse of {a!r} and {m!r} does not exist'''
raise ValueError(lowerCamelCase_ )
__a , __a , __a : Dict = 1, 0, a
__a , __a , __a : List[str] = 0, 1, m
while va != 0:
__a : Union[str, Any] = ua // va
__a , __a , __a , __a , __a , __a : List[str] = (ua - q * va), (ua - q * va), (ua - q * va), va, va, va
return ua % m
| 47 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_bridgetower''': [
'''BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''BridgeTowerConfig''',
'''BridgeTowerTextConfig''',
'''BridgeTowerVisionConfig''',
],
'''processing_bridgetower''': ['''BridgeTowerProcessor'''],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''BridgeTowerImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''BridgeTowerForContrastiveLearning''',
'''BridgeTowerForImageAndTextRetrieval''',
'''BridgeTowerForMaskedLM''',
'''BridgeTowerModel''',
'''BridgeTowerPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_bridgetower import (
BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP,
BridgeTowerConfig,
BridgeTowerTextConfig,
BridgeTowerVisionConfig,
)
from .processing_bridgetower import BridgeTowerProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_bridgetower import BridgeTowerImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_bridgetower import (
BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST,
BridgeTowerForContrastiveLearning,
BridgeTowerForImageAndTextRetrieval,
BridgeTowerForMaskedLM,
BridgeTowerModel,
BridgeTowerPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure)
| 47 | 1 |
import unittest
import numpy as np
from diffusers import LMSDiscreteScheduler, OnnxStableDiffusionInpaintPipeline
from diffusers.utils.testing_utils import (
is_onnx_available,
load_image,
nightly,
require_onnxruntime,
require_torch_gpu,
)
from ..test_pipelines_onnx_common import OnnxPipelineTesterMixin
if is_onnx_available():
import onnxruntime as ort
class _UpperCamelCase( __lowerCamelCase , unittest.TestCase ):
# FIXME: add fast tests
pass
@nightly
@require_onnxruntime
@require_torch_gpu
class _UpperCamelCase( unittest.TestCase ):
@property
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return (
"CUDAExecutionProvider",
{
"gpu_mem_limit": "15000000000", # 15GB
"arena_extend_strategy": "kSameAsRequested",
},
)
@property
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : int = ort.SessionOptions()
__a : str = False
return options
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : str = load_image(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/in_paint/overture-creations-5sI6fQgYIuo.png' )
__a : int = load_image(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/in_paint/overture-creations-5sI6fQgYIuo_mask.png' )
__a : str = OnnxStableDiffusionInpaintPipeline.from_pretrained(
'runwayml/stable-diffusion-inpainting' , revision='onnx' , safety_checker=SCREAMING_SNAKE_CASE__ , feature_extractor=SCREAMING_SNAKE_CASE__ , provider=self.gpu_provider , sess_options=self.gpu_options , )
pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
__a : int = 'A red cat sitting on a park bench'
__a : Union[str, Any] = np.random.RandomState(0 )
__a : Any = pipe(
prompt=SCREAMING_SNAKE_CASE__ , image=SCREAMING_SNAKE_CASE__ , mask_image=SCREAMING_SNAKE_CASE__ , guidance_scale=7.5 , num_inference_steps=1_0 , generator=SCREAMING_SNAKE_CASE__ , output_type='np' , )
__a : Dict = output.images
__a : List[str] = images[0, 2_5_5:2_5_8, 2_5_5:2_5_8, -1]
assert images.shape == (1, 5_1_2, 5_1_2, 3)
__a : Dict = np.array([0.2_514, 0.3_007, 0.3_517, 0.1_790, 0.2_382, 0.3_167, 0.1_944, 0.2_273, 0.2_464] )
assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-3
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : List[Any] = load_image(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/in_paint/overture-creations-5sI6fQgYIuo.png' )
__a : List[str] = load_image(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/in_paint/overture-creations-5sI6fQgYIuo_mask.png' )
__a : Optional[int] = LMSDiscreteScheduler.from_pretrained(
'runwayml/stable-diffusion-inpainting' , subfolder='scheduler' , revision='onnx' )
__a : Union[str, Any] = OnnxStableDiffusionInpaintPipeline.from_pretrained(
'runwayml/stable-diffusion-inpainting' , revision='onnx' , scheduler=SCREAMING_SNAKE_CASE__ , safety_checker=SCREAMING_SNAKE_CASE__ , feature_extractor=SCREAMING_SNAKE_CASE__ , provider=self.gpu_provider , sess_options=self.gpu_options , )
pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = 'A red cat sitting on a park bench'
__a : Optional[int] = np.random.RandomState(0 )
__a : List[str] = pipe(
prompt=SCREAMING_SNAKE_CASE__ , image=SCREAMING_SNAKE_CASE__ , mask_image=SCREAMING_SNAKE_CASE__ , guidance_scale=7.5 , num_inference_steps=2_0 , generator=SCREAMING_SNAKE_CASE__ , output_type='np' , )
__a : Tuple = output.images
__a : Union[str, Any] = images[0, 2_5_5:2_5_8, 2_5_5:2_5_8, -1]
assert images.shape == (1, 5_1_2, 5_1_2, 3)
__a : Optional[Any] = np.array([0.0_086, 0.0_077, 0.0_083, 0.0_093, 0.0_107, 0.0_139, 0.0_094, 0.0_097, 0.0_125] )
assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-3
| 47 |
from string import ascii_lowercase, ascii_uppercase
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
if not sentence:
return ""
__a : Union[str, Any] = dict(zip(lowerCamelCase_ , lowerCamelCase_ ) )
return lower_to_upper.get(sentence[0] , sentence[0] ) + sentence[1:]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 47 | 1 |
import re
import jax.numpy as jnp
from flax.traverse_util import flatten_dict, unflatten_dict
from jax.random import PRNGKey
from ..utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
__a : Tuple = R'\w+[.]\d+'
__a : Tuple = re.findall(lowerCamelCase_ , lowerCamelCase_ )
for pat in pats:
__a : List[Any] = key.replace(lowerCamelCase_ , '_'.join(pat.split('.' ) ) )
return key
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Optional[int] , lowerCamelCase_ : str ):
__a : Optional[int] = pt_tuple_key[:-1] + ('scale',)
if (
any('norm' in str_ for str_ in pt_tuple_key )
and (pt_tuple_key[-1] == "bias")
and (pt_tuple_key[:-1] + ("bias",) not in random_flax_state_dict)
and (pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict)
):
__a : Tuple = pt_tuple_key[:-1] + ('scale',)
return renamed_pt_tuple_key, pt_tensor
elif pt_tuple_key[-1] in ["weight", "gamma"] and pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict:
__a : str = pt_tuple_key[:-1] + ('scale',)
return renamed_pt_tuple_key, pt_tensor
# embedding
if pt_tuple_key[-1] == "weight" and pt_tuple_key[:-1] + ("embedding",) in random_flax_state_dict:
__a : Optional[Any] = pt_tuple_key[:-1] + ('embedding',)
return renamed_pt_tuple_key, pt_tensor
# conv layer
__a : Any = pt_tuple_key[:-1] + ('kernel',)
if pt_tuple_key[-1] == "weight" and pt_tensor.ndim == 4:
__a : Any = pt_tensor.transpose(2 , 3 , 1 , 0 )
return renamed_pt_tuple_key, pt_tensor
# linear layer
__a : Optional[int] = pt_tuple_key[:-1] + ('kernel',)
if pt_tuple_key[-1] == "weight":
__a : Any = pt_tensor.T
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm weight
__a : Union[str, Any] = pt_tuple_key[:-1] + ('weight',)
if pt_tuple_key[-1] == "gamma":
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm bias
__a : Dict = pt_tuple_key[:-1] + ('bias',)
if pt_tuple_key[-1] == "beta":
return renamed_pt_tuple_key, pt_tensor
return pt_tuple_key, pt_tensor
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[Any]=4_2 ):
# Step 1: Convert pytorch tensor to numpy
__a : Any = {k: v.numpy() for k, v in pt_state_dict.items()}
# Step 2: Since the model is stateless, get random Flax params
__a : Dict = flax_model.init_weights(PRNGKey(lowerCamelCase_ ) )
__a : List[str] = flatten_dict(lowerCamelCase_ )
__a : Optional[Any] = {}
# Need to change some parameters name to match Flax names
for pt_key, pt_tensor in pt_state_dict.items():
__a : Any = rename_key(lowerCamelCase_ )
__a : List[Any] = tuple(renamed_pt_key.split('.' ) )
# Correctly rename weight parameters
__a , __a : Union[str, Any] = rename_key_and_reshape_tensor(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if flax_key in random_flax_state_dict:
if flax_tensor.shape != random_flax_state_dict[flax_key].shape:
raise ValueError(
f'''PyTorch checkpoint seems to be incorrect. Weight {pt_key} was expected to be of shape '''
f'''{random_flax_state_dict[flax_key].shape}, but is {flax_tensor.shape}.''' )
# also add unexpected weight so that warning is thrown
__a : List[Any] = jnp.asarray(lowerCamelCase_ )
return unflatten_dict(lowerCamelCase_ )
| 47 |
import functools
import operator
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''asapp/sew-d-tiny-100k''': '''https://huggingface.co/asapp/sew-d-tiny-100k/resolve/main/config.json''',
# See all SEW-D models at https://huggingface.co/models?filter=sew-d
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = '''sew-d'''
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : Dict=3_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=7_6_8 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : str=3_0_7_2 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : List[str]=2_5_6 , SCREAMING_SNAKE_CASE__ : Optional[Any]=True , SCREAMING_SNAKE_CASE__ : Tuple=True , SCREAMING_SNAKE_CASE__ : List[str]=("p2c", "c2p") , SCREAMING_SNAKE_CASE__ : str="layer_norm" , SCREAMING_SNAKE_CASE__ : Tuple="gelu_python" , SCREAMING_SNAKE_CASE__ : Tuple=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.0 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.02 , SCREAMING_SNAKE_CASE__ : int=1e-7 , SCREAMING_SNAKE_CASE__ : Any=1e-5 , SCREAMING_SNAKE_CASE__ : Optional[int]="group" , SCREAMING_SNAKE_CASE__ : Optional[Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=(6_4, 1_2_8, 1_2_8, 1_2_8, 1_2_8, 2_5_6, 2_5_6, 2_5_6, 2_5_6, 5_1_2, 5_1_2, 5_1_2, 5_1_2) , SCREAMING_SNAKE_CASE__ : List[Any]=(5, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : str=(1_0, 3, 1, 3, 1, 3, 1, 3, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : Optional[int]=False , SCREAMING_SNAKE_CASE__ : Optional[int]=1_2_8 , SCREAMING_SNAKE_CASE__ : Tuple=1_6 , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[Any]=0.05 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , SCREAMING_SNAKE_CASE__ : int=0.0 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : Optional[int]=0 , SCREAMING_SNAKE_CASE__ : Optional[int]="mean" , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : List[str]=False , SCREAMING_SNAKE_CASE__ : str=2_5_6 , SCREAMING_SNAKE_CASE__ : str=0 , SCREAMING_SNAKE_CASE__ : List[Any]=1 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ , pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = hidden_size
__a : Optional[Any] = feat_extract_norm
__a : List[str] = feat_extract_activation
__a : Dict = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = list(SCREAMING_SNAKE_CASE__ )
__a : List[str] = list(SCREAMING_SNAKE_CASE__ )
__a : int = conv_bias
__a : Tuple = num_conv_pos_embeddings
__a : List[str] = num_conv_pos_embedding_groups
__a : Optional[Any] = len(self.conv_dim )
__a : Union[str, Any] = num_hidden_layers
__a : Optional[Any] = intermediate_size
__a : Union[str, Any] = squeeze_factor
__a : List[Any] = max_position_embeddings
__a : Tuple = position_buckets
__a : Optional[int] = share_att_key
__a : List[str] = relative_attention
__a : Any = norm_rel_ebd
__a : Any = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = hidden_act
__a : str = num_attention_heads
__a : Union[str, Any] = hidden_dropout
__a : Optional[int] = attention_dropout
__a : List[str] = activation_dropout
__a : int = feat_proj_dropout
__a : int = final_dropout
__a : Dict = layer_norm_eps
__a : Tuple = feature_layer_norm_eps
__a : str = initializer_range
__a : Tuple = vocab_size
if (
(len(self.conv_stride ) != self.num_feat_extract_layers)
or (len(self.conv_kernel ) != self.num_feat_extract_layers)
or (len(self.conv_dim ) != self.num_feat_extract_layers)
):
raise ValueError(
'Configuration for convolutional layers is incorrect.'
'It is required that `len(config.conv_dim)` == `len(config.conv_stride)` == `len(config.conv_kernel)`,'
f'''but is `len(config.conv_dim) = {len(self.conv_dim )}`, `len(config.conv_stride)'''
f'''= {len(self.conv_stride )}`, `len(config.conv_kernel) = {len(self.conv_kernel )}`.''' )
# fine-tuning config parameters for SpecAugment: https://arxiv.org/abs/1904.08779
__a : Tuple = apply_spec_augment
__a : Optional[Any] = mask_time_prob
__a : Any = mask_time_length
__a : List[str] = mask_time_min_masks
__a : List[str] = mask_feature_prob
__a : Tuple = mask_feature_length
__a : Any = mask_feature_min_masks
# ctc loss
__a : Optional[int] = ctc_loss_reduction
__a : List[Any] = ctc_zero_infinity
# sequence classification
__a : Dict = use_weighted_layer_sum
__a : Optional[Any] = classifier_proj_size
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return functools.reduce(operator.mul , self.conv_stride , 1 )
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_tokenizers_available, is_torch_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_mvp''': ['''MVP_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''MvpConfig''', '''MvpOnnxConfig'''],
'''tokenization_mvp''': ['''MvpTokenizer'''],
}
try:
if not is_tokenizers_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''MvpTokenizerFast''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''MVP_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''MvpForCausalLM''',
'''MvpForConditionalGeneration''',
'''MvpForQuestionAnswering''',
'''MvpForSequenceClassification''',
'''MvpModel''',
'''MvpPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_mvp import MVP_PRETRAINED_CONFIG_ARCHIVE_MAP, MvpConfig, MvpOnnxConfig
from .tokenization_mvp import MvpTokenizer
try:
if not is_tokenizers_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_mvp_fast import MvpTokenizerFast
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_mvp import (
MVP_PRETRAINED_MODEL_ARCHIVE_LIST,
MvpForCausalLM,
MvpForConditionalGeneration,
MvpForQuestionAnswering,
MvpForSequenceClassification,
MvpModel,
MvpPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
from __future__ import annotations
from sys import maxsize
from typing import Generic, TypeVar
SCREAMING_SNAKE_CASE__ = TypeVar('''T''')
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (position - 1) // 2
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 1
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 2
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[str] ):
'''simple docstring'''
__a : list[tuple[T, int]] = []
__a : dict[T, int] = {}
__a : int = 0
def __len__( self : Any ):
'''simple docstring'''
return self.elements
def __repr__( self : Any ):
'''simple docstring'''
return str(self.heap )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.elements == 0
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.heap.append((elem, weight) )
__a : List[Any] = self.elements
self.elements += 1
self._bubble_up(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self.elements > 1:
self._swap_nodes(0 , self.elements - 1 )
__a , __a : Union[str, Any] = self.heap.pop()
del self.position_map[elem]
self.elements -= 1
if self.elements > 0:
__a , __a : Dict = self.heap[0]
self._bubble_down(SCREAMING_SNAKE_CASE__ )
return elem
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
__a : str = (elem, weight)
if position > 0:
__a : Tuple = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = self.heap[parent_position]
if parent_weight > weight:
self._bubble_up(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
if curr_pos == 0:
return None
__a : List[str] = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : str = self.heap[curr_pos]
__a , __a : Optional[int] = self.heap[parent_position]
if parent_weight > weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_up(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : int = self.position_map[elem]
__a , __a : Optional[Any] = self.heap[curr_pos]
__a : Tuple = get_child_left_position(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = get_child_right_position(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements and child_right_position < self.elements:
__a , __a : str = self.heap[child_left_position]
__a , __a : List[str] = self.heap[child_right_position]
if child_right_weight < child_left_weight and child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements:
__a , __a : Any = self.heap[child_left_position]
if child_left_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
return None
if child_right_position < self.elements:
__a , __a : Union[str, Any] = self.heap[child_right_position]
if child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Optional[Any] = self.heap[nodea_pos][0]
__a : str = self.heap[nodea_pos][0]
__a , __a : int = (
self.heap[nodea_pos],
self.heap[nodea_pos],
)
__a : str = nodea_pos
__a : Optional[int] = nodea_pos
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[Any] ):
'''simple docstring'''
__a : dict[T, dict[T, int]] = {}
__a : int = 0
def __repr__( self : Tuple ):
'''simple docstring'''
return str(self.connections )
def __len__( self : Dict ):
'''simple docstring'''
return self.nodes
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
if node not in self.connections:
__a : Tuple = {}
self.nodes += 1
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.add_node(SCREAMING_SNAKE_CASE__ )
self.add_node(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = weight
__a : Any = weight
def UpperCAmelCase__ ( lowerCamelCase_ : GraphUndirectedWeighted[T] , ):
__a : dict[T, int] = {node: maxsize for node in graph.connections}
__a : dict[T, T | None] = {node: None for node in graph.connections}
__a : MinPriorityQueue[T] = MinPriorityQueue()
for node, weight in dist.items():
priority_queue.push(lowerCamelCase_ , lowerCamelCase_ )
if priority_queue.is_empty():
return dist, parent
# initialization
__a : Optional[int] = priority_queue.extract_min()
__a : int = 0
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : str = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Optional[int] = node
# running prim's algorithm
while not priority_queue.is_empty():
__a : Any = priority_queue.extract_min()
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : Tuple = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Dict = node
return dist, parent
| 47 | 1 |
import argparse
import intel_extension_for_pytorch as ipex
import torch
from diffusers import DPMSolverMultistepScheduler, StableDiffusionPipeline
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser('''Stable Diffusion script with intel optimization''', add_help=False)
parser.add_argument('''--dpm''', action='''store_true''', help='''Enable DPMSolver or not''')
parser.add_argument('''--steps''', default=None, type=int, help='''Num inference steps''')
SCREAMING_SNAKE_CASE__ = parser.parse_args()
SCREAMING_SNAKE_CASE__ = '''cpu'''
SCREAMING_SNAKE_CASE__ = '''a lovely <dicoo> in red dress and hat, in the snowly and brightly night, with many brighly buildings'''
SCREAMING_SNAKE_CASE__ = '''path-to-your-trained-model'''
SCREAMING_SNAKE_CASE__ = StableDiffusionPipeline.from_pretrained(model_id)
if args.dpm:
SCREAMING_SNAKE_CASE__ = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
SCREAMING_SNAKE_CASE__ = pipe.to(device)
# to channels last
SCREAMING_SNAKE_CASE__ = pipe.unet.to(memory_format=torch.channels_last)
SCREAMING_SNAKE_CASE__ = pipe.vae.to(memory_format=torch.channels_last)
SCREAMING_SNAKE_CASE__ = pipe.text_encoder.to(memory_format=torch.channels_last)
if pipe.requires_safety_checker:
SCREAMING_SNAKE_CASE__ = pipe.safety_checker.to(memory_format=torch.channels_last)
# optimize with ipex
SCREAMING_SNAKE_CASE__ = torch.randn(2, 4, 64, 64)
SCREAMING_SNAKE_CASE__ = torch.rand(1) * 999
SCREAMING_SNAKE_CASE__ = torch.randn(2, 77, 768)
SCREAMING_SNAKE_CASE__ = (sample, timestep, encoder_hidden_status)
try:
SCREAMING_SNAKE_CASE__ = ipex.optimize(pipe.unet.eval(), dtype=torch.bfloataa, inplace=True, sample_input=input_example)
except Exception:
SCREAMING_SNAKE_CASE__ = ipex.optimize(pipe.unet.eval(), dtype=torch.bfloataa, inplace=True)
SCREAMING_SNAKE_CASE__ = ipex.optimize(pipe.vae.eval(), dtype=torch.bfloataa, inplace=True)
SCREAMING_SNAKE_CASE__ = ipex.optimize(pipe.text_encoder.eval(), dtype=torch.bfloataa, inplace=True)
if pipe.requires_safety_checker:
SCREAMING_SNAKE_CASE__ = ipex.optimize(pipe.safety_checker.eval(), dtype=torch.bfloataa, inplace=True)
# compute
SCREAMING_SNAKE_CASE__ = 666
SCREAMING_SNAKE_CASE__ = torch.Generator(device).manual_seed(seed)
SCREAMING_SNAKE_CASE__ = {'''generator''': generator}
if args.steps is not None:
SCREAMING_SNAKE_CASE__ = args.steps
with torch.cpu.amp.autocast(enabled=True, dtype=torch.bfloataa):
SCREAMING_SNAKE_CASE__ = pipe(prompt, **generate_kwargs).images[0]
# save image
image.save('''generated.png''')
| 47 |
from collections.abc import Sequence
from queue import Queue
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : int=None , SCREAMING_SNAKE_CASE__ : Tuple=None ):
'''simple docstring'''
__a : Tuple = start
__a : Dict = end
__a : List[str] = val
__a : List[Any] = (start + end) // 2
__a : Optional[Any] = left
__a : List[str] = right
def __repr__( self : Dict ):
'''simple docstring'''
return f'''SegmentTreeNode(start={self.start}, end={self.end}, val={self.val})'''
class _UpperCamelCase:
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : Sequence , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Tuple = collection
__a : Dict = function
if self.collection:
__a : int = self._build_tree(0 , len(SCREAMING_SNAKE_CASE__ ) - 1 )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self._update_tree(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
return self._query_range(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if start == end:
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.collection[start] )
__a : Tuple = (start + end) // 2
__a : Optional[int] = self._build_tree(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = self._build_tree(mid + 1 , SCREAMING_SNAKE_CASE__ )
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.fn(left.val , right.val ) , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if node.start == i and node.end == i:
__a : Optional[Any] = val
return
if i <= node.mid:
self._update_tree(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
self._update_tree(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = self.fn(node.left.val , node.right.val )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
if node.start == i and node.end == j:
return node.val
if i <= node.mid:
if j <= node.mid:
# range in left child tree
return self._query_range(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
# range in left child tree and right child tree
return self.fn(
self._query_range(node.left , SCREAMING_SNAKE_CASE__ , node.mid ) , self._query_range(node.right , node.mid + 1 , SCREAMING_SNAKE_CASE__ ) , )
else:
# range in right child tree
return self._query_range(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
if self.root is not None:
__a : Tuple = Queue()
queue.put(self.root )
while not queue.empty():
__a : Tuple = queue.get()
yield node
if node.left is not None:
queue.put(node.left )
if node.right is not None:
queue.put(node.right )
if __name__ == "__main__":
import operator
for fn in [operator.add, max, min]:
print('''*''' * 50)
SCREAMING_SNAKE_CASE__ = SegmentTree([2, 1, 5, 3, 4], fn)
for node in arr.traverse():
print(node)
print()
arr.update(1, 5)
for node in arr.traverse():
print(node)
print()
print(arr.query_range(3, 4)) # 7
print(arr.query_range(2, 2)) # 5
print(arr.query_range(1, 3)) # 13
print()
| 47 | 1 |
# Lint as: python3
import os
import re
import urllib.parse
from pathlib import Path
from typing import Callable, List, Optional, Union
from zipfile import ZipFile
from ..utils.file_utils import cached_path, hf_github_url
from ..utils.logging import get_logger
from ..utils.version import Version
SCREAMING_SNAKE_CASE__ = get_logger(__name__)
class _UpperCamelCase:
__SCREAMING_SNAKE_CASE : List[str] = '''dummy_data'''
__SCREAMING_SNAKE_CASE : List[Any] = '''datasets'''
__SCREAMING_SNAKE_CASE : Tuple = False
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Union[Version, str] , SCREAMING_SNAKE_CASE__ : Optional[str] = None , SCREAMING_SNAKE_CASE__ : bool = False , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : Optional[List[Callable]] = None , ):
'''simple docstring'''
__a : Optional[int] = 0
__a : List[Any] = dataset_name
__a : List[str] = cache_dir
__a : Any = use_local_dummy_data
__a : Optional[int] = config
# download_callbacks take a single url as input
__a : List[Callable] = download_callbacks or []
# if False, it doesn't load existing files and it returns the paths of the dummy files relative
# to the dummy_data zip file root
__a : List[Any] = load_existing_dummy_data
# TODO(PVP, QL) might need to make this more general
__a : int = str(SCREAMING_SNAKE_CASE__ )
# to be downloaded
__a : Any = None
__a : Union[str, Any] = None
@property
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self._dummy_file is None:
__a : Optional[Any] = self.download_dummy_data()
return self._dummy_file
@property
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self.config is not None:
# structure is dummy / config_name / version_name
return os.path.join('dummy' , self.config.name , self.version_name )
# structure is dummy / version_name
return os.path.join('dummy' , self.version_name )
@property
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
return os.path.join(self.dummy_data_folder , 'dummy_data.zip' )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : Dict = (
self.local_path_to_dummy_data if self.use_local_dummy_data is True else self.github_path_to_dummy_data
)
__a : Dict = cached_path(
SCREAMING_SNAKE_CASE__ , cache_dir=self.cache_dir , extract_compressed_file=SCREAMING_SNAKE_CASE__ , force_extract=SCREAMING_SNAKE_CASE__ )
return os.path.join(SCREAMING_SNAKE_CASE__ , self.dummy_file_name )
@property
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return os.path.join(self.datasets_scripts_dir , self.dataset_name , self.dummy_zip_file )
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
if self._bucket_url is None:
__a : int = hf_github_url(self.dataset_name , self.dummy_zip_file.replace(os.sep , '/' ) )
return self._bucket_url
@property
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
if os.path.isdir(self.dummy_file ):
return self.dummy_file
# else cut off path to file -> example `xsum`.
return "/".join(self.dummy_file.replace(os.sep , '/' ).split('/' )[:-1] )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Tuple , *SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if self.load_existing_dummy_data:
# dummy data is downloaded and tested
__a : int = self.dummy_file
else:
# dummy data cannot be downloaded and only the path to dummy file is returned
__a : List[str] = self.dummy_file_name
# special case when data_url is a dict
if isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
return self.create_dummy_data_dict(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
elif isinstance(SCREAMING_SNAKE_CASE__ , (list, tuple) ):
return self.create_dummy_data_list(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
return self.create_dummy_data_single(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : int , *SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
return self.download_and_extract(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
return self.download_and_extract(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Union[str, Any] , *SCREAMING_SNAKE_CASE__ : str , **SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
return path
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
return {}
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : str = {}
for key, single_urls in data_url.items():
for download_callback in self.download_callbacks:
if isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
for single_url in single_urls:
download_callback(SCREAMING_SNAKE_CASE__ )
else:
__a : Union[str, Any] = single_urls
download_callback(SCREAMING_SNAKE_CASE__ )
# we force the name of each key to be the last file / folder name of the url path
# if the url has arguments, we need to encode them with urllib.parse.quote_plus
if isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
__a : Any = [os.path.join(SCREAMING_SNAKE_CASE__ , urllib.parse.quote_plus(Path(SCREAMING_SNAKE_CASE__ ).name ) ) for x in single_urls]
else:
__a : Tuple = single_urls
__a : Optional[Any] = os.path.join(SCREAMING_SNAKE_CASE__ , urllib.parse.quote_plus(Path(SCREAMING_SNAKE_CASE__ ).name ) )
__a : List[str] = value
# make sure that values are unique
if all(isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) for i in dummy_data_dict.values() ) and len(set(dummy_data_dict.values() ) ) < len(
dummy_data_dict.values() ):
# append key to value to make its name unique
__a : int = {key: value + key for key, value in dummy_data_dict.items()}
return dummy_data_dict
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
__a : Optional[int] = []
# trick: if there are many shards named like `data.txt-000001-of-00300`, only use the first one
__a : Any = all(bool(re.findall('[0-9]{3,}-of-[0-9]{3,}' , SCREAMING_SNAKE_CASE__ ) ) for url in data_url )
__a : Tuple = all(
url.startswith('https://ftp.ncbi.nlm.nih.gov/pubmed/baseline/pubmed' ) for url in data_url )
if data_url and (is_tf_records or is_pubmed_records):
__a : List[str] = [data_url[0]] * len(SCREAMING_SNAKE_CASE__ )
for single_url in data_url:
for download_callback in self.download_callbacks:
download_callback(SCREAMING_SNAKE_CASE__ )
# we force the name of each key to be the last file / folder name of the url path
# if the url has arguments, we need to encode them with urllib.parse.quote_plus
__a : str = os.path.join(SCREAMING_SNAKE_CASE__ , urllib.parse.quote_plus(single_url.split('/' )[-1] ) )
dummy_data_list.append(SCREAMING_SNAKE_CASE__ )
return dummy_data_list
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
for download_callback in self.download_callbacks:
download_callback(SCREAMING_SNAKE_CASE__ )
# we force the name of each key to be the last file / folder name of the url path
# if the url has arguments, we need to encode them with urllib.parse.quote_plus
__a : Any = os.path.join(SCREAMING_SNAKE_CASE__ , urllib.parse.quote_plus(data_url.split('/' )[-1] ) )
if os.path.exists(SCREAMING_SNAKE_CASE__ ) or not self.load_existing_dummy_data:
return value
else:
# Backward compatibility, maybe deprecate at one point.
# For many datasets with single url calls to dl_manager.download_and_extract,
# the dummy_data.zip file is actually the zipped downloaded file
# while now we expected the dummy_data.zip file to be a directory containing
# the downloaded file.
return path_to_dummy_data
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
pass
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
pass
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
def _iter_archive_members(SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
# this preserves the order of the members inside the ZIP archive
__a : Optional[int] = Path(self.dummy_file ).parent
__a : str = path.relative_to(SCREAMING_SNAKE_CASE__ )
with ZipFile(self.local_path_to_dummy_data ) as zip_file:
__a : List[str] = zip_file.namelist()
for member in members:
if member.startswith(relative_path.as_posix() ):
yield dummy_parent_path.joinpath(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = Path(SCREAMING_SNAKE_CASE__ )
__a : Any = _iter_archive_members(SCREAMING_SNAKE_CASE__ ) if self.use_local_dummy_data else path.rglob('*' )
for file_path in file_paths:
if file_path.is_file() and not file_path.name.startswith(('.', '__') ):
yield file_path.relative_to(SCREAMING_SNAKE_CASE__ ).as_posix(), file_path.open('rb' )
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
if not isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
__a : List[Any] = [paths]
for path in paths:
if os.path.isfile(SCREAMING_SNAKE_CASE__ ):
if os.path.basename(SCREAMING_SNAKE_CASE__ ).startswith(('.', '__') ):
return
yield path
else:
for dirpath, dirnames, filenames in os.walk(SCREAMING_SNAKE_CASE__ ):
if os.path.basename(SCREAMING_SNAKE_CASE__ ).startswith(('.', '__') ):
continue
dirnames.sort()
for filename in sorted(SCREAMING_SNAKE_CASE__ ):
if filename.startswith(('.', '__') ):
continue
yield os.path.join(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
| 47 |
import os
import posixpath
import uuid
from dataclasses import dataclass
from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, Union
import numpy as np
import pyarrow as pa
import datasets
from datasets.arrow_writer import ArrowWriter, ParquetWriter
from datasets.config import MAX_SHARD_SIZE
from datasets.filesystems import (
is_remote_filesystem,
rename,
)
from datasets.iterable_dataset import _BaseExamplesIterable
from datasets.utils.py_utils import convert_file_size_to_int
SCREAMING_SNAKE_CASE__ = datasets.utils.logging.get_logger(__name__)
if TYPE_CHECKING:
import pyspark
@dataclass
class _UpperCamelCase( datasets.BuilderConfig ):
__SCREAMING_SNAKE_CASE : Optional[datasets.Features] = None
def UpperCAmelCase__ ( lowerCamelCase_ : "pyspark.sql.DataFrame" , lowerCamelCase_ : List[int] , ):
import pyspark
def generate_fn():
__a : List[Any] = df.select('*' , pyspark.sql.functions.spark_partition_id().alias('part_id' ) )
for partition_id in partition_order:
__a : Optional[int] = df_with_partition_id.select('*' ).where(f'''part_id = {partition_id}''' ).drop('part_id' )
__a : Optional[Any] = partition_df.collect()
__a : Union[str, Any] = 0
for row in rows:
yield f'''{partition_id}_{row_id}''', row.asDict()
row_id += 1
return generate_fn
class _UpperCamelCase( _BaseExamplesIterable ):
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : Dict=None , ):
'''simple docstring'''
__a : List[str] = df
__a : Tuple = partition_order or range(self.df.rdd.getNumPartitions() )
__a : List[Any] = _generate_iterable_examples(self.df , self.partition_order )
def __iter__( self : Tuple ):
'''simple docstring'''
yield from self.generate_examples_fn()
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : np.random.Generator ):
'''simple docstring'''
__a : Union[str, Any] = list(range(self.df.rdd.getNumPartitions() ) )
generator.shuffle(SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Union[str, Any] = self.split_shard_indices_by_worker(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return len(self.partition_order )
class _UpperCamelCase( datasets.DatasetBuilder ):
__SCREAMING_SNAKE_CASE : List[str] = SparkConfig
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : str = None , SCREAMING_SNAKE_CASE__ : str = None , **SCREAMING_SNAKE_CASE__ : Optional[int] , ):
'''simple docstring'''
import pyspark
__a : int = pyspark.sql.SparkSession.builder.getOrCreate()
__a : Optional[int] = df
__a : List[Any] = working_dir
super().__init__(
cache_dir=SCREAMING_SNAKE_CASE__ , config_name=str(self.df.semanticHash() ) , **SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
def create_cache_and_write_probe(SCREAMING_SNAKE_CASE__ : List[str] ):
# makedirs with exist_ok will recursively create the directory. It will not throw an error if directories
# already exist.
os.makedirs(self._cache_dir , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(self._cache_dir , 'fs_test' + uuid.uuida().hex )
# Opening the file in append mode will create a new file unless it already exists, in which case it will not
# change the file contents.
open(SCREAMING_SNAKE_CASE__ , 'a' )
return [probe_file]
if self._spark.conf.get('spark.master' , '' ).startswith('local' ):
return
# If the cluster is multi-node, make sure that the user provided a cache_dir and that it is on an NFS
# accessible to the driver.
# TODO: Stream batches to the driver using ArrowCollectSerializer instead of throwing an error.
if self._cache_dir:
__a : List[Any] = (
self._spark.sparkContext.parallelize(range(1 ) , 1 ).mapPartitions(SCREAMING_SNAKE_CASE__ ).collect()
)
if os.path.isfile(probe[0] ):
return
raise ValueError(
'When using Dataset.from_spark on a multi-node cluster, the driver and all workers should be able to access cache_dir' )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
return datasets.DatasetInfo(features=self.config.features )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : datasets.download.download_manager.DownloadManager ):
'''simple docstring'''
return [datasets.SplitGenerator(name=datasets.Split.TRAIN )]
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
import pyspark
def get_arrow_batch_size(SCREAMING_SNAKE_CASE__ : int ):
for batch in it:
yield pa.RecordBatch.from_pydict({'batch_bytes': [batch.nbytes]} )
__a : List[str] = self.df.count()
__a : Dict = df_num_rows if df_num_rows <= 1_0_0 else 1_0_0
# Approximate the size of each row (in Arrow format) by averaging over a max-100-row sample.
__a : List[str] = (
self.df.limit(SCREAMING_SNAKE_CASE__ )
.repartition(1 )
.mapInArrow(SCREAMING_SNAKE_CASE__ , 'batch_bytes: long' )
.agg(pyspark.sql.functions.sum('batch_bytes' ).alias('sample_bytes' ) )
.collect()[0]
.sample_bytes
/ sample_num_rows
)
__a : Dict = approx_bytes_per_row * df_num_rows
if approx_total_size > max_shard_size:
# Make sure there is at least one row per partition.
__a : Union[str, Any] = min(SCREAMING_SNAKE_CASE__ , int(approx_total_size / max_shard_size ) )
__a : int = self.df.repartition(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : int , ):
'''simple docstring'''
import pyspark
__a : Any = ParquetWriter if file_format == 'parquet' else ArrowWriter
__a : Union[str, Any] = os.path.join(self._working_dir , os.path.basename(SCREAMING_SNAKE_CASE__ ) ) if self._working_dir else fpath
__a : Optional[int] = file_format == 'parquet'
# Define these so that we don't reference self in write_arrow, which will result in a pickling error due to
# pickling the SparkContext.
__a : List[str] = self.config.features
__a : int = self._writer_batch_size
__a : Union[str, Any] = self._fs.storage_options
def write_arrow(SCREAMING_SNAKE_CASE__ : Optional[int] ):
# Within the same SparkContext, no two task attempts will share the same attempt ID.
__a : Any = pyspark.TaskContext().taskAttemptId()
__a : str = next(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
if first_batch is None:
# Some partitions might not receive any data.
return pa.RecordBatch.from_arrays(
[[task_id], [0], [0]] , names=['task_id', 'num_examples', 'num_bytes'] , )
__a : Any = 0
__a : List[str] = writer_class(
features=SCREAMING_SNAKE_CASE__ , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Optional[Any] = pa.Table.from_batches([first_batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
for batch in it:
if max_shard_size is not None and writer._num_bytes >= max_shard_size:
__a , __a : Optional[int] = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
shard_id += 1
__a : Optional[Any] = writer_class(
features=writer._features , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Union[str, Any] = pa.Table.from_batches([batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
if writer._num_bytes > 0:
__a , __a : str = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
if working_fpath != fpath:
for file in os.listdir(os.path.dirname(SCREAMING_SNAKE_CASE__ ) ):
__a : Any = os.path.join(os.path.dirname(SCREAMING_SNAKE_CASE__ ) , os.path.basename(SCREAMING_SNAKE_CASE__ ) )
shutil.move(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Dict = (
self.df.mapInArrow(SCREAMING_SNAKE_CASE__ , 'task_id: long, num_examples: long, num_bytes: long' )
.groupBy('task_id' )
.agg(
pyspark.sql.functions.sum('num_examples' ).alias('total_num_examples' ) , pyspark.sql.functions.sum('num_bytes' ).alias('total_num_bytes' ) , pyspark.sql.functions.count('num_bytes' ).alias('num_shards' ) , pyspark.sql.functions.collect_list('num_examples' ).alias('shard_lengths' ) , )
.collect()
)
for row in stats:
yield row.task_id, (row.total_num_examples, row.total_num_bytes, row.num_shards, row.shard_lengths)
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , SCREAMING_SNAKE_CASE__ : str = "arrow" , SCREAMING_SNAKE_CASE__ : Optional[Union[str, int]] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
self._validate_cache_dir()
__a : List[str] = convert_file_size_to_int(max_shard_size or MAX_SHARD_SIZE )
self._repartition_df_if_needed(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = not is_remote_filesystem(self._fs )
__a : Optional[Any] = os.path.join if is_local else posixpath.join
__a : Any = '-TTTTT-SSSSS-of-NNNNN'
__a : Union[str, Any] = f'''{self.name}-{split_generator.name}{SUFFIX}.{file_format}'''
__a : Any = path_join(self._output_dir , SCREAMING_SNAKE_CASE__ )
__a : Any = 0
__a : Dict = 0
__a : int = 0
__a : List[str] = []
__a : Optional[int] = []
for task_id, content in self._prepare_split_single(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : Optional[int] = content
if num_bytes > 0:
total_num_examples += num_examples
total_num_bytes += num_bytes
total_shards += num_shards
task_id_and_num_shards.append((task_id, num_shards) )
all_shard_lengths.extend(SCREAMING_SNAKE_CASE__ )
__a : List[str] = total_num_examples
__a : Optional[int] = total_num_bytes
# should rename everything at the end
logger.debug(f'''Renaming {total_shards} shards.''' )
if total_shards > 1:
__a : Any = all_shard_lengths
# Define fs outside of _rename_shard so that we don't reference self in the function, which will result in a
# pickling error due to pickling the SparkContext.
__a : Dict = self._fs
# use the -SSSSS-of-NNNNN pattern
def _rename_shard(
SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , ):
rename(
SCREAMING_SNAKE_CASE__ , fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace('TTTTT-SSSSS' , f'''{global_shard_id:05d}''' ).replace('NNNNN' , f'''{total_shards:05d}''' ) , )
__a : Union[str, Any] = []
__a : List[str] = 0
for i in range(len(SCREAMING_SNAKE_CASE__ ) ):
__a , __a : Union[str, Any] = task_id_and_num_shards[i]
for shard_id in range(SCREAMING_SNAKE_CASE__ ):
args.append([task_id, shard_id, global_shard_id] )
global_shard_id += 1
self._spark.sparkContext.parallelize(SCREAMING_SNAKE_CASE__ , len(SCREAMING_SNAKE_CASE__ ) ).map(lambda SCREAMING_SNAKE_CASE__ : _rename_shard(*SCREAMING_SNAKE_CASE__ ) ).collect()
else:
# don't use any pattern
__a : List[Any] = 0
__a : Any = task_id_and_num_shards[0][0]
self._rename(
fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace(SCREAMING_SNAKE_CASE__ , '' ) , )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , ):
'''simple docstring'''
return SparkExamplesIterable(self.df )
| 47 | 1 |
import argparse
import requests
import torch
from PIL import Image
from torchvision.transforms import Compose, Normalize, Resize, ToTensor
from transformers import SwinaSRConfig, SwinaSRForImageSuperResolution, SwinaSRImageProcessor
def UpperCAmelCase__ ( lowerCamelCase_ : Any ):
__a : int = SwinaSRConfig()
if "Swin2SR_ClassicalSR_X4_64" in checkpoint_url:
__a : Dict = 4
elif "Swin2SR_CompressedSR_X4_48" in checkpoint_url:
__a : Tuple = 4
__a : List[str] = 4_8
__a : Optional[Any] = 'pixelshuffle_aux'
elif "Swin2SR_Lightweight_X2_64" in checkpoint_url:
__a : Optional[int] = [6, 6, 6, 6]
__a : List[str] = 6_0
__a : Dict = [6, 6, 6, 6]
__a : Tuple = 'pixelshuffledirect'
elif "Swin2SR_RealworldSR_X4_64_BSRGAN_PSNR" in checkpoint_url:
__a : List[str] = 4
__a : List[str] = 'nearest+conv'
elif "Swin2SR_Jpeg_dynamic" in checkpoint_url:
__a : Tuple = 1
__a : List[str] = 1
__a : Dict = 1_2_6
__a : Optional[int] = 7
__a : Dict = 255.0
__a : Tuple = ''
return config
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : List[Any] ):
if "patch_embed.proj" in name and "layers" not in name:
__a : str = name.replace('patch_embed.proj' , 'embeddings.patch_embeddings.projection' )
if "patch_embed.norm" in name:
__a : str = name.replace('patch_embed.norm' , 'embeddings.patch_embeddings.layernorm' )
if "layers" in name:
__a : Union[str, Any] = name.replace('layers' , 'encoder.stages' )
if "residual_group.blocks" in name:
__a : int = name.replace('residual_group.blocks' , 'layers' )
if "attn.proj" in name:
__a : str = name.replace('attn.proj' , 'attention.output.dense' )
if "attn" in name:
__a : List[str] = name.replace('attn' , 'attention.self' )
if "norm1" in name:
__a : Tuple = name.replace('norm1' , 'layernorm_before' )
if "norm2" in name:
__a : Dict = name.replace('norm2' , 'layernorm_after' )
if "mlp.fc1" in name:
__a : Any = name.replace('mlp.fc1' , 'intermediate.dense' )
if "mlp.fc2" in name:
__a : Optional[int] = name.replace('mlp.fc2' , 'output.dense' )
if "q_bias" in name:
__a : Union[str, Any] = name.replace('q_bias' , 'query.bias' )
if "k_bias" in name:
__a : Union[str, Any] = name.replace('k_bias' , 'key.bias' )
if "v_bias" in name:
__a : List[str] = name.replace('v_bias' , 'value.bias' )
if "cpb_mlp" in name:
__a : Optional[Any] = name.replace('cpb_mlp' , 'continuous_position_bias_mlp' )
if "patch_embed.proj" in name:
__a : Optional[int] = name.replace('patch_embed.proj' , 'patch_embed.projection' )
if name == "norm.weight":
__a : List[Any] = 'layernorm.weight'
if name == "norm.bias":
__a : Tuple = 'layernorm.bias'
if "conv_first" in name:
__a : Optional[Any] = name.replace('conv_first' , 'first_convolution' )
if (
"upsample" in name
or "conv_before_upsample" in name
or "conv_bicubic" in name
or "conv_up" in name
or "conv_hr" in name
or "conv_last" in name
or "aux" in name
):
# heads
if "conv_last" in name:
__a : List[str] = name.replace('conv_last' , 'final_convolution' )
if config.upsampler in ["pixelshuffle", "pixelshuffle_aux", "nearest+conv"]:
if "conv_before_upsample.0" in name:
__a : List[str] = name.replace('conv_before_upsample.0' , 'conv_before_upsample' )
if "upsample.0" in name:
__a : Tuple = name.replace('upsample.0' , 'upsample.convolution_0' )
if "upsample.2" in name:
__a : Dict = name.replace('upsample.2' , 'upsample.convolution_1' )
__a : Optional[int] = 'upsample.' + name
elif config.upsampler == "pixelshuffledirect":
__a : Optional[Any] = name.replace('upsample.0.weight' , 'upsample.conv.weight' )
__a : Dict = name.replace('upsample.0.bias' , 'upsample.conv.bias' )
else:
pass
else:
__a : List[Any] = 'swin2sr.' + name
return name
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : Optional[int] ):
for key in orig_state_dict.copy().keys():
__a : Dict = orig_state_dict.pop(lowerCamelCase_ )
if "qkv" in key:
__a : Any = key.split('.' )
__a : Optional[Any] = int(key_split[1] )
__a : List[str] = int(key_split[4] )
__a : Optional[Any] = config.embed_dim
if "weight" in key:
__a : int = val[:dim, :]
__a : Optional[int] = val[dim : dim * 2, :]
__a : Tuple = val[-dim:, :]
else:
__a : Optional[Any] = val[:dim]
__a : int = val[dim : dim * 2]
__a : str = val[-dim:]
pass
else:
__a : int = val
return orig_state_dict
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Tuple ):
__a : int = get_config(lowerCamelCase_ )
__a : List[Any] = SwinaSRForImageSuperResolution(lowerCamelCase_ )
model.eval()
__a : List[Any] = torch.hub.load_state_dict_from_url(lowerCamelCase_ , map_location='cpu' )
__a : Optional[Any] = convert_state_dict(lowerCamelCase_ , lowerCamelCase_ )
__a , __a : Dict = model.load_state_dict(lowerCamelCase_ , strict=lowerCamelCase_ )
if len(lowerCamelCase_ ) > 0:
raise ValueError('Missing keys when converting: {}'.format(lowerCamelCase_ ) )
for key in unexpected_keys:
if not ("relative_position_index" in key or "relative_coords_table" in key or "self_mask" in key):
raise ValueError(f'''Unexpected key {key} in state_dict''' )
# verify values
__a : Union[str, Any] = 'https://github.com/mv-lab/swin2sr/blob/main/testsets/real-inputs/shanghai.jpg?raw=true'
__a : Any = Image.open(requests.get(lowerCamelCase_ , stream=lowerCamelCase_ ).raw ).convert('RGB' )
__a : str = SwinaSRImageProcessor()
# pixel_values = processor(image, return_tensors="pt").pixel_values
__a : Union[str, Any] = 1_2_6 if 'Jpeg' in checkpoint_url else 2_5_6
__a : Tuple = Compose(
[
Resize((image_size, image_size) ),
ToTensor(),
Normalize(mean=[0.485, 0.456, 0.406] , std=[0.229, 0.224, 0.225] ),
] )
__a : Optional[int] = transforms(lowerCamelCase_ ).unsqueeze(0 )
if config.num_channels == 1:
__a : List[Any] = pixel_values[:, 0, :, :].unsqueeze(1 )
__a : str = model(lowerCamelCase_ )
# assert values
if "Swin2SR_ClassicalSR_X2_64" in checkpoint_url:
__a : Union[str, Any] = torch.Size([1, 3, 5_1_2, 5_1_2] )
__a : List[str] = torch.tensor(
[[-0.7087, -0.7138, -0.6721], [-0.8340, -0.8095, -0.7298], [-0.9149, -0.8414, -0.7940]] )
elif "Swin2SR_ClassicalSR_X4_64" in checkpoint_url:
__a : Optional[int] = torch.Size([1, 3, 1_0_2_4, 1_0_2_4] )
__a : Union[str, Any] = torch.tensor(
[[-0.7775, -0.8105, -0.8933], [-0.7764, -0.8356, -0.9225], [-0.7976, -0.8686, -0.9579]] )
elif "Swin2SR_CompressedSR_X4_48" in checkpoint_url:
# TODO values didn't match exactly here
__a : Optional[int] = torch.Size([1, 3, 1_0_2_4, 1_0_2_4] )
__a : List[str] = torch.tensor(
[[-0.8035, -0.7504, -0.7491], [-0.8538, -0.8124, -0.7782], [-0.8804, -0.8651, -0.8493]] )
elif "Swin2SR_Lightweight_X2_64" in checkpoint_url:
__a : List[str] = torch.Size([1, 3, 5_1_2, 5_1_2] )
__a : Optional[int] = torch.tensor(
[[-0.7669, -0.8662, -0.8767], [-0.8810, -0.9962, -0.9820], [-0.9340, -1.0322, -1.1149]] )
elif "Swin2SR_RealworldSR_X4_64_BSRGAN_PSNR" in checkpoint_url:
__a : Dict = torch.Size([1, 3, 1_0_2_4, 1_0_2_4] )
__a : Optional[Any] = torch.tensor(
[[-0.5238, -0.5557, -0.6321], [-0.6016, -0.5903, -0.6391], [-0.6244, -0.6334, -0.6889]] )
assert (
outputs.reconstruction.shape == expected_shape
), f'''Shape of reconstruction should be {expected_shape}, but is {outputs.reconstruction.shape}'''
assert torch.allclose(outputs.reconstruction[0, 0, :3, :3] , lowerCamelCase_ , atol=1e-3 )
print('Looks ok!' )
__a : int = {
'https://github.com/mv-lab/swin2sr/releases/download/v0.0.1/Swin2SR_ClassicalSR_X2_64.pth': (
'swin2SR-classical-sr-x2-64'
),
'https://github.com/mv-lab/swin2sr/releases/download/v0.0.1/Swin2SR_ClassicalSR_X4_64.pth': (
'swin2SR-classical-sr-x4-64'
),
'https://github.com/mv-lab/swin2sr/releases/download/v0.0.1/Swin2SR_CompressedSR_X4_48.pth': (
'swin2SR-compressed-sr-x4-48'
),
'https://github.com/mv-lab/swin2sr/releases/download/v0.0.1/Swin2SR_Lightweight_X2_64.pth': (
'swin2SR-lightweight-x2-64'
),
'https://github.com/mv-lab/swin2sr/releases/download/v0.0.1/Swin2SR_RealworldSR_X4_64_BSRGAN_PSNR.pth': (
'swin2SR-realworld-sr-x4-64-bsrgan-psnr'
),
}
__a : int = url_to_name[checkpoint_url]
if pytorch_dump_folder_path is not None:
print(f'''Saving model {model_name} to {pytorch_dump_folder_path}''' )
model.save_pretrained(lowerCamelCase_ )
print(f'''Saving image processor to {pytorch_dump_folder_path}''' )
processor.save_pretrained(lowerCamelCase_ )
if push_to_hub:
model.push_to_hub(f'''caidas/{model_name}''' )
processor.push_to_hub(f'''caidas/{model_name}''' )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--checkpoint_url''',
default='''https://github.com/mv-lab/swin2sr/releases/download/v0.0.1/Swin2SR_ClassicalSR_X2_64.pth''',
type=str,
help='''URL of the original Swin2SR checkpoint you\'d like to convert.''',
)
parser.add_argument(
'''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.'''
)
parser.add_argument('''--push_to_hub''', action='''store_true''', help='''Whether to push the converted model to the hub.''')
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_swinasr_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path, args.push_to_hub)
| 47 |
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, TensorDataset
from tqdm import tqdm
from transformers import GPTaLMHeadModel
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : int ):
# save results
if os.path.exists(lowerCamelCase_ ):
if os.path.exists(os.path.join(lowerCamelCase_ , 'config.json' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'config.json' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'config.json' ) )
if os.path.exists(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) )
else:
os.makedirs(lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Any=False ):
__a : Dict = 2
if unlogit:
__a : Optional[Any] = torch.pow(lowerCamelCase_ , lowerCamelCase_ )
__a : Any = p * torch.log(lowerCamelCase_ )
__a : Union[str, Any] = 0
return -plogp.sum(dim=-1 )
def UpperCAmelCase__ ( lowerCamelCase_ : Any ):
logger.info('lv, h >\t' + '\t'.join(f'''{x + 1}''' for x in range(len(lowerCamelCase_ ) ) ) )
for row in range(len(lowerCamelCase_ ) ):
if tensor.dtype != torch.long:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:.5f}''' for x in tensor[row].cpu().data ) )
else:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:d}''' for x in tensor[row].cpu().data ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Any , lowerCamelCase_ : int , lowerCamelCase_ : int=True , lowerCamelCase_ : Optional[Any]=True , lowerCamelCase_ : List[Any]=None , lowerCamelCase_ : List[Any]=False ):
__a , __a : Optional[int] = model.config.num_hidden_layers, model.config.num_attention_heads
__a : str = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
__a : int = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
if head_mask is None:
__a : Union[str, Any] = torch.ones(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
head_mask.requires_grad_(requires_grad=lowerCamelCase_ )
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
__a : Any = None
__a : Optional[int] = 0.0
__a : Optional[Any] = 0.0
for step, inputs in enumerate(tqdm(lowerCamelCase_ , desc='Iteration' , disable=args.local_rank not in [-1, 0] ) ):
__a : Dict = tuple(t.to(args.device ) for t in inputs )
((__a) , ) : Dict = inputs
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
__a : List[Any] = model(lowerCamelCase_ , labels=lowerCamelCase_ , head_mask=lowerCamelCase_ )
# (loss), lm_logits, presents, (all hidden_states), (attentions)
__a , __a , __a : int = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
total_loss += loss.detach().cpu().numpy()
if compute_entropy:
for layer, attn in enumerate(lowerCamelCase_ ):
__a : List[str] = entropy(attn.detach() , lowerCamelCase_ )
attn_entropy[layer] += masked_entropy.sum(-1 ).sum(0 ).sum(0 ).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
tot_tokens += torch.ones_like(lowerCamelCase_ ).float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
__a : Optional[Any] = 2
__a : Union[str, Any] = torch.pow(torch.pow(lowerCamelCase_ , lowerCamelCase_ ).sum(-1 ) , 1 / exponent )
head_importance /= norm_by_layer.unsqueeze(-1 ) + 1e-20
if not args.dont_normalize_global_importance:
__a : List[str] = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print matrices
if compute_entropy:
logger.info('Attention entropies' )
print_ad_tensor(lowerCamelCase_ )
if compute_importance:
logger.info('Head importance scores' )
print_ad_tensor(lowerCamelCase_ )
logger.info('Head ranked by importance scores' )
__a : Optional[Any] = torch.zeros(head_importance.numel() , dtype=torch.long , device=args.device )
__a : str = torch.arange(
head_importance.numel() , device=args.device )
__a : Tuple = head_ranks.view_as(lowerCamelCase_ )
print_ad_tensor(lowerCamelCase_ )
return attn_entropy, head_importance, total_loss
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Tuple , lowerCamelCase_ : int ):
__a , __a , __a : Optional[int] = compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ )
__a : Tuple = 1 / loss # instead of downsteam score use the LM loss
logger.info('Pruning: original score: %f, threshold: %f' , lowerCamelCase_ , original_score * args.masking_threshold )
__a : Tuple = torch.ones_like(lowerCamelCase_ )
__a : int = max(1 , int(new_head_mask.numel() * args.masking_amount ) )
__a : Tuple = original_score
while current_score >= original_score * args.masking_threshold:
__a : Optional[Any] = new_head_mask.clone().detach() # save current head mask
# heads from least important to most - keep only not-masked heads
__a : List[str] = float('Inf' )
__a : List[Any] = head_importance.view(-1 ).sort()[1]
if len(lowerCamelCase_ ) <= num_to_mask:
print('BREAK BY num_to_mask' )
break
# mask heads
__a : Any = current_heads_to_mask[:num_to_mask]
logger.info('Heads to mask: %s' , str(current_heads_to_mask.tolist() ) )
__a : int = new_head_mask.view(-1 )
__a : Tuple = 0.0
__a : int = new_head_mask.view_as(lowerCamelCase_ )
__a : Optional[int] = new_head_mask.clone().detach()
print_ad_tensor(lowerCamelCase_ )
# Compute metric and head importance again
__a , __a , __a : int = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[Any] = 1 / loss
logger.info(
'Masking: current score: %f, remaining heads %d (%.1f percents)' , lowerCamelCase_ , new_head_mask.sum() , new_head_mask.sum() / new_head_mask.numel() * 1_0_0 , )
logger.info('Final head mask' )
print_ad_tensor(lowerCamelCase_ )
np.save(os.path.join(args.output_dir , 'head_mask.npy' ) , head_mask.detach().cpu().numpy() )
return head_mask
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Union[str, Any] ):
__a : List[Any] = datetime.now()
__a , __a , __a : List[str] = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[str] = 1 / loss
__a : List[Any] = datetime.now() - before_time
__a : List[str] = sum(p.numel() for p in model.parameters() )
__a : Dict = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(lowerCamelCase_ ) )
}
for k, v in heads_to_prune.items():
if isinstance(lowerCamelCase_ , lowerCamelCase_ ):
__a : Tuple = [
v,
]
assert sum(len(lowerCamelCase_ ) for h in heads_to_prune.values() ) == (1 - head_mask.long()).sum().item()
model.prune_heads(lowerCamelCase_ )
__a : Optional[Any] = sum(p.numel() for p in model.parameters() )
__a : Tuple = datetime.now()
__a , __a , __a : Tuple = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ , actually_pruned=lowerCamelCase_ , )
__a : Optional[Any] = 1 / loss
__a : List[Any] = datetime.now() - before_time
logger.info(
'Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)' , lowerCamelCase_ , lowerCamelCase_ , pruned_num_params / original_num_params * 1_0_0 , )
logger.info('Pruning: score with masking: %f score with pruning: %f' , lowerCamelCase_ , lowerCamelCase_ )
logger.info('Pruning: speed ratio (original timing / new timing): %f percents' , original_time / new_time * 1_0_0 )
save_model(lowerCamelCase_ , args.output_dir )
def UpperCAmelCase__ ( ):
__a : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'--data_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The input data dir. Should contain the .tsv files (or other data files) for the task.' , )
parser.add_argument(
'--model_name_or_path' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='Path to pretrained model or model identifier from huggingface.co/models' , )
parser.add_argument(
'--output_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
# Other parameters
parser.add_argument(
'--config_name' , default='' , type=lowerCamelCase_ , help='Pretrained config name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--tokenizer_name' , default='' , type=lowerCamelCase_ , help='Pretrained tokenizer name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--cache_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , help='Where do you want to store the pre-trained models downloaded from s3' , )
parser.add_argument(
'--data_subset' , type=lowerCamelCase_ , default=-1 , help='If > 0: limit the data to a subset of data_subset instances.' )
parser.add_argument(
'--overwrite_output_dir' , action='store_true' , help='Whether to overwrite data in output directory' )
parser.add_argument(
'--overwrite_cache' , action='store_true' , help='Overwrite the cached training and evaluation sets' )
parser.add_argument(
'--dont_normalize_importance_by_layer' , action='store_true' , help='Don\'t normalize importance score by layers' )
parser.add_argument(
'--dont_normalize_global_importance' , action='store_true' , help='Don\'t normalize all importance scores between 0 and 1' , )
parser.add_argument(
'--try_masking' , action='store_true' , help='Whether to try to mask head until a threshold of accuracy.' )
parser.add_argument(
'--masking_threshold' , default=0.9 , type=lowerCamelCase_ , help='masking threshold in term of metrics (stop masking when metric < threshold * original metric value).' , )
parser.add_argument(
'--masking_amount' , default=0.1 , type=lowerCamelCase_ , help='Amount to heads to masking at each masking step.' )
parser.add_argument('--metric_name' , default='acc' , type=lowerCamelCase_ , help='Metric to use for head masking.' )
parser.add_argument(
'--max_seq_length' , default=1_2_8 , type=lowerCamelCase_ , help=(
'The maximum total input sequence length after WordPiece tokenization. \n'
'Sequences longer than this will be truncated, sequences shorter padded.'
) , )
parser.add_argument('--batch_size' , default=1 , type=lowerCamelCase_ , help='Batch size.' )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 )
parser.add_argument('--local_rank' , type=lowerCamelCase_ , default=-1 , help='local_rank for distributed training on gpus' )
parser.add_argument('--no_cuda' , action='store_true' , help='Whether not to use CUDA when available' )
parser.add_argument('--server_ip' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
parser.add_argument('--server_port' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
__a : Optional[Any] = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print('Waiting for debugger attach' )
ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=lowerCamelCase_ )
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
__a : List[str] = torch.device('cuda' if torch.cuda.is_available() and not args.no_cuda else 'cpu' )
__a : Tuple = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank )
__a : Union[str, Any] = torch.device('cuda' , args.local_rank )
__a : Any = 1
torch.distributed.init_process_group(backend='nccl' ) # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN )
logger.info('device: {} n_gpu: {}, distributed: {}'.format(args.device , args.n_gpu , bool(args.local_rank != -1 ) ) )
__a : Optional[Any] = GPTaLMHeadModel.from_pretrained(args.model_name_or_path )
# Distributed and parallel training
model.to(args.device )
if args.local_rank != -1:
__a : List[Any] = nn.parallel.DistributedDataParallel(
lowerCamelCase_ , device_ids=[args.local_rank] , output_device=args.local_rank , find_unused_parameters=lowerCamelCase_ )
elif args.n_gpu > 1:
__a : Union[str, Any] = nn.DataParallel(lowerCamelCase_ )
# Print/save training arguments
os.makedirs(args.output_dir , exist_ok=lowerCamelCase_ )
torch.save(lowerCamelCase_ , os.path.join(args.output_dir , 'run_args.bin' ) )
logger.info('Training/evaluation parameters %s' , lowerCamelCase_ )
# Prepare dataset
__a : Tuple = np.concatenate(
[
np.loadtxt(args.data_dir , dtype=np.intaa ),
] )
__a : str = (torch.from_numpy(lowerCamelCase_ ),)
__a : List[str] = TensorDataset(*lowerCamelCase_ )
__a : Optional[Any] = RandomSampler(lowerCamelCase_ )
__a : Union[str, Any] = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.batch_size )
# Compute head entropy and importance score
compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
__a : Union[str, Any] = mask_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
prune_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 | 1 |
import math
import torch
from torch import nn
from ..configuration_utils import ConfigMixin, register_to_config
from .attention_processor import Attention
from .embeddings import get_timestep_embedding
from .modeling_utils import ModelMixin
class _UpperCamelCase( __lowerCamelCase , __lowerCamelCase ):
@register_to_config
def __init__( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : int = 1_2_8 , SCREAMING_SNAKE_CASE__ : int = 2_5_6 , SCREAMING_SNAKE_CASE__ : float = 2_000.0 , SCREAMING_SNAKE_CASE__ : int = 7_6_8 , SCREAMING_SNAKE_CASE__ : int = 1_2 , SCREAMING_SNAKE_CASE__ : int = 1_2 , SCREAMING_SNAKE_CASE__ : int = 6_4 , SCREAMING_SNAKE_CASE__ : int = 2_0_4_8 , SCREAMING_SNAKE_CASE__ : float = 0.1 , ):
'''simple docstring'''
super().__init__()
__a : List[str] = nn.Sequential(
nn.Linear(SCREAMING_SNAKE_CASE__ , d_model * 4 , bias=SCREAMING_SNAKE_CASE__ ) , nn.SiLU() , nn.Linear(d_model * 4 , d_model * 4 , bias=SCREAMING_SNAKE_CASE__ ) , nn.SiLU() , )
__a : Optional[Any] = nn.Embedding(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = False
__a : Tuple = nn.Linear(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , bias=SCREAMING_SNAKE_CASE__ )
__a : Tuple = nn.Dropout(p=SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = nn.ModuleList()
for lyr_num in range(SCREAMING_SNAKE_CASE__ ):
# FiLM conditional T5 decoder
__a : List[Any] = DecoderLayer(d_model=SCREAMING_SNAKE_CASE__ , d_kv=SCREAMING_SNAKE_CASE__ , num_heads=SCREAMING_SNAKE_CASE__ , d_ff=SCREAMING_SNAKE_CASE__ , dropout_rate=SCREAMING_SNAKE_CASE__ )
self.decoders.append(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = TaLayerNorm(SCREAMING_SNAKE_CASE__ )
__a : Any = nn.Dropout(p=SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = nn.Linear(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , bias=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Any = torch.mul(query_input.unsqueeze(-1 ) , key_input.unsqueeze(-2 ) )
return mask.unsqueeze(-3 )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a , __a , __a : Optional[int] = decoder_input_tokens.shape
assert decoder_noise_time.shape == (batch,)
# decoder_noise_time is in [0, 1), so rescale to expected timing range.
__a : Optional[int] = get_timestep_embedding(
decoder_noise_time * self.config.max_decoder_noise_time , embedding_dim=self.config.d_model , max_period=self.config.max_decoder_noise_time , ).to(dtype=self.dtype )
__a : List[Any] = self.conditioning_emb(SCREAMING_SNAKE_CASE__ ).unsqueeze(1 )
assert conditioning_emb.shape == (batch, 1, self.config.d_model * 4)
__a : Optional[Any] = decoder_input_tokens.shape[1]
# If we want to use relative positions for audio context, we can just offset
# this sequence by the length of encodings_and_masks.
__a : int = torch.broadcast_to(
torch.arange(SCREAMING_SNAKE_CASE__ , device=decoder_input_tokens.device ) , (batch, seq_length) , )
__a : Any = self.position_encoding(SCREAMING_SNAKE_CASE__ )
__a : Any = self.continuous_inputs_projection(SCREAMING_SNAKE_CASE__ )
inputs += position_encodings
__a : Union[str, Any] = self.dropout(SCREAMING_SNAKE_CASE__ )
# decoder: No padding present.
__a : Optional[int] = torch.ones(
decoder_input_tokens.shape[:2] , device=decoder_input_tokens.device , dtype=inputs.dtype )
# Translate encoding masks to encoder-decoder masks.
__a : List[Any] = [(x, self.encoder_decoder_mask(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )) for x, y in encodings_and_masks]
# cross attend style: concat encodings
__a : Optional[Any] = torch.cat([x[0] for x in encodings_and_encdec_masks] , dim=1 )
__a : List[str] = torch.cat([x[1] for x in encodings_and_encdec_masks] , dim=-1 )
for lyr in self.decoders:
__a : Optional[Any] = lyr(
SCREAMING_SNAKE_CASE__ , conditioning_emb=SCREAMING_SNAKE_CASE__ , encoder_hidden_states=SCREAMING_SNAKE_CASE__ , encoder_attention_mask=SCREAMING_SNAKE_CASE__ , )[0]
__a : Optional[int] = self.decoder_norm(SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = self.post_dropout(SCREAMING_SNAKE_CASE__ )
__a : str = self.spec_out(SCREAMING_SNAKE_CASE__ )
return spec_out
class _UpperCamelCase( nn.Module ):
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1e-6 ):
'''simple docstring'''
super().__init__()
__a : str = nn.ModuleList()
# cond self attention: layer 0
self.layer.append(
TaLayerSelfAttentionCond(d_model=SCREAMING_SNAKE_CASE__ , d_kv=SCREAMING_SNAKE_CASE__ , num_heads=SCREAMING_SNAKE_CASE__ , dropout_rate=SCREAMING_SNAKE_CASE__ ) )
# cross attention: layer 1
self.layer.append(
TaLayerCrossAttention(
d_model=SCREAMING_SNAKE_CASE__ , d_kv=SCREAMING_SNAKE_CASE__ , num_heads=SCREAMING_SNAKE_CASE__ , dropout_rate=SCREAMING_SNAKE_CASE__ , layer_norm_epsilon=SCREAMING_SNAKE_CASE__ , ) )
# Film Cond MLP + dropout: last layer
self.layer.append(
TaLayerFFCond(d_model=SCREAMING_SNAKE_CASE__ , d_ff=SCREAMING_SNAKE_CASE__ , dropout_rate=SCREAMING_SNAKE_CASE__ , layer_norm_epsilon=SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[int]=None , SCREAMING_SNAKE_CASE__ : Union[str, Any]=None , SCREAMING_SNAKE_CASE__ : Optional[Any]=None , SCREAMING_SNAKE_CASE__ : Any=None , SCREAMING_SNAKE_CASE__ : int=None , ):
'''simple docstring'''
__a : List[str] = self.layer[0](
SCREAMING_SNAKE_CASE__ , conditioning_emb=SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , )
if encoder_hidden_states is not None:
__a : List[str] = torch.where(encoder_attention_mask > 0 , 0 , -1e10 ).to(
encoder_hidden_states.dtype )
__a : str = self.layer[1](
SCREAMING_SNAKE_CASE__ , key_value_states=SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , )
# Apply Film Conditional Feed Forward layer
__a : List[str] = self.layer[-1](SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return (hidden_states,)
class _UpperCamelCase( nn.Module ):
def __init__( self : int , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
super().__init__()
__a : int = TaLayerNorm(SCREAMING_SNAKE_CASE__ )
__a : str = TaFiLMLayer(in_features=d_model * 4 , out_features=SCREAMING_SNAKE_CASE__ )
__a : int = Attention(query_dim=SCREAMING_SNAKE_CASE__ , heads=SCREAMING_SNAKE_CASE__ , dim_head=SCREAMING_SNAKE_CASE__ , out_bias=SCREAMING_SNAKE_CASE__ , scale_qk=SCREAMING_SNAKE_CASE__ )
__a : int = nn.Dropout(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Optional[Any]=None , SCREAMING_SNAKE_CASE__ : int=None , ):
'''simple docstring'''
__a : List[Any] = self.layer_norm(SCREAMING_SNAKE_CASE__ )
if conditioning_emb is not None:
__a : int = self.FiLMLayer(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# Self-attention block
__a : List[str] = self.attention(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = hidden_states + self.dropout(SCREAMING_SNAKE_CASE__ )
return hidden_states
class _UpperCamelCase( nn.Module ):
def __init__( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
super().__init__()
__a : int = Attention(query_dim=SCREAMING_SNAKE_CASE__ , heads=SCREAMING_SNAKE_CASE__ , dim_head=SCREAMING_SNAKE_CASE__ , out_bias=SCREAMING_SNAKE_CASE__ , scale_qk=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = TaLayerNorm(SCREAMING_SNAKE_CASE__ , eps=SCREAMING_SNAKE_CASE__ )
__a : Dict = nn.Dropout(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Tuple=None , SCREAMING_SNAKE_CASE__ : Tuple=None , ):
'''simple docstring'''
__a : Optional[Any] = self.layer_norm(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = self.attention(
SCREAMING_SNAKE_CASE__ , encoder_hidden_states=SCREAMING_SNAKE_CASE__ , attention_mask=attention_mask.squeeze(1 ) , )
__a : int = hidden_states + self.dropout(SCREAMING_SNAKE_CASE__ )
return layer_output
class _UpperCamelCase( nn.Module ):
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
super().__init__()
__a : str = TaDenseGatedActDense(d_model=SCREAMING_SNAKE_CASE__ , d_ff=SCREAMING_SNAKE_CASE__ , dropout_rate=SCREAMING_SNAKE_CASE__ )
__a : Tuple = TaFiLMLayer(in_features=d_model * 4 , out_features=SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = TaLayerNorm(SCREAMING_SNAKE_CASE__ , eps=SCREAMING_SNAKE_CASE__ )
__a : Dict = nn.Dropout(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[str]=None ):
'''simple docstring'''
__a : int = self.layer_norm(SCREAMING_SNAKE_CASE__ )
if conditioning_emb is not None:
__a : Optional[int] = self.film(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : str = self.DenseReluDense(SCREAMING_SNAKE_CASE__ )
__a : Dict = hidden_states + self.dropout(SCREAMING_SNAKE_CASE__ )
return hidden_states
class _UpperCamelCase( nn.Module ):
def __init__( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
super().__init__()
__a : Any = nn.Linear(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , bias=SCREAMING_SNAKE_CASE__ )
__a : List[str] = nn.Linear(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , bias=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = nn.Linear(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , bias=SCREAMING_SNAKE_CASE__ )
__a : Any = nn.Dropout(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = NewGELUActivation()
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a : List[Any] = self.act(self.wi_a(SCREAMING_SNAKE_CASE__ ) )
__a : List[str] = self.wi_a(SCREAMING_SNAKE_CASE__ )
__a : int = hidden_gelu * hidden_linear
__a : str = self.dropout(SCREAMING_SNAKE_CASE__ )
__a : List[str] = self.wo(SCREAMING_SNAKE_CASE__ )
return hidden_states
class _UpperCamelCase( nn.Module ):
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : int=1e-6 ):
'''simple docstring'''
super().__init__()
__a : int = nn.Parameter(torch.ones(SCREAMING_SNAKE_CASE__ ) )
__a : Any = eps
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : str = hidden_states.to(torch.floataa ).pow(2 ).mean(-1 , keepdim=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = hidden_states * torch.rsqrt(variance + self.variance_epsilon )
# convert into half-precision if necessary
if self.weight.dtype in [torch.floataa, torch.bfloataa]:
__a : Union[str, Any] = hidden_states.to(self.weight.dtype )
return self.weight * hidden_states
class _UpperCamelCase( nn.Module ):
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : torch.Tensor ):
'''simple docstring'''
return 0.5 * input * (1.0 + torch.tanh(math.sqrt(2.0 / math.pi ) * (input + 0.044_715 * torch.pow(SCREAMING_SNAKE_CASE__ , 3.0 )) ))
class _UpperCamelCase( nn.Module ):
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
super().__init__()
__a : Any = nn.Linear(SCREAMING_SNAKE_CASE__ , out_features * 2 , bias=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
__a : Dict = self.scale_bias(SCREAMING_SNAKE_CASE__ )
__a , __a : List[str] = torch.chunk(SCREAMING_SNAKE_CASE__ , 2 , -1 )
__a : List[Any] = x * (1 + scale) + shift
return x
| 47 |
import argparse
import os
import gluonnlp as nlp
import mxnet as mx
import numpy as np
import torch
from gluonnlp.base import get_home_dir
from gluonnlp.model.bert import BERTEncoder
from gluonnlp.model.utils import _load_vocab
from gluonnlp.vocab import Vocab
from packaging import version
from torch import nn
from transformers import BertConfig, BertForMaskedLM, BertModel, RobertaTokenizer
from transformers.models.bert.modeling_bert import (
BertIntermediate,
BertLayer,
BertOutput,
BertSelfAttention,
BertSelfOutput,
)
from transformers.utils import logging
if version.parse(nlp.__version__) != version.parse('''0.8.3'''):
raise Exception('''requires gluonnlp == 0.8.3''')
if version.parse(mx.__version__) != version.parse('''1.5.0'''):
raise Exception('''requires mxnet == 1.5.0''')
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''The Nymphenburg Palace is a beautiful palace in Munich!'''
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
__a : List[Any] = {
'attention_cell': 'multi_head',
'num_layers': 4,
'units': 1_0_2_4,
'hidden_size': 7_6_8,
'max_length': 5_1_2,
'num_heads': 8,
'scaled': True,
'dropout': 0.1,
'use_residual': True,
'embed_size': 1_0_2_4,
'embed_dropout': 0.1,
'word_embed': None,
'layer_norm_eps': 1e-5,
'token_type_vocab_size': 2,
}
__a : Optional[int] = bort_4_8_768_1024_hparams
# Let's construct the original Bort model here
# Taken from official BERT implementation, see:
# https://github.com/alexa/bort/blob/master/bort/bort.py
__a : List[str] = BERTEncoder(
attention_cell=predefined_args['attention_cell'] , num_layers=predefined_args['num_layers'] , units=predefined_args['units'] , hidden_size=predefined_args['hidden_size'] , max_length=predefined_args['max_length'] , num_heads=predefined_args['num_heads'] , scaled=predefined_args['scaled'] , dropout=predefined_args['dropout'] , output_attention=lowerCamelCase_ , output_all_encodings=lowerCamelCase_ , use_residual=predefined_args['use_residual'] , activation=predefined_args.get('activation' , 'gelu' ) , layer_norm_eps=predefined_args.get('layer_norm_eps' , lowerCamelCase_ ) , )
# Vocab information needs to be fetched first
# It's the same as RoBERTa, so RobertaTokenizer can be used later
__a : int = 'openwebtext_ccnews_stories_books_cased'
# Specify download folder to Gluonnlp's vocab
__a : Optional[Any] = os.path.join(get_home_dir() , 'models' )
__a : Optional[Any] = _load_vocab(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , cls=lowerCamelCase_ )
__a : Any = nlp.model.BERTModel(
lowerCamelCase_ , len(lowerCamelCase_ ) , units=predefined_args['units'] , embed_size=predefined_args['embed_size'] , embed_dropout=predefined_args['embed_dropout'] , word_embed=predefined_args['word_embed'] , use_pooler=lowerCamelCase_ , use_token_type_embed=lowerCamelCase_ , token_type_vocab_size=predefined_args['token_type_vocab_size'] , use_classifier=lowerCamelCase_ , use_decoder=lowerCamelCase_ , )
original_bort.load_parameters(lowerCamelCase_ , cast_dtype=lowerCamelCase_ , ignore_extra=lowerCamelCase_ )
__a : Dict = original_bort._collect_params_with_prefix()
# Build our config 🤗
__a : Optional[Any] = {
'architectures': ['BertForMaskedLM'],
'attention_probs_dropout_prob': predefined_args['dropout'],
'hidden_act': 'gelu',
'hidden_dropout_prob': predefined_args['dropout'],
'hidden_size': predefined_args['embed_size'],
'initializer_range': 0.02,
'intermediate_size': predefined_args['hidden_size'],
'layer_norm_eps': predefined_args['layer_norm_eps'],
'max_position_embeddings': predefined_args['max_length'],
'model_type': 'bort',
'num_attention_heads': predefined_args['num_heads'],
'num_hidden_layers': predefined_args['num_layers'],
'pad_token_id': 1, # 2 = BERT, 1 = RoBERTa
'type_vocab_size': 1, # 2 = BERT, 1 = RoBERTa
'vocab_size': len(lowerCamelCase_ ),
}
__a : str = BertConfig.from_dict(lowerCamelCase_ )
__a : Optional[int] = BertForMaskedLM(lowerCamelCase_ )
hf_bort_model.eval()
# Parameter mapping table (Gluonnlp to Transformers)
# * denotes layer index
#
# | Gluon Parameter | Transformers Parameter
# | -------------------------------------------------------------- | ----------------------
# | `encoder.layer_norm.beta` | `bert.embeddings.LayerNorm.bias`
# | `encoder.layer_norm.gamma` | `bert.embeddings.LayerNorm.weight`
# | `encoder.position_weight` | `bert.embeddings.position_embeddings.weight`
# | `word_embed.0.weight` | `bert.embeddings.word_embeddings.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_key.bias` | `bert.encoder.layer.*.attention.self.key.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_key.weight` | `bert.encoder.layer.*.attention.self.key.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_query.bias` | `bert.encoder.layer.*.attention.self.query.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_query.weight` | `bert.encoder.layer.*.attention.self.query.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_value.bias` | `bert.encoder.layer.*.attention.self.value.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_value.weight` | `bert.encoder.layer.*.attention.self.value.weight`
# | `encoder.transformer_cells.*.ffn.ffn_2.bias` | `bert.encoder.layer.*.attention.output.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_2.weight` | `bert.encoder.layer.*.attention.output.dense.weight`
# | `encoder.transformer_cells.*.layer_norm.beta` | `bert.encoder.layer.*.attention.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.layer_norm.gamma` | `bert.encoder.layer.*.attention.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.ffn.ffn_1.bias` | `bert.encoder.layer.*.intermediate.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_1.weight` | `bert.encoder.layer.*.intermediate.dense.weight`
# | `encoder.transformer_cells.*.ffn.layer_norm.beta` | `bert.encoder.layer.*.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.ffn.layer_norm.gamma` | `bert.encoder.layer.*.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.proj.bias` | `bert.encoder.layer.*.output.dense.bias`
# | `encoder.transformer_cells.*.proj.weight` | `bert.encoder.layer.*.output.dense.weight`
# Helper function to convert MXNET Arrays to PyTorch
def to_torch(lowerCamelCase_ : Optional[Any] ) -> nn.Parameter:
return nn.Parameter(torch.FloatTensor(mx_array.data().asnumpy() ) )
# Check param shapes and map new HF param back
def check_and_map_params(lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[str] ):
__a : Optional[int] = hf_param.shape
__a : int = to_torch(params[gluon_param] )
__a : int = gluon_param.shape
assert (
shape_hf == shape_gluon
), f'''The gluon parameter {gluon_param} has shape {shape_gluon}, but expects shape {shape_hf} for Transformers'''
return gluon_param
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.word_embeddings.weight , 'word_embed.0.weight' )
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.position_embeddings.weight , 'encoder.position_weight' )
__a : Tuple = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.bias , 'encoder.layer_norm.beta' )
__a : Union[str, Any] = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.weight , 'encoder.layer_norm.gamma' )
# Inspired by RoBERTa conversion script, we just zero them out (Bort does not use them)
__a : Union[str, Any] = torch.zeros_like(
hf_bort_model.bert.embeddings.token_type_embeddings.weight.data )
for i in range(hf_bort_config.num_hidden_layers ):
__a : BertLayer = hf_bort_model.bert.encoder.layer[i]
# self attention
__a : BertSelfAttention = layer.attention.self
__a : Optional[int] = check_and_map_params(
self_attn.key.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.bias''' )
__a : str = check_and_map_params(
self_attn.key.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.weight''' )
__a : List[str] = check_and_map_params(
self_attn.query.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.bias''' )
__a : str = check_and_map_params(
self_attn.query.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.weight''' )
__a : Dict = check_and_map_params(
self_attn.value.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.bias''' )
__a : str = check_and_map_params(
self_attn.value.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.weight''' )
# self attention output
__a : BertSelfOutput = layer.attention.output
__a : Tuple = check_and_map_params(
self_output.dense.bias , f'''encoder.transformer_cells.{i}.proj.bias''' )
__a : Dict = check_and_map_params(
self_output.dense.weight , f'''encoder.transformer_cells.{i}.proj.weight''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.layer_norm.beta''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.layer_norm.gamma''' )
# intermediate
__a : BertIntermediate = layer.intermediate
__a : List[str] = check_and_map_params(
intermediate.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_1.bias''' )
__a : Optional[Any] = check_and_map_params(
intermediate.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_1.weight''' )
# output
__a : BertOutput = layer.output
__a : str = check_and_map_params(
bert_output.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_2.bias''' )
__a : List[Any] = check_and_map_params(
bert_output.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_2.weight''' )
__a : str = check_and_map_params(
bert_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.ffn.layer_norm.beta''' )
__a : List[str] = check_and_map_params(
bert_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.ffn.layer_norm.gamma''' )
# Save space and energy 🎄
hf_bort_model.half()
# Compare output of both models
__a : Union[str, Any] = RobertaTokenizer.from_pretrained('roberta-base' )
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ )['input_ids']
# Get gluon output
__a : Optional[int] = mx.nd.array([input_ids] )
__a : Tuple = original_bort(inputs=lowerCamelCase_ , token_types=[] )
# Get Transformer output (save and reload model again)
hf_bort_model.save_pretrained(lowerCamelCase_ )
__a : Optional[Any] = BertModel.from_pretrained(lowerCamelCase_ )
hf_bort_model.eval()
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ , return_tensors='pt' )
__a : int = hf_bort_model(**lowerCamelCase_ )[0]
__a : Dict = output_gluon[0].asnumpy()
__a : str = output_hf[0].detach().numpy()
__a : List[Any] = np.max(np.abs(hf_layer - gluon_layer ) ).item()
__a : str = np.allclose(lowerCamelCase_ , lowerCamelCase_ , atol=1e-3 )
if success:
print('✔️ Both model do output the same tensors' )
else:
print('❌ Both model do **NOT** output the same tensors' )
print('Absolute difference is:' , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--bort_checkpoint_path''', default=None, type=str, required=True, help='''Path the official Bort params file.'''
)
parser.add_argument(
'''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.'''
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_bort_checkpoint_to_pytorch(args.bort_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_sentencepiece_available
SCREAMING_SNAKE_CASE__ = {}
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''BartphoTokenizer''']
if TYPE_CHECKING:
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_bartpho import BartphoTokenizer
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[str] ):
__a : Any = ''
for i in table:
res += inp[i - 1]
return res
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
return data[1:] + data[0]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] ):
__a : Optional[int] = ''
for i in range(len(lowerCamelCase_ ) ):
if a[i] == b[i]:
res += "0"
else:
res += "1"
return res
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : str ):
__a : List[str] = int('0b' + data[0] + data[-1] , 2 )
__a : List[str] = int('0b' + data[1:3] , 2 )
return bin(s[row][col] )[2:]
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : List[str] , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] ):
__a : List[Any] = message[:4]
__a : str = message[4:]
__a : Any = apply_table(lowerCamelCase_ , lowerCamelCase_ )
__a : int = xor(lowerCamelCase_ , lowerCamelCase_ )
__a : Dict = apply_sbox(lowerCamelCase_ , temp[:4] ) # noqa: E741
__a : Tuple = apply_sbox(lowerCamelCase_ , temp[4:] )
__a : List[Any] = '0' * (2 - len(lowerCamelCase_ )) + l # noqa: E741
__a : List[str] = '0' * (2 - len(lowerCamelCase_ )) + r
__a : List[Any] = apply_table(l + r , lowerCamelCase_ )
__a : Dict = xor(lowerCamelCase_ , lowerCamelCase_ )
return temp + right
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter 10 bit key: ''')
SCREAMING_SNAKE_CASE__ = input('''Enter 8 bit message: ''')
SCREAMING_SNAKE_CASE__ = [6, 3, 7, 4, 8, 5, 10, 9]
SCREAMING_SNAKE_CASE__ = [3, 5, 2, 7, 4, 10, 1, 9, 8, 6]
SCREAMING_SNAKE_CASE__ = [2, 4, 3, 1]
SCREAMING_SNAKE_CASE__ = [2, 6, 3, 1, 4, 8, 5, 7]
SCREAMING_SNAKE_CASE__ = [4, 1, 3, 5, 7, 2, 8, 6]
SCREAMING_SNAKE_CASE__ = [4, 1, 2, 3, 2, 3, 4, 1]
SCREAMING_SNAKE_CASE__ = [[1, 0, 3, 2], [3, 2, 1, 0], [0, 2, 1, 3], [3, 1, 3, 2]]
SCREAMING_SNAKE_CASE__ = [[0, 1, 2, 3], [2, 0, 1, 3], [3, 0, 1, 0], [2, 1, 0, 3]]
# key generation
SCREAMING_SNAKE_CASE__ = apply_table(key, paa_table)
SCREAMING_SNAKE_CASE__ = temp[:5]
SCREAMING_SNAKE_CASE__ = temp[5:]
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
# encryption
SCREAMING_SNAKE_CASE__ = apply_table(message, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Cipher text is:''', CT)
# decryption
SCREAMING_SNAKE_CASE__ = apply_table(CT, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Plain text after decypting is:''', PT)
| 47 | 1 |
from typing import Dict, List, Optional, Union
import numpy as np
from ...image_processing_utils import BaseImageProcessor, BatchFeature, get_size_dict
from ...image_transforms import (
center_crop,
get_resize_output_image_size,
normalize,
rescale,
resize,
to_channel_dimension_format,
)
from ...image_utils import (
IMAGENET_STANDARD_MEAN,
IMAGENET_STANDARD_STD,
ChannelDimension,
ImageInput,
PILImageResampling,
make_list_of_images,
to_numpy_array,
valid_images,
)
from ...utils import TensorType, logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = ['''pixel_values''']
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, int]] = None , SCREAMING_SNAKE_CASE__ : PILImageResampling = PILImageResampling.BILINEAR , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : Dict[str, int] = None , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : Union[int, float] = 1 / 2_5_5 , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : Optional[Union[float, List[float]]] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[float, List[float]]] = None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = size if size is not None else {'shortest_edge': 2_5_6}
__a : str = get_size_dict(SCREAMING_SNAKE_CASE__ , default_to_square=SCREAMING_SNAKE_CASE__ )
__a : int = crop_size if crop_size is not None else {'height': 2_2_4, 'width': 2_2_4}
__a : Union[str, Any] = get_size_dict(SCREAMING_SNAKE_CASE__ )
__a : Dict = do_resize
__a : Union[str, Any] = size
__a : Any = resample
__a : List[Any] = do_center_crop
__a : Any = crop_size
__a : Tuple = do_rescale
__a : Optional[int] = rescale_factor
__a : Optional[int] = do_normalize
__a : int = image_mean if image_mean is not None else IMAGENET_STANDARD_MEAN
__a : Optional[Any] = image_std if image_std is not None else IMAGENET_STANDARD_STD
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : np.ndarray , SCREAMING_SNAKE_CASE__ : Dict[str, int] , SCREAMING_SNAKE_CASE__ : PILImageResampling = PILImageResampling.BICUBIC , SCREAMING_SNAKE_CASE__ : Optional[Union[str, ChannelDimension]] = None , **SCREAMING_SNAKE_CASE__ : List[Any] , ):
'''simple docstring'''
__a : List[Any] = get_size_dict(SCREAMING_SNAKE_CASE__ , default_to_square=SCREAMING_SNAKE_CASE__ )
if "shortest_edge" not in size:
raise ValueError(f'''The `size` parameter must contain the key `shortest_edge`. Got {size.keys()}''' )
__a : int = get_resize_output_image_size(SCREAMING_SNAKE_CASE__ , size=size['shortest_edge'] , default_to_square=SCREAMING_SNAKE_CASE__ )
return resize(SCREAMING_SNAKE_CASE__ , size=SCREAMING_SNAKE_CASE__ , resample=SCREAMING_SNAKE_CASE__ , data_format=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : np.ndarray , SCREAMING_SNAKE_CASE__ : Dict[str, int] , SCREAMING_SNAKE_CASE__ : Optional[Union[str, ChannelDimension]] = None , **SCREAMING_SNAKE_CASE__ : List[str] , ):
'''simple docstring'''
__a : str = get_size_dict(SCREAMING_SNAKE_CASE__ )
return center_crop(SCREAMING_SNAKE_CASE__ , size=(size['height'], size['width']) , data_format=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : np.ndarray , SCREAMING_SNAKE_CASE__ : float , SCREAMING_SNAKE_CASE__ : Optional[Union[str, ChannelDimension]] = None , **SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
return rescale(SCREAMING_SNAKE_CASE__ , scale=SCREAMING_SNAKE_CASE__ , data_format=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : np.ndarray , SCREAMING_SNAKE_CASE__ : Union[float, List[float]] , SCREAMING_SNAKE_CASE__ : Union[float, List[float]] , SCREAMING_SNAKE_CASE__ : Optional[Union[str, ChannelDimension]] = None , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
return normalize(SCREAMING_SNAKE_CASE__ , mean=SCREAMING_SNAKE_CASE__ , std=SCREAMING_SNAKE_CASE__ , data_format=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : ImageInput , SCREAMING_SNAKE_CASE__ : Optional[bool] = None , SCREAMING_SNAKE_CASE__ : Dict[str, int] = None , SCREAMING_SNAKE_CASE__ : PILImageResampling = None , SCREAMING_SNAKE_CASE__ : bool = None , SCREAMING_SNAKE_CASE__ : Dict[str, int] = None , SCREAMING_SNAKE_CASE__ : Optional[bool] = None , SCREAMING_SNAKE_CASE__ : Optional[float] = None , SCREAMING_SNAKE_CASE__ : Optional[bool] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[float, List[float]]] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[float, List[float]]] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[str, TensorType]] = None , SCREAMING_SNAKE_CASE__ : Union[str, ChannelDimension] = ChannelDimension.FIRST , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Tuple = do_resize if do_resize is not None else self.do_resize
__a : Optional[int] = size if size is not None else self.size
__a : int = get_size_dict(SCREAMING_SNAKE_CASE__ , default_to_square=SCREAMING_SNAKE_CASE__ )
__a : Dict = resample if resample is not None else self.resample
__a : Optional[int] = do_center_crop if do_center_crop is not None else self.do_center_crop
__a : Optional[int] = crop_size if crop_size is not None else self.crop_size
__a : List[Any] = get_size_dict(SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = do_rescale if do_rescale is not None else self.do_rescale
__a : Optional[Any] = rescale_factor if rescale_factor is not None else self.rescale_factor
__a : Optional[Any] = do_normalize if do_normalize is not None else self.do_normalize
__a : Optional[Any] = image_mean if image_mean is not None else self.image_mean
__a : List[Any] = image_std if image_std is not None else self.image_std
__a : int = make_list_of_images(SCREAMING_SNAKE_CASE__ )
if not valid_images(SCREAMING_SNAKE_CASE__ ):
raise ValueError(
'Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, '
'torch.Tensor, tf.Tensor or jax.ndarray.' )
if do_resize and size is None:
raise ValueError('Size must be specified if do_resize is True.' )
if do_center_crop and crop_size is None:
raise ValueError('Crop size must be specified if do_center_crop is True.' )
if do_rescale and rescale_factor is None:
raise ValueError('Rescale factor must be specified if do_rescale is True.' )
if do_normalize and (image_mean is None or image_std is None):
raise ValueError('Image mean and std must be specified if do_normalize is True.' )
# All transformations expect numpy arrays.
__a : Any = [to_numpy_array(SCREAMING_SNAKE_CASE__ ) for image in images]
if do_resize:
__a : Any = [self.resize(image=SCREAMING_SNAKE_CASE__ , size=SCREAMING_SNAKE_CASE__ , resample=SCREAMING_SNAKE_CASE__ ) for image in images]
if do_center_crop:
__a : str = [self.center_crop(image=SCREAMING_SNAKE_CASE__ , size=SCREAMING_SNAKE_CASE__ ) for image in images]
if do_rescale:
__a : List[Any] = [self.rescale(image=SCREAMING_SNAKE_CASE__ , scale=SCREAMING_SNAKE_CASE__ ) for image in images]
if do_normalize:
__a : Any = [self.normalize(image=SCREAMING_SNAKE_CASE__ , mean=SCREAMING_SNAKE_CASE__ , std=SCREAMING_SNAKE_CASE__ ) for image in images]
__a : Tuple = [to_channel_dimension_format(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) for image in images]
__a : int = {'pixel_values': images}
return BatchFeature(data=SCREAMING_SNAKE_CASE__ , tensor_type=SCREAMING_SNAKE_CASE__ )
| 47 |
import unittest
from transformers import is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from transformers import GradientAccumulator, create_optimizer
@require_tf
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : List[Any] = GradientAccumulator()
accumulator([tf.constant([1.0, 2.0] )] )
accumulator([tf.constant([-2.0, 1.0] )] )
accumulator([tf.constant([-1.0, 2.0] )] )
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
accumulator([tf.constant([1.0, 1.0] ), tf.constant([2.0, 2.0] )] )
self.assertEqual(accumulator.step , 3 )
self.assertEqual(len(accumulator.gradients ) , 1 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [-2.0, 5.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [0.0, 0.0] , tol=1e-2 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : int = None
ops.enable_eager_execution_internal()
__a : Optional[Any] = tf.config.list_physical_devices('CPU' )
if len(SCREAMING_SNAKE_CASE__ ) == 1:
tf.config.set_logical_device_configuration(
physical_devices[0] , [tf.config.LogicalDeviceConfiguration(), tf.config.LogicalDeviceConfiguration()] )
__a : int = tf.config.list_logical_devices(device_type='CPU' )
__a : str = tf.distribute.MirroredStrategy(devices=devices[:2] )
with strategy.scope():
__a : List[str] = GradientAccumulator()
__a : Tuple = tf.Variable([4.0, 3.0] )
__a , __a : int = create_optimizer(5e-5 , 1_0 , 5 )
__a : List[Any] = tf.Variable([0.0, 0.0] , trainable=SCREAMING_SNAKE_CASE__ )
def accumulate_on_replica(SCREAMING_SNAKE_CASE__ : Optional[Any] ):
accumulator([gradient] )
def apply_on_replica():
optimizer.apply_gradients(list(zip(accumulator.gradients , [variable] ) ) )
@tf.function
def accumulate(SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
with strategy.scope():
__a : Optional[Any] = strategy.experimental_local_results(SCREAMING_SNAKE_CASE__ )
local_variables[0].assign(SCREAMING_SNAKE_CASE__ )
local_variables[1].assign(SCREAMING_SNAKE_CASE__ )
strategy.run(SCREAMING_SNAKE_CASE__ , args=(gradient_placeholder,) )
@tf.function
def apply_grad():
with strategy.scope():
strategy.run(SCREAMING_SNAKE_CASE__ )
def _check_local_values(SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int ):
__a : Union[str, Any] = strategy.experimental_local_results(accumulator._gradients[0] )
self.assertListAlmostEqual(values[0].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
self.assertListAlmostEqual(values[1].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
accumulate([1.0, 2.0] , [-1.0, 1.0] )
accumulate([3.0, -1.0] , [-1.0, -1.0] )
accumulate([-2.0, 2.0] , [3.0, -2.0] )
self.assertEqual(accumulator.step , 3 )
_check_local_values([2.0, 3.0] , [1.0, -2.0] )
apply_grad()
self.assertListAlmostEqual(variable.value() , [4.0, 3.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
_check_local_values([0.0, 0.0] , [0.0, 0.0] )
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import (
OptionalDependencyNotAvailable,
_LazyModule,
is_sentencepiece_available,
is_tf_available,
is_tokenizers_available,
is_torch_available,
)
SCREAMING_SNAKE_CASE__ = {'''configuration_xlnet''': ['''XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''XLNetConfig''']}
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''XLNetTokenizer''']
try:
if not is_tokenizers_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''XLNetTokenizerFast''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''XLNET_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''XLNetForMultipleChoice''',
'''XLNetForQuestionAnswering''',
'''XLNetForQuestionAnsweringSimple''',
'''XLNetForSequenceClassification''',
'''XLNetForTokenClassification''',
'''XLNetLMHeadModel''',
'''XLNetModel''',
'''XLNetPreTrainedModel''',
'''load_tf_weights_in_xlnet''',
]
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''TF_XLNET_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''TFXLNetForMultipleChoice''',
'''TFXLNetForQuestionAnsweringSimple''',
'''TFXLNetForSequenceClassification''',
'''TFXLNetForTokenClassification''',
'''TFXLNetLMHeadModel''',
'''TFXLNetMainLayer''',
'''TFXLNetModel''',
'''TFXLNetPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_xlnet import XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLNetConfig
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_xlnet import XLNetTokenizer
try:
if not is_tokenizers_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_xlnet_fast import XLNetTokenizerFast
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_xlnet import (
XLNET_PRETRAINED_MODEL_ARCHIVE_LIST,
XLNetForMultipleChoice,
XLNetForQuestionAnswering,
XLNetForQuestionAnsweringSimple,
XLNetForSequenceClassification,
XLNetForTokenClassification,
XLNetLMHeadModel,
XLNetModel,
XLNetPreTrainedModel,
load_tf_weights_in_xlnet,
)
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_tf_xlnet import (
TF_XLNET_PRETRAINED_MODEL_ARCHIVE_LIST,
TFXLNetForMultipleChoice,
TFXLNetForQuestionAnsweringSimple,
TFXLNetForSequenceClassification,
TFXLNetForTokenClassification,
TFXLNetLMHeadModel,
TFXLNetMainLayer,
TFXLNetModel,
TFXLNetPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''roberta-base''': '''https://huggingface.co/roberta-base/resolve/main/config.json''',
'''roberta-large''': '''https://huggingface.co/roberta-large/resolve/main/config.json''',
'''roberta-large-mnli''': '''https://huggingface.co/roberta-large-mnli/resolve/main/config.json''',
'''distilroberta-base''': '''https://huggingface.co/distilroberta-base/resolve/main/config.json''',
'''roberta-base-openai-detector''': '''https://huggingface.co/roberta-base-openai-detector/resolve/main/config.json''',
'''roberta-large-openai-detector''': '''https://huggingface.co/roberta-large-openai-detector/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = '''roberta'''
def __init__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=5_0_2_6_5 , SCREAMING_SNAKE_CASE__ : Optional[int]=7_6_8 , SCREAMING_SNAKE_CASE__ : str=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Any="gelu" , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.1 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=2 , SCREAMING_SNAKE_CASE__ : Any=0.02 , SCREAMING_SNAKE_CASE__ : List[str]=1e-12 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0 , SCREAMING_SNAKE_CASE__ : List[str]=2 , SCREAMING_SNAKE_CASE__ : Tuple="absolute" , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[str]=None , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = vocab_size
__a : Tuple = hidden_size
__a : List[str] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : str = hidden_act
__a : Optional[Any] = intermediate_size
__a : Dict = hidden_dropout_prob
__a : List[str] = attention_probs_dropout_prob
__a : Optional[Any] = max_position_embeddings
__a : Dict = type_vocab_size
__a : str = initializer_range
__a : List[str] = layer_norm_eps
__a : Optional[int] = position_embedding_type
__a : Union[str, Any] = use_cache
__a : str = classifier_dropout
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
if self.task == "multiple-choice":
__a : List[str] = {0: 'batch', 1: 'choice', 2: 'sequence'}
else:
__a : Dict = {0: 'batch', 1: 'sequence'}
return OrderedDict(
[
('input_ids', dynamic_axis),
('attention_mask', dynamic_axis),
] )
| 47 | 1 |
from __future__ import annotations
from collections.abc import MutableSequence
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : MutableSequence[float] ):
'''simple docstring'''
if len(SCREAMING_SNAKE_CASE__ ) != degree + 1:
raise ValueError(
'The number of coefficients should be equal to the degree + 1.' )
__a : list[float] = list(SCREAMING_SNAKE_CASE__ )
__a : List[str] = degree
def __add__( self : str , SCREAMING_SNAKE_CASE__ : Polynomial ):
'''simple docstring'''
if self.degree > polynomial_a.degree:
__a : List[Any] = self.coefficients[:]
for i in range(polynomial_a.degree + 1 ):
coefficients[i] += polynomial_a.coefficients[i]
return Polynomial(self.degree , SCREAMING_SNAKE_CASE__ )
else:
__a : List[str] = polynomial_a.coefficients[:]
for i in range(self.degree + 1 ):
coefficients[i] += self.coefficients[i]
return Polynomial(polynomial_a.degree , SCREAMING_SNAKE_CASE__ )
def __sub__( self : str , SCREAMING_SNAKE_CASE__ : Polynomial ):
'''simple docstring'''
return self + polynomial_a * Polynomial(0 , [-1] )
def __neg__( self : List[Any] ):
'''simple docstring'''
return Polynomial(self.degree , [-c for c in self.coefficients] )
def __mul__( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Polynomial ):
'''simple docstring'''
__a : list[float] = [0] * (self.degree + polynomial_a.degree + 1)
for i in range(self.degree + 1 ):
for j in range(polynomial_a.degree + 1 ):
coefficients[i + j] += (
self.coefficients[i] * polynomial_a.coefficients[j]
)
return Polynomial(self.degree + polynomial_a.degree , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : int | float ):
'''simple docstring'''
__a : int | float = 0
for i in range(self.degree + 1 ):
result += self.coefficients[i] * (substitution**i)
return result
def __str__( self : Any ):
'''simple docstring'''
__a : Any = ''
for i in range(self.degree , -1 , -1 ):
if self.coefficients[i] == 0:
continue
elif self.coefficients[i] > 0:
if polynomial:
polynomial += " + "
else:
polynomial += " - "
if i == 0:
polynomial += str(abs(self.coefficients[i] ) )
elif i == 1:
polynomial += str(abs(self.coefficients[i] ) ) + "x"
else:
polynomial += str(abs(self.coefficients[i] ) ) + "x^" + str(SCREAMING_SNAKE_CASE__ )
return polynomial
def __repr__( self : Tuple ):
'''simple docstring'''
return self.__str__()
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : list[float] = [0] * self.degree
for i in range(self.degree ):
__a : List[Any] = self.coefficients[i + 1] * (i + 1)
return Polynomial(self.degree - 1 , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : int | float = 0 ):
'''simple docstring'''
__a : list[float] = [0] * (self.degree + 2)
__a : Optional[int] = constant
for i in range(self.degree + 1 ):
__a : List[Any] = self.coefficients[i] / (i + 1)
return Polynomial(self.degree + 1 , SCREAMING_SNAKE_CASE__ )
def __eq__( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : object ):
'''simple docstring'''
if not isinstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
return False
if self.degree != polynomial_a.degree:
return False
for i in range(self.degree + 1 ):
if self.coefficients[i] != polynomial_a.coefficients[i]:
return False
return True
def __ne__( self : Any , SCREAMING_SNAKE_CASE__ : object ):
'''simple docstring'''
return not self.__eq__(SCREAMING_SNAKE_CASE__ )
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''▁'''
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''sentencepiece.bpe.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''facebook/xglm-564M''': '''https://huggingface.co/facebook/xglm-564M/resolve/main/sentencepiece.bpe.model''',
}
}
SCREAMING_SNAKE_CASE__ = {
'''facebook/xglm-564M''': 2048,
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : Union[str, Any] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : str = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : Any = ['''input_ids''', '''attention_mask''']
def __init__( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : List[str]="</s>" , SCREAMING_SNAKE_CASE__ : Optional[Any]="</s>" , SCREAMING_SNAKE_CASE__ : Union[str, Any]="<s>" , SCREAMING_SNAKE_CASE__ : str="<unk>" , SCREAMING_SNAKE_CASE__ : Dict="<pad>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : List[str] , ):
'''simple docstring'''
__a : Union[str, Any] = {} if sp_model_kwargs is None else sp_model_kwargs
# Compatibility with the original tokenizer
__a : Any = 7
__a : Union[str, Any] = [f'''<madeupword{i}>''' for i in range(self.num_madeup_words )]
__a : Union[str, Any] = kwargs.get('additional_special_tokens' , [] )
kwargs["additional_special_tokens"] += [
word for word in madeup_words if word not in kwargs["additional_special_tokens"]
]
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , cls_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : List[str] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(str(SCREAMING_SNAKE_CASE__ ) )
__a : str = vocab_file
# Original fairseq vocab and spm vocab must be "aligned":
# Vocab | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
# -------- | ------- | ------- | ------ | ------- | --- | --- | --- | ----- | ----- | ----
# fairseq | '<s>' | '<pad>' | '</s>' | '<unk>' | ',' | '.' | '▁' | 's' | '▁de' | '-'
# spm | '<unk>' | '<s>' | '</s>' | ',' | '.' | '▁' | 's' | '▁de' | '-' | '▁a'
# The first "real" token "," has position 4 in the original fairseq vocab and position 3 in the spm vocab
__a : Any = 1
# Mimic fairseq token-to-id alignment for the first 4 token
__a : str = {'<s>': 0, '<pad>': 1, '</s>': 2, '<unk>': 3}
__a : List[str] = len(self.sp_model )
__a : Optional[int] = {f'''<madeupword{i}>''': sp_size + i + self.fairseq_offset for i in range(self.num_madeup_words )}
self.fairseq_tokens_to_ids.update(SCREAMING_SNAKE_CASE__ )
__a : Dict = {v: k for k, v in self.fairseq_tokens_to_ids.items()}
def __getstate__( self : List[str] ):
'''simple docstring'''
__a : Tuple = self.__dict__.copy()
__a : List[str] = None
__a : Optional[int] = self.sp_model.serialized_model_proto()
return state
def __setstate__( self : List[str] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : int = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : Dict = {}
__a : Tuple = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.LoadFromSerializedProto(self.sp_model_proto )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
if token_ids_a is None:
return [self.sep_token_id] + token_ids_a
__a : Optional[Any] = [self.sep_token_id]
return sep + token_ids_a + sep + sep + token_ids_a
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : bool = False ):
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=SCREAMING_SNAKE_CASE__ , token_ids_a=SCREAMING_SNAKE_CASE__ , already_has_special_tokens=SCREAMING_SNAKE_CASE__ )
if token_ids_a is None:
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
return [1] + ([0] * len(SCREAMING_SNAKE_CASE__ )) + [1, 1] + ([0] * len(SCREAMING_SNAKE_CASE__ ))
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
__a : Optional[int] = [self.sep_token_id]
if token_ids_a is None:
return len(sep + token_ids_a ) * [0]
return len(sep + token_ids_a + sep + sep + token_ids_a ) * [0]
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return len(self.sp_model ) + self.fairseq_offset + self.num_madeup_words
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : str = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if token in self.fairseq_tokens_to_ids:
return self.fairseq_tokens_to_ids[token]
__a : List[str] = self.sp_model.PieceToId(SCREAMING_SNAKE_CASE__ )
# Need to return unknown token if the SP model returned 0
return spm_id + self.fairseq_offset if spm_id else self.unk_token_id
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if index in self.fairseq_ids_to_tokens:
return self.fairseq_ids_to_tokens[index]
return self.sp_model.IdToPiece(index - self.fairseq_offset )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
__a : Optional[int] = ''.join(SCREAMING_SNAKE_CASE__ ).replace(SCREAMING_SNAKE_CASE__ , ' ' ).strip()
return out_string
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Any = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[Any] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''tiiuae/falcon-40b''': '''https://huggingface.co/tiiuae/falcon-40b/resolve/main/config.json''',
'''tiiuae/falcon-7b''': '''https://huggingface.co/tiiuae/falcon-7b/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[str] = '''falcon'''
__SCREAMING_SNAKE_CASE : Optional[Any] = ['''past_key_values''']
def __init__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Any=6_5_0_2_4 , SCREAMING_SNAKE_CASE__ : Any=4_5_4_4 , SCREAMING_SNAKE_CASE__ : Any=3_2 , SCREAMING_SNAKE_CASE__ : List[str]=7_1 , SCREAMING_SNAKE_CASE__ : int=1e-5 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0.02 , SCREAMING_SNAKE_CASE__ : str=True , SCREAMING_SNAKE_CASE__ : List[Any]=0.0 , SCREAMING_SNAKE_CASE__ : str=0.0 , SCREAMING_SNAKE_CASE__ : Optional[Any]=None , SCREAMING_SNAKE_CASE__ : Optional[Any]=False , SCREAMING_SNAKE_CASE__ : List[str]=False , SCREAMING_SNAKE_CASE__ : Optional[int]=True , SCREAMING_SNAKE_CASE__ : List[str]=True , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : str=1_1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_1 , **SCREAMING_SNAKE_CASE__ : Union[str, Any] , ):
'''simple docstring'''
__a : Any = vocab_size
# Backward compatibility with n_embed kwarg
__a : int = kwargs.pop('n_embed' , SCREAMING_SNAKE_CASE__ )
__a : str = hidden_size if n_embed is None else n_embed
__a : Any = num_hidden_layers
__a : Dict = num_attention_heads
__a : Union[str, Any] = layer_norm_epsilon
__a : List[str] = initializer_range
__a : Optional[Any] = use_cache
__a : List[str] = hidden_dropout
__a : Dict = attention_dropout
__a : Any = bos_token_id
__a : Optional[int] = eos_token_id
__a : List[Any] = num_attention_heads if num_kv_heads is None else num_kv_heads
__a : int = alibi
__a : Tuple = new_decoder_architecture
__a : str = multi_query # Ignored when new_decoder_architecture is True
__a : Any = parallel_attn
__a : Dict = bias
super().__init__(bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.hidden_size // self.num_attention_heads
@property
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return not self.alibi
| 47 |
import argparse
from collections import OrderedDict
from pathlib import Path
import torch
from transformers import (
VisualBertConfig,
VisualBertForMultipleChoice,
VisualBertForPreTraining,
VisualBertForQuestionAnswering,
VisualBertForVisualReasoning,
)
from transformers.utils import logging
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = [
('''bert.bert''', '''visual_bert'''),
('''bert.cls''', '''cls'''),
('''bert.classifier''', '''cls'''),
('''token_type_embeddings_visual''', '''visual_token_type_embeddings'''),
('''position_embeddings_visual''', '''visual_position_embeddings'''),
('''projection''', '''visual_projection'''),
]
SCREAMING_SNAKE_CASE__ = [
'''nlvr2_coco_pre_trained.th''',
'''nlvr2_fine_tuned.th''',
'''nlvr2_pre_trained.th''',
'''vcr_coco_pre_train.th''',
'''vcr_fine_tune.th''',
'''vcr_pre_train.th''',
'''vqa_coco_pre_trained.th''',
'''vqa_fine_tuned.th''',
'''vqa_pre_trained.th''',
]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ):
__a : str = torch.load(lowerCamelCase_ , map_location='cpu' )
return sd
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : int , lowerCamelCase_ : Dict=rename_keys_prefix ):
__a : Optional[Any] = OrderedDict()
__a : Any = torch.arange(config.max_position_embeddings ).expand((1, -1) )
# detector_d = OrderedDict()
for key in d:
if "detector" in key:
# detector_d[key.replace('detector.','')] = d[key]
continue
__a : List[Any] = key
for name_pair in rename_keys_prefix:
__a : List[str] = new_key.replace(name_pair[0] , name_pair[1] )
__a : Any = d[key]
if key == "bert.cls.predictions.decoder.weight":
# Old bert code didn't have `decoder.bias`, but was added separately
__a : int = new_d['cls.predictions.bias']
return new_d
@torch.no_grad()
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : Any ):
assert (
checkpoint_path.split('/' )[-1] in ACCEPTABLE_CHECKPOINTS
), f'''The checkpoint provided must be in {ACCEPTABLE_CHECKPOINTS}.'''
# Get Config
if "pre" in checkpoint_path:
__a : Dict = 'pretraining'
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
elif "vqa_advanced" in checkpoint_path:
__a : int = {'visual_embedding_dim': 2_0_4_8}
elif "vqa" in checkpoint_path:
__a : Tuple = {'visual_embedding_dim': 2_0_4_8}
elif "nlvr" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 1_0_2_4}
else:
raise NotImplementedError(f'''No implementation found for `{checkpoint_path}`.''' )
else:
if "vcr" in checkpoint_path:
__a : int = {'visual_embedding_dim': 5_1_2}
__a : Any = 'multichoice'
elif "vqa_advanced" in checkpoint_path:
__a : Any = {'visual_embedding_dim': 2_0_4_8}
__a : List[str] = 'vqa_advanced'
elif "vqa" in checkpoint_path:
__a : List[Any] = {'visual_embedding_dim': 2_0_4_8, 'num_labels': 3_1_2_9}
__a : List[Any] = 'vqa'
elif "nlvr" in checkpoint_path:
__a : Optional[int] = {
'visual_embedding_dim': 1_0_2_4,
'num_labels': 2,
}
__a : Optional[Any] = 'nlvr'
__a : str = VisualBertConfig(**lowerCamelCase_ )
# Load State Dict
__a : str = load_state_dict(lowerCamelCase_ )
__a : str = get_new_dict(lowerCamelCase_ , lowerCamelCase_ )
if model_type == "pretraining":
__a : Optional[Any] = VisualBertForPreTraining(lowerCamelCase_ )
elif model_type == "vqa":
__a : Any = VisualBertForQuestionAnswering(lowerCamelCase_ )
elif model_type == "nlvr":
__a : int = VisualBertForVisualReasoning(lowerCamelCase_ )
elif model_type == "multichoice":
__a : Optional[int] = VisualBertForMultipleChoice(lowerCamelCase_ )
model.load_state_dict(lowerCamelCase_ )
# Save Checkpoints
Path(lowerCamelCase_ ).mkdir(exist_ok=lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument('''orig_checkpoint_path''', type=str, help='''A path to .th on local filesystem.''')
parser.add_argument('''pytorch_dump_folder_path''', type=str, help='''Path to the output PyTorch model.''')
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_visual_bert_checkpoint(args.orig_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
import gc
import random
import unittest
import numpy as np
import torch
from PIL import Image
from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer
from diffusers import AutoencoderKL, DDIMScheduler, DDPMScheduler, StableDiffusionUpscalePipeline, UNetaDConditionModel
from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device
from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu
enable_full_determinism()
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
super().tearDown()
gc.collect()
torch.cuda.empty_cache()
@property
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : List[str] = 1
__a : List[Any] = 3
__a : str = (3_2, 3_2)
__a : List[Any] = floats_tensor((batch_size, num_channels) + sizes , rng=random.Random(0 ) ).to(SCREAMING_SNAKE_CASE__ )
return image
@property
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
torch.manual_seed(0 )
__a : Optional[int] = UNetaDConditionModel(
block_out_channels=(3_2, 3_2, 6_4) , layers_per_block=2 , sample_size=3_2 , in_channels=7 , out_channels=4 , down_block_types=('DownBlock2D', 'CrossAttnDownBlock2D', 'CrossAttnDownBlock2D') , up_block_types=('CrossAttnUpBlock2D', 'CrossAttnUpBlock2D', 'UpBlock2D') , cross_attention_dim=3_2 , attention_head_dim=8 , use_linear_projection=SCREAMING_SNAKE_CASE__ , only_cross_attention=(True, True, False) , num_class_embeds=1_0_0 , )
return model
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
torch.manual_seed(0 )
__a : Optional[Any] = AutoencoderKL(
block_out_channels=[3_2, 3_2, 6_4] , in_channels=3 , out_channels=3 , down_block_types=['DownEncoderBlock2D', 'DownEncoderBlock2D', 'DownEncoderBlock2D'] , up_block_types=['UpDecoderBlock2D', 'UpDecoderBlock2D', 'UpDecoderBlock2D'] , latent_channels=4 , )
return model
@property
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
torch.manual_seed(0 )
__a : Dict = CLIPTextConfig(
bos_token_id=0 , eos_token_id=2 , hidden_size=3_2 , intermediate_size=3_7 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1_0_0_0 , hidden_act='gelu' , projection_dim=5_1_2 , )
return CLIPTextModel(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Tuple = 'cpu' # ensure determinism for the device-dependent torch.Generator
__a : int = self.dummy_cond_unet_upscale
__a : Any = DDPMScheduler()
__a : Any = DDIMScheduler(prediction_type='v_prediction' )
__a : Any = self.dummy_vae
__a : List[str] = self.dummy_text_encoder
__a : List[Any] = CLIPTokenizer.from_pretrained('hf-internal-testing/tiny-random-clip' )
__a : Optional[int] = self.dummy_image.cpu().permute(0 , 2 , 3 , 1 )[0]
__a : Optional[Any] = Image.fromarray(np.uinta(SCREAMING_SNAKE_CASE__ ) ).convert('RGB' ).resize((6_4, 6_4) )
# make sure here that pndm scheduler skips prk
__a : str = StableDiffusionUpscalePipeline(
unet=SCREAMING_SNAKE_CASE__ , low_res_scheduler=SCREAMING_SNAKE_CASE__ , scheduler=SCREAMING_SNAKE_CASE__ , vae=SCREAMING_SNAKE_CASE__ , text_encoder=SCREAMING_SNAKE_CASE__ , tokenizer=SCREAMING_SNAKE_CASE__ , max_noise_level=3_5_0 , )
__a : List[Any] = sd_pipe.to(SCREAMING_SNAKE_CASE__ )
sd_pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = 'A painting of a squirrel eating a burger'
__a : Any = torch.Generator(device=SCREAMING_SNAKE_CASE__ ).manual_seed(0 )
__a : List[str] = sd_pipe(
[prompt] , image=SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ , guidance_scale=6.0 , noise_level=2_0 , num_inference_steps=2 , output_type='np' , )
__a : Tuple = output.images
__a : Optional[Any] = torch.Generator(device=SCREAMING_SNAKE_CASE__ ).manual_seed(0 )
__a : int = sd_pipe(
[prompt] , image=SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ , guidance_scale=6.0 , noise_level=2_0 , num_inference_steps=2 , output_type='np' , return_dict=SCREAMING_SNAKE_CASE__ , )[0]
__a : Dict = image[0, -3:, -3:, -1]
__a : Union[str, Any] = image_from_tuple[0, -3:, -3:, -1]
__a : Any = low_res_image.size[0] * 4
assert image.shape == (1, expected_height_width, expected_height_width, 3)
__a : Dict = np.array([0.3_113, 0.3_910, 0.4_272, 0.4_859, 0.5_061, 0.4_652, 0.5_362, 0.5_715, 0.5_661] )
assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2
assert np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
__a : List[str] = 'cpu' # ensure determinism for the device-dependent torch.Generator
__a : List[str] = self.dummy_cond_unet_upscale
__a : Dict = DDPMScheduler()
__a : Dict = DDIMScheduler(prediction_type='v_prediction' )
__a : int = self.dummy_vae
__a : Optional[int] = self.dummy_text_encoder
__a : str = CLIPTokenizer.from_pretrained('hf-internal-testing/tiny-random-clip' )
__a : Optional[Any] = self.dummy_image.cpu().permute(0 , 2 , 3 , 1 )[0]
__a : int = Image.fromarray(np.uinta(SCREAMING_SNAKE_CASE__ ) ).convert('RGB' ).resize((6_4, 6_4) )
# make sure here that pndm scheduler skips prk
__a : Dict = StableDiffusionUpscalePipeline(
unet=SCREAMING_SNAKE_CASE__ , low_res_scheduler=SCREAMING_SNAKE_CASE__ , scheduler=SCREAMING_SNAKE_CASE__ , vae=SCREAMING_SNAKE_CASE__ , text_encoder=SCREAMING_SNAKE_CASE__ , tokenizer=SCREAMING_SNAKE_CASE__ , max_noise_level=3_5_0 , )
__a : int = sd_pipe.to(SCREAMING_SNAKE_CASE__ )
sd_pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
__a : List[str] = 'A painting of a squirrel eating a burger'
__a : str = sd_pipe(
2 * [prompt] , image=2 * [low_res_image] , guidance_scale=6.0 , noise_level=2_0 , num_inference_steps=2 , output_type='np' , )
__a : Any = output.images
assert image.shape[0] == 2
__a : List[Any] = torch.Generator(device=SCREAMING_SNAKE_CASE__ ).manual_seed(0 )
__a : int = sd_pipe(
[prompt] , image=SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ , num_images_per_prompt=2 , guidance_scale=6.0 , noise_level=2_0 , num_inference_steps=2 , output_type='np' , )
__a : Optional[Any] = output.images
assert image.shape[0] == 2
@unittest.skipIf(torch_device != 'cuda' , 'This test requires a GPU' )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Dict = self.dummy_cond_unet_upscale
__a : Any = DDPMScheduler()
__a : Tuple = DDIMScheduler(prediction_type='v_prediction' )
__a : Union[str, Any] = self.dummy_vae
__a : Dict = self.dummy_text_encoder
__a : List[Any] = CLIPTokenizer.from_pretrained('hf-internal-testing/tiny-random-clip' )
__a : Optional[int] = self.dummy_image.cpu().permute(0 , 2 , 3 , 1 )[0]
__a : Dict = Image.fromarray(np.uinta(SCREAMING_SNAKE_CASE__ ) ).convert('RGB' ).resize((6_4, 6_4) )
# put models in fp16, except vae as it overflows in fp16
__a : str = unet.half()
__a : Any = text_encoder.half()
# make sure here that pndm scheduler skips prk
__a : Optional[Any] = StableDiffusionUpscalePipeline(
unet=SCREAMING_SNAKE_CASE__ , low_res_scheduler=SCREAMING_SNAKE_CASE__ , scheduler=SCREAMING_SNAKE_CASE__ , vae=SCREAMING_SNAKE_CASE__ , text_encoder=SCREAMING_SNAKE_CASE__ , tokenizer=SCREAMING_SNAKE_CASE__ , max_noise_level=3_5_0 , )
__a : int = sd_pipe.to(SCREAMING_SNAKE_CASE__ )
sd_pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
__a : str = 'A painting of a squirrel eating a burger'
__a : Tuple = torch.manual_seed(0 )
__a : int = sd_pipe(
[prompt] , image=SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ , num_inference_steps=2 , output_type='np' , ).images
__a : List[str] = low_res_image.size[0] * 4
assert image.shape == (1, expected_height_width, expected_height_width, 3)
@slow
@require_torch_gpu
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
super().tearDown()
gc.collect()
torch.cuda.empty_cache()
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Optional[int] = load_image(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/sd2-upscale/low_res_cat.png' )
__a : int = load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-upscale'
'/upsampled_cat.npy' )
__a : List[Any] = 'stabilityai/stable-diffusion-x4-upscaler'
__a : List[Any] = StableDiffusionUpscalePipeline.from_pretrained(SCREAMING_SNAKE_CASE__ )
pipe.to(SCREAMING_SNAKE_CASE__ )
pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
pipe.enable_attention_slicing()
__a : Any = 'a cat sitting on a park bench'
__a : List[str] = torch.manual_seed(0 )
__a : List[Any] = pipe(
prompt=SCREAMING_SNAKE_CASE__ , image=SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ , output_type='np' , )
__a : Union[str, Any] = output.images[0]
assert image.shape == (5_1_2, 5_1_2, 3)
assert np.abs(expected_image - image ).max() < 1e-3
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : List[Any] = load_image(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/sd2-upscale/low_res_cat.png' )
__a : List[Any] = load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-upscale'
'/upsampled_cat_fp16.npy' )
__a : Dict = 'stabilityai/stable-diffusion-x4-upscaler'
__a : Any = StableDiffusionUpscalePipeline.from_pretrained(
SCREAMING_SNAKE_CASE__ , torch_dtype=torch.floataa , )
pipe.to(SCREAMING_SNAKE_CASE__ )
pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
pipe.enable_attention_slicing()
__a : Optional[Any] = 'a cat sitting on a park bench'
__a : Any = torch.manual_seed(0 )
__a : List[Any] = pipe(
prompt=SCREAMING_SNAKE_CASE__ , image=SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ , output_type='np' , )
__a : Tuple = output.images[0]
assert image.shape == (5_1_2, 5_1_2, 3)
assert np.abs(expected_image - image ).max() < 5e-1
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
torch.cuda.empty_cache()
torch.cuda.reset_max_memory_allocated()
torch.cuda.reset_peak_memory_stats()
__a : int = load_image(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/sd2-upscale/low_res_cat.png' )
__a : Dict = 'stabilityai/stable-diffusion-x4-upscaler'
__a : str = StableDiffusionUpscalePipeline.from_pretrained(
SCREAMING_SNAKE_CASE__ , torch_dtype=torch.floataa , )
pipe.to(SCREAMING_SNAKE_CASE__ )
pipe.set_progress_bar_config(disable=SCREAMING_SNAKE_CASE__ )
pipe.enable_attention_slicing(1 )
pipe.enable_sequential_cpu_offload()
__a : List[Any] = 'a cat sitting on a park bench'
__a : Union[str, Any] = torch.manual_seed(0 )
__a : Dict = pipe(
prompt=SCREAMING_SNAKE_CASE__ , image=SCREAMING_SNAKE_CASE__ , generator=SCREAMING_SNAKE_CASE__ , num_inference_steps=5 , output_type='np' , )
__a : Optional[int] = torch.cuda.max_memory_allocated()
# make sure that less than 2.9 GB is allocated
assert mem_bytes < 2.9 * 1_0**9
| 47 |
print((lambda quine: quine % quine)('''print((lambda quine: quine %% quine)(%r))'''))
| 47 | 1 |
import os
import tempfile
import unittest
from transformers import NezhaConfig, is_torch_available
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_torch_gpu, slow, torch_device
from ...generation.test_utils import GenerationTesterMixin
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor, random_attention_mask
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import (
MODEL_FOR_PRETRAINING_MAPPING,
NezhaForMaskedLM,
NezhaForMultipleChoice,
NezhaForNextSentencePrediction,
NezhaForPreTraining,
NezhaForQuestionAnswering,
NezhaForSequenceClassification,
NezhaForTokenClassification,
NezhaModel,
)
from transformers.models.nezha.modeling_nezha import NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST
class _UpperCamelCase:
def __init__( self : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_3 , SCREAMING_SNAKE_CASE__ : Optional[int]=7 , SCREAMING_SNAKE_CASE__ : List[Any]=True , SCREAMING_SNAKE_CASE__ : List[Any]=True , SCREAMING_SNAKE_CASE__ : int=True , SCREAMING_SNAKE_CASE__ : List[Any]=True , SCREAMING_SNAKE_CASE__ : Any=9_9 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5 , SCREAMING_SNAKE_CASE__ : int=4 , SCREAMING_SNAKE_CASE__ : Dict=3_7 , SCREAMING_SNAKE_CASE__ : Union[str, Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : int=1_2_8 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_6 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Any=0.02 , SCREAMING_SNAKE_CASE__ : Optional[int]=3 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=4 , SCREAMING_SNAKE_CASE__ : Optional[int]=None , ):
'''simple docstring'''
__a : Union[str, Any] = parent
__a : str = batch_size
__a : List[Any] = seq_length
__a : List[Any] = is_training
__a : str = use_input_mask
__a : Union[str, Any] = use_token_type_ids
__a : Union[str, Any] = use_labels
__a : Dict = vocab_size
__a : Tuple = hidden_size
__a : List[str] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : Union[str, Any] = intermediate_size
__a : Tuple = hidden_act
__a : int = hidden_dropout_prob
__a : Dict = attention_probs_dropout_prob
__a : Any = max_position_embeddings
__a : Any = type_vocab_size
__a : Dict = type_sequence_label_size
__a : Optional[int] = initializer_range
__a : Tuple = num_labels
__a : Any = num_choices
__a : Any = scope
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Optional[int] = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a : List[str] = None
if self.use_input_mask:
__a : List[Any] = random_attention_mask([self.batch_size, self.seq_length] )
__a : str = None
if self.use_token_type_ids:
__a : List[str] = ids_tensor([self.batch_size, self.seq_length] , self.type_vocab_size )
__a : Dict = None
__a : Union[str, Any] = None
__a : Union[str, Any] = None
if self.use_labels:
__a : Any = ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a : List[str] = ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a : List[str] = ids_tensor([self.batch_size] , self.num_choices )
__a : Optional[Any] = self.get_config()
return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
return NezhaConfig(
vocab_size=self.vocab_size , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , max_position_embeddings=self.max_position_embeddings , type_vocab_size=self.type_vocab_size , is_decoder=SCREAMING_SNAKE_CASE__ , initializer_range=self.initializer_range , )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : Tuple = self.prepare_config_and_inputs()
__a : str = True
__a : Any = floats_tensor([self.batch_size, self.seq_length, self.hidden_size] )
__a : int = ids_tensor([self.batch_size, self.seq_length] , vocab_size=2 )
return (
config,
input_ids,
token_type_ids,
input_mask,
sequence_labels,
token_labels,
choice_labels,
encoder_hidden_states,
encoder_attention_mask,
)
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Optional[Any] = NezhaModel(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Optional[Any] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ )
__a : str = model(SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ )
__a : Any = model(SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
self.parent.assertEqual(result.pooler_output.shape , (self.batch_size, self.hidden_size) )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : int , ):
'''simple docstring'''
__a : Optional[Any] = True
__a : List[Any] = NezhaModel(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Optional[Any] = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , encoder_hidden_states=SCREAMING_SNAKE_CASE__ , encoder_attention_mask=SCREAMING_SNAKE_CASE__ , )
__a : Any = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , encoder_hidden_states=SCREAMING_SNAKE_CASE__ , )
__a : str = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
self.parent.assertEqual(result.pooler_output.shape , (self.batch_size, self.hidden_size) )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
__a : Dict = NezhaForMaskedLM(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : str = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : Tuple = NezhaForNextSentencePrediction(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Dict = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, 2) )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Union[str, Any] = NezhaForPreTraining(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : List[str] = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ , next_sentence_label=SCREAMING_SNAKE_CASE__ , )
self.parent.assertEqual(result.prediction_logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
self.parent.assertEqual(result.seq_relationship_logits.shape , (self.batch_size, 2) )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a : Any = NezhaForQuestionAnswering(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : int = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , start_positions=SCREAMING_SNAKE_CASE__ , end_positions=SCREAMING_SNAKE_CASE__ , )
self.parent.assertEqual(result.start_logits.shape , (self.batch_size, self.seq_length) )
self.parent.assertEqual(result.end_logits.shape , (self.batch_size, self.seq_length) )
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
__a : int = self.num_labels
__a : Any = NezhaForSequenceClassification(SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Any = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
__a : Any = self.num_labels
__a : Tuple = NezhaForTokenClassification(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Optional[Any] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
__a : List[str] = self.num_choices
__a : str = NezhaForMultipleChoice(config=SCREAMING_SNAKE_CASE__ )
model.to(SCREAMING_SNAKE_CASE__ )
model.eval()
__a : Dict = input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a : Tuple = token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a : Any = input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a : str = model(
SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ , token_type_ids=SCREAMING_SNAKE_CASE__ , labels=SCREAMING_SNAKE_CASE__ , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Tuple = self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : int = config_and_inputs
__a : List[Any] = {'input_ids': input_ids, 'token_type_ids': token_type_ids, 'attention_mask': input_mask}
return config, inputs_dict
@require_torch
class _UpperCamelCase( __lowerCamelCase , __lowerCamelCase , __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : str = (
(
NezhaModel,
NezhaForMaskedLM,
NezhaForMultipleChoice,
NezhaForNextSentencePrediction,
NezhaForPreTraining,
NezhaForQuestionAnswering,
NezhaForSequenceClassification,
NezhaForTokenClassification,
)
if is_torch_available()
else ()
)
__SCREAMING_SNAKE_CASE : Dict = (
{
'''feature-extraction''': NezhaModel,
'''fill-mask''': NezhaForMaskedLM,
'''question-answering''': NezhaForQuestionAnswering,
'''text-classification''': NezhaForSequenceClassification,
'''token-classification''': NezhaForTokenClassification,
'''zero-shot''': NezhaForSequenceClassification,
}
if is_torch_available()
else {}
)
__SCREAMING_SNAKE_CASE : Optional[Any] = True
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : str=False ):
'''simple docstring'''
__a : List[Any] = super()._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , return_labels=SCREAMING_SNAKE_CASE__ )
if return_labels:
if model_class in get_values(SCREAMING_SNAKE_CASE__ ):
__a : List[str] = torch.zeros(
(self.model_tester.batch_size, self.model_tester.seq_length) , dtype=torch.long , device=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=SCREAMING_SNAKE_CASE__ )
return inputs_dict
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : int = NezhaModelTester(self )
__a : List[Any] = ConfigTester(self , config_class=SCREAMING_SNAKE_CASE__ , hidden_size=3_7 )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
self.config_tester.run_common_tests()
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : str = self.model_tester.prepare_config_and_inputs_for_decoder()
self.model_tester.create_and_check_model_as_decoder(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : List[Any] = self.model_tester.prepare_config_and_inputs_for_decoder()
__a : Dict = None
self.model_tester.create_and_check_model_as_decoder(
SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : Union[str, Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_multiple_choice(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
__a : Any = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_next_sequence_prediction(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_pretraining(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Any = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_question_answering(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Optional[Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_sequence_classification(*SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_token_classification(*SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for model_name in NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a : Optional[Any] = NezhaModel.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@slow
@require_torch_gpu
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
__a , __a : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
# NezhaForMultipleChoice behaves incorrectly in JIT environments.
if model_class == NezhaForMultipleChoice:
return
__a : Union[str, Any] = True
__a : Tuple = model_class(config=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = self._prepare_for_class(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = torch.jit.trace(
SCREAMING_SNAKE_CASE__ , (inputs_dict['input_ids'].to('cpu' ), inputs_dict['attention_mask'].to('cpu' )) )
with tempfile.TemporaryDirectory() as tmp:
torch.jit.save(SCREAMING_SNAKE_CASE__ , os.path.join(SCREAMING_SNAKE_CASE__ , 'bert.pt' ) )
__a : str = torch.jit.load(os.path.join(SCREAMING_SNAKE_CASE__ , 'bert.pt' ) , map_location=SCREAMING_SNAKE_CASE__ )
loaded(inputs_dict['input_ids'].to(SCREAMING_SNAKE_CASE__ ) , inputs_dict['attention_mask'].to(SCREAMING_SNAKE_CASE__ ) )
@require_torch
class _UpperCamelCase( unittest.TestCase ):
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : List[str] = NezhaModel.from_pretrained('sijunhe/nezha-cn-base' )
__a : Dict = torch.tensor([[0, 1, 2, 3, 4, 5]] )
__a : Optional[int] = torch.tensor([[0, 1, 1, 1, 1, 1]] )
with torch.no_grad():
__a : List[str] = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ )[0]
__a : Dict = torch.Size((1, 6, 7_6_8) )
self.assertEqual(output.shape , SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = torch.tensor([[[0.0_685, 0.2_441, 0.1_102], [0.0_600, 0.1_906, 0.1_349], [0.0_221, 0.0_819, 0.0_586]]] )
self.assertTrue(torch.allclose(output[:, 1:4, 1:4] , SCREAMING_SNAKE_CASE__ , atol=1e-4 ) )
@slow
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Any = NezhaForMaskedLM.from_pretrained('sijunhe/nezha-cn-base' )
__a : Optional[Any] = torch.tensor([[0, 1, 2, 3, 4, 5]] )
__a : int = torch.tensor([[1, 1, 1, 1, 1, 1]] )
with torch.no_grad():
__a : Dict = model(SCREAMING_SNAKE_CASE__ , attention_mask=SCREAMING_SNAKE_CASE__ )[0]
__a : int = torch.Size((1, 6, 2_1_1_2_8) )
self.assertEqual(output.shape , SCREAMING_SNAKE_CASE__ )
__a : str = torch.tensor(
[[-2.7_939, -1.7_902, -2.2_189], [-2.8_585, -1.8_908, -2.3_723], [-2.6_499, -1.7_750, -2.2_558]] )
self.assertTrue(torch.allclose(output[:, 1:4, 1:4] , SCREAMING_SNAKE_CASE__ , atol=1e-4 ) )
| 47 |
import json
import os
import shutil
import tempfile
from unittest import TestCase
from transformers import BartTokenizer, BartTokenizerFast, DPRQuestionEncoderTokenizer, DPRQuestionEncoderTokenizerFast
from transformers.models.bart.configuration_bart import BartConfig
from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES
from transformers.models.dpr.configuration_dpr import DPRConfig
from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
from transformers.testing_utils import require_faiss, require_tokenizers, require_torch, slow
from transformers.utils import is_datasets_available, is_faiss_available, is_torch_available
if is_torch_available() and is_datasets_available() and is_faiss_available():
from transformers.models.rag.configuration_rag import RagConfig
from transformers.models.rag.tokenization_rag import RagTokenizer
@require_faiss
@require_torch
class _UpperCamelCase( __lowerCamelCase ):
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[Any] = tempfile.mkdtemp()
__a : int = 8
# DPR tok
__a : Dict = [
'[UNK]',
'[CLS]',
'[SEP]',
'[PAD]',
'[MASK]',
'want',
'##want',
'##ed',
'wa',
'un',
'runn',
'##ing',
',',
'low',
'lowest',
]
__a : int = os.path.join(self.tmpdirname , 'dpr_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , DPR_VOCAB_FILES_NAMES['vocab_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as vocab_writer:
vocab_writer.write(''.join([x + '\n' for x in vocab_tokens] ) )
# BART tok
__a : str = [
'l',
'o',
'w',
'e',
'r',
's',
't',
'i',
'd',
'n',
'\u0120',
'\u0120l',
'\u0120n',
'\u0120lo',
'\u0120low',
'er',
'\u0120lowest',
'\u0120newer',
'\u0120wider',
'<unk>',
]
__a : Optional[int] = dict(zip(SCREAMING_SNAKE_CASE__ , range(len(SCREAMING_SNAKE_CASE__ ) ) ) )
__a : List[str] = ['#version: 0.2', '\u0120 l', '\u0120l o', '\u0120lo w', 'e r', '']
__a : List[str] = {'unk_token': '<unk>'}
__a : Dict = os.path.join(self.tmpdirname , 'bart_tokenizer' )
os.makedirs(SCREAMING_SNAKE_CASE__ , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['vocab_file'] )
__a : Dict = os.path.join(SCREAMING_SNAKE_CASE__ , BART_VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(SCREAMING_SNAKE_CASE__ ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'dpr_tokenizer' ) )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'bart_tokenizer' ) )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
shutil.rmtree(self.tmpdirname )
@require_tokenizers
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : Tuple = os.path.join(self.tmpdirname , 'rag_tokenizer' )
__a : Optional[Any] = RagConfig(question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() )
__a : Optional[Any] = RagTokenizer(question_encoder=self.get_dpr_tokenizer() , generator=self.get_bart_tokenizer() )
rag_config.save_pretrained(SCREAMING_SNAKE_CASE__ )
rag_tokenizer.save_pretrained(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = RagTokenizer.from_pretrained(SCREAMING_SNAKE_CASE__ , config=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(new_rag_tokenizer.question_encoder , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.question_encoder.get_vocab() , rag_tokenizer.question_encoder.get_vocab() )
self.assertIsInstance(new_rag_tokenizer.generator , SCREAMING_SNAKE_CASE__ )
self.assertEqual(new_rag_tokenizer.generator.get_vocab() , rag_tokenizer.generator.get_vocab() )
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Optional[Any] = RagTokenizer.from_pretrained('facebook/rag-token-nq' )
__a : List[Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : Tuple = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Any = RagTokenizer.from_pretrained('facebook/rag-sequence-nq' )
__a : Union[str, Any] = [
'who got the first nobel prize in physics',
'when is the next deadpool movie being released',
'which mode is used for short wave broadcast service',
'who is the owner of reading football club',
'when is the next scandal episode coming out',
'when is the last time the philadelphia won the superbowl',
'what is the most current adobe flash player version',
'how many episodes are there in dragon ball z',
'what is the first step in the evolution of the eye',
'where is gall bladder situated in human body',
'what is the main mineral in lithium batteries',
'who is the president of usa right now',
'where do the greasers live in the outsiders',
'panda is a national animal of which country',
'what is the name of manchester united stadium',
]
__a : str = tokenizer(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
| 47 | 1 |
import os
import re
import warnings
from shutil import copyfile
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
if TYPE_CHECKING:
from ...tokenization_utils_base import TextInput
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''spiece.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''t5-small''': '''https://huggingface.co/t5-small/resolve/main/spiece.model''',
'''t5-base''': '''https://huggingface.co/t5-base/resolve/main/spiece.model''',
'''t5-large''': '''https://huggingface.co/t5-large/resolve/main/spiece.model''',
'''t5-3b''': '''https://huggingface.co/t5-3b/resolve/main/spiece.model''',
'''t5-11b''': '''https://huggingface.co/t5-11b/resolve/main/spiece.model''',
}
}
# TODO(PVP) - this should be removed in Transformers v5
SCREAMING_SNAKE_CASE__ = {
'''t5-small''': 512,
'''t5-base''': 512,
'''t5-large''': 512,
'''t5-3b''': 512,
'''t5-11b''': 512,
}
SCREAMING_SNAKE_CASE__ = '''▁'''
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : int = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : int = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : Tuple = ['''input_ids''', '''attention_mask''']
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Dict="</s>" , SCREAMING_SNAKE_CASE__ : List[Any]="<unk>" , SCREAMING_SNAKE_CASE__ : List[Any]="<pad>" , SCREAMING_SNAKE_CASE__ : List[str]=1_0_0 , SCREAMING_SNAKE_CASE__ : Tuple=None , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , SCREAMING_SNAKE_CASE__ : List[str]=True , **SCREAMING_SNAKE_CASE__ : str , ):
'''simple docstring'''
if extra_ids > 0 and additional_special_tokens is None:
__a : int = [f'''<extra_id_{i}>''' for i in range(SCREAMING_SNAKE_CASE__ )]
elif extra_ids > 0 and additional_special_tokens is not None:
# Check that we have the right number of extra_id special tokens
__a : List[Any] = len(set(filter(lambda SCREAMING_SNAKE_CASE__ : bool('extra_id' in str(SCREAMING_SNAKE_CASE__ ) ) , SCREAMING_SNAKE_CASE__ ) ) )
if extra_tokens != extra_ids:
raise ValueError(
f'''Both extra_ids ({extra_ids}) and additional_special_tokens ({additional_special_tokens}) are'''
' provided to T5Tokenizer. In this case the additional_special_tokens must include the extra_ids'
' tokens' )
if legacy:
logger.warning_once(
f'''You are using the legacy behaviour of the {self.__class__}. This means that tokens that come after special tokens will not be properly handled. We recommend you to'''
' read the related pull request available at https://github.com/huggingface/transformers/pull/24565' )
__a : Union[str, Any] = legacy
__a : List[Any] = {} if sp_model_kwargs is None else sp_model_kwargs
super().__init__(
eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , extra_ids=SCREAMING_SNAKE_CASE__ , additional_special_tokens=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , legacy=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ , )
__a : Optional[int] = vocab_file
__a : List[Any] = extra_ids
__a : Union[str, Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(SCREAMING_SNAKE_CASE__ )
@staticmethod
def __lowerCAmelCase ( SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if pretrained_model_name_or_path in TaTokenizer.max_model_input_sizes:
__a : Union[str, Any] = TaTokenizer.max_model_input_sizes[pretrained_model_name_or_path]
if init_max_model_length is not None and init_max_model_length != max_model_length:
return init_max_model_length
elif init_max_model_length is None:
warnings.warn(
'This tokenizer was incorrectly instantiated with a model max length of'
f''' {deprecated_max_model_length} which will be corrected in Transformers v5.\nFor now, this'''
' behavior is kept to avoid breaking backwards compatibility when padding/encoding with'
' `truncation is True`.\n- Be aware that you SHOULD NOT rely on'
f''' {pretrained_model_name_or_path} automatically truncating your input to'''
f''' {deprecated_max_model_length} when padding/encoding.\n- If you want to encode/pad to sequences'''
f''' longer than {deprecated_max_model_length} you can either instantiate this tokenizer with'''
' `model_max_length` or pass `max_length` when encoding/padding.\n- To avoid this warning, please'
' instantiate this tokenizer with `model_max_length` set to your preferred value.' , SCREAMING_SNAKE_CASE__ , )
return max_model_length
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return self.sp_model.get_piece_size() + self._extra_ids
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : List[Any] = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : bool = False ):
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=SCREAMING_SNAKE_CASE__ , token_ids_a=SCREAMING_SNAKE_CASE__ , already_has_special_tokens=SCREAMING_SNAKE_CASE__ )
# normal case: some special tokens
if token_ids_a is None:
return ([0] * len(SCREAMING_SNAKE_CASE__ )) + [1]
return ([0] * len(SCREAMING_SNAKE_CASE__ )) + [1] + ([0] * len(SCREAMING_SNAKE_CASE__ )) + [1]
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return list(
set(filter(lambda SCREAMING_SNAKE_CASE__ : bool(re.search(r'<extra_id_\d+>' , SCREAMING_SNAKE_CASE__ ) ) is not None , self.additional_special_tokens ) ) )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
return [self._convert_token_to_id(SCREAMING_SNAKE_CASE__ ) for token in self.get_sentinel_tokens()]
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : List[int] ):
'''simple docstring'''
if len(SCREAMING_SNAKE_CASE__ ) > 0 and token_ids[-1] == self.eos_token_id:
warnings.warn(
f'''This sequence already has {self.eos_token}. In future versions this behavior may lead to duplicated'''
' eos tokens being added.' )
return token_ids
else:
return token_ids + [self.eos_token_id]
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
__a : str = [self.eos_token_id]
if token_ids_a is None:
return len(token_ids_a + eos ) * [0]
return len(token_ids_a + eos + token_ids_a + eos ) * [0]
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : List[int] , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None ):
'''simple docstring'''
__a : str = self._add_eos_if_not_present(SCREAMING_SNAKE_CASE__ )
if token_ids_a is None:
return token_ids_a
else:
__a : int = self._add_eos_if_not_present(SCREAMING_SNAKE_CASE__ )
return token_ids_a + token_ids_a
def __getstate__( self : str ):
'''simple docstring'''
__a : Optional[Any] = self.__dict__.copy()
__a : Union[str, Any] = None
return state
def __setstate__( self : Any , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Optional[int] = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : Dict = {}
__a : Any = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(self.vocab_file )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : "TextInput" , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
if not self.legacy:
__a : Tuple = SPIECE_UNDERLINE + text.replace(SCREAMING_SNAKE_CASE__ , ' ' )
return super().tokenize(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
if not self.legacy:
__a : Optional[int] = text.startswith(SCREAMING_SNAKE_CASE__ )
if is_first:
__a : List[Any] = text[1:]
__a : List[Any] = self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
if not self.legacy and not is_first and not text.startswith(' ' ) and tokens[0].startswith(SCREAMING_SNAKE_CASE__ ):
__a : Union[str, Any] = ([tokens[0][1:]] if len(tokens[0] ) > 1 else []) + tokens[1:]
return tokens
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
if token.startswith('<extra_id_' ):
__a : Dict = re.match(r'<extra_id_(\d+)>' , SCREAMING_SNAKE_CASE__ )
__a : List[Any] = int(match.group(1 ) )
return self.vocab_size - num - 1
return self.sp_model.piece_to_id(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if index < self.sp_model.get_piece_size():
__a : Union[str, Any] = self.sp_model.IdToPiece(SCREAMING_SNAKE_CASE__ )
else:
__a : List[str] = f'''<extra_id_{self.vocab_size - 1 - index}>'''
return token
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Tuple = []
__a : List[Any] = ''
__a : Optional[int] = False
for token in tokens:
# make sure that special tokens are not decoded using sentencepiece model
if token in self.all_special_tokens:
if not prev_is_special:
out_string += " "
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ ) + token
__a : str = True
__a : str = []
else:
current_sub_tokens.append(SCREAMING_SNAKE_CASE__ )
__a : Dict = False
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ )
return out_string.strip()
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Optional[int] = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[Any] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 |
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {'''vocab_file''': '''spiece.model'''}
SCREAMING_SNAKE_CASE__ = {
'''vocab_file''': {
'''bert_for_seq_generation''': (
'''https://huggingface.co/google/bert_for_seq_generation_L-24_bbc_encoder/resolve/main/spiece.model'''
),
}
}
SCREAMING_SNAKE_CASE__ = {'''bert_for_seq_generation''': 512}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = VOCAB_FILES_NAMES
__SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_VOCAB_FILES_MAP
__SCREAMING_SNAKE_CASE : Any = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
__SCREAMING_SNAKE_CASE : List[int] = []
__SCREAMING_SNAKE_CASE : int = ['''input_ids''', '''attention_mask''']
def __init__( self : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : List[Any]="<s>" , SCREAMING_SNAKE_CASE__ : Tuple="</s>" , SCREAMING_SNAKE_CASE__ : Any="<unk>" , SCREAMING_SNAKE_CASE__ : int="<pad>" , SCREAMING_SNAKE_CASE__ : List[str]="<::::>" , SCREAMING_SNAKE_CASE__ : Optional[Dict[str, Any]] = None , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Tuple = {} if sp_model_kwargs is None else sp_model_kwargs
# Add extra_ids to the special token list
super().__init__(
bos_token=SCREAMING_SNAKE_CASE__ , eos_token=SCREAMING_SNAKE_CASE__ , unk_token=SCREAMING_SNAKE_CASE__ , pad_token=SCREAMING_SNAKE_CASE__ , sep_token=SCREAMING_SNAKE_CASE__ , sp_model_kwargs=self.sp_model_kwargs , **SCREAMING_SNAKE_CASE__ , )
__a : int = vocab_file
__a : Union[str, Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
return self.sp_model.get_piece_size()
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Dict = {self.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __getstate__( self : Optional[Any] ):
'''simple docstring'''
__a : Union[str, Any] = self.__dict__.copy()
__a : Any = None
return state
def __setstate__( self : int , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
__a : str = d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a : str = {}
__a : List[Any] = spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(self.vocab_file )
def __lowerCAmelCase ( self : Dict , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
return self.sp_model.encode(SCREAMING_SNAKE_CASE__ , out_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
return self.sp_model.piece_to_id(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
__a : int = self.sp_model.IdToPiece(SCREAMING_SNAKE_CASE__ )
return token
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Optional[Any] = []
__a : Optional[int] = ''
for token in tokens:
# make sure that special tokens are not decoded using sentencepiece model
if token in self.all_special_tokens:
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ ) + token
__a : Dict = []
else:
current_sub_tokens.append(SCREAMING_SNAKE_CASE__ )
out_string += self.sp_model.decode(SCREAMING_SNAKE_CASE__ )
return out_string.strip()
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Optional[str] = None ):
'''simple docstring'''
if not os.path.isdir(SCREAMING_SNAKE_CASE__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
__a : Tuple = os.path.join(
SCREAMING_SNAKE_CASE__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(SCREAMING_SNAKE_CASE__ ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , SCREAMING_SNAKE_CASE__ )
elif not os.path.isfile(self.vocab_file ):
with open(SCREAMING_SNAKE_CASE__ , 'wb' ) as fi:
__a : List[str] = self.sp_model.serialized_model_proto()
fi.write(SCREAMING_SNAKE_CASE__ )
return (out_vocab_file,)
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_bridgetower''': [
'''BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''BridgeTowerConfig''',
'''BridgeTowerTextConfig''',
'''BridgeTowerVisionConfig''',
],
'''processing_bridgetower''': ['''BridgeTowerProcessor'''],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''BridgeTowerImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''BridgeTowerForContrastiveLearning''',
'''BridgeTowerForImageAndTextRetrieval''',
'''BridgeTowerForMaskedLM''',
'''BridgeTowerModel''',
'''BridgeTowerPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_bridgetower import (
BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP,
BridgeTowerConfig,
BridgeTowerTextConfig,
BridgeTowerVisionConfig,
)
from .processing_bridgetower import BridgeTowerProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_bridgetower import BridgeTowerImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_bridgetower import (
BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST,
BridgeTowerForContrastiveLearning,
BridgeTowerForImageAndTextRetrieval,
BridgeTowerForMaskedLM,
BridgeTowerModel,
BridgeTowerPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure)
| 47 |
from ..utils import DummyObject, requires_backends
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Dict , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Tuple , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Any = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Tuple , *SCREAMING_SNAKE_CASE__ : List[str] , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : str , *SCREAMING_SNAKE_CASE__ : Any , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Dict , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Dict = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Any , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Any , *SCREAMING_SNAKE_CASE__ : int , **SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : List[str] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : List[Any] , **SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
class _UpperCamelCase( metaclass=__lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[int] = ['''torch''', '''transformers''', '''onnx''']
def __init__( self : Optional[int] , *SCREAMING_SNAKE_CASE__ : Dict , **SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
requires_backends(self , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : Optional[Any] , *SCREAMING_SNAKE_CASE__ : Tuple , **SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
@classmethod
def __lowerCAmelCase ( cls : int , *SCREAMING_SNAKE_CASE__ : Optional[Any] , **SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
requires_backends(cls , ['torch', 'transformers', 'onnx'] )
| 47 | 1 |
import argparse
import csv
import logging
import os
import random
import numpy as np
import torch
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
from tqdm import tqdm, trange
from transformers import (
CONFIG_NAME,
WEIGHTS_NAME,
AdamW,
OpenAIGPTDoubleHeadsModel,
OpenAIGPTTokenizer,
get_linear_schedule_with_warmup,
)
logging.basicConfig(
format='''%(asctime)s - %(levelname)s - %(name)s - %(message)s''', datefmt='''%m/%d/%Y %H:%M:%S''', level=logging.INFO
)
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : List[Any] ):
__a : Any = np.argmax(lowerCamelCase_ , axis=1 )
return np.sum(outputs == labels )
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple ):
with open(lowerCamelCase_ , encoding='utf_8' ) as f:
__a : Optional[Any] = csv.reader(lowerCamelCase_ )
__a : Optional[int] = []
next(lowerCamelCase_ ) # skip the first line
for line in tqdm(lowerCamelCase_ ):
output.append((' '.join(line[1:5] ), line[5], line[6], int(line[-1] ) - 1) )
return output
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : Dict , lowerCamelCase_ : Dict , lowerCamelCase_ : List[Any] , lowerCamelCase_ : str , lowerCamelCase_ : Optional[Any] ):
__a : int = []
for dataset in encoded_datasets:
__a : List[Any] = len(lowerCamelCase_ )
__a : Optional[Any] = np.zeros((n_batch, 2, input_len) , dtype=np.intaa )
__a : Optional[int] = np.zeros((n_batch, 2) , dtype=np.intaa )
__a : str = np.full((n_batch, 2, input_len) , fill_value=-1_0_0 , dtype=np.intaa )
__a : List[Any] = np.zeros((n_batch,) , dtype=np.intaa )
for (
i,
(story, conta, conta, mc_label),
) in enumerate(lowerCamelCase_ ):
__a : str = [start_token] + story[:cap_length] + [delimiter_token] + conta[:cap_length] + [clf_token]
__a : int = [start_token] + story[:cap_length] + [delimiter_token] + conta[:cap_length] + [clf_token]
__a : List[str] = with_conta
__a : Optional[Any] = with_conta
__a : Any = len(lowerCamelCase_ ) - 1
__a : Any = len(lowerCamelCase_ ) - 1
__a : Optional[int] = with_conta
__a : Tuple = with_conta
__a : Optional[int] = mc_label
__a : Union[str, Any] = (input_ids, mc_token_ids, lm_labels, mc_labels)
tensor_datasets.append(tuple(torch.tensor(lowerCamelCase_ ) for t in all_inputs ) )
return tensor_datasets
def UpperCAmelCase__ ( ):
__a : List[str] = argparse.ArgumentParser()
parser.add_argument('--model_name' , type=lowerCamelCase_ , default='openai-gpt' , help='pretrained model name' )
parser.add_argument('--do_train' , action='store_true' , help='Whether to run training.' )
parser.add_argument('--do_eval' , action='store_true' , help='Whether to run eval on the dev set.' )
parser.add_argument(
'--output_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
parser.add_argument('--train_dataset' , type=lowerCamelCase_ , default='' )
parser.add_argument('--eval_dataset' , type=lowerCamelCase_ , default='' )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 )
parser.add_argument('--num_train_epochs' , type=lowerCamelCase_ , default=3 )
parser.add_argument('--train_batch_size' , type=lowerCamelCase_ , default=8 )
parser.add_argument('--eval_batch_size' , type=lowerCamelCase_ , default=1_6 )
parser.add_argument('--adam_epsilon' , default=1e-8 , type=lowerCamelCase_ , help='Epsilon for Adam optimizer.' )
parser.add_argument('--max_grad_norm' , type=lowerCamelCase_ , default=1 )
parser.add_argument(
'--max_steps' , default=-1 , type=lowerCamelCase_ , help=(
'If > 0: set total number of training steps to perform. Override num_train_epochs.'
) , )
parser.add_argument(
'--gradient_accumulation_steps' , type=lowerCamelCase_ , default=1 , help='Number of updates steps to accumulate before performing a backward/update pass.' , )
parser.add_argument('--learning_rate' , type=lowerCamelCase_ , default=6.25e-5 )
parser.add_argument('--warmup_steps' , default=0 , type=lowerCamelCase_ , help='Linear warmup over warmup_steps.' )
parser.add_argument('--lr_schedule' , type=lowerCamelCase_ , default='warmup_linear' )
parser.add_argument('--weight_decay' , type=lowerCamelCase_ , default=0.01 )
parser.add_argument('--lm_coef' , type=lowerCamelCase_ , default=0.9 )
parser.add_argument('--n_valid' , type=lowerCamelCase_ , default=3_7_4 )
parser.add_argument('--server_ip' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
parser.add_argument('--server_port' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
__a : int = parser.parse_args()
print(lowerCamelCase_ )
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print('Waiting for debugger attach' )
ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=lowerCamelCase_ )
ptvsd.wait_for_attach()
random.seed(args.seed )
np.random.seed(args.seed )
torch.manual_seed(args.seed )
torch.cuda.manual_seed_all(args.seed )
__a : Any = torch.device('cuda' if torch.cuda.is_available() else 'cpu' )
__a : Union[str, Any] = torch.cuda.device_count()
logger.info('device: {}, n_gpu {}'.format(lowerCamelCase_ , lowerCamelCase_ ) )
if not args.do_train and not args.do_eval:
raise ValueError('At least one of `do_train` or `do_eval` must be True.' )
if not os.path.exists(args.output_dir ):
os.makedirs(args.output_dir )
# Load tokenizer and model
# This loading functions also add new tokens and embeddings called `special tokens`
# These new embeddings will be fine-tuned on the RocStories dataset
__a : int = ['_start_', '_delimiter_', '_classify_']
__a : Optional[int] = OpenAIGPTTokenizer.from_pretrained(args.model_name )
tokenizer.add_tokens(lowerCamelCase_ )
__a : Optional[int] = tokenizer.convert_tokens_to_ids(lowerCamelCase_ )
__a : int = OpenAIGPTDoubleHeadsModel.from_pretrained(args.model_name )
model.resize_token_embeddings(len(lowerCamelCase_ ) )
model.to(lowerCamelCase_ )
# Load and encode the datasets
def tokenize_and_encode(lowerCamelCase_ : Optional[Any] ):
if isinstance(lowerCamelCase_ , lowerCamelCase_ ):
return tokenizer.convert_tokens_to_ids(tokenizer.tokenize(lowerCamelCase_ ) )
elif isinstance(lowerCamelCase_ , lowerCamelCase_ ):
return obj
return [tokenize_and_encode(lowerCamelCase_ ) for o in obj]
logger.info('Encoding dataset...' )
__a : Tuple = load_rocstories_dataset(args.train_dataset )
__a : List[Any] = load_rocstories_dataset(args.eval_dataset )
__a : str = (train_dataset, eval_dataset)
__a : str = tokenize_and_encode(lowerCamelCase_ )
# Compute the max input length for the Transformer
__a : Any = model.config.n_positions // 2 - 2
__a : Dict = max(
len(story[:max_length] ) + max(len(conta[:max_length] ) , len(conta[:max_length] ) ) + 3
for dataset in encoded_datasets
for story, conta, conta, _ in dataset )
__a : Dict = min(lowerCamelCase_ , model.config.n_positions ) # Max size of input for the pre-trained model
# Prepare inputs tensors and dataloaders
__a : Dict = pre_process_datasets(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , *lowerCamelCase_ )
__a , __a : str = tensor_datasets[0], tensor_datasets[1]
__a : str = TensorDataset(*lowerCamelCase_ )
__a : int = RandomSampler(lowerCamelCase_ )
__a : str = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.train_batch_size )
__a : List[Any] = TensorDataset(*lowerCamelCase_ )
__a : str = SequentialSampler(lowerCamelCase_ )
__a : Dict = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.eval_batch_size )
# Prepare optimizer
if args.do_train:
if args.max_steps > 0:
__a : Optional[Any] = args.max_steps
__a : int = args.max_steps // (len(lowerCamelCase_ ) // args.gradient_accumulation_steps) + 1
else:
__a : List[Any] = len(lowerCamelCase_ ) // args.gradient_accumulation_steps * args.num_train_epochs
__a : Any = list(model.named_parameters() )
__a : Optional[int] = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
__a : int = [
{
'params': [p for n, p in param_optimizer if not any(nd in n for nd in no_decay )],
'weight_decay': args.weight_decay,
},
{'params': [p for n, p in param_optimizer if any(nd in n for nd in no_decay )], 'weight_decay': 0.0},
]
__a : List[str] = AdamW(lowerCamelCase_ , lr=args.learning_rate , eps=args.adam_epsilon )
__a : Optional[Any] = get_linear_schedule_with_warmup(
lowerCamelCase_ , num_warmup_steps=args.warmup_steps , num_training_steps=lowerCamelCase_ )
if args.do_train:
__a , __a , __a : List[Any] = 0, 0, None
model.train()
for _ in trange(int(args.num_train_epochs ) , desc='Epoch' ):
__a : Union[str, Any] = 0
__a : int = 0
__a : str = tqdm(lowerCamelCase_ , desc='Training' )
for step, batch in enumerate(lowerCamelCase_ ):
__a : int = tuple(t.to(lowerCamelCase_ ) for t in batch )
__a , __a , __a , __a : Any = batch
__a : List[Any] = model(lowerCamelCase_ , mc_token_ids=lowerCamelCase_ , lm_labels=lowerCamelCase_ , mc_labels=lowerCamelCase_ )
__a : Union[str, Any] = args.lm_coef * losses[0] + losses[1]
loss.backward()
optimizer.step()
scheduler.step()
optimizer.zero_grad()
tr_loss += loss.item()
__a : Optional[int] = (
loss.item() if exp_average_loss is None else 0.7 * exp_average_loss + 0.3 * loss.item()
)
nb_tr_steps += 1
__a : Optional[Any] = 'Training loss: {:.2e} lr: {:.2e}'.format(lowerCamelCase_ , scheduler.get_lr()[0] )
# Save a trained model
if args.do_train:
# Save a trained model, configuration and tokenizer
__a : List[str] = model.module if hasattr(lowerCamelCase_ , 'module' ) else model # Only save the model itself
# If we save using the predefined names, we can load using `from_pretrained`
__a : Union[str, Any] = os.path.join(args.output_dir , lowerCamelCase_ )
__a : List[Any] = os.path.join(args.output_dir , lowerCamelCase_ )
torch.save(model_to_save.state_dict() , lowerCamelCase_ )
model_to_save.config.to_json_file(lowerCamelCase_ )
tokenizer.save_vocabulary(args.output_dir )
# Load a trained model and vocabulary that you have fine-tuned
__a : str = OpenAIGPTDoubleHeadsModel.from_pretrained(args.output_dir )
__a : Tuple = OpenAIGPTTokenizer.from_pretrained(args.output_dir )
model.to(lowerCamelCase_ )
if args.do_eval:
model.eval()
__a , __a : Union[str, Any] = 0, 0
__a , __a : List[Any] = 0, 0
for batch in tqdm(lowerCamelCase_ , desc='Evaluating' ):
__a : int = tuple(t.to(lowerCamelCase_ ) for t in batch )
__a , __a , __a , __a : Dict = batch
with torch.no_grad():
__a , __a , __a , __a : Dict = model(
lowerCamelCase_ , mc_token_ids=lowerCamelCase_ , lm_labels=lowerCamelCase_ , mc_labels=lowerCamelCase_ )
__a : int = mc_logits.detach().cpu().numpy()
__a : List[str] = mc_labels.to('cpu' ).numpy()
__a : str = accuracy(lowerCamelCase_ , lowerCamelCase_ )
eval_loss += mc_loss.mean().item()
eval_accuracy += tmp_eval_accuracy
nb_eval_examples += input_ids.size(0 )
nb_eval_steps += 1
__a : Optional[int] = eval_loss / nb_eval_steps
__a : Dict = eval_accuracy / nb_eval_examples
__a : Dict = tr_loss / nb_tr_steps if args.do_train else None
__a : Dict = {'eval_loss': eval_loss, 'eval_accuracy': eval_accuracy, 'train_loss': train_loss}
__a : str = os.path.join(args.output_dir , 'eval_results.txt' )
with open(lowerCamelCase_ , 'w' ) as writer:
logger.info('***** Eval results *****' )
for key in sorted(result.keys() ):
logger.info(' %s = %s' , lowerCamelCase_ , str(result[key] ) )
writer.write('%s = %s\n' % (key, str(result[key] )) )
if __name__ == "__main__":
main()
| 47 |
import math
from datetime import datetime, timedelta
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
__a : Union[str, Any] = year % 1_9
__a : int = year % 4
__a : Optional[int] = year % 7
__a : Dict = math.floor(year / 1_0_0 )
__a : Optional[Any] = math.floor((1_3 + 8 * leap_day_inhibits) / 2_5 )
__a : Union[str, Any] = leap_day_inhibits / 4
__a : str = (
1_5 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 3_0
__a : Union[str, Any] = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
__a : List[Any] = (1_9 * metonic_cycle + secular_moon_shift) % 3_0
# PHM -> Paschal Full Moon
__a : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 2_9 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_9 )
elif days_to_add == 2_8 and days_from_phm_to_sunday == 6:
return datetime(lowerCamelCase_ , 4 , 1_8 )
else:
return datetime(lowerCamelCase_ , 3 , 2_2 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
SCREAMING_SNAKE_CASE__ = '''will be''' if year > datetime.now().year else '''was'''
print(F"Easter in {year} {tense} {gauss_easter(year)}")
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
if len(lowerCamelCase_ ) != len(lowerCamelCase_ ):
raise ValueError('String lengths must match!' )
__a : List[Any] = 0
for chara, chara in zip(lowerCamelCase_ , lowerCamelCase_ ):
if chara != chara:
count += 1
return count
if __name__ == "__main__":
import doctest
doctest.testmod()
| 47 |
from typing import List, Optional, Union
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''huggingface/informer-tourism-monthly''': (
'''https://huggingface.co/huggingface/informer-tourism-monthly/resolve/main/config.json'''
),
# See all Informer models at https://huggingface.co/models?filter=informer
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : List[Any] = '''informer'''
__SCREAMING_SNAKE_CASE : List[Any] = {
'''hidden_size''': '''d_model''',
'''num_attention_heads''': '''encoder_attention_heads''',
'''num_hidden_layers''': '''encoder_layers''',
}
def __init__( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , SCREAMING_SNAKE_CASE__ : str = "student_t" , SCREAMING_SNAKE_CASE__ : str = "nll" , SCREAMING_SNAKE_CASE__ : int = 1 , SCREAMING_SNAKE_CASE__ : List[int] = None , SCREAMING_SNAKE_CASE__ : Optional[Union[str, bool]] = "mean" , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : int = 0 , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : Optional[List[int]] = None , SCREAMING_SNAKE_CASE__ : int = 6_4 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 3_2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : int = 2 , SCREAMING_SNAKE_CASE__ : bool = True , SCREAMING_SNAKE_CASE__ : str = "gelu" , SCREAMING_SNAKE_CASE__ : float = 0.05 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : float = 0.1 , SCREAMING_SNAKE_CASE__ : int = 1_0_0 , SCREAMING_SNAKE_CASE__ : float = 0.02 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=True , SCREAMING_SNAKE_CASE__ : str = "prob" , SCREAMING_SNAKE_CASE__ : int = 5 , SCREAMING_SNAKE_CASE__ : bool = True , **SCREAMING_SNAKE_CASE__ : Tuple , ):
'''simple docstring'''
__a : Dict = prediction_length
__a : Tuple = context_length or prediction_length
__a : Tuple = distribution_output
__a : Tuple = loss
__a : str = input_size
__a : Dict = num_time_features
__a : Optional[int] = lags_sequence if lags_sequence is not None else [1, 2, 3, 4, 5, 6, 7]
__a : str = scaling
__a : Tuple = num_dynamic_real_features
__a : int = num_static_real_features
__a : Dict = num_static_categorical_features
# set cardinality
if cardinality and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The cardinality should be a list of the same length as `num_static_categorical_features`' )
__a : Optional[Any] = cardinality
else:
__a : Optional[int] = [0]
# set embedding_dimension
if embedding_dimension and num_static_categorical_features > 0:
if len(SCREAMING_SNAKE_CASE__ ) != num_static_categorical_features:
raise ValueError(
'The embedding dimension should be a list of the same length as `num_static_categorical_features`' )
__a : int = embedding_dimension
else:
__a : List[Any] = [min(5_0 , (cat + 1) // 2 ) for cat in self.cardinality]
__a : int = num_parallel_samples
# Transformer architecture configuration
__a : str = input_size * len(self.lags_sequence ) + self._number_of_features
__a : Optional[int] = d_model
__a : Union[str, Any] = encoder_attention_heads
__a : int = decoder_attention_heads
__a : Any = encoder_ffn_dim
__a : Union[str, Any] = decoder_ffn_dim
__a : List[Any] = encoder_layers
__a : Optional[int] = decoder_layers
__a : int = dropout
__a : Optional[Any] = attention_dropout
__a : Dict = activation_dropout
__a : Union[str, Any] = encoder_layerdrop
__a : Optional[int] = decoder_layerdrop
__a : List[str] = activation_function
__a : str = init_std
__a : Optional[int] = use_cache
# Informer
__a : Union[str, Any] = attention_type
__a : str = sampling_factor
__a : Dict = distil
super().__init__(is_encoder_decoder=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
return (
sum(self.embedding_dimension )
+ self.num_dynamic_real_features
+ self.num_time_features
+ self.num_static_real_features
+ self.input_size * 2 # the log1p(abs(loc)) and log(scale) features
)
| 47 | 1 |
from collections import OrderedDict
from ...utils import logging
from .auto_factory import _BaseAutoModelClass, _LazyAutoMapping, auto_class_update
from .configuration_auto import CONFIG_MAPPING_NAMES
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Base model mapping
('''albert''', '''FlaxAlbertModel'''),
('''bart''', '''FlaxBartModel'''),
('''beit''', '''FlaxBeitModel'''),
('''bert''', '''FlaxBertModel'''),
('''big_bird''', '''FlaxBigBirdModel'''),
('''blenderbot''', '''FlaxBlenderbotModel'''),
('''blenderbot-small''', '''FlaxBlenderbotSmallModel'''),
('''clip''', '''FlaxCLIPModel'''),
('''distilbert''', '''FlaxDistilBertModel'''),
('''electra''', '''FlaxElectraModel'''),
('''gpt-sw3''', '''FlaxGPT2Model'''),
('''gpt2''', '''FlaxGPT2Model'''),
('''gpt_neo''', '''FlaxGPTNeoModel'''),
('''gptj''', '''FlaxGPTJModel'''),
('''longt5''', '''FlaxLongT5Model'''),
('''marian''', '''FlaxMarianModel'''),
('''mbart''', '''FlaxMBartModel'''),
('''mt5''', '''FlaxMT5Model'''),
('''opt''', '''FlaxOPTModel'''),
('''pegasus''', '''FlaxPegasusModel'''),
('''regnet''', '''FlaxRegNetModel'''),
('''resnet''', '''FlaxResNetModel'''),
('''roberta''', '''FlaxRobertaModel'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormModel'''),
('''roformer''', '''FlaxRoFormerModel'''),
('''t5''', '''FlaxT5Model'''),
('''vision-text-dual-encoder''', '''FlaxVisionTextDualEncoderModel'''),
('''vit''', '''FlaxViTModel'''),
('''wav2vec2''', '''FlaxWav2Vec2Model'''),
('''whisper''', '''FlaxWhisperModel'''),
('''xglm''', '''FlaxXGLMModel'''),
('''xlm-roberta''', '''FlaxXLMRobertaModel'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for pre-training mapping
('''albert''', '''FlaxAlbertForPreTraining'''),
('''bart''', '''FlaxBartForConditionalGeneration'''),
('''bert''', '''FlaxBertForPreTraining'''),
('''big_bird''', '''FlaxBigBirdForPreTraining'''),
('''electra''', '''FlaxElectraForPreTraining'''),
('''longt5''', '''FlaxLongT5ForConditionalGeneration'''),
('''mbart''', '''FlaxMBartForConditionalGeneration'''),
('''mt5''', '''FlaxMT5ForConditionalGeneration'''),
('''roberta''', '''FlaxRobertaForMaskedLM'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormForMaskedLM'''),
('''roformer''', '''FlaxRoFormerForMaskedLM'''),
('''t5''', '''FlaxT5ForConditionalGeneration'''),
('''wav2vec2''', '''FlaxWav2Vec2ForPreTraining'''),
('''whisper''', '''FlaxWhisperForConditionalGeneration'''),
('''xlm-roberta''', '''FlaxXLMRobertaForMaskedLM'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Masked LM mapping
('''albert''', '''FlaxAlbertForMaskedLM'''),
('''bart''', '''FlaxBartForConditionalGeneration'''),
('''bert''', '''FlaxBertForMaskedLM'''),
('''big_bird''', '''FlaxBigBirdForMaskedLM'''),
('''distilbert''', '''FlaxDistilBertForMaskedLM'''),
('''electra''', '''FlaxElectraForMaskedLM'''),
('''mbart''', '''FlaxMBartForConditionalGeneration'''),
('''roberta''', '''FlaxRobertaForMaskedLM'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormForMaskedLM'''),
('''roformer''', '''FlaxRoFormerForMaskedLM'''),
('''xlm-roberta''', '''FlaxXLMRobertaForMaskedLM'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Seq2Seq Causal LM mapping
('''bart''', '''FlaxBartForConditionalGeneration'''),
('''blenderbot''', '''FlaxBlenderbotForConditionalGeneration'''),
('''blenderbot-small''', '''FlaxBlenderbotSmallForConditionalGeneration'''),
('''encoder-decoder''', '''FlaxEncoderDecoderModel'''),
('''longt5''', '''FlaxLongT5ForConditionalGeneration'''),
('''marian''', '''FlaxMarianMTModel'''),
('''mbart''', '''FlaxMBartForConditionalGeneration'''),
('''mt5''', '''FlaxMT5ForConditionalGeneration'''),
('''pegasus''', '''FlaxPegasusForConditionalGeneration'''),
('''t5''', '''FlaxT5ForConditionalGeneration'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Image-classsification
('''beit''', '''FlaxBeitForImageClassification'''),
('''regnet''', '''FlaxRegNetForImageClassification'''),
('''resnet''', '''FlaxResNetForImageClassification'''),
('''vit''', '''FlaxViTForImageClassification'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
('''vision-encoder-decoder''', '''FlaxVisionEncoderDecoderModel'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Causal LM mapping
('''bart''', '''FlaxBartForCausalLM'''),
('''bert''', '''FlaxBertForCausalLM'''),
('''big_bird''', '''FlaxBigBirdForCausalLM'''),
('''electra''', '''FlaxElectraForCausalLM'''),
('''gpt-sw3''', '''FlaxGPT2LMHeadModel'''),
('''gpt2''', '''FlaxGPT2LMHeadModel'''),
('''gpt_neo''', '''FlaxGPTNeoForCausalLM'''),
('''gptj''', '''FlaxGPTJForCausalLM'''),
('''opt''', '''FlaxOPTForCausalLM'''),
('''roberta''', '''FlaxRobertaForCausalLM'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormForCausalLM'''),
('''xglm''', '''FlaxXGLMForCausalLM'''),
('''xlm-roberta''', '''FlaxXLMRobertaForCausalLM'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Sequence Classification mapping
('''albert''', '''FlaxAlbertForSequenceClassification'''),
('''bart''', '''FlaxBartForSequenceClassification'''),
('''bert''', '''FlaxBertForSequenceClassification'''),
('''big_bird''', '''FlaxBigBirdForSequenceClassification'''),
('''distilbert''', '''FlaxDistilBertForSequenceClassification'''),
('''electra''', '''FlaxElectraForSequenceClassification'''),
('''mbart''', '''FlaxMBartForSequenceClassification'''),
('''roberta''', '''FlaxRobertaForSequenceClassification'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormForSequenceClassification'''),
('''roformer''', '''FlaxRoFormerForSequenceClassification'''),
('''xlm-roberta''', '''FlaxXLMRobertaForSequenceClassification'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Question Answering mapping
('''albert''', '''FlaxAlbertForQuestionAnswering'''),
('''bart''', '''FlaxBartForQuestionAnswering'''),
('''bert''', '''FlaxBertForQuestionAnswering'''),
('''big_bird''', '''FlaxBigBirdForQuestionAnswering'''),
('''distilbert''', '''FlaxDistilBertForQuestionAnswering'''),
('''electra''', '''FlaxElectraForQuestionAnswering'''),
('''mbart''', '''FlaxMBartForQuestionAnswering'''),
('''roberta''', '''FlaxRobertaForQuestionAnswering'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormForQuestionAnswering'''),
('''roformer''', '''FlaxRoFormerForQuestionAnswering'''),
('''xlm-roberta''', '''FlaxXLMRobertaForQuestionAnswering'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Token Classification mapping
('''albert''', '''FlaxAlbertForTokenClassification'''),
('''bert''', '''FlaxBertForTokenClassification'''),
('''big_bird''', '''FlaxBigBirdForTokenClassification'''),
('''distilbert''', '''FlaxDistilBertForTokenClassification'''),
('''electra''', '''FlaxElectraForTokenClassification'''),
('''roberta''', '''FlaxRobertaForTokenClassification'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormForTokenClassification'''),
('''roformer''', '''FlaxRoFormerForTokenClassification'''),
('''xlm-roberta''', '''FlaxXLMRobertaForTokenClassification'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
# Model for Multiple Choice mapping
('''albert''', '''FlaxAlbertForMultipleChoice'''),
('''bert''', '''FlaxBertForMultipleChoice'''),
('''big_bird''', '''FlaxBigBirdForMultipleChoice'''),
('''distilbert''', '''FlaxDistilBertForMultipleChoice'''),
('''electra''', '''FlaxElectraForMultipleChoice'''),
('''roberta''', '''FlaxRobertaForMultipleChoice'''),
('''roberta-prelayernorm''', '''FlaxRobertaPreLayerNormForMultipleChoice'''),
('''roformer''', '''FlaxRoFormerForMultipleChoice'''),
('''xlm-roberta''', '''FlaxXLMRobertaForMultipleChoice'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
('''bert''', '''FlaxBertForNextSentencePrediction'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
('''speech-encoder-decoder''', '''FlaxSpeechEncoderDecoderModel'''),
('''whisper''', '''FlaxWhisperForConditionalGeneration'''),
]
)
SCREAMING_SNAKE_CASE__ = OrderedDict(
[
('''whisper''', '''FlaxWhisperForAudioClassification'''),
]
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(CONFIG_MAPPING_NAMES, FLAX_MODEL_MAPPING_NAMES)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_PRETRAINING_MAPPING_NAMES)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_MASKED_LM_MAPPING_NAMES)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_VISION_2_SEQ_MAPPING_NAMES)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_CAUSAL_LM_MAPPING_NAMES)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_MULTIPLE_CHOICE_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES
)
SCREAMING_SNAKE_CASE__ = _LazyAutoMapping(
CONFIG_MAPPING_NAMES, FLAX_MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMES
)
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : str = FLAX_MODEL_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(FlaxAutoModel)
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : Tuple = FLAX_MODEL_FOR_PRETRAINING_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(FlaxAutoModelForPreTraining, head_doc='''pretraining''')
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : Tuple = FLAX_MODEL_FOR_CAUSAL_LM_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(FlaxAutoModelForCausalLM, head_doc='''causal language modeling''')
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : Any = FLAX_MODEL_FOR_MASKED_LM_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(FlaxAutoModelForMaskedLM, head_doc='''masked language modeling''')
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : List[str] = FLAX_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(
FlaxAutoModelForSeqaSeqLM, head_doc='''sequence-to-sequence language modeling''', checkpoint_for_example='''t5-base'''
)
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : int = FLAX_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(
FlaxAutoModelForSequenceClassification, head_doc='''sequence classification'''
)
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : int = FLAX_MODEL_FOR_QUESTION_ANSWERING_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(FlaxAutoModelForQuestionAnswering, head_doc='''question answering''')
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : Any = FLAX_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(
FlaxAutoModelForTokenClassification, head_doc='''token classification'''
)
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : Any = FLAX_MODEL_FOR_MULTIPLE_CHOICE_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(FlaxAutoModelForMultipleChoice, head_doc='''multiple choice''')
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : Any = FLAX_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(
FlaxAutoModelForNextSentencePrediction, head_doc='''next sentence prediction'''
)
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : str = FLAX_MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(
FlaxAutoModelForImageClassification, head_doc='''image classification'''
)
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : List[str] = FLAX_MODEL_FOR_VISION_2_SEQ_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(FlaxAutoModelForVisionaSeq, head_doc='''vision-to-text modeling''')
class _UpperCamelCase( _BaseAutoModelClass ):
__SCREAMING_SNAKE_CASE : List[Any] = FLAX_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING
SCREAMING_SNAKE_CASE__ = auto_class_update(
FlaxAutoModelForSpeechSeqaSeq, head_doc='''sequence-to-sequence speech-to-text modeling'''
)
| 47 |
import torch
from diffusers import DDIMParallelScheduler
from .test_schedulers import SchedulerCommonTest
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : int = (DDIMParallelScheduler,)
__SCREAMING_SNAKE_CASE : Union[str, Any] = (('''eta''', 0.0), ('''num_inference_steps''', 50))
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : List[Any] = {
'num_train_timesteps': 1_0_0_0,
'beta_start': 0.0_001,
'beta_end': 0.02,
'beta_schedule': 'linear',
'clip_sample': True,
}
config.update(**SCREAMING_SNAKE_CASE__ )
return config
def __lowerCAmelCase ( self : str , **SCREAMING_SNAKE_CASE__ : Optional[int] ):
'''simple docstring'''
__a : Tuple = self.scheduler_classes[0]
__a : Optional[Any] = self.get_scheduler_config(**SCREAMING_SNAKE_CASE__ )
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : List[str] = 1_0, 0.0
__a : Dict = self.dummy_model()
__a : str = self.dummy_sample_deter
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
for t in scheduler.timesteps:
__a : str = model(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[str] = scheduler.step(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ).prev_sample
return sample
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
for timesteps in [1_0_0, 5_0_0, 1_0_0_0]:
self.check_over_configs(num_train_timesteps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
for steps_offset in [0, 1]:
self.check_over_configs(steps_offset=SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config(steps_offset=1 )
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
scheduler.set_timesteps(5 )
assert torch.equal(scheduler.timesteps , torch.LongTensor([8_0_1, 6_0_1, 4_0_1, 2_0_1, 1] ) )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for beta_start, beta_end in zip([0.0_001, 0.001, 0.01, 0.1] , [0.002, 0.02, 0.2, 2] ):
self.check_over_configs(beta_start=SCREAMING_SNAKE_CASE__ , beta_end=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for schedule in ["linear", "squaredcos_cap_v2"]:
self.check_over_configs(beta_schedule=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(prediction_type=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for clip_sample in [True, False]:
self.check_over_configs(clip_sample=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for timestep_spacing in ["trailing", "leading"]:
self.check_over_configs(timestep_spacing=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
for rescale_betas_zero_snr in [True, False]:
self.check_over_configs(rescale_betas_zero_snr=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
self.check_over_configs(thresholding=SCREAMING_SNAKE_CASE__ )
for threshold in [0.5, 1.0, 2.0]:
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(
thresholding=SCREAMING_SNAKE_CASE__ , prediction_type=SCREAMING_SNAKE_CASE__ , sample_max_value=SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for t in [1, 1_0, 4_9]:
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
for t, num_inference_steps in zip([1, 1_0, 5_0] , [1_0, 5_0, 5_0_0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , num_inference_steps=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for t, eta in zip([1, 1_0, 4_9] , [0.0, 0.5, 1.0] ):
self.check_over_forward(time_step=SCREAMING_SNAKE_CASE__ , eta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : Union[str, Any] = self.get_scheduler_config()
__a : Any = scheduler_class(**SCREAMING_SNAKE_CASE__ )
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_2_0 , 4_0_0 ) - 0.14_771 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_8_0 , 9_6_0 ) - 0.32_460 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(0 , 0 ) - 0.0 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(4_8_7 , 4_8_6 ) - 0.00_979 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(9_9_9 , 9_9_8 ) - 0.02 ) ) < 1e-5
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : List[str] = self.scheduler_classes[0]
__a : List[str] = self.get_scheduler_config()
__a : Optional[Any] = scheduler_class(**SCREAMING_SNAKE_CASE__ )
__a , __a : Any = 1_0, 0.0
scheduler.set_timesteps(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = self.dummy_model()
__a : int = self.dummy_sample_deter
__a : List[Any] = self.dummy_sample_deter + 0.1
__a : List[str] = self.dummy_sample_deter - 0.1
__a : Optional[Any] = samplea.shape[0]
__a : Optional[Any] = torch.stack([samplea, samplea, samplea] , dim=0 )
__a : Union[str, Any] = torch.arange(SCREAMING_SNAKE_CASE__ )[0:3, None].repeat(1 , SCREAMING_SNAKE_CASE__ )
__a : int = model(samples.flatten(0 , 1 ) , timesteps.flatten(0 , 1 ) )
__a : int = scheduler.batch_step_no_noise(SCREAMING_SNAKE_CASE__ , timesteps.flatten(0 , 1 ) , samples.flatten(0 , 1 ) , SCREAMING_SNAKE_CASE__ )
__a : Dict = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 1_147.7_904 ) < 1e-2
assert abs(result_mean.item() - 0.4_982 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : List[str] = self.full_loop()
__a : Tuple = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : int = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 172.0_067 ) < 1e-2
assert abs(result_mean.item() - 0.223_967 ) < 1e-3
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Optional[int] = self.full_loop(prediction_type='v_prediction' )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Union[str, Any] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 52.5_302 ) < 1e-2
assert abs(result_mean.item() - 0.0_684 ) < 1e-3
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Union[str, Any] = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : Optional[int] = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Optional[int] = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.8_295 ) < 1e-2
assert abs(result_mean.item() - 0.1_951 ) < 1e-3
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Dict = self.full_loop(set_alpha_to_one=SCREAMING_SNAKE_CASE__ , beta_start=0.01 )
__a : str = torch.sum(torch.abs(SCREAMING_SNAKE_CASE__ ) )
__a : Tuple = torch.mean(torch.abs(SCREAMING_SNAKE_CASE__ ) )
assert abs(result_sum.item() - 149.0_784 ) < 1e-2
assert abs(result_mean.item() - 0.1_941 ) < 1e-3
| 47 | 1 |
import shutil
import tempfile
import unittest
from transformers import SPIECE_UNDERLINE, BatchEncoding, MBartTokenizer, MBartTokenizerFast, is_torch_available
from transformers.testing_utils import (
get_tests_dir,
nested_simplify,
require_sentencepiece,
require_tokenizers,
require_torch,
)
from ...test_tokenization_common import TokenizerTesterMixin
SCREAMING_SNAKE_CASE__ = get_tests_dir('''fixtures/test_sentencepiece.model''')
if is_torch_available():
from transformers.models.mbart.modeling_mbart import shift_tokens_right
SCREAMING_SNAKE_CASE__ = 25_0004
SCREAMING_SNAKE_CASE__ = 25_0020
@require_sentencepiece
@require_tokenizers
class _UpperCamelCase( __lowerCamelCase , unittest.TestCase ):
__SCREAMING_SNAKE_CASE : List[str] = MBartTokenizer
__SCREAMING_SNAKE_CASE : Tuple = MBartTokenizerFast
__SCREAMING_SNAKE_CASE : Optional[int] = True
__SCREAMING_SNAKE_CASE : Tuple = True
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
super().setUp()
# We have a SentencePiece fixture for testing
__a : Any = MBartTokenizer(SCREAMING_SNAKE_CASE__ , keep_accents=SCREAMING_SNAKE_CASE__ )
tokenizer.save_pretrained(self.tmpdirname )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : List[Any] = MBartTokenizer(SCREAMING_SNAKE_CASE__ , keep_accents=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = tokenizer.tokenize('This is a test' )
self.assertListEqual(SCREAMING_SNAKE_CASE__ , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(SCREAMING_SNAKE_CASE__ ) , [value + tokenizer.fairseq_offset for value in [2_8_5, 4_6, 1_0, 1_7_0, 3_8_2]] , )
__a : Optional[int] = tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
SCREAMING_SNAKE_CASE__ , [
SPIECE_UNDERLINE + 'I',
SPIECE_UNDERLINE + 'was',
SPIECE_UNDERLINE + 'b',
'or',
'n',
SPIECE_UNDERLINE + 'in',
SPIECE_UNDERLINE + '',
'9',
'2',
'0',
'0',
'0',
',',
SPIECE_UNDERLINE + 'and',
SPIECE_UNDERLINE + 'this',
SPIECE_UNDERLINE + 'is',
SPIECE_UNDERLINE + 'f',
'al',
's',
'é',
'.',
] , )
__a : Optional[Any] = tokenizer.convert_tokens_to_ids(SCREAMING_SNAKE_CASE__ )
self.assertListEqual(
SCREAMING_SNAKE_CASE__ , [
value + tokenizer.fairseq_offset
for value in [8, 2_1, 8_4, 5_5, 2_4, 1_9, 7, 2, 6_0_2, 3_4_7, 3_4_7, 3_4_7, 3, 1_2, 6_6, 4_6, 7_2, 8_0, 6, 2, 4]
# ^ unk: 2 + 1 = 3 unk: 2 + 1 = 3 ^
] , )
__a : List[str] = tokenizer.convert_ids_to_tokens(SCREAMING_SNAKE_CASE__ )
self.assertListEqual(
SCREAMING_SNAKE_CASE__ , [
SPIECE_UNDERLINE + 'I',
SPIECE_UNDERLINE + 'was',
SPIECE_UNDERLINE + 'b',
'or',
'n',
SPIECE_UNDERLINE + 'in',
SPIECE_UNDERLINE + '',
'<unk>',
'2',
'0',
'0',
'0',
',',
SPIECE_UNDERLINE + 'and',
SPIECE_UNDERLINE + 'this',
SPIECE_UNDERLINE + 'is',
SPIECE_UNDERLINE + 'f',
'al',
's',
'<unk>',
'.',
] , )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
if not self.test_slow_tokenizer:
# as we don't have a slow version, we can't compare the outputs between slow and fast versions
return
__a : Tuple = (self.rust_tokenizer_class, 'hf-internal-testing/tiny-random-mbart', {})
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
__a : Tuple = self.rust_tokenizer_class.from_pretrained(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Dict = self.tokenizer_class.from_pretrained(SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : str = tempfile.mkdtemp()
__a : Any = tokenizer_r.save_pretrained(SCREAMING_SNAKE_CASE__ )
__a : Dict = tokenizer_p.save_pretrained(SCREAMING_SNAKE_CASE__ )
# Checks it save with the same files + the tokenizer.json file for the fast one
self.assertTrue(any('tokenizer.json' in f for f in tokenizer_r_files ) )
__a : List[str] = tuple(f for f in tokenizer_r_files if 'tokenizer.json' not in f )
self.assertSequenceEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# Checks everything loads correctly in the same way
__a : Dict = tokenizer_r.from_pretrained(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = tokenizer_p.from_pretrained(SCREAMING_SNAKE_CASE__ )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
# self.assertEqual(getattr(tokenizer_rp, key), getattr(tokenizer_pp, key))
# self.assertEqual(getattr(tokenizer_rp, key + "_id"), getattr(tokenizer_pp, key + "_id"))
shutil.rmtree(SCREAMING_SNAKE_CASE__ )
# Save tokenizer rust, legacy_format=True
__a : Optional[Any] = tempfile.mkdtemp()
__a : Optional[Any] = tokenizer_r.save_pretrained(SCREAMING_SNAKE_CASE__ , legacy_format=SCREAMING_SNAKE_CASE__ )
__a : int = tokenizer_p.save_pretrained(SCREAMING_SNAKE_CASE__ )
# Checks it save with the same files
self.assertSequenceEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# Checks everything loads correctly in the same way
__a : Optional[Any] = tokenizer_r.from_pretrained(SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = tokenizer_p.from_pretrained(SCREAMING_SNAKE_CASE__ )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
shutil.rmtree(SCREAMING_SNAKE_CASE__ )
# Save tokenizer rust, legacy_format=False
__a : Tuple = tempfile.mkdtemp()
__a : List[Any] = tokenizer_r.save_pretrained(SCREAMING_SNAKE_CASE__ , legacy_format=SCREAMING_SNAKE_CASE__ )
__a : Tuple = tokenizer_p.save_pretrained(SCREAMING_SNAKE_CASE__ )
# Checks it saved the tokenizer.json file
self.assertTrue(any('tokenizer.json' in f for f in tokenizer_r_files ) )
# Checks everything loads correctly in the same way
__a : Union[str, Any] = tokenizer_r.from_pretrained(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = tokenizer_p.from_pretrained(SCREAMING_SNAKE_CASE__ )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
shutil.rmtree(SCREAMING_SNAKE_CASE__ )
@require_torch
@require_sentencepiece
@require_tokenizers
class _UpperCamelCase( unittest.TestCase ):
__SCREAMING_SNAKE_CASE : List[Any] = '''facebook/mbart-large-en-ro'''
__SCREAMING_SNAKE_CASE : Optional[Any] = [
''' UN Chief Says There Is No Military Solution in Syria''',
''' Secretary-General Ban Ki-moon says his response to Russia\'s stepped up military support for Syria is that "there is no military solution" to the nearly five-year conflict and more weapons will only worsen the violence and misery for millions of people.''',
]
__SCREAMING_SNAKE_CASE : Dict = [
'''Şeful ONU declară că nu există o soluţie militară în Siria''',
'''Secretarul General Ban Ki-moon declară că răspunsul său la intensificarea sprijinului militar al Rusiei'''
''' pentru Siria este că "nu există o soluţie militară" la conflictul de aproape cinci ani şi că noi arme nu vor'''
''' face decât să înrăutăţească violenţele şi mizeria pentru milioane de oameni.''',
]
__SCREAMING_SNAKE_CASE : Optional[int] = [8274, 12_7873, 2_5916, 7, 8622, 2071, 438, 6_7485, 53, 18_7895, 23, 5_1712, 2, EN_CODE]
@classmethod
def __lowerCAmelCase ( cls : List[str] ):
'''simple docstring'''
__a : MBartTokenizer = MBartTokenizer.from_pretrained(
cls.checkpoint_name , src_lang='en_XX' , tgt_lang='ro_RO' )
__a : Dict = 1
return cls
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ar_AR'] , 2_5_0_0_0_1 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['en_EN'] , 2_5_0_0_0_4 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ro_RO'] , 2_5_0_0_2_0 )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
__a : Dict = self.tokenizer.batch_encode_plus(self.src_text ).input_ids[0]
self.assertListEqual(self.expected_src_tokens , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
self.assertIn(SCREAMING_SNAKE_CASE__ , self.tokenizer.all_special_ids )
__a : List[str] = [RO_CODE, 8_8_4, 9_0_1_9, 9_6, 9, 9_1_6, 8_6_7_9_2, 3_6, 1_8_7_4_3, 1_5_5_9_6, 5, 2]
__a : Union[str, Any] = self.tokenizer.decode(SCREAMING_SNAKE_CASE__ , skip_special_tokens=SCREAMING_SNAKE_CASE__ )
__a : Tuple = self.tokenizer.decode(generated_ids[1:] , skip_special_tokens=SCREAMING_SNAKE_CASE__ )
self.assertEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertNotIn(self.tokenizer.eos_token , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : Any = ['this is gunna be a long sentence ' * 2_0]
assert isinstance(src_text[0] , SCREAMING_SNAKE_CASE__ )
__a : int = 1_0
__a : Dict = self.tokenizer(SCREAMING_SNAKE_CASE__ , max_length=SCREAMING_SNAKE_CASE__ , truncation=SCREAMING_SNAKE_CASE__ ).input_ids[0]
self.assertEqual(ids[-2] , 2 )
self.assertEqual(ids[-1] , SCREAMING_SNAKE_CASE__ )
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
self.assertListEqual(self.tokenizer.convert_tokens_to_ids(['<mask>', 'ar_AR'] ) , [2_5_0_0_2_6, 2_5_0_0_0_1] )
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
__a : Tuple = tempfile.mkdtemp()
__a : Optional[Any] = self.tokenizer.fairseq_tokens_to_ids
self.tokenizer.save_pretrained(SCREAMING_SNAKE_CASE__ )
__a : List[Any] = MBartTokenizer.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertDictEqual(new_tok.fairseq_tokens_to_ids , SCREAMING_SNAKE_CASE__ )
@require_torch
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
__a : str = self.tokenizer(self.src_text , text_target=self.tgt_text , padding=SCREAMING_SNAKE_CASE__ , return_tensors='pt' )
__a : Any = shift_tokens_right(batch['labels'] , self.tokenizer.pad_token_id )
# fairseq batch: https://gist.github.com/sshleifer/cba08bc2109361a74ac3760a7e30e4f4
assert batch.input_ids[1][-2:].tolist() == [2, EN_CODE]
assert batch.decoder_input_ids[1][0].tolist() == RO_CODE
assert batch.decoder_input_ids[1][-1] == 2
assert batch.labels[1][-2:].tolist() == [2, RO_CODE]
@require_torch
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a : str = self.tokenizer(
self.src_text , text_target=self.tgt_text , padding=SCREAMING_SNAKE_CASE__ , truncation=SCREAMING_SNAKE_CASE__ , max_length=len(self.expected_src_tokens ) , return_tensors='pt' , )
__a : Optional[int] = shift_tokens_right(batch['labels'] , self.tokenizer.pad_token_id )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual((2, 1_4) , batch.input_ids.shape )
self.assertEqual((2, 1_4) , batch.attention_mask.shape )
__a : List[Any] = batch.input_ids.tolist()[0]
self.assertListEqual(self.expected_src_tokens , SCREAMING_SNAKE_CASE__ )
self.assertEqual(2 , batch.decoder_input_ids[0, -1] ) # EOS
# Test that special tokens are reset
self.assertEqual(self.tokenizer.prefix_tokens , [] )
self.assertEqual(self.tokenizer.suffix_tokens , [self.tokenizer.eos_token_id, EN_CODE] )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : str = self.tokenizer(self.src_text , padding=SCREAMING_SNAKE_CASE__ , truncation=SCREAMING_SNAKE_CASE__ , max_length=3 , return_tensors='pt' )
__a : Optional[int] = self.tokenizer(
text_target=self.tgt_text , padding=SCREAMING_SNAKE_CASE__ , truncation=SCREAMING_SNAKE_CASE__ , max_length=1_0 , return_tensors='pt' )
__a : Optional[Any] = targets['input_ids']
__a : Dict = shift_tokens_right(SCREAMING_SNAKE_CASE__ , self.tokenizer.pad_token_id )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.decoder_input_ids.shape[1] , 1_0 )
@require_torch
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Dict = self.tokenizer._build_translation_inputs(
'A test' , return_tensors='pt' , src_lang='en_XX' , tgt_lang='ar_AR' )
self.assertEqual(
nested_simplify(SCREAMING_SNAKE_CASE__ ) , {
# A, test, EOS, en_XX
'input_ids': [[6_2, 3_0_3_4, 2, 2_5_0_0_0_4]],
'attention_mask': [[1, 1, 1, 1]],
# ar_AR
'forced_bos_token_id': 2_5_0_0_0_1,
} , )
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : list[int] , lowerCamelCase_ : list[int] ):
# Check if the input is valid
if not len(lowerCamelCase_ ) == len(lowerCamelCase_ ) == 3:
raise ValueError('Please enter a valid equation.' )
if equationa[0] == equationa[1] == equationa[0] == equationa[1] == 0:
raise ValueError('Both a & b of two equations can\'t be zero.' )
# Extract the coefficients
__a , __a , __a : Optional[Any] = equationa
__a , __a , __a : Optional[int] = equationa
# Calculate the determinants of the matrices
__a : str = aa * ba - aa * ba
__a : Tuple = ca * ba - ca * ba
__a : Union[str, Any] = aa * ca - aa * ca
# Check if the system of linear equations has a solution (using Cramer's rule)
if determinant == 0:
if determinant_x == determinant_y == 0:
raise ValueError('Infinite solutions. (Consistent system)' )
else:
raise ValueError('No solution. (Inconsistent system)' )
else:
if determinant_x == determinant_y == 0:
# Trivial solution (Inconsistent system)
return (0.0, 0.0)
else:
__a : Any = determinant_x / determinant
__a : Optional[Any] = determinant_y / determinant
# Non-Trivial Solution (Consistent system)
return (x, y)
| 47 | 1 |
from __future__ import annotations
import unittest
from transformers import is_tf_available, is_torch_available
from transformers.testing_utils import DUMMY_UNKNOWN_IDENTIFIER, SMALL_MODEL_IDENTIFIER, is_pt_tf_cross_test, slow
if is_tf_available():
from transformers import (
AutoConfig,
BertConfig,
GPTaConfig,
TaConfig,
TFAutoModel,
TFAutoModelForCausalLM,
TFAutoModelForMaskedLM,
TFAutoModelForPreTraining,
TFAutoModelForQuestionAnswering,
TFAutoModelForSeqaSeqLM,
TFAutoModelForSequenceClassification,
TFAutoModelWithLMHead,
TFBertForMaskedLM,
TFBertForPreTraining,
TFBertForQuestionAnswering,
TFBertForSequenceClassification,
TFBertModel,
TFGPTaLMHeadModel,
TFRobertaForMaskedLM,
TFTaForConditionalGeneration,
)
from transformers.models.bert.modeling_tf_bert import TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST
from transformers.models.gpta.modeling_tf_gpta import TF_GPT2_PRETRAINED_MODEL_ARCHIVE_LIST
from transformers.models.ta.modeling_tf_ta import TF_T5_PRETRAINED_MODEL_ARCHIVE_LIST
if is_torch_available():
from transformers import (
AutoModel,
AutoModelForCausalLM,
AutoModelForMaskedLM,
AutoModelForPreTraining,
AutoModelForQuestionAnswering,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoModelWithLMHead,
BertForMaskedLM,
BertForPreTraining,
BertForQuestionAnswering,
BertForSequenceClassification,
BertModel,
GPTaLMHeadModel,
RobertaForMaskedLM,
TaForConditionalGeneration,
)
@is_pt_tf_cross_test
class _UpperCamelCase( unittest.TestCase ):
@slow
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
for model_name in ["bert-base-uncased"]:
__a : Dict = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Dict = TFAutoModel.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = AutoModel.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for model_name in ["bert-base-uncased"]:
__a : Optional[Any] = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : str = TFAutoModelForPreTraining.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[Any] = AutoModelForPreTraining.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
for model_name in TF_GPT2_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a : Dict = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = TFAutoModelForCausalLM.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
__a , __a : Any = TFAutoModelForCausalLM.from_pretrained(
SCREAMING_SNAKE_CASE__ , output_loading_info=SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = AutoModelForCausalLM.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
__a , __a : str = AutoModelForCausalLM.from_pretrained(
SCREAMING_SNAKE_CASE__ , output_loading_info=SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for model_name in TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a : List[Any] = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[Any] = TFAutoModelWithLMHead.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[str] = AutoModelWithLMHead.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : str ):
'''simple docstring'''
for model_name in TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a : List[str] = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : str = TFAutoModelForMaskedLM.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
__a , __a : Union[str, Any] = TFAutoModelForMaskedLM.from_pretrained(
SCREAMING_SNAKE_CASE__ , output_loading_info=SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = AutoModelForMaskedLM.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
__a , __a : List[str] = AutoModelForMaskedLM.from_pretrained(
SCREAMING_SNAKE_CASE__ , output_loading_info=SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
for model_name in TF_T5_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a : List[Any] = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = TFAutoModelForSeqaSeqLM.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
__a , __a : Any = TFAutoModelForSeqaSeqLM.from_pretrained(
SCREAMING_SNAKE_CASE__ , output_loading_info=SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : str = AutoModelForSeqaSeqLM.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
__a , __a : str = AutoModelForSeqaSeqLM.from_pretrained(
SCREAMING_SNAKE_CASE__ , output_loading_info=SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
for model_name in ["bert-base-uncased"]:
__a : List[Any] = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = TFAutoModelForSequenceClassification.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = AutoModelForSequenceClassification.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
@slow
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
for model_name in ["bert-base-uncased"]:
__a : Optional[int] = AutoConfig.from_pretrained(SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Dict = TFAutoModelForQuestionAnswering.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : List[Any] = AutoModelForQuestionAnswering.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsNotNone(SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : List[Any] = TFAutoModelWithLMHead.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual(model.num_parameters() , 1_4_4_1_0 )
self.assertEqual(model.num_parameters(only_trainable=SCREAMING_SNAKE_CASE__ ) , 1_4_4_1_0 )
__a : int = AutoModelWithLMHead.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual(model.num_parameters() , 1_4_4_1_0 )
self.assertEqual(model.num_parameters(only_trainable=SCREAMING_SNAKE_CASE__ ) , 1_4_4_1_0 )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Union[str, Any] = TFAutoModelWithLMHead.from_pretrained(SCREAMING_SNAKE_CASE__ , from_pt=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual(model.num_parameters() , 1_4_4_1_0 )
self.assertEqual(model.num_parameters(only_trainable=SCREAMING_SNAKE_CASE__ ) , 1_4_4_1_0 )
__a : List[str] = AutoModelWithLMHead.from_pretrained(SCREAMING_SNAKE_CASE__ , from_tf=SCREAMING_SNAKE_CASE__ )
self.assertIsInstance(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
self.assertEqual(model.num_parameters() , 1_4_4_1_0 )
self.assertEqual(model.num_parameters(only_trainable=SCREAMING_SNAKE_CASE__ ) , 1_4_4_1_0 )
| 47 |
from ...utils import is_note_seq_available, is_transformers_available, is_torch_available
from ...utils import OptionalDependencyNotAvailable
try:
if not (is_transformers_available() and is_torch_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_torch_and_transformers_objects import * # noqa F403
else:
from .notes_encoder import SpectrogramNotesEncoder
from .continous_encoder import SpectrogramContEncoder
from .pipeline_spectrogram_diffusion import (
SpectrogramContEncoder,
SpectrogramDiffusionPipeline,
TaFilmDecoder,
)
try:
if not (is_transformers_available() and is_torch_available() and is_note_seq_available()):
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
from ...utils.dummy_transformers_and_torch_and_note_seq_objects import * # noqa F403
else:
from .midi_utils import MidiProcessor
| 47 | 1 |
import torch
def UpperCAmelCase__ ( ):
if torch.cuda.is_available():
__a : str = torch.cuda.device_count()
else:
__a : List[str] = 0
print(f'''Successfully ran on {num_gpus} GPUs''' )
if __name__ == "__main__":
main()
| 47 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_bridgetower''': [
'''BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''BridgeTowerConfig''',
'''BridgeTowerTextConfig''',
'''BridgeTowerVisionConfig''',
],
'''processing_bridgetower''': ['''BridgeTowerProcessor'''],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''BridgeTowerImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''BridgeTowerForContrastiveLearning''',
'''BridgeTowerForImageAndTextRetrieval''',
'''BridgeTowerForMaskedLM''',
'''BridgeTowerModel''',
'''BridgeTowerPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_bridgetower import (
BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP,
BridgeTowerConfig,
BridgeTowerTextConfig,
BridgeTowerVisionConfig,
)
from .processing_bridgetower import BridgeTowerProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_bridgetower import BridgeTowerImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_bridgetower import (
BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST,
BridgeTowerForContrastiveLearning,
BridgeTowerForImageAndTextRetrieval,
BridgeTowerForMaskedLM,
BridgeTowerModel,
BridgeTowerPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure)
| 47 | 1 |
import logging
from pathlib import Path
import numpy as np
import pytorch_lightning as pl
import torch
from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint
from pytorch_lightning.utilities import rank_zero_only
from utils_rag import save_json
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] ):
__a : Dict = filter(lambda lowerCamelCase_ : p.requires_grad , model.parameters() )
__a : Union[str, Any] = sum([np.prod(p.size() ) for p in model_parameters] )
return params
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : Tuple , lowerCamelCase_ : Optional[Any] ):
if metric == "rouge2":
__a : int = '{val_avg_rouge2:.4f}-{step_count}'
elif metric == "bleu":
__a : Optional[Any] = '{val_avg_bleu:.4f}-{step_count}'
elif metric == "em":
__a : Any = '{val_avg_em:.4f}-{step_count}'
elif metric == "loss":
__a : int = '{val_avg_loss:.4f}-{step_count}'
else:
raise NotImplementedError(
f'''seq2seq callbacks only support rouge2 and bleu, got {metric}, You can make your own by adding to this'''
' function.' )
__a : Union[str, Any] = ModelCheckpoint(
dirpath=lowerCamelCase_ , filename=lowerCamelCase_ , monitor=f'''val_{metric}''' , mode='max' , save_top_k=1 , every_n_epochs=1 , )
return checkpoint_callback
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Any ):
return EarlyStopping(
monitor=f'''val_{metric}''' , mode='min' if 'loss' in metric else 'max' , patience=lowerCamelCase_ , verbose=lowerCamelCase_ , )
class _UpperCamelCase( pl.Callback ):
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Tuple ):
'''simple docstring'''
__a : List[str] = {f'''lr_group_{i}''': param['lr'] for i, param in enumerate(pl_module.trainer.optimizers[0].param_groups )}
pl_module.logger.log_metrics(SCREAMING_SNAKE_CASE__ )
@rank_zero_only
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : pl.Trainer , SCREAMING_SNAKE_CASE__ : pl.LightningModule , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : Any=True ):
'''simple docstring'''
logger.info(f'''***** {type_path} results at step {trainer.global_step:05d} *****''' )
__a : Optional[int] = trainer.callback_metrics
trainer.logger.log_metrics({k: v for k, v in metrics.items() if k not in ['log', 'progress_bar', 'preds']} )
# Log results
__a : Tuple = Path(pl_module.hparams.output_dir )
if type_path == "test":
__a : int = od / 'test_results.txt'
__a : str = od / 'test_generations.txt'
else:
# this never gets hit. I prefer not to save intermediate generations, and results are in metrics.json
# If people want this it will be easy enough to add back.
__a : Tuple = od / f'''{type_path}_results/{trainer.global_step:05d}.txt'''
__a : List[str] = od / f'''{type_path}_generations/{trainer.global_step:05d}.txt'''
results_file.parent.mkdir(exist_ok=SCREAMING_SNAKE_CASE__ )
generations_file.parent.mkdir(exist_ok=SCREAMING_SNAKE_CASE__ )
with open(SCREAMING_SNAKE_CASE__ , 'a+' ) as writer:
for key in sorted(SCREAMING_SNAKE_CASE__ ):
if key in ["log", "progress_bar", "preds"]:
continue
__a : Tuple = metrics[key]
if isinstance(SCREAMING_SNAKE_CASE__ , torch.Tensor ):
__a : Union[str, Any] = val.item()
__a : List[str] = f'''{key}: {val:.6f}\n'''
writer.write(SCREAMING_SNAKE_CASE__ )
if not save_generations:
return
if "preds" in metrics:
__a : List[Any] = '\n'.join(metrics['preds'] )
generations_file.open('w+' ).write(SCREAMING_SNAKE_CASE__ )
@rank_zero_only
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : str ):
'''simple docstring'''
try:
__a : Optional[int] = pl_module.model.model.num_parameters()
except AttributeError:
__a : List[Any] = pl_module.model.num_parameters()
__a : Dict = count_trainable_parameters(SCREAMING_SNAKE_CASE__ )
# mp stands for million parameters
trainer.logger.log_metrics({'n_params': npars, 'mp': npars / 1e6, 'grad_mp': n_trainable_pars / 1e6} )
@rank_zero_only
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : pl.Trainer , SCREAMING_SNAKE_CASE__ : pl.LightningModule ):
'''simple docstring'''
save_json(pl_module.metrics , pl_module.metrics_save_path )
return self._write_logs(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , 'test' )
@rank_zero_only
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : pl.Trainer , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
save_json(pl_module.metrics , pl_module.metrics_save_path )
# Uncommenting this will save val generations
# return self._write_logs(trainer, pl_module, "valid")
| 47 |
from string import ascii_lowercase, ascii_uppercase
def UpperCAmelCase__ ( lowerCamelCase_ : str ):
if not sentence:
return ""
__a : Union[str, Any] = dict(zip(lowerCamelCase_ , lowerCamelCase_ ) )
return lower_to_upper.get(sentence[0] , sentence[0] ) + sentence[1:]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : list ):
if len(lowerCamelCase_ ) < 2:
return collection
def circle_sort_util(lowerCamelCase_ : list , lowerCamelCase_ : int , lowerCamelCase_ : int ) -> bool:
__a : Optional[Any] = False
if low == high:
return swapped
__a : Tuple = low
__a : int = high
while left < right:
if collection[left] > collection[right]:
__a , __a : Optional[Any] = (
collection[right],
collection[left],
)
__a : Tuple = True
left += 1
right -= 1
if left == right and collection[left] > collection[right + 1]:
__a , __a : Dict = (
collection[right + 1],
collection[left],
)
__a : Tuple = True
__a : int = low + int((high - low) / 2 )
__a : Tuple = circle_sort_util(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
__a : int = circle_sort_util(lowerCamelCase_ , mid + 1 , lowerCamelCase_ )
return swapped or left_swap or right_swap
__a : Optional[int] = True
while is_not_sorted is True:
__a : Tuple = circle_sort_util(lowerCamelCase_ , 0 , len(lowerCamelCase_ ) - 1 )
return collection
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter numbers separated by a comma:\n''').strip()
SCREAMING_SNAKE_CASE__ = [int(item) for item in user_input.split(''',''')]
print(circle_sort(unsorted))
| 47 |
import functools
import operator
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''asapp/sew-d-tiny-100k''': '''https://huggingface.co/asapp/sew-d-tiny-100k/resolve/main/config.json''',
# See all SEW-D models at https://huggingface.co/models?filter=sew-d
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Optional[Any] = '''sew-d'''
def __init__( self : Dict , SCREAMING_SNAKE_CASE__ : Dict=3_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=7_6_8 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : str=3_0_7_2 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : List[str]=2_5_6 , SCREAMING_SNAKE_CASE__ : Optional[Any]=True , SCREAMING_SNAKE_CASE__ : Tuple=True , SCREAMING_SNAKE_CASE__ : List[str]=("p2c", "c2p") , SCREAMING_SNAKE_CASE__ : str="layer_norm" , SCREAMING_SNAKE_CASE__ : Tuple="gelu_python" , SCREAMING_SNAKE_CASE__ : Tuple=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.0 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[int]=0.02 , SCREAMING_SNAKE_CASE__ : int=1e-7 , SCREAMING_SNAKE_CASE__ : Any=1e-5 , SCREAMING_SNAKE_CASE__ : Optional[int]="group" , SCREAMING_SNAKE_CASE__ : Optional[Any]="gelu" , SCREAMING_SNAKE_CASE__ : Optional[int]=(6_4, 1_2_8, 1_2_8, 1_2_8, 1_2_8, 2_5_6, 2_5_6, 2_5_6, 2_5_6, 5_1_2, 5_1_2, 5_1_2, 5_1_2) , SCREAMING_SNAKE_CASE__ : List[Any]=(5, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : str=(1_0, 3, 1, 3, 1, 3, 1, 3, 1, 2, 1, 2, 1) , SCREAMING_SNAKE_CASE__ : Optional[int]=False , SCREAMING_SNAKE_CASE__ : Optional[int]=1_2_8 , SCREAMING_SNAKE_CASE__ : Tuple=1_6 , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[Any]=0.05 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , SCREAMING_SNAKE_CASE__ : int=0.0 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_0 , SCREAMING_SNAKE_CASE__ : Optional[int]=0 , SCREAMING_SNAKE_CASE__ : Optional[int]="mean" , SCREAMING_SNAKE_CASE__ : List[Any]=False , SCREAMING_SNAKE_CASE__ : List[str]=False , SCREAMING_SNAKE_CASE__ : str=2_5_6 , SCREAMING_SNAKE_CASE__ : str=0 , SCREAMING_SNAKE_CASE__ : List[Any]=1 , SCREAMING_SNAKE_CASE__ : List[Any]=2 , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(**SCREAMING_SNAKE_CASE__ , pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = hidden_size
__a : Optional[Any] = feat_extract_norm
__a : List[str] = feat_extract_activation
__a : Dict = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = list(SCREAMING_SNAKE_CASE__ )
__a : List[str] = list(SCREAMING_SNAKE_CASE__ )
__a : int = conv_bias
__a : Tuple = num_conv_pos_embeddings
__a : List[str] = num_conv_pos_embedding_groups
__a : Optional[Any] = len(self.conv_dim )
__a : Union[str, Any] = num_hidden_layers
__a : Optional[Any] = intermediate_size
__a : Union[str, Any] = squeeze_factor
__a : List[Any] = max_position_embeddings
__a : Tuple = position_buckets
__a : Optional[int] = share_att_key
__a : List[str] = relative_attention
__a : Any = norm_rel_ebd
__a : Any = list(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = hidden_act
__a : str = num_attention_heads
__a : Union[str, Any] = hidden_dropout
__a : Optional[int] = attention_dropout
__a : List[str] = activation_dropout
__a : int = feat_proj_dropout
__a : int = final_dropout
__a : Dict = layer_norm_eps
__a : Tuple = feature_layer_norm_eps
__a : str = initializer_range
__a : Tuple = vocab_size
if (
(len(self.conv_stride ) != self.num_feat_extract_layers)
or (len(self.conv_kernel ) != self.num_feat_extract_layers)
or (len(self.conv_dim ) != self.num_feat_extract_layers)
):
raise ValueError(
'Configuration for convolutional layers is incorrect.'
'It is required that `len(config.conv_dim)` == `len(config.conv_stride)` == `len(config.conv_kernel)`,'
f'''but is `len(config.conv_dim) = {len(self.conv_dim )}`, `len(config.conv_stride)'''
f'''= {len(self.conv_stride )}`, `len(config.conv_kernel) = {len(self.conv_kernel )}`.''' )
# fine-tuning config parameters for SpecAugment: https://arxiv.org/abs/1904.08779
__a : Tuple = apply_spec_augment
__a : Optional[Any] = mask_time_prob
__a : Any = mask_time_length
__a : List[str] = mask_time_min_masks
__a : List[str] = mask_feature_prob
__a : Tuple = mask_feature_length
__a : Any = mask_feature_min_masks
# ctc loss
__a : Optional[int] = ctc_loss_reduction
__a : List[Any] = ctc_zero_infinity
# sequence classification
__a : Dict = use_weighted_layer_sum
__a : Optional[Any] = classifier_proj_size
@property
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return functools.reduce(operator.mul , self.conv_stride , 1 )
| 47 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_nllb_moe''': [
'''NLLB_MOE_PRETRAINED_CONFIG_ARCHIVE_MAP''',
'''NllbMoeConfig''',
]
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''NLLB_MOE_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''NllbMoeForConditionalGeneration''',
'''NllbMoeModel''',
'''NllbMoePreTrainedModel''',
'''NllbMoeTop2Router''',
'''NllbMoeSparseMLP''',
]
if TYPE_CHECKING:
from .configuration_nllb_moe import (
NLLB_MOE_PRETRAINED_CONFIG_ARCHIVE_MAP,
NllbMoeConfig,
)
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_nllb_moe import (
NLLB_MOE_PRETRAINED_MODEL_ARCHIVE_LIST,
NllbMoeForConditionalGeneration,
NllbMoeModel,
NllbMoePreTrainedModel,
NllbMoeSparseMLP,
NllbMoeTopaRouter,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
from __future__ import annotations
from sys import maxsize
from typing import Generic, TypeVar
SCREAMING_SNAKE_CASE__ = TypeVar('''T''')
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (position - 1) // 2
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 1
def UpperCAmelCase__ ( lowerCamelCase_ : int ):
return (2 * position) + 2
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[str] ):
'''simple docstring'''
__a : list[tuple[T, int]] = []
__a : dict[T, int] = {}
__a : int = 0
def __len__( self : Any ):
'''simple docstring'''
return self.elements
def __repr__( self : Any ):
'''simple docstring'''
return str(self.heap )
def __lowerCAmelCase ( self : Optional[Any] ):
'''simple docstring'''
return self.elements == 0
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.heap.append((elem, weight) )
__a : List[Any] = self.elements
self.elements += 1
self._bubble_up(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
if self.elements > 1:
self._swap_nodes(0 , self.elements - 1 )
__a , __a : Union[str, Any] = self.heap.pop()
del self.position_map[elem]
self.elements -= 1
if self.elements > 0:
__a , __a : Dict = self.heap[0]
self._bubble_down(SCREAMING_SNAKE_CASE__ )
return elem
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
__a : str = (elem, weight)
if position > 0:
__a : Tuple = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : Dict = self.heap[parent_position]
if parent_weight > weight:
self._bubble_up(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
self._bubble_down(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : List[Any] = self.position_map[elem]
if curr_pos == 0:
return None
__a : List[str] = get_parent_position(SCREAMING_SNAKE_CASE__ )
__a , __a : str = self.heap[curr_pos]
__a , __a : Optional[int] = self.heap[parent_position]
if parent_weight > weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_up(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : Optional[Any] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
__a : int = self.position_map[elem]
__a , __a : Optional[Any] = self.heap[curr_pos]
__a : Tuple = get_child_left_position(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = get_child_right_position(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements and child_right_position < self.elements:
__a , __a : str = self.heap[child_left_position]
__a , __a : List[str] = self.heap[child_right_position]
if child_right_weight < child_left_weight and child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
if child_left_position < self.elements:
__a , __a : Any = self.heap[child_left_position]
if child_left_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
else:
return None
if child_right_position < self.elements:
__a , __a : Union[str, Any] = self.heap[child_right_position]
if child_right_weight < weight:
self._swap_nodes(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return self._bubble_down(SCREAMING_SNAKE_CASE__ )
return None
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Optional[Any] = self.heap[nodea_pos][0]
__a : str = self.heap[nodea_pos][0]
__a , __a : int = (
self.heap[nodea_pos],
self.heap[nodea_pos],
)
__a : str = nodea_pos
__a : Optional[int] = nodea_pos
class _UpperCamelCase( Generic[T] ):
def __init__( self : List[Any] ):
'''simple docstring'''
__a : dict[T, dict[T, int]] = {}
__a : int = 0
def __repr__( self : Tuple ):
'''simple docstring'''
return str(self.connections )
def __len__( self : Dict ):
'''simple docstring'''
return self.nodes
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : T ):
'''simple docstring'''
if node not in self.connections:
__a : Tuple = {}
self.nodes += 1
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : T , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self.add_node(SCREAMING_SNAKE_CASE__ )
self.add_node(SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = weight
__a : Any = weight
def UpperCAmelCase__ ( lowerCamelCase_ : GraphUndirectedWeighted[T] , ):
__a : dict[T, int] = {node: maxsize for node in graph.connections}
__a : dict[T, T | None] = {node: None for node in graph.connections}
__a : MinPriorityQueue[T] = MinPriorityQueue()
for node, weight in dist.items():
priority_queue.push(lowerCamelCase_ , lowerCamelCase_ )
if priority_queue.is_empty():
return dist, parent
# initialization
__a : Optional[int] = priority_queue.extract_min()
__a : int = 0
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : str = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Optional[int] = node
# running prim's algorithm
while not priority_queue.is_empty():
__a : Any = priority_queue.extract_min()
for neighbour in graph.connections[node]:
if dist[neighbour] > dist[node] + graph.connections[node][neighbour]:
__a : Tuple = dist[node] + graph.connections[node][neighbour]
priority_queue.update_key(lowerCamelCase_ , dist[neighbour] )
__a : Dict = node
return dist, parent
| 47 | 1 |
import argparse
import os
import gluonnlp as nlp
import mxnet as mx
import numpy as np
import torch
from gluonnlp.base import get_home_dir
from gluonnlp.model.bert import BERTEncoder
from gluonnlp.model.utils import _load_vocab
from gluonnlp.vocab import Vocab
from packaging import version
from torch import nn
from transformers import BertConfig, BertForMaskedLM, BertModel, RobertaTokenizer
from transformers.models.bert.modeling_bert import (
BertIntermediate,
BertLayer,
BertOutput,
BertSelfAttention,
BertSelfOutput,
)
from transformers.utils import logging
if version.parse(nlp.__version__) != version.parse('''0.8.3'''):
raise Exception('''requires gluonnlp == 0.8.3''')
if version.parse(mx.__version__) != version.parse('''1.5.0'''):
raise Exception('''requires mxnet == 1.5.0''')
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''The Nymphenburg Palace is a beautiful palace in Munich!'''
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
__a : List[Any] = {
'attention_cell': 'multi_head',
'num_layers': 4,
'units': 1_0_2_4,
'hidden_size': 7_6_8,
'max_length': 5_1_2,
'num_heads': 8,
'scaled': True,
'dropout': 0.1,
'use_residual': True,
'embed_size': 1_0_2_4,
'embed_dropout': 0.1,
'word_embed': None,
'layer_norm_eps': 1e-5,
'token_type_vocab_size': 2,
}
__a : Optional[int] = bort_4_8_768_1024_hparams
# Let's construct the original Bort model here
# Taken from official BERT implementation, see:
# https://github.com/alexa/bort/blob/master/bort/bort.py
__a : List[str] = BERTEncoder(
attention_cell=predefined_args['attention_cell'] , num_layers=predefined_args['num_layers'] , units=predefined_args['units'] , hidden_size=predefined_args['hidden_size'] , max_length=predefined_args['max_length'] , num_heads=predefined_args['num_heads'] , scaled=predefined_args['scaled'] , dropout=predefined_args['dropout'] , output_attention=lowerCamelCase_ , output_all_encodings=lowerCamelCase_ , use_residual=predefined_args['use_residual'] , activation=predefined_args.get('activation' , 'gelu' ) , layer_norm_eps=predefined_args.get('layer_norm_eps' , lowerCamelCase_ ) , )
# Vocab information needs to be fetched first
# It's the same as RoBERTa, so RobertaTokenizer can be used later
__a : int = 'openwebtext_ccnews_stories_books_cased'
# Specify download folder to Gluonnlp's vocab
__a : Optional[Any] = os.path.join(get_home_dir() , 'models' )
__a : Optional[Any] = _load_vocab(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , cls=lowerCamelCase_ )
__a : Any = nlp.model.BERTModel(
lowerCamelCase_ , len(lowerCamelCase_ ) , units=predefined_args['units'] , embed_size=predefined_args['embed_size'] , embed_dropout=predefined_args['embed_dropout'] , word_embed=predefined_args['word_embed'] , use_pooler=lowerCamelCase_ , use_token_type_embed=lowerCamelCase_ , token_type_vocab_size=predefined_args['token_type_vocab_size'] , use_classifier=lowerCamelCase_ , use_decoder=lowerCamelCase_ , )
original_bort.load_parameters(lowerCamelCase_ , cast_dtype=lowerCamelCase_ , ignore_extra=lowerCamelCase_ )
__a : Dict = original_bort._collect_params_with_prefix()
# Build our config 🤗
__a : Optional[Any] = {
'architectures': ['BertForMaskedLM'],
'attention_probs_dropout_prob': predefined_args['dropout'],
'hidden_act': 'gelu',
'hidden_dropout_prob': predefined_args['dropout'],
'hidden_size': predefined_args['embed_size'],
'initializer_range': 0.02,
'intermediate_size': predefined_args['hidden_size'],
'layer_norm_eps': predefined_args['layer_norm_eps'],
'max_position_embeddings': predefined_args['max_length'],
'model_type': 'bort',
'num_attention_heads': predefined_args['num_heads'],
'num_hidden_layers': predefined_args['num_layers'],
'pad_token_id': 1, # 2 = BERT, 1 = RoBERTa
'type_vocab_size': 1, # 2 = BERT, 1 = RoBERTa
'vocab_size': len(lowerCamelCase_ ),
}
__a : str = BertConfig.from_dict(lowerCamelCase_ )
__a : Optional[int] = BertForMaskedLM(lowerCamelCase_ )
hf_bort_model.eval()
# Parameter mapping table (Gluonnlp to Transformers)
# * denotes layer index
#
# | Gluon Parameter | Transformers Parameter
# | -------------------------------------------------------------- | ----------------------
# | `encoder.layer_norm.beta` | `bert.embeddings.LayerNorm.bias`
# | `encoder.layer_norm.gamma` | `bert.embeddings.LayerNorm.weight`
# | `encoder.position_weight` | `bert.embeddings.position_embeddings.weight`
# | `word_embed.0.weight` | `bert.embeddings.word_embeddings.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_key.bias` | `bert.encoder.layer.*.attention.self.key.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_key.weight` | `bert.encoder.layer.*.attention.self.key.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_query.bias` | `bert.encoder.layer.*.attention.self.query.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_query.weight` | `bert.encoder.layer.*.attention.self.query.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_value.bias` | `bert.encoder.layer.*.attention.self.value.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_value.weight` | `bert.encoder.layer.*.attention.self.value.weight`
# | `encoder.transformer_cells.*.ffn.ffn_2.bias` | `bert.encoder.layer.*.attention.output.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_2.weight` | `bert.encoder.layer.*.attention.output.dense.weight`
# | `encoder.transformer_cells.*.layer_norm.beta` | `bert.encoder.layer.*.attention.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.layer_norm.gamma` | `bert.encoder.layer.*.attention.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.ffn.ffn_1.bias` | `bert.encoder.layer.*.intermediate.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_1.weight` | `bert.encoder.layer.*.intermediate.dense.weight`
# | `encoder.transformer_cells.*.ffn.layer_norm.beta` | `bert.encoder.layer.*.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.ffn.layer_norm.gamma` | `bert.encoder.layer.*.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.proj.bias` | `bert.encoder.layer.*.output.dense.bias`
# | `encoder.transformer_cells.*.proj.weight` | `bert.encoder.layer.*.output.dense.weight`
# Helper function to convert MXNET Arrays to PyTorch
def to_torch(lowerCamelCase_ : Optional[Any] ) -> nn.Parameter:
return nn.Parameter(torch.FloatTensor(mx_array.data().asnumpy() ) )
# Check param shapes and map new HF param back
def check_and_map_params(lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[str] ):
__a : Optional[int] = hf_param.shape
__a : int = to_torch(params[gluon_param] )
__a : int = gluon_param.shape
assert (
shape_hf == shape_gluon
), f'''The gluon parameter {gluon_param} has shape {shape_gluon}, but expects shape {shape_hf} for Transformers'''
return gluon_param
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.word_embeddings.weight , 'word_embed.0.weight' )
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.position_embeddings.weight , 'encoder.position_weight' )
__a : Tuple = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.bias , 'encoder.layer_norm.beta' )
__a : Union[str, Any] = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.weight , 'encoder.layer_norm.gamma' )
# Inspired by RoBERTa conversion script, we just zero them out (Bort does not use them)
__a : Union[str, Any] = torch.zeros_like(
hf_bort_model.bert.embeddings.token_type_embeddings.weight.data )
for i in range(hf_bort_config.num_hidden_layers ):
__a : BertLayer = hf_bort_model.bert.encoder.layer[i]
# self attention
__a : BertSelfAttention = layer.attention.self
__a : Optional[int] = check_and_map_params(
self_attn.key.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.bias''' )
__a : str = check_and_map_params(
self_attn.key.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.weight''' )
__a : List[str] = check_and_map_params(
self_attn.query.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.bias''' )
__a : str = check_and_map_params(
self_attn.query.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.weight''' )
__a : Dict = check_and_map_params(
self_attn.value.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.bias''' )
__a : str = check_and_map_params(
self_attn.value.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.weight''' )
# self attention output
__a : BertSelfOutput = layer.attention.output
__a : Tuple = check_and_map_params(
self_output.dense.bias , f'''encoder.transformer_cells.{i}.proj.bias''' )
__a : Dict = check_and_map_params(
self_output.dense.weight , f'''encoder.transformer_cells.{i}.proj.weight''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.layer_norm.beta''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.layer_norm.gamma''' )
# intermediate
__a : BertIntermediate = layer.intermediate
__a : List[str] = check_and_map_params(
intermediate.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_1.bias''' )
__a : Optional[Any] = check_and_map_params(
intermediate.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_1.weight''' )
# output
__a : BertOutput = layer.output
__a : str = check_and_map_params(
bert_output.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_2.bias''' )
__a : List[Any] = check_and_map_params(
bert_output.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_2.weight''' )
__a : str = check_and_map_params(
bert_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.ffn.layer_norm.beta''' )
__a : List[str] = check_and_map_params(
bert_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.ffn.layer_norm.gamma''' )
# Save space and energy 🎄
hf_bort_model.half()
# Compare output of both models
__a : Union[str, Any] = RobertaTokenizer.from_pretrained('roberta-base' )
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ )['input_ids']
# Get gluon output
__a : Optional[int] = mx.nd.array([input_ids] )
__a : Tuple = original_bort(inputs=lowerCamelCase_ , token_types=[] )
# Get Transformer output (save and reload model again)
hf_bort_model.save_pretrained(lowerCamelCase_ )
__a : Optional[Any] = BertModel.from_pretrained(lowerCamelCase_ )
hf_bort_model.eval()
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ , return_tensors='pt' )
__a : int = hf_bort_model(**lowerCamelCase_ )[0]
__a : Dict = output_gluon[0].asnumpy()
__a : str = output_hf[0].detach().numpy()
__a : List[Any] = np.max(np.abs(hf_layer - gluon_layer ) ).item()
__a : str = np.allclose(lowerCamelCase_ , lowerCamelCase_ , atol=1e-3 )
if success:
print('✔️ Both model do output the same tensors' )
else:
print('❌ Both model do **NOT** output the same tensors' )
print('Absolute difference is:' , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--bort_checkpoint_path''', default=None, type=str, required=True, help='''Path the official Bort params file.'''
)
parser.add_argument(
'''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.'''
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_bort_checkpoint_to_pytorch(args.bort_checkpoint_path, args.pytorch_dump_folder_path)
| 47 |
from collections.abc import Sequence
from queue import Queue
class _UpperCamelCase:
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : List[str] , SCREAMING_SNAKE_CASE__ : List[Any] , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : int=None , SCREAMING_SNAKE_CASE__ : Tuple=None ):
'''simple docstring'''
__a : Tuple = start
__a : Dict = end
__a : List[str] = val
__a : List[Any] = (start + end) // 2
__a : Optional[Any] = left
__a : List[str] = right
def __repr__( self : Dict ):
'''simple docstring'''
return f'''SegmentTreeNode(start={self.start}, end={self.end}, val={self.val})'''
class _UpperCamelCase:
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : Sequence , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
__a : Tuple = collection
__a : Dict = function
if self.collection:
__a : int = self._build_tree(0 , len(SCREAMING_SNAKE_CASE__ ) - 1 )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
self._update_tree(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : Any , SCREAMING_SNAKE_CASE__ : List[str] ):
'''simple docstring'''
return self._query_range(self.root , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
if start == end:
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.collection[start] )
__a : Tuple = (start + end) // 2
__a : Optional[int] = self._build_tree(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Tuple = self._build_tree(mid + 1 , SCREAMING_SNAKE_CASE__ )
return SegmentTreeNode(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , self.fn(left.val , right.val ) , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] , SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : List[Any] ):
'''simple docstring'''
if node.start == i and node.end == i:
__a : Optional[Any] = val
return
if i <= node.mid:
self._update_tree(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
self._update_tree(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : int = self.fn(node.left.val , node.right.val )
def __lowerCAmelCase ( self : Any , SCREAMING_SNAKE_CASE__ : Dict , SCREAMING_SNAKE_CASE__ : Tuple , SCREAMING_SNAKE_CASE__ : Optional[Any] ):
'''simple docstring'''
if node.start == i and node.end == j:
return node.val
if i <= node.mid:
if j <= node.mid:
# range in left child tree
return self._query_range(node.left , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
else:
# range in left child tree and right child tree
return self.fn(
self._query_range(node.left , SCREAMING_SNAKE_CASE__ , node.mid ) , self._query_range(node.right , node.mid + 1 , SCREAMING_SNAKE_CASE__ ) , )
else:
# range in right child tree
return self._query_range(node.right , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
if self.root is not None:
__a : Tuple = Queue()
queue.put(self.root )
while not queue.empty():
__a : Tuple = queue.get()
yield node
if node.left is not None:
queue.put(node.left )
if node.right is not None:
queue.put(node.right )
if __name__ == "__main__":
import operator
for fn in [operator.add, max, min]:
print('''*''' * 50)
SCREAMING_SNAKE_CASE__ = SegmentTree([2, 1, 5, 3, 4], fn)
for node in arr.traverse():
print(node)
print()
arr.update(1, 5)
for node in arr.traverse():
print(node)
print()
print(arr.query_range(3, 4)) # 7
print(arr.query_range(2, 2)) # 5
print(arr.query_range(1, 3)) # 13
print()
| 47 | 1 |
from typing import TYPE_CHECKING
# rely on isort to merge the imports
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
SCREAMING_SNAKE_CASE__ = {'''configuration_focalnet''': ['''FOCALNET_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''FocalNetConfig''']}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''FOCALNET_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''FocalNetForImageClassification''',
'''FocalNetForMaskedImageModeling''',
'''FocalNetBackbone''',
'''FocalNetModel''',
'''FocalNetPreTrainedModel''',
]
if TYPE_CHECKING:
from .configuration_focalnet import FOCALNET_PRETRAINED_CONFIG_ARCHIVE_MAP, FocalNetConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_focalnet import (
FOCALNET_PRETRAINED_MODEL_ARCHIVE_LIST,
FocalNetBackbone,
FocalNetForImageClassification,
FocalNetForMaskedImageModeling,
FocalNetModel,
FocalNetPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
import os
import posixpath
import uuid
from dataclasses import dataclass
from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, Union
import numpy as np
import pyarrow as pa
import datasets
from datasets.arrow_writer import ArrowWriter, ParquetWriter
from datasets.config import MAX_SHARD_SIZE
from datasets.filesystems import (
is_remote_filesystem,
rename,
)
from datasets.iterable_dataset import _BaseExamplesIterable
from datasets.utils.py_utils import convert_file_size_to_int
SCREAMING_SNAKE_CASE__ = datasets.utils.logging.get_logger(__name__)
if TYPE_CHECKING:
import pyspark
@dataclass
class _UpperCamelCase( datasets.BuilderConfig ):
__SCREAMING_SNAKE_CASE : Optional[datasets.Features] = None
def UpperCAmelCase__ ( lowerCamelCase_ : "pyspark.sql.DataFrame" , lowerCamelCase_ : List[int] , ):
import pyspark
def generate_fn():
__a : List[Any] = df.select('*' , pyspark.sql.functions.spark_partition_id().alias('part_id' ) )
for partition_id in partition_order:
__a : Optional[int] = df_with_partition_id.select('*' ).where(f'''part_id = {partition_id}''' ).drop('part_id' )
__a : Optional[Any] = partition_df.collect()
__a : Union[str, Any] = 0
for row in rows:
yield f'''{partition_id}_{row_id}''', row.asDict()
row_id += 1
return generate_fn
class _UpperCamelCase( _BaseExamplesIterable ):
def __init__( self : Any , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : Dict=None , ):
'''simple docstring'''
__a : List[str] = df
__a : Tuple = partition_order or range(self.df.rdd.getNumPartitions() )
__a : List[Any] = _generate_iterable_examples(self.df , self.partition_order )
def __iter__( self : Tuple ):
'''simple docstring'''
yield from self.generate_examples_fn()
def __lowerCAmelCase ( self : Optional[int] , SCREAMING_SNAKE_CASE__ : np.random.Generator ):
'''simple docstring'''
__a : Union[str, Any] = list(range(self.df.rdd.getNumPartitions() ) )
generator.shuffle(SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int ):
'''simple docstring'''
__a : Union[str, Any] = self.split_shard_indices_by_worker(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
return SparkExamplesIterable(self.df , partition_order=SCREAMING_SNAKE_CASE__ )
@property
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
return len(self.partition_order )
class _UpperCamelCase( datasets.DatasetBuilder ):
__SCREAMING_SNAKE_CASE : List[str] = SparkConfig
def __init__( self : List[str] , SCREAMING_SNAKE_CASE__ : "pyspark.sql.DataFrame" , SCREAMING_SNAKE_CASE__ : str = None , SCREAMING_SNAKE_CASE__ : str = None , **SCREAMING_SNAKE_CASE__ : Optional[int] , ):
'''simple docstring'''
import pyspark
__a : int = pyspark.sql.SparkSession.builder.getOrCreate()
__a : Optional[int] = df
__a : List[Any] = working_dir
super().__init__(
cache_dir=SCREAMING_SNAKE_CASE__ , config_name=str(self.df.semanticHash() ) , **SCREAMING_SNAKE_CASE__ , )
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
def create_cache_and_write_probe(SCREAMING_SNAKE_CASE__ : List[str] ):
# makedirs with exist_ok will recursively create the directory. It will not throw an error if directories
# already exist.
os.makedirs(self._cache_dir , exist_ok=SCREAMING_SNAKE_CASE__ )
__a : List[Any] = os.path.join(self._cache_dir , 'fs_test' + uuid.uuida().hex )
# Opening the file in append mode will create a new file unless it already exists, in which case it will not
# change the file contents.
open(SCREAMING_SNAKE_CASE__ , 'a' )
return [probe_file]
if self._spark.conf.get('spark.master' , '' ).startswith('local' ):
return
# If the cluster is multi-node, make sure that the user provided a cache_dir and that it is on an NFS
# accessible to the driver.
# TODO: Stream batches to the driver using ArrowCollectSerializer instead of throwing an error.
if self._cache_dir:
__a : List[Any] = (
self._spark.sparkContext.parallelize(range(1 ) , 1 ).mapPartitions(SCREAMING_SNAKE_CASE__ ).collect()
)
if os.path.isfile(probe[0] ):
return
raise ValueError(
'When using Dataset.from_spark on a multi-node cluster, the driver and all workers should be able to access cache_dir' )
def __lowerCAmelCase ( self : List[str] ):
'''simple docstring'''
return datasets.DatasetInfo(features=self.config.features )
def __lowerCAmelCase ( self : int , SCREAMING_SNAKE_CASE__ : datasets.download.download_manager.DownloadManager ):
'''simple docstring'''
return [datasets.SplitGenerator(name=datasets.Split.TRAIN )]
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Any ):
'''simple docstring'''
import pyspark
def get_arrow_batch_size(SCREAMING_SNAKE_CASE__ : int ):
for batch in it:
yield pa.RecordBatch.from_pydict({'batch_bytes': [batch.nbytes]} )
__a : List[str] = self.df.count()
__a : Dict = df_num_rows if df_num_rows <= 1_0_0 else 1_0_0
# Approximate the size of each row (in Arrow format) by averaging over a max-100-row sample.
__a : List[str] = (
self.df.limit(SCREAMING_SNAKE_CASE__ )
.repartition(1 )
.mapInArrow(SCREAMING_SNAKE_CASE__ , 'batch_bytes: long' )
.agg(pyspark.sql.functions.sum('batch_bytes' ).alias('sample_bytes' ) )
.collect()[0]
.sample_bytes
/ sample_num_rows
)
__a : Dict = approx_bytes_per_row * df_num_rows
if approx_total_size > max_shard_size:
# Make sure there is at least one row per partition.
__a : Union[str, Any] = min(SCREAMING_SNAKE_CASE__ , int(approx_total_size / max_shard_size ) )
__a : int = self.df.repartition(SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : str , SCREAMING_SNAKE_CASE__ : int , ):
'''simple docstring'''
import pyspark
__a : Any = ParquetWriter if file_format == 'parquet' else ArrowWriter
__a : Union[str, Any] = os.path.join(self._working_dir , os.path.basename(SCREAMING_SNAKE_CASE__ ) ) if self._working_dir else fpath
__a : Optional[int] = file_format == 'parquet'
# Define these so that we don't reference self in write_arrow, which will result in a pickling error due to
# pickling the SparkContext.
__a : List[str] = self.config.features
__a : int = self._writer_batch_size
__a : Union[str, Any] = self._fs.storage_options
def write_arrow(SCREAMING_SNAKE_CASE__ : Optional[int] ):
# Within the same SparkContext, no two task attempts will share the same attempt ID.
__a : Any = pyspark.TaskContext().taskAttemptId()
__a : str = next(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
if first_batch is None:
# Some partitions might not receive any data.
return pa.RecordBatch.from_arrays(
[[task_id], [0], [0]] , names=['task_id', 'num_examples', 'num_bytes'] , )
__a : Any = 0
__a : List[str] = writer_class(
features=SCREAMING_SNAKE_CASE__ , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Optional[Any] = pa.Table.from_batches([first_batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
for batch in it:
if max_shard_size is not None and writer._num_bytes >= max_shard_size:
__a , __a : Optional[int] = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
shard_id += 1
__a : Optional[Any] = writer_class(
features=writer._features , path=working_fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , writer_batch_size=SCREAMING_SNAKE_CASE__ , storage_options=SCREAMING_SNAKE_CASE__ , embed_local_files=SCREAMING_SNAKE_CASE__ , )
__a : Union[str, Any] = pa.Table.from_batches([batch] )
writer.write_table(SCREAMING_SNAKE_CASE__ )
if writer._num_bytes > 0:
__a , __a : str = writer.finalize()
writer.close()
yield pa.RecordBatch.from_arrays(
[[task_id], [num_examples], [num_bytes]] , names=['task_id', 'num_examples', 'num_bytes'] , )
if working_fpath != fpath:
for file in os.listdir(os.path.dirname(SCREAMING_SNAKE_CASE__ ) ):
__a : Any = os.path.join(os.path.dirname(SCREAMING_SNAKE_CASE__ ) , os.path.basename(SCREAMING_SNAKE_CASE__ ) )
shutil.move(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
__a : Dict = (
self.df.mapInArrow(SCREAMING_SNAKE_CASE__ , 'task_id: long, num_examples: long, num_bytes: long' )
.groupBy('task_id' )
.agg(
pyspark.sql.functions.sum('num_examples' ).alias('total_num_examples' ) , pyspark.sql.functions.sum('num_bytes' ).alias('total_num_bytes' ) , pyspark.sql.functions.count('num_bytes' ).alias('num_shards' ) , pyspark.sql.functions.collect_list('num_examples' ).alias('shard_lengths' ) , )
.collect()
)
for row in stats:
yield row.task_id, (row.total_num_examples, row.total_num_bytes, row.num_shards, row.shard_lengths)
def __lowerCAmelCase ( self : Union[str, Any] , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , SCREAMING_SNAKE_CASE__ : str = "arrow" , SCREAMING_SNAKE_CASE__ : Optional[Union[str, int]] = None , SCREAMING_SNAKE_CASE__ : Optional[int] = None , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
self._validate_cache_dir()
__a : List[str] = convert_file_size_to_int(max_shard_size or MAX_SHARD_SIZE )
self._repartition_df_if_needed(SCREAMING_SNAKE_CASE__ )
__a : Union[str, Any] = not is_remote_filesystem(self._fs )
__a : Optional[Any] = os.path.join if is_local else posixpath.join
__a : Any = '-TTTTT-SSSSS-of-NNNNN'
__a : Union[str, Any] = f'''{self.name}-{split_generator.name}{SUFFIX}.{file_format}'''
__a : Any = path_join(self._output_dir , SCREAMING_SNAKE_CASE__ )
__a : Any = 0
__a : Dict = 0
__a : int = 0
__a : List[str] = []
__a : Optional[int] = []
for task_id, content in self._prepare_split_single(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) : Optional[int] = content
if num_bytes > 0:
total_num_examples += num_examples
total_num_bytes += num_bytes
total_shards += num_shards
task_id_and_num_shards.append((task_id, num_shards) )
all_shard_lengths.extend(SCREAMING_SNAKE_CASE__ )
__a : List[str] = total_num_examples
__a : Optional[int] = total_num_bytes
# should rename everything at the end
logger.debug(f'''Renaming {total_shards} shards.''' )
if total_shards > 1:
__a : Any = all_shard_lengths
# Define fs outside of _rename_shard so that we don't reference self in the function, which will result in a
# pickling error due to pickling the SparkContext.
__a : Dict = self._fs
# use the -SSSSS-of-NNNNN pattern
def _rename_shard(
SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , SCREAMING_SNAKE_CASE__ : int , ):
rename(
SCREAMING_SNAKE_CASE__ , fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace('TTTTT-SSSSS' , f'''{global_shard_id:05d}''' ).replace('NNNNN' , f'''{total_shards:05d}''' ) , )
__a : Union[str, Any] = []
__a : List[str] = 0
for i in range(len(SCREAMING_SNAKE_CASE__ ) ):
__a , __a : Union[str, Any] = task_id_and_num_shards[i]
for shard_id in range(SCREAMING_SNAKE_CASE__ ):
args.append([task_id, shard_id, global_shard_id] )
global_shard_id += 1
self._spark.sparkContext.parallelize(SCREAMING_SNAKE_CASE__ , len(SCREAMING_SNAKE_CASE__ ) ).map(lambda SCREAMING_SNAKE_CASE__ : _rename_shard(*SCREAMING_SNAKE_CASE__ ) ).collect()
else:
# don't use any pattern
__a : List[Any] = 0
__a : Any = task_id_and_num_shards[0][0]
self._rename(
fpath.replace('SSSSS' , f'''{shard_id:05d}''' ).replace('TTTTT' , f'''{task_id:05d}''' ) , fpath.replace(SCREAMING_SNAKE_CASE__ , '' ) , )
def __lowerCAmelCase ( self : str , SCREAMING_SNAKE_CASE__ : "datasets.SplitGenerator" , ):
'''simple docstring'''
return SparkExamplesIterable(self.df )
| 47 | 1 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''roberta-base''': '''https://huggingface.co/roberta-base/resolve/main/config.json''',
'''roberta-large''': '''https://huggingface.co/roberta-large/resolve/main/config.json''',
'''roberta-large-mnli''': '''https://huggingface.co/roberta-large-mnli/resolve/main/config.json''',
'''distilroberta-base''': '''https://huggingface.co/distilroberta-base/resolve/main/config.json''',
'''roberta-base-openai-detector''': '''https://huggingface.co/roberta-base-openai-detector/resolve/main/config.json''',
'''roberta-large-openai-detector''': '''https://huggingface.co/roberta-large-openai-detector/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = '''roberta'''
def __init__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=5_0_2_6_5 , SCREAMING_SNAKE_CASE__ : Optional[int]=7_6_8 , SCREAMING_SNAKE_CASE__ : str=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Any="gelu" , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.1 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=2 , SCREAMING_SNAKE_CASE__ : Any=0.02 , SCREAMING_SNAKE_CASE__ : List[str]=1e-12 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0 , SCREAMING_SNAKE_CASE__ : List[str]=2 , SCREAMING_SNAKE_CASE__ : Tuple="absolute" , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[str]=None , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = vocab_size
__a : Tuple = hidden_size
__a : List[str] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : str = hidden_act
__a : Optional[Any] = intermediate_size
__a : Dict = hidden_dropout_prob
__a : List[str] = attention_probs_dropout_prob
__a : Optional[Any] = max_position_embeddings
__a : Dict = type_vocab_size
__a : str = initializer_range
__a : List[str] = layer_norm_eps
__a : Optional[int] = position_embedding_type
__a : Union[str, Any] = use_cache
__a : str = classifier_dropout
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
if self.task == "multiple-choice":
__a : List[str] = {0: 'batch', 1: 'choice', 2: 'sequence'}
else:
__a : Dict = {0: 'batch', 1: 'sequence'}
return OrderedDict(
[
('input_ids', dynamic_axis),
('attention_mask', dynamic_axis),
] )
| 47 |
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, TensorDataset
from tqdm import tqdm
from transformers import GPTaLMHeadModel
SCREAMING_SNAKE_CASE__ = logging.getLogger(__name__)
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : int ):
# save results
if os.path.exists(lowerCamelCase_ ):
if os.path.exists(os.path.join(lowerCamelCase_ , 'config.json' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'config.json' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'config.json' ) )
if os.path.exists(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ) and os.path.isfile(
os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) ):
os.remove(os.path.join(lowerCamelCase_ , 'pytorch_model.bin' ) )
else:
os.makedirs(lowerCamelCase_ )
model.save_pretrained(lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : int , lowerCamelCase_ : Any=False ):
__a : Dict = 2
if unlogit:
__a : Optional[Any] = torch.pow(lowerCamelCase_ , lowerCamelCase_ )
__a : Any = p * torch.log(lowerCamelCase_ )
__a : Union[str, Any] = 0
return -plogp.sum(dim=-1 )
def UpperCAmelCase__ ( lowerCamelCase_ : Any ):
logger.info('lv, h >\t' + '\t'.join(f'''{x + 1}''' for x in range(len(lowerCamelCase_ ) ) ) )
for row in range(len(lowerCamelCase_ ) ):
if tensor.dtype != torch.long:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:.5f}''' for x in tensor[row].cpu().data ) )
else:
logger.info(f'''layer {row + 1}:\t''' + '\t'.join(f'''{x:d}''' for x in tensor[row].cpu().data ) )
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : Any , lowerCamelCase_ : int , lowerCamelCase_ : int=True , lowerCamelCase_ : Optional[Any]=True , lowerCamelCase_ : List[Any]=None , lowerCamelCase_ : List[Any]=False ):
__a , __a : Optional[int] = model.config.num_hidden_layers, model.config.num_attention_heads
__a : str = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
__a : int = torch.zeros(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
if head_mask is None:
__a : Union[str, Any] = torch.ones(lowerCamelCase_ , lowerCamelCase_ ).to(args.device )
head_mask.requires_grad_(requires_grad=lowerCamelCase_ )
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
__a : Any = None
__a : Optional[int] = 0.0
__a : Optional[Any] = 0.0
for step, inputs in enumerate(tqdm(lowerCamelCase_ , desc='Iteration' , disable=args.local_rank not in [-1, 0] ) ):
__a : Dict = tuple(t.to(args.device ) for t in inputs )
((__a) , ) : Dict = inputs
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
__a : List[Any] = model(lowerCamelCase_ , labels=lowerCamelCase_ , head_mask=lowerCamelCase_ )
# (loss), lm_logits, presents, (all hidden_states), (attentions)
__a , __a , __a : int = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
total_loss += loss.detach().cpu().numpy()
if compute_entropy:
for layer, attn in enumerate(lowerCamelCase_ ):
__a : List[str] = entropy(attn.detach() , lowerCamelCase_ )
attn_entropy[layer] += masked_entropy.sum(-1 ).sum(0 ).sum(0 ).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
tot_tokens += torch.ones_like(lowerCamelCase_ ).float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
__a : Optional[Any] = 2
__a : Union[str, Any] = torch.pow(torch.pow(lowerCamelCase_ , lowerCamelCase_ ).sum(-1 ) , 1 / exponent )
head_importance /= norm_by_layer.unsqueeze(-1 ) + 1e-20
if not args.dont_normalize_global_importance:
__a : List[str] = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print matrices
if compute_entropy:
logger.info('Attention entropies' )
print_ad_tensor(lowerCamelCase_ )
if compute_importance:
logger.info('Head importance scores' )
print_ad_tensor(lowerCamelCase_ )
logger.info('Head ranked by importance scores' )
__a : Optional[Any] = torch.zeros(head_importance.numel() , dtype=torch.long , device=args.device )
__a : str = torch.arange(
head_importance.numel() , device=args.device )
__a : Tuple = head_ranks.view_as(lowerCamelCase_ )
print_ad_tensor(lowerCamelCase_ )
return attn_entropy, head_importance, total_loss
def UpperCAmelCase__ ( lowerCamelCase_ : Union[str, Any] , lowerCamelCase_ : Tuple , lowerCamelCase_ : int ):
__a , __a , __a : Optional[int] = compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ )
__a : Tuple = 1 / loss # instead of downsteam score use the LM loss
logger.info('Pruning: original score: %f, threshold: %f' , lowerCamelCase_ , original_score * args.masking_threshold )
__a : Tuple = torch.ones_like(lowerCamelCase_ )
__a : int = max(1 , int(new_head_mask.numel() * args.masking_amount ) )
__a : Tuple = original_score
while current_score >= original_score * args.masking_threshold:
__a : Optional[Any] = new_head_mask.clone().detach() # save current head mask
# heads from least important to most - keep only not-masked heads
__a : List[str] = float('Inf' )
__a : List[Any] = head_importance.view(-1 ).sort()[1]
if len(lowerCamelCase_ ) <= num_to_mask:
print('BREAK BY num_to_mask' )
break
# mask heads
__a : Any = current_heads_to_mask[:num_to_mask]
logger.info('Heads to mask: %s' , str(current_heads_to_mask.tolist() ) )
__a : int = new_head_mask.view(-1 )
__a : Tuple = 0.0
__a : int = new_head_mask.view_as(lowerCamelCase_ )
__a : Optional[int] = new_head_mask.clone().detach()
print_ad_tensor(lowerCamelCase_ )
# Compute metric and head importance again
__a , __a , __a : int = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[Any] = 1 / loss
logger.info(
'Masking: current score: %f, remaining heads %d (%.1f percents)' , lowerCamelCase_ , new_head_mask.sum() , new_head_mask.sum() / new_head_mask.numel() * 1_0_0 , )
logger.info('Final head mask' )
print_ad_tensor(lowerCamelCase_ )
np.save(os.path.join(args.output_dir , 'head_mask.npy' ) , head_mask.detach().cpu().numpy() )
return head_mask
def UpperCAmelCase__ ( lowerCamelCase_ : Dict , lowerCamelCase_ : Tuple , lowerCamelCase_ : Any , lowerCamelCase_ : Union[str, Any] ):
__a : List[Any] = datetime.now()
__a , __a , __a : List[str] = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ )
__a : List[str] = 1 / loss
__a : List[Any] = datetime.now() - before_time
__a : List[str] = sum(p.numel() for p in model.parameters() )
__a : Dict = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(lowerCamelCase_ ) )
}
for k, v in heads_to_prune.items():
if isinstance(lowerCamelCase_ , lowerCamelCase_ ):
__a : Tuple = [
v,
]
assert sum(len(lowerCamelCase_ ) for h in heads_to_prune.values() ) == (1 - head_mask.long()).sum().item()
model.prune_heads(lowerCamelCase_ )
__a : Optional[Any] = sum(p.numel() for p in model.parameters() )
__a : Tuple = datetime.now()
__a , __a , __a : Tuple = compute_heads_importance(
lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , compute_entropy=lowerCamelCase_ , compute_importance=lowerCamelCase_ , head_mask=lowerCamelCase_ , actually_pruned=lowerCamelCase_ , )
__a : Optional[Any] = 1 / loss
__a : List[Any] = datetime.now() - before_time
logger.info(
'Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)' , lowerCamelCase_ , lowerCamelCase_ , pruned_num_params / original_num_params * 1_0_0 , )
logger.info('Pruning: score with masking: %f score with pruning: %f' , lowerCamelCase_ , lowerCamelCase_ )
logger.info('Pruning: speed ratio (original timing / new timing): %f percents' , original_time / new_time * 1_0_0 )
save_model(lowerCamelCase_ , args.output_dir )
def UpperCAmelCase__ ( ):
__a : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'--data_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The input data dir. Should contain the .tsv files (or other data files) for the task.' , )
parser.add_argument(
'--model_name_or_path' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='Path to pretrained model or model identifier from huggingface.co/models' , )
parser.add_argument(
'--output_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , required=lowerCamelCase_ , help='The output directory where the model predictions and checkpoints will be written.' , )
# Other parameters
parser.add_argument(
'--config_name' , default='' , type=lowerCamelCase_ , help='Pretrained config name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--tokenizer_name' , default='' , type=lowerCamelCase_ , help='Pretrained tokenizer name or path if not the same as model_name_or_path' , )
parser.add_argument(
'--cache_dir' , default=lowerCamelCase_ , type=lowerCamelCase_ , help='Where do you want to store the pre-trained models downloaded from s3' , )
parser.add_argument(
'--data_subset' , type=lowerCamelCase_ , default=-1 , help='If > 0: limit the data to a subset of data_subset instances.' )
parser.add_argument(
'--overwrite_output_dir' , action='store_true' , help='Whether to overwrite data in output directory' )
parser.add_argument(
'--overwrite_cache' , action='store_true' , help='Overwrite the cached training and evaluation sets' )
parser.add_argument(
'--dont_normalize_importance_by_layer' , action='store_true' , help='Don\'t normalize importance score by layers' )
parser.add_argument(
'--dont_normalize_global_importance' , action='store_true' , help='Don\'t normalize all importance scores between 0 and 1' , )
parser.add_argument(
'--try_masking' , action='store_true' , help='Whether to try to mask head until a threshold of accuracy.' )
parser.add_argument(
'--masking_threshold' , default=0.9 , type=lowerCamelCase_ , help='masking threshold in term of metrics (stop masking when metric < threshold * original metric value).' , )
parser.add_argument(
'--masking_amount' , default=0.1 , type=lowerCamelCase_ , help='Amount to heads to masking at each masking step.' )
parser.add_argument('--metric_name' , default='acc' , type=lowerCamelCase_ , help='Metric to use for head masking.' )
parser.add_argument(
'--max_seq_length' , default=1_2_8 , type=lowerCamelCase_ , help=(
'The maximum total input sequence length after WordPiece tokenization. \n'
'Sequences longer than this will be truncated, sequences shorter padded.'
) , )
parser.add_argument('--batch_size' , default=1 , type=lowerCamelCase_ , help='Batch size.' )
parser.add_argument('--seed' , type=lowerCamelCase_ , default=4_2 )
parser.add_argument('--local_rank' , type=lowerCamelCase_ , default=-1 , help='local_rank for distributed training on gpus' )
parser.add_argument('--no_cuda' , action='store_true' , help='Whether not to use CUDA when available' )
parser.add_argument('--server_ip' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
parser.add_argument('--server_port' , type=lowerCamelCase_ , default='' , help='Can be used for distant debugging.' )
__a : Optional[Any] = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print('Waiting for debugger attach' )
ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=lowerCamelCase_ )
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
__a : List[str] = torch.device('cuda' if torch.cuda.is_available() and not args.no_cuda else 'cpu' )
__a : Tuple = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank )
__a : Union[str, Any] = torch.device('cuda' , args.local_rank )
__a : Any = 1
torch.distributed.init_process_group(backend='nccl' ) # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN )
logger.info('device: {} n_gpu: {}, distributed: {}'.format(args.device , args.n_gpu , bool(args.local_rank != -1 ) ) )
__a : Optional[Any] = GPTaLMHeadModel.from_pretrained(args.model_name_or_path )
# Distributed and parallel training
model.to(args.device )
if args.local_rank != -1:
__a : List[Any] = nn.parallel.DistributedDataParallel(
lowerCamelCase_ , device_ids=[args.local_rank] , output_device=args.local_rank , find_unused_parameters=lowerCamelCase_ )
elif args.n_gpu > 1:
__a : Union[str, Any] = nn.DataParallel(lowerCamelCase_ )
# Print/save training arguments
os.makedirs(args.output_dir , exist_ok=lowerCamelCase_ )
torch.save(lowerCamelCase_ , os.path.join(args.output_dir , 'run_args.bin' ) )
logger.info('Training/evaluation parameters %s' , lowerCamelCase_ )
# Prepare dataset
__a : Tuple = np.concatenate(
[
np.loadtxt(args.data_dir , dtype=np.intaa ),
] )
__a : str = (torch.from_numpy(lowerCamelCase_ ),)
__a : List[str] = TensorDataset(*lowerCamelCase_ )
__a : Optional[Any] = RandomSampler(lowerCamelCase_ )
__a : Union[str, Any] = DataLoader(lowerCamelCase_ , sampler=lowerCamelCase_ , batch_size=args.batch_size )
# Compute head entropy and importance score
compute_heads_importance(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
__a : Union[str, Any] = mask_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
prune_heads(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
if __name__ == "__main__":
main()
| 47 | 1 |
# Copyright 2023 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import TYPE_CHECKING
# rely on isort to merge the imports
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
SCREAMING_SNAKE_CASE__ = {
'''configuration_vivit''': ['''VIVIT_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''VivitConfig'''],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = ['''VivitImageProcessor''']
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
SCREAMING_SNAKE_CASE__ = [
'''VIVIT_PRETRAINED_MODEL_ARCHIVE_LIST''',
'''VivitModel''',
'''VivitPreTrainedModel''',
'''VivitForVideoClassification''',
]
if TYPE_CHECKING:
from .configuration_vivit import VIVIT_PRETRAINED_CONFIG_ARCHIVE_MAP, VivitConfig
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_vivit import VivitImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_vivit import (
VIVIT_PRETRAINED_MODEL_ARCHIVE_LIST,
VivitForVideoClassification,
VivitModel,
VivitPreTrainedModel,
)
else:
import sys
SCREAMING_SNAKE_CASE__ = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
| 47 |
import argparse
import os
import gluonnlp as nlp
import mxnet as mx
import numpy as np
import torch
from gluonnlp.base import get_home_dir
from gluonnlp.model.bert import BERTEncoder
from gluonnlp.model.utils import _load_vocab
from gluonnlp.vocab import Vocab
from packaging import version
from torch import nn
from transformers import BertConfig, BertForMaskedLM, BertModel, RobertaTokenizer
from transformers.models.bert.modeling_bert import (
BertIntermediate,
BertLayer,
BertOutput,
BertSelfAttention,
BertSelfOutput,
)
from transformers.utils import logging
if version.parse(nlp.__version__) != version.parse('''0.8.3'''):
raise Exception('''requires gluonnlp == 0.8.3''')
if version.parse(mx.__version__) != version.parse('''1.5.0'''):
raise Exception('''requires mxnet == 1.5.0''')
logging.set_verbosity_info()
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = '''The Nymphenburg Palace is a beautiful palace in Munich!'''
def UpperCAmelCase__ ( lowerCamelCase_ : str , lowerCamelCase_ : str ):
__a : List[Any] = {
'attention_cell': 'multi_head',
'num_layers': 4,
'units': 1_0_2_4,
'hidden_size': 7_6_8,
'max_length': 5_1_2,
'num_heads': 8,
'scaled': True,
'dropout': 0.1,
'use_residual': True,
'embed_size': 1_0_2_4,
'embed_dropout': 0.1,
'word_embed': None,
'layer_norm_eps': 1e-5,
'token_type_vocab_size': 2,
}
__a : Optional[int] = bort_4_8_768_1024_hparams
# Let's construct the original Bort model here
# Taken from official BERT implementation, see:
# https://github.com/alexa/bort/blob/master/bort/bort.py
__a : List[str] = BERTEncoder(
attention_cell=predefined_args['attention_cell'] , num_layers=predefined_args['num_layers'] , units=predefined_args['units'] , hidden_size=predefined_args['hidden_size'] , max_length=predefined_args['max_length'] , num_heads=predefined_args['num_heads'] , scaled=predefined_args['scaled'] , dropout=predefined_args['dropout'] , output_attention=lowerCamelCase_ , output_all_encodings=lowerCamelCase_ , use_residual=predefined_args['use_residual'] , activation=predefined_args.get('activation' , 'gelu' ) , layer_norm_eps=predefined_args.get('layer_norm_eps' , lowerCamelCase_ ) , )
# Vocab information needs to be fetched first
# It's the same as RoBERTa, so RobertaTokenizer can be used later
__a : int = 'openwebtext_ccnews_stories_books_cased'
# Specify download folder to Gluonnlp's vocab
__a : Optional[Any] = os.path.join(get_home_dir() , 'models' )
__a : Optional[Any] = _load_vocab(lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ , cls=lowerCamelCase_ )
__a : Any = nlp.model.BERTModel(
lowerCamelCase_ , len(lowerCamelCase_ ) , units=predefined_args['units'] , embed_size=predefined_args['embed_size'] , embed_dropout=predefined_args['embed_dropout'] , word_embed=predefined_args['word_embed'] , use_pooler=lowerCamelCase_ , use_token_type_embed=lowerCamelCase_ , token_type_vocab_size=predefined_args['token_type_vocab_size'] , use_classifier=lowerCamelCase_ , use_decoder=lowerCamelCase_ , )
original_bort.load_parameters(lowerCamelCase_ , cast_dtype=lowerCamelCase_ , ignore_extra=lowerCamelCase_ )
__a : Dict = original_bort._collect_params_with_prefix()
# Build our config 🤗
__a : Optional[Any] = {
'architectures': ['BertForMaskedLM'],
'attention_probs_dropout_prob': predefined_args['dropout'],
'hidden_act': 'gelu',
'hidden_dropout_prob': predefined_args['dropout'],
'hidden_size': predefined_args['embed_size'],
'initializer_range': 0.02,
'intermediate_size': predefined_args['hidden_size'],
'layer_norm_eps': predefined_args['layer_norm_eps'],
'max_position_embeddings': predefined_args['max_length'],
'model_type': 'bort',
'num_attention_heads': predefined_args['num_heads'],
'num_hidden_layers': predefined_args['num_layers'],
'pad_token_id': 1, # 2 = BERT, 1 = RoBERTa
'type_vocab_size': 1, # 2 = BERT, 1 = RoBERTa
'vocab_size': len(lowerCamelCase_ ),
}
__a : str = BertConfig.from_dict(lowerCamelCase_ )
__a : Optional[int] = BertForMaskedLM(lowerCamelCase_ )
hf_bort_model.eval()
# Parameter mapping table (Gluonnlp to Transformers)
# * denotes layer index
#
# | Gluon Parameter | Transformers Parameter
# | -------------------------------------------------------------- | ----------------------
# | `encoder.layer_norm.beta` | `bert.embeddings.LayerNorm.bias`
# | `encoder.layer_norm.gamma` | `bert.embeddings.LayerNorm.weight`
# | `encoder.position_weight` | `bert.embeddings.position_embeddings.weight`
# | `word_embed.0.weight` | `bert.embeddings.word_embeddings.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_key.bias` | `bert.encoder.layer.*.attention.self.key.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_key.weight` | `bert.encoder.layer.*.attention.self.key.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_query.bias` | `bert.encoder.layer.*.attention.self.query.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_query.weight` | `bert.encoder.layer.*.attention.self.query.weight`
# | `encoder.transformer_cells.*.attention_cell.proj_value.bias` | `bert.encoder.layer.*.attention.self.value.bias`
# | `encoder.transformer_cells.*.attention_cell.proj_value.weight` | `bert.encoder.layer.*.attention.self.value.weight`
# | `encoder.transformer_cells.*.ffn.ffn_2.bias` | `bert.encoder.layer.*.attention.output.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_2.weight` | `bert.encoder.layer.*.attention.output.dense.weight`
# | `encoder.transformer_cells.*.layer_norm.beta` | `bert.encoder.layer.*.attention.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.layer_norm.gamma` | `bert.encoder.layer.*.attention.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.ffn.ffn_1.bias` | `bert.encoder.layer.*.intermediate.dense.bias`
# | `encoder.transformer_cells.*.ffn.ffn_1.weight` | `bert.encoder.layer.*.intermediate.dense.weight`
# | `encoder.transformer_cells.*.ffn.layer_norm.beta` | `bert.encoder.layer.*.output.LayerNorm.bias`
# | `encoder.transformer_cells.*.ffn.layer_norm.gamma` | `bert.encoder.layer.*.output.LayerNorm.weight`
# | `encoder.transformer_cells.*.proj.bias` | `bert.encoder.layer.*.output.dense.bias`
# | `encoder.transformer_cells.*.proj.weight` | `bert.encoder.layer.*.output.dense.weight`
# Helper function to convert MXNET Arrays to PyTorch
def to_torch(lowerCamelCase_ : Optional[Any] ) -> nn.Parameter:
return nn.Parameter(torch.FloatTensor(mx_array.data().asnumpy() ) )
# Check param shapes and map new HF param back
def check_and_map_params(lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : List[str] ):
__a : Optional[int] = hf_param.shape
__a : int = to_torch(params[gluon_param] )
__a : int = gluon_param.shape
assert (
shape_hf == shape_gluon
), f'''The gluon parameter {gluon_param} has shape {shape_gluon}, but expects shape {shape_hf} for Transformers'''
return gluon_param
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.word_embeddings.weight , 'word_embed.0.weight' )
__a : str = check_and_map_params(
hf_bort_model.bert.embeddings.position_embeddings.weight , 'encoder.position_weight' )
__a : Tuple = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.bias , 'encoder.layer_norm.beta' )
__a : Union[str, Any] = check_and_map_params(
hf_bort_model.bert.embeddings.LayerNorm.weight , 'encoder.layer_norm.gamma' )
# Inspired by RoBERTa conversion script, we just zero them out (Bort does not use them)
__a : Union[str, Any] = torch.zeros_like(
hf_bort_model.bert.embeddings.token_type_embeddings.weight.data )
for i in range(hf_bort_config.num_hidden_layers ):
__a : BertLayer = hf_bort_model.bert.encoder.layer[i]
# self attention
__a : BertSelfAttention = layer.attention.self
__a : Optional[int] = check_and_map_params(
self_attn.key.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.bias''' )
__a : str = check_and_map_params(
self_attn.key.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_key.weight''' )
__a : List[str] = check_and_map_params(
self_attn.query.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.bias''' )
__a : str = check_and_map_params(
self_attn.query.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_query.weight''' )
__a : Dict = check_and_map_params(
self_attn.value.bias.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.bias''' )
__a : str = check_and_map_params(
self_attn.value.weight.data , f'''encoder.transformer_cells.{i}.attention_cell.proj_value.weight''' )
# self attention output
__a : BertSelfOutput = layer.attention.output
__a : Tuple = check_and_map_params(
self_output.dense.bias , f'''encoder.transformer_cells.{i}.proj.bias''' )
__a : Dict = check_and_map_params(
self_output.dense.weight , f'''encoder.transformer_cells.{i}.proj.weight''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.layer_norm.beta''' )
__a : Optional[Any] = check_and_map_params(
self_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.layer_norm.gamma''' )
# intermediate
__a : BertIntermediate = layer.intermediate
__a : List[str] = check_and_map_params(
intermediate.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_1.bias''' )
__a : Optional[Any] = check_and_map_params(
intermediate.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_1.weight''' )
# output
__a : BertOutput = layer.output
__a : str = check_and_map_params(
bert_output.dense.bias , f'''encoder.transformer_cells.{i}.ffn.ffn_2.bias''' )
__a : List[Any] = check_and_map_params(
bert_output.dense.weight , f'''encoder.transformer_cells.{i}.ffn.ffn_2.weight''' )
__a : str = check_and_map_params(
bert_output.LayerNorm.bias , f'''encoder.transformer_cells.{i}.ffn.layer_norm.beta''' )
__a : List[str] = check_and_map_params(
bert_output.LayerNorm.weight , f'''encoder.transformer_cells.{i}.ffn.layer_norm.gamma''' )
# Save space and energy 🎄
hf_bort_model.half()
# Compare output of both models
__a : Union[str, Any] = RobertaTokenizer.from_pretrained('roberta-base' )
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ )['input_ids']
# Get gluon output
__a : Optional[int] = mx.nd.array([input_ids] )
__a : Tuple = original_bort(inputs=lowerCamelCase_ , token_types=[] )
# Get Transformer output (save and reload model again)
hf_bort_model.save_pretrained(lowerCamelCase_ )
__a : Optional[Any] = BertModel.from_pretrained(lowerCamelCase_ )
hf_bort_model.eval()
__a : Union[str, Any] = tokenizer.encode_plus(lowerCamelCase_ , return_tensors='pt' )
__a : int = hf_bort_model(**lowerCamelCase_ )[0]
__a : Dict = output_gluon[0].asnumpy()
__a : str = output_hf[0].detach().numpy()
__a : List[Any] = np.max(np.abs(hf_layer - gluon_layer ) ).item()
__a : str = np.allclose(lowerCamelCase_ , lowerCamelCase_ , atol=1e-3 )
if success:
print('✔️ Both model do output the same tensors' )
else:
print('❌ Both model do **NOT** output the same tensors' )
print('Absolute difference is:' , lowerCamelCase_ )
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
'''--bort_checkpoint_path''', default=None, type=str, required=True, help='''Path the official Bort params file.'''
)
parser.add_argument(
'''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.'''
)
SCREAMING_SNAKE_CASE__ = parser.parse_args()
convert_bort_checkpoint_to_pytorch(args.bort_checkpoint_path, args.pytorch_dump_folder_path)
| 47 | 1 |
from ...configuration_utils import PretrainedConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''facebook/s2t-small-librispeech-asr''': (
'''https://huggingface.co/facebook/s2t-small-librispeech-asr/resolve/main/config.json'''
),
# See all Speech2Text models at https://huggingface.co/models?filter=speech_to_text
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : str = '''speech_to_text'''
__SCREAMING_SNAKE_CASE : Any = ['''past_key_values''']
__SCREAMING_SNAKE_CASE : Optional[Any] = {'''num_attention_heads''': '''encoder_attention_heads''', '''hidden_size''': '''d_model'''}
def __init__( self : Tuple , SCREAMING_SNAKE_CASE__ : Tuple=1_0_0_0_0 , SCREAMING_SNAKE_CASE__ : Dict=1_2 , SCREAMING_SNAKE_CASE__ : Optional[int]=2_0_4_8 , SCREAMING_SNAKE_CASE__ : str=4 , SCREAMING_SNAKE_CASE__ : Optional[int]=6 , SCREAMING_SNAKE_CASE__ : List[str]=2_0_4_8 , SCREAMING_SNAKE_CASE__ : Dict=4 , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.0 , SCREAMING_SNAKE_CASE__ : List[str]=0.0 , SCREAMING_SNAKE_CASE__ : List[str]=True , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : Dict="relu" , SCREAMING_SNAKE_CASE__ : Any=2_5_6 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0.1 , SCREAMING_SNAKE_CASE__ : Any=0.0 , SCREAMING_SNAKE_CASE__ : int=0.0 , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.02 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : List[Any]=True , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1 , SCREAMING_SNAKE_CASE__ : Any=0 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Dict=6_0_0_0 , SCREAMING_SNAKE_CASE__ : Any=1_0_2_4 , SCREAMING_SNAKE_CASE__ : str=2 , SCREAMING_SNAKE_CASE__ : Any=(5, 5) , SCREAMING_SNAKE_CASE__ : List[str]=1_0_2_4 , SCREAMING_SNAKE_CASE__ : Any=8_0 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1 , **SCREAMING_SNAKE_CASE__ : Optional[Any] , ):
'''simple docstring'''
__a : Optional[int] = vocab_size
__a : Optional[int] = d_model
__a : List[str] = encoder_ffn_dim
__a : Optional[Any] = encoder_layers
__a : Optional[Any] = encoder_attention_heads
__a : Union[str, Any] = decoder_ffn_dim
__a : Union[str, Any] = decoder_layers
__a : Optional[Any] = decoder_attention_heads
__a : str = dropout
__a : int = attention_dropout
__a : Optional[Any] = activation_dropout
__a : Dict = activation_function
__a : List[str] = init_std
__a : int = encoder_layerdrop
__a : Tuple = decoder_layerdrop
__a : Optional[int] = use_cache
__a : int = encoder_layers
__a : int = scale_embedding # scale factor will be sqrt(d_model) if True
__a : List[str] = max_source_positions
__a : Optional[Any] = max_target_positions
__a : Tuple = num_conv_layers
__a : Optional[int] = list(SCREAMING_SNAKE_CASE__ )
__a : Dict = conv_channels
__a : str = input_feat_per_channel
__a : Any = input_channels
if len(self.conv_kernel_sizes ) != self.num_conv_layers:
raise ValueError(
'Configuration for convolutional module is incorrect. '
'It is required that `len(config.conv_kernel_sizes)` == `config.num_conv_layers` '
f'''but is `len(config.conv_kernel_sizes) = {len(self.conv_kernel_sizes )}`, '''
f'''`config.num_conv_layers = {self.num_conv_layers}`.''' )
super().__init__(
pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , is_encoder_decoder=SCREAMING_SNAKE_CASE__ , decoder_start_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ , )
| 47 |
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[str] ):
__a : Any = ''
for i in table:
res += inp[i - 1]
return res
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] ):
return data[1:] + data[0]
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[Any] , lowerCamelCase_ : Optional[int] ):
__a : Optional[int] = ''
for i in range(len(lowerCamelCase_ ) ):
if a[i] == b[i]:
res += "0"
else:
res += "1"
return res
def UpperCAmelCase__ ( lowerCamelCase_ : List[Any] , lowerCamelCase_ : str ):
__a : List[str] = int('0b' + data[0] + data[-1] , 2 )
__a : List[str] = int('0b' + data[1:3] , 2 )
return bin(s[row][col] )[2:]
def UpperCAmelCase__ ( lowerCamelCase_ : Any , lowerCamelCase_ : List[str] , lowerCamelCase_ : int , lowerCamelCase_ : int , lowerCamelCase_ : Optional[Any] ):
__a : List[Any] = message[:4]
__a : str = message[4:]
__a : Any = apply_table(lowerCamelCase_ , lowerCamelCase_ )
__a : int = xor(lowerCamelCase_ , lowerCamelCase_ )
__a : Dict = apply_sbox(lowerCamelCase_ , temp[:4] ) # noqa: E741
__a : Tuple = apply_sbox(lowerCamelCase_ , temp[4:] )
__a : List[Any] = '0' * (2 - len(lowerCamelCase_ )) + l # noqa: E741
__a : List[str] = '0' * (2 - len(lowerCamelCase_ )) + r
__a : List[Any] = apply_table(l + r , lowerCamelCase_ )
__a : Dict = xor(lowerCamelCase_ , lowerCamelCase_ )
return temp + right
if __name__ == "__main__":
SCREAMING_SNAKE_CASE__ = input('''Enter 10 bit key: ''')
SCREAMING_SNAKE_CASE__ = input('''Enter 8 bit message: ''')
SCREAMING_SNAKE_CASE__ = [6, 3, 7, 4, 8, 5, 10, 9]
SCREAMING_SNAKE_CASE__ = [3, 5, 2, 7, 4, 10, 1, 9, 8, 6]
SCREAMING_SNAKE_CASE__ = [2, 4, 3, 1]
SCREAMING_SNAKE_CASE__ = [2, 6, 3, 1, 4, 8, 5, 7]
SCREAMING_SNAKE_CASE__ = [4, 1, 3, 5, 7, 2, 8, 6]
SCREAMING_SNAKE_CASE__ = [4, 1, 2, 3, 2, 3, 4, 1]
SCREAMING_SNAKE_CASE__ = [[1, 0, 3, 2], [3, 2, 1, 0], [0, 2, 1, 3], [3, 1, 3, 2]]
SCREAMING_SNAKE_CASE__ = [[0, 1, 2, 3], [2, 0, 1, 3], [3, 0, 1, 0], [2, 1, 0, 3]]
# key generation
SCREAMING_SNAKE_CASE__ = apply_table(key, paa_table)
SCREAMING_SNAKE_CASE__ = temp[:5]
SCREAMING_SNAKE_CASE__ = temp[5:]
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = left_shift(left)
SCREAMING_SNAKE_CASE__ = left_shift(right)
SCREAMING_SNAKE_CASE__ = apply_table(left + right, pa_table)
# encryption
SCREAMING_SNAKE_CASE__ = apply_table(message, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Cipher text is:''', CT)
# decryption
SCREAMING_SNAKE_CASE__ = apply_table(CT, IP)
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = temp[4:] + temp[:4]
SCREAMING_SNAKE_CASE__ = function(expansion, sa, sa, keya, temp)
SCREAMING_SNAKE_CASE__ = apply_table(temp, IP_inv)
print('''Plain text after decypting is:''', PT)
| 47 | 1 |
def UpperCAmelCase__ ( lowerCamelCase_ : Optional[int] , lowerCamelCase_ : List[Any] , lowerCamelCase_ : List[str] , lowerCamelCase_ : Optional[int] ):
if height >= 1:
move_tower(height - 1 , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
move_disk(lowerCamelCase_ , lowerCamelCase_ )
move_tower(height - 1 , lowerCamelCase_ , lowerCamelCase_ , lowerCamelCase_ )
def UpperCAmelCase__ ( lowerCamelCase_ : List[str] , lowerCamelCase_ : Any ):
print('moving disk from' , lowerCamelCase_ , 'to' , lowerCamelCase_ )
def UpperCAmelCase__ ( ):
__a : str = int(input('Height of hanoi: ' ).strip() )
move_tower(lowerCamelCase_ , 'A' , 'B' , 'C' )
if __name__ == "__main__":
main()
| 47 |
import unittest
from transformers import is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from transformers import GradientAccumulator, create_optimizer
@require_tf
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : List[str] , SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Optional[int] , SCREAMING_SNAKE_CASE__ : Union[str, Any] ):
'''simple docstring'''
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , len(SCREAMING_SNAKE_CASE__ ) )
for a, b in zip(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ):
self.assertAlmostEqual(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , delta=SCREAMING_SNAKE_CASE__ )
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
__a : List[Any] = GradientAccumulator()
accumulator([tf.constant([1.0, 2.0] )] )
accumulator([tf.constant([-2.0, 1.0] )] )
accumulator([tf.constant([-1.0, 2.0] )] )
with self.assertRaises(SCREAMING_SNAKE_CASE__ ):
accumulator([tf.constant([1.0, 1.0] ), tf.constant([2.0, 2.0] )] )
self.assertEqual(accumulator.step , 3 )
self.assertEqual(len(accumulator.gradients ) , 1 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [-2.0, 5.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [0.0, 0.0] , tol=1e-2 )
def __lowerCAmelCase ( self : Dict ):
'''simple docstring'''
__a : int = None
ops.enable_eager_execution_internal()
__a : Optional[Any] = tf.config.list_physical_devices('CPU' )
if len(SCREAMING_SNAKE_CASE__ ) == 1:
tf.config.set_logical_device_configuration(
physical_devices[0] , [tf.config.LogicalDeviceConfiguration(), tf.config.LogicalDeviceConfiguration()] )
__a : int = tf.config.list_logical_devices(device_type='CPU' )
__a : str = tf.distribute.MirroredStrategy(devices=devices[:2] )
with strategy.scope():
__a : List[str] = GradientAccumulator()
__a : Tuple = tf.Variable([4.0, 3.0] )
__a , __a : int = create_optimizer(5e-5 , 1_0 , 5 )
__a : List[Any] = tf.Variable([0.0, 0.0] , trainable=SCREAMING_SNAKE_CASE__ )
def accumulate_on_replica(SCREAMING_SNAKE_CASE__ : Optional[Any] ):
accumulator([gradient] )
def apply_on_replica():
optimizer.apply_gradients(list(zip(accumulator.gradients , [variable] ) ) )
@tf.function
def accumulate(SCREAMING_SNAKE_CASE__ : Union[str, Any] , SCREAMING_SNAKE_CASE__ : Tuple ):
with strategy.scope():
__a : Optional[Any] = strategy.experimental_local_results(SCREAMING_SNAKE_CASE__ )
local_variables[0].assign(SCREAMING_SNAKE_CASE__ )
local_variables[1].assign(SCREAMING_SNAKE_CASE__ )
strategy.run(SCREAMING_SNAKE_CASE__ , args=(gradient_placeholder,) )
@tf.function
def apply_grad():
with strategy.scope():
strategy.run(SCREAMING_SNAKE_CASE__ )
def _check_local_values(SCREAMING_SNAKE_CASE__ : Optional[Any] , SCREAMING_SNAKE_CASE__ : int ):
__a : Union[str, Any] = strategy.experimental_local_results(accumulator._gradients[0] )
self.assertListAlmostEqual(values[0].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
self.assertListAlmostEqual(values[1].value() , SCREAMING_SNAKE_CASE__ , tol=1e-2 )
accumulate([1.0, 2.0] , [-1.0, 1.0] )
accumulate([3.0, -1.0] , [-1.0, -1.0] )
accumulate([-2.0, 2.0] , [3.0, -2.0] )
self.assertEqual(accumulator.step , 3 )
_check_local_values([2.0, 3.0] , [1.0, -2.0] )
apply_grad()
self.assertListAlmostEqual(variable.value() , [4.0, 3.0] , tol=1e-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
_check_local_values([0.0, 0.0] , [0.0, 0.0] )
| 47 | 1 |
import time
import unittest
from transformers import is_torch_available
from transformers.testing_utils import require_torch, torch_device
from ..test_modeling_common import ids_tensor
if is_torch_available():
import torch
from transformers.generation import (
MaxLengthCriteria,
MaxNewTokensCriteria,
MaxTimeCriteria,
StoppingCriteriaList,
validate_stopping_criteria,
)
@require_torch
class _UpperCamelCase( unittest.TestCase ):
def __lowerCAmelCase ( self : Tuple , SCREAMING_SNAKE_CASE__ : Dict ):
'''simple docstring'''
__a : Any = 3
__a : int = 2_5_0
__a : Union[str, Any] = ids_tensor((batch_size, length) , SCREAMING_SNAKE_CASE__ )
__a : Optional[int] = torch.ones((batch_size, length) , device=SCREAMING_SNAKE_CASE__ , dtype=torch.float ) / length
return input_ids, scores
def __lowerCAmelCase ( self : Optional[int] ):
'''simple docstring'''
__a , __a : List[str] = self._get_tensors(5 )
__a : List[str] = StoppingCriteriaList(
[
MaxLengthCriteria(max_length=1_0 ),
MaxTimeCriteria(max_time=0.1 ),
] )
self.assertFalse(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a , __a : str = self._get_tensors(9 )
self.assertFalse(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a , __a : List[Any] = self._get_tensors(1_0 )
self.assertTrue(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : List[Any] ):
'''simple docstring'''
__a : Any = MaxLengthCriteria(max_length=1_0 )
__a , __a : Optional[Any] = self._get_tensors(5 )
self.assertFalse(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a , __a : Any = self._get_tensors(9 )
self.assertFalse(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a , __a : List[str] = self._get_tensors(1_0 )
self.assertTrue(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : int ):
'''simple docstring'''
__a : Dict = MaxNewTokensCriteria(start_length=5 , max_new_tokens=5 )
__a , __a : List[Any] = self._get_tensors(5 )
self.assertFalse(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a , __a : Union[str, Any] = self._get_tensors(9 )
self.assertFalse(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a , __a : Any = self._get_tensors(1_0 )
self.assertTrue(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a : Union[str, Any] = StoppingCriteriaList([criteria] )
self.assertEqual(criteria_list.max_length , 1_0 )
def __lowerCAmelCase ( self : Union[str, Any] ):
'''simple docstring'''
__a , __a : Tuple = self._get_tensors(5 )
__a : Dict = MaxTimeCriteria(max_time=0.1 )
self.assertFalse(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
__a : List[Any] = MaxTimeCriteria(max_time=0.1 , initial_timestamp=time.time() - 0.2 )
self.assertTrue(criteria(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ ) )
def __lowerCAmelCase ( self : Tuple ):
'''simple docstring'''
validate_stopping_criteria(StoppingCriteriaList([MaxLengthCriteria(1_0 )] ) , 1_0 )
with self.assertWarns(SCREAMING_SNAKE_CASE__ ):
validate_stopping_criteria(StoppingCriteriaList([MaxLengthCriteria(1_0 )] ) , 1_1 )
__a : List[str] = validate_stopping_criteria(StoppingCriteriaList() , 1_1 )
self.assertEqual(len(SCREAMING_SNAKE_CASE__ ) , 1 )
| 47 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
SCREAMING_SNAKE_CASE__ = logging.get_logger(__name__)
SCREAMING_SNAKE_CASE__ = {
'''roberta-base''': '''https://huggingface.co/roberta-base/resolve/main/config.json''',
'''roberta-large''': '''https://huggingface.co/roberta-large/resolve/main/config.json''',
'''roberta-large-mnli''': '''https://huggingface.co/roberta-large-mnli/resolve/main/config.json''',
'''distilroberta-base''': '''https://huggingface.co/distilroberta-base/resolve/main/config.json''',
'''roberta-base-openai-detector''': '''https://huggingface.co/roberta-base-openai-detector/resolve/main/config.json''',
'''roberta-large-openai-detector''': '''https://huggingface.co/roberta-large-openai-detector/resolve/main/config.json''',
}
class _UpperCamelCase( __lowerCamelCase ):
__SCREAMING_SNAKE_CASE : Union[str, Any] = '''roberta'''
def __init__( self : List[Any] , SCREAMING_SNAKE_CASE__ : Union[str, Any]=5_0_2_6_5 , SCREAMING_SNAKE_CASE__ : Optional[int]=7_6_8 , SCREAMING_SNAKE_CASE__ : str=1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=1_2 , SCREAMING_SNAKE_CASE__ : Optional[Any]=3_0_7_2 , SCREAMING_SNAKE_CASE__ : Any="gelu" , SCREAMING_SNAKE_CASE__ : Optional[Any]=0.1 , SCREAMING_SNAKE_CASE__ : str=0.1 , SCREAMING_SNAKE_CASE__ : Optional[Any]=5_1_2 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=2 , SCREAMING_SNAKE_CASE__ : Any=0.02 , SCREAMING_SNAKE_CASE__ : List[str]=1e-12 , SCREAMING_SNAKE_CASE__ : Optional[Any]=1 , SCREAMING_SNAKE_CASE__ : Union[str, Any]=0 , SCREAMING_SNAKE_CASE__ : List[str]=2 , SCREAMING_SNAKE_CASE__ : Tuple="absolute" , SCREAMING_SNAKE_CASE__ : Any=True , SCREAMING_SNAKE_CASE__ : List[str]=None , **SCREAMING_SNAKE_CASE__ : Any , ):
'''simple docstring'''
super().__init__(pad_token_id=SCREAMING_SNAKE_CASE__ , bos_token_id=SCREAMING_SNAKE_CASE__ , eos_token_id=SCREAMING_SNAKE_CASE__ , **SCREAMING_SNAKE_CASE__ )
__a : Optional[Any] = vocab_size
__a : Tuple = hidden_size
__a : List[str] = num_hidden_layers
__a : List[Any] = num_attention_heads
__a : str = hidden_act
__a : Optional[Any] = intermediate_size
__a : Dict = hidden_dropout_prob
__a : List[str] = attention_probs_dropout_prob
__a : Optional[Any] = max_position_embeddings
__a : Dict = type_vocab_size
__a : str = initializer_range
__a : List[str] = layer_norm_eps
__a : Optional[int] = position_embedding_type
__a : Union[str, Any] = use_cache
__a : str = classifier_dropout
class _UpperCamelCase( __lowerCamelCase ):
@property
def __lowerCAmelCase ( self : Any ):
'''simple docstring'''
if self.task == "multiple-choice":
__a : List[str] = {0: 'batch', 1: 'choice', 2: 'sequence'}
else:
__a : Dict = {0: 'batch', 1: 'sequence'}
return OrderedDict(
[
('input_ids', dynamic_axis),
('attention_mask', dynamic_axis),
] )
| 47 | 1 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.