code
stringlengths 86
54.5k
| code_codestyle
int64 0
371
| style_context
stringlengths 87
49.2k
| style_context_codestyle
int64 0
349
| label
int64 0
1
|
|---|---|---|---|---|
import warnings
from contextlib import contextmanager
from ...processing_utils import ProcessorMixin
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'Speech2TextFeatureExtractor'
SCREAMING_SNAKE_CASE = 'Speech2TextTokenizer'
def __init__( self , __snake_case , __snake_case ) -> Any:
'''simple docstring'''
super().__init__(__snake_case , __snake_case )
__a =self.feature_extractor
__a =False
def __call__( self , *__snake_case , **__snake_case ) -> Optional[Any]:
'''simple docstring'''
# For backward compatibility
if self._in_target_context_manager:
return self.current_processor(*__snake_case , **__snake_case )
if "raw_speech" in kwargs:
warnings.warn('Using `raw_speech` as a keyword argument is deprecated. Use `audio` instead.' )
__a =kwargs.pop('raw_speech' )
else:
__a =kwargs.pop('audio' , __snake_case )
__a =kwargs.pop('sampling_rate' , __snake_case )
__a =kwargs.pop('text' , __snake_case )
if len(__snake_case ) > 0:
__a =args[0]
__a =args[1:]
if audio is None and text is None:
raise ValueError('You need to specify either an `audio` or `text` input to process.' )
if audio is not None:
__a =self.feature_extractor(__snake_case , *__snake_case , sampling_rate=__snake_case , **__snake_case )
if text is not None:
__a =self.tokenizer(__snake_case , **__snake_case )
if text is None:
return inputs
elif audio is None:
return encodings
else:
__a =encodings['input_ids']
return inputs
def __magic_name__ ( self , *__snake_case , **__snake_case ) -> Any:
'''simple docstring'''
return self.tokenizer.batch_decode(*__snake_case , **__snake_case )
def __magic_name__ ( self , *__snake_case , **__snake_case ) -> Optional[int]:
'''simple docstring'''
return self.tokenizer.decode(*__snake_case , **__snake_case )
@contextmanager
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
warnings.warn(
'`as_target_processor` is deprecated and will be removed in v5 of Transformers. You can process your '
'labels by using the argument `text` of the regular `__call__` method (either in the same call as '
'your audio inputs, or in a separate call.' )
__a =True
__a =self.tokenizer
yield
__a =self.feature_extractor
__a =False
| 308
|
from collections import OrderedDict
from typing import Mapping
from packaging import version
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
_lowerCAmelCase : List[str] = logging.get_logger(__name__)
_lowerCAmelCase : List[Any] = {
"hustvl/yolos-small": "https://huggingface.co/hustvl/yolos-small/resolve/main/config.json",
# See all YOLOS models at https://huggingface.co/models?filter=yolos
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'yolos'
def __init__( self , __snake_case=768 , __snake_case=12 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=[512, 864] , __snake_case=16 , __snake_case=3 , __snake_case=True , __snake_case=100 , __snake_case=True , __snake_case=False , __snake_case=1 , __snake_case=5 , __snake_case=2 , __snake_case=5 , __snake_case=2 , __snake_case=0.1 , **__snake_case , ) -> str:
'''simple docstring'''
super().__init__(**__snake_case )
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =initializer_range
__a =layer_norm_eps
__a =image_size
__a =patch_size
__a =num_channels
__a =qkv_bias
__a =num_detection_tokens
__a =use_mid_position_embeddings
__a =auxiliary_loss
# Hungarian matcher
__a =class_cost
__a =bbox_cost
__a =giou_cost
# Loss coefficients
__a =bbox_loss_coefficient
__a =giou_loss_coefficient
__a =eos_coefficient
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = version.parse('1.11' )
@property
def __magic_name__ ( self ) -> Mapping[str, Mapping[int, str]]:
'''simple docstring'''
return OrderedDict(
[
('pixel_values', {0: 'batch', 1: 'num_channels', 2: 'height', 3: 'width'}),
] )
@property
def __magic_name__ ( self ) -> float:
'''simple docstring'''
return 1e-4
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return 12
| 308
| 1
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
_lowerCAmelCase : Optional[int] = {
"configuration_chinese_clip": [
"CHINESE_CLIP_PRETRAINED_CONFIG_ARCHIVE_MAP",
"ChineseCLIPConfig",
"ChineseCLIPOnnxConfig",
"ChineseCLIPTextConfig",
"ChineseCLIPVisionConfig",
],
"processing_chinese_clip": ["ChineseCLIPProcessor"],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Dict = ["ChineseCLIPFeatureExtractor"]
_lowerCAmelCase : str = ["ChineseCLIPImageProcessor"]
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : str = [
"CHINESE_CLIP_PRETRAINED_MODEL_ARCHIVE_LIST",
"ChineseCLIPModel",
"ChineseCLIPPreTrainedModel",
"ChineseCLIPTextModel",
"ChineseCLIPVisionModel",
]
if TYPE_CHECKING:
from .configuration_chinese_clip import (
CHINESE_CLIP_PRETRAINED_CONFIG_ARCHIVE_MAP,
ChineseCLIPConfig,
ChineseCLIPOnnxConfig,
ChineseCLIPTextConfig,
ChineseCLIPVisionConfig,
)
from .processing_chinese_clip import ChineseCLIPProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .feature_extraction_chinese_clip import ChineseCLIPFeatureExtractor, ChineseCLIPImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_chinese_clip import (
CHINESE_CLIP_PRETRAINED_MODEL_ARCHIVE_LIST,
ChineseCLIPModel,
ChineseCLIPPreTrainedModel,
ChineseCLIPTextModel,
ChineseCLIPVisionModel,
)
else:
import sys
_lowerCAmelCase : Any = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
|
import argparse
import torch
from datasets import load_dataset
from donut import DonutModel
from transformers import (
DonutImageProcessor,
DonutProcessor,
DonutSwinConfig,
DonutSwinModel,
MBartConfig,
MBartForCausalLM,
VisionEncoderDecoderModel,
XLMRobertaTokenizerFast,
)
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =model.config
__a =DonutSwinConfig(
image_size=original_config.input_size , patch_size=4 , depths=original_config.encoder_layer , num_heads=[4, 8, 16, 32] , window_size=original_config.window_size , embed_dim=128 , )
__a =MBartConfig(
is_decoder=_snake_case , is_encoder_decoder=_snake_case , add_cross_attention=_snake_case , decoder_layers=original_config.decoder_layer , max_position_embeddings=original_config.max_position_embeddings , vocab_size=len(
model.decoder.tokenizer ) , scale_embedding=_snake_case , add_final_layer_norm=_snake_case , )
return encoder_config, decoder_config
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
if "encoder.model" in name:
__a =name.replace('encoder.model' , 'encoder' )
if "decoder.model" in name:
__a =name.replace('decoder.model' , 'decoder' )
if "patch_embed.proj" in name:
__a =name.replace('patch_embed.proj' , 'embeddings.patch_embeddings.projection' )
if "patch_embed.norm" in name:
__a =name.replace('patch_embed.norm' , 'embeddings.norm' )
if name.startswith('encoder' ):
if "layers" in name:
__a ='encoder.' + name
if "attn.proj" in name:
__a =name.replace('attn.proj' , 'attention.output.dense' )
if "attn" in name and "mask" not in name:
__a =name.replace('attn' , 'attention.self' )
if "norm1" in name:
__a =name.replace('norm1' , 'layernorm_before' )
if "norm2" in name:
__a =name.replace('norm2' , 'layernorm_after' )
if "mlp.fc1" in name:
__a =name.replace('mlp.fc1' , 'intermediate.dense' )
if "mlp.fc2" in name:
__a =name.replace('mlp.fc2' , 'output.dense' )
if name == "encoder.norm.weight":
__a ='encoder.layernorm.weight'
if name == "encoder.norm.bias":
__a ='encoder.layernorm.bias'
return name
def UpperCamelCase_( _snake_case : Tuple , _snake_case : str ):
"""simple docstring"""
for key in orig_state_dict.copy().keys():
__a =orig_state_dict.pop(_snake_case )
if "qkv" in key:
__a =key.split('.' )
__a =int(key_split[3] )
__a =int(key_split[5] )
__a =model.encoder.encoder.layers[layer_num].blocks[block_num].attention.self.all_head_size
if "weight" in key:
__a =val[:dim, :]
__a =val[dim : dim * 2, :]
__a =val[-dim:, :]
else:
__a =val[:dim]
__a =val[dim : dim * 2]
__a =val[-dim:]
elif "attn_mask" in key or key in ["encoder.model.norm.weight", "encoder.model.norm.bias"]:
# HuggingFace implementation doesn't use attn_mask buffer
# and model doesn't use final LayerNorms for the encoder
pass
else:
__a =val
return orig_state_dict
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Union[str, Any]=None , _snake_case : List[Any]=False ):
"""simple docstring"""
__a =DonutModel.from_pretrained(_snake_case ).eval()
# load HuggingFace model
__a , __a =get_configs(_snake_case )
__a =DonutSwinModel(_snake_case )
__a =MBartForCausalLM(_snake_case )
__a =VisionEncoderDecoderModel(encoder=_snake_case , decoder=_snake_case )
model.eval()
__a =original_model.state_dict()
__a =convert_state_dict(_snake_case , _snake_case )
model.load_state_dict(_snake_case )
# verify results on scanned document
__a =load_dataset('hf-internal-testing/example-documents' )
__a =dataset['test'][0]['image'].convert('RGB' )
__a =XLMRobertaTokenizerFast.from_pretrained(_snake_case , from_slow=_snake_case )
__a =DonutImageProcessor(
do_align_long_axis=original_model.config.align_long_axis , size=original_model.config.input_size[::-1] )
__a =DonutProcessor(_snake_case , _snake_case )
__a =processor(_snake_case , return_tensors='pt' ).pixel_values
if model_name == "naver-clova-ix/donut-base-finetuned-docvqa":
__a ='<s_docvqa><s_question>{user_input}</s_question><s_answer>'
__a ='When is the coffee break?'
__a =task_prompt.replace('{user_input}' , _snake_case )
elif model_name == "naver-clova-ix/donut-base-finetuned-rvlcdip":
__a ='<s_rvlcdip>'
elif model_name in [
"naver-clova-ix/donut-base-finetuned-cord-v1",
"naver-clova-ix/donut-base-finetuned-cord-v1-2560",
]:
__a ='<s_cord>'
elif model_name == "naver-clova-ix/donut-base-finetuned-cord-v2":
__a ='s_cord-v2>'
elif model_name == "naver-clova-ix/donut-base-finetuned-zhtrainticket":
__a ='<s_zhtrainticket>'
elif model_name in ["naver-clova-ix/donut-proto", "naver-clova-ix/donut-base"]:
# use a random prompt
__a ='hello world'
else:
raise ValueError('Model name not supported' )
__a =original_model.decoder.tokenizer(_snake_case , add_special_tokens=_snake_case , return_tensors='pt' )[
'input_ids'
]
__a =original_model.encoder.model.patch_embed(_snake_case )
__a , __a =model.encoder.embeddings(_snake_case )
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
# verify encoder hidden states
__a =original_model.encoder(_snake_case )
__a =model.encoder(_snake_case ).last_hidden_state
assert torch.allclose(_snake_case , _snake_case , atol=1e-2 )
# verify decoder hidden states
__a =original_model(_snake_case , _snake_case , _snake_case ).logits
__a =model(_snake_case , decoder_input_ids=_snake_case ).logits
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
print('Looks ok!' )
if pytorch_dump_folder_path is not None:
print(F'Saving model and processor to {pytorch_dump_folder_path}' )
model.save_pretrained(_snake_case )
processor.save_pretrained(_snake_case )
if push_to_hub:
model.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
processor.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
if __name__ == "__main__":
_lowerCAmelCase : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_name",
default="naver-clova-ix/donut-base-finetuned-docvqa",
required=False,
type=str,
help="Name of the original model you'd like to convert.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
required=False,
type=str,
help="Path to the output PyTorch model directory.",
)
parser.add_argument(
"--push_to_hub",
action="store_true",
help="Whether or not to push the converted model and processor to the 🤗 hub.",
)
_lowerCAmelCase : List[Any] = parser.parse_args()
convert_donut_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 308
| 1
|
from __future__ import annotations
from collections.abc import Iterator
class __magic_name__ :
def __init__( self , __snake_case ) -> None:
'''simple docstring'''
__a =value
__a =None
__a =None
class __magic_name__ :
def __init__( self , __snake_case ) -> None:
'''simple docstring'''
__a =tree
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
if node is None:
return 0
return node.value + (
self.depth_first_search(node.left ) + self.depth_first_search(node.right )
)
def __iter__( self ) -> Iterator[int]:
'''simple docstring'''
yield self.depth_first_search(self.tree )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
|
import json
import os
import unittest
from transformers.models.blenderbot_small.tokenization_blenderbot_small import (
VOCAB_FILES_NAMES,
BlenderbotSmallTokenizer,
)
from ...test_tokenization_common import TokenizerTesterMixin
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = BlenderbotSmallTokenizer
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
super().setUp()
__a =['__start__', 'adapt', 'act', 'ap@@', 'te', '__end__', '__unk__']
__a =dict(zip(__snake_case , range(len(__snake_case ) ) ) )
__a =['#version: 0.2', 'a p', 't e</w>', 'ap t</w>', 'a d', 'ad apt</w>', 'a c', 'ac t</w>', '']
__a ={'unk_token': '__unk__', 'bos_token': '__start__', 'eos_token': '__end__'}
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(__snake_case ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(__snake_case ) )
def __magic_name__ ( self , **__snake_case ) -> Any:
'''simple docstring'''
kwargs.update(self.special_tokens_map )
return BlenderbotSmallTokenizer.from_pretrained(self.tmpdirname , **__snake_case )
def __magic_name__ ( self , __snake_case ) -> List[Any]:
'''simple docstring'''
__a ='adapt act apte'
__a ='adapt act apte'
return input_text, output_text
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
__a ='adapt act apte'
__a =['adapt', 'act', 'ap@@', 'te']
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
__a =[tokenizer.bos_token] + tokens + [tokenizer.eos_token]
__a =[0, 1, 2, 3, 4, 5]
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
assert tok('sam' ).input_ids == [1384]
__a ='I am a small frog.'
__a =tok([src_text] , padding=__snake_case , truncation=__snake_case )['input_ids']
__a =tok.batch_decode(__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case )[0]
assert src_text != decoded # I wish it did!
assert decoded == "i am a small frog ."
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
__a ='I am a small frog .'
__a ='.'
__a =tok(__snake_case )['input_ids']
__a =tok(__snake_case )['input_ids']
assert encoded[-1] == encoded_dot[0]
| 308
| 1
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
_lowerCAmelCase : Tuple = {
"configuration_megatron_bert": ["MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MegatronBertConfig"],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : List[str] = [
"MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"MegatronBertForCausalLM",
"MegatronBertForMaskedLM",
"MegatronBertForMultipleChoice",
"MegatronBertForNextSentencePrediction",
"MegatronBertForPreTraining",
"MegatronBertForQuestionAnswering",
"MegatronBertForSequenceClassification",
"MegatronBertForTokenClassification",
"MegatronBertModel",
"MegatronBertPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_megatron_bert import MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MegatronBertConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_megatron_bert import (
MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
MegatronBertForCausalLM,
MegatronBertForMaskedLM,
MegatronBertForMultipleChoice,
MegatronBertForNextSentencePrediction,
MegatronBertForPreTraining,
MegatronBertForQuestionAnswering,
MegatronBertForSequenceClassification,
MegatronBertForTokenClassification,
MegatronBertModel,
MegatronBertPreTrainedModel,
)
else:
import sys
_lowerCAmelCase : Optional[int] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
|
import unittest
from transformers import load_tool
from transformers.utils import is_torch_available
if is_torch_available():
import torch
from transformers.testing_utils import require_torch
from .test_tools_common import ToolTesterMixin
@require_torch
class __magic_name__ ( unittest.TestCase , lowerCAmelCase_ ):
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =load_tool('text-to-speech' )
self.tool.setup()
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
| 308
| 1
|
from collections.abc import Callable
from math import pi, sqrt
from random import uniform
from statistics import mean
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
def is_in_circle(_snake_case : float , _snake_case : float ) -> bool:
__a =sqrt((x**2) + (y**2) )
# Our circle has a radius of 1, so a distance
# greater than 1 would land outside the circle.
return distance_from_centre <= 1
# The proportion of guesses that landed in the circle
__a =mean(
int(is_in_circle(uniform(-1.0 , 1.0 ) , uniform(-1.0 , 1.0 ) ) )
for _ in range(_snake_case ) )
# The ratio of the area for circle to square is pi/4.
__a =proportion * 4
print(F'The estimated value of pi is {pi_estimate}' )
print(F'The numpy value of pi is {pi}' )
print(F'The total error is {abs(pi - pi_estimate )}' )
def UpperCamelCase_( _snake_case : int , _snake_case : Callable[[float], float] , _snake_case : float = 0.0 , _snake_case : float = 1.0 , ):
"""simple docstring"""
return mean(
function_to_integrate(uniform(_snake_case , _snake_case ) ) for _ in range(_snake_case ) ) * (max_value - min_value)
def UpperCamelCase_( _snake_case : int , _snake_case : float = 0.0 , _snake_case : float = 1.0 ):
"""simple docstring"""
def identity_function(_snake_case : float ) -> float:
return x
__a =area_under_curve_estimator(
_snake_case , _snake_case , _snake_case , _snake_case )
__a =(max_value * max_value - min_value * min_value) / 2
print('******************' )
print(F'Estimating area under y=x where x varies from {min_value} to {max_value}' )
print(F'Estimated value is {estimated_value}' )
print(F'Expected value is {expected_value}' )
print(F'Total error is {abs(estimated_value - expected_value )}' )
print('******************' )
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
def function_to_integrate(_snake_case : float ) -> float:
return sqrt(4.0 - x * x )
__a =area_under_curve_estimator(
_snake_case , _snake_case , 0.0 , 2.0 )
print('******************' )
print('Estimating pi using area_under_curve_estimator' )
print(F'Estimated value is {estimated_value}' )
print(F'Expected value is {pi}' )
print(F'Total error is {abs(estimated_value - pi )}' )
print('******************' )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
|
import flax.linen as nn
import jax.numpy as jnp
from .attention_flax import FlaxTransformeraDModel
from .resnet_flax import FlaxDownsampleaD, FlaxResnetBlockaD, FlaxUpsampleaD
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> Optional[Any]:
'''simple docstring'''
__a =()
for resnet, attn in zip(self.resnets , self.attentions ):
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case=True ) -> Optional[int]:
'''simple docstring'''
__a =()
for resnet in self.resnets:
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet, attn in zip(self.resnets , self.attentions ):
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet in self.resnets:
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# there is always at least one resnet
__a =[
FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
]
__a =[]
for _ in range(self.num_layers ):
__a =FlaxTransformeraDModel(
in_channels=self.in_channels , n_heads=self.num_attention_heads , d_head=self.in_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
__a =attentions
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[str]:
'''simple docstring'''
__a =self.resnets[0](__snake_case , __snake_case )
for attn, resnet in zip(self.attentions , self.resnets[1:] ):
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
return hidden_states
| 308
| 1
|
import torch
from diffusers import UnCLIPScheduler
from .test_schedulers import SchedulerCommonTest
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = (UnCLIPScheduler,)
def __magic_name__ ( self , **__snake_case ) -> Dict:
'''simple docstring'''
__a ={
'num_train_timesteps': 1000,
'variance_type': 'fixed_small_log',
'clip_sample': True,
'clip_sample_range': 1.0,
'prediction_type': 'epsilon',
}
config.update(**__snake_case )
return config
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
for timesteps in [1, 5, 100, 1000]:
self.check_over_configs(num_train_timesteps=__snake_case )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
for variance in ["fixed_small_log", "learned_range"]:
self.check_over_configs(variance_type=__snake_case )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
for clip_sample in [True, False]:
self.check_over_configs(clip_sample=__snake_case )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
for clip_sample_range in [1, 5, 10, 20]:
self.check_over_configs(clip_sample_range=__snake_case )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
for prediction_type in ["epsilon", "sample"]:
self.check_over_configs(prediction_type=__snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
for time_step in [0, 500, 999]:
for prev_timestep in [None, 5, 100, 250, 500, 750]:
if prev_timestep is not None and prev_timestep >= time_step:
continue
self.check_over_forward(time_step=__snake_case , prev_timestep=__snake_case )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.scheduler_classes[0]
__a =self.get_scheduler_config(variance_type='fixed_small_log' )
__a =scheduler_class(**__snake_case )
assert torch.sum(torch.abs(scheduler._get_variance(0 ) - 1.0_000e-10 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(487 ) - 0.054_9625 ) ) < 1e-5
assert torch.sum(torch.abs(scheduler._get_variance(999 ) - 0.999_4987 ) ) < 1e-5
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.scheduler_classes[0]
__a =self.get_scheduler_config(variance_type='learned_range' )
__a =scheduler_class(**__snake_case )
__a =0.5
assert scheduler._get_variance(1 , predicted_variance=__snake_case ) - -10.171_2790 < 1e-5
assert scheduler._get_variance(487 , predicted_variance=__snake_case ) - -5.799_8052 < 1e-5
assert scheduler._get_variance(999 , predicted_variance=__snake_case ) - -0.001_0011 < 1e-5
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.scheduler_classes[0]
__a =self.get_scheduler_config()
__a =scheduler_class(**__snake_case )
__a =scheduler.timesteps
__a =self.dummy_model()
__a =self.dummy_sample_deter
__a =torch.manual_seed(0 )
for i, t in enumerate(__snake_case ):
# 1. predict noise residual
__a =model(__snake_case , __snake_case )
# 2. predict previous mean of sample x_t-1
__a =scheduler.step(__snake_case , __snake_case , __snake_case , generator=__snake_case ).prev_sample
__a =pred_prev_sample
__a =torch.sum(torch.abs(__snake_case ) )
__a =torch.mean(torch.abs(__snake_case ) )
assert abs(result_sum.item() - 252.268_2495 ) < 1e-2
assert abs(result_mean.item() - 0.328_4743 ) < 1e-3
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.scheduler_classes[0]
__a =self.get_scheduler_config()
__a =scheduler_class(**__snake_case )
scheduler.set_timesteps(25 )
__a =scheduler.timesteps
__a =self.dummy_model()
__a =self.dummy_sample_deter
__a =torch.manual_seed(0 )
for i, t in enumerate(__snake_case ):
# 1. predict noise residual
__a =model(__snake_case , __snake_case )
if i + 1 == timesteps.shape[0]:
__a =None
else:
__a =timesteps[i + 1]
# 2. predict previous mean of sample x_t-1
__a =scheduler.step(
__snake_case , __snake_case , __snake_case , prev_timestep=__snake_case , generator=__snake_case ).prev_sample
__a =pred_prev_sample
__a =torch.sum(torch.abs(__snake_case ) )
__a =torch.mean(torch.abs(__snake_case ) )
assert abs(result_sum.item() - 258.204_4983 ) < 1e-2
assert abs(result_mean.item() - 0.336_2038 ) < 1e-3
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
pass
| 308
|
import json
import os
import subprocess
import unittest
from ast import literal_eval
import pytest
from parameterized import parameterized, parameterized_class
from . import is_sagemaker_available
if is_sagemaker_available():
from sagemaker import Session, TrainingJobAnalytics
from sagemaker.huggingface import HuggingFace
@pytest.mark.skipif(
literal_eval(os.getenv('TEST_SAGEMAKER' , 'False' ) ) is not True , reason='Skipping test because should only be run when releasing minor transformers version' , )
@pytest.mark.usefixtures('sm_env' )
@parameterized_class(
[
{
'framework': 'pytorch',
'script': 'run_glue.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_5_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'pytorch',
'script': 'run_ddp.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'tensorflow',
'script': 'run_tf_dist.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.6, 'eval_loss': 0.7},
},
] )
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.framework == "pytorch":
subprocess.run(
f'cp ./examples/pytorch/text-classification/run_glue.py {self.env.test_path}/run_glue.py'.split() , encoding='utf-8' , check=__snake_case , )
assert hasattr(self , 'env' )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
__a =f'{self.env.base_job_name}-{instance_count}-{"ddp" if "ddp" in self.script else "smd"}'
# distributed data settings
__a ={'smdistributed': {'dataparallel': {'enabled': True}}} if self.script != 'run_ddp.py' else None
# creates estimator
return HuggingFace(
entry_point=self.script , source_dir=self.env.test_path , role=self.env.role , image_uri=self.env.image_uri , base_job_name=__snake_case , instance_count=__snake_case , instance_type=self.instance_type , debugger_hook_config=__snake_case , hyperparameters={**self.env.distributed_hyperparameters, 'model_name_or_path': self.model_name_or_path} , metric_definitions=self.env.metric_definitions , distribution=__snake_case , py_version='py36' , )
def __magic_name__ ( self , __snake_case ) -> Optional[Any]:
'''simple docstring'''
TrainingJobAnalytics(__snake_case ).export_csv(f'{self.env.test_path}/{job_name}_metrics.csv' )
@parameterized.expand([(2,)] )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
# create estimator
__a =self.create_estimator(__snake_case )
# run training
estimator.fit()
# result dataframe
__a =TrainingJobAnalytics(estimator.latest_training_job.name ).dataframe()
# extract kpis
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_accuracy']['value'] )
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_loss']['value'] )
# get train time from SageMaker job, this includes starting, preprocessing, stopping
__a =(
Session().describe_training_job(estimator.latest_training_job.name ).get('TrainingTimeInSeconds' , 99_9999 )
)
# assert kpis
assert train_runtime <= self.results["train_runtime"]
assert all(t >= self.results['eval_accuracy'] for t in eval_accuracy )
assert all(t <= self.results['eval_loss'] for t in eval_loss )
# dump tests result into json file to share in PR
with open(f'{estimator.latest_training_job.name}.json' , 'w' ) as outfile:
json.dump({'train_time': train_runtime, 'eval_accuracy': eval_accuracy, 'eval_loss': eval_loss} , __snake_case )
| 308
| 1
|
import os
import unittest
from transformers import LayoutLMTokenizer, LayoutLMTokenizerFast
from transformers.models.layoutlm.tokenization_layoutlm import VOCAB_FILES_NAMES
from transformers.testing_utils import require_tokenizers
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = LayoutLMTokenizer
SCREAMING_SNAKE_CASE = LayoutLMTokenizerFast
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = True
def __magic_name__ ( self ) -> int:
'''simple docstring'''
super().setUp()
__a =[
'[UNK]',
'[CLS]',
'[SEP]',
'want',
'##want',
'##ed',
'wa',
'un',
'runn',
'##ing',
',',
'low',
'lowest',
]
__a =os.path.join(self.tmpdirname , 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] ) )
def __magic_name__ ( self , **__snake_case ) -> List[str]:
'''simple docstring'''
return LayoutLMTokenizer.from_pretrained(self.tmpdirname , **__snake_case )
def __magic_name__ ( self , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a ='UNwant\u00E9d,running'
__a ='unwanted, running'
return input_text, output_text
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer_class(self.vocab_file )
__a =tokenizer.tokenize('UNwant\u00E9d,running' )
self.assertListEqual(__snake_case , ['un', '##want', '##ed', ',', 'runn', '##ing'] )
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , [7, 4, 5, 10, 8, 9] )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
pass
| 308
|
import argparse
import logging
from collections import namedtuple
import torch
from model_bertabs import BertAbsSummarizer
from models.model_builder import AbsSummarizer # The authors' implementation
from transformers import BertTokenizer
logging.basicConfig(level=logging.INFO)
_lowerCAmelCase : List[Any] = logging.getLogger(__name__)
_lowerCAmelCase : Optional[Any] = "Hello world! cécé herlolip"
_lowerCAmelCase : str = namedtuple(
"BertAbsConfig",
[
"temp_dir",
"large",
"use_bert_emb",
"finetune_bert",
"encoder",
"share_emb",
"max_pos",
"enc_layers",
"enc_hidden_size",
"enc_heads",
"enc_ff_size",
"enc_dropout",
"dec_layers",
"dec_hidden_size",
"dec_heads",
"dec_ff_size",
"dec_dropout",
],
)
def UpperCamelCase_( _snake_case : str , _snake_case : List[Any] ):
"""simple docstring"""
__a =BertAbsConfig(
temp_dir='.' , finetune_bert=_snake_case , large=_snake_case , share_emb=_snake_case , use_bert_emb=_snake_case , encoder='bert' , max_pos=512 , enc_layers=6 , enc_hidden_size=512 , enc_heads=8 , enc_ff_size=512 , enc_dropout=0.2 , dec_layers=6 , dec_hidden_size=768 , dec_heads=8 , dec_ff_size=2048 , dec_dropout=0.2 , )
__a =torch.load(_snake_case , lambda _snake_case , _snake_case : storage )
__a =AbsSummarizer(_snake_case , torch.device('cpu' ) , _snake_case )
original.eval()
__a =BertAbsSummarizer(_snake_case , torch.device('cpu' ) )
new_model.eval()
# -------------------
# Convert the weights
# -------------------
logging.info('convert the model' )
new_model.bert.load_state_dict(original.bert.state_dict() )
new_model.decoder.load_state_dict(original.decoder.state_dict() )
new_model.generator.load_state_dict(original.generator.state_dict() )
# ----------------------------------
# Make sure the outpus are identical
# ----------------------------------
logging.info('Make sure that the models\' outputs are identical' )
__a =BertTokenizer.from_pretrained('bert-base-uncased' )
# prepare the model inputs
__a =tokenizer.encode('This is sample éàalj\'-.' )
encoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
__a =tokenizer.encode('This is sample 3 éàalj\'-.' )
decoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
# failsafe to make sure the weights reset does not affect the
# loaded weights.
assert torch.max(torch.abs(original.generator[0].weight - new_model.generator[0].weight ) ) == 0
# forward pass
__a =encoder_input_ids
__a =decoder_input_ids
__a =__a =None
__a =None
__a =__a =None
__a =__a =None
__a =None
# The original model does not apply the geneator layer immediatly but rather in
# the beam search (where it combines softmax + linear layer). Since we already
# apply the softmax in our generation process we only apply the linear layer here.
# We make sure that the outputs of the full stack are identical
__a =original(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =original.generator(_snake_case )
__a =new_model(
_snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =new_model.generator(_snake_case )
__a =torch.max(torch.abs(output_converted_model - output_original_model ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.max(torch.abs(output_converted_generator - output_original_generator ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.allclose(_snake_case , _snake_case , atol=1e-3 )
if are_identical:
logging.info('all weights are equal up to 1e-3' )
else:
raise ValueError('the weights are different. The new model is likely different from the original one.' )
# The model has been saved with torch.save(model) and this is bound to the exact
# directory structure. We save the state_dict instead.
logging.info('saving the model\'s state dictionary' )
torch.save(
new_model.state_dict() , './bertabs-finetuned-cnndm-extractive-abstractive-summarization/pytorch_model.bin' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
parser.add_argument(
"--bertabs_checkpoint_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch dump.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
type=str,
required=True,
help="Path to the output PyTorch model.",
)
_lowerCAmelCase : Optional[Any] = parser.parse_args()
convert_bertabs_checkpoints(
args.bertabs_checkpoint_path,
args.pytorch_dump_folder_path,
)
| 308
| 1
|
from typing import List, Optional, Tuple
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import logging
from .tokenization_herbert import HerbertTokenizer
_lowerCAmelCase : List[Any] = logging.get_logger(__name__)
_lowerCAmelCase : Optional[int] = {"vocab_file": "vocab.json", "merges_file": "merges.txt", "tokenizer_file": "tokenizer.json"}
_lowerCAmelCase : List[str] = {
"vocab_file": {
"allegro/herbert-base-cased": "https://huggingface.co/allegro/herbert-base-cased/resolve/main/vocab.json"
},
"merges_file": {
"allegro/herbert-base-cased": "https://huggingface.co/allegro/herbert-base-cased/resolve/main/merges.txt"
},
}
_lowerCAmelCase : str = {"allegro/herbert-base-cased": 514}
_lowerCAmelCase : Optional[Any] = {}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = VOCAB_FILES_NAMES
SCREAMING_SNAKE_CASE = PRETRAINED_VOCAB_FILES_MAP
SCREAMING_SNAKE_CASE = PRETRAINED_INIT_CONFIGURATION
SCREAMING_SNAKE_CASE = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
SCREAMING_SNAKE_CASE = HerbertTokenizer
def __init__( self , __snake_case=None , __snake_case=None , __snake_case=None , __snake_case="<s>" , __snake_case="<unk>" , __snake_case="<pad>" , __snake_case="<mask>" , __snake_case="</s>" , **__snake_case , ) -> Optional[Any]:
'''simple docstring'''
super().__init__(
__snake_case , __snake_case , tokenizer_file=__snake_case , cls_token=__snake_case , unk_token=__snake_case , pad_token=__snake_case , mask_token=__snake_case , sep_token=__snake_case , **__snake_case , )
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> List[int]:
'''simple docstring'''
__a =[self.cls_token_id]
__a =[self.sep_token_id]
if token_ids_a is None:
return cls + token_ids_a + sep
return cls + token_ids_a + sep + token_ids_a + sep
def __magic_name__ ( self , __snake_case , __snake_case = None , __snake_case = False ) -> List[int]:
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=__snake_case , token_ids_a=__snake_case , already_has_special_tokens=__snake_case )
if token_ids_a is None:
return [1] + ([0] * len(__snake_case )) + [1]
return [1] + ([0] * len(__snake_case )) + [1] + ([0] * len(__snake_case )) + [1]
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> List[int]:
'''simple docstring'''
__a =[self.sep_token_id]
__a =[self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep ) * [0] + len(token_ids_a + sep ) * [1]
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> Tuple[str]:
'''simple docstring'''
__a =self._tokenizer.model.save(__snake_case , name=__snake_case )
return tuple(__snake_case )
| 308
|
import unittest
from transformers import XLMConfig, is_torch_available
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 (
XLMForMultipleChoice,
XLMForQuestionAnswering,
XLMForQuestionAnsweringSimple,
XLMForSequenceClassification,
XLMForTokenClassification,
XLMModel,
XLMWithLMHeadModel,
)
from transformers.models.xlm.modeling_xlm import XLM_PRETRAINED_MODEL_ARCHIVE_LIST
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=13 , __snake_case=7 , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=False , __snake_case=False , __snake_case=False , __snake_case=2 , __snake_case=99 , __snake_case=0 , __snake_case=32 , __snake_case=5 , __snake_case=4 , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=2 , __snake_case=0.02 , __snake_case=2 , __snake_case=4 , __snake_case="last" , __snake_case=True , __snake_case=None , __snake_case=0 , ) -> Optional[Any]:
'''simple docstring'''
__a =parent
__a =batch_size
__a =seq_length
__a =is_training
__a =use_input_lengths
__a =use_token_type_ids
__a =use_labels
__a =gelu_activation
__a =sinusoidal_embeddings
__a =causal
__a =asm
__a =n_langs
__a =vocab_size
__a =n_special
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_sequence_label_size
__a =initializer_range
__a =num_labels
__a =num_choices
__a =summary_type
__a =use_proj
__a =scope
__a =bos_token_id
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a =random_attention_mask([self.batch_size, self.seq_length] )
__a =None
if self.use_input_lengths:
__a =(
ids_tensor([self.batch_size] , vocab_size=2 ) + self.seq_length - 2
) # small variation of seq_length
__a =None
if self.use_token_type_ids:
__a =ids_tensor([self.batch_size, self.seq_length] , self.n_langs )
__a =None
__a =None
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a =ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a =ids_tensor([self.batch_size] , 2 ).float()
__a =ids_tensor([self.batch_size] , self.num_choices )
__a =self.get_config()
return (
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
)
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return XLMConfig(
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 , num_labels=self.num_labels , bos_token_id=self.bos_token_id , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMModel(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , lengths=__snake_case , langs=__snake_case )
__a =model(__snake_case , langs=__snake_case )
__a =model(__snake_case )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[int]:
'''simple docstring'''
__a =XLMWithLMHeadModel(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , token_type_ids=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Dict:
'''simple docstring'''
__a =XLMForQuestionAnsweringSimple(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
__a =outputs
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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMForQuestionAnswering(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , p_mask=__snake_case , )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , )
((__a) , ) =result_with_labels.to_tuple()
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
((__a) , ) =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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[Any]:
'''simple docstring'''
__a =XLMForSequenceClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Any:
'''simple docstring'''
__a =self.num_labels
__a =XLMForTokenClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , attention_mask=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Tuple:
'''simple docstring'''
__a =self.num_choices
__a =XLMForMultipleChoice(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =model(
__snake_case , attention_mask=__snake_case , token_type_ids=__snake_case , labels=__snake_case , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) =config_and_inputs
__a ={'input_ids': input_ids, 'token_type_ids': token_type_ids, 'lengths': input_lengths}
return config, inputs_dict
@require_torch
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ , lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
XLMModel,
XLMWithLMHeadModel,
XLMForQuestionAnswering,
XLMForSequenceClassification,
XLMForQuestionAnsweringSimple,
XLMForTokenClassification,
XLMForMultipleChoice,
)
if is_torch_available()
else ()
)
SCREAMING_SNAKE_CASE = (
(XLMWithLMHeadModel,) if is_torch_available() else ()
) # TODO (PVP): Check other models whether language generation is also applicable
SCREAMING_SNAKE_CASE = (
{
'feature-extraction': XLMModel,
'fill-mask': XLMWithLMHeadModel,
'question-answering': XLMForQuestionAnsweringSimple,
'text-classification': XLMForSequenceClassification,
'text-generation': XLMWithLMHeadModel,
'token-classification': XLMForTokenClassification,
'zero-shot': XLMForSequenceClassification,
}
if is_torch_available()
else {}
)
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> int:
'''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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case=False ) -> str:
'''simple docstring'''
__a =super()._prepare_for_class(__snake_case , __snake_case , return_labels=__snake_case )
if return_labels:
if model_class.__name__ == "XLMForQuestionAnswering":
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
return inputs_dict
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMModelTester(self )
__a =ConfigTester(self , config_class=__snake_case , emb_dim=37 )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.config_tester.run_common_tests()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_model(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_lm_head(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_simple_qa(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_qa(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_sequence_classif(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_token_classif(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_for_multiple_choice(*__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Optional[Any]:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_attentions in attentions] , [True] * len(__snake_case ) )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_attentions in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =min_length + idx + 1
__a =(
batch_size * num_beam_groups,
config.num_attention_heads,
tgt_len,
src_len,
)
# check attn size
self.assertListEqual(
[layer_attention.shape for layer_attention in iter_attentions] , [expected_shape] * len(__snake_case ) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Dict:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_hidden_states in hidden_states] , [True] * len(__snake_case ) , )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_hidden_states in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =(batch_size * num_beam_groups, seq_len, config.hidden_size)
# check hidden size
self.assertListEqual(
[layer_hidden_states.shape for layer_hidden_states in iter_hidden_states] , [expected_shape] * len(__snake_case ) , )
pass
@slow
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
for model_name in XLM_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =XLMModel.from_pretrained(__snake_case )
self.assertIsNotNone(__snake_case )
@require_torch
class __magic_name__ ( unittest.TestCase ):
@slow
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMWithLMHeadModel.from_pretrained('xlm-mlm-en-2048' )
model.to(__snake_case )
__a =torch.tensor([[14, 447]] , dtype=torch.long , device=__snake_case ) # the president
__a =[
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
] # the president the president the president the president the president the president the president the president the president the president
# TODO(PVP): this and other input_ids I tried for generation give pretty bad results. Not sure why. Model might just not be made for auto-regressive inference
__a =model.generate(__snake_case , do_sample=__snake_case )
self.assertListEqual(output_ids[0].cpu().numpy().tolist() , __snake_case )
| 308
| 1
|
import inspect
import unittest
from transformers import ConvNextConfig
from transformers.testing_utils import require_torch, require_vision, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_vision_available
from ...test_backbone_common import BackboneTesterMixin
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import ConvNextBackbone, ConvNextForImageClassification, ConvNextModel
from transformers.models.convnext.modeling_convnext import CONVNEXT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
from PIL import Image
from transformers import AutoImageProcessor
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=13 , __snake_case=32 , __snake_case=3 , __snake_case=4 , __snake_case=[10, 20, 30, 40] , __snake_case=[2, 2, 3, 2] , __snake_case=True , __snake_case=True , __snake_case=37 , __snake_case="gelu" , __snake_case=10 , __snake_case=0.02 , __snake_case=["stage2", "stage3", "stage4"] , __snake_case=[2, 3, 4] , __snake_case=None , ) -> Union[str, Any]:
'''simple docstring'''
__a =parent
__a =batch_size
__a =image_size
__a =num_channels
__a =num_stages
__a =hidden_sizes
__a =depths
__a =is_training
__a =use_labels
__a =intermediate_size
__a =hidden_act
__a =num_labels
__a =initializer_range
__a =out_features
__a =out_indices
__a =scope
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.num_labels )
__a =self.get_config()
return config, pixel_values, labels
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
return ConvNextConfig(
num_channels=self.num_channels , hidden_sizes=self.hidden_sizes , depths=self.depths , num_stages=self.num_stages , hidden_act=self.hidden_act , is_decoder=__snake_case , initializer_range=self.initializer_range , out_features=self.out_features , out_indices=self.out_indices , num_labels=self.num_labels , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> str:
'''simple docstring'''
__a =ConvNextModel(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
# expected last hidden states: B, C, H // 32, W // 32
self.parent.assertEqual(
result.last_hidden_state.shape , (self.batch_size, self.hidden_sizes[-1], self.image_size // 32, self.image_size // 32) , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
__a =ConvNextForImageClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , labels=__snake_case )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> int:
'''simple docstring'''
__a =ConvNextBackbone(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
# verify hidden states
self.parent.assertEqual(len(result.feature_maps ) , len(config.out_features ) )
self.parent.assertListEqual(list(result.feature_maps[0].shape ) , [self.batch_size, self.hidden_sizes[1], 4, 4] )
# verify channels
self.parent.assertEqual(len(model.channels ) , len(config.out_features ) )
self.parent.assertListEqual(model.channels , config.hidden_sizes[1:] )
# verify backbone works with out_features=None
__a =None
__a =ConvNextBackbone(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
# verify feature maps
self.parent.assertEqual(len(result.feature_maps ) , 1 )
self.parent.assertListEqual(list(result.feature_maps[0].shape ) , [self.batch_size, self.hidden_sizes[-1], 1, 1] )
# verify channels
self.parent.assertEqual(len(model.channels ) , 1 )
self.parent.assertListEqual(model.channels , [config.hidden_sizes[-1]] )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
__a , __a , __a =config_and_inputs
__a ={'pixel_values': pixel_values}
return config, inputs_dict
@require_torch
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
ConvNextModel,
ConvNextForImageClassification,
ConvNextBackbone,
)
if is_torch_available()
else ()
)
SCREAMING_SNAKE_CASE = (
{'feature-extraction': ConvNextModel, 'image-classification': ConvNextForImageClassification}
if is_torch_available()
else {}
)
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =ConvNextModelTester(self )
__a =ConfigTester(self , config_class=__snake_case , has_text_modality=__snake_case , hidden_size=37 )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
self.create_and_test_config_common_properties()
self.config_tester.create_and_test_config_to_json_string()
self.config_tester.create_and_test_config_to_json_file()
self.config_tester.create_and_test_config_from_and_save_pretrained()
self.config_tester.create_and_test_config_with_num_labels()
self.config_tester.check_config_can_be_init_without_params()
self.config_tester.check_config_arguments_init()
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
return
@unittest.skip(reason='ConvNext does not use inputs_embeds' )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
pass
@unittest.skip(reason='ConvNext does not support input and output embeddings' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
pass
@unittest.skip(reason='ConvNext does not use feedforward chunking' )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a , __a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
__a =model_class(__snake_case )
__a =inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
__a =[*signature.parameters.keys()]
__a =['pixel_values']
self.assertListEqual(arg_names[:1] , __snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_backbone(*__snake_case )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
def check_hidden_states_output(__snake_case , __snake_case , __snake_case ):
__a =model_class(__snake_case )
model.to(__snake_case )
model.eval()
with torch.no_grad():
__a =model(**self._prepare_for_class(__snake_case , __snake_case ) )
__a =outputs.encoder_hidden_states if config.is_encoder_decoder else outputs.hidden_states
__a =self.model_tester.num_stages
self.assertEqual(len(__snake_case ) , expected_num_stages + 1 )
# ConvNext's feature maps are of shape (batch_size, num_channels, height, width)
self.assertListEqual(
list(hidden_states[0].shape[-2:] ) , [self.model_tester.image_size // 4, self.model_tester.image_size // 4] , )
__a , __a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
__a =True
check_hidden_states_output(__snake_case , __snake_case , __snake_case )
# check that output_hidden_states also work using config
del inputs_dict["output_hidden_states"]
__a =True
check_hidden_states_output(__snake_case , __snake_case , __snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*__snake_case )
@slow
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
for model_name in CONVNEXT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =ConvNextModel.from_pretrained(__snake_case )
self.assertIsNotNone(__snake_case )
def UpperCamelCase_( ):
"""simple docstring"""
__a =Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' )
return image
@require_torch
@require_vision
class __magic_name__ ( unittest.TestCase ):
@cached_property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return AutoImageProcessor.from_pretrained('facebook/convnext-tiny-224' ) if is_vision_available() else None
@slow
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =ConvNextForImageClassification.from_pretrained('facebook/convnext-tiny-224' ).to(__snake_case )
__a =self.default_image_processor
__a =prepare_img()
__a =image_processor(images=__snake_case , return_tensors='pt' ).to(__snake_case )
# forward pass
with torch.no_grad():
__a =model(**__snake_case )
# verify the logits
__a =torch.Size((1, 1000) )
self.assertEqual(outputs.logits.shape , __snake_case )
__a =torch.tensor([-0.0260, -0.4739, 0.1911] ).to(__snake_case )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , __snake_case , atol=1e-4 ) )
@require_torch
class __magic_name__ ( unittest.TestCase , lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = (ConvNextBackbone,) if is_torch_available() else ()
SCREAMING_SNAKE_CASE = ConvNextConfig
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =ConvNextModelTester(self )
| 308
|
from __future__ import annotations
import matplotlib.pyplot as plt # type: ignore
import numpy
# initial triangle of Koch snowflake
_lowerCAmelCase : Optional[Any] = numpy.array([0, 0])
_lowerCAmelCase : Dict = numpy.array([0.5, 0.8660254])
_lowerCAmelCase : Any = numpy.array([1, 0])
_lowerCAmelCase : int = [VECTOR_1, VECTOR_2, VECTOR_3, VECTOR_1]
def UpperCamelCase_( _snake_case : list[numpy.ndarray] , _snake_case : int ):
"""simple docstring"""
__a =initial_vectors
for _ in range(_snake_case ):
__a =iteration_step(_snake_case )
return vectors
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =[]
for i, start_vector in enumerate(vectors[:-1] ):
__a =vectors[i + 1]
new_vectors.append(_snake_case )
__a =end_vector - start_vector
new_vectors.append(start_vector + difference_vector / 3 )
new_vectors.append(
start_vector + difference_vector / 3 + rotate(difference_vector / 3 , 60 ) )
new_vectors.append(start_vector + difference_vector * 2 / 3 )
new_vectors.append(vectors[-1] )
return new_vectors
def UpperCamelCase_( _snake_case : numpy.ndarray , _snake_case : float ):
"""simple docstring"""
__a =numpy.radians(_snake_case )
__a , __a =numpy.cos(_snake_case ), numpy.sin(_snake_case )
__a =numpy.array(((c, -s), (s, c)) )
return numpy.dot(_snake_case , _snake_case )
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =plt.gca()
axes.set_aspect('equal' )
# matplotlib.pyplot.plot takes a list of all x-coordinates and a list of all
# y-coordinates as inputs, which are constructed from the vector-list using
# zip()
__a , __a =zip(*_snake_case )
plt.plot(_snake_case , _snake_case )
plt.show()
if __name__ == "__main__":
import doctest
doctest.testmod()
_lowerCAmelCase : List[Any] = iterate(INITIAL_VECTORS, 5)
plot(processed_vectors)
| 308
| 1
|
from heapq import heappop, heappush
import numpy as np
def UpperCamelCase_( _snake_case : np.ndarray , _snake_case : tuple[int, int] , _snake_case : tuple[int, int] , _snake_case : bool , ):
"""simple docstring"""
__a , __a =grid.shape
__a =[-1, 1, 0, 0]
__a =[0, 0, -1, 1]
if allow_diagonal:
dx += [-1, -1, 1, 1]
dy += [-1, 1, -1, 1]
__a , __a =[(0, source)], set()
__a =np.full((rows, cols) , np.inf )
__a =0
__a =np.empty((rows, cols) , dtype=_snake_case )
__a =None
while queue:
((__a) , (__a)) =heappop(_snake_case )
if (x, y) in visited:
continue
visited.add((x, y) )
if (x, y) == destination:
__a =[]
while (x, y) != source:
path.append((x, y) )
__a , __a =predecessors[x, y]
path.append(_snake_case ) # add the source manually
path.reverse()
return matrix[destination], path
for i in range(len(_snake_case ) ):
__a , __a =x + dx[i], y + dy[i]
if 0 <= nx < rows and 0 <= ny < cols:
__a =grid[nx][ny]
if next_node == 1 and matrix[nx, ny] > dist + 1:
heappush(_snake_case , (dist + 1, (nx, ny)) )
__a =dist + 1
__a =(x, y)
return np.inf, []
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
_lowerCAmelCase : Tuple = {
"configuration_megatron_bert": ["MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MegatronBertConfig"],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : List[str] = [
"MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"MegatronBertForCausalLM",
"MegatronBertForMaskedLM",
"MegatronBertForMultipleChoice",
"MegatronBertForNextSentencePrediction",
"MegatronBertForPreTraining",
"MegatronBertForQuestionAnswering",
"MegatronBertForSequenceClassification",
"MegatronBertForTokenClassification",
"MegatronBertModel",
"MegatronBertPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_megatron_bert import MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MegatronBertConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_megatron_bert import (
MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
MegatronBertForCausalLM,
MegatronBertForMaskedLM,
MegatronBertForMultipleChoice,
MegatronBertForNextSentencePrediction,
MegatronBertForPreTraining,
MegatronBertForQuestionAnswering,
MegatronBertForSequenceClassification,
MegatronBertForTokenClassification,
MegatronBertModel,
MegatronBertPreTrainedModel,
)
else:
import sys
_lowerCAmelCase : Optional[int] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 1
|
import os
import pickle
import unittest
from transformers import AutoTokenizer
from transformers.models.bert.tokenization_bert import BertTokenizer
from transformers.models.bert_japanese.tokenization_bert_japanese import (
VOCAB_FILES_NAMES,
BertJapaneseTokenizer,
CharacterTokenizer,
JumanppTokenizer,
MecabTokenizer,
SudachiTokenizer,
WordpieceTokenizer,
)
from transformers.testing_utils import custom_tokenizers, require_jumanpp, require_sudachi
from ...test_tokenization_common import TokenizerTesterMixin
@custom_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = BertJapaneseTokenizer
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = True
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
super().setUp()
__a =[
'[UNK]',
'[CLS]',
'[SEP]',
'こんにちは',
'こん',
'にちは',
'ばんは',
'##こん',
'##にちは',
'##ばんは',
'世界',
'##世界',
'、',
'##、',
'。',
'##。',
]
__a =os.path.join(self.tmpdirname , 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] ) )
def __magic_name__ ( self , __snake_case ) -> List[Any]:
'''simple docstring'''
__a ='こんにちは、世界。 \nこんばんは、世界。'
__a ='こんにちは 、 世界 。 こんばんは 、 世界 。'
return input_text, output_text
def __magic_name__ ( self , __snake_case ) -> List[Any]:
'''simple docstring'''
__a , __a =self.get_input_output_texts(__snake_case )
__a =tokenizer.encode(__snake_case , add_special_tokens=__snake_case )
__a =tokenizer.decode(__snake_case , clean_up_tokenization_spaces=__snake_case )
return text, ids
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
pass # TODO add if relevant
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
pass # TODO add if relevant
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
pass # TODO add if relevant
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.tokenizer_class(self.vocab_file )
__a =tokenizer.tokenize('こんにちは、世界。\nこんばんは、世界。' )
self.assertListEqual(__snake_case , ['こんにちは', '、', '世界', '。', 'こん', '##ばんは', '、', '世界', '。'] )
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , [3, 12, 10, 14, 4, 9, 12, 10, 14] )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.tokenizer_class(self.vocab_file , word_tokenizer_type='mecab' )
self.assertIsNotNone(__snake_case )
__a ='こんにちは、世界。\nこんばんは、世界。'
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , ['こんにちは', '、', '世界', '。', 'こん', '##ばんは', '、', '世界', '。'] )
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , [3, 12, 10, 14, 4, 9, 12, 10, 14] )
__a =os.path.join(self.tmpdirname , 'tokenizer.bin' )
with open(__snake_case , 'wb' ) as handle:
pickle.dump(__snake_case , __snake_case )
with open(__snake_case , 'rb' ) as handle:
__a =pickle.load(__snake_case )
__a =tokenizer_new.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =MecabTokenizer(mecab_dic='ipadic' )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップルストア', 'で', 'iPhone', '8', 'が', '発売', 'さ', 'れ', 'た', '。'] , )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
try:
__a =MecabTokenizer(mecab_dic='unidic_lite' )
except ModuleNotFoundError:
return
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップル', 'ストア', 'で', 'iPhone', '8', 'が', '発売', 'さ', 'れ', 'た', '。'] , )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
try:
__a =MecabTokenizer(mecab_dic='unidic' )
except ModuleNotFoundError:
return
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップル', 'ストア', 'で', 'iPhone', '8', 'が', '発売', 'さ', 'れ', 'た', '。'] , )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =MecabTokenizer(do_lower_case=__snake_case , mecab_dic='ipadic' )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップルストア', 'で', 'iphone', '8', 'が', '発売', 'さ', 'れ', 'た', '。'] , )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
try:
__a =MecabTokenizer(
do_lower_case=__snake_case , normalize_text=__snake_case , mecab_option='-d /usr/local/lib/mecab/dic/jumandic' )
except RuntimeError:
# if dict doesn't exist in the system, previous code raises this error.
return
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['ｱｯﾌﾟﾙストア', 'で', 'iPhone', '８', 'が', '発売', 'さ', 'れた', '\u3000', '。'] , )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =MecabTokenizer(normalize_text=__snake_case , mecab_dic='ipadic' )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['ｱｯﾌﾟﾙストア', 'で', 'iPhone', '８', 'が', '発売', 'さ', 'れ', 'た', ' ', '。'] , )
@require_sudachi
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.tokenizer_class(self.vocab_file , word_tokenizer_type='sudachi' )
self.assertIsNotNone(__snake_case )
__a ='こんにちは、世界。\nこんばんは、世界。'
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , ['こんにちは', '、', '世界', '。', 'こん', '##ばんは', '、', '世界', '。'] )
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , [3, 12, 10, 14, 4, 9, 12, 10, 14] )
__a =os.path.join(self.tmpdirname , 'tokenizer.bin' )
with open(__snake_case , 'wb' ) as handle:
pickle.dump(__snake_case , __snake_case )
with open(__snake_case , 'rb' ) as handle:
__a =pickle.load(__snake_case )
__a =tokenizer_new.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
@require_sudachi
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =SudachiTokenizer(sudachi_dict_type='core' )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , [' ', '\t', 'アップル', 'ストア', 'で', 'iPhone', '8', ' ', 'が', ' ', ' ', '\n ', '発売', 'さ', 'れ', 'た', ' ', '。', ' ', ' '] , )
@require_sudachi
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =SudachiTokenizer(sudachi_dict_type='core' , sudachi_split_mode='A' )
self.assertListEqual(tokenizer.tokenize('外国人参政権' ) , ['外国', '人', '参政', '権'] )
@require_sudachi
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =SudachiTokenizer(sudachi_dict_type='core' , sudachi_split_mode='B' )
self.assertListEqual(tokenizer.tokenize('外国人参政権' ) , ['外国人', '参政権'] )
@require_sudachi
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =SudachiTokenizer(sudachi_dict_type='core' , sudachi_split_mode='C' )
self.assertListEqual(tokenizer.tokenize('外国人参政権' ) , ['外国人参政権'] )
@require_sudachi
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =SudachiTokenizer(do_lower_case=__snake_case , sudachi_dict_type='core' )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , [' ', '\t', 'アップル', 'ストア', 'で', 'iphone', '8', ' ', 'が', ' ', ' ', '\n ', '発売', 'さ', 'れ', 'た', ' ', '。', ' ', ' '] , )
@require_sudachi
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =SudachiTokenizer(normalize_text=__snake_case , sudachi_dict_type='core' )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , [' ', '\t', 'ｱｯﾌﾟﾙ', 'ストア', 'で', 'iPhone', '８', ' ', 'が', ' ', ' ', '\n ', '発売', 'さ', 'れ', 'た', '\u3000', '。', ' ', ' '] , )
@require_sudachi
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =SudachiTokenizer(trim_whitespace=__snake_case , sudachi_dict_type='core' )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップル', 'ストア', 'で', 'iPhone', '8', 'が', '発売', 'さ', 'れ', 'た', '。'] , )
@require_jumanpp
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer_class(self.vocab_file , word_tokenizer_type='jumanpp' )
self.assertIsNotNone(__snake_case )
__a ='こんにちは、世界。\nこんばんは、世界。'
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , ['こんにちは', '、', '世界', '。', 'こん', '##ばんは', '、', '世界', '。'] )
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , [3, 12, 10, 14, 4, 9, 12, 10, 14] )
__a =os.path.join(self.tmpdirname , 'tokenizer.bin' )
with open(__snake_case , 'wb' ) as handle:
pickle.dump(__snake_case , __snake_case )
with open(__snake_case , 'rb' ) as handle:
__a =pickle.load(__snake_case )
__a =tokenizer_new.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
@require_jumanpp
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =JumanppTokenizer()
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップル', 'ストア', 'で', 'iPhone', '8', '\u3000', 'が', '\u3000', '\u3000', '\u3000', '発売', 'さ', 'れた', '\u3000', '。'] , )
@require_jumanpp
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =JumanppTokenizer(do_lower_case=__snake_case )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップル', 'ストア', 'で', 'iphone', '8', '\u3000', 'が', '\u3000', '\u3000', '\u3000', '発売', 'さ', 'れた', '\u3000', '。'] , )
@require_jumanpp
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =JumanppTokenizer(normalize_text=__snake_case )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['ｱ', 'ｯ', 'ﾌ', 'ﾟ', 'ﾙ', 'ストア', 'で', 'iPhone', '８', '\u3000', 'が', '\u3000', '\u3000', '\u3000', '発売', 'さ', 'れた', '\u3000', '。'] , )
@require_jumanpp
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =JumanppTokenizer(trim_whitespace=__snake_case )
self.assertListEqual(
tokenizer.tokenize(' \tｱｯﾌﾟﾙストアでiPhone８ が \n 発売された 。 ' ) , ['アップル', 'ストア', 'で', 'iPhone', '8', 'が', '発売', 'さ', 'れた', '。'] , )
@require_jumanpp
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =JumanppTokenizer()
self.assertListEqual(
tokenizer.tokenize('ありがとうございますm(_ _)ｍ見つけるのが大変です。' ) , ['ありがとう', 'ございます', 'm(_ _)m', '見つける', 'の', 'が', '大変です', '。'] , )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =['[UNK]', '[CLS]', '[SEP]', 'こんにちは', 'こん', 'にちは', 'ばんは', '##こん', '##にちは', '##ばんは']
__a ={}
for i, token in enumerate(__snake_case ):
__a =i
__a =WordpieceTokenizer(vocab=__snake_case , unk_token='[UNK]' )
self.assertListEqual(tokenizer.tokenize('' ) , [] )
self.assertListEqual(tokenizer.tokenize('こんにちは' ) , ['こんにちは'] )
self.assertListEqual(tokenizer.tokenize('こんばんは' ) , ['こん', '##ばんは'] )
self.assertListEqual(tokenizer.tokenize('こんばんは こんばんにちは こんにちは' ) , ['こん', '##ばんは', '[UNK]', 'こんにちは'] )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =BertJapaneseTokenizer.from_pretrained('nlp-waseda/roberta-base-japanese-with-auto-jumanpp' )
__a =tokenizer.subword_tokenizer
__a =subword_tokenizer.tokenize('国境 の 長い トンネル を 抜ける と 雪国 であった 。' )
self.assertListEqual(__snake_case , ['▁国境', '▁の', '▁長い', '▁トンネル', '▁を', '▁抜ける', '▁と', '▁雪', '国', '▁であった', '▁。'] )
__a =subword_tokenizer.tokenize('こんばんは こんばん にち は こんにちは' )
self.assertListEqual(__snake_case , ['▁こん', 'ばん', 'は', '▁こん', 'ばん', '▁に', 'ち', '▁は', '▁こんにちは'] )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.tokenizer_class.from_pretrained('cl-tohoku/bert-base-japanese' )
__a =tokenizer.encode('ありがとう。' , add_special_tokens=__snake_case )
__a =tokenizer.encode('どういたしまして。' , add_special_tokens=__snake_case )
__a =tokenizer.build_inputs_with_special_tokens(__snake_case )
__a =tokenizer.build_inputs_with_special_tokens(__snake_case , __snake_case )
# 2 is for "[CLS]", 3 is for "[SEP]"
assert encoded_sentence == [2] + text + [3]
assert encoded_pair == [2] + text + [3] + text_a + [3]
@custom_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = BertJapaneseTokenizer
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> str:
'''simple docstring'''
super().setUp()
__a =['[UNK]', '[CLS]', '[SEP]', 'こ', 'ん', 'に', 'ち', 'は', 'ば', '世', '界', '、', '。']
__a =os.path.join(self.tmpdirname , 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] ) )
def __magic_name__ ( self , **__snake_case ) -> Any:
'''simple docstring'''
return BertJapaneseTokenizer.from_pretrained(self.tmpdirname , subword_tokenizer_type='character' , **__snake_case )
def __magic_name__ ( self , __snake_case ) -> Tuple:
'''simple docstring'''
__a ='こんにちは、世界。 \nこんばんは、世界。'
__a ='こ ん に ち は 、 世 界 。 こ ん ば ん は 、 世 界 。'
return input_text, output_text
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
pass # TODO add if relevant
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
pass # TODO add if relevant
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
pass # TODO add if relevant
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =self.tokenizer_class(self.vocab_file , subword_tokenizer_type='character' )
__a =tokenizer.tokenize('こんにちは、世界。 \nこんばんは、世界。' )
self.assertListEqual(
__snake_case , ['こ', 'ん', 'に', 'ち', 'は', '、', '世', '界', '。', 'こ', 'ん', 'ば', 'ん', 'は', '、', '世', '界', '。'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(__snake_case ) , [3, 4, 5, 6, 7, 11, 9, 10, 12, 3, 4, 8, 4, 7, 11, 9, 10, 12] )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =['[UNK]', '[CLS]', '[SEP]', 'こ', 'ん', 'に', 'ち', 'は', 'ば', '世', '界', '、', '。']
__a ={}
for i, token in enumerate(__snake_case ):
__a =i
__a =CharacterTokenizer(vocab=__snake_case , unk_token='[UNK]' )
self.assertListEqual(tokenizer.tokenize('' ) , [] )
self.assertListEqual(tokenizer.tokenize('こんにちは' ) , ['こ', 'ん', 'に', 'ち', 'は'] )
self.assertListEqual(tokenizer.tokenize('こんにちほ' ) , ['こ', 'ん', 'に', 'ち', '[UNK]'] )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer_class.from_pretrained('cl-tohoku/bert-base-japanese-char' )
__a =tokenizer.encode('ありがとう。' , add_special_tokens=__snake_case )
__a =tokenizer.encode('どういたしまして。' , add_special_tokens=__snake_case )
__a =tokenizer.build_inputs_with_special_tokens(__snake_case )
__a =tokenizer.build_inputs_with_special_tokens(__snake_case , __snake_case )
# 2 is for "[CLS]", 3 is for "[SEP]"
assert encoded_sentence == [2] + text + [3]
assert encoded_pair == [2] + text + [3] + text_a + [3]
@custom_tokenizers
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a ='cl-tohoku/bert-base-japanese'
__a =AutoTokenizer.from_pretrained(__snake_case )
self.assertIsInstance(__snake_case , __snake_case )
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a ='cl-tohoku/bert-base-japanese'
with self.assertLogs('transformers' , level='WARNING' ) as cm:
BertTokenizer.from_pretrained(__snake_case )
self.assertTrue(
cm.records[0].message.startswith(
'The tokenizer class you load from this checkpoint is not the same type as the class this function'
' is called from.' ) )
__a ='bert-base-cased'
with self.assertLogs('transformers' , level='WARNING' ) as cm:
BertJapaneseTokenizer.from_pretrained(__snake_case )
self.assertTrue(
cm.records[0].message.startswith(
'The tokenizer class you load from this checkpoint is not the same type as the class this function'
' is called from.' ) )
| 308
|
import hashlib
import unittest
from transformers import MODEL_FOR_DEPTH_ESTIMATION_MAPPING, is_torch_available, is_vision_available
from transformers.pipelines import DepthEstimationPipeline, pipeline
from transformers.testing_utils import (
is_pipeline_test,
nested_simplify,
require_tf,
require_timm,
require_torch,
require_vision,
slow,
)
from .test_pipelines_common import ANY
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
else:
class __magic_name__ :
@staticmethod
def __magic_name__ ( *__snake_case , **__snake_case ) -> List[str]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : Image ):
"""simple docstring"""
__a =hashlib.mda(image.tobytes() )
return m.hexdigest()
@is_pipeline_test
@require_vision
@require_timm
@require_torch
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = MODEL_FOR_DEPTH_ESTIMATION_MAPPING
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Dict:
'''simple docstring'''
__a =DepthEstimationPipeline(model=__snake_case , image_processor=__snake_case )
return depth_estimator, [
"./tests/fixtures/tests_samples/COCO/000000039769.png",
"./tests/fixtures/tests_samples/COCO/000000039769.png",
]
def __magic_name__ ( self , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
__a =depth_estimator('./tests/fixtures/tests_samples/COCO/000000039769.png' )
self.assertEqual({'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )} , __snake_case )
import datasets
__a =datasets.load_dataset('hf-internal-testing/fixtures_image_utils' , 'image' , split='test' )
__a =depth_estimator(
[
Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' ),
'http://images.cocodataset.org/val2017/000000039769.jpg',
# RGBA
dataset[0]['file'],
# LA
dataset[1]['file'],
# L
dataset[2]['file'],
] )
self.assertEqual(
[
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
] , __snake_case , )
@require_tf
@unittest.skip('Depth estimation is not implemented in TF' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
pass
@slow
@require_torch
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a ='Intel/dpt-large'
__a =pipeline('depth-estimation' , model=__snake_case )
__a =depth_estimator('http://images.cocodataset.org/val2017/000000039769.jpg' )
__a =hashimage(outputs['depth'] )
# This seems flaky.
# self.assertEqual(outputs["depth"], "1a39394e282e9f3b0741a90b9f108977")
self.assertEqual(nested_simplify(outputs['predicted_depth'].max().item() ) , 29.304 )
self.assertEqual(nested_simplify(outputs['predicted_depth'].min().item() ) , 2.662 )
@require_torch
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
# This is highly irregular to have no small tests.
self.skipTest('There is not hf-internal-testing tiny model for either GLPN nor DPT' )
| 308
| 1
|
def UpperCamelCase_( _snake_case : int , _snake_case : int ):
"""simple docstring"""
if a < 0 or b < 0:
raise ValueError('the value of both inputs must be positive' )
__a =str(bin(_snake_case ) )[2:] # remove the leading "0b"
__a =str(bin(_snake_case ) )[2:] # remove the leading "0b"
__a =max(len(_snake_case ) , len(_snake_case ) )
return "0b" + "".join(
str(int(char_a != char_b ) )
for char_a, char_b in zip(a_binary.zfill(_snake_case ) , b_binary.zfill(_snake_case ) ) )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
|
import argparse
import dataclasses
import json
import logging
import os
import shutil
from typing import List, Optional
import datasets
from accelerate import Accelerator
from datasets import load_dataset
from finetuning import finetune
from tqdm.auto import tqdm
import transformers
from transformers import AutoConfig, set_seed
from transformers.trainer_utils import IntervalStrategy
_lowerCAmelCase : Optional[int] = logging.getLogger(__name__)
_lowerCAmelCase : Any = "pytorch_model.bin"
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co.'} , )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the data to predict on.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The name of the task to train on.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The list of labels for the task.'} )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'The output directory where the model predictions and checkpoints will be written.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='accuracy' , metadata={'help': 'The evaluation metric used for the task.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='no' , metadata={
'help': 'The evaluation strategy to adopt during training. Possible values are: ["no", "step", "epoch]'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={
'help': 'How much the specified evaluation metric must improve to satisfy early stopping conditions.'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the confidence score.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the validation performance.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to fine-tune on labeled data after pseudo training.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={'help': 'Confidence threshold for pseudo-labeled data filtering.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Random seed for initialization.'} , )
def UpperCamelCase_( _snake_case : int , _snake_case : str , _snake_case : Optional[int] , _snake_case : str , _snake_case : Union[str, Any] , _snake_case : List[Any] ):
"""simple docstring"""
__a =datasets.concatenate_datasets([infer_input, infer_output] , axis=1 )
if args.do_filter_by_confidence:
__a =dataset.filter(lambda _snake_case : example["probability"] > args.confidence_threshold )
if args.do_filter_by_val_performance:
assert eval_result >= 0.0 and eval_result <= 1.0
__a =int(eval_result * len(_snake_case ) )
print(_snake_case )
__a =dataset.sort('probability' , reverse=_snake_case )
__a =dataset.select(range(_snake_case ) )
__a =dataset.remove_columns(['label', 'probability'] )
__a =dataset.rename_column('prediction' , 'label' )
__a =dataset.map(lambda _snake_case : {"label": idalabel[example["label"]]} )
__a =dataset.shuffle(seed=args.seed )
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.data_file_extension == "csv":
dataset.to_csv(_snake_case , index=_snake_case )
else:
dataset.to_json(_snake_case )
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : str , _snake_case : int , _snake_case : Optional[int] , **_snake_case : List[str] ):
"""simple docstring"""
__a =Accelerator()
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , level=logging.INFO , )
logger.info(accelerator.state )
# Setup logging, we only want one process per machine to log things on the
# screen. accelerator.is_local_main_process is only True for one process per
# machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR )
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
__a =STModelArguments(model_name_or_path=_snake_case )
__a =STDataArguments(train_file=_snake_case , infer_file=_snake_case )
__a =STTrainingArguments(output_dir=_snake_case )
__a =argparse.Namespace()
for arg_class in (model_args, data_args, training_args):
for key, value in vars(_snake_case ).items():
setattr(_snake_case , _snake_case , _snake_case )
for key, value in kwargs.items():
if hasattr(_snake_case , _snake_case ):
setattr(_snake_case , _snake_case , _snake_case )
# Sanity checks
__a ={}
__a =None
# You need to provide the training data and the data to predict on
assert args.train_file is not None
assert args.infer_file is not None
__a =args.train_file
__a =args.infer_file
if args.evaluation_strategy != IntervalStrategy.NO.value:
assert args.eval_file is not None
__a =args.eval_file
for key in data_files:
__a =data_files[key].split('.' )[-1]
assert extension in ["csv", "json"], F'`{key}_file` should be a csv or a json file.'
if args.data_file_extension is None:
__a =extension
else:
assert extension == args.data_file_extension, F'`{key}_file` should be a {args.data_file_extension} file`.'
assert (
args.eval_metric in datasets.list_metrics()
), F'{args.eval_metric} not in the list of supported metrics {datasets.list_metrics()}.'
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed )
logger.info('Creating the initial data directory for self-training...' )
__a =F'{args.output_dir}/self-train_iter-{{}}'.format
__a =data_dir_format(0 )
if accelerator.is_main_process:
if args.output_dir is not None:
os.makedirs(args.output_dir , exist_ok=_snake_case )
os.makedirs(_snake_case , exist_ok=_snake_case )
accelerator.wait_for_everyone()
__a =None
__a =None
__a =0
__a =False
# Show the progress bar
__a =tqdm(range(args.max_selftrain_iterations ) , disable=not accelerator.is_local_main_process )
# Self-train
for iteration in range(0 , int(args.max_selftrain_iterations ) ):
__a =data_dir_format(_snake_case )
assert os.path.exists(_snake_case )
# Stage 1: initial fine-tuning for iteration = 0 or pseudo-training for
# iteration > 0
__a =os.path.join(_snake_case , 'stage-1' )
__a ={
'accelerator': accelerator,
'model_name_or_path': args.model_name_or_path,
'cache_dir': args.cache_dir,
'do_train': True,
'train_file': data_files['train'] if iteration == 0 else data_files['train_pseudo'],
'do_eval': True if args.eval_file is not None else False,
'eval_file': data_files['eval'],
'do_predict': True,
'infer_file': data_files['infer'],
'task_name': args.task_name,
'label_list': args.label_list,
'output_dir': current_output_dir,
'eval_metric': args.eval_metric,
'evaluation_strategy': args.evaluation_strategy,
'early_stopping_patience': args.early_stopping_patience,
'early_stopping_threshold': args.early_stopping_threshold,
'seed': args.seed,
}
# Add additional training arguments
for key, value in kwargs.items():
if key not in arguments_dict and not hasattr(_snake_case , _snake_case ):
arguments_dict.update({key: value} )
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 1.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 1 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 1.' , _snake_case )
if iteration > 0 and args.finetune_on_labeled_data:
# Stage 2 (optional): fine-tuning on the original labeled data
__a =os.path.join(_snake_case , 'best-checkpoint' )
__a =os.path.join(_snake_case , 'stage-2' )
# Update arguments_dict
__a =model_path
__a =data_files['train']
__a =current_output_dir
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 2.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 2 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 2.' , _snake_case )
__a =iteration
__a =data_dir_format(iteration + 1 )
__a =AutoConfig.from_pretrained(os.path.join(_snake_case , 'best-checkpoint' ) )
__a =config.idalabel
__a =os.path.join(_snake_case , 'eval_results_best-checkpoint.json' )
__a =os.path.join(_snake_case , 'test_results_best-checkpoint.json' )
assert os.path.exists(_snake_case )
with open(_snake_case , 'r' ) as f:
__a =float(json.load(_snake_case )[args.eval_metric] )
__a =os.path.join(_snake_case , 'infer_output_best-checkpoint.csv' )
assert os.path.exists(_snake_case )
# Loading the dataset from local csv or json files.
__a =load_dataset(args.data_file_extension , data_files={'data': data_files['infer']} )['data']
__a =load_dataset('csv' , data_files={'data': infer_output_file} )['data']
if accelerator.is_main_process:
os.makedirs(_snake_case , exist_ok=_snake_case )
shutil.copy(_snake_case , os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) )
if os.path.exists(_snake_case ):
shutil.copy(_snake_case , os.path.join(_snake_case , F'test_results_iter-{iteration}.json' ) )
create_pseudo_labeled_data(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )
accelerator.wait_for_everyone()
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.evaluation_strategy != IntervalStrategy.NO.value:
__a =eval_result
if best_iteration is None:
__a =new_iteration
__a =new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
__a =new_iteration
__a =new_eval_result
__a =0
else:
if new_eval_result == best_eval_result:
__a =new_iteration
__a =new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
__a =True
progress_bar.update(1 )
if should_training_stop:
break
if best_iteration is not None:
# Save the best iteration
logger.info('Best iteration: %d' , _snake_case )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
else:
# Assume that the last iteration is the best
logger.info('Best iteration: %d' , args.max_selftrain_iterations - 1 )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{args.max_selftrain_iterations - 1}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
| 308
| 1
|
import logging
import os
import sys
from dataclasses import dataclass, field
from typing import Optional
import torch
from datasets import load_dataset
from torchvision.transforms import Compose, Lambda, Normalize, RandomHorizontalFlip, RandomResizedCrop, ToTensor
from torchvision.transforms.functional import InterpolationMode
import transformers
from transformers import (
HfArgumentParser,
Trainer,
TrainingArguments,
ViTImageProcessor,
ViTMAEConfig,
ViTMAEForPreTraining,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version, send_example_telemetry
from transformers.utils.versions import require_version
_lowerCAmelCase : List[str] = logging.getLogger(__name__)
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.31.0")
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/image-pretraining/requirements.txt")
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default='cifar10' , metadata={'help': 'Name of a dataset from the datasets package'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'The column name of the images in the files.'} )
SCREAMING_SNAKE_CASE = field(default=lowerCAmelCase_ , metadata={'help': 'A folder containing the training data.'} )
SCREAMING_SNAKE_CASE = field(default=lowerCAmelCase_ , metadata={'help': 'A folder containing the validation data.'} )
SCREAMING_SNAKE_CASE = field(
default=0.1_5 , metadata={'help': 'Percent to split off of train for validation.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of training examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of evaluation examples to this '
'value if set.'
)
} , )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a ={}
if self.train_dir is not None:
__a =self.train_dir
if self.validation_dir is not None:
__a =self.validation_dir
__a =data_files if data_files else None
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'The model checkpoint for weights initialization.Don\'t set if you want to train a model from scratch.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained config name or path if not the same as model_name_or_path'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Override some existing default config settings when a model is trained from scratch. Example: '
'n_embd=10,resid_pdrop=0.2,scale_attn_weights=false,summary_type=cls_index'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from s3'} )
SCREAMING_SNAKE_CASE = field(
default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , )
SCREAMING_SNAKE_CASE = field(default=lowerCAmelCase_ , metadata={'help': 'Name or path of preprocessor config.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Will use the token generated when running `huggingface-cli login` (necessary to use this script '
'with private models).'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=0.7_5 , metadata={'help': 'The ratio of the number of masked tokens in the input sequence.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Whether or not to train with normalized pixel values as target.'} )
@dataclass
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = field(
default=1E-3 , metadata={'help': 'Base learning rate: absolute_lr = base_lr * total_batch_size / 256.'} )
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =torch.stack([example['pixel_values'] for example in examples] )
return {"pixel_values": pixel_values}
def UpperCamelCase_( ):
"""simple docstring"""
__a =HfArgumentParser((ModelArguments, DataTrainingArguments, CustomTrainingArguments) )
if len(sys.argv ) == 2 and sys.argv[1].endswith('.json' ):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
__a , __a , __a =parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) )
else:
__a , __a , __a =parser.parse_args_into_dataclasses()
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry('run_mae' , _snake_case , _snake_case )
# Setup logging
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , handlers=[logging.StreamHandler(sys.stdout )] , )
if training_args.should_log:
# The default of training_args.log_level is passive, so we set log level at info here to have that default.
transformers.utils.logging.set_verbosity_info()
__a =training_args.get_process_log_level()
logger.setLevel(_snake_case )
transformers.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Log on each process the small summary:
logger.warning(
F'Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}'
+ F'distributed training: {bool(training_args.local_rank != -1 )}, 16-bits training: {training_args.fpaa}' )
logger.info(F'Training/evaluation parameters {training_args}' )
# Detecting last checkpoint.
__a =None
if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir:
__a =get_last_checkpoint(training_args.output_dir )
if last_checkpoint is None and len(os.listdir(training_args.output_dir ) ) > 0:
raise ValueError(
F'Output directory ({training_args.output_dir}) already exists and is not empty. '
'Use --overwrite_output_dir to overcome.' )
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
F'Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change '
'the `--output_dir` or add `--overwrite_output_dir` to train from scratch.' )
# Initialize our dataset.
__a =load_dataset(
data_args.dataset_name , data_args.dataset_config_name , data_files=data_args.data_files , cache_dir=model_args.cache_dir , use_auth_token=True if model_args.use_auth_token else None , )
# If we don't have a validation split, split off a percentage of train as validation.
__a =None if 'validation' in ds.keys() else data_args.train_val_split
if isinstance(data_args.train_val_split , _snake_case ) and data_args.train_val_split > 0.0:
__a =ds['train'].train_test_split(data_args.train_val_split )
__a =split['train']
__a =split['test']
# Load pretrained model and image processor
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
__a ={
'cache_dir': model_args.cache_dir,
'revision': model_args.model_revision,
'use_auth_token': True if model_args.use_auth_token else None,
}
if model_args.config_name:
__a =ViTMAEConfig.from_pretrained(model_args.config_name , **_snake_case )
elif model_args.model_name_or_path:
__a =ViTMAEConfig.from_pretrained(model_args.model_name_or_path , **_snake_case )
else:
__a =ViTMAEConfig()
logger.warning('You are instantiating a new config instance from scratch.' )
if model_args.config_overrides is not None:
logger.info(F'Overriding config: {model_args.config_overrides}' )
config.update_from_string(model_args.config_overrides )
logger.info(F'New config: {config}' )
# adapt config
config.update(
{
'mask_ratio': model_args.mask_ratio,
'norm_pix_loss': model_args.norm_pix_loss,
} )
# create image processor
if model_args.image_processor_name:
__a =ViTImageProcessor.from_pretrained(model_args.image_processor_name , **_snake_case )
elif model_args.model_name_or_path:
__a =ViTImageProcessor.from_pretrained(model_args.model_name_or_path , **_snake_case )
else:
__a =ViTImageProcessor()
# create model
if model_args.model_name_or_path:
__a =ViTMAEForPreTraining.from_pretrained(
model_args.model_name_or_path , from_tf=bool('.ckpt' in model_args.model_name_or_path ) , config=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
else:
logger.info('Training new model from scratch' )
__a =ViTMAEForPreTraining(_snake_case )
if training_args.do_train:
__a =ds['train'].column_names
else:
__a =ds['validation'].column_names
if data_args.image_column_name is not None:
__a =data_args.image_column_name
elif "image" in column_names:
__a ='image'
elif "img" in column_names:
__a ='img'
else:
__a =column_names[0]
# transformations as done in original MAE paper
# source: https://github.com/facebookresearch/mae/blob/main/main_pretrain.py
if "shortest_edge" in image_processor.size:
__a =image_processor.size['shortest_edge']
else:
__a =(image_processor.size['height'], image_processor.size['width'])
__a =Compose(
[
Lambda(lambda _snake_case : img.convert('RGB' ) if img.mode != "RGB" else img ),
RandomResizedCrop(_snake_case , scale=(0.2, 1.0) , interpolation=InterpolationMode.BICUBIC ),
RandomHorizontalFlip(),
ToTensor(),
Normalize(mean=image_processor.image_mean , std=image_processor.image_std ),
] )
def preprocess_images(_snake_case : Any ):
__a =[transforms(_snake_case ) for image in examples[image_column_name]]
return examples
if training_args.do_train:
if "train" not in ds:
raise ValueError('--do_train requires a train dataset' )
if data_args.max_train_samples is not None:
__a =ds['train'].shuffle(seed=training_args.seed ).select(range(data_args.max_train_samples ) )
# Set the training transforms
ds["train"].set_transform(_snake_case )
if training_args.do_eval:
if "validation" not in ds:
raise ValueError('--do_eval requires a validation dataset' )
if data_args.max_eval_samples is not None:
__a =(
ds['validation'].shuffle(seed=training_args.seed ).select(range(data_args.max_eval_samples ) )
)
# Set the validation transforms
ds["validation"].set_transform(_snake_case )
# Compute absolute learning rate
__a =(
training_args.train_batch_size * training_args.gradient_accumulation_steps * training_args.world_size
)
if training_args.base_learning_rate is not None:
__a =training_args.base_learning_rate * total_train_batch_size / 256
# Initialize our trainer
__a =Trainer(
model=_snake_case , args=_snake_case , train_dataset=ds['train'] if training_args.do_train else None , eval_dataset=ds['validation'] if training_args.do_eval else None , tokenizer=_snake_case , data_collator=_snake_case , )
# Training
if training_args.do_train:
__a =None
if training_args.resume_from_checkpoint is not None:
__a =training_args.resume_from_checkpoint
elif last_checkpoint is not None:
__a =last_checkpoint
__a =trainer.train(resume_from_checkpoint=_snake_case )
trainer.save_model()
trainer.log_metrics('train' , train_result.metrics )
trainer.save_metrics('train' , train_result.metrics )
trainer.save_state()
# Evaluation
if training_args.do_eval:
__a =trainer.evaluate()
trainer.log_metrics('eval' , _snake_case )
trainer.save_metrics('eval' , _snake_case )
# Write model card and (optionally) push to hub
__a ={
'tasks': 'masked-auto-encoding',
'dataset': data_args.dataset_name,
'tags': ['masked-auto-encoding'],
}
if training_args.push_to_hub:
trainer.push_to_hub(**_snake_case )
else:
trainer.create_model_card(**_snake_case )
def UpperCamelCase_( _snake_case : List[str] ):
"""simple docstring"""
main()
if __name__ == "__main__":
main()
| 308
|
import shutil
import tempfile
import unittest
from transformers import (
SPIECE_UNDERLINE,
AddedToken,
BatchEncoding,
NllbTokenizer,
NllbTokenizerFast,
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
_lowerCAmelCase : Any = get_tests_dir("fixtures/test_sentencepiece.model")
if is_torch_available():
from transformers.models.mam_aaa.modeling_mam_aaa import shift_tokens_right
_lowerCAmelCase : List[Any] = 256_047
_lowerCAmelCase : Dict = 256_145
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = NllbTokenizer
SCREAMING_SNAKE_CASE = NllbTokenizerFast
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = {}
def __magic_name__ ( self ) -> int:
'''simple docstring'''
super().setUp()
# We have a SentencePiece fixture for testing
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
tokenizer.save_pretrained(self.tmpdirname )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
__a =tokenizer.tokenize('This is a test' )
self.assertListEqual(__snake_case , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(__snake_case ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
__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 =tokenizer.convert_tokens_to_ids(__snake_case )
self.assertListEqual(
__snake_case , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
__a =tokenizer.convert_ids_to_tokens(__snake_case )
self.assertListEqual(
__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 __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =(self.rust_tokenizer_class, 'hf-internal-testing/tiny-random-nllb', {})
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =self.rust_tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =self.tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tuple(f for f in tokenizer_r_files if 'tokenizer.json' not in f )
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=True
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__snake_case )
# Checks it save with the same files
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=False
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
if not self.test_seqaseq:
return
__a =self.get_tokenizers()
for tokenizer in tokenizers:
with self.subTest(f'{tokenizer.__class__.__name__}' ):
# Longer text that will definitely require truncation.
__a =[
' 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.',
]
__a =[
'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.',
]
try:
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , tgt_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='ron_Latn' , )
except NotImplementedError:
return
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 10 )
# max_target_length will default to max_length if not specified
__a =tokenizer.prepare_seqaseq_batch(
__snake_case , tgt_texts=__snake_case , max_length=3 , return_tensors='pt' )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 3 )
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' )
self.assertEqual(batch_encoder_only.input_ids.shape[1] , 3 )
self.assertEqual(batch_encoder_only.attention_mask.shape[1] , 3 )
self.assertNotIn('decoder_input_ids' , __snake_case )
@unittest.skip('Unfortunately way too slow to build a BPE with SentencePiece.' )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =[AddedToken('<special>' , lstrip=__snake_case )]
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_r.encode('Hey this is a <special> token' )
__a =tokenizer_r.encode('<special>' , add_special_tokens=__snake_case )[0]
self.assertTrue(special_token_id in r_output )
if self.test_slow_tokenizer:
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case , )
__a =self.tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_p.encode('Hey this is a <special> token' )
__a =tokenizer_cr.encode('Hey this is a <special> token' )
self.assertEqual(__snake_case , __snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertTrue(special_token_id in p_output )
self.assertTrue(special_token_id in cr_output )
@require_torch
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = 'facebook/nllb-200-distilled-600M'
SCREAMING_SNAKE_CASE = [
' 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 = [
'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 = [
2_5_6_0_4_7,
1_6_2_9_7,
1_3_4_4_0_8,
8_1_6_5,
2_4_8_0_6_6,
1_4_7_3_4,
9_5_0,
1_1_3_5,
1_0_5_7_2_1,
3_5_7_3,
8_3,
2_7_3_5_2,
1_0_8,
4_9_4_8_6,
2,
]
@classmethod
def __magic_name__ ( cls ) -> Tuple:
'''simple docstring'''
__a =NllbTokenizer.from_pretrained(
cls.checkpoint_name , src_lang='eng_Latn' , tgt_lang='ron_Latn' )
__a =1
return cls
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Arab'] , 25_6001 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Latn'] , 25_6002 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['fra_Latn'] , 25_6057 )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer.batch_encode_plus(self.src_text ).input_ids[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
self.assertIn(__snake_case , self.tokenizer.all_special_ids )
# fmt: off
__a =[RO_CODE, 4254, 9_8068, 11_2923, 3_9072, 3909, 713, 10_2767, 26, 1_7314, 3_5642, 1_4683, 3_3118, 2022, 6_6987, 2, 25_6047]
# fmt: on
__a =self.tokenizer.decode(__snake_case , skip_special_tokens=__snake_case )
__a =self.tokenizer.decode(generated_ids[1:] , skip_special_tokens=__snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertNotIn(self.tokenizer.eos_token , __snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =['this is gunna be a long sentence ' * 20]
assert isinstance(src_text[0] , __snake_case )
__a =10
__a =self.tokenizer(__snake_case , max_length=__snake_case , truncation=__snake_case ).input_ids[0]
self.assertEqual(ids[-1] , 2 )
self.assertEqual(ids[0] , __snake_case )
self.assertEqual(len(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
self.assertListEqual(self.tokenizer.convert_tokens_to_ids(['<mask>', 'ar_AR'] ) , [25_6203, 3] )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =tempfile.mkdtemp()
__a =self.tokenizer.fairseq_tokens_to_ids
self.tokenizer.save_pretrained(__snake_case )
__a =NllbTokenizer.from_pretrained(__snake_case )
self.assertDictEqual(new_tok.fairseq_tokens_to_ids , __snake_case )
@require_torch
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.tokenizer(
self.src_text , text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=len(self.expected_src_tokens ) , return_tensors='pt' , )
__a =shift_tokens_right(
batch['labels'] , self.tokenizer.pad_token_id , self.tokenizer.lang_code_to_id['ron_Latn'] )
self.assertIsInstance(__snake_case , __snake_case )
self.assertEqual((2, 15) , batch.input_ids.shape )
self.assertEqual((2, 15) , batch.attention_mask.shape )
__a =batch.input_ids.tolist()[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
self.assertEqual(__snake_case , batch.decoder_input_ids[0, 0] ) # EOS
# Test that special tokens are reset
self.assertEqual(self.tokenizer.prefix_tokens , [EN_CODE] )
self.assertEqual(self.tokenizer.suffix_tokens , [self.tokenizer.eos_token_id] )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =self.tokenizer(self.src_text , padding=__snake_case , truncation=__snake_case , max_length=3 , return_tensors='pt' )
__a =self.tokenizer(
text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=10 , return_tensors='pt' )
__a =targets['input_ids']
__a =shift_tokens_right(
__snake_case , self.tokenizer.pad_token_id , decoder_start_token_id=self.tokenizer.lang_code_to_id[self.tokenizer.tgt_lang] , )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.decoder_input_ids.shape[1] , 10 )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.tokenizer._build_translation_inputs(
'A test' , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
nested_simplify(__snake_case ) , {
# A, test, EOS, en_XX
'input_ids': [[25_6047, 70, 7356, 2]],
'attention_mask': [[1, 1, 1, 1]],
# ar_AR
'forced_bos_token_id': 25_6057,
} , )
@require_torch
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =True
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2, 25_6047] )
__a =False
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [25_6047, 1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2] )
| 308
| 1
|
# Copyright 2021 The HuggingFace Inc. 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 packaging import version
from .. import __version__
from .constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD, IMAGENET_STANDARD_MEAN, IMAGENET_STANDARD_STD
from .doc import (
add_code_sample_docstrings,
add_end_docstrings,
add_start_docstrings,
add_start_docstrings_to_model_forward,
copy_func,
replace_return_docstrings,
)
from .generic import (
ContextManagers,
ExplicitEnum,
ModelOutput,
PaddingStrategy,
TensorType,
add_model_info_to_auto_map,
cached_property,
can_return_loss,
expand_dims,
find_labels,
flatten_dict,
infer_framework,
is_jax_tensor,
is_numpy_array,
is_tensor,
is_tf_symbolic_tensor,
is_tf_tensor,
is_torch_device,
is_torch_dtype,
is_torch_tensor,
reshape,
squeeze,
strtobool,
tensor_size,
to_numpy,
to_py_obj,
transpose,
working_or_temp_dir,
)
from .hub import (
CLOUDFRONT_DISTRIB_PREFIX,
DISABLE_TELEMETRY,
HF_MODULES_CACHE,
HUGGINGFACE_CO_PREFIX,
HUGGINGFACE_CO_RESOLVE_ENDPOINT,
PYTORCH_PRETRAINED_BERT_CACHE,
PYTORCH_TRANSFORMERS_CACHE,
S3_BUCKET_PREFIX,
TRANSFORMERS_CACHE,
TRANSFORMERS_DYNAMIC_MODULE_NAME,
EntryNotFoundError,
PushToHubMixin,
RepositoryNotFoundError,
RevisionNotFoundError,
cached_file,
default_cache_path,
define_sagemaker_information,
download_url,
extract_commit_hash,
get_cached_models,
get_file_from_repo,
get_full_repo_name,
has_file,
http_user_agent,
is_offline_mode,
is_remote_url,
move_cache,
send_example_telemetry,
try_to_load_from_cache,
)
from .import_utils import (
ENV_VARS_TRUE_AND_AUTO_VALUES,
ENV_VARS_TRUE_VALUES,
TORCH_FX_REQUIRED_VERSION,
USE_JAX,
USE_TF,
USE_TORCH,
DummyObject,
OptionalDependencyNotAvailable,
_LazyModule,
ccl_version,
direct_transformers_import,
get_torch_version,
is_accelerate_available,
is_apex_available,
is_bitsandbytes_available,
is_bsa_available,
is_coloredlogs_available,
is_cython_available,
is_datasets_available,
is_decord_available,
is_detectrona_available,
is_faiss_available,
is_flax_available,
is_ftfy_available,
is_in_notebook,
is_ipex_available,
is_jieba_available,
is_jumanpp_available,
is_kenlm_available,
is_keras_nlp_available,
is_librosa_available,
is_natten_available,
is_ninja_available,
is_onnx_available,
is_openai_available,
is_optimum_available,
is_pandas_available,
is_peft_available,
is_phonemizer_available,
is_protobuf_available,
is_psutil_available,
is_pyanvml_available,
is_pyctcdecode_available,
is_pytesseract_available,
is_pytest_available,
is_pytorch_quantization_available,
is_rjieba_available,
is_sacremoses_available,
is_safetensors_available,
is_sagemaker_dp_enabled,
is_sagemaker_mp_enabled,
is_scipy_available,
is_sentencepiece_available,
is_seqio_available,
is_sklearn_available,
is_soundfile_availble,
is_spacy_available,
is_speech_available,
is_sudachi_available,
is_tensorflow_probability_available,
is_tensorflow_text_available,
is_tfaonnx_available,
is_tf_available,
is_timm_available,
is_tokenizers_available,
is_torch_available,
is_torch_bfaa_available,
is_torch_bfaa_cpu_available,
is_torch_bfaa_gpu_available,
is_torch_compile_available,
is_torch_cuda_available,
is_torch_fx_available,
is_torch_fx_proxy,
is_torch_mps_available,
is_torch_neuroncore_available,
is_torch_tensorrt_fx_available,
is_torch_tfaa_available,
is_torch_tpu_available,
is_torchaudio_available,
is_torchdistx_available,
is_torchdynamo_available,
is_torchvision_available,
is_training_run_on_sagemaker,
is_vision_available,
requires_backends,
torch_only_method,
)
_lowerCAmelCase : Optional[Any] = "pytorch_model.bin"
_lowerCAmelCase : Dict = "pytorch_model.bin.index.json"
_lowerCAmelCase : Optional[Any] = "adapter_config.json"
_lowerCAmelCase : List[str] = "adapter_model.bin"
_lowerCAmelCase : str = "adapter_model.safetensors"
_lowerCAmelCase : List[str] = "tf_model.h5"
_lowerCAmelCase : int = "tf_model.h5.index.json"
_lowerCAmelCase : int = "model.ckpt"
_lowerCAmelCase : Union[str, Any] = "flax_model.msgpack"
_lowerCAmelCase : Any = "flax_model.msgpack.index.json"
_lowerCAmelCase : Dict = "model.safetensors"
_lowerCAmelCase : Any = "model.safetensors.index.json"
_lowerCAmelCase : str = "config.json"
_lowerCAmelCase : List[Any] = "preprocessor_config.json"
_lowerCAmelCase : Dict = FEATURE_EXTRACTOR_NAME
_lowerCAmelCase : Any = "generation_config.json"
_lowerCAmelCase : Dict = "modelcard.json"
_lowerCAmelCase : Any = "▁"
_lowerCAmelCase : str = SENTENCEPIECE_UNDERLINE # Kept for backward compatibility
_lowerCAmelCase : Optional[Any] = [
[[0, 1, 0, 1], [1, 0, 0, 1]]
] * 2 # Needs to have 0s and 1s only since XLM uses it for langs too.
_lowerCAmelCase : str = [[7, 6, 0, 0, 1], [1, 2, 3, 0, 0], [0, 0, 0, 4, 5]]
_lowerCAmelCase : int = [[1, 1, 1, 1, 1], [1, 1, 1, 0, 0], [0, 0, 0, 1, 1]]
def UpperCamelCase_( _snake_case : Union[str, Any] ):
"""simple docstring"""
if version.parse(_snake_case ) < version.parse(_snake_case ):
if "dev" in min_version:
__a =(
'This example requires a source install from HuggingFace Transformers (see '
'`https://huggingface.co/docs/transformers/installation#install-from-source`),'
)
else:
__a =F'This example requires a minimum version of {min_version},'
error_message += F' but the version found is {__version__}.\n'
raise ImportError(
error_message
+ 'Check out https://github.com/huggingface/transformers/tree/main/examples#important-note for the examples corresponding to other '
'versions of HuggingFace Transformers.' )
| 308
|
def UpperCamelCase_( _snake_case : str , _snake_case : int ):
"""simple docstring"""
return [sentence[i : i + ngram_size] for i in range(len(_snake_case ) - ngram_size + 1 )]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 308
| 1
|
from ...configuration_utils import PretrainedConfig
from ...utils import logging
_lowerCAmelCase : Dict = logging.get_logger(__name__)
_lowerCAmelCase : Optional[Any] = {
"SCUT-DLVCLab/lilt-roberta-en-base": (
"https://huggingface.co/SCUT-DLVCLab/lilt-roberta-en-base/resolve/main/config.json"
),
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'lilt'
def __init__( self , __snake_case=3_0522 , __snake_case=768 , __snake_case=12 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=2 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=0 , __snake_case="absolute" , __snake_case=None , __snake_case=4 , __snake_case=1024 , **__snake_case , ) -> Any:
'''simple docstring'''
super().__init__(pad_token_id=__snake_case , **__snake_case )
__a =vocab_size
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_act
__a =intermediate_size
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_vocab_size
__a =initializer_range
__a =layer_norm_eps
__a =position_embedding_type
__a =classifier_dropout
__a =channel_shrink_ratio
__a =max_ad_position_embeddings
| 308
|
import argparse
import pytorch_lightning as pl
import torch
from torch import nn
from transformers import LongformerForQuestionAnswering, LongformerModel
class __magic_name__ ( pl.LightningModule ):
def __init__( self , __snake_case ) -> List[Any]:
'''simple docstring'''
super().__init__()
__a =model
__a =2
__a =nn.Linear(self.model.config.hidden_size , self.num_labels )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : str , _snake_case : str , _snake_case : str ):
"""simple docstring"""
__a =LongformerModel.from_pretrained(_snake_case )
__a =LightningModel(_snake_case )
__a =torch.load(_snake_case , map_location=torch.device('cpu' ) )
lightning_model.load_state_dict(ckpt['state_dict'] )
# init longformer question answering model
__a =LongformerForQuestionAnswering.from_pretrained(_snake_case )
# transfer weights
longformer_for_qa.longformer.load_state_dict(lightning_model.model.state_dict() )
longformer_for_qa.qa_outputs.load_state_dict(lightning_model.qa_outputs.state_dict() )
longformer_for_qa.eval()
# save model
longformer_for_qa.save_pretrained(_snake_case )
print(F'Conversion successful. Model saved under {pytorch_dump_folder_path}' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--longformer_model",
default=None,
type=str,
required=True,
help="model identifier of longformer. Should be either `longformer-base-4096` or `longformer-large-4096`.",
)
parser.add_argument(
"--longformer_question_answering_ckpt_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch Lightning Checkpoint.",
)
parser.add_argument(
"--pytorch_dump_folder_path", default=None, type=str, required=True, help="Path to the output PyTorch model."
)
_lowerCAmelCase : Optional[int] = parser.parse_args()
convert_longformer_qa_checkpoint_to_pytorch(
args.longformer_model, args.longformer_question_answering_ckpt_path, args.pytorch_dump_folder_path
)
| 308
| 1
|
import os
import re
import unicodedata
from shutil import copyfile
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union
import sentencepiece as spm
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import is_torch_available, logging
if is_torch_available():
import torch
if TYPE_CHECKING:
from transformers.pipelines.conversational import Conversation
_lowerCAmelCase : Optional[int] = logging.get_logger(__name__)
_lowerCAmelCase : Dict = {"vocab_file": "spiece.model"}
_lowerCAmelCase : int = {
"vocab_file": {
"AI-Sweden/gpt-sw3-126m": "https://huggingface.co/AI-Sweden/gpt-sw3-126m/resolve/main/spiece.model",
"AI-Sweden/gpt-sw3-350m": "https://huggingface.co/AI-Sweden/gpt-sw3-350m/resolve/main/spiece.model",
"AI-Sweden/gpt-sw3-1.6b": "https://huggingface.co/AI-Sweden/gpt-sw3-1.6b/resolve/main/spiece.model",
"AI-Sweden/gpt-sw3-6.7b": "https://huggingface.co/AI-Sweden/gpt-sw3-6.7b/resolve/main/spiece.model",
"AI-Sweden/gpt-sw3-20b": "https://huggingface.co/AI-Sweden/gpt-sw3-20b/resolve/main/spiece.model",
}
}
_lowerCAmelCase : int = {
"AI-Sweden/gpt-sw3-126m": 2_048,
"AI-Sweden/gpt-sw3-350m": 2_048,
"AI-Sweden/gpt-sw3-1.6b": 2_048,
"AI-Sweden/gpt-sw3-6.7b": 2_048,
"AI-Sweden/gpt-sw3-20b": 2_048,
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = VOCAB_FILES_NAMES
SCREAMING_SNAKE_CASE = PRETRAINED_VOCAB_FILES_MAP
SCREAMING_SNAKE_CASE = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
SCREAMING_SNAKE_CASE = ['input_ids', 'attention_mask']
def __init__( self , __snake_case , __snake_case=False , __snake_case=False , __snake_case=False , __snake_case=None , __snake_case=None , __snake_case=None , __snake_case=None , __snake_case = None , **__snake_case , ) -> None:
'''simple docstring'''
__a ={} if sp_model_kwargs is None else sp_model_kwargs
__a =kwargs.get('name_or_path' )
if name_or_path is None:
logger.warning(
'name_or_path not provided, will work for all GPTSw3 models except gpt-sw3-7b,'
' you are testing the model, this can safely be ignored' )
__a ='None'
# Default definitions for our 2 tokenizer versions, with None-checks to enable proper testing
__a ='<|endoftext|>' if eos_token is None else eos_token
__a ='<unk>' if unk_token is None else unk_token
if "gpt-sw3-7b" in name_or_path:
__a =unk_token if pad_token is None else pad_token
__a =eos_token if bos_token is None else bos_token
else:
__a ='<pad>' if pad_token is None else pad_token
__a ='<s>' if bos_token is None else bos_token
super().__init__(
do_lower_case=__snake_case , remove_space=__snake_case , keep_accents=__snake_case , bos_token=__snake_case , eos_token=__snake_case , unk_token=__snake_case , pad_token=__snake_case , sp_model_kwargs=self.sp_model_kwargs , **__snake_case , )
__a =do_lower_case
__a =remove_space
__a =keep_accents
__a =vocab_file
__a =spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(__snake_case )
# Used for whitespace normalization in input texts
# fmt : off
__a ={' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', '', ''}
# fmt : on
# Regular expression to remove non-printing characters (e.g. some unicode control chars) in preprocessing
__a =re.compile(
f'[{"".join(map(__snake_case , list(range(0 , 9 ) ) + list(range(11 , 32 ) ) + list(range(127 , 160 ) ) + [160, 173, 8203] ) )}]' )
def __getstate__( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.__dict__.copy()
__a =None
return state
def __setstate__( self , __snake_case ) -> Any:
'''simple docstring'''
__a =d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a ={}
__a =spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(self.vocab_file )
@property
# Copied from transformers.models.albert.tokenization_albert.AlbertTokenizer.vocab_size
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return len(self.sp_model )
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
__a =self.non_printing_characters_re.sub('' , __snake_case )
# Normalize whitespaces
__a =''.join([char if char not in self.whitespaces else ' ' for char in text] )
# NFC Unicode normalization
__a =unicodedata.normalize('NFC' , __snake_case )
return text
def __magic_name__ ( self , __snake_case , **__snake_case ) -> List[str]:
'''simple docstring'''
__a =self.preprocess_text(__snake_case )
return self.sp_model.encode(__snake_case , out_type=__snake_case )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
return self.sp_model.PieceToId(__snake_case )
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
return self.sp_model.IdToPiece(__snake_case )
@staticmethod
def __magic_name__ ( __snake_case ) -> str:
'''simple docstring'''
return out_string
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
__a =[]
__a =''
__a =False
for token in tokens:
# make sure that special tokens are not decoded using sentencepiece model
if token in self.all_special_tokens:
# TODO: Check if this is needed, as it ensures that decode(encode(doc)) != doc by adding extra whitespace in the decoded document
if not prev_is_special:
out_string += " "
out_string += self.sp_model.decode(__snake_case ) + token
__a =True
__a =[]
else:
current_sub_tokens.append(__snake_case )
__a =False
out_string += self.sp_model.decode(__snake_case )
return out_string
def __magic_name__ ( self ) -> Dict[str, int]:
'''simple docstring'''
__a ={self.convert_ids_to_tokens(__snake_case ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> Tuple[str]:
'''simple docstring'''
if not os.path.isdir(__snake_case ):
logger.error(f'Vocabulary path ({save_directory}) should be a directory' )
return
__a =os.path.join(
__snake_case , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(__snake_case ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , __snake_case )
elif not os.path.isfile(self.vocab_file ):
with open(__snake_case , 'wb' ) as fi:
__a =self.sp_model.serialized_model_proto()
fi.write(__snake_case )
return (out_vocab_file,)
def __magic_name__ ( self , __snake_case , __snake_case = False ) -> Union[List[int], List[List[int]], "torch.Tensor"]:
'''simple docstring'''
if isinstance(__snake_case , __snake_case ):
__a =self.preprocess_text(__snake_case )
__a =self.sp_model.encode(__snake_case )
else:
__a =[self.preprocess_text(__snake_case ) for t in text]
__a =self.sp_model.encode(__snake_case )
if return_tensors is True or return_tensors == "pt":
__a =torch.tensor(__snake_case )
return token_ids
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
return self.sp_model.decode(__snake_case )
def __magic_name__ ( self , __snake_case ) -> List[int]:
'''simple docstring'''
__a =[f'User: {text}' if is_user else f'Bot: {text}' for is_user, text in conversation.iter_texts()]
__a =(
f'{self.eos_token}{self.bos_token}' + f'{self.bos_token}'.join(__snake_case ) + f'{self.bos_token}Bot:'
)
return self.encode(text=__snake_case )
| 308
|
import logging
import os
import random
import sys
from dataclasses import dataclass, field
from typing import Optional
import datasets
import numpy as np
import pandas as pd
from datasets import load_dataset
import transformers
from transformers import (
AutoConfig,
BartForSequenceClassification,
DataCollatorWithPadding,
EvalPrediction,
HfArgumentParser,
TapexTokenizer,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
_lowerCAmelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The name of the dataset to use (via the datasets library).'} )
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} , )
SCREAMING_SNAKE_CASE = field(
default=1_0_2_4 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Overwrite the cached preprocessed datasets or not.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Whether to pad all samples to `max_seq_length`. '
'If False, will pad the samples dynamically when batching to the maximum length in the batch.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of training examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of evaluation examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of prediction examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = field(default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the test data.'} )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
if self.dataset_name is not None:
pass
elif self.train_file is None or self.validation_file is None:
raise ValueError('Need either a GLUE task, a training/validation file or a dataset name.' )
else:
__a =self.train_file.split('.' )[-1]
assert train_extension in ["csv", "json"], "`train_file` should be a csv or a json file."
__a =self.validation_file.split('.' )[-1]
assert (
validation_extension == train_extension
), "`validation_file` should have the same extension (csv or json) as `train_file`."
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to use one of the fast tokenizer (backed by the tokenizers library) or not.'} , )
SCREAMING_SNAKE_CASE = field(
default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Will use the token generated when running `huggingface-cli login` (necessary to use this script '
'with private models).'
)
} , )
def UpperCamelCase_( ):
"""simple docstring"""
__a =HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments) )
if len(sys.argv ) == 2 and sys.argv[1].endswith('.json' ):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
__a , __a , __a =parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) )
else:
__a , __a , __a =parser.parse_args_into_dataclasses()
# Setup logging
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , handlers=[logging.StreamHandler(sys.stdout )] , )
__a =training_args.get_process_log_level()
logger.setLevel(_snake_case )
datasets.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Log on each process the small summary:
logger.warning(
F'Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}'
+ F'distributed training: {bool(training_args.local_rank != -1 )}, 16-bits training: {training_args.fpaa}' )
logger.info(F'Training/evaluation parameters {training_args}' )
# Detecting last checkpoint.
__a =None
if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir:
__a =get_last_checkpoint(training_args.output_dir )
if last_checkpoint is None and len(os.listdir(training_args.output_dir ) ) > 0:
raise ValueError(
F'Output directory ({training_args.output_dir}) already exists and is not empty. '
'Use --overwrite_output_dir to overcome.' )
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
F'Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change '
'the `--output_dir` or add `--overwrite_output_dir` to train from scratch.' )
# Set seed before initializing model.
set_seed(training_args.seed )
# Get the datasets: you can either provide your own CSV/JSON training and evaluation files (see below)
# or specify a GLUE benchmark task (the dataset will be downloaded automatically from the datasets Hub).
#
# For JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# If the CSVs/JSONs contain only one non-label column, the script does single sentence classification on this
# single column. You can easily tweak this behavior (see below)
#
# In distributed training, the load_dataset function guarantee that only one local process can concurrently
# download the dataset.
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
__a =load_dataset(
data_args.dataset_name , data_args.dataset_config_name , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from your local files.
# CSV/JSON training and evaluation files are needed.
__a ={'train': data_args.train_file, 'validation': data_args.validation_file}
# Get the test dataset: you can provide your own CSV/JSON test file (see below)
# when you use `do_predict` without specifying a GLUE benchmark task.
if training_args.do_predict:
if data_args.test_file is not None:
__a =data_args.train_file.split('.' )[-1]
__a =data_args.test_file.split('.' )[-1]
assert (
test_extension == train_extension
), "`test_file` should have the same extension (csv or json) as `train_file`."
__a =data_args.test_file
else:
raise ValueError('Need either a GLUE task or a test file for `do_predict`.' )
for key in data_files.keys():
logger.info(F'load a local file for {key}: {data_files[key]}' )
if data_args.train_file.endswith('.csv' ):
# Loading a dataset from local csv files
__a =load_dataset('csv' , data_files=_snake_case , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from local json files
__a =load_dataset('json' , data_files=_snake_case , cache_dir=model_args.cache_dir )
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.html.
# Labels
__a =raw_datasets['train'].features['label'].names
__a =len(_snake_case )
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
__a =AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# load tapex tokenizer
__a =TapexTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , use_fast=model_args.use_fast_tokenizer , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , add_prefix_space=_snake_case , )
__a =BartForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_tf=bool('.ckpt' in model_args.model_name_or_path ) , config=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# Padding strategy
if data_args.pad_to_max_length:
__a ='max_length'
else:
# We will pad later, dynamically at batch creation, to the max sequence length in each batch
__a =False
# Some models have set the order of the labels to use, so let's make sure we do use it.
__a ={'Refused': 0, 'Entailed': 1}
__a ={0: 'Refused', 1: 'Entailed'}
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warning(
F'The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the'
F'model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}.' )
__a =min(data_args.max_seq_length , tokenizer.model_max_length )
def preprocess_tabfact_function(_snake_case : Tuple ):
# Tokenize the texts
def _convert_table_text_to_pandas(_snake_case : Optional[Any] ):
__a =[_table_row.split('#' ) for _table_row in _table_text.strip('\n' ).split('\n' )]
__a =pd.DataFrame.from_records(_table_content[1:] , columns=_table_content[0] )
return _table_pd
__a =examples['statement']
__a =list(map(_convert_table_text_to_pandas , examples['table_text'] ) )
__a =tokenizer(_snake_case , _snake_case , padding=_snake_case , max_length=_snake_case , truncation=_snake_case )
__a =examples['label']
return result
with training_args.main_process_first(desc='dataset map pre-processing' ):
__a =raw_datasets.map(
_snake_case , batched=_snake_case , load_from_cache_file=not data_args.overwrite_cache , desc='Running tokenizer on dataset' , )
if training_args.do_train:
if "train" not in raw_datasets:
raise ValueError('--do_train requires a train dataset' )
__a =raw_datasets['train']
if data_args.max_train_samples is not None:
__a =train_dataset.select(range(data_args.max_train_samples ) )
if training_args.do_eval:
if "validation" not in raw_datasets and "validation_matched" not in raw_datasets:
raise ValueError('--do_eval requires a validation dataset' )
__a =raw_datasets['validation']
if data_args.max_eval_samples is not None:
__a =eval_dataset.select(range(data_args.max_eval_samples ) )
if training_args.do_predict or data_args.test_file is not None:
if "test" not in raw_datasets and "test_matched" not in raw_datasets:
raise ValueError('--do_predict requires a test dataset' )
__a =raw_datasets['test']
if data_args.max_predict_samples is not None:
__a =predict_dataset.select(range(data_args.max_predict_samples ) )
# Log a few random samples from the training set:
if training_args.do_train:
for index in random.sample(range(len(_snake_case ) ) , 3 ):
logger.info(F'Sample {index} of the training set: {train_dataset[index]}.' )
# You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
# predictions and label_ids field) and has to return a dictionary string to float.
def compute_metrics(_snake_case : EvalPrediction ):
__a =p.predictions[0] if isinstance(p.predictions , _snake_case ) else p.predictions
__a =np.argmax(_snake_case , axis=1 )
return {"accuracy": (preds == p.label_ids).astype(np.floataa ).mean().item()}
# Data collator will default to DataCollatorWithPadding, so we change it if we already did the padding.
if data_args.pad_to_max_length:
__a =default_data_collator
elif training_args.fpaa:
__a =DataCollatorWithPadding(_snake_case , pad_to_multiple_of=8 )
else:
__a =None
# Initialize our Trainer
__a =Trainer(
model=_snake_case , args=_snake_case , train_dataset=train_dataset if training_args.do_train else None , eval_dataset=eval_dataset if training_args.do_eval else None , compute_metrics=_snake_case , tokenizer=_snake_case , data_collator=_snake_case , )
# Training
if training_args.do_train:
__a =None
if training_args.resume_from_checkpoint is not None:
__a =training_args.resume_from_checkpoint
elif last_checkpoint is not None:
__a =last_checkpoint
__a =trainer.train(resume_from_checkpoint=_snake_case )
__a =train_result.metrics
__a =(
data_args.max_train_samples if data_args.max_train_samples is not None else len(_snake_case )
)
__a =min(_snake_case , len(_snake_case ) )
trainer.save_model() # Saves the tokenizer too for easy upload
trainer.log_metrics('train' , _snake_case )
trainer.save_metrics('train' , _snake_case )
trainer.save_state()
# Evaluation
if training_args.do_eval:
logger.info('*** Evaluate ***' )
__a =trainer.evaluate(eval_dataset=_snake_case )
__a =data_args.max_eval_samples if data_args.max_eval_samples is not None else len(_snake_case )
__a =min(_snake_case , len(_snake_case ) )
trainer.log_metrics('eval' , _snake_case )
trainer.save_metrics('eval' , _snake_case )
if training_args.do_predict:
logger.info('*** Predict ***' )
# Removing the `label` columns because it contains -1 and Trainer won't like that.
__a =predict_dataset.remove_columns('label' )
__a =trainer.predict(_snake_case , metric_key_prefix='predict' ).predictions
__a =np.argmax(_snake_case , axis=1 )
__a =os.path.join(training_args.output_dir , 'predict_results_tabfact.txt' )
if trainer.is_world_process_zero():
with open(_snake_case , 'w' ) as writer:
logger.info('***** Predict Results *****' )
writer.write('index\tprediction\n' )
for index, item in enumerate(_snake_case ):
__a =label_list[item]
writer.write(F'{index}\t{item}\n' )
__a ={'finetuned_from': model_args.model_name_or_path, 'tasks': 'text-classification'}
if training_args.push_to_hub:
trainer.push_to_hub(**_snake_case )
else:
trainer.create_model_card(**_snake_case )
def UpperCamelCase_( _snake_case : Union[str, Any] ):
"""simple docstring"""
main()
if __name__ == "__main__":
main()
| 308
| 1
|
import json
import os
import subprocess
import unittest
from ast import literal_eval
import pytest
from parameterized import parameterized, parameterized_class
from . import is_sagemaker_available
if is_sagemaker_available():
from sagemaker import Session, TrainingJobAnalytics
from sagemaker.huggingface import HuggingFace
@pytest.mark.skipif(
literal_eval(os.getenv('TEST_SAGEMAKER' , 'False' ) ) is not True , reason='Skipping test because should only be run when releasing minor transformers version' , )
@pytest.mark.usefixtures('sm_env' )
@parameterized_class(
[
{
'framework': 'pytorch',
'script': 'run_glue.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_5_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'pytorch',
'script': 'run_ddp.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'tensorflow',
'script': 'run_tf_dist.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.6, 'eval_loss': 0.7},
},
] )
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.framework == "pytorch":
subprocess.run(
f'cp ./examples/pytorch/text-classification/run_glue.py {self.env.test_path}/run_glue.py'.split() , encoding='utf-8' , check=__snake_case , )
assert hasattr(self , 'env' )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
__a =f'{self.env.base_job_name}-{instance_count}-{"ddp" if "ddp" in self.script else "smd"}'
# distributed data settings
__a ={'smdistributed': {'dataparallel': {'enabled': True}}} if self.script != 'run_ddp.py' else None
# creates estimator
return HuggingFace(
entry_point=self.script , source_dir=self.env.test_path , role=self.env.role , image_uri=self.env.image_uri , base_job_name=__snake_case , instance_count=__snake_case , instance_type=self.instance_type , debugger_hook_config=__snake_case , hyperparameters={**self.env.distributed_hyperparameters, 'model_name_or_path': self.model_name_or_path} , metric_definitions=self.env.metric_definitions , distribution=__snake_case , py_version='py36' , )
def __magic_name__ ( self , __snake_case ) -> Optional[Any]:
'''simple docstring'''
TrainingJobAnalytics(__snake_case ).export_csv(f'{self.env.test_path}/{job_name}_metrics.csv' )
@parameterized.expand([(2,)] )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
# create estimator
__a =self.create_estimator(__snake_case )
# run training
estimator.fit()
# result dataframe
__a =TrainingJobAnalytics(estimator.latest_training_job.name ).dataframe()
# extract kpis
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_accuracy']['value'] )
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_loss']['value'] )
# get train time from SageMaker job, this includes starting, preprocessing, stopping
__a =(
Session().describe_training_job(estimator.latest_training_job.name ).get('TrainingTimeInSeconds' , 99_9999 )
)
# assert kpis
assert train_runtime <= self.results["train_runtime"]
assert all(t >= self.results['eval_accuracy'] for t in eval_accuracy )
assert all(t <= self.results['eval_loss'] for t in eval_loss )
# dump tests result into json file to share in PR
with open(f'{estimator.latest_training_job.name}.json' , 'w' ) as outfile:
json.dump({'train_time': train_runtime, 'eval_accuracy': eval_accuracy, 'eval_loss': eval_loss} , __snake_case )
| 308
|
from __future__ import annotations
import time
import numpy as np
_lowerCAmelCase : List[str] = [8, 5, 9, 7]
_lowerCAmelCase : List[str] = [
[2, 0, 1, 1],
[0, 1, 2, 1],
[4, 0, 0, 3],
[0, 2, 1, 0],
[1, 0, 3, 0],
]
_lowerCAmelCase : List[Any] = [
[3, 2, 1, 4],
[0, 2, 5, 2],
[5, 1, 0, 5],
[1, 5, 3, 0],
[3, 0, 3, 3],
]
class __magic_name__ :
def __init__( self , __snake_case , __snake_case , __snake_case , ) -> None:
'''simple docstring'''
__a =claim_vector
__a =allocated_resources_table
__a =maximum_claim_table
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return [
sum(p_item[i] for p_item in self.__allocated_resources_table )
for i in range(len(self.__allocated_resources_table[0] ) )
]
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return np.array(self.__claim_vector ) - np.array(
self.__processes_resource_summation() )
def __magic_name__ ( self ) -> list[list[int]]:
'''simple docstring'''
return [
list(np.array(self.__maximum_claim_table[i] ) - np.array(__snake_case ) )
for i, allocated_resource in enumerate(self.__allocated_resources_table )
]
def __magic_name__ ( self ) -> dict[int, list[int]]:
'''simple docstring'''
return {self.__need().index(__snake_case ): i for i in self.__need()}
def __magic_name__ ( self , **__snake_case ) -> None:
'''simple docstring'''
__a =self.__need()
__a =self.__allocated_resources_table
__a =self.__available_resources()
__a =self.__need_index_manager()
for kw, val in kwargs.items():
if kw and val is True:
self.__pretty_data()
print('_' * 50 + '\n' )
while need_list:
__a =False
for each_need in need_list:
__a =True
for index, need in enumerate(__snake_case ):
if need > available_resources[index]:
__a =False
break
if execution:
__a =True
# get the original index of the process from ind_ctrl db
for original_need_index, need_clone in need_index_manager.items():
if each_need == need_clone:
__a =original_need_index
print(f'Process {process_number + 1} is executing.' )
# remove the process run from stack
need_list.remove(__snake_case )
# update available/freed resources stack
__a =np.array(__snake_case ) + np.array(
alloc_resources_table[process_number] )
print(
'Updated available resource stack for processes: '
+ ' '.join([str(__snake_case ) for x in available_resources] ) )
break
if safe:
print('The process is in a safe state.\n' )
else:
print('System in unsafe state. Aborting...\n' )
break
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
print(' ' * 9 + 'Allocated Resource Table' )
for item in self.__allocated_resources_table:
print(
f'P{self.__allocated_resources_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(' ' * 9 + 'System Resource Table' )
for item in self.__maximum_claim_table:
print(
f'P{self.__maximum_claim_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(
'Current Usage by Active Processes: '
+ ' '.join(str(__snake_case ) for x in self.__claim_vector ) )
print(
'Initial Available Resources: '
+ ' '.join(str(__snake_case ) for x in self.__available_resources() ) )
time.sleep(1 )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 1
|
import collections
import importlib.util
import os
import re
from pathlib import Path
_lowerCAmelCase : Tuple = "src/transformers"
# Matches is_xxx_available()
_lowerCAmelCase : Tuple = re.compile(r"is\_([a-z_]*)_available()")
# Catches a one-line _import_struct = {xxx}
_lowerCAmelCase : Optional[int] = re.compile(r"^_import_structure\s+=\s+\{([^\}]+)\}")
# Catches a line with a key-values pattern: "bla": ["foo", "bar"]
_lowerCAmelCase : Tuple = re.compile(r"\s+\"\S*\":\s+\[([^\]]*)\]")
# Catches a line if not is_foo_available
_lowerCAmelCase : Optional[int] = re.compile(r"^\s*if\s+not\s+is\_[a-z_]*\_available\(\)")
# Catches a line _import_struct["bla"].append("foo")
_lowerCAmelCase : int = re.compile(r"^\s*_import_structure\[\"\S*\"\]\.append\(\"(\S*)\"\)")
# Catches a line _import_struct["bla"].extend(["foo", "bar"]) or _import_struct["bla"] = ["foo", "bar"]
_lowerCAmelCase : Tuple = re.compile(r"^\s*_import_structure\[\S*\](?:\.extend\(|\s*=\s+)\[([^\]]*)\]")
# Catches a line with an object between quotes and a comma: "MyModel",
_lowerCAmelCase : Any = re.compile("^\s+\"([^\"]+)\",")
# Catches a line with objects between brackets only: ["foo", "bar"],
_lowerCAmelCase : Optional[Any] = re.compile("^\s+\[([^\]]+)\]")
# Catches a line with from foo import bar, bla, boo
_lowerCAmelCase : int = re.compile(r"\s+from\s+\S*\s+import\s+([^\(\s].*)\n")
# Catches a line with try:
_lowerCAmelCase : Optional[int] = re.compile(r"^\s*try:")
# Catches a line with else:
_lowerCAmelCase : Optional[Any] = re.compile(r"^\s*else:")
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
if _re_test_backend.search(_snake_case ) is None:
return None
__a =[b[0] for b in _re_backend.findall(_snake_case )]
backends.sort()
return "_and_".join(_snake_case )
def UpperCamelCase_( _snake_case : List[Any] ):
"""simple docstring"""
with open(_snake_case , 'r' , encoding='utf-8' , newline='\n' ) as f:
__a =f.readlines()
__a =0
while line_index < len(_snake_case ) and not lines[line_index].startswith('_import_structure = {' ):
line_index += 1
# If this is a traditional init, just return.
if line_index >= len(_snake_case ):
return None
# First grab the objects without a specific backend in _import_structure
__a =[]
while not lines[line_index].startswith('if TYPE_CHECKING' ) and find_backend(lines[line_index] ) is None:
__a =lines[line_index]
# If we have everything on a single line, let's deal with it.
if _re_one_line_import_struct.search(_snake_case ):
__a =_re_one_line_import_struct.search(_snake_case ).groups()[0]
__a =re.findall('\[([^\]]+)\]' , _snake_case )
for imp in imports:
objects.extend([obj[1:-1] for obj in imp.split(', ' )] )
line_index += 1
continue
__a =_re_import_struct_key_value.search(_snake_case )
if single_line_import_search is not None:
__a =[obj[1:-1] for obj in single_line_import_search.groups()[0].split(', ' ) if len(_snake_case ) > 0]
objects.extend(_snake_case )
elif line.startswith(' ' * 8 + '"' ):
objects.append(line[9:-3] )
line_index += 1
__a ={'none': objects}
# Let's continue with backend-specific objects in _import_structure
while not lines[line_index].startswith('if TYPE_CHECKING' ):
# If the line is an if not is_backend_available, we grab all objects associated.
__a =find_backend(lines[line_index] )
# Check if the backend declaration is inside a try block:
if _re_try.search(lines[line_index - 1] ) is None:
__a =None
if backend is not None:
line_index += 1
# Scroll until we hit the else block of try-except-else
while _re_else.search(lines[line_index] ) is None:
line_index += 1
line_index += 1
__a =[]
# Until we unindent, add backend objects to the list
while len(lines[line_index] ) <= 1 or lines[line_index].startswith(' ' * 4 ):
__a =lines[line_index]
if _re_import_struct_add_one.search(_snake_case ) is not None:
objects.append(_re_import_struct_add_one.search(_snake_case ).groups()[0] )
elif _re_import_struct_add_many.search(_snake_case ) is not None:
__a =_re_import_struct_add_many.search(_snake_case ).groups()[0].split(', ' )
__a =[obj[1:-1] for obj in imports if len(_snake_case ) > 0]
objects.extend(_snake_case )
elif _re_between_brackets.search(_snake_case ) is not None:
__a =_re_between_brackets.search(_snake_case ).groups()[0].split(', ' )
__a =[obj[1:-1] for obj in imports if len(_snake_case ) > 0]
objects.extend(_snake_case )
elif _re_quote_object.search(_snake_case ) is not None:
objects.append(_re_quote_object.search(_snake_case ).groups()[0] )
elif line.startswith(' ' * 8 + '"' ):
objects.append(line[9:-3] )
elif line.startswith(' ' * 12 + '"' ):
objects.append(line[13:-3] )
line_index += 1
__a =objects
else:
line_index += 1
# At this stage we are in the TYPE_CHECKING part, first grab the objects without a specific backend
__a =[]
while (
line_index < len(_snake_case )
and find_backend(lines[line_index] ) is None
and not lines[line_index].startswith('else' )
):
__a =lines[line_index]
__a =_re_import.search(_snake_case )
if single_line_import_search is not None:
objects.extend(single_line_import_search.groups()[0].split(', ' ) )
elif line.startswith(' ' * 8 ):
objects.append(line[8:-2] )
line_index += 1
__a ={'none': objects}
# Let's continue with backend-specific objects
while line_index < len(_snake_case ):
# If the line is an if is_backend_available, we grab all objects associated.
__a =find_backend(lines[line_index] )
# Check if the backend declaration is inside a try block:
if _re_try.search(lines[line_index - 1] ) is None:
__a =None
if backend is not None:
line_index += 1
# Scroll until we hit the else block of try-except-else
while _re_else.search(lines[line_index] ) is None:
line_index += 1
line_index += 1
__a =[]
# Until we unindent, add backend objects to the list
while len(lines[line_index] ) <= 1 or lines[line_index].startswith(' ' * 8 ):
__a =lines[line_index]
__a =_re_import.search(_snake_case )
if single_line_import_search is not None:
objects.extend(single_line_import_search.groups()[0].split(', ' ) )
elif line.startswith(' ' * 12 ):
objects.append(line[12:-2] )
line_index += 1
__a =objects
else:
line_index += 1
return import_dict_objects, type_hint_objects
def UpperCamelCase_( _snake_case : List[str] , _snake_case : Dict ):
"""simple docstring"""
def find_duplicates(_snake_case : Tuple ):
return [k for k, v in collections.Counter(_snake_case ).items() if v > 1]
if list(import_dict_objects.keys() ) != list(type_hint_objects.keys() ):
return ["Both sides of the init do not have the same backends!"]
__a =[]
for key in import_dict_objects.keys():
__a =find_duplicates(import_dict_objects[key] )
if duplicate_imports:
errors.append(F'Duplicate _import_structure definitions for: {duplicate_imports}' )
__a =find_duplicates(type_hint_objects[key] )
if duplicate_type_hints:
errors.append(F'Duplicate TYPE_CHECKING objects for: {duplicate_type_hints}' )
if sorted(set(import_dict_objects[key] ) ) != sorted(set(type_hint_objects[key] ) ):
__a ='base imports' if key == 'none' else F'{key} backend'
errors.append(F'Differences for {name}:' )
for a in type_hint_objects[key]:
if a not in import_dict_objects[key]:
errors.append(F' {a} in TYPE_HINT but not in _import_structure.' )
for a in import_dict_objects[key]:
if a not in type_hint_objects[key]:
errors.append(F' {a} in _import_structure but not in TYPE_HINT.' )
return errors
def UpperCamelCase_( ):
"""simple docstring"""
__a =[]
for root, _, files in os.walk(_snake_case ):
if "__init__.py" in files:
__a =os.path.join(_snake_case , '__init__.py' )
__a =parse_init(_snake_case )
if objects is not None:
__a =analyze_results(*_snake_case )
if len(_snake_case ) > 0:
__a =F'Problem in {fname}, both halves do not define the same objects.\n{errors[0]}'
failures.append('\n'.join(_snake_case ) )
if len(_snake_case ) > 0:
raise ValueError('\n\n'.join(_snake_case ) )
def UpperCamelCase_( ):
"""simple docstring"""
__a =[]
for path, directories, files in os.walk(_snake_case ):
for folder in directories:
# Ignore private modules
if folder.startswith('_' ):
directories.remove(_snake_case )
continue
# Ignore leftovers from branches (empty folders apart from pycache)
if len(list((Path(_snake_case ) / folder).glob('*.py' ) ) ) == 0:
continue
__a =str((Path(_snake_case ) / folder).relative_to(_snake_case ) )
__a =short_path.replace(os.path.sep , '.' )
submodules.append(_snake_case )
for fname in files:
if fname == "__init__.py":
continue
__a =str((Path(_snake_case ) / fname).relative_to(_snake_case ) )
__a =short_path.replace('.py' , '' ).replace(os.path.sep , '.' )
if len(submodule.split('.' ) ) == 1:
submodules.append(_snake_case )
return submodules
_lowerCAmelCase : List[str] = [
"convert_pytorch_checkpoint_to_tf2",
"modeling_flax_pytorch_utils",
]
def UpperCamelCase_( ):
"""simple docstring"""
__a =importlib.util.spec_from_file_location(
'transformers' , os.path.join(_snake_case , '__init__.py' ) , submodule_search_locations=[PATH_TO_TRANSFORMERS] , )
__a =spec.loader.load_module()
__a =[
module
for module in get_transformers_submodules()
if module not in IGNORE_SUBMODULES and module not in transformers._import_structure.keys()
]
if len(_snake_case ) > 0:
__a ='\n'.join(F'- {module}' for module in module_not_registered )
raise ValueError(
'The following submodules are not properly registered in the main init of Transformers:\n'
F'{list_of_modules}\n'
'Make sure they appear somewhere in the keys of `_import_structure` with an empty list as value.' )
if __name__ == "__main__":
check_all_inits()
check_submodules()
| 308
|
import string
# frequency taken from https://en.wikipedia.org/wiki/Letter_frequency
_lowerCAmelCase : Tuple = {
"E": 12.70,
"T": 9.06,
"A": 8.17,
"O": 7.51,
"I": 6.97,
"N": 6.75,
"S": 6.33,
"H": 6.09,
"R": 5.99,
"D": 4.25,
"L": 4.03,
"C": 2.78,
"U": 2.76,
"M": 2.41,
"W": 2.36,
"F": 2.23,
"G": 2.02,
"Y": 1.97,
"P": 1.93,
"B": 1.29,
"V": 0.98,
"K": 0.77,
"J": 0.15,
"X": 0.15,
"Q": 0.10,
"Z": 0.07,
}
_lowerCAmelCase : Optional[int] = "ETAOINSHRDLCUMWFGYPBVKJXQZ"
_lowerCAmelCase : Any = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a ={letter: 0 for letter in string.ascii_uppercase}
for letter in message.upper():
if letter in LETTERS:
letter_count[letter] += 1
return letter_count
def UpperCamelCase_( _snake_case : tuple ):
"""simple docstring"""
return x[0]
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_letter_count(_snake_case )
__a ={
freq: [] for letter, freq in letter_to_freq.items()
}
for letter in LETTERS:
freq_to_letter[letter_to_freq[letter]].append(_snake_case )
__a ={}
for freq in freq_to_letter:
freq_to_letter[freq].sort(key=ETAOIN.find , reverse=_snake_case )
__a =''.join(freq_to_letter[freq] )
__a =list(freq_to_letter_str.items() )
freq_pairs.sort(key=_snake_case , reverse=_snake_case )
__a =[freq_pair[1] for freq_pair in freq_pairs]
return "".join(_snake_case )
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_frequency_order(_snake_case )
__a =0
for common_letter in ETAOIN[:6]:
if common_letter in freq_order[:6]:
match_score += 1
for uncommon_letter in ETAOIN[-6:]:
if uncommon_letter in freq_order[-6:]:
match_score += 1
return match_score
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 1
|
from typing import Dict, Optional, Union
import numpy as np
from ...image_processing_utils import BaseImageProcessor, BatchFeature, get_size_dict
from ...image_transforms import flip_channel_order, resize, to_channel_dimension_format, to_pil_image
from ...image_utils import (
ChannelDimension,
ImageInput,
PILImageResampling,
make_list_of_images,
to_numpy_array,
valid_images,
)
from ...utils import TensorType, is_pytesseract_available, is_vision_available, logging, requires_backends
if is_vision_available():
import PIL
# soft dependency
if is_pytesseract_available():
import pytesseract
_lowerCAmelCase : Union[str, Any] = logging.get_logger(__name__)
def UpperCamelCase_( _snake_case : Optional[Any] , _snake_case : List[Any] , _snake_case : Union[str, Any] ):
"""simple docstring"""
return [
int(1000 * (box[0] / width) ),
int(1000 * (box[1] / height) ),
int(1000 * (box[2] / width) ),
int(1000 * (box[3] / height) ),
]
def UpperCamelCase_( _snake_case : np.ndarray , _snake_case : Optional[str] , _snake_case : Optional[str] = None ):
"""simple docstring"""
__a =tesseract_config if tesseract_config is not None else ''
# apply OCR
__a =to_pil_image(_snake_case )
__a , __a =pil_image.size
__a =pytesseract.image_to_data(_snake_case , lang=_snake_case , output_type='dict' , config=_snake_case )
__a , __a , __a , __a , __a =data['text'], data['left'], data['top'], data['width'], data['height']
# filter empty words and corresponding coordinates
__a =[idx for idx, word in enumerate(_snake_case ) if not word.strip()]
__a =[word for idx, word in enumerate(_snake_case ) if idx not in irrelevant_indices]
__a =[coord for idx, coord in enumerate(_snake_case ) if idx not in irrelevant_indices]
__a =[coord for idx, coord in enumerate(_snake_case ) if idx not in irrelevant_indices]
__a =[coord for idx, coord in enumerate(_snake_case ) if idx not in irrelevant_indices]
__a =[coord for idx, coord in enumerate(_snake_case ) if idx not in irrelevant_indices]
# turn coordinates into (left, top, left+width, top+height) format
__a =[]
for x, y, w, h in zip(_snake_case , _snake_case , _snake_case , _snake_case ):
__a =[x, y, x + w, y + h]
actual_boxes.append(_snake_case )
# finally, normalize the bounding boxes
__a =[]
for box in actual_boxes:
normalized_boxes.append(normalize_box(_snake_case , _snake_case , _snake_case ) )
assert len(_snake_case ) == len(_snake_case ), "Not as many words as there are bounding boxes"
return words, normalized_boxes
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = ['pixel_values']
def __init__( self , __snake_case = True , __snake_case = None , __snake_case = PILImageResampling.BILINEAR , __snake_case = True , __snake_case = None , __snake_case = "" , **__snake_case , ) -> None:
'''simple docstring'''
super().__init__(**__snake_case )
__a =size if size is not None else {'height': 224, 'width': 224}
__a =get_size_dict(__snake_case )
__a =do_resize
__a =size
__a =resample
__a =apply_ocr
__a =ocr_lang
__a =tesseract_config
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case = PILImageResampling.BILINEAR , __snake_case = None , **__snake_case , ) -> np.ndarray:
'''simple docstring'''
__a =get_size_dict(__snake_case )
if "height" not in size or "width" not in size:
raise ValueError(f'The size dictionary must contain the keys \'height\' and \'width\'. Got {size.keys()}' )
__a =(size['height'], size['width'])
return resize(__snake_case , size=__snake_case , resample=__snake_case , data_format=__snake_case , **__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = ChannelDimension.FIRST , **__snake_case , ) -> PIL.Image.Image:
'''simple docstring'''
__a =do_resize if do_resize is not None else self.do_resize
__a =size if size is not None else self.size
__a =get_size_dict(__snake_case )
__a =resample if resample is not None else self.resample
__a =apply_ocr if apply_ocr is not None else self.apply_ocr
__a =ocr_lang if ocr_lang is not None else self.ocr_lang
__a =tesseract_config if tesseract_config is not None else self.tesseract_config
__a =make_list_of_images(__snake_case )
if not valid_images(__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.' )
# All transformations expect numpy arrays.
__a =[to_numpy_array(__snake_case ) for image in images]
if apply_ocr:
requires_backends(self , 'pytesseract' )
__a =[]
__a =[]
for image in images:
__a , __a =apply_tesseract(__snake_case , __snake_case , __snake_case )
words_batch.append(__snake_case )
boxes_batch.append(__snake_case )
if do_resize:
__a =[self.resize(image=__snake_case , size=__snake_case , resample=__snake_case ) for image in images]
# flip color channels from RGB to BGR (as Detectron2 requires this)
__a =[flip_channel_order(__snake_case ) for image in images]
__a =[to_channel_dimension_format(__snake_case , __snake_case ) for image in images]
__a =BatchFeature(data={'pixel_values': images} , tensor_type=__snake_case )
if apply_ocr:
__a =words_batch
__a =boxes_batch
return data
| 308
|
from ...configuration_utils import PretrainedConfig
from ...utils import logging
_lowerCAmelCase : Any = logging.get_logger(__name__)
_lowerCAmelCase : int = {
"caidas/swin2sr-classicalsr-x2-64": (
"https://huggingface.co/caidas/swin2sr-classicalsr-x2-64/resolve/main/config.json"
),
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'swin2sr'
SCREAMING_SNAKE_CASE = {
'hidden_size': 'embed_dim',
'num_attention_heads': 'num_heads',
'num_hidden_layers': 'num_layers',
}
def __init__( self , __snake_case=64 , __snake_case=1 , __snake_case=3 , __snake_case=180 , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=8 , __snake_case=2.0 , __snake_case=True , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.1 , __snake_case="gelu" , __snake_case=False , __snake_case=0.02 , __snake_case=1e-5 , __snake_case=2 , __snake_case=1.0 , __snake_case="1conv" , __snake_case="pixelshuffle" , **__snake_case , ) -> Dict:
'''simple docstring'''
super().__init__(**__snake_case )
__a =image_size
__a =patch_size
__a =num_channels
__a =embed_dim
__a =depths
__a =len(__snake_case )
__a =num_heads
__a =window_size
__a =mlp_ratio
__a =qkv_bias
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =drop_path_rate
__a =hidden_act
__a =use_absolute_embeddings
__a =layer_norm_eps
__a =initializer_range
__a =upscale
__a =img_range
__a =resi_connection
__a =upsampler
| 308
| 1
|
import enum
import warnings
from ..tokenization_utils import TruncationStrategy
from ..utils import add_end_docstrings, is_tf_available, is_torch_available, logging
from .base import PIPELINE_INIT_ARGS, Pipeline
if is_tf_available():
import tensorflow as tf
from ..models.auto.modeling_tf_auto import TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING
if is_torch_available():
from ..models.auto.modeling_auto import MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING
_lowerCAmelCase : Union[str, Any] = logging.get_logger(__name__)
class __magic_name__ ( enum.Enum ):
SCREAMING_SNAKE_CASE = 0
SCREAMING_SNAKE_CASE = 1
@add_end_docstrings(lowerCAmelCase_ )
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'generated'
def __init__( self , *__snake_case , **__snake_case ) -> int:
'''simple docstring'''
super().__init__(*__snake_case , **__snake_case )
self.check_model_type(
TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING
if self.framework == 'tf'
else MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING )
def __magic_name__ ( self , __snake_case=None , __snake_case=None , __snake_case=None , __snake_case=None , __snake_case=None , __snake_case=None , **__snake_case , ) -> Any:
'''simple docstring'''
__a ={}
if truncation is not None:
__a =truncation
__a =generate_kwargs
__a ={}
if return_tensors is not None and return_type is None:
__a =ReturnType.TENSORS if return_tensors else ReturnType.TEXT
if return_type is not None:
__a =return_type
if clean_up_tokenization_spaces is not None:
__a =clean_up_tokenization_spaces
if stop_sequence is not None:
__a =self.tokenizer.encode(__snake_case , add_special_tokens=__snake_case )
if len(__snake_case ) > 1:
warnings.warn(
'Stopping on a multiple token sequence is not yet supported on transformers. The first token of'
' the stop sequence will be used as the stop sequence string in the interim.' )
__a =stop_sequence_ids[0]
return preprocess_params, forward_params, postprocess_params
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Optional[int]:
'''simple docstring'''
return True
def __magic_name__ ( self , *__snake_case , __snake_case ) -> Optional[Any]:
'''simple docstring'''
__a =self.model.config.prefix if self.model.config.prefix is not None else ''
if isinstance(args[0] , __snake_case ):
if self.tokenizer.pad_token_id is None:
raise ValueError('Please make sure that the tokenizer has a pad_token_id when using a batch input' )
__a =([prefix + arg for arg in args[0]],)
__a =True
elif isinstance(args[0] , __snake_case ):
__a =(prefix + args[0],)
__a =False
else:
raise ValueError(
f' `args[0]`: {args[0]} have the wrong format. The should be either of type `str` or type `list`' )
__a =self.tokenizer(*__snake_case , padding=__snake_case , truncation=__snake_case , return_tensors=self.framework )
# This is produced by tokenizers but is an invalid generate kwargs
if "token_type_ids" in inputs:
del inputs["token_type_ids"]
return inputs
def __call__( self , *__snake_case , **__snake_case ) -> List[str]:
'''simple docstring'''
__a =super().__call__(*__snake_case , **__snake_case )
if (
isinstance(args[0] , __snake_case )
and all(isinstance(__snake_case , __snake_case ) for el in args[0] )
and all(len(__snake_case ) == 1 for res in result )
):
return [res[0] for res in result]
return result
def __magic_name__ ( self , __snake_case , __snake_case=TruncationStrategy.DO_NOT_TRUNCATE , **__snake_case ) -> List[str]:
'''simple docstring'''
__a =self._parse_and_tokenize(__snake_case , truncation=__snake_case , **__snake_case )
return inputs
def __magic_name__ ( self , __snake_case , **__snake_case ) -> Union[str, Any]:
'''simple docstring'''
if self.framework == "pt":
__a , __a =model_inputs['input_ids'].shape
elif self.framework == "tf":
__a , __a =tf.shape(model_inputs['input_ids'] ).numpy()
__a =generate_kwargs.get('min_length' , self.model.config.min_length )
__a =generate_kwargs.get('max_length' , self.model.config.max_length )
self.check_inputs(__snake_case , generate_kwargs['min_length'] , generate_kwargs['max_length'] )
__a =self.model.generate(**__snake_case , **__snake_case )
__a =output_ids.shape[0]
if self.framework == "pt":
__a =output_ids.reshape(__snake_case , out_b // in_b , *output_ids.shape[1:] )
elif self.framework == "tf":
__a =tf.reshape(__snake_case , (in_b, out_b // in_b, *output_ids.shape[1:]) )
return {"output_ids": output_ids}
def __magic_name__ ( self , __snake_case , __snake_case=ReturnType.TEXT , __snake_case=False ) -> Any:
'''simple docstring'''
__a =[]
for output_ids in model_outputs["output_ids"][0]:
if return_type == ReturnType.TENSORS:
__a ={f'{self.return_name}_token_ids': output_ids}
elif return_type == ReturnType.TEXT:
__a ={
f'{self.return_name}_text': self.tokenizer.decode(
__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case , )
}
records.append(__snake_case )
return records
@add_end_docstrings(lowerCAmelCase_ )
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'summary'
def __call__( self , *__snake_case , **__snake_case ) -> Tuple:
'''simple docstring'''
return super().__call__(*__snake_case , **__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> bool:
'''simple docstring'''
if max_length < min_length:
logger.warning(f'Your min_length={min_length} must be inferior than your max_length={max_length}.' )
if input_length < max_length:
logger.warning(
f'Your max_length is set to {max_length}, but your input_length is only {input_length}. Since this is '
'a summarization task, where outputs shorter than the input are typically wanted, you might '
f'consider decreasing max_length manually, e.g. summarizer(\'...\', max_length={input_length//2})' )
@add_end_docstrings(lowerCAmelCase_ )
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'translation'
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> List[str]:
'''simple docstring'''
if input_length > 0.9 * max_length:
logger.warning(
f'Your input_length: {input_length} is bigger than 0.9 * max_length: {max_length}. You might consider '
'increasing your max_length manually, e.g. translator(\'...\', max_length=400)' )
return True
def __magic_name__ ( self , *__snake_case , __snake_case=TruncationStrategy.DO_NOT_TRUNCATE , __snake_case=None , __snake_case=None ) -> Tuple:
'''simple docstring'''
if getattr(self.tokenizer , '_build_translation_inputs' , __snake_case ):
return self.tokenizer._build_translation_inputs(
*__snake_case , return_tensors=self.framework , truncation=__snake_case , src_lang=__snake_case , tgt_lang=__snake_case )
else:
return super()._parse_and_tokenize(*__snake_case , truncation=__snake_case )
def __magic_name__ ( self , __snake_case=None , __snake_case=None , **__snake_case ) -> Optional[Any]:
'''simple docstring'''
__a , __a , __a =super()._sanitize_parameters(**__snake_case )
if src_lang is not None:
__a =src_lang
if tgt_lang is not None:
__a =tgt_lang
if src_lang is None and tgt_lang is None:
# Backward compatibility, direct arguments use is preferred.
__a =kwargs.get('task' , self.task )
__a =task.split('_' )
if task and len(__snake_case ) == 4:
# translation, XX, to YY
__a =items[1]
__a =items[3]
return preprocess_params, forward_params, postprocess_params
def __call__( self , *__snake_case , **__snake_case ) -> Optional[Any]:
'''simple docstring'''
return super().__call__(*__snake_case , **__snake_case )
| 308
|
import os
def UpperCamelCase_( _snake_case : str = "input.txt" ):
"""simple docstring"""
with open(os.path.join(os.path.dirname(_snake_case ) , _snake_case ) ) as input_file:
__a =[
[int(_snake_case ) for element in line.split(',' )]
for line in input_file.readlines()
]
__a =len(_snake_case )
__a =len(matrix[0] )
__a =[[-1 for _ in range(_snake_case )] for _ in range(_snake_case )]
for i in range(_snake_case ):
__a =matrix[i][0]
for j in range(1 , _snake_case ):
for i in range(_snake_case ):
__a =minimal_path_sums[i][j - 1] + matrix[i][j]
for i in range(1 , _snake_case ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i - 1][j] + matrix[i][j] )
for i in range(rows - 2 , -1 , -1 ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i + 1][j] + matrix[i][j] )
return min(minimal_path_sums_row[-1] for minimal_path_sums_row in minimal_path_sums )
if __name__ == "__main__":
print(f'''{solution() = }''')
| 308
| 1
|
import time
import warnings
from abc import ABC
from copy import deepcopy
from typing import Optional
import torch
from ..utils import add_start_docstrings, logging
_lowerCAmelCase : str = logging.get_logger(__name__)
_lowerCAmelCase : List[str] = r"\n Args:\n input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):\n Indices of input sequence tokens in the vocabulary.\n\n Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and\n [`PreTrainedTokenizer.__call__`] for details.\n\n [What are input IDs?](../glossary#input-ids)\n scores (`torch.FloatTensor` of shape `(batch_size, config.vocab_size)`):\n Prediction scores of a language modeling head. These can be scores for each vocabulary token before SoftMax\n or scores for each vocabulary token after SoftMax.\n kwargs (`Dict[str, Any]`, *optional*):\n Additional stopping criteria specific kwargs.\n\n Return:\n `bool`. `False` indicates we should continue, `True` indicates we should stop.\n\n"
class __magic_name__ ( lowerCAmelCase_ ):
@add_start_docstrings(__snake_case )
def __call__( self , __snake_case , __snake_case , **__snake_case ) -> bool:
'''simple docstring'''
raise NotImplementedError('StoppingCriteria needs to be subclassed' )
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , __snake_case , __snake_case = None ) -> int:
'''simple docstring'''
__a =max_length
__a =max_position_embeddings
@add_start_docstrings(__snake_case )
def __call__( self , __snake_case , __snake_case , **__snake_case ) -> bool:
'''simple docstring'''
__a =input_ids.shape[-1]
__a =cur_len >= self.max_length
if self.max_position_embeddings is not None and not is_done and cur_len >= self.max_position_embeddings:
logger.warning_once(
'This is a friendly reminder - the current text generation call will exceed the model\'s predefined '
f'maximum length ({self.max_position_embeddings}). Depending on the model, you may observe '
'exceptions, performance degradation, or nothing at all.' )
return is_done
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
warnings.warn(
'The class `MaxNewTokensCriteria` is deprecated. '
f'Please use `MaxLengthCriteria(max_length={start_length + max_new_tokens})` '
'with `max_length = start_length + max_new_tokens` instead.' , __snake_case , )
__a =start_length
__a =max_new_tokens
__a =start_length + max_new_tokens
@add_start_docstrings(__snake_case )
def __call__( self , __snake_case , __snake_case , **__snake_case ) -> bool:
'''simple docstring'''
return input_ids.shape[-1] >= self.max_length
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , __snake_case , __snake_case = None ) -> int:
'''simple docstring'''
__a =max_time
__a =time.time() if initial_timestamp is None else initial_timestamp
@add_start_docstrings(__snake_case )
def __call__( self , __snake_case , __snake_case , **__snake_case ) -> bool:
'''simple docstring'''
return time.time() - self.initial_timestamp > self.max_time
class __magic_name__ ( lowerCAmelCase_ ):
@add_start_docstrings(__snake_case )
def __call__( self , __snake_case , __snake_case , **__snake_case ) -> bool:
'''simple docstring'''
return any(criteria(__snake_case , __snake_case ) for criteria in self )
@property
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
for stopping_criterium in self:
if isinstance(__snake_case , __snake_case ):
return stopping_criterium.max_length
elif isinstance(__snake_case , __snake_case ):
return stopping_criterium.max_length
return None
def UpperCamelCase_( _snake_case : StoppingCriteriaList , _snake_case : int ):
"""simple docstring"""
__a =stopping_criteria.max_length
__a =deepcopy(_snake_case )
if stopping_max_length is not None and stopping_max_length != max_length:
warnings.warn('You set different `max_length` for stopping criteria and `max_length` parameter' , _snake_case )
elif stopping_max_length is None:
new_stopping_criteria.append(MaxLengthCriteria(max_length=_snake_case ) )
return new_stopping_criteria
| 308
|
import warnings
from ...utils import logging
from .image_processing_mobilevit import MobileViTImageProcessor
_lowerCAmelCase : Any = logging.get_logger(__name__)
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , *__snake_case , **__snake_case ) -> None:
'''simple docstring'''
warnings.warn(
'The class MobileViTFeatureExtractor is deprecated and will be removed in version 5 of Transformers.'
' Please use MobileViTImageProcessor instead.' , __snake_case , )
super().__init__(*__snake_case , **__snake_case )
| 308
| 1
|
# Copyright 2021 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.
import argparse
import os
from accelerate.utils import ComputeEnvironment
from .cluster import get_cluster_input
from .config_args import cache_dir, default_config_file, default_yaml_config_file, load_config_from_file # noqa: F401
from .config_utils import _ask_field, _ask_options, _convert_compute_environment # noqa: F401
from .sagemaker import get_sagemaker_input
_lowerCAmelCase : str = "Launches a series of prompts to create and save a `default_config.yaml` configuration file for your training system. Should always be ran first on your machine"
def UpperCamelCase_( ):
"""simple docstring"""
__a =_ask_options(
'In which compute environment are you running?' , ['This machine', 'AWS (Amazon SageMaker)'] , _convert_compute_environment , )
if compute_environment == ComputeEnvironment.AMAZON_SAGEMAKER:
__a =get_sagemaker_input()
else:
__a =get_cluster_input()
return config
def UpperCamelCase_( _snake_case : int=None ):
"""simple docstring"""
if subparsers is not None:
__a =subparsers.add_parser('config' , description=_snake_case )
else:
__a =argparse.ArgumentParser('Accelerate config command' , description=_snake_case )
parser.add_argument(
'--config_file' , default=_snake_case , help=(
'The path to use to store the config file. Will default to a file named default_config.yaml in the cache '
'location, which is the content of the environment `HF_HOME` suffixed with \'accelerate\', or if you don\'t have '
'such an environment variable, your cache directory (\'~/.cache\' or the content of `XDG_CACHE_HOME`) suffixed '
'with \'huggingface\'.'
) , )
if subparsers is not None:
parser.set_defaults(func=_snake_case )
return parser
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_user_input()
if args.config_file is not None:
__a =args.config_file
else:
if not os.path.isdir(_snake_case ):
os.makedirs(_snake_case )
__a =default_yaml_config_file
if config_file.endswith('.json' ):
config.to_json_file(_snake_case )
else:
config.to_yaml_file(_snake_case )
print(F'accelerate configuration saved at {config_file}' )
def UpperCamelCase_( ):
"""simple docstring"""
__a =config_command_parser()
__a =parser.parse_args()
config_command(_snake_case )
if __name__ == "__main__":
main()
| 308
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_sentencepiece_available
_lowerCAmelCase : int = {}
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Dict = ["BartphoTokenizer"]
if TYPE_CHECKING:
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_bartpho import BartphoTokenizer
else:
import sys
_lowerCAmelCase : Optional[Any] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 1
|
from __future__ import annotations
def UpperCamelCase_( _snake_case : list[int] ):
"""simple docstring"""
__a =len(_snake_case ) // 2
# choose the middle 3 elements
__a =lst[m - 1 : m + 2]
# if middle element is peak
if three[1] > three[0] and three[1] > three[2]:
return three[1]
# if increasing, recurse on right
elif three[0] < three[2]:
if len(lst[:m] ) == 2:
m -= 1
return peak(lst[m:] )
# decreasing
else:
if len(lst[:m] ) == 2:
m += 1
return peak(lst[:m] )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
|
from collections import OrderedDict
from typing import Mapping
from packaging import version
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
_lowerCAmelCase : List[str] = logging.get_logger(__name__)
_lowerCAmelCase : List[Any] = {
"hustvl/yolos-small": "https://huggingface.co/hustvl/yolos-small/resolve/main/config.json",
# See all YOLOS models at https://huggingface.co/models?filter=yolos
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'yolos'
def __init__( self , __snake_case=768 , __snake_case=12 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=[512, 864] , __snake_case=16 , __snake_case=3 , __snake_case=True , __snake_case=100 , __snake_case=True , __snake_case=False , __snake_case=1 , __snake_case=5 , __snake_case=2 , __snake_case=5 , __snake_case=2 , __snake_case=0.1 , **__snake_case , ) -> str:
'''simple docstring'''
super().__init__(**__snake_case )
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =initializer_range
__a =layer_norm_eps
__a =image_size
__a =patch_size
__a =num_channels
__a =qkv_bias
__a =num_detection_tokens
__a =use_mid_position_embeddings
__a =auxiliary_loss
# Hungarian matcher
__a =class_cost
__a =bbox_cost
__a =giou_cost
# Loss coefficients
__a =bbox_loss_coefficient
__a =giou_loss_coefficient
__a =eos_coefficient
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = version.parse('1.11' )
@property
def __magic_name__ ( self ) -> Mapping[str, Mapping[int, str]]:
'''simple docstring'''
return OrderedDict(
[
('pixel_values', {0: 'batch', 1: 'num_channels', 2: 'height', 3: 'width'}),
] )
@property
def __magic_name__ ( self ) -> float:
'''simple docstring'''
return 1e-4
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return 12
| 308
| 1
|
import warnings
from ...utils import logging
from .image_processing_mobilevit import MobileViTImageProcessor
_lowerCAmelCase : Any = logging.get_logger(__name__)
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , *__snake_case , **__snake_case ) -> None:
'''simple docstring'''
warnings.warn(
'The class MobileViTFeatureExtractor is deprecated and will be removed in version 5 of Transformers.'
' Please use MobileViTImageProcessor instead.' , __snake_case , )
super().__init__(*__snake_case , **__snake_case )
| 308
|
import argparse
import torch
from datasets import load_dataset
from donut import DonutModel
from transformers import (
DonutImageProcessor,
DonutProcessor,
DonutSwinConfig,
DonutSwinModel,
MBartConfig,
MBartForCausalLM,
VisionEncoderDecoderModel,
XLMRobertaTokenizerFast,
)
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =model.config
__a =DonutSwinConfig(
image_size=original_config.input_size , patch_size=4 , depths=original_config.encoder_layer , num_heads=[4, 8, 16, 32] , window_size=original_config.window_size , embed_dim=128 , )
__a =MBartConfig(
is_decoder=_snake_case , is_encoder_decoder=_snake_case , add_cross_attention=_snake_case , decoder_layers=original_config.decoder_layer , max_position_embeddings=original_config.max_position_embeddings , vocab_size=len(
model.decoder.tokenizer ) , scale_embedding=_snake_case , add_final_layer_norm=_snake_case , )
return encoder_config, decoder_config
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
if "encoder.model" in name:
__a =name.replace('encoder.model' , 'encoder' )
if "decoder.model" in name:
__a =name.replace('decoder.model' , 'decoder' )
if "patch_embed.proj" in name:
__a =name.replace('patch_embed.proj' , 'embeddings.patch_embeddings.projection' )
if "patch_embed.norm" in name:
__a =name.replace('patch_embed.norm' , 'embeddings.norm' )
if name.startswith('encoder' ):
if "layers" in name:
__a ='encoder.' + name
if "attn.proj" in name:
__a =name.replace('attn.proj' , 'attention.output.dense' )
if "attn" in name and "mask" not in name:
__a =name.replace('attn' , 'attention.self' )
if "norm1" in name:
__a =name.replace('norm1' , 'layernorm_before' )
if "norm2" in name:
__a =name.replace('norm2' , 'layernorm_after' )
if "mlp.fc1" in name:
__a =name.replace('mlp.fc1' , 'intermediate.dense' )
if "mlp.fc2" in name:
__a =name.replace('mlp.fc2' , 'output.dense' )
if name == "encoder.norm.weight":
__a ='encoder.layernorm.weight'
if name == "encoder.norm.bias":
__a ='encoder.layernorm.bias'
return name
def UpperCamelCase_( _snake_case : Tuple , _snake_case : str ):
"""simple docstring"""
for key in orig_state_dict.copy().keys():
__a =orig_state_dict.pop(_snake_case )
if "qkv" in key:
__a =key.split('.' )
__a =int(key_split[3] )
__a =int(key_split[5] )
__a =model.encoder.encoder.layers[layer_num].blocks[block_num].attention.self.all_head_size
if "weight" in key:
__a =val[:dim, :]
__a =val[dim : dim * 2, :]
__a =val[-dim:, :]
else:
__a =val[:dim]
__a =val[dim : dim * 2]
__a =val[-dim:]
elif "attn_mask" in key or key in ["encoder.model.norm.weight", "encoder.model.norm.bias"]:
# HuggingFace implementation doesn't use attn_mask buffer
# and model doesn't use final LayerNorms for the encoder
pass
else:
__a =val
return orig_state_dict
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Union[str, Any]=None , _snake_case : List[Any]=False ):
"""simple docstring"""
__a =DonutModel.from_pretrained(_snake_case ).eval()
# load HuggingFace model
__a , __a =get_configs(_snake_case )
__a =DonutSwinModel(_snake_case )
__a =MBartForCausalLM(_snake_case )
__a =VisionEncoderDecoderModel(encoder=_snake_case , decoder=_snake_case )
model.eval()
__a =original_model.state_dict()
__a =convert_state_dict(_snake_case , _snake_case )
model.load_state_dict(_snake_case )
# verify results on scanned document
__a =load_dataset('hf-internal-testing/example-documents' )
__a =dataset['test'][0]['image'].convert('RGB' )
__a =XLMRobertaTokenizerFast.from_pretrained(_snake_case , from_slow=_snake_case )
__a =DonutImageProcessor(
do_align_long_axis=original_model.config.align_long_axis , size=original_model.config.input_size[::-1] )
__a =DonutProcessor(_snake_case , _snake_case )
__a =processor(_snake_case , return_tensors='pt' ).pixel_values
if model_name == "naver-clova-ix/donut-base-finetuned-docvqa":
__a ='<s_docvqa><s_question>{user_input}</s_question><s_answer>'
__a ='When is the coffee break?'
__a =task_prompt.replace('{user_input}' , _snake_case )
elif model_name == "naver-clova-ix/donut-base-finetuned-rvlcdip":
__a ='<s_rvlcdip>'
elif model_name in [
"naver-clova-ix/donut-base-finetuned-cord-v1",
"naver-clova-ix/donut-base-finetuned-cord-v1-2560",
]:
__a ='<s_cord>'
elif model_name == "naver-clova-ix/donut-base-finetuned-cord-v2":
__a ='s_cord-v2>'
elif model_name == "naver-clova-ix/donut-base-finetuned-zhtrainticket":
__a ='<s_zhtrainticket>'
elif model_name in ["naver-clova-ix/donut-proto", "naver-clova-ix/donut-base"]:
# use a random prompt
__a ='hello world'
else:
raise ValueError('Model name not supported' )
__a =original_model.decoder.tokenizer(_snake_case , add_special_tokens=_snake_case , return_tensors='pt' )[
'input_ids'
]
__a =original_model.encoder.model.patch_embed(_snake_case )
__a , __a =model.encoder.embeddings(_snake_case )
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
# verify encoder hidden states
__a =original_model.encoder(_snake_case )
__a =model.encoder(_snake_case ).last_hidden_state
assert torch.allclose(_snake_case , _snake_case , atol=1e-2 )
# verify decoder hidden states
__a =original_model(_snake_case , _snake_case , _snake_case ).logits
__a =model(_snake_case , decoder_input_ids=_snake_case ).logits
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
print('Looks ok!' )
if pytorch_dump_folder_path is not None:
print(F'Saving model and processor to {pytorch_dump_folder_path}' )
model.save_pretrained(_snake_case )
processor.save_pretrained(_snake_case )
if push_to_hub:
model.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
processor.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
if __name__ == "__main__":
_lowerCAmelCase : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_name",
default="naver-clova-ix/donut-base-finetuned-docvqa",
required=False,
type=str,
help="Name of the original model you'd like to convert.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
required=False,
type=str,
help="Path to the output PyTorch model directory.",
)
parser.add_argument(
"--push_to_hub",
action="store_true",
help="Whether or not to push the converted model and processor to the 🤗 hub.",
)
_lowerCAmelCase : List[Any] = parser.parse_args()
convert_donut_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 308
| 1
|
from typing import Callable, List, Optional, Tuple, Union
import torch
from transformers import CLIPTextModel, CLIPTokenizer
from ...configuration_utils import ConfigMixin, register_to_config
from ...models import ModelMixin, TransformeraDModel, VQModel
from ...schedulers import VQDiffusionScheduler
from ...utils import logging
from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
_lowerCAmelCase : Tuple = logging.get_logger(__name__) # pylint: disable=invalid-name
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ ):
@register_to_config
def __init__( self , __snake_case , __snake_case = None , __snake_case = None ) -> Dict:
'''simple docstring'''
super().__init__()
__a =learnable
if self.learnable:
assert hidden_size is not None, "learnable=True requires `hidden_size` to be set"
assert length is not None, "learnable=True requires `length` to be set"
__a =torch.zeros(__snake_case , __snake_case )
else:
__a =None
__a =torch.nn.Parameter(__snake_case )
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Any:
'''simple docstring'''
super().__init__()
self.register_modules(
vqvae=__snake_case , transformer=__snake_case , text_encoder=__snake_case , tokenizer=__snake_case , scheduler=__snake_case , learned_classifier_free_sampling_embeddings=__snake_case , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> str:
'''simple docstring'''
__a =len(__snake_case ) if isinstance(__snake_case , __snake_case ) else 1
# get prompt text embeddings
__a =self.tokenizer(
__snake_case , padding='max_length' , max_length=self.tokenizer.model_max_length , return_tensors='pt' , )
__a =text_inputs.input_ids
if text_input_ids.shape[-1] > self.tokenizer.model_max_length:
__a =self.tokenizer.batch_decode(text_input_ids[:, self.tokenizer.model_max_length :] )
logger.warning(
'The following part of your input was truncated because CLIP can only handle sequences up to'
f' {self.tokenizer.model_max_length} tokens: {removed_text}' )
__a =text_input_ids[:, : self.tokenizer.model_max_length]
__a =self.text_encoder(text_input_ids.to(self.device ) )[0]
# NOTE: This additional step of normalizing the text embeddings is from VQ-Diffusion.
# While CLIP does normalize the pooled output of the text transformer when combining
# the image and text embeddings, CLIP does not directly normalize the last hidden state.
#
# CLIP normalizing the pooled output.
# https://github.com/huggingface/transformers/blob/d92e22d1f28324f513f3080e5c47c071a3916721/src/transformers/models/clip/modeling_clip.py#L1052-L1053
__a =prompt_embeds / prompt_embeds.norm(dim=-1 , keepdim=__snake_case )
# duplicate text embeddings for each generation per prompt
__a =prompt_embeds.repeat_interleave(__snake_case , dim=0 )
if do_classifier_free_guidance:
if self.learned_classifier_free_sampling_embeddings.learnable:
__a =self.learned_classifier_free_sampling_embeddings.embeddings
__a =negative_prompt_embeds.unsqueeze(0 ).repeat(__snake_case , 1 , 1 )
else:
__a =[''] * batch_size
__a =text_input_ids.shape[-1]
__a =self.tokenizer(
__snake_case , padding='max_length' , max_length=__snake_case , truncation=__snake_case , return_tensors='pt' , )
__a =self.text_encoder(uncond_input.input_ids.to(self.device ) )[0]
# See comment for normalizing text embeddings
__a =negative_prompt_embeds / negative_prompt_embeds.norm(dim=-1 , keepdim=__snake_case )
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
__a =negative_prompt_embeds.shape[1]
__a =negative_prompt_embeds.repeat(1 , __snake_case , 1 )
__a =negative_prompt_embeds.view(batch_size * num_images_per_prompt , __snake_case , -1 )
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch
# to avoid doing two forward passes
__a =torch.cat([negative_prompt_embeds, prompt_embeds] )
return prompt_embeds
@torch.no_grad()
def __call__( self , __snake_case , __snake_case = 100 , __snake_case = 5.0 , __snake_case = 1.0 , __snake_case = 1 , __snake_case = None , __snake_case = None , __snake_case = "pil" , __snake_case = True , __snake_case = None , __snake_case = 1 , ) -> Union[ImagePipelineOutput, Tuple]:
'''simple docstring'''
if isinstance(__snake_case , __snake_case ):
__a =1
elif isinstance(__snake_case , __snake_case ):
__a =len(__snake_case )
else:
raise ValueError(f'`prompt` has to be of type `str` or `list` but is {type(__snake_case )}' )
__a =batch_size * num_images_per_prompt
__a =guidance_scale > 1.0
__a =self._encode_prompt(__snake_case , __snake_case , __snake_case )
if (callback_steps is None) or (
callback_steps is not None and (not isinstance(__snake_case , __snake_case ) or callback_steps <= 0)
):
raise ValueError(
f'`callback_steps` has to be a positive integer but is {callback_steps} of type'
f' {type(__snake_case )}.' )
# get the initial completely masked latents unless the user supplied it
__a =(batch_size, self.transformer.num_latent_pixels)
if latents is None:
__a =self.transformer.num_vector_embeds - 1
__a =torch.full(__snake_case , __snake_case ).to(self.device )
else:
if latents.shape != latents_shape:
raise ValueError(f'Unexpected latents shape, got {latents.shape}, expected {latents_shape}' )
if (latents < 0).any() or (latents >= self.transformer.num_vector_embeds).any():
raise ValueError(
'Unexpected latents value(s). All latents be valid embedding indices i.e. in the range 0,'
f' {self.transformer.num_vector_embeds - 1} (inclusive).' )
__a =latents.to(self.device )
# set timesteps
self.scheduler.set_timesteps(__snake_case , device=self.device )
__a =self.scheduler.timesteps.to(self.device )
__a =latents
for i, t in enumerate(self.progress_bar(__snake_case ) ):
# expand the sample if we are doing classifier free guidance
__a =torch.cat([sample] * 2 ) if do_classifier_free_guidance else sample
# predict the un-noised image
# model_output == `log_p_x_0`
__a =self.transformer(__snake_case , encoder_hidden_states=__snake_case , timestep=__snake_case ).sample
if do_classifier_free_guidance:
__a , __a =model_output.chunk(2 )
__a =model_output_uncond + guidance_scale * (model_output_text - model_output_uncond)
model_output -= torch.logsumexp(__snake_case , dim=1 , keepdim=__snake_case )
__a =self.truncate(__snake_case , __snake_case )
# remove `log(0)`'s (`-inf`s)
__a =model_output.clamp(-70 )
# compute the previous noisy sample x_t -> x_t-1
__a =self.scheduler.step(__snake_case , timestep=__snake_case , sample=__snake_case , generator=__snake_case ).prev_sample
# call the callback, if provided
if callback is not None and i % callback_steps == 0:
callback(__snake_case , __snake_case , __snake_case )
__a =self.vqvae.config.vq_embed_dim
__a =(batch_size, self.transformer.height, self.transformer.width, embedding_channels)
__a =self.vqvae.quantize.get_codebook_entry(__snake_case , shape=__snake_case )
__a =self.vqvae.decode(__snake_case , force_not_quantize=__snake_case ).sample
__a =(image / 2 + 0.5).clamp(0 , 1 )
__a =image.cpu().permute(0 , 2 , 3 , 1 ).numpy()
if output_type == "pil":
__a =self.numpy_to_pil(__snake_case )
if not return_dict:
return (image,)
return ImagePipelineOutput(images=__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case ) -> torch.FloatTensor:
'''simple docstring'''
__a , __a =torch.sort(__snake_case , 1 , descending=__snake_case )
__a =torch.exp(__snake_case )
__a =sorted_p_x_0.cumsum(dim=1 ) < truncation_rate
# Ensure that at least the largest probability is not zeroed out
__a =torch.full_like(keep_mask[:, 0:1, :] , __snake_case )
__a =torch.cat((all_true, keep_mask) , dim=1 )
__a =keep_mask[:, :-1, :]
__a =keep_mask.gather(1 , indices.argsort(1 ) )
__a =log_p_x_0.clone()
__a =-torch.inf # -inf = log(0)
return rv
| 308
|
import json
import os
import unittest
from transformers.models.blenderbot_small.tokenization_blenderbot_small import (
VOCAB_FILES_NAMES,
BlenderbotSmallTokenizer,
)
from ...test_tokenization_common import TokenizerTesterMixin
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = BlenderbotSmallTokenizer
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
super().setUp()
__a =['__start__', 'adapt', 'act', 'ap@@', 'te', '__end__', '__unk__']
__a =dict(zip(__snake_case , range(len(__snake_case ) ) ) )
__a =['#version: 0.2', 'a p', 't e</w>', 'ap t</w>', 'a d', 'ad apt</w>', 'a c', 'ac t</w>', '']
__a ={'unk_token': '__unk__', 'bos_token': '__start__', 'eos_token': '__end__'}
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(__snake_case ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(__snake_case ) )
def __magic_name__ ( self , **__snake_case ) -> Any:
'''simple docstring'''
kwargs.update(self.special_tokens_map )
return BlenderbotSmallTokenizer.from_pretrained(self.tmpdirname , **__snake_case )
def __magic_name__ ( self , __snake_case ) -> List[Any]:
'''simple docstring'''
__a ='adapt act apte'
__a ='adapt act apte'
return input_text, output_text
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
__a ='adapt act apte'
__a =['adapt', 'act', 'ap@@', 'te']
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
__a =[tokenizer.bos_token] + tokens + [tokenizer.eos_token]
__a =[0, 1, 2, 3, 4, 5]
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
assert tok('sam' ).input_ids == [1384]
__a ='I am a small frog.'
__a =tok([src_text] , padding=__snake_case , truncation=__snake_case )['input_ids']
__a =tok.batch_decode(__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case )[0]
assert src_text != decoded # I wish it did!
assert decoded == "i am a small frog ."
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
__a ='I am a small frog .'
__a ='.'
__a =tok(__snake_case )['input_ids']
__a =tok(__snake_case )['input_ids']
assert encoded[-1] == encoded_dot[0]
| 308
| 1
|
from math import isqrt
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
return all(number % divisor != 0 for divisor in range(2 , isqrt(_snake_case ) + 1 ) )
def UpperCamelCase_( _snake_case : int = 10**6 ):
"""simple docstring"""
__a =0
__a =1
__a =7
while prime_candidate < max_prime:
primes_count += is_prime(_snake_case )
cube_index += 1
prime_candidate += 6 * cube_index
return primes_count
if __name__ == "__main__":
print(f'''{solution() = }''')
| 308
|
import unittest
from transformers import load_tool
from transformers.utils import is_torch_available
if is_torch_available():
import torch
from transformers.testing_utils import require_torch
from .test_tools_common import ToolTesterMixin
@require_torch
class __magic_name__ ( unittest.TestCase , lowerCAmelCase_ ):
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =load_tool('text-to-speech' )
self.tool.setup()
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
| 308
| 1
|
import copy
import os
from typing import TYPE_CHECKING, List, Union
if TYPE_CHECKING:
pass
from ...configuration_utils import PretrainedConfig
from ...utils import logging
_lowerCAmelCase : List[str] = logging.get_logger(__name__)
_lowerCAmelCase : List[str] = {
"kakaobrain/align-base": "https://huggingface.co/kakaobrain/align-base/resolve/main/config.json",
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'align_text_model'
def __init__( self , __snake_case=3_0522 , __snake_case=768 , __snake_case=12 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=2 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=0 , __snake_case="absolute" , __snake_case=True , **__snake_case , ) -> Any:
'''simple docstring'''
super().__init__(**__snake_case )
__a =vocab_size
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_act
__a =intermediate_size
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_vocab_size
__a =initializer_range
__a =layer_norm_eps
__a =position_embedding_type
__a =use_cache
__a =pad_token_id
@classmethod
def __magic_name__ ( cls , __snake_case , **__snake_case ) -> "PretrainedConfig":
'''simple docstring'''
cls._set_token_in_kwargs(__snake_case )
__a , __a =cls.get_config_dict(__snake_case , **__snake_case )
# get the text config dict if we are loading from AlignConfig
if config_dict.get('model_type' ) == "align":
__a =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(__snake_case , **__snake_case )
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'align_vision_model'
def __init__( self , __snake_case = 3 , __snake_case = 600 , __snake_case = 2.0 , __snake_case = 3.1 , __snake_case = 8 , __snake_case = [3, 3, 5, 3, 5, 5, 3] , __snake_case = [32, 16, 24, 40, 80, 112, 192] , __snake_case = [16, 24, 40, 80, 112, 192, 320] , __snake_case = [] , __snake_case = [1, 2, 2, 2, 1, 2, 1] , __snake_case = [1, 2, 2, 3, 3, 4, 1] , __snake_case = [1, 6, 6, 6, 6, 6, 6] , __snake_case = 0.25 , __snake_case = "swish" , __snake_case = 2560 , __snake_case = "mean" , __snake_case = 0.02 , __snake_case = 0.001 , __snake_case = 0.99 , __snake_case = 0.2 , **__snake_case , ) -> Union[str, Any]:
'''simple docstring'''
super().__init__(**__snake_case )
__a =num_channels
__a =image_size
__a =width_coefficient
__a =depth_coefficient
__a =depth_divisor
__a =kernel_sizes
__a =in_channels
__a =out_channels
__a =depthwise_padding
__a =strides
__a =num_block_repeats
__a =expand_ratios
__a =squeeze_expansion_ratio
__a =hidden_act
__a =hidden_dim
__a =pooling_type
__a =initializer_range
__a =batch_norm_eps
__a =batch_norm_momentum
__a =drop_connect_rate
__a =sum(__snake_case ) * 4
@classmethod
def __magic_name__ ( cls , __snake_case , **__snake_case ) -> "PretrainedConfig":
'''simple docstring'''
cls._set_token_in_kwargs(__snake_case )
__a , __a =cls.get_config_dict(__snake_case , **__snake_case )
# get the vision config dict if we are loading from AlignConfig
if config_dict.get('model_type' ) == "align":
__a =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(__snake_case , **__snake_case )
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'align'
SCREAMING_SNAKE_CASE = True
def __init__( self , __snake_case=None , __snake_case=None , __snake_case=640 , __snake_case=1.0 , __snake_case=0.02 , **__snake_case , ) -> Union[str, Any]:
'''simple docstring'''
super().__init__(**__snake_case )
if text_config is None:
__a ={}
logger.info('text_config is None. Initializing the AlignTextConfig with default values.' )
if vision_config is None:
__a ={}
logger.info('vision_config is None. Initializing the AlignVisionConfig with default values.' )
__a =AlignTextConfig(**__snake_case )
__a =AlignVisionConfig(**__snake_case )
__a =projection_dim
__a =temperature_init_value
__a =initializer_range
@classmethod
def __magic_name__ ( cls , __snake_case , __snake_case , **__snake_case ) -> int:
'''simple docstring'''
return cls(text_config=text_config.to_dict() , vision_config=vision_config.to_dict() , **__snake_case )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =copy.deepcopy(self.__dict__ )
__a =self.text_config.to_dict()
__a =self.vision_config.to_dict()
__a =self.__class__.model_type
return output
| 308
|
import flax.linen as nn
import jax.numpy as jnp
from .attention_flax import FlaxTransformeraDModel
from .resnet_flax import FlaxDownsampleaD, FlaxResnetBlockaD, FlaxUpsampleaD
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> Optional[Any]:
'''simple docstring'''
__a =()
for resnet, attn in zip(self.resnets , self.attentions ):
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case=True ) -> Optional[int]:
'''simple docstring'''
__a =()
for resnet in self.resnets:
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet, attn in zip(self.resnets , self.attentions ):
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet in self.resnets:
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# there is always at least one resnet
__a =[
FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
]
__a =[]
for _ in range(self.num_layers ):
__a =FlaxTransformeraDModel(
in_channels=self.in_channels , n_heads=self.num_attention_heads , d_head=self.in_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
__a =attentions
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[str]:
'''simple docstring'''
__a =self.resnets[0](__snake_case , __snake_case )
for attn, resnet in zip(self.attentions , self.resnets[1:] ):
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
return hidden_states
| 308
| 1
|
import argparse
import json
from collections import OrderedDict
from pathlib import Path
import requests
import torch
from huggingface_hub import hf_hub_download
from PIL import Image
from transformers import (
SegformerConfig,
SegformerForImageClassification,
SegformerForSemanticSegmentation,
SegformerImageProcessor,
)
from transformers.utils import logging
logging.set_verbosity_info()
_lowerCAmelCase : List[str] = logging.get_logger(__name__)
def UpperCamelCase_( _snake_case : Tuple , _snake_case : int=False ):
"""simple docstring"""
__a =OrderedDict()
for key, value in state_dict.items():
if encoder_only and not key.startswith('head' ):
__a ='segformer.encoder.' + key
if key.startswith('backbone' ):
__a =key.replace('backbone' , 'segformer.encoder' )
if "patch_embed" in key:
# replace for example patch_embed1 by patch_embeddings.0
__a =key[key.find('patch_embed' ) + len('patch_embed' )]
__a =key.replace(F'patch_embed{idx}' , F'patch_embeddings.{int(_snake_case )-1}' )
if "norm" in key:
__a =key.replace('norm' , 'layer_norm' )
if "segformer.encoder.layer_norm" in key:
# replace for example layer_norm1 by layer_norm.0
__a =key[key.find('segformer.encoder.layer_norm' ) + len('segformer.encoder.layer_norm' )]
__a =key.replace(F'layer_norm{idx}' , F'layer_norm.{int(_snake_case )-1}' )
if "layer_norm1" in key:
__a =key.replace('layer_norm1' , 'layer_norm_1' )
if "layer_norm2" in key:
__a =key.replace('layer_norm2' , 'layer_norm_2' )
if "block" in key:
# replace for example block1 by block.0
__a =key[key.find('block' ) + len('block' )]
__a =key.replace(F'block{idx}' , F'block.{int(_snake_case )-1}' )
if "attn.q" in key:
__a =key.replace('attn.q' , 'attention.self.query' )
if "attn.proj" in key:
__a =key.replace('attn.proj' , 'attention.output.dense' )
if "attn" in key:
__a =key.replace('attn' , 'attention.self' )
if "fc1" in key:
__a =key.replace('fc1' , 'dense1' )
if "fc2" in key:
__a =key.replace('fc2' , 'dense2' )
if "linear_pred" in key:
__a =key.replace('linear_pred' , 'classifier' )
if "linear_fuse" in key:
__a =key.replace('linear_fuse.conv' , 'linear_fuse' )
__a =key.replace('linear_fuse.bn' , 'batch_norm' )
if "linear_c" in key:
# replace for example linear_c4 by linear_c.3
__a =key[key.find('linear_c' ) + len('linear_c' )]
__a =key.replace(F'linear_c{idx}' , F'linear_c.{int(_snake_case )-1}' )
if key.startswith('head' ):
__a =key.replace('head' , 'classifier' )
__a =value
return new_state_dict
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : Union[str, Any] ):
"""simple docstring"""
for i in range(config.num_encoder_blocks ):
for j in range(config.depths[i] ):
# read in weights + bias of keys and values (which is a single matrix in the original implementation)
__a =state_dict.pop(F'segformer.encoder.block.{i}.{j}.attention.self.kv.weight' )
__a =state_dict.pop(F'segformer.encoder.block.{i}.{j}.attention.self.kv.bias' )
# next, add keys and values (in that order) to the state dict
__a =kv_weight[
: config.hidden_sizes[i], :
]
__a =kv_bias[: config.hidden_sizes[i]]
__a =kv_weight[
config.hidden_sizes[i] :, :
]
__a =kv_bias[
config.hidden_sizes[i] :
]
def UpperCamelCase_( ):
"""simple docstring"""
__a ='http://images.cocodataset.org/val2017/000000039769.jpg'
__a =Image.open(requests.get(_snake_case , stream=_snake_case ).raw )
return image
@torch.no_grad()
def UpperCamelCase_( _snake_case : Optional[int] , _snake_case : Union[str, Any] , _snake_case : Optional[int] ):
"""simple docstring"""
__a =SegformerConfig()
__a =False
# set attributes based on model_name
__a ='huggingface/label-files'
if "segformer" in model_name:
__a =model_name[len('segformer.' ) : len('segformer.' ) + 2]
if "ade" in model_name:
__a =150
__a ='ade20k-id2label.json'
__a =(1, 150, 128, 128)
elif "city" in model_name:
__a =19
__a ='cityscapes-id2label.json'
__a =(1, 19, 128, 128)
else:
raise ValueError(F'Model {model_name} not supported' )
elif "mit" in model_name:
__a =True
__a =model_name[4:6]
__a =1000
__a ='imagenet-1k-id2label.json'
__a =(1, 1000)
else:
raise ValueError(F'Model {model_name} not supported' )
# set config attributes
__a =json.load(open(hf_hub_download(_snake_case , _snake_case , repo_type='dataset' ) , 'r' ) )
__a ={int(_snake_case ): v for k, v in idalabel.items()}
__a =idalabel
__a ={v: k for k, v in idalabel.items()}
if size == "b0":
pass
elif size == "b1":
__a =[64, 128, 320, 512]
__a =256
elif size == "b2":
__a =[64, 128, 320, 512]
__a =768
__a =[3, 4, 6, 3]
elif size == "b3":
__a =[64, 128, 320, 512]
__a =768
__a =[3, 4, 18, 3]
elif size == "b4":
__a =[64, 128, 320, 512]
__a =768
__a =[3, 8, 27, 3]
elif size == "b5":
__a =[64, 128, 320, 512]
__a =768
__a =[3, 6, 40, 3]
else:
raise ValueError(F'Size {size} not supported' )
# load image processor (only resize + normalize)
__a =SegformerImageProcessor(
image_scale=(512, 512) , keep_ratio=_snake_case , align=_snake_case , do_random_crop=_snake_case )
# prepare image
__a =prepare_img()
__a =image_processor(images=_snake_case , return_tensors='pt' ).pixel_values
logger.info(F'Converting model {model_name}...' )
# load original state dict
if encoder_only:
__a =torch.load(_snake_case , map_location=torch.device('cpu' ) )
else:
__a =torch.load(_snake_case , map_location=torch.device('cpu' ) )['state_dict']
# rename keys
__a =rename_keys(_snake_case , encoder_only=_snake_case )
if not encoder_only:
del state_dict["decode_head.conv_seg.weight"]
del state_dict["decode_head.conv_seg.bias"]
# key and value matrices need special treatment
read_in_k_v(_snake_case , _snake_case )
# create HuggingFace model and load state dict
if encoder_only:
__a =False
__a =SegformerForImageClassification(_snake_case )
else:
__a =SegformerForSemanticSegmentation(_snake_case )
model.load_state_dict(_snake_case )
model.eval()
# forward pass
__a =model(_snake_case )
__a =outputs.logits
# set expected_slice based on model name
# ADE20k checkpoints
if model_name == "segformer.b0.512x512.ade.160k":
__a =torch.tensor(
[
[[-4.6_310, -5.5_232, -6.2_356], [-5.1_921, -6.1_444, -6.5_996], [-5.4_424, -6.2_790, -6.7_574]],
[[-12.1_391, -13.3_122, -13.9_554], [-12.8_732, -13.9_352, -14.3_563], [-12.9_438, -13.8_226, -14.2_513]],
[[-12.5_134, -13.4_686, -14.4_915], [-12.8_669, -14.4_343, -14.7_758], [-13.2_523, -14.5_819, -15.0_694]],
] )
elif model_name == "segformer.b1.512x512.ade.160k":
__a =torch.tensor(
[
[[-7.5_820, -8.7_231, -8.3_215], [-8.0_600, -10.3_529, -10.0_304], [-7.5_208, -9.4_103, -9.6_239]],
[[-12.6_918, -13.8_994, -13.7_137], [-13.3_196, -15.7_523, -15.4_789], [-12.9_343, -14.8_757, -14.9_689]],
[[-11.1_911, -11.9_421, -11.3_243], [-11.3_342, -13.6_839, -13.3_581], [-10.3_909, -12.1_832, -12.4_858]],
] )
elif model_name == "segformer.b2.512x512.ade.160k":
__a =torch.tensor(
[
[[-11.8_173, -14.3_850, -16.3_128], [-14.5_648, -16.5_804, -18.6_568], [-14.7_223, -15.7_387, -18.4_218]],
[[-15.7_290, -17.9_171, -19.4_423], [-18.3_105, -19.9_448, -21.4_661], [-17.9_296, -18.6_497, -20.7_910]],
[[-15.0_783, -17.0_336, -18.2_789], [-16.8_771, -18.6_870, -20.1_612], [-16.2_454, -17.1_426, -19.5_055]],
] )
elif model_name == "segformer.b3.512x512.ade.160k":
__a =torch.tensor(
[
[[-9.0_878, -10.2_081, -10.1_891], [-9.3_144, -10.7_941, -10.9_843], [-9.2_294, -10.3_855, -10.5_704]],
[[-12.2_316, -13.9_068, -13.6_102], [-12.9_161, -14.3_702, -14.3_235], [-12.5_233, -13.7_174, -13.7_932]],
[[-14.6_275, -15.2_490, -14.9_727], [-14.3_400, -15.9_687, -16.2_827], [-14.1_484, -15.4_033, -15.8_937]],
] )
elif model_name == "segformer.b4.512x512.ade.160k":
__a =torch.tensor(
[
[[-12.3_144, -13.2_447, -14.0_802], [-13.3_614, -14.5_816, -15.6_117], [-13.3_340, -14.4_433, -16.2_219]],
[[-19.2_781, -20.4_128, -20.7_506], [-20.6_153, -21.6_566, -22.0_998], [-19.9_800, -21.0_430, -22.1_494]],
[[-18.8_739, -19.7_804, -21.1_834], [-20.1_233, -21.6_765, -23.2_944], [-20.0_315, -21.2_641, -23.6_944]],
] )
elif model_name == "segformer.b5.640x640.ade.160k":
__a =torch.tensor(
[
[[-9.5_524, -12.0_835, -11.7_348], [-10.5_229, -13.6_446, -14.5_662], [-9.5_842, -12.8_851, -13.9_414]],
[[-15.3_432, -17.5_323, -17.0_818], [-16.3_330, -18.9_255, -19.2_101], [-15.1_340, -17.7_848, -18.3_971]],
[[-12.6_072, -14.9_486, -14.6_631], [-13.7_629, -17.0_907, -17.7_745], [-12.7_899, -16.1_695, -17.1_671]],
] )
# Cityscapes checkpoints
elif model_name == "segformer.b0.1024x1024.city.160k":
__a =torch.tensor(
[
[[-11.9_295, -13.4_057, -14.8_106], [-13.3_431, -14.8_179, -15.3_781], [-14.2_836, -15.5_942, -16.1_588]],
[[-11.4_906, -12.8_067, -13.6_564], [-13.1_189, -14.0_500, -14.1_543], [-13.8_748, -14.5_136, -14.8_789]],
[[0.5_374, 0.1_067, -0.4_742], [0.1_141, -0.2_255, -0.7_099], [-0.3_000, -0.5_924, -1.3_105]],
] )
elif model_name == "segformer.b0.512x1024.city.160k":
__a =torch.tensor(
[
[[-7.8_217, -9.8_767, -10.1_717], [-9.4_438, -10.9_058, -11.4_047], [-9.7_939, -12.3_495, -12.1_079]],
[[-7.1_514, -9.5_336, -10.0_860], [-9.7_776, -11.6_822, -11.8_439], [-10.1_411, -12.7_655, -12.8_972]],
[[0.3_021, 0.0_805, -0.2_310], [-0.0_328, -0.1_605, -0.2_714], [-0.1_408, -0.5_477, -0.6_976]],
] )
elif model_name == "segformer.b0.640x1280.city.160k":
__a =torch.tensor(
[
[
[-1.1_3_7_2e0_1, -1.2_7_8_7e0_1, -1.3_4_7_7e0_1],
[-1.2_5_3_6e0_1, -1.4_1_9_4e0_1, -1.4_4_0_9e0_1],
[-1.3_2_1_7e0_1, -1.4_8_8_8e0_1, -1.5_3_2_7e0_1],
],
[
[-1.4_7_9_1e0_1, -1.7_1_2_2e0_1, -1.8_2_7_7e0_1],
[-1.7_1_6_3e0_1, -1.9_1_9_2e0_1, -1.9_5_3_3e0_1],
[-1.7_8_9_7e0_1, -1.9_9_9_1e0_1, -2.0_3_1_5e0_1],
],
[
[7.6_7_2_3e-0_1, 4.1_9_2_1e-0_1, -7.7_8_7_8e-0_2],
[4.7_7_7_2e-0_1, 9.5_5_5_7e-0_3, -2.8_0_8_2e-0_1],
[3.6_0_3_2e-0_1, -2.4_8_2_6e-0_1, -5.1_1_6_8e-0_1],
],
] )
elif model_name == "segformer.b0.768x768.city.160k":
__a =torch.tensor(
[
[[-9.4_959, -11.3_087, -11.7_479], [-11.0_025, -12.6_540, -12.3_319], [-11.4_064, -13.0_487, -12.9_905]],
[[-9.8_905, -11.3_084, -12.0_854], [-11.1_726, -12.7_698, -12.9_583], [-11.5_985, -13.3_278, -14.1_774]],
[[0.2_213, 0.0_192, -0.2_466], [-0.1_731, -0.4_213, -0.4_874], [-0.3_126, -0.6_541, -1.1_389]],
] )
elif model_name == "segformer.b1.1024x1024.city.160k":
__a =torch.tensor(
[
[[-13.5_748, -13.9_111, -12.6_500], [-14.3_500, -15.3_683, -14.2_328], [-14.7_532, -16.0_424, -15.6_087]],
[[-17.1_651, -15.8_725, -12.9_653], [-17.2_580, -17.3_718, -14.8_223], [-16.6_058, -16.8_783, -16.7_452]],
[[-3.6_456, -3.0_209, -1.4_203], [-3.0_797, -3.1_959, -2.0_000], [-1.8_757, -1.9_217, -1.6_997]],
] )
elif model_name == "segformer.b2.1024x1024.city.160k":
__a =torch.tensor(
[
[[-16.0_976, -16.4_856, -17.3_962], [-16.6_234, -19.0_342, -19.7_685], [-16.0_900, -18.0_661, -19.1_180]],
[[-18.4_750, -18.8_488, -19.5_074], [-19.4_030, -22.1_570, -22.5_977], [-19.1_191, -20.8_486, -22.3_783]],
[[-4.5_178, -5.5_037, -6.5_109], [-5.0_884, -7.2_174, -8.0_334], [-4.4_156, -5.8_117, -7.2_970]],
] )
elif model_name == "segformer.b3.1024x1024.city.160k":
__a =torch.tensor(
[
[[-14.2_081, -14.4_732, -14.1_977], [-14.5_867, -16.4_423, -16.6_356], [-13.4_441, -14.9_685, -16.8_696]],
[[-14.4_576, -14.7_073, -15.0_451], [-15.0_816, -17.6_237, -17.9_873], [-14.4_213, -16.0_199, -18.5_992]],
[[-4.7_349, -4.9_588, -5.0_966], [-4.3_210, -6.9_325, -7.2_591], [-3.4_312, -4.7_484, -7.1_917]],
] )
elif model_name == "segformer.b4.1024x1024.city.160k":
__a =torch.tensor(
[
[[-11.7_737, -11.9_526, -11.3_273], [-13.6_692, -14.4_574, -13.8_878], [-13.8_937, -14.6_924, -15.9_345]],
[[-14.6_706, -14.5_330, -14.1_306], [-16.1_502, -16.8_180, -16.4_269], [-16.8_338, -17.8_939, -20.1_746]],
[[1.0_491, 0.8_289, 1.0_310], [1.1_044, 0.5_219, 0.8_055], [1.0_899, 0.6_926, 0.5_590]],
] )
elif model_name == "segformer.b5.1024x1024.city.160k":
__a =torch.tensor(
[
[[-12.5_641, -13.4_777, -13.0_684], [-13.9_587, -15.8_983, -16.6_557], [-13.3_109, -15.7_350, -16.3_141]],
[[-14.7_074, -15.4_352, -14.5_944], [-16.6_353, -18.1_663, -18.6_120], [-15.1_702, -18.0_329, -18.1_547]],
[[-1.7_990, -2.0_951, -1.7_784], [-2.6_397, -3.8_245, -3.9_686], [-1.5_264, -2.8_126, -2.9_316]],
] )
else:
__a =logits.argmax(-1 ).item()
print('Predicted class:' , model.config.idalabel[predicted_class_idx] )
# verify logits
if not encoder_only:
assert logits.shape == expected_shape
assert torch.allclose(logits[0, :3, :3, :3] , _snake_case , atol=1e-2 )
# finally, save model and image processor
logger.info(F'Saving PyTorch model and image processor to {pytorch_dump_folder_path}...' )
Path(_snake_case ).mkdir(exist_ok=_snake_case )
model.save_pretrained(_snake_case )
image_processor.save_pretrained(_snake_case )
if __name__ == "__main__":
_lowerCAmelCase : List[str] = argparse.ArgumentParser()
parser.add_argument(
"--model_name",
default="segformer.b0.512x512.ade.160k",
type=str,
help="Name of the model you'd like to convert.",
)
parser.add_argument(
"--checkpoint_path", default=None, type=str, help="Path to the original PyTorch checkpoint (.pth file)."
)
parser.add_argument(
"--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model."
)
_lowerCAmelCase : Union[str, Any] = parser.parse_args()
convert_segformer_checkpoint(args.model_name, args.checkpoint_path, args.pytorch_dump_folder_path)
| 308
|
import json
import os
import subprocess
import unittest
from ast import literal_eval
import pytest
from parameterized import parameterized, parameterized_class
from . import is_sagemaker_available
if is_sagemaker_available():
from sagemaker import Session, TrainingJobAnalytics
from sagemaker.huggingface import HuggingFace
@pytest.mark.skipif(
literal_eval(os.getenv('TEST_SAGEMAKER' , 'False' ) ) is not True , reason='Skipping test because should only be run when releasing minor transformers version' , )
@pytest.mark.usefixtures('sm_env' )
@parameterized_class(
[
{
'framework': 'pytorch',
'script': 'run_glue.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_5_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'pytorch',
'script': 'run_ddp.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'tensorflow',
'script': 'run_tf_dist.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.6, 'eval_loss': 0.7},
},
] )
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.framework == "pytorch":
subprocess.run(
f'cp ./examples/pytorch/text-classification/run_glue.py {self.env.test_path}/run_glue.py'.split() , encoding='utf-8' , check=__snake_case , )
assert hasattr(self , 'env' )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
__a =f'{self.env.base_job_name}-{instance_count}-{"ddp" if "ddp" in self.script else "smd"}'
# distributed data settings
__a ={'smdistributed': {'dataparallel': {'enabled': True}}} if self.script != 'run_ddp.py' else None
# creates estimator
return HuggingFace(
entry_point=self.script , source_dir=self.env.test_path , role=self.env.role , image_uri=self.env.image_uri , base_job_name=__snake_case , instance_count=__snake_case , instance_type=self.instance_type , debugger_hook_config=__snake_case , hyperparameters={**self.env.distributed_hyperparameters, 'model_name_or_path': self.model_name_or_path} , metric_definitions=self.env.metric_definitions , distribution=__snake_case , py_version='py36' , )
def __magic_name__ ( self , __snake_case ) -> Optional[Any]:
'''simple docstring'''
TrainingJobAnalytics(__snake_case ).export_csv(f'{self.env.test_path}/{job_name}_metrics.csv' )
@parameterized.expand([(2,)] )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
# create estimator
__a =self.create_estimator(__snake_case )
# run training
estimator.fit()
# result dataframe
__a =TrainingJobAnalytics(estimator.latest_training_job.name ).dataframe()
# extract kpis
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_accuracy']['value'] )
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_loss']['value'] )
# get train time from SageMaker job, this includes starting, preprocessing, stopping
__a =(
Session().describe_training_job(estimator.latest_training_job.name ).get('TrainingTimeInSeconds' , 99_9999 )
)
# assert kpis
assert train_runtime <= self.results["train_runtime"]
assert all(t >= self.results['eval_accuracy'] for t in eval_accuracy )
assert all(t <= self.results['eval_loss'] for t in eval_loss )
# dump tests result into json file to share in PR
with open(f'{estimator.latest_training_job.name}.json' , 'w' ) as outfile:
json.dump({'train_time': train_runtime, 'eval_accuracy': eval_accuracy, 'eval_loss': eval_loss} , __snake_case )
| 308
| 1
|
import argparse
import torch
from datasets import load_dataset
from donut import DonutModel
from transformers import (
DonutImageProcessor,
DonutProcessor,
DonutSwinConfig,
DonutSwinModel,
MBartConfig,
MBartForCausalLM,
VisionEncoderDecoderModel,
XLMRobertaTokenizerFast,
)
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =model.config
__a =DonutSwinConfig(
image_size=original_config.input_size , patch_size=4 , depths=original_config.encoder_layer , num_heads=[4, 8, 16, 32] , window_size=original_config.window_size , embed_dim=128 , )
__a =MBartConfig(
is_decoder=_snake_case , is_encoder_decoder=_snake_case , add_cross_attention=_snake_case , decoder_layers=original_config.decoder_layer , max_position_embeddings=original_config.max_position_embeddings , vocab_size=len(
model.decoder.tokenizer ) , scale_embedding=_snake_case , add_final_layer_norm=_snake_case , )
return encoder_config, decoder_config
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
if "encoder.model" in name:
__a =name.replace('encoder.model' , 'encoder' )
if "decoder.model" in name:
__a =name.replace('decoder.model' , 'decoder' )
if "patch_embed.proj" in name:
__a =name.replace('patch_embed.proj' , 'embeddings.patch_embeddings.projection' )
if "patch_embed.norm" in name:
__a =name.replace('patch_embed.norm' , 'embeddings.norm' )
if name.startswith('encoder' ):
if "layers" in name:
__a ='encoder.' + name
if "attn.proj" in name:
__a =name.replace('attn.proj' , 'attention.output.dense' )
if "attn" in name and "mask" not in name:
__a =name.replace('attn' , 'attention.self' )
if "norm1" in name:
__a =name.replace('norm1' , 'layernorm_before' )
if "norm2" in name:
__a =name.replace('norm2' , 'layernorm_after' )
if "mlp.fc1" in name:
__a =name.replace('mlp.fc1' , 'intermediate.dense' )
if "mlp.fc2" in name:
__a =name.replace('mlp.fc2' , 'output.dense' )
if name == "encoder.norm.weight":
__a ='encoder.layernorm.weight'
if name == "encoder.norm.bias":
__a ='encoder.layernorm.bias'
return name
def UpperCamelCase_( _snake_case : Tuple , _snake_case : str ):
"""simple docstring"""
for key in orig_state_dict.copy().keys():
__a =orig_state_dict.pop(_snake_case )
if "qkv" in key:
__a =key.split('.' )
__a =int(key_split[3] )
__a =int(key_split[5] )
__a =model.encoder.encoder.layers[layer_num].blocks[block_num].attention.self.all_head_size
if "weight" in key:
__a =val[:dim, :]
__a =val[dim : dim * 2, :]
__a =val[-dim:, :]
else:
__a =val[:dim]
__a =val[dim : dim * 2]
__a =val[-dim:]
elif "attn_mask" in key or key in ["encoder.model.norm.weight", "encoder.model.norm.bias"]:
# HuggingFace implementation doesn't use attn_mask buffer
# and model doesn't use final LayerNorms for the encoder
pass
else:
__a =val
return orig_state_dict
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Union[str, Any]=None , _snake_case : List[Any]=False ):
"""simple docstring"""
__a =DonutModel.from_pretrained(_snake_case ).eval()
# load HuggingFace model
__a , __a =get_configs(_snake_case )
__a =DonutSwinModel(_snake_case )
__a =MBartForCausalLM(_snake_case )
__a =VisionEncoderDecoderModel(encoder=_snake_case , decoder=_snake_case )
model.eval()
__a =original_model.state_dict()
__a =convert_state_dict(_snake_case , _snake_case )
model.load_state_dict(_snake_case )
# verify results on scanned document
__a =load_dataset('hf-internal-testing/example-documents' )
__a =dataset['test'][0]['image'].convert('RGB' )
__a =XLMRobertaTokenizerFast.from_pretrained(_snake_case , from_slow=_snake_case )
__a =DonutImageProcessor(
do_align_long_axis=original_model.config.align_long_axis , size=original_model.config.input_size[::-1] )
__a =DonutProcessor(_snake_case , _snake_case )
__a =processor(_snake_case , return_tensors='pt' ).pixel_values
if model_name == "naver-clova-ix/donut-base-finetuned-docvqa":
__a ='<s_docvqa><s_question>{user_input}</s_question><s_answer>'
__a ='When is the coffee break?'
__a =task_prompt.replace('{user_input}' , _snake_case )
elif model_name == "naver-clova-ix/donut-base-finetuned-rvlcdip":
__a ='<s_rvlcdip>'
elif model_name in [
"naver-clova-ix/donut-base-finetuned-cord-v1",
"naver-clova-ix/donut-base-finetuned-cord-v1-2560",
]:
__a ='<s_cord>'
elif model_name == "naver-clova-ix/donut-base-finetuned-cord-v2":
__a ='s_cord-v2>'
elif model_name == "naver-clova-ix/donut-base-finetuned-zhtrainticket":
__a ='<s_zhtrainticket>'
elif model_name in ["naver-clova-ix/donut-proto", "naver-clova-ix/donut-base"]:
# use a random prompt
__a ='hello world'
else:
raise ValueError('Model name not supported' )
__a =original_model.decoder.tokenizer(_snake_case , add_special_tokens=_snake_case , return_tensors='pt' )[
'input_ids'
]
__a =original_model.encoder.model.patch_embed(_snake_case )
__a , __a =model.encoder.embeddings(_snake_case )
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
# verify encoder hidden states
__a =original_model.encoder(_snake_case )
__a =model.encoder(_snake_case ).last_hidden_state
assert torch.allclose(_snake_case , _snake_case , atol=1e-2 )
# verify decoder hidden states
__a =original_model(_snake_case , _snake_case , _snake_case ).logits
__a =model(_snake_case , decoder_input_ids=_snake_case ).logits
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
print('Looks ok!' )
if pytorch_dump_folder_path is not None:
print(F'Saving model and processor to {pytorch_dump_folder_path}' )
model.save_pretrained(_snake_case )
processor.save_pretrained(_snake_case )
if push_to_hub:
model.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
processor.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
if __name__ == "__main__":
_lowerCAmelCase : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_name",
default="naver-clova-ix/donut-base-finetuned-docvqa",
required=False,
type=str,
help="Name of the original model you'd like to convert.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
required=False,
type=str,
help="Path to the output PyTorch model directory.",
)
parser.add_argument(
"--push_to_hub",
action="store_true",
help="Whether or not to push the converted model and processor to the 🤗 hub.",
)
_lowerCAmelCase : List[Any] = parser.parse_args()
convert_donut_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 308
|
import argparse
import logging
from collections import namedtuple
import torch
from model_bertabs import BertAbsSummarizer
from models.model_builder import AbsSummarizer # The authors' implementation
from transformers import BertTokenizer
logging.basicConfig(level=logging.INFO)
_lowerCAmelCase : List[Any] = logging.getLogger(__name__)
_lowerCAmelCase : Optional[Any] = "Hello world! cécé herlolip"
_lowerCAmelCase : str = namedtuple(
"BertAbsConfig",
[
"temp_dir",
"large",
"use_bert_emb",
"finetune_bert",
"encoder",
"share_emb",
"max_pos",
"enc_layers",
"enc_hidden_size",
"enc_heads",
"enc_ff_size",
"enc_dropout",
"dec_layers",
"dec_hidden_size",
"dec_heads",
"dec_ff_size",
"dec_dropout",
],
)
def UpperCamelCase_( _snake_case : str , _snake_case : List[Any] ):
"""simple docstring"""
__a =BertAbsConfig(
temp_dir='.' , finetune_bert=_snake_case , large=_snake_case , share_emb=_snake_case , use_bert_emb=_snake_case , encoder='bert' , max_pos=512 , enc_layers=6 , enc_hidden_size=512 , enc_heads=8 , enc_ff_size=512 , enc_dropout=0.2 , dec_layers=6 , dec_hidden_size=768 , dec_heads=8 , dec_ff_size=2048 , dec_dropout=0.2 , )
__a =torch.load(_snake_case , lambda _snake_case , _snake_case : storage )
__a =AbsSummarizer(_snake_case , torch.device('cpu' ) , _snake_case )
original.eval()
__a =BertAbsSummarizer(_snake_case , torch.device('cpu' ) )
new_model.eval()
# -------------------
# Convert the weights
# -------------------
logging.info('convert the model' )
new_model.bert.load_state_dict(original.bert.state_dict() )
new_model.decoder.load_state_dict(original.decoder.state_dict() )
new_model.generator.load_state_dict(original.generator.state_dict() )
# ----------------------------------
# Make sure the outpus are identical
# ----------------------------------
logging.info('Make sure that the models\' outputs are identical' )
__a =BertTokenizer.from_pretrained('bert-base-uncased' )
# prepare the model inputs
__a =tokenizer.encode('This is sample éàalj\'-.' )
encoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
__a =tokenizer.encode('This is sample 3 éàalj\'-.' )
decoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
# failsafe to make sure the weights reset does not affect the
# loaded weights.
assert torch.max(torch.abs(original.generator[0].weight - new_model.generator[0].weight ) ) == 0
# forward pass
__a =encoder_input_ids
__a =decoder_input_ids
__a =__a =None
__a =None
__a =__a =None
__a =__a =None
__a =None
# The original model does not apply the geneator layer immediatly but rather in
# the beam search (where it combines softmax + linear layer). Since we already
# apply the softmax in our generation process we only apply the linear layer here.
# We make sure that the outputs of the full stack are identical
__a =original(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =original.generator(_snake_case )
__a =new_model(
_snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =new_model.generator(_snake_case )
__a =torch.max(torch.abs(output_converted_model - output_original_model ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.max(torch.abs(output_converted_generator - output_original_generator ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.allclose(_snake_case , _snake_case , atol=1e-3 )
if are_identical:
logging.info('all weights are equal up to 1e-3' )
else:
raise ValueError('the weights are different. The new model is likely different from the original one.' )
# The model has been saved with torch.save(model) and this is bound to the exact
# directory structure. We save the state_dict instead.
logging.info('saving the model\'s state dictionary' )
torch.save(
new_model.state_dict() , './bertabs-finetuned-cnndm-extractive-abstractive-summarization/pytorch_model.bin' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
parser.add_argument(
"--bertabs_checkpoint_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch dump.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
type=str,
required=True,
help="Path to the output PyTorch model.",
)
_lowerCAmelCase : Optional[Any] = parser.parse_args()
convert_bertabs_checkpoints(
args.bertabs_checkpoint_path,
args.pytorch_dump_folder_path,
)
| 308
| 1
|
from typing import TYPE_CHECKING
from ...utils import (
OptionalDependencyNotAvailable,
_LazyModule,
is_flax_available,
is_tf_available,
is_torch_available,
)
_lowerCAmelCase : List[Any] = {"configuration_vit_mae": ["VIT_MAE_PRETRAINED_CONFIG_ARCHIVE_MAP", "ViTMAEConfig"]}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Union[str, Any] = [
"VIT_MAE_PRETRAINED_MODEL_ARCHIVE_LIST",
"ViTMAEForPreTraining",
"ViTMAELayer",
"ViTMAEModel",
"ViTMAEPreTrainedModel",
]
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Union[str, Any] = [
"TFViTMAEForPreTraining",
"TFViTMAEModel",
"TFViTMAEPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_vit_mae import VIT_MAE_PRETRAINED_CONFIG_ARCHIVE_MAP, ViTMAEConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_vit_mae import (
VIT_MAE_PRETRAINED_MODEL_ARCHIVE_LIST,
ViTMAEForPreTraining,
ViTMAELayer,
ViTMAEModel,
ViTMAEPreTrainedModel,
)
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_tf_vit_mae import TFViTMAEForPreTraining, TFViTMAEModel, TFViTMAEPreTrainedModel
else:
import sys
_lowerCAmelCase : Optional[Any] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
|
import unittest
from transformers import XLMConfig, is_torch_available
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 (
XLMForMultipleChoice,
XLMForQuestionAnswering,
XLMForQuestionAnsweringSimple,
XLMForSequenceClassification,
XLMForTokenClassification,
XLMModel,
XLMWithLMHeadModel,
)
from transformers.models.xlm.modeling_xlm import XLM_PRETRAINED_MODEL_ARCHIVE_LIST
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=13 , __snake_case=7 , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=False , __snake_case=False , __snake_case=False , __snake_case=2 , __snake_case=99 , __snake_case=0 , __snake_case=32 , __snake_case=5 , __snake_case=4 , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=2 , __snake_case=0.02 , __snake_case=2 , __snake_case=4 , __snake_case="last" , __snake_case=True , __snake_case=None , __snake_case=0 , ) -> Optional[Any]:
'''simple docstring'''
__a =parent
__a =batch_size
__a =seq_length
__a =is_training
__a =use_input_lengths
__a =use_token_type_ids
__a =use_labels
__a =gelu_activation
__a =sinusoidal_embeddings
__a =causal
__a =asm
__a =n_langs
__a =vocab_size
__a =n_special
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_sequence_label_size
__a =initializer_range
__a =num_labels
__a =num_choices
__a =summary_type
__a =use_proj
__a =scope
__a =bos_token_id
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a =random_attention_mask([self.batch_size, self.seq_length] )
__a =None
if self.use_input_lengths:
__a =(
ids_tensor([self.batch_size] , vocab_size=2 ) + self.seq_length - 2
) # small variation of seq_length
__a =None
if self.use_token_type_ids:
__a =ids_tensor([self.batch_size, self.seq_length] , self.n_langs )
__a =None
__a =None
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a =ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a =ids_tensor([self.batch_size] , 2 ).float()
__a =ids_tensor([self.batch_size] , self.num_choices )
__a =self.get_config()
return (
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
)
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return XLMConfig(
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 , num_labels=self.num_labels , bos_token_id=self.bos_token_id , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMModel(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , lengths=__snake_case , langs=__snake_case )
__a =model(__snake_case , langs=__snake_case )
__a =model(__snake_case )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[int]:
'''simple docstring'''
__a =XLMWithLMHeadModel(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , token_type_ids=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Dict:
'''simple docstring'''
__a =XLMForQuestionAnsweringSimple(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
__a =outputs
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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMForQuestionAnswering(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , p_mask=__snake_case , )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , )
((__a) , ) =result_with_labels.to_tuple()
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
((__a) , ) =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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[Any]:
'''simple docstring'''
__a =XLMForSequenceClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Any:
'''simple docstring'''
__a =self.num_labels
__a =XLMForTokenClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , attention_mask=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Tuple:
'''simple docstring'''
__a =self.num_choices
__a =XLMForMultipleChoice(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =model(
__snake_case , attention_mask=__snake_case , token_type_ids=__snake_case , labels=__snake_case , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) =config_and_inputs
__a ={'input_ids': input_ids, 'token_type_ids': token_type_ids, 'lengths': input_lengths}
return config, inputs_dict
@require_torch
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ , lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
XLMModel,
XLMWithLMHeadModel,
XLMForQuestionAnswering,
XLMForSequenceClassification,
XLMForQuestionAnsweringSimple,
XLMForTokenClassification,
XLMForMultipleChoice,
)
if is_torch_available()
else ()
)
SCREAMING_SNAKE_CASE = (
(XLMWithLMHeadModel,) if is_torch_available() else ()
) # TODO (PVP): Check other models whether language generation is also applicable
SCREAMING_SNAKE_CASE = (
{
'feature-extraction': XLMModel,
'fill-mask': XLMWithLMHeadModel,
'question-answering': XLMForQuestionAnsweringSimple,
'text-classification': XLMForSequenceClassification,
'text-generation': XLMWithLMHeadModel,
'token-classification': XLMForTokenClassification,
'zero-shot': XLMForSequenceClassification,
}
if is_torch_available()
else {}
)
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> int:
'''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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case=False ) -> str:
'''simple docstring'''
__a =super()._prepare_for_class(__snake_case , __snake_case , return_labels=__snake_case )
if return_labels:
if model_class.__name__ == "XLMForQuestionAnswering":
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
return inputs_dict
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMModelTester(self )
__a =ConfigTester(self , config_class=__snake_case , emb_dim=37 )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.config_tester.run_common_tests()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_model(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_lm_head(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_simple_qa(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_qa(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_sequence_classif(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_token_classif(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_for_multiple_choice(*__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Optional[Any]:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_attentions in attentions] , [True] * len(__snake_case ) )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_attentions in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =min_length + idx + 1
__a =(
batch_size * num_beam_groups,
config.num_attention_heads,
tgt_len,
src_len,
)
# check attn size
self.assertListEqual(
[layer_attention.shape for layer_attention in iter_attentions] , [expected_shape] * len(__snake_case ) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Dict:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_hidden_states in hidden_states] , [True] * len(__snake_case ) , )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_hidden_states in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =(batch_size * num_beam_groups, seq_len, config.hidden_size)
# check hidden size
self.assertListEqual(
[layer_hidden_states.shape for layer_hidden_states in iter_hidden_states] , [expected_shape] * len(__snake_case ) , )
pass
@slow
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
for model_name in XLM_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =XLMModel.from_pretrained(__snake_case )
self.assertIsNotNone(__snake_case )
@require_torch
class __magic_name__ ( unittest.TestCase ):
@slow
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMWithLMHeadModel.from_pretrained('xlm-mlm-en-2048' )
model.to(__snake_case )
__a =torch.tensor([[14, 447]] , dtype=torch.long , device=__snake_case ) # the president
__a =[
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
] # the president the president the president the president the president the president the president the president the president the president
# TODO(PVP): this and other input_ids I tried for generation give pretty bad results. Not sure why. Model might just not be made for auto-regressive inference
__a =model.generate(__snake_case , do_sample=__snake_case )
self.assertListEqual(output_ids[0].cpu().numpy().tolist() , __snake_case )
| 308
| 1
|
import tempfile
import unittest
from transformers import SPIECE_UNDERLINE, BatchEncoding, PLBartTokenizer, 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
_lowerCAmelCase : Optional[Any] = get_tests_dir("fixtures/test_sentencepiece.model")
if is_torch_available():
from transformers.models.plbart.modeling_plbart import shift_tokens_right
_lowerCAmelCase : Optional[Any] = 50_003
_lowerCAmelCase : List[str] = 50_002
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = PLBartTokenizer
SCREAMING_SNAKE_CASE = None
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
super().setUp()
# We have a SentencePiece fixture for testing
__a =PLBartTokenizer(__snake_case , language_codes='base' , keep_accents=__snake_case )
tokenizer.save_pretrained(self.tmpdirname )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =PLBartTokenizer(__snake_case , language_codes='base' , keep_accents=__snake_case )
__a =tokenizer.tokenize('This is a test' )
self.assertListEqual(__snake_case , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(__snake_case ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
__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 =tokenizer.convert_tokens_to_ids(__snake_case )
self.assertListEqual(
__snake_case , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
__a =tokenizer.convert_ids_to_tokens(__snake_case )
self.assertListEqual(
__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>',
'.',
] , )
__a =tokenizer.vocab_size
__a =[tokenizer.convert_ids_to_tokens(__snake_case ) for x in range(end - 4 , __snake_case )]
self.assertListEqual(__snake_case , ['__java__', '__python__', '__en_XX__', '<mask>'] )
__a ='java.lang.Exception, python.lang.Exception, javascript, php, ruby, go'
__a =tokenizer(__snake_case ).input_ids
self.assertEqual(
tokenizer.decode(__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case ) , __snake_case , )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =PLBartTokenizer(__snake_case , language_codes='multi' , keep_accents=__snake_case )
__a =tokenizer.tokenize('This is a test' )
self.assertListEqual(__snake_case , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(__snake_case ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
__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 =tokenizer.convert_tokens_to_ids(__snake_case )
self.assertListEqual(
__snake_case , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
__a =tokenizer.convert_ids_to_tokens(__snake_case )
self.assertListEqual(
__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>',
'.',
] , )
__a =tokenizer.vocab_size
__a =[tokenizer.convert_ids_to_tokens(__snake_case ) for x in range(end - 7 , __snake_case )]
self.assertListEqual(
__snake_case , ['__java__', '__python__', '__en_XX__', '__javascript__', '__php__', '__ruby__', '__go__'] )
__a ='java.lang.Exception, python.lang.Exception, javascript, php, ruby, go'
__a =tokenizer(__snake_case ).input_ids
self.assertEqual(
tokenizer.decode(__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case ) , __snake_case , )
@require_torch
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = 'uclanlp/plbart-python-en_XX'
SCREAMING_SNAKE_CASE = [
'def maximum(a,b,c):NEW_LINE_INDENTreturn max([a,b,c])',
'def sum(a,b,c):NEW_LINE_INDENTreturn sum([a,b,c])',
]
SCREAMING_SNAKE_CASE = [
'Returns the maximum value of a b c.',
'Sums the values of a b c.',
]
SCREAMING_SNAKE_CASE = [
1_3_4,
5_4_5_2,
3_3_4_6_0,
3_3_4_4_1,
3_3_4_6_3,
3_3_4_6_5,
3_3_4_6_3,
3_3_4_4_9,
9_8_8,
2_0,
3_3_4_5_6,
1_9,
3_3_4_5_6,
7_7_1,
3_9,
4_2_5_8,
8_8_9,
3_3_1_8,
3_3_4_4_1,
3_3_4_6_3,
3_3_4_6_5,
3_3_4_6_3,
3_3_4_4_9,
2_4_7_1,
2,
PYTHON_CODE,
]
@classmethod
def __magic_name__ ( cls ) -> Dict:
'''simple docstring'''
__a =PLBartTokenizer.from_pretrained(
cls.checkpoint_name , language_codes='base' , src_lang='python' , tgt_lang='en_XX' )
__a =1
return cls
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['__java__'] , 5_0001 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['__python__'] , 5_0002 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['__en_XX__'] , 5_0003 )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.tokenizer.batch_encode_plus(self.src_text ).input_ids[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
self.assertIn(__snake_case , self.tokenizer.all_special_ids )
__a =[EN_CODE, 9037, 3_3442, 57, 752, 153, 14, 56, 18, 9, 2]
__a =self.tokenizer.decode(__snake_case , skip_special_tokens=__snake_case )
__a =self.tokenizer.decode(generated_ids[1:] , skip_special_tokens=__snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertNotIn(self.tokenizer.eos_token , __snake_case )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =['def sum(a,b,c):NEW_LINE_INDENTreturn sum([a,b,c])' * 20]
self.assertIsInstance(src_text[0] , __snake_case )
__a =10
__a =self.tokenizer(__snake_case , max_length=__snake_case , truncation=__snake_case ).input_ids[0]
self.assertEqual(ids[-2] , 2 )
self.assertEqual(ids[-1] , __snake_case )
self.assertEqual(len(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.assertListEqual(self.tokenizer.convert_tokens_to_ids(['<mask>', '__java__'] ) , [5_0004, 5_0001] )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =tempfile.mkdtemp()
__a =self.tokenizer.fairseq_tokens_to_ids
self.tokenizer.save_pretrained(__snake_case )
__a =PLBartTokenizer.from_pretrained(__snake_case )
self.assertDictEqual(new_tok.fairseq_tokens_to_ids , __snake_case )
@require_torch
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer(self.src_text , text_target=self.tgt_text , padding=__snake_case , return_tensors='pt' )
__a =shift_tokens_right(batch['labels'] , self.tokenizer.pad_token_id )
# fairseq batch: https://gist.github.com/sshleifer/cba08bc2109361a74ac3760a7e30e4f4
self.assertEqual(batch.input_ids[1][-2:].tolist() , [2, PYTHON_CODE] )
self.assertEqual(batch.decoder_input_ids[1][0] , __snake_case )
self.assertEqual(batch.decoder_input_ids[1][-1] , 2 )
self.assertEqual(batch.labels[1][-2:].tolist() , [2, EN_CODE] )
@require_torch
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.tokenizer(
self.src_text , text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=len(self.expected_src_tokens ) , return_tensors='pt' , )
__a =shift_tokens_right(batch['labels'] , self.tokenizer.pad_token_id )
self.assertIsInstance(__snake_case , __snake_case )
self.assertEqual((2, 26) , batch.input_ids.shape )
self.assertEqual((2, 26) , batch.attention_mask.shape )
__a =batch.input_ids.tolist()[0]
self.assertListEqual(self.expected_src_tokens , __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, PYTHON_CODE] )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.tokenizer(self.src_text , padding=__snake_case , truncation=__snake_case , max_length=3 , return_tensors='pt' )
__a =self.tokenizer(
text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=10 , return_tensors='pt' )
__a =targets['input_ids']
__a =shift_tokens_right(__snake_case , self.tokenizer.pad_token_id )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.decoder_input_ids.shape[1] , 10 )
@require_torch
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.tokenizer._build_translation_inputs(
'A test' , return_tensors='pt' , src_lang='en_XX' , tgt_lang='java' )
self.assertEqual(
nested_simplify(__snake_case ) , {
# A, test, EOS, en_XX
'input_ids': [[150, 242, 2, 5_0003]],
'attention_mask': [[1, 1, 1, 1]],
# java
'forced_bos_token_id': 5_0001,
} , )
| 308
|
from __future__ import annotations
import matplotlib.pyplot as plt # type: ignore
import numpy
# initial triangle of Koch snowflake
_lowerCAmelCase : Optional[Any] = numpy.array([0, 0])
_lowerCAmelCase : Dict = numpy.array([0.5, 0.8660254])
_lowerCAmelCase : Any = numpy.array([1, 0])
_lowerCAmelCase : int = [VECTOR_1, VECTOR_2, VECTOR_3, VECTOR_1]
def UpperCamelCase_( _snake_case : list[numpy.ndarray] , _snake_case : int ):
"""simple docstring"""
__a =initial_vectors
for _ in range(_snake_case ):
__a =iteration_step(_snake_case )
return vectors
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =[]
for i, start_vector in enumerate(vectors[:-1] ):
__a =vectors[i + 1]
new_vectors.append(_snake_case )
__a =end_vector - start_vector
new_vectors.append(start_vector + difference_vector / 3 )
new_vectors.append(
start_vector + difference_vector / 3 + rotate(difference_vector / 3 , 60 ) )
new_vectors.append(start_vector + difference_vector * 2 / 3 )
new_vectors.append(vectors[-1] )
return new_vectors
def UpperCamelCase_( _snake_case : numpy.ndarray , _snake_case : float ):
"""simple docstring"""
__a =numpy.radians(_snake_case )
__a , __a =numpy.cos(_snake_case ), numpy.sin(_snake_case )
__a =numpy.array(((c, -s), (s, c)) )
return numpy.dot(_snake_case , _snake_case )
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =plt.gca()
axes.set_aspect('equal' )
# matplotlib.pyplot.plot takes a list of all x-coordinates and a list of all
# y-coordinates as inputs, which are constructed from the vector-list using
# zip()
__a , __a =zip(*_snake_case )
plt.plot(_snake_case , _snake_case )
plt.show()
if __name__ == "__main__":
import doctest
doctest.testmod()
_lowerCAmelCase : List[Any] = iterate(INITIAL_VECTORS, 5)
plot(processed_vectors)
| 308
| 1
|
def SCREAMING_SNAKE_CASE_( _snake_case : str = "The quick brown fox jumps over the lazy dog" , ):
"""simple docstring"""
__a =set()
# Replace all the whitespace in our sentence
__a =input_str.replace(' ' , '' )
for alpha in input_str:
if "a" <= alpha.lower() <= "z":
frequency.add(alpha.lower() )
return len(_UpperCAmelCase ) == 26
def SCREAMING_SNAKE_CASE_( _snake_case : str = "The quick brown fox jumps over the lazy dog" , ):
"""simple docstring"""
__a =[False] * 26
for char in input_str:
if char.islower():
__a =True
elif char.isupper():
__a =True
return all(_UpperCAmelCase )
def SCREAMING_SNAKE_CASE_( _snake_case : str = "The quick brown fox jumps over the lazy dog" , ):
"""simple docstring"""
return len({char for char in input_str.lower() if char.isalpha()} ) == 26
def SCREAMING_SNAKE_CASE_( ):
"""simple docstring"""
from timeit import timeit
__a ='from __main__ import is_pangram, is_pangram_faster, is_pangram_fastest'
print(timeit('is_pangram()' , setup=_UpperCAmelCase ) )
print(timeit('is_pangram_faster()' , setup=_UpperCAmelCase ) )
print(timeit('is_pangram_fastest()' , setup=_UpperCAmelCase ) )
# 5.348480500048026, 2.6477354579837993, 1.8470395830227062
# 5.036091582966037, 2.644472333951853, 1.8869528750656173
if __name__ == "__main__":
import doctest
doctest.testmod()
benchmark()
| 350
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
_lowerCAmelCase : Tuple = {
"configuration_megatron_bert": ["MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MegatronBertConfig"],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : List[str] = [
"MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"MegatronBertForCausalLM",
"MegatronBertForMaskedLM",
"MegatronBertForMultipleChoice",
"MegatronBertForNextSentencePrediction",
"MegatronBertForPreTraining",
"MegatronBertForQuestionAnswering",
"MegatronBertForSequenceClassification",
"MegatronBertForTokenClassification",
"MegatronBertModel",
"MegatronBertPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_megatron_bert import MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MegatronBertConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_megatron_bert import (
MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
MegatronBertForCausalLM,
MegatronBertForMaskedLM,
MegatronBertForMultipleChoice,
MegatronBertForNextSentencePrediction,
MegatronBertForPreTraining,
MegatronBertForQuestionAnswering,
MegatronBertForSequenceClassification,
MegatronBertForTokenClassification,
MegatronBertModel,
MegatronBertPreTrainedModel,
)
else:
import sys
_lowerCAmelCase : Optional[int] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 0
|
import os
from typing import Any, Callable, Dict, List, Optional, Tuple, Union
import torch
from torch import nn
from ...models.controlnet import ControlNetModel, ControlNetOutput
from ...models.modeling_utils import ModelMixin
from ...utils import logging
_lowerCAmelCase : Optional[Any] = logging.get_logger(__name__)
class __magic_name__ ( A__ ):
def __init__( self , __snake_case ) -> Tuple:
'''simple docstring'''
super().__init__()
__a =nn.ModuleList(__A )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = False , __snake_case = True , ) -> Union[ControlNetOutput, Tuple]:
'''simple docstring'''
for i, (image, scale, controlnet) in enumerate(zip(__A , __A , self.nets ) ):
__a =controlnet(
__A , __A , __A , __A , __A , __A , __A , __A , __A , __A , __A , )
# merge samples
if i == 0:
__a =down_samples, mid_sample
else:
__a =[
samples_prev + samples_curr
for samples_prev, samples_curr in zip(__A , __A )
]
mid_block_res_sample += mid_sample
return down_block_res_samples, mid_block_res_sample
def __magic_name__ ( self , __snake_case , __snake_case = True , __snake_case = None , __snake_case = False , __snake_case = None , ) -> Dict:
'''simple docstring'''
__a =0
__a =save_directory
for controlnet in self.nets:
controlnet.save_pretrained(
__A , is_main_process=__A , save_function=__A , safe_serialization=__A , variant=__A , )
idx += 1
__a =model_path_to_save + f'_{idx}'
@classmethod
def __magic_name__ ( cls , __snake_case , **__snake_case ) -> int:
'''simple docstring'''
__a =0
__a =[]
# load controlnet and append to list until no controlnet directory exists anymore
# first controlnet has to be saved under `./mydirectory/controlnet` to be compliant with `DiffusionPipeline.from_prertained`
# second, third, ... controlnets have to be saved under `./mydirectory/controlnet_1`, `./mydirectory/controlnet_2`, ...
__a =pretrained_model_path
while os.path.isdir(__A ):
__a =ControlNetModel.from_pretrained(__A , **__A )
controlnets.append(__A )
idx += 1
__a =pretrained_model_path + f'_{idx}'
logger.info(f'{len(__A )} controlnets loaded from {pretrained_model_path}.' )
if len(__A ) == 0:
raise ValueError(
f'No ControlNets found under {os.path.dirname(__A )}. Expected at least {pretrained_model_path + "_0"}.' )
return cls(__A )
| 351
|
import hashlib
import unittest
from transformers import MODEL_FOR_DEPTH_ESTIMATION_MAPPING, is_torch_available, is_vision_available
from transformers.pipelines import DepthEstimationPipeline, pipeline
from transformers.testing_utils import (
is_pipeline_test,
nested_simplify,
require_tf,
require_timm,
require_torch,
require_vision,
slow,
)
from .test_pipelines_common import ANY
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
else:
class __magic_name__ :
@staticmethod
def __magic_name__ ( *__snake_case , **__snake_case ) -> List[str]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : Image ):
"""simple docstring"""
__a =hashlib.mda(image.tobytes() )
return m.hexdigest()
@is_pipeline_test
@require_vision
@require_timm
@require_torch
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = MODEL_FOR_DEPTH_ESTIMATION_MAPPING
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Dict:
'''simple docstring'''
__a =DepthEstimationPipeline(model=__snake_case , image_processor=__snake_case )
return depth_estimator, [
"./tests/fixtures/tests_samples/COCO/000000039769.png",
"./tests/fixtures/tests_samples/COCO/000000039769.png",
]
def __magic_name__ ( self , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
__a =depth_estimator('./tests/fixtures/tests_samples/COCO/000000039769.png' )
self.assertEqual({'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )} , __snake_case )
import datasets
__a =datasets.load_dataset('hf-internal-testing/fixtures_image_utils' , 'image' , split='test' )
__a =depth_estimator(
[
Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' ),
'http://images.cocodataset.org/val2017/000000039769.jpg',
# RGBA
dataset[0]['file'],
# LA
dataset[1]['file'],
# L
dataset[2]['file'],
] )
self.assertEqual(
[
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
] , __snake_case , )
@require_tf
@unittest.skip('Depth estimation is not implemented in TF' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
pass
@slow
@require_torch
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a ='Intel/dpt-large'
__a =pipeline('depth-estimation' , model=__snake_case )
__a =depth_estimator('http://images.cocodataset.org/val2017/000000039769.jpg' )
__a =hashimage(outputs['depth'] )
# This seems flaky.
# self.assertEqual(outputs["depth"], "1a39394e282e9f3b0741a90b9f108977")
self.assertEqual(nested_simplify(outputs['predicted_depth'].max().item() ) , 29.304 )
self.assertEqual(nested_simplify(outputs['predicted_depth'].min().item() ) , 2.662 )
@require_torch
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
# This is highly irregular to have no small tests.
self.skipTest('There is not hf-internal-testing tiny model for either GLPN nor DPT' )
| 308
| 0
|
import math
from datetime import datetime, timedelta
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
__a =year % 19
__a =year % 4
__a =year % 7
__a =math.floor(year / 100 )
__a =math.floor((13 + 8 * leap_day_inhibits) / 25 )
__a =leap_day_inhibits / 4
__a =(
15 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 30
__a =(4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
__a =(19 * metonic_cycle + secular_moon_shift) % 30
# PHM -> Paschal Full Moon
__a =(
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 29 and days_from_phm_to_sunday == 6:
return datetime(__UpperCAmelCase , 4 , 19 )
elif days_to_add == 28 and days_from_phm_to_sunday == 6:
return datetime(__UpperCAmelCase , 4 , 18 )
else:
return datetime(__UpperCAmelCase , 3 , 22 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1_994, 2_000, 2_010, 2_021, 2_023):
_lowerCAmelCase : Optional[Any] = "will be" if year > datetime.now().year else "was"
print(f'''Easter in {year} {tense} {gauss_easter(year)}''')
| 352
|
import argparse
import dataclasses
import json
import logging
import os
import shutil
from typing import List, Optional
import datasets
from accelerate import Accelerator
from datasets import load_dataset
from finetuning import finetune
from tqdm.auto import tqdm
import transformers
from transformers import AutoConfig, set_seed
from transformers.trainer_utils import IntervalStrategy
_lowerCAmelCase : Optional[int] = logging.getLogger(__name__)
_lowerCAmelCase : Any = "pytorch_model.bin"
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co.'} , )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the data to predict on.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The name of the task to train on.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The list of labels for the task.'} )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'The output directory where the model predictions and checkpoints will be written.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='accuracy' , metadata={'help': 'The evaluation metric used for the task.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='no' , metadata={
'help': 'The evaluation strategy to adopt during training. Possible values are: ["no", "step", "epoch]'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={
'help': 'How much the specified evaluation metric must improve to satisfy early stopping conditions.'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the confidence score.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the validation performance.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to fine-tune on labeled data after pseudo training.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={'help': 'Confidence threshold for pseudo-labeled data filtering.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Random seed for initialization.'} , )
def UpperCamelCase_( _snake_case : int , _snake_case : str , _snake_case : Optional[int] , _snake_case : str , _snake_case : Union[str, Any] , _snake_case : List[Any] ):
"""simple docstring"""
__a =datasets.concatenate_datasets([infer_input, infer_output] , axis=1 )
if args.do_filter_by_confidence:
__a =dataset.filter(lambda _snake_case : example["probability"] > args.confidence_threshold )
if args.do_filter_by_val_performance:
assert eval_result >= 0.0 and eval_result <= 1.0
__a =int(eval_result * len(_snake_case ) )
print(_snake_case )
__a =dataset.sort('probability' , reverse=_snake_case )
__a =dataset.select(range(_snake_case ) )
__a =dataset.remove_columns(['label', 'probability'] )
__a =dataset.rename_column('prediction' , 'label' )
__a =dataset.map(lambda _snake_case : {"label": idalabel[example["label"]]} )
__a =dataset.shuffle(seed=args.seed )
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.data_file_extension == "csv":
dataset.to_csv(_snake_case , index=_snake_case )
else:
dataset.to_json(_snake_case )
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : str , _snake_case : int , _snake_case : Optional[int] , **_snake_case : List[str] ):
"""simple docstring"""
__a =Accelerator()
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , level=logging.INFO , )
logger.info(accelerator.state )
# Setup logging, we only want one process per machine to log things on the
# screen. accelerator.is_local_main_process is only True for one process per
# machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR )
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
__a =STModelArguments(model_name_or_path=_snake_case )
__a =STDataArguments(train_file=_snake_case , infer_file=_snake_case )
__a =STTrainingArguments(output_dir=_snake_case )
__a =argparse.Namespace()
for arg_class in (model_args, data_args, training_args):
for key, value in vars(_snake_case ).items():
setattr(_snake_case , _snake_case , _snake_case )
for key, value in kwargs.items():
if hasattr(_snake_case , _snake_case ):
setattr(_snake_case , _snake_case , _snake_case )
# Sanity checks
__a ={}
__a =None
# You need to provide the training data and the data to predict on
assert args.train_file is not None
assert args.infer_file is not None
__a =args.train_file
__a =args.infer_file
if args.evaluation_strategy != IntervalStrategy.NO.value:
assert args.eval_file is not None
__a =args.eval_file
for key in data_files:
__a =data_files[key].split('.' )[-1]
assert extension in ["csv", "json"], F'`{key}_file` should be a csv or a json file.'
if args.data_file_extension is None:
__a =extension
else:
assert extension == args.data_file_extension, F'`{key}_file` should be a {args.data_file_extension} file`.'
assert (
args.eval_metric in datasets.list_metrics()
), F'{args.eval_metric} not in the list of supported metrics {datasets.list_metrics()}.'
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed )
logger.info('Creating the initial data directory for self-training...' )
__a =F'{args.output_dir}/self-train_iter-{{}}'.format
__a =data_dir_format(0 )
if accelerator.is_main_process:
if args.output_dir is not None:
os.makedirs(args.output_dir , exist_ok=_snake_case )
os.makedirs(_snake_case , exist_ok=_snake_case )
accelerator.wait_for_everyone()
__a =None
__a =None
__a =0
__a =False
# Show the progress bar
__a =tqdm(range(args.max_selftrain_iterations ) , disable=not accelerator.is_local_main_process )
# Self-train
for iteration in range(0 , int(args.max_selftrain_iterations ) ):
__a =data_dir_format(_snake_case )
assert os.path.exists(_snake_case )
# Stage 1: initial fine-tuning for iteration = 0 or pseudo-training for
# iteration > 0
__a =os.path.join(_snake_case , 'stage-1' )
__a ={
'accelerator': accelerator,
'model_name_or_path': args.model_name_or_path,
'cache_dir': args.cache_dir,
'do_train': True,
'train_file': data_files['train'] if iteration == 0 else data_files['train_pseudo'],
'do_eval': True if args.eval_file is not None else False,
'eval_file': data_files['eval'],
'do_predict': True,
'infer_file': data_files['infer'],
'task_name': args.task_name,
'label_list': args.label_list,
'output_dir': current_output_dir,
'eval_metric': args.eval_metric,
'evaluation_strategy': args.evaluation_strategy,
'early_stopping_patience': args.early_stopping_patience,
'early_stopping_threshold': args.early_stopping_threshold,
'seed': args.seed,
}
# Add additional training arguments
for key, value in kwargs.items():
if key not in arguments_dict and not hasattr(_snake_case , _snake_case ):
arguments_dict.update({key: value} )
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 1.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 1 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 1.' , _snake_case )
if iteration > 0 and args.finetune_on_labeled_data:
# Stage 2 (optional): fine-tuning on the original labeled data
__a =os.path.join(_snake_case , 'best-checkpoint' )
__a =os.path.join(_snake_case , 'stage-2' )
# Update arguments_dict
__a =model_path
__a =data_files['train']
__a =current_output_dir
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 2.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 2 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 2.' , _snake_case )
__a =iteration
__a =data_dir_format(iteration + 1 )
__a =AutoConfig.from_pretrained(os.path.join(_snake_case , 'best-checkpoint' ) )
__a =config.idalabel
__a =os.path.join(_snake_case , 'eval_results_best-checkpoint.json' )
__a =os.path.join(_snake_case , 'test_results_best-checkpoint.json' )
assert os.path.exists(_snake_case )
with open(_snake_case , 'r' ) as f:
__a =float(json.load(_snake_case )[args.eval_metric] )
__a =os.path.join(_snake_case , 'infer_output_best-checkpoint.csv' )
assert os.path.exists(_snake_case )
# Loading the dataset from local csv or json files.
__a =load_dataset(args.data_file_extension , data_files={'data': data_files['infer']} )['data']
__a =load_dataset('csv' , data_files={'data': infer_output_file} )['data']
if accelerator.is_main_process:
os.makedirs(_snake_case , exist_ok=_snake_case )
shutil.copy(_snake_case , os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) )
if os.path.exists(_snake_case ):
shutil.copy(_snake_case , os.path.join(_snake_case , F'test_results_iter-{iteration}.json' ) )
create_pseudo_labeled_data(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )
accelerator.wait_for_everyone()
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.evaluation_strategy != IntervalStrategy.NO.value:
__a =eval_result
if best_iteration is None:
__a =new_iteration
__a =new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
__a =new_iteration
__a =new_eval_result
__a =0
else:
if new_eval_result == best_eval_result:
__a =new_iteration
__a =new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
__a =True
progress_bar.update(1 )
if should_training_stop:
break
if best_iteration is not None:
# Save the best iteration
logger.info('Best iteration: %d' , _snake_case )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
else:
# Assume that the last iteration is the best
logger.info('Best iteration: %d' , args.max_selftrain_iterations - 1 )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{args.max_selftrain_iterations - 1}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
| 308
| 0
|
from __future__ import annotations
import typing
from collections import Counter
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
__a =Counter()
for base in range(1 , max_perimeter + 1 ):
for perpendicular in range(__lowerCAmelCase , max_perimeter + 1 ):
__a =(base * base + perpendicular * perpendicular) ** 0.5
if hypotenuse == int(__lowerCAmelCase ):
__a =int(base + perpendicular + hypotenuse )
if perimeter > max_perimeter:
continue
triplets[perimeter] += 1
return triplets
def UpperCamelCase_( _snake_case : int = 1000 ):
"""simple docstring"""
__a =pythagorean_triple(__lowerCAmelCase )
return triplets.most_common(1 )[0][0]
if __name__ == "__main__":
print(f'''Perimeter {solution()} has maximum solutions''')
| 353
|
import shutil
import tempfile
import unittest
from transformers import (
SPIECE_UNDERLINE,
AddedToken,
BatchEncoding,
NllbTokenizer,
NllbTokenizerFast,
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
_lowerCAmelCase : Any = get_tests_dir("fixtures/test_sentencepiece.model")
if is_torch_available():
from transformers.models.mam_aaa.modeling_mam_aaa import shift_tokens_right
_lowerCAmelCase : List[Any] = 256_047
_lowerCAmelCase : Dict = 256_145
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = NllbTokenizer
SCREAMING_SNAKE_CASE = NllbTokenizerFast
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = {}
def __magic_name__ ( self ) -> int:
'''simple docstring'''
super().setUp()
# We have a SentencePiece fixture for testing
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
tokenizer.save_pretrained(self.tmpdirname )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
__a =tokenizer.tokenize('This is a test' )
self.assertListEqual(__snake_case , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(__snake_case ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
__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 =tokenizer.convert_tokens_to_ids(__snake_case )
self.assertListEqual(
__snake_case , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
__a =tokenizer.convert_ids_to_tokens(__snake_case )
self.assertListEqual(
__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 __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =(self.rust_tokenizer_class, 'hf-internal-testing/tiny-random-nllb', {})
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =self.rust_tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =self.tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tuple(f for f in tokenizer_r_files if 'tokenizer.json' not in f )
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=True
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__snake_case )
# Checks it save with the same files
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=False
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
if not self.test_seqaseq:
return
__a =self.get_tokenizers()
for tokenizer in tokenizers:
with self.subTest(f'{tokenizer.__class__.__name__}' ):
# Longer text that will definitely require truncation.
__a =[
' 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.',
]
__a =[
'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.',
]
try:
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , tgt_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='ron_Latn' , )
except NotImplementedError:
return
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 10 )
# max_target_length will default to max_length if not specified
__a =tokenizer.prepare_seqaseq_batch(
__snake_case , tgt_texts=__snake_case , max_length=3 , return_tensors='pt' )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 3 )
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' )
self.assertEqual(batch_encoder_only.input_ids.shape[1] , 3 )
self.assertEqual(batch_encoder_only.attention_mask.shape[1] , 3 )
self.assertNotIn('decoder_input_ids' , __snake_case )
@unittest.skip('Unfortunately way too slow to build a BPE with SentencePiece.' )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =[AddedToken('<special>' , lstrip=__snake_case )]
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_r.encode('Hey this is a <special> token' )
__a =tokenizer_r.encode('<special>' , add_special_tokens=__snake_case )[0]
self.assertTrue(special_token_id in r_output )
if self.test_slow_tokenizer:
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case , )
__a =self.tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_p.encode('Hey this is a <special> token' )
__a =tokenizer_cr.encode('Hey this is a <special> token' )
self.assertEqual(__snake_case , __snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertTrue(special_token_id in p_output )
self.assertTrue(special_token_id in cr_output )
@require_torch
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = 'facebook/nllb-200-distilled-600M'
SCREAMING_SNAKE_CASE = [
' 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 = [
'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 = [
2_5_6_0_4_7,
1_6_2_9_7,
1_3_4_4_0_8,
8_1_6_5,
2_4_8_0_6_6,
1_4_7_3_4,
9_5_0,
1_1_3_5,
1_0_5_7_2_1,
3_5_7_3,
8_3,
2_7_3_5_2,
1_0_8,
4_9_4_8_6,
2,
]
@classmethod
def __magic_name__ ( cls ) -> Tuple:
'''simple docstring'''
__a =NllbTokenizer.from_pretrained(
cls.checkpoint_name , src_lang='eng_Latn' , tgt_lang='ron_Latn' )
__a =1
return cls
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Arab'] , 25_6001 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Latn'] , 25_6002 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['fra_Latn'] , 25_6057 )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer.batch_encode_plus(self.src_text ).input_ids[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
self.assertIn(__snake_case , self.tokenizer.all_special_ids )
# fmt: off
__a =[RO_CODE, 4254, 9_8068, 11_2923, 3_9072, 3909, 713, 10_2767, 26, 1_7314, 3_5642, 1_4683, 3_3118, 2022, 6_6987, 2, 25_6047]
# fmt: on
__a =self.tokenizer.decode(__snake_case , skip_special_tokens=__snake_case )
__a =self.tokenizer.decode(generated_ids[1:] , skip_special_tokens=__snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertNotIn(self.tokenizer.eos_token , __snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =['this is gunna be a long sentence ' * 20]
assert isinstance(src_text[0] , __snake_case )
__a =10
__a =self.tokenizer(__snake_case , max_length=__snake_case , truncation=__snake_case ).input_ids[0]
self.assertEqual(ids[-1] , 2 )
self.assertEqual(ids[0] , __snake_case )
self.assertEqual(len(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
self.assertListEqual(self.tokenizer.convert_tokens_to_ids(['<mask>', 'ar_AR'] ) , [25_6203, 3] )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =tempfile.mkdtemp()
__a =self.tokenizer.fairseq_tokens_to_ids
self.tokenizer.save_pretrained(__snake_case )
__a =NllbTokenizer.from_pretrained(__snake_case )
self.assertDictEqual(new_tok.fairseq_tokens_to_ids , __snake_case )
@require_torch
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.tokenizer(
self.src_text , text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=len(self.expected_src_tokens ) , return_tensors='pt' , )
__a =shift_tokens_right(
batch['labels'] , self.tokenizer.pad_token_id , self.tokenizer.lang_code_to_id['ron_Latn'] )
self.assertIsInstance(__snake_case , __snake_case )
self.assertEqual((2, 15) , batch.input_ids.shape )
self.assertEqual((2, 15) , batch.attention_mask.shape )
__a =batch.input_ids.tolist()[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
self.assertEqual(__snake_case , batch.decoder_input_ids[0, 0] ) # EOS
# Test that special tokens are reset
self.assertEqual(self.tokenizer.prefix_tokens , [EN_CODE] )
self.assertEqual(self.tokenizer.suffix_tokens , [self.tokenizer.eos_token_id] )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =self.tokenizer(self.src_text , padding=__snake_case , truncation=__snake_case , max_length=3 , return_tensors='pt' )
__a =self.tokenizer(
text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=10 , return_tensors='pt' )
__a =targets['input_ids']
__a =shift_tokens_right(
__snake_case , self.tokenizer.pad_token_id , decoder_start_token_id=self.tokenizer.lang_code_to_id[self.tokenizer.tgt_lang] , )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.decoder_input_ids.shape[1] , 10 )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.tokenizer._build_translation_inputs(
'A test' , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
nested_simplify(__snake_case ) , {
# A, test, EOS, en_XX
'input_ids': [[25_6047, 70, 7356, 2]],
'attention_mask': [[1, 1, 1, 1]],
# ar_AR
'forced_bos_token_id': 25_6057,
} , )
@require_torch
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =True
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2, 25_6047] )
__a =False
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [25_6047, 1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2] )
| 308
| 0
|
def UpperCamelCase_( _snake_case : list ):
"""simple docstring"""
if len(lowerCamelCase__ ) <= 1:
return [tuple(lowerCamelCase__ )]
__a =[]
def generate(_snake_case : int , _snake_case : list ):
if k == 1:
res.append(tuple(arr[:] ) )
return
generate(k - 1 , lowerCamelCase__ )
for i in range(k - 1 ):
if k % 2 == 0: # k is even
__a =arr[k - 1], arr[i]
else: # k is odd
__a =arr[k - 1], arr[0]
generate(k - 1 , lowerCamelCase__ )
generate(len(lowerCamelCase__ ) , lowerCamelCase__ )
return res
if __name__ == "__main__":
_lowerCAmelCase : Dict = input("Enter numbers separated by a comma:\n").strip()
_lowerCAmelCase : List[Any] = [int(item) for item in user_input.split(",")]
print(heaps(arr))
| 354
|
def UpperCamelCase_( _snake_case : str , _snake_case : int ):
"""simple docstring"""
return [sentence[i : i + ngram_size] for i in range(len(_snake_case ) - ngram_size + 1 )]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 308
| 0
|
import argparse
import glob
import importlib.util
import os
import re
import black
from doc_builder.style_doc import style_docstrings_in_code
# All paths are set with the intent you should run this script from the root of the repo with the command
# python utils/check_copies.py
_lowerCAmelCase : List[str] = "src/diffusers"
_lowerCAmelCase : List[Any] = "."
# This is to make sure the diffusers module imported is the one in the repo.
_lowerCAmelCase : Optional[Any] = importlib.util.spec_from_file_location(
"diffusers",
os.path.join(DIFFUSERS_PATH, "__init__.py"),
submodule_search_locations=[DIFFUSERS_PATH],
)
_lowerCAmelCase : int = spec.loader.load_module()
def UpperCamelCase_( _snake_case : Dict , _snake_case : str ):
"""simple docstring"""
return line.startswith(__a ) or len(__a ) <= 1 or re.search(r'^\s*\)(\s*->.*:|:)\s*$' , __a ) is not None
def UpperCamelCase_( _snake_case : List[str] ):
"""simple docstring"""
__a =object_name.split('.' )
__a =0
# First let's find the module where our object lives.
__a =parts[i]
while i < len(__a ) and not os.path.isfile(os.path.join(__a , F'{module}.py' ) ):
i += 1
if i < len(__a ):
__a =os.path.join(__a , parts[i] )
if i >= len(__a ):
raise ValueError(F'`object_name` should begin with the name of a module of diffusers but got {object_name}.' )
with open(os.path.join(__a , F'{module}.py' ) , 'r' , encoding='utf-8' , newline='\n' ) as f:
__a =f.readlines()
# Now let's find the class / func in the code!
__a =''
__a =0
for name in parts[i + 1 :]:
while (
line_index < len(__a ) and re.search(rF'^{indent}(class|def)\s+{name}(\(|\:)' , lines[line_index] ) is None
):
line_index += 1
indent += " "
line_index += 1
if line_index >= len(__a ):
raise ValueError(F' {object_name} does not match any function or class in {module}.' )
# We found the beginning of the class / func, now let's find the end (when the indent diminishes).
__a =line_index
while line_index < len(__a ) and _should_continue(lines[line_index] , __a ):
line_index += 1
# Clean up empty lines at the end (if any).
while len(lines[line_index - 1] ) <= 1:
line_index -= 1
__a =lines[start_index:line_index]
return "".join(__a )
_lowerCAmelCase : Tuple = re.compile(r"^(\s*)#\s*Copied from\s+diffusers\.(\S+\.\S+)\s*($|\S.*$)")
_lowerCAmelCase : Union[str, Any] = re.compile(r"^\s*(\S+)->(\S+)(\s+.*|$)")
_lowerCAmelCase : List[Any] = re.compile(r"<FILL\s+[^>]*>")
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
__a =code.split('\n' )
__a =0
while idx < len(__a ) and len(lines[idx] ) == 0:
idx += 1
if idx < len(__a ):
return re.search(r'^(\s*)\S' , lines[idx] ).groups()[0]
return ""
def UpperCamelCase_( _snake_case : List[str] ):
"""simple docstring"""
__a =len(get_indent(__a ) ) > 0
if has_indent:
__a =F'class Bla:\n{code}'
__a =black.Mode(target_versions={black.TargetVersion.PYaa} , line_length=119 , preview=__a )
__a =black.format_str(__a , mode=__a )
__a =style_docstrings_in_code(__a )
return result[len('class Bla:\n' ) :] if has_indent else result
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : Optional[int]=False ):
"""simple docstring"""
with open(__a , 'r' , encoding='utf-8' , newline='\n' ) as f:
__a =f.readlines()
__a =[]
__a =0
# Not a for loop cause `lines` is going to change (if `overwrite=True`).
while line_index < len(__a ):
__a =_re_copy_warning.search(lines[line_index] )
if search is None:
line_index += 1
continue
# There is some copied code here, let's retrieve the original.
__a =search.groups()
__a =find_code_in_diffusers(__a )
__a =get_indent(__a )
__a =line_index + 1 if indent == theoretical_indent else line_index + 2
__a =theoretical_indent
__a =start_index
# Loop to check the observed code, stop when indentation diminishes or if we see a End copy comment.
__a =True
while line_index < len(__a ) and should_continue:
line_index += 1
if line_index >= len(__a ):
break
__a =lines[line_index]
__a =_should_continue(__a , __a ) and re.search(F'^{indent}# End copy' , __a ) is None
# Clean up empty lines at the end (if any).
while len(lines[line_index - 1] ) <= 1:
line_index -= 1
__a =lines[start_index:line_index]
__a =''.join(__a )
# Remove any nested `Copied from` comments to avoid circular copies
__a =[line for line in theoretical_code.split('\n' ) if _re_copy_warning.search(__a ) is None]
__a ='\n'.join(__a )
# Before comparing, use the `replace_pattern` on the original code.
if len(__a ) > 0:
__a =replace_pattern.replace('with' , '' ).split(',' )
__a =[_re_replace_pattern.search(__a ) for p in patterns]
for pattern in patterns:
if pattern is None:
continue
__a =pattern.groups()
__a =re.sub(__a , __a , __a )
if option.strip() == "all-casing":
__a =re.sub(obja.lower() , obja.lower() , __a )
__a =re.sub(obja.upper() , obja.upper() , __a )
# Blackify after replacement. To be able to do that, we need the header (class or function definition)
# from the previous line
__a =blackify(lines[start_index - 1] + theoretical_code )
__a =theoretical_code[len(lines[start_index - 1] ) :]
# Test for a diff and act accordingly.
if observed_code != theoretical_code:
diffs.append([object_name, start_index] )
if overwrite:
__a =lines[:start_index] + [theoretical_code] + lines[line_index:]
__a =start_index + 1
if overwrite and len(__a ) > 0:
# Warn the user a file has been modified.
print(F'Detected changes, rewriting {filename}.' )
with open(__a , 'w' , encoding='utf-8' , newline='\n' ) as f:
f.writelines(__a )
return diffs
def UpperCamelCase_( _snake_case : bool = False ):
"""simple docstring"""
__a =glob.glob(os.path.join(__a , '**/*.py' ) , recursive=__a )
__a =[]
for filename in all_files:
__a =is_copy_consistent(__a , __a )
diffs += [F'- {filename}: copy does not match {d[0]} at line {d[1]}' for d in new_diffs]
if not overwrite and len(__a ) > 0:
__a ='\n'.join(__a )
raise Exception(
'Found the following copy inconsistencies:\n'
+ diff
+ '\nRun `make fix-copies` or `python utils/check_copies.py --fix_and_overwrite` to fix them.' )
if __name__ == "__main__":
_lowerCAmelCase : Tuple = argparse.ArgumentParser()
parser.add_argument("--fix_and_overwrite", action="store_true", help="Whether to fix inconsistencies.")
_lowerCAmelCase : Optional[Any] = parser.parse_args()
check_copies(args.fix_and_overwrite)
| 355
|
import argparse
import pytorch_lightning as pl
import torch
from torch import nn
from transformers import LongformerForQuestionAnswering, LongformerModel
class __magic_name__ ( pl.LightningModule ):
def __init__( self , __snake_case ) -> List[Any]:
'''simple docstring'''
super().__init__()
__a =model
__a =2
__a =nn.Linear(self.model.config.hidden_size , self.num_labels )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : str , _snake_case : str , _snake_case : str ):
"""simple docstring"""
__a =LongformerModel.from_pretrained(_snake_case )
__a =LightningModel(_snake_case )
__a =torch.load(_snake_case , map_location=torch.device('cpu' ) )
lightning_model.load_state_dict(ckpt['state_dict'] )
# init longformer question answering model
__a =LongformerForQuestionAnswering.from_pretrained(_snake_case )
# transfer weights
longformer_for_qa.longformer.load_state_dict(lightning_model.model.state_dict() )
longformer_for_qa.qa_outputs.load_state_dict(lightning_model.qa_outputs.state_dict() )
longformer_for_qa.eval()
# save model
longformer_for_qa.save_pretrained(_snake_case )
print(F'Conversion successful. Model saved under {pytorch_dump_folder_path}' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--longformer_model",
default=None,
type=str,
required=True,
help="model identifier of longformer. Should be either `longformer-base-4096` or `longformer-large-4096`.",
)
parser.add_argument(
"--longformer_question_answering_ckpt_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch Lightning Checkpoint.",
)
parser.add_argument(
"--pytorch_dump_folder_path", default=None, type=str, required=True, help="Path to the output PyTorch model."
)
_lowerCAmelCase : Optional[int] = parser.parse_args()
convert_longformer_qa_checkpoint_to_pytorch(
args.longformer_model, args.longformer_question_answering_ckpt_path, args.pytorch_dump_folder_path
)
| 308
| 0
|
import math
from collections import defaultdict
from typing import List, Optional, Tuple, Union
import numpy as np
import torch
from ..configuration_utils import ConfigMixin, register_to_config
from .scheduling_utils import KarrasDiffusionSchedulers, SchedulerMixin, SchedulerOutput
def UpperCamelCase_( _snake_case : Union[str, Any] , _snake_case : Optional[int]=0.999 , _snake_case : List[str]="cosine" , ):
"""simple docstring"""
if alpha_transform_type == "cosine":
def alpha_bar_fn(_snake_case : Tuple ):
return math.cos((t + 0.008) / 1.008 * math.pi / 2 ) ** 2
elif alpha_transform_type == "exp":
def alpha_bar_fn(_snake_case : Any ):
return math.exp(t * -12.0 )
else:
raise ValueError(F'Unsupported alpha_tranform_type: {alpha_transform_type}' )
__a =[]
for i in range(__snake_case ):
__a =i / num_diffusion_timesteps
__a =(i + 1) / num_diffusion_timesteps
betas.append(min(1 - alpha_bar_fn(__snake_case ) / alpha_bar_fn(__snake_case ) , __snake_case ) )
return torch.tensor(__snake_case , dtype=torch.floataa )
class __magic_name__ ( lowerCamelCase__ , lowerCamelCase__ ):
SCREAMING_SNAKE_CASE = [e.name for e in KarrasDiffusionSchedulers]
SCREAMING_SNAKE_CASE = 2
@register_to_config
def __init__( self , __snake_case = 1000 , __snake_case = 0.0_0085 , __snake_case = 0.012 , __snake_case = "linear" , __snake_case = None , __snake_case = "epsilon" , __snake_case = False , __snake_case = False , __snake_case = 1.0 , __snake_case = "linspace" , __snake_case = 0 , ) -> List[str]:
'''simple docstring'''
if trained_betas is not None:
__a =torch.tensor(__snake_case , dtype=torch.floataa )
elif beta_schedule == "linear":
__a =torch.linspace(__snake_case , __snake_case , __snake_case , dtype=torch.floataa )
elif beta_schedule == "scaled_linear":
# this schedule is very specific to the latent diffusion model.
__a =(
torch.linspace(beta_start**0.5 , beta_end**0.5 , __snake_case , dtype=torch.floataa ) ** 2
)
elif beta_schedule == "squaredcos_cap_v2":
# Glide cosine schedule
__a =betas_for_alpha_bar(__snake_case , alpha_transform_type='cosine' )
elif beta_schedule == "exp":
__a =betas_for_alpha_bar(__snake_case , alpha_transform_type='exp' )
else:
raise NotImplementedError(f'{beta_schedule} does is not implemented for {self.__class__}' )
__a =1.0 - self.betas
__a =torch.cumprod(self.alphas , dim=0 )
# set all values
self.set_timesteps(__snake_case , __snake_case , __snake_case )
__a =use_karras_sigmas
def __magic_name__ ( self , __snake_case , __snake_case=None ) -> Optional[int]:
'''simple docstring'''
if schedule_timesteps is None:
__a =self.timesteps
__a =(schedule_timesteps == timestep).nonzero()
# The sigma index that is taken for the **very** first `step`
# is always the second index (or the last index if there is only 1)
# This way we can ensure we don't accidentally skip a sigma in
# case we start in the middle of the denoising schedule (e.g. for image-to-image)
if len(self._index_counter ) == 0:
__a =1 if len(__snake_case ) > 1 else 0
else:
__a =timestep.cpu().item() if torch.is_tensor(__snake_case ) else timestep
__a =self._index_counter[timestep_int]
return indices[pos].item()
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.config.timestep_spacing in ["linspace", "trailing"]:
return self.sigmas.max()
return (self.sigmas.max() ** 2 + 1) ** 0.5
def __magic_name__ ( self , __snake_case , __snake_case , ) -> List[str]:
'''simple docstring'''
__a =self.index_for_timestep(__snake_case )
__a =self.sigmas[step_index]
__a =sample / ((sigma**2 + 1) ** 0.5)
return sample
def __magic_name__ ( self , __snake_case , __snake_case = None , __snake_case = None , ) -> Any:
'''simple docstring'''
__a =num_inference_steps
__a =num_train_timesteps or self.config.num_train_timesteps
# "linspace", "leading", "trailing" corresponds to annotation of Table 2. of https://arxiv.org/abs/2305.08891
if self.config.timestep_spacing == "linspace":
__a =np.linspace(0 , num_train_timesteps - 1 , __snake_case , dtype=__snake_case )[::-1].copy()
elif self.config.timestep_spacing == "leading":
__a =num_train_timesteps // self.num_inference_steps
# creates integer timesteps by multiplying by ratio
# casting to int to avoid issues when num_inference_step is power of 3
__a =(np.arange(0 , __snake_case ) * step_ratio).round()[::-1].copy().astype(__snake_case )
timesteps += self.config.steps_offset
elif self.config.timestep_spacing == "trailing":
__a =num_train_timesteps / self.num_inference_steps
# creates integer timesteps by multiplying by ratio
# casting to int to avoid issues when num_inference_step is power of 3
__a =(np.arange(__snake_case , 0 , -step_ratio )).round().copy().astype(__snake_case )
timesteps -= 1
else:
raise ValueError(
f'{self.config.timestep_spacing} is not supported. Please make sure to choose one of \'linspace\', \'leading\' or \'trailing\'.' )
__a =np.array(((1 - self.alphas_cumprod) / self.alphas_cumprod) ** 0.5 )
__a =np.log(__snake_case )
__a =np.interp(__snake_case , np.arange(0 , len(__snake_case ) ) , __snake_case )
if self.config.use_karras_sigmas:
__a =self._convert_to_karras(in_sigmas=__snake_case , num_inference_steps=self.num_inference_steps )
__a =np.array([self._sigma_to_t(__snake_case , __snake_case ) for sigma in sigmas] )
__a =np.concatenate([sigmas, [0.0]] ).astype(np.floataa )
__a =torch.from_numpy(__snake_case ).to(device=__snake_case )
__a =torch.cat([sigmas[:1], sigmas[1:-1].repeat_interleave(2 ), sigmas[-1:]] )
__a =torch.from_numpy(__snake_case )
__a =torch.cat([timesteps[:1], timesteps[1:].repeat_interleave(2 )] )
if str(__snake_case ).startswith('mps' ):
# mps does not support float64
__a =timesteps.to(__snake_case , dtype=torch.floataa )
else:
__a =timesteps.to(device=__snake_case )
# empty dt and derivative
__a =None
__a =None
# for exp beta schedules, such as the one for `pipeline_shap_e.py`
# we need an index counter
__a =defaultdict(__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =np.log(__snake_case )
# get distribution
__a =log_sigma - log_sigmas[:, np.newaxis]
# get sigmas range
__a =np.cumsum((dists >= 0) , axis=0 ).argmax(axis=0 ).clip(max=log_sigmas.shape[0] - 2 )
__a =low_idx + 1
__a =log_sigmas[low_idx]
__a =log_sigmas[high_idx]
# interpolate sigmas
__a =(low - log_sigma) / (low - high)
__a =np.clip(__snake_case , 0 , 1 )
# transform interpolation to time range
__a =(1 - w) * low_idx + w * high_idx
__a =t.reshape(sigma.shape )
return t
def __magic_name__ ( self , __snake_case , __snake_case ) -> List[str]:
'''simple docstring'''
__a =in_sigmas[-1].item()
__a =in_sigmas[0].item()
__a =7.0 # 7.0 is the value used in the paper
__a =np.linspace(0 , 1 , __snake_case )
__a =sigma_min ** (1 / rho)
__a =sigma_max ** (1 / rho)
__a =(max_inv_rho + ramp * (min_inv_rho - max_inv_rho)) ** rho
return sigmas
@property
def __magic_name__ ( self ) -> str:
'''simple docstring'''
return self.dt is None
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case = True , ) -> List[str]:
'''simple docstring'''
__a =self.index_for_timestep(__snake_case )
# advance index counter by 1
__a =timestep.cpu().item() if torch.is_tensor(__snake_case ) else timestep
self._index_counter[timestep_int] += 1
if self.state_in_first_order:
__a =self.sigmas[step_index]
__a =self.sigmas[step_index + 1]
else:
# 2nd order / Heun's method
__a =self.sigmas[step_index - 1]
__a =self.sigmas[step_index]
# currently only gamma=0 is supported. This usually works best anyways.
# We can support gamma in the future but then need to scale the timestep before
# passing it to the model which requires a change in API
__a =0
__a =sigma * (gamma + 1) # Note: sigma_hat == sigma for now
# 1. compute predicted original sample (x_0) from sigma-scaled predicted noise
if self.config.prediction_type == "epsilon":
__a =sigma_hat if self.state_in_first_order else sigma_next
__a =sample - sigma_input * model_output
elif self.config.prediction_type == "v_prediction":
__a =sigma_hat if self.state_in_first_order else sigma_next
__a =model_output * (-sigma_input / (sigma_input**2 + 1) ** 0.5) + (
sample / (sigma_input**2 + 1)
)
elif self.config.prediction_type == "sample":
__a =model_output
else:
raise ValueError(
f'prediction_type given as {self.config.prediction_type} must be one of `epsilon`, or `v_prediction`' )
if self.config.clip_sample:
__a =pred_original_sample.clamp(
-self.config.clip_sample_range , self.config.clip_sample_range )
if self.state_in_first_order:
# 2. Convert to an ODE derivative for 1st order
__a =(sample - pred_original_sample) / sigma_hat
# 3. delta timestep
__a =sigma_next - sigma_hat
# store for 2nd order step
__a =derivative
__a =dt
__a =sample
else:
# 2. 2nd order / Heun's method
__a =(sample - pred_original_sample) / sigma_next
__a =(self.prev_derivative + derivative) / 2
# 3. take prev timestep & sample
__a =self.dt
__a =self.sample
# free dt and derivative
# Note, this puts the scheduler in "first order mode"
__a =None
__a =None
__a =None
__a =sample + derivative * dt
if not return_dict:
return (prev_sample,)
return SchedulerOutput(prev_sample=__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , ) -> Optional[int]:
'''simple docstring'''
__a =self.sigmas.to(device=original_samples.device , dtype=original_samples.dtype )
if original_samples.device.type == "mps" and torch.is_floating_point(__snake_case ):
# mps does not support float64
__a =self.timesteps.to(original_samples.device , dtype=torch.floataa )
__a =timesteps.to(original_samples.device , dtype=torch.floataa )
else:
__a =self.timesteps.to(original_samples.device )
__a =timesteps.to(original_samples.device )
__a =[self.index_for_timestep(__snake_case , __snake_case ) for t in timesteps]
__a =sigmas[step_indices].flatten()
while len(sigma.shape ) < len(original_samples.shape ):
__a =sigma.unsqueeze(-1 )
__a =original_samples + noise * sigma
return noisy_samples
def __len__( self ) -> Any:
'''simple docstring'''
return self.config.num_train_timesteps
| 356
|
import logging
import os
import random
import sys
from dataclasses import dataclass, field
from typing import Optional
import datasets
import numpy as np
import pandas as pd
from datasets import load_dataset
import transformers
from transformers import (
AutoConfig,
BartForSequenceClassification,
DataCollatorWithPadding,
EvalPrediction,
HfArgumentParser,
TapexTokenizer,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
_lowerCAmelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The name of the dataset to use (via the datasets library).'} )
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} , )
SCREAMING_SNAKE_CASE = field(
default=1_0_2_4 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Overwrite the cached preprocessed datasets or not.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Whether to pad all samples to `max_seq_length`. '
'If False, will pad the samples dynamically when batching to the maximum length in the batch.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of training examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of evaluation examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of prediction examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = field(default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the test data.'} )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
if self.dataset_name is not None:
pass
elif self.train_file is None or self.validation_file is None:
raise ValueError('Need either a GLUE task, a training/validation file or a dataset name.' )
else:
__a =self.train_file.split('.' )[-1]
assert train_extension in ["csv", "json"], "`train_file` should be a csv or a json file."
__a =self.validation_file.split('.' )[-1]
assert (
validation_extension == train_extension
), "`validation_file` should have the same extension (csv or json) as `train_file`."
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to use one of the fast tokenizer (backed by the tokenizers library) or not.'} , )
SCREAMING_SNAKE_CASE = field(
default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Will use the token generated when running `huggingface-cli login` (necessary to use this script '
'with private models).'
)
} , )
def UpperCamelCase_( ):
"""simple docstring"""
__a =HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments) )
if len(sys.argv ) == 2 and sys.argv[1].endswith('.json' ):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
__a , __a , __a =parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) )
else:
__a , __a , __a =parser.parse_args_into_dataclasses()
# Setup logging
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , handlers=[logging.StreamHandler(sys.stdout )] , )
__a =training_args.get_process_log_level()
logger.setLevel(_snake_case )
datasets.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Log on each process the small summary:
logger.warning(
F'Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}'
+ F'distributed training: {bool(training_args.local_rank != -1 )}, 16-bits training: {training_args.fpaa}' )
logger.info(F'Training/evaluation parameters {training_args}' )
# Detecting last checkpoint.
__a =None
if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir:
__a =get_last_checkpoint(training_args.output_dir )
if last_checkpoint is None and len(os.listdir(training_args.output_dir ) ) > 0:
raise ValueError(
F'Output directory ({training_args.output_dir}) already exists and is not empty. '
'Use --overwrite_output_dir to overcome.' )
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
F'Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change '
'the `--output_dir` or add `--overwrite_output_dir` to train from scratch.' )
# Set seed before initializing model.
set_seed(training_args.seed )
# Get the datasets: you can either provide your own CSV/JSON training and evaluation files (see below)
# or specify a GLUE benchmark task (the dataset will be downloaded automatically from the datasets Hub).
#
# For JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# If the CSVs/JSONs contain only one non-label column, the script does single sentence classification on this
# single column. You can easily tweak this behavior (see below)
#
# In distributed training, the load_dataset function guarantee that only one local process can concurrently
# download the dataset.
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
__a =load_dataset(
data_args.dataset_name , data_args.dataset_config_name , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from your local files.
# CSV/JSON training and evaluation files are needed.
__a ={'train': data_args.train_file, 'validation': data_args.validation_file}
# Get the test dataset: you can provide your own CSV/JSON test file (see below)
# when you use `do_predict` without specifying a GLUE benchmark task.
if training_args.do_predict:
if data_args.test_file is not None:
__a =data_args.train_file.split('.' )[-1]
__a =data_args.test_file.split('.' )[-1]
assert (
test_extension == train_extension
), "`test_file` should have the same extension (csv or json) as `train_file`."
__a =data_args.test_file
else:
raise ValueError('Need either a GLUE task or a test file for `do_predict`.' )
for key in data_files.keys():
logger.info(F'load a local file for {key}: {data_files[key]}' )
if data_args.train_file.endswith('.csv' ):
# Loading a dataset from local csv files
__a =load_dataset('csv' , data_files=_snake_case , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from local json files
__a =load_dataset('json' , data_files=_snake_case , cache_dir=model_args.cache_dir )
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.html.
# Labels
__a =raw_datasets['train'].features['label'].names
__a =len(_snake_case )
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
__a =AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# load tapex tokenizer
__a =TapexTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , use_fast=model_args.use_fast_tokenizer , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , add_prefix_space=_snake_case , )
__a =BartForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_tf=bool('.ckpt' in model_args.model_name_or_path ) , config=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# Padding strategy
if data_args.pad_to_max_length:
__a ='max_length'
else:
# We will pad later, dynamically at batch creation, to the max sequence length in each batch
__a =False
# Some models have set the order of the labels to use, so let's make sure we do use it.
__a ={'Refused': 0, 'Entailed': 1}
__a ={0: 'Refused', 1: 'Entailed'}
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warning(
F'The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the'
F'model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}.' )
__a =min(data_args.max_seq_length , tokenizer.model_max_length )
def preprocess_tabfact_function(_snake_case : Tuple ):
# Tokenize the texts
def _convert_table_text_to_pandas(_snake_case : Optional[Any] ):
__a =[_table_row.split('#' ) for _table_row in _table_text.strip('\n' ).split('\n' )]
__a =pd.DataFrame.from_records(_table_content[1:] , columns=_table_content[0] )
return _table_pd
__a =examples['statement']
__a =list(map(_convert_table_text_to_pandas , examples['table_text'] ) )
__a =tokenizer(_snake_case , _snake_case , padding=_snake_case , max_length=_snake_case , truncation=_snake_case )
__a =examples['label']
return result
with training_args.main_process_first(desc='dataset map pre-processing' ):
__a =raw_datasets.map(
_snake_case , batched=_snake_case , load_from_cache_file=not data_args.overwrite_cache , desc='Running tokenizer on dataset' , )
if training_args.do_train:
if "train" not in raw_datasets:
raise ValueError('--do_train requires a train dataset' )
__a =raw_datasets['train']
if data_args.max_train_samples is not None:
__a =train_dataset.select(range(data_args.max_train_samples ) )
if training_args.do_eval:
if "validation" not in raw_datasets and "validation_matched" not in raw_datasets:
raise ValueError('--do_eval requires a validation dataset' )
__a =raw_datasets['validation']
if data_args.max_eval_samples is not None:
__a =eval_dataset.select(range(data_args.max_eval_samples ) )
if training_args.do_predict or data_args.test_file is not None:
if "test" not in raw_datasets and "test_matched" not in raw_datasets:
raise ValueError('--do_predict requires a test dataset' )
__a =raw_datasets['test']
if data_args.max_predict_samples is not None:
__a =predict_dataset.select(range(data_args.max_predict_samples ) )
# Log a few random samples from the training set:
if training_args.do_train:
for index in random.sample(range(len(_snake_case ) ) , 3 ):
logger.info(F'Sample {index} of the training set: {train_dataset[index]}.' )
# You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
# predictions and label_ids field) and has to return a dictionary string to float.
def compute_metrics(_snake_case : EvalPrediction ):
__a =p.predictions[0] if isinstance(p.predictions , _snake_case ) else p.predictions
__a =np.argmax(_snake_case , axis=1 )
return {"accuracy": (preds == p.label_ids).astype(np.floataa ).mean().item()}
# Data collator will default to DataCollatorWithPadding, so we change it if we already did the padding.
if data_args.pad_to_max_length:
__a =default_data_collator
elif training_args.fpaa:
__a =DataCollatorWithPadding(_snake_case , pad_to_multiple_of=8 )
else:
__a =None
# Initialize our Trainer
__a =Trainer(
model=_snake_case , args=_snake_case , train_dataset=train_dataset if training_args.do_train else None , eval_dataset=eval_dataset if training_args.do_eval else None , compute_metrics=_snake_case , tokenizer=_snake_case , data_collator=_snake_case , )
# Training
if training_args.do_train:
__a =None
if training_args.resume_from_checkpoint is not None:
__a =training_args.resume_from_checkpoint
elif last_checkpoint is not None:
__a =last_checkpoint
__a =trainer.train(resume_from_checkpoint=_snake_case )
__a =train_result.metrics
__a =(
data_args.max_train_samples if data_args.max_train_samples is not None else len(_snake_case )
)
__a =min(_snake_case , len(_snake_case ) )
trainer.save_model() # Saves the tokenizer too for easy upload
trainer.log_metrics('train' , _snake_case )
trainer.save_metrics('train' , _snake_case )
trainer.save_state()
# Evaluation
if training_args.do_eval:
logger.info('*** Evaluate ***' )
__a =trainer.evaluate(eval_dataset=_snake_case )
__a =data_args.max_eval_samples if data_args.max_eval_samples is not None else len(_snake_case )
__a =min(_snake_case , len(_snake_case ) )
trainer.log_metrics('eval' , _snake_case )
trainer.save_metrics('eval' , _snake_case )
if training_args.do_predict:
logger.info('*** Predict ***' )
# Removing the `label` columns because it contains -1 and Trainer won't like that.
__a =predict_dataset.remove_columns('label' )
__a =trainer.predict(_snake_case , metric_key_prefix='predict' ).predictions
__a =np.argmax(_snake_case , axis=1 )
__a =os.path.join(training_args.output_dir , 'predict_results_tabfact.txt' )
if trainer.is_world_process_zero():
with open(_snake_case , 'w' ) as writer:
logger.info('***** Predict Results *****' )
writer.write('index\tprediction\n' )
for index, item in enumerate(_snake_case ):
__a =label_list[item]
writer.write(F'{index}\t{item}\n' )
__a ={'finetuned_from': model_args.model_name_or_path, 'tasks': 'text-classification'}
if training_args.push_to_hub:
trainer.push_to_hub(**_snake_case )
else:
trainer.create_model_card(**_snake_case )
def UpperCamelCase_( _snake_case : Union[str, Any] ):
"""simple docstring"""
main()
if __name__ == "__main__":
main()
| 308
| 0
|
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from ...tokenization_utils import AddedToken, PreTrainedTokenizer
from ...utils import logging
_lowerCAmelCase : Union[str, Any] = logging.get_logger(__name__)
_lowerCAmelCase : Dict = "▁"
_lowerCAmelCase : str = {"vocab_file": "sentencepiece.bpe.model", "monolingual_vocab_file": "dict.txt"}
_lowerCAmelCase : List[Any] = {
"vocab_file": {
"vinai/bartpho-syllable": "https://huggingface.co/vinai/bartpho-syllable/resolve/main/sentencepiece.bpe.model",
},
"monolingual_vocab_file": {
"vinai/bartpho-syllable": "https://huggingface.co/vinai/bartpho-syllable/resolve/main/dict.txt",
},
}
_lowerCAmelCase : Dict = {"vinai/bartpho-syllable": 1_024}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = VOCAB_FILES_NAMES
SCREAMING_SNAKE_CASE = PRETRAINED_VOCAB_FILES_MAP
SCREAMING_SNAKE_CASE = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
SCREAMING_SNAKE_CASE = ['''input_ids''', '''attention_mask''']
def __init__( self , __snake_case , __snake_case , __snake_case="<s>" , __snake_case="</s>" , __snake_case="</s>" , __snake_case="<s>" , __snake_case="<unk>" , __snake_case="<pad>" , __snake_case="<mask>" , __snake_case = None , **__snake_case , ) -> int:
'''simple docstring'''
__a =AddedToken(_snake_case , lstrip=_snake_case , rstrip=_snake_case ) if isinstance(_snake_case , _snake_case ) else mask_token
__a ={} if sp_model_kwargs is None else sp_model_kwargs
super().__init__(
bos_token=_snake_case , eos_token=_snake_case , unk_token=_snake_case , sep_token=_snake_case , cls_token=_snake_case , pad_token=_snake_case , mask_token=_snake_case , sp_model_kwargs=self.sp_model_kwargs , **_snake_case , )
__a =vocab_file
__a =monolingual_vocab_file
__a =spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(str(_snake_case ) )
# Load the reduced vocab
# Keep order of special tokens for backward compatibility
__a ={}
__a =0
for token in [bos_token, pad_token, eos_token, unk_token, sep_token, cls_token]:
if str(_snake_case ) not in self.fairseq_tokens_to_ids:
__a =cnt
cnt += 1
with open(_snake_case , 'r' , encoding='utf-8' ) as f:
for line in f.readlines():
__a =line.strip().split()[0]
__a =len(self.fairseq_tokens_to_ids )
if str(_snake_case ) not in self.fairseq_tokens_to_ids:
__a =len(self.fairseq_tokens_to_ids )
__a ={v: k for k, v in self.fairseq_tokens_to_ids.items()}
def __getstate__( self ) -> List[str]:
'''simple docstring'''
__a =self.__dict__.copy()
__a =None
__a =self.sp_model.serialized_model_proto()
return state
def __setstate__( self , __snake_case ) -> Tuple:
'''simple docstring'''
__a =d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a ={}
__a =spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.LoadFromSerializedProto(self.sp_model_proto )
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> Any:
'''simple docstring'''
if token_ids_a is None:
return [self.cls_token_id] + token_ids_a + [self.sep_token_id]
__a =[self.cls_token_id]
__a =[self.sep_token_id]
return cls + token_ids_a + sep + sep + token_ids_a + sep
def __magic_name__ ( self , __snake_case , __snake_case = None , __snake_case = False ) -> Dict:
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=_snake_case , token_ids_a=_snake_case , already_has_special_tokens=_snake_case )
if token_ids_a is None:
return [1] + ([0] * len(_snake_case )) + [1]
return [1] + ([0] * len(_snake_case )) + [1, 1] + ([0] * len(_snake_case )) + [1]
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> str:
'''simple docstring'''
__a =[self.sep_token_id]
__a =[self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
@property
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
return len(self.fairseq_ids_to_tokens )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a ={self.convert_ids_to_tokens(_snake_case ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __magic_name__ ( self , __snake_case ) -> Tuple:
'''simple docstring'''
return self.sp_model.encode(_snake_case , out_type=_snake_case )
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
if token in self.fairseq_tokens_to_ids:
return self.fairseq_tokens_to_ids[token]
else:
return self.unk_token_id
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
return self.fairseq_ids_to_tokens[index]
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
__a =''.join(_snake_case ).replace(_snake_case , ' ' ).strip()
return out_string
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> str:
'''simple docstring'''
if not os.path.isdir(_snake_case ):
logger.error(f'Vocabulary path ({save_directory}) should be a directory' )
return
__a =os.path.join(
_snake_case , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(
_snake_case , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['monolingual_vocab_file'] , )
if os.path.abspath(self.vocab_file ) != os.path.abspath(_snake_case ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , _snake_case )
elif not os.path.isfile(self.vocab_file ):
with open(_snake_case , 'wb' ) as fi:
__a =self.sp_model.serialized_model_proto()
fi.write(_snake_case )
if os.path.abspath(self.monolingual_vocab_file ) != os.path.abspath(
_snake_case ) and os.path.isfile(self.monolingual_vocab_file ):
copyfile(self.monolingual_vocab_file , _snake_case )
elif not os.path.isfile(self.monolingual_vocab_file ):
with open(_snake_case , 'w' , encoding='utf-8' ) as fp:
for token in self.fairseq_tokens_to_ids:
if token not in self.all_special_tokens:
fp.write(f'{str(_snake_case )} \n' )
return out_vocab_file, out_monolingual_vocab_file
| 357
|
from __future__ import annotations
import time
import numpy as np
_lowerCAmelCase : List[str] = [8, 5, 9, 7]
_lowerCAmelCase : List[str] = [
[2, 0, 1, 1],
[0, 1, 2, 1],
[4, 0, 0, 3],
[0, 2, 1, 0],
[1, 0, 3, 0],
]
_lowerCAmelCase : List[Any] = [
[3, 2, 1, 4],
[0, 2, 5, 2],
[5, 1, 0, 5],
[1, 5, 3, 0],
[3, 0, 3, 3],
]
class __magic_name__ :
def __init__( self , __snake_case , __snake_case , __snake_case , ) -> None:
'''simple docstring'''
__a =claim_vector
__a =allocated_resources_table
__a =maximum_claim_table
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return [
sum(p_item[i] for p_item in self.__allocated_resources_table )
for i in range(len(self.__allocated_resources_table[0] ) )
]
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return np.array(self.__claim_vector ) - np.array(
self.__processes_resource_summation() )
def __magic_name__ ( self ) -> list[list[int]]:
'''simple docstring'''
return [
list(np.array(self.__maximum_claim_table[i] ) - np.array(__snake_case ) )
for i, allocated_resource in enumerate(self.__allocated_resources_table )
]
def __magic_name__ ( self ) -> dict[int, list[int]]:
'''simple docstring'''
return {self.__need().index(__snake_case ): i for i in self.__need()}
def __magic_name__ ( self , **__snake_case ) -> None:
'''simple docstring'''
__a =self.__need()
__a =self.__allocated_resources_table
__a =self.__available_resources()
__a =self.__need_index_manager()
for kw, val in kwargs.items():
if kw and val is True:
self.__pretty_data()
print('_' * 50 + '\n' )
while need_list:
__a =False
for each_need in need_list:
__a =True
for index, need in enumerate(__snake_case ):
if need > available_resources[index]:
__a =False
break
if execution:
__a =True
# get the original index of the process from ind_ctrl db
for original_need_index, need_clone in need_index_manager.items():
if each_need == need_clone:
__a =original_need_index
print(f'Process {process_number + 1} is executing.' )
# remove the process run from stack
need_list.remove(__snake_case )
# update available/freed resources stack
__a =np.array(__snake_case ) + np.array(
alloc_resources_table[process_number] )
print(
'Updated available resource stack for processes: '
+ ' '.join([str(__snake_case ) for x in available_resources] ) )
break
if safe:
print('The process is in a safe state.\n' )
else:
print('System in unsafe state. Aborting...\n' )
break
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
print(' ' * 9 + 'Allocated Resource Table' )
for item in self.__allocated_resources_table:
print(
f'P{self.__allocated_resources_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(' ' * 9 + 'System Resource Table' )
for item in self.__maximum_claim_table:
print(
f'P{self.__maximum_claim_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(
'Current Usage by Active Processes: '
+ ' '.join(str(__snake_case ) for x in self.__claim_vector ) )
print(
'Initial Available Resources: '
+ ' '.join(str(__snake_case ) for x in self.__available_resources() ) )
time.sleep(1 )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 0
|
"""simple docstring"""
def UpperCamelCase_( _snake_case : int , _snake_case : int ):
"""simple docstring"""
return int((input_a, input_a).count(0 ) == 0 )
def UpperCamelCase_( ):
"""simple docstring"""
assert and_gate(0 , 0 ) == 0
assert and_gate(0 , 1 ) == 0
assert and_gate(1 , 0 ) == 0
assert and_gate(1 , 1 ) == 1
if __name__ == "__main__":
test_and_gate()
print(and_gate(1, 0))
print(and_gate(0, 0))
print(and_gate(0, 1))
print(and_gate(1, 1))
| 358
|
import string
# frequency taken from https://en.wikipedia.org/wiki/Letter_frequency
_lowerCAmelCase : Tuple = {
"E": 12.70,
"T": 9.06,
"A": 8.17,
"O": 7.51,
"I": 6.97,
"N": 6.75,
"S": 6.33,
"H": 6.09,
"R": 5.99,
"D": 4.25,
"L": 4.03,
"C": 2.78,
"U": 2.76,
"M": 2.41,
"W": 2.36,
"F": 2.23,
"G": 2.02,
"Y": 1.97,
"P": 1.93,
"B": 1.29,
"V": 0.98,
"K": 0.77,
"J": 0.15,
"X": 0.15,
"Q": 0.10,
"Z": 0.07,
}
_lowerCAmelCase : Optional[int] = "ETAOINSHRDLCUMWFGYPBVKJXQZ"
_lowerCAmelCase : Any = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a ={letter: 0 for letter in string.ascii_uppercase}
for letter in message.upper():
if letter in LETTERS:
letter_count[letter] += 1
return letter_count
def UpperCamelCase_( _snake_case : tuple ):
"""simple docstring"""
return x[0]
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_letter_count(_snake_case )
__a ={
freq: [] for letter, freq in letter_to_freq.items()
}
for letter in LETTERS:
freq_to_letter[letter_to_freq[letter]].append(_snake_case )
__a ={}
for freq in freq_to_letter:
freq_to_letter[freq].sort(key=ETAOIN.find , reverse=_snake_case )
__a =''.join(freq_to_letter[freq] )
__a =list(freq_to_letter_str.items() )
freq_pairs.sort(key=_snake_case , reverse=_snake_case )
__a =[freq_pair[1] for freq_pair in freq_pairs]
return "".join(_snake_case )
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_frequency_order(_snake_case )
__a =0
for common_letter in ETAOIN[:6]:
if common_letter in freq_order[:6]:
match_score += 1
for uncommon_letter in ETAOIN[-6:]:
if uncommon_letter in freq_order[-6:]:
match_score += 1
return match_score
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 0
|
import collections.abc
from typing import Optional, Tuple, Union
import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
from ...activations import ACTaFN
from ...modeling_outputs import BaseModelOutputWithNoAttention, ImageClassifierOutputWithNoAttention
from ...modeling_utils import PreTrainedModel
from ...utils import add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward, logging
from .configuration_poolformer import PoolFormerConfig
_lowerCAmelCase : int = logging.get_logger(__name__)
# General docstring
_lowerCAmelCase : List[Any] = """PoolFormerConfig"""
# Base docstring
_lowerCAmelCase : Optional[int] = """sail/poolformer_s12"""
_lowerCAmelCase : Union[str, Any] = [1, 512, 7, 7]
# Image classification docstring
_lowerCAmelCase : Tuple = """sail/poolformer_s12"""
_lowerCAmelCase : Tuple = """tabby, tabby cat"""
_lowerCAmelCase : str = [
"""sail/poolformer_s12""",
# See all PoolFormer models at https://huggingface.co/models?filter=poolformer
]
def UpperCamelCase_( _snake_case : str , _snake_case : Union[str, Any] = 0.0 , _snake_case : str = False ):
"""simple docstring"""
if drop_prob == 0.0 or not training:
return input
__a =1 - drop_prob
__a =(input.shape[0],) + (1,) * (input.ndim - 1) # work with diff dim tensors, not just 2D ConvNets
__a =keep_prob + torch.rand(_snake_case , dtype=input.dtype , device=input.device )
random_tensor.floor_() # binarize
__a =input.div(_snake_case ) * random_tensor
return output
class __magic_name__ ( nn.Module ):
def __init__( self , __snake_case = None ) -> Optional[Any]:
'''simple docstring'''
super().__init__()
__a =drop_prob
def __magic_name__ ( self , __snake_case ) -> Tuple:
'''simple docstring'''
return drop_path(__lowercase , self.drop_prob , self.training )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return "p={}".format(self.drop_prob )
class __magic_name__ ( nn.Module ):
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=None ) -> Tuple:
'''simple docstring'''
super().__init__()
__a =patch_size if isinstance(__lowercase , collections.abc.Iterable ) else (patch_size, patch_size)
__a =stride if isinstance(__lowercase , collections.abc.Iterable ) else (stride, stride)
__a =padding if isinstance(__lowercase , collections.abc.Iterable ) else (padding, padding)
__a =nn.Convad(__lowercase , __lowercase , kernel_size=__lowercase , stride=__lowercase , padding=__lowercase )
__a =norm_layer(__lowercase ) if norm_layer else nn.Identity()
def __magic_name__ ( self , __snake_case ) -> Dict:
'''simple docstring'''
__a =self.projection(__lowercase )
__a =self.norm(__lowercase )
return embeddings
class __magic_name__ ( nn.GroupNorm ):
def __init__( self , __snake_case , **__snake_case ) -> Optional[Any]:
'''simple docstring'''
super().__init__(1 , __lowercase , **__lowercase )
class __magic_name__ ( nn.Module ):
def __init__( self , __snake_case ) -> int:
'''simple docstring'''
super().__init__()
__a =nn.AvgPoolad(__lowercase , stride=1 , padding=pool_size // 2 , count_include_pad=__lowercase )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
return self.pool(__lowercase ) - hidden_states
class __magic_name__ ( nn.Module ):
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
super().__init__()
__a =nn.Convad(__lowercase , __lowercase , 1 )
__a =nn.Convad(__lowercase , __lowercase , 1 )
__a =PoolFormerDropPath(__lowercase )
if isinstance(config.hidden_act , __lowercase ):
__a =ACTaFN[config.hidden_act]
else:
__a =config.hidden_act
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
__a =self.conva(__lowercase )
__a =self.act_fn(__lowercase )
__a =self.drop(__lowercase )
__a =self.conva(__lowercase )
__a =self.drop(__lowercase )
return hidden_states
class __magic_name__ ( nn.Module ):
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
super().__init__()
__a =PoolFormerPooling(__lowercase )
__a =PoolFormerOutput(__lowercase , __lowercase , __lowercase , __lowercase )
__a =PoolFormerGroupNorm(__lowercase )
__a =PoolFormerGroupNorm(__lowercase )
# Useful for training neural nets
__a =PoolFormerDropPath(__lowercase ) if drop_path > 0.0 else nn.Identity()
__a =config.use_layer_scale
if config.use_layer_scale:
__a =nn.Parameter(
config.layer_scale_init_value * torch.ones((__lowercase) ) , requires_grad=__lowercase )
__a =nn.Parameter(
config.layer_scale_init_value * torch.ones((__lowercase) ) , requires_grad=__lowercase )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
if self.use_layer_scale:
__a =self.pooling(self.before_norm(__lowercase ) )
__a =self.layer_scale_a.unsqueeze(-1 ).unsqueeze(-1 ) * pooling_output
# First residual connection
__a =hidden_states + self.drop_path(__lowercase )
__a =()
__a =self.output(self.after_norm(__lowercase ) )
__a =self.layer_scale_a.unsqueeze(-1 ).unsqueeze(-1 ) * layer_output
# Second residual connection
__a =hidden_states + self.drop_path(__lowercase )
__a =(output,) + outputs
return outputs
else:
__a =self.drop_path(self.pooling(self.before_norm(__lowercase ) ) )
# First residual connection
__a =pooling_output + hidden_states
__a =()
# Second residual connection inside the PoolFormerOutput block
__a =self.drop_path(self.output(self.after_norm(__lowercase ) ) )
__a =hidden_states + layer_output
__a =(output,) + outputs
return outputs
class __magic_name__ ( nn.Module ):
def __init__( self , __snake_case ) -> Dict:
'''simple docstring'''
super().__init__()
__a =config
# stochastic depth decay rule
__a =[x.item() for x in torch.linspace(0 , config.drop_path_rate , sum(config.depths ) )]
# patch embeddings
__a =[]
for i in range(config.num_encoder_blocks ):
embeddings.append(
PoolFormerEmbeddings(
patch_size=config.patch_sizes[i] , stride=config.strides[i] , padding=config.padding[i] , num_channels=config.num_channels if i == 0 else config.hidden_sizes[i - 1] , hidden_size=config.hidden_sizes[i] , ) )
__a =nn.ModuleList(__lowercase )
# Transformer blocks
__a =[]
__a =0
for i in range(config.num_encoder_blocks ):
# each block consists of layers
__a =[]
if i != 0:
cur += config.depths[i - 1]
for j in range(config.depths[i] ):
layers.append(
PoolFormerLayer(
__lowercase , num_channels=config.hidden_sizes[i] , pool_size=config.pool_size , hidden_size=config.hidden_sizes[i] , intermediate_size=int(config.hidden_sizes[i] * config.mlp_ratio ) , drop_path=dpr[cur + j] , ) )
blocks.append(nn.ModuleList(__lowercase ) )
__a =nn.ModuleList(__lowercase )
def __magic_name__ ( self , __snake_case , __snake_case=False , __snake_case=True ) -> Dict:
'''simple docstring'''
__a =() if output_hidden_states else None
__a =pixel_values
for idx, layers in enumerate(zip(self.patch_embeddings , self.block ) ):
__a , __a =layers
# Get patch embeddings from hidden_states
__a =embedding_layer(__lowercase )
# Send the embeddings through the blocks
for _, blk in enumerate(__lowercase ):
__a =blk(__lowercase )
__a =layer_outputs[0]
if output_hidden_states:
__a =all_hidden_states + (hidden_states,)
if not return_dict:
return tuple(v for v in [hidden_states, all_hidden_states] if v is not None )
return BaseModelOutputWithNoAttention(last_hidden_state=__lowercase , hidden_states=__lowercase )
class __magic_name__ ( __A ):
SCREAMING_SNAKE_CASE = PoolFormerConfig
SCREAMING_SNAKE_CASE = 'poolformer'
SCREAMING_SNAKE_CASE = 'pixel_values'
SCREAMING_SNAKE_CASE = True
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
if isinstance(__lowercase , (nn.Linear, nn.Convad) ):
module.weight.data.normal_(mean=0.0 , std=self.config.initializer_range )
if module.bias is not None:
module.bias.data.zero_()
elif isinstance(__lowercase , nn.LayerNorm ):
module.bias.data.zero_()
module.weight.data.fill_(1.0 )
def __magic_name__ ( self , __snake_case , __snake_case=False ) -> str:
'''simple docstring'''
if isinstance(__lowercase , __lowercase ):
__a =value
_lowerCAmelCase : Any = r"""
This model is a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) sub-class. Use
it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
behavior.
Parameters:
config ([`PoolFormerConfig`]): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the
configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
"""
_lowerCAmelCase : List[Any] = r"""
Args:
pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
Pixel values. Pixel values can be obtained using [`AutoImageProcessor`]. See
[`PoolFormerImageProcessor.__call__`] for details.
"""
@add_start_docstrings(
'The bare PoolFormer Model transformer outputting raw hidden-states without any specific head on top.' , __A , )
class __magic_name__ ( __A ):
def __init__( self , __snake_case ) -> Dict:
'''simple docstring'''
super().__init__(__lowercase )
__a =config
__a =PoolFormerEncoder(__lowercase )
# Initialize weights and apply final processing
self.post_init()
def __magic_name__ ( self ) -> str:
'''simple docstring'''
return self.embeddings.patch_embeddings
@add_start_docstrings_to_model_forward(__lowercase )
@add_code_sample_docstrings(
checkpoint=_CHECKPOINT_FOR_DOC , output_type=__lowercase , config_class=_CONFIG_FOR_DOC , modality='vision' , expected_output=_EXPECTED_OUTPUT_SHAPE , )
def __magic_name__ ( self , __snake_case = None , __snake_case = None , __snake_case = None , ) -> str:
'''simple docstring'''
__a =(
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
)
__a =return_dict if return_dict is not None else self.config.use_return_dict
if pixel_values is None:
raise ValueError('You have to specify pixel_values' )
__a =self.encoder(
__lowercase , output_hidden_states=__lowercase , return_dict=__lowercase , )
__a =encoder_outputs[0]
if not return_dict:
return (sequence_output, None) + encoder_outputs[1:]
return BaseModelOutputWithNoAttention(
last_hidden_state=__lowercase , hidden_states=encoder_outputs.hidden_states , )
class __magic_name__ ( nn.Module ):
def __init__( self , __snake_case ) -> str:
'''simple docstring'''
super().__init__()
__a =nn.Linear(config.hidden_size , config.hidden_size )
def __magic_name__ ( self , __snake_case ) -> Any:
'''simple docstring'''
__a =self.dense(__lowercase )
return output
@add_start_docstrings(
'\n PoolFormer Model transformer with an image classification head on top\n ' , __A , )
class __magic_name__ ( __A ):
def __init__( self , __snake_case ) -> int:
'''simple docstring'''
super().__init__(__lowercase )
__a =config.num_labels
__a =PoolFormerModel(__lowercase )
# Final norm
__a =PoolFormerGroupNorm(config.hidden_sizes[-1] )
# Classifier head
__a =(
nn.Linear(config.hidden_sizes[-1] , config.num_labels ) if config.num_labels > 0 else nn.Identity()
)
# Initialize weights and apply final processing
self.post_init()
@add_start_docstrings_to_model_forward(__lowercase )
@add_code_sample_docstrings(
checkpoint=_IMAGE_CLASS_CHECKPOINT , output_type=__lowercase , config_class=_CONFIG_FOR_DOC , expected_output=_IMAGE_CLASS_EXPECTED_OUTPUT , )
def __magic_name__ ( self , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , ) -> Union[str, Any]:
'''simple docstring'''
__a =return_dict if return_dict is not None else self.config.use_return_dict
__a =self.poolformer(
__lowercase , output_hidden_states=__lowercase , return_dict=__lowercase , )
__a =outputs[0]
__a =self.classifier(self.norm(__lowercase ).mean([-2, -1] ) )
__a =None
if labels is not None:
if self.config.problem_type is None:
if self.num_labels == 1:
__a ='regression'
elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
__a ='single_label_classification'
else:
__a ='multi_label_classification'
if self.config.problem_type == "regression":
__a =MSELoss()
if self.num_labels == 1:
__a =loss_fct(logits.squeeze() , labels.squeeze() )
else:
__a =loss_fct(__lowercase , __lowercase )
elif self.config.problem_type == "single_label_classification":
__a =CrossEntropyLoss()
__a =loss_fct(logits.view(-1 , self.num_labels ) , labels.view(-1 ) )
elif self.config.problem_type == "multi_label_classification":
__a =BCEWithLogitsLoss()
__a =loss_fct(__lowercase , __lowercase )
if not return_dict:
__a =(logits,) + outputs[2:]
return ((loss,) + output) if loss is not None else output
return ImageClassifierOutputWithNoAttention(loss=__lowercase , logits=__lowercase , hidden_states=outputs.hidden_states )
| 359
|
from ...configuration_utils import PretrainedConfig
from ...utils import logging
_lowerCAmelCase : Any = logging.get_logger(__name__)
_lowerCAmelCase : int = {
"caidas/swin2sr-classicalsr-x2-64": (
"https://huggingface.co/caidas/swin2sr-classicalsr-x2-64/resolve/main/config.json"
),
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'swin2sr'
SCREAMING_SNAKE_CASE = {
'hidden_size': 'embed_dim',
'num_attention_heads': 'num_heads',
'num_hidden_layers': 'num_layers',
}
def __init__( self , __snake_case=64 , __snake_case=1 , __snake_case=3 , __snake_case=180 , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=8 , __snake_case=2.0 , __snake_case=True , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.1 , __snake_case="gelu" , __snake_case=False , __snake_case=0.02 , __snake_case=1e-5 , __snake_case=2 , __snake_case=1.0 , __snake_case="1conv" , __snake_case="pixelshuffle" , **__snake_case , ) -> Dict:
'''simple docstring'''
super().__init__(**__snake_case )
__a =image_size
__a =patch_size
__a =num_channels
__a =embed_dim
__a =depths
__a =len(__snake_case )
__a =num_heads
__a =window_size
__a =mlp_ratio
__a =qkv_bias
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =drop_path_rate
__a =hidden_act
__a =use_absolute_embeddings
__a =layer_norm_eps
__a =initializer_range
__a =upscale
__a =img_range
__a =resi_connection
__a =upsampler
| 308
| 0
|
import unittest
from transformers import SPIECE_UNDERLINE, XLNetTokenizer, XLNetTokenizerFast
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, slow
from ...test_tokenization_common import TokenizerTesterMixin
_lowerCAmelCase : List[Any] = get_tests_dir("fixtures/test_sentencepiece.model")
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( A_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = XLNetTokenizer
SCREAMING_SNAKE_CASE = XLNetTokenizerFast
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = True
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
super().setUp()
# We have a SentencePiece fixture for testing
__a =XLNetTokenizer(snake_case__ , keep_accents=snake_case__ )
tokenizer.sanitize_special_tokens()
tokenizer.save_pretrained(self.tmpdirname )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a ="<s>"
__a =1
self.assertEqual(self.get_tokenizer()._convert_token_to_id(snake_case__ ) , snake_case__ )
self.assertEqual(self.get_tokenizer()._convert_id_to_token(snake_case__ ) , snake_case__ )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =list(self.get_tokenizer().get_vocab().keys() )
self.assertEqual(vocab_keys[0] , '<unk>' )
self.assertEqual(vocab_keys[1] , '<s>' )
self.assertEqual(vocab_keys[-1] , '<eod>' )
self.assertEqual(len(snake_case__ ) , 1006 )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
self.assertEqual(self.get_tokenizer().vocab_size , 1000 )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =XLNetTokenizer(snake_case__ , keep_accents=snake_case__ )
__a =tokenizer.tokenize('This is a test' )
self.assertListEqual(snake_case__ , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(tokenizer.convert_tokens_to_ids(snake_case__ ) , [285, 46, 10, 170, 382] )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
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 =tokenizer.convert_tokens_to_ids(snake_case__ )
self.assertListEqual(snake_case__ , [8, 21, 84, 55, 24, 19, 7, 0, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 0, 4] )
__a =tokenizer.convert_ids_to_tokens(snake_case__ )
self.assertListEqual(
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 __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =XLNetTokenizer(snake_case__ , do_lower_case=snake_case__ )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
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',
'se',
'.',
] , )
self.assertListEqual(tokenizer.tokenize('H\u00E9llo' ) , ['▁he', 'll', 'o'] )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =XLNetTokenizer(snake_case__ , do_lower_case=snake_case__ )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
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',
'se',
'.',
] , )
@slow
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =XLNetTokenizer.from_pretrained('xlnet-base-cased' )
__a =tokenizer.encode('sequence builders' , add_special_tokens=snake_case__ )
__a =tokenizer.encode('multi-sequence build' , add_special_tokens=snake_case__ )
__a =tokenizer.build_inputs_with_special_tokens(snake_case__ )
__a =tokenizer.build_inputs_with_special_tokens(snake_case__ , snake_case__ )
assert encoded_sentence == text + [4, 3]
assert encoded_pair == text + [4] + text_a + [4, 3]
@slow
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a ={"input_ids": [[17, 2_1442, 270, 17, 10, 1_4645, 318, 34, 17, 4546, 3145, 787, 13, 7752, 2_2018, 23, 21, 17, 4546, 3145, 787, 13, 3352, 1_4431, 13, 5500, 11, 1176, 580, 13, 1_6819, 4797, 23, 17, 10, 1_7135, 658, 19, 457, 7932, 13, 184, 19, 3154, 1_7135, 6468, 19, 1404, 1_2269, 19, 4229, 5356, 1_6264, 46, 19, 17, 2_0545, 1_0395, 9, 9, 9, 11, 28, 6421, 9531, 2_0729, 17, 10, 353, 1_7022, 11, 21, 6421, 9531, 1_6949, 17, 10, 1_1509, 753, 11, 33, 95, 2421, 7385, 956, 1_4431, 2626, 25, 842, 7385, 4836, 21, 1429, 2272, 9855, 3120, 161, 2_4738, 19, 1_3203, 658, 218, 787, 21, 430, 1_8482, 847, 2637, 9, 4, 3], [5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 322, 2_2178, 27, 1064, 22, 956, 13, 1_1101, 1429, 5854, 2_4313, 1_8953, 40, 422, 2_4366, 68, 1758, 37, 1_0483, 1_4257, 31, 207, 263, 21, 203, 3773, 25, 71, 9735, 9, 4, 3], [5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 32, 2049, 3442, 17, 1_3894, 3380, 23, 95, 18, 1_7634, 2288, 9, 4, 3]], "token_type_ids": [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2], [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2], [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]], "attention_mask": [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=snake_case__ , model_name='xlnet-base-cased' , revision='c841166438c31ec7ca9a106dee7bb312b73ae511' , )
| 360
|
import os
def UpperCamelCase_( _snake_case : str = "input.txt" ):
"""simple docstring"""
with open(os.path.join(os.path.dirname(_snake_case ) , _snake_case ) ) as input_file:
__a =[
[int(_snake_case ) for element in line.split(',' )]
for line in input_file.readlines()
]
__a =len(_snake_case )
__a =len(matrix[0] )
__a =[[-1 for _ in range(_snake_case )] for _ in range(_snake_case )]
for i in range(_snake_case ):
__a =matrix[i][0]
for j in range(1 , _snake_case ):
for i in range(_snake_case ):
__a =minimal_path_sums[i][j - 1] + matrix[i][j]
for i in range(1 , _snake_case ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i - 1][j] + matrix[i][j] )
for i in range(rows - 2 , -1 , -1 ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i + 1][j] + matrix[i][j] )
return min(minimal_path_sums_row[-1] for minimal_path_sums_row in minimal_path_sums )
if __name__ == "__main__":
print(f'''{solution() = }''')
| 308
| 0
|
import os
import tempfile
import unittest
from pathlib import Path
from transformers import AutoConfig, is_torch_available
from transformers.testing_utils import require_torch, torch_device
if is_torch_available():
from transformers import PyTorchBenchmark, PyTorchBenchmarkArguments
@require_torch
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self , __snake_case ) -> Any:
'''simple docstring'''
for model_result in results.values():
for batch_size, sequence_length in zip(model_result['bs'] , model_result['ss'] ):
__a =model_result['result'][batch_size][sequence_length]
self.assertIsNotNone(_a )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result )
self.check_results_dict_not_empty(results.memory_inference_result )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a ='sgugger/tiny-distilbert-classification'
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , only_pretrain_model=_a , )
__a =PyTorchBenchmark(_a )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result )
self.check_results_dict_not_empty(results.memory_inference_result )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , torchscript=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result )
self.check_results_dict_not_empty(results.memory_inference_result )
@unittest.skipIf(torch_device == 'cpu' , 'Cant do half precision' )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , fpaa=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result )
self.check_results_dict_not_empty(results.memory_inference_result )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =AutoConfig.from_pretrained(_a )
# set architectures equal to `None`
__a =None
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a , configs=[config] )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result )
self.check_results_dict_not_empty(results.memory_inference_result )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_train_result )
self.check_results_dict_not_empty(results.memory_train_result )
@unittest.skipIf(torch_device == 'cpu' , 'Can\'t do half precision' )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , fpaa=_a , multi_process=_a , )
__a =PyTorchBenchmark(_a )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_train_result )
self.check_results_dict_not_empty(results.memory_train_result )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =AutoConfig.from_pretrained(_a )
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a , configs=[config] )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result )
self.check_results_dict_not_empty(results.memory_inference_result )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a ='sshleifer/tinier_bart'
__a =AutoConfig.from_pretrained(_a )
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a , configs=[config] )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result )
self.check_results_dict_not_empty(results.memory_inference_result )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
__a =AutoConfig.from_pretrained(_a )
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a , configs=[config] )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_train_result )
self.check_results_dict_not_empty(results.memory_train_result )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a ='sshleifer/tinier_bart'
__a =AutoConfig.from_pretrained(_a )
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , multi_process=_a , )
__a =PyTorchBenchmark(_a , configs=[config] )
__a =benchmark.run()
self.check_results_dict_not_empty(results.time_train_result )
self.check_results_dict_not_empty(results.memory_train_result )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
with tempfile.TemporaryDirectory() as tmp_dir:
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , save_to_csv=_a , sequence_lengths=[8] , batch_sizes=[1] , inference_time_csv_file=os.path.join(_a , 'inf_time.csv' ) , train_memory_csv_file=os.path.join(_a , 'train_mem.csv' ) , inference_memory_csv_file=os.path.join(_a , 'inf_mem.csv' ) , train_time_csv_file=os.path.join(_a , 'train_time.csv' ) , env_info_csv_file=os.path.join(_a , 'env.csv' ) , multi_process=_a , )
__a =PyTorchBenchmark(_a )
benchmark.run()
self.assertTrue(Path(os.path.join(_a , 'inf_time.csv' ) ).exists() )
self.assertTrue(Path(os.path.join(_a , 'train_time.csv' ) ).exists() )
self.assertTrue(Path(os.path.join(_a , 'inf_mem.csv' ) ).exists() )
self.assertTrue(Path(os.path.join(_a , 'train_mem.csv' ) ).exists() )
self.assertTrue(Path(os.path.join(_a , 'env.csv' ) ).exists() )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a ='sshleifer/tiny-gpt2'
def _check_summary_is_not_empty(__snake_case ):
self.assertTrue(hasattr(_a , 'sequential' ) )
self.assertTrue(hasattr(_a , 'cumulative' ) )
self.assertTrue(hasattr(_a , 'current' ) )
self.assertTrue(hasattr(_a , 'total' ) )
with tempfile.TemporaryDirectory() as tmp_dir:
__a =PyTorchBenchmarkArguments(
models=[MODEL_ID] , training=_a , inference=_a , sequence_lengths=[8] , batch_sizes=[1] , log_filename=os.path.join(_a , 'log.txt' ) , log_print=_a , trace_memory_line_by_line=_a , multi_process=_a , )
__a =PyTorchBenchmark(_a )
__a =benchmark.run()
_check_summary_is_not_empty(result.inference_summary )
_check_summary_is_not_empty(result.train_summary )
self.assertTrue(Path(os.path.join(_a , 'log.txt' ) ).exists() )
| 361
|
import warnings
from ...utils import logging
from .image_processing_mobilevit import MobileViTImageProcessor
_lowerCAmelCase : Any = logging.get_logger(__name__)
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , *__snake_case , **__snake_case ) -> None:
'''simple docstring'''
warnings.warn(
'The class MobileViTFeatureExtractor is deprecated and will be removed in version 5 of Transformers.'
' Please use MobileViTImageProcessor instead.' , __snake_case , )
super().__init__(*__snake_case , **__snake_case )
| 308
| 0
|
import unittest
from transformers import BigBirdConfig, is_flax_available
from transformers.testing_utils import require_flax, slow
from ...test_modeling_flax_common import FlaxModelTesterMixin, ids_tensor, random_attention_mask
if is_flax_available():
import jax
from transformers.models.big_bird.modeling_flax_big_bird import (
FlaxBigBirdForCausalLM,
FlaxBigBirdForMaskedLM,
FlaxBigBirdForMultipleChoice,
FlaxBigBirdForPreTraining,
FlaxBigBirdForQuestionAnswering,
FlaxBigBirdForSequenceClassification,
FlaxBigBirdForTokenClassification,
FlaxBigBirdModel,
)
class __magic_name__ ( unittest.TestCase ):
def __init__( self , __snake_case , __snake_case=2 , __snake_case=56 , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=99 , __snake_case=32 , __snake_case=2 , __snake_case=2 , __snake_case=7 , __snake_case="gelu_new" , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=16 , __snake_case=2 , __snake_case=0.02 , __snake_case=4 , __snake_case="block_sparse" , __snake_case=True , __snake_case=False , __snake_case=2 , __snake_case=3 , ) -> Tuple:
'''simple docstring'''
__a =parent
__a =batch_size
__a =seq_length
__a =is_training
__a =use_attention_mask
__a =use_token_type_ids
__a =use_labels
__a =vocab_size
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_vocab_size
__a =type_sequence_label_size
__a =initializer_range
__a =num_choices
__a =rescale_embeddings
__a =attention_type
__a =use_bias
__a =block_size
__a =num_random_blocks
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a =None
if self.use_attention_mask:
__a =random_attention_mask([self.batch_size, self.seq_length] )
__a =None
if self.use_token_type_ids:
__a =ids_tensor([self.batch_size, self.seq_length] , self.type_vocab_size )
__a =BigBirdConfig(
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=__snake_case , initializer_range=self.initializer_range , attention_type=self.attention_type , block_size=self.block_size , num_random_blocks=self.num_random_blocks , use_bias=self.use_bias , rescale_embeddings=self.rescale_embeddings , )
return config, input_ids, token_type_ids, attention_mask
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
__a =config_and_inputs
__a ={
'input_ids': input_ids,
'token_type_ids': token_type_ids,
'attention_mask': attention_mask,
}
return config, inputs_dict
@require_flax
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
FlaxBigBirdForCausalLM,
FlaxBigBirdModel,
FlaxBigBirdForPreTraining,
FlaxBigBirdForMaskedLM,
FlaxBigBirdForMultipleChoice,
FlaxBigBirdForQuestionAnswering,
FlaxBigBirdForSequenceClassification,
FlaxBigBirdForTokenClassification,
)
if is_flax_available()
else ()
)
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =FlaxBigBirdModelTester(self )
@slow
# copied from `test_modeling_flax_common` because it takes much longer than other models
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
super().test_from_pretrained_save_pretrained()
@slow
# copied from `test_modeling_flax_common` because it takes much longer than other models
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
super().test_from_pretrained_with_no_automatic_init()
@slow
# copied from `test_modeling_flax_common` because it takes much longer than other models
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
super().test_no_automatic_init()
@slow
# copied from `test_modeling_flax_common` because it takes much longer than other models
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
super().test_hidden_states_output()
@slow
def __magic_name__ ( self ) -> str:
'''simple docstring'''
for model_class_name in self.all_model_classes:
__a =model_class_name.from_pretrained('google/bigbird-roberta-base' )
self.assertIsNotNone(__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
if self.test_attn_probs:
super().test_attention_outputs()
@slow
# copied from `test_modeling_flax_common` because it takes much longer than other models
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
with self.subTest(model_class.__name__ ):
__a =self._prepare_for_class(__snake_case , __snake_case )
__a =model_class(__snake_case )
@jax.jit
def model_jitted(__snake_case , __snake_case=None , **__snake_case ):
return model(input_ids=__snake_case , attention_mask=__snake_case , **__snake_case )
with self.subTest('JIT Enabled' ):
__a =model_jitted(**__snake_case ).to_tuple()
with self.subTest('JIT Disabled' ):
with jax.disable_jit():
__a =model_jitted(**__snake_case ).to_tuple()
self.assertEqual(len(__snake_case ) , len(__snake_case ) )
for jitted_output, output in zip(__snake_case , __snake_case ):
self.assertEqual(jitted_output.shape , output.shape )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case=1e-5 , __snake_case="outputs" , __snake_case=None ) -> Union[str, Any]:
'''simple docstring'''
# `bigbird_block_sparse_attention` in `FlaxBigBird` returns `attention_probs = None`, while in PyTorch version,
# an effort was done to return `attention_probs` (yet to be verified).
if name.startswith('outputs.attentions' ):
return
else:
super().check_pt_flax_outputs(__snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case )
| 362
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_sentencepiece_available
_lowerCAmelCase : int = {}
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Dict = ["BartphoTokenizer"]
if TYPE_CHECKING:
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_bartpho import BartphoTokenizer
else:
import sys
_lowerCAmelCase : Optional[Any] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 0
|
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
__a =len(_lowerCamelCase )
for i in range(length - 1 ):
__a =i
for k in range(i + 1 , _lowerCamelCase ):
if collection[k] < collection[least]:
__a =k
if least != i:
__a =(collection[i], collection[least])
return collection
if __name__ == "__main__":
_lowerCAmelCase : Dict = input("Enter numbers separated by a comma:\n").strip()
_lowerCAmelCase : Optional[int] = [int(item) for item in user_input.split(",")]
print(selection_sort(unsorted))
| 363
|
from collections import OrderedDict
from typing import Mapping
from packaging import version
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
_lowerCAmelCase : List[str] = logging.get_logger(__name__)
_lowerCAmelCase : List[Any] = {
"hustvl/yolos-small": "https://huggingface.co/hustvl/yolos-small/resolve/main/config.json",
# See all YOLOS models at https://huggingface.co/models?filter=yolos
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'yolos'
def __init__( self , __snake_case=768 , __snake_case=12 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=[512, 864] , __snake_case=16 , __snake_case=3 , __snake_case=True , __snake_case=100 , __snake_case=True , __snake_case=False , __snake_case=1 , __snake_case=5 , __snake_case=2 , __snake_case=5 , __snake_case=2 , __snake_case=0.1 , **__snake_case , ) -> str:
'''simple docstring'''
super().__init__(**__snake_case )
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =initializer_range
__a =layer_norm_eps
__a =image_size
__a =patch_size
__a =num_channels
__a =qkv_bias
__a =num_detection_tokens
__a =use_mid_position_embeddings
__a =auxiliary_loss
# Hungarian matcher
__a =class_cost
__a =bbox_cost
__a =giou_cost
# Loss coefficients
__a =bbox_loss_coefficient
__a =giou_loss_coefficient
__a =eos_coefficient
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = version.parse('1.11' )
@property
def __magic_name__ ( self ) -> Mapping[str, Mapping[int, str]]:
'''simple docstring'''
return OrderedDict(
[
('pixel_values', {0: 'batch', 1: 'num_channels', 2: 'height', 3: 'width'}),
] )
@property
def __magic_name__ ( self ) -> float:
'''simple docstring'''
return 1e-4
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return 12
| 308
| 0
|
_lowerCAmelCase : Optional[int] = "\n# Transformers 설치 방법\n! pip install transformers datasets\n# 마지막 릴리스 대신 소스에서 설치하려면, 위 명령을 주석으로 바꾸고 아래 명령을 해제하세요.\n# ! pip install git+https://github.com/huggingface/transformers.git\n"
_lowerCAmelCase : Tuple = [{"type": "code", "content": INSTALL_CONTENT}]
_lowerCAmelCase : Optional[Any] = {
"{processor_class}": "FakeProcessorClass",
"{model_class}": "FakeModelClass",
"{object_class}": "FakeObjectClass",
}
| 364
|
import argparse
import torch
from datasets import load_dataset
from donut import DonutModel
from transformers import (
DonutImageProcessor,
DonutProcessor,
DonutSwinConfig,
DonutSwinModel,
MBartConfig,
MBartForCausalLM,
VisionEncoderDecoderModel,
XLMRobertaTokenizerFast,
)
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =model.config
__a =DonutSwinConfig(
image_size=original_config.input_size , patch_size=4 , depths=original_config.encoder_layer , num_heads=[4, 8, 16, 32] , window_size=original_config.window_size , embed_dim=128 , )
__a =MBartConfig(
is_decoder=_snake_case , is_encoder_decoder=_snake_case , add_cross_attention=_snake_case , decoder_layers=original_config.decoder_layer , max_position_embeddings=original_config.max_position_embeddings , vocab_size=len(
model.decoder.tokenizer ) , scale_embedding=_snake_case , add_final_layer_norm=_snake_case , )
return encoder_config, decoder_config
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
if "encoder.model" in name:
__a =name.replace('encoder.model' , 'encoder' )
if "decoder.model" in name:
__a =name.replace('decoder.model' , 'decoder' )
if "patch_embed.proj" in name:
__a =name.replace('patch_embed.proj' , 'embeddings.patch_embeddings.projection' )
if "patch_embed.norm" in name:
__a =name.replace('patch_embed.norm' , 'embeddings.norm' )
if name.startswith('encoder' ):
if "layers" in name:
__a ='encoder.' + name
if "attn.proj" in name:
__a =name.replace('attn.proj' , 'attention.output.dense' )
if "attn" in name and "mask" not in name:
__a =name.replace('attn' , 'attention.self' )
if "norm1" in name:
__a =name.replace('norm1' , 'layernorm_before' )
if "norm2" in name:
__a =name.replace('norm2' , 'layernorm_after' )
if "mlp.fc1" in name:
__a =name.replace('mlp.fc1' , 'intermediate.dense' )
if "mlp.fc2" in name:
__a =name.replace('mlp.fc2' , 'output.dense' )
if name == "encoder.norm.weight":
__a ='encoder.layernorm.weight'
if name == "encoder.norm.bias":
__a ='encoder.layernorm.bias'
return name
def UpperCamelCase_( _snake_case : Tuple , _snake_case : str ):
"""simple docstring"""
for key in orig_state_dict.copy().keys():
__a =orig_state_dict.pop(_snake_case )
if "qkv" in key:
__a =key.split('.' )
__a =int(key_split[3] )
__a =int(key_split[5] )
__a =model.encoder.encoder.layers[layer_num].blocks[block_num].attention.self.all_head_size
if "weight" in key:
__a =val[:dim, :]
__a =val[dim : dim * 2, :]
__a =val[-dim:, :]
else:
__a =val[:dim]
__a =val[dim : dim * 2]
__a =val[-dim:]
elif "attn_mask" in key or key in ["encoder.model.norm.weight", "encoder.model.norm.bias"]:
# HuggingFace implementation doesn't use attn_mask buffer
# and model doesn't use final LayerNorms for the encoder
pass
else:
__a =val
return orig_state_dict
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Union[str, Any]=None , _snake_case : List[Any]=False ):
"""simple docstring"""
__a =DonutModel.from_pretrained(_snake_case ).eval()
# load HuggingFace model
__a , __a =get_configs(_snake_case )
__a =DonutSwinModel(_snake_case )
__a =MBartForCausalLM(_snake_case )
__a =VisionEncoderDecoderModel(encoder=_snake_case , decoder=_snake_case )
model.eval()
__a =original_model.state_dict()
__a =convert_state_dict(_snake_case , _snake_case )
model.load_state_dict(_snake_case )
# verify results on scanned document
__a =load_dataset('hf-internal-testing/example-documents' )
__a =dataset['test'][0]['image'].convert('RGB' )
__a =XLMRobertaTokenizerFast.from_pretrained(_snake_case , from_slow=_snake_case )
__a =DonutImageProcessor(
do_align_long_axis=original_model.config.align_long_axis , size=original_model.config.input_size[::-1] )
__a =DonutProcessor(_snake_case , _snake_case )
__a =processor(_snake_case , return_tensors='pt' ).pixel_values
if model_name == "naver-clova-ix/donut-base-finetuned-docvqa":
__a ='<s_docvqa><s_question>{user_input}</s_question><s_answer>'
__a ='When is the coffee break?'
__a =task_prompt.replace('{user_input}' , _snake_case )
elif model_name == "naver-clova-ix/donut-base-finetuned-rvlcdip":
__a ='<s_rvlcdip>'
elif model_name in [
"naver-clova-ix/donut-base-finetuned-cord-v1",
"naver-clova-ix/donut-base-finetuned-cord-v1-2560",
]:
__a ='<s_cord>'
elif model_name == "naver-clova-ix/donut-base-finetuned-cord-v2":
__a ='s_cord-v2>'
elif model_name == "naver-clova-ix/donut-base-finetuned-zhtrainticket":
__a ='<s_zhtrainticket>'
elif model_name in ["naver-clova-ix/donut-proto", "naver-clova-ix/donut-base"]:
# use a random prompt
__a ='hello world'
else:
raise ValueError('Model name not supported' )
__a =original_model.decoder.tokenizer(_snake_case , add_special_tokens=_snake_case , return_tensors='pt' )[
'input_ids'
]
__a =original_model.encoder.model.patch_embed(_snake_case )
__a , __a =model.encoder.embeddings(_snake_case )
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
# verify encoder hidden states
__a =original_model.encoder(_snake_case )
__a =model.encoder(_snake_case ).last_hidden_state
assert torch.allclose(_snake_case , _snake_case , atol=1e-2 )
# verify decoder hidden states
__a =original_model(_snake_case , _snake_case , _snake_case ).logits
__a =model(_snake_case , decoder_input_ids=_snake_case ).logits
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
print('Looks ok!' )
if pytorch_dump_folder_path is not None:
print(F'Saving model and processor to {pytorch_dump_folder_path}' )
model.save_pretrained(_snake_case )
processor.save_pretrained(_snake_case )
if push_to_hub:
model.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
processor.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
if __name__ == "__main__":
_lowerCAmelCase : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_name",
default="naver-clova-ix/donut-base-finetuned-docvqa",
required=False,
type=str,
help="Name of the original model you'd like to convert.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
required=False,
type=str,
help="Path to the output PyTorch model directory.",
)
parser.add_argument(
"--push_to_hub",
action="store_true",
help="Whether or not to push the converted model and processor to the 🤗 hub.",
)
_lowerCAmelCase : List[Any] = parser.parse_args()
convert_donut_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 308
| 0
|
import os
import unittest
from huggingface_hub.utils import are_progress_bars_disabled
import transformers.models.bart.tokenization_bart
from transformers import logging
from transformers.testing_utils import CaptureLogger, mockenv, mockenv_context
from transformers.utils.logging import disable_progress_bar, enable_progress_bar
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =logging.get_logger()
# the current default level is logging.WARNING
__a =logging.get_verbosity()
logging.set_verbosity_error()
self.assertEqual(logger.getEffectiveLevel() , logging.get_verbosity() )
logging.set_verbosity_warning()
self.assertEqual(logger.getEffectiveLevel() , logging.get_verbosity() )
logging.set_verbosity_info()
self.assertEqual(logger.getEffectiveLevel() , logging.get_verbosity() )
logging.set_verbosity_debug()
self.assertEqual(logger.getEffectiveLevel() , logging.get_verbosity() )
# restore to the original level
logging.set_verbosity(_a )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =logging.get_verbosity()
__a =logging.get_logger('transformers.models.bart.tokenization_bart' )
__a ='''Testing 1, 2, 3'''
# should be able to log warnings (if default settings weren't overridden by `pytest --log-level-all`)
if level_origin <= logging.WARNING:
with CaptureLogger(_a ) as cl:
logger.warning(_a )
self.assertEqual(cl.out , msg + '\n' )
# this is setting the level for all of `transformers.*` loggers
logging.set_verbosity_error()
# should not be able to log warnings
with CaptureLogger(_a ) as cl:
logger.warning(_a )
self.assertEqual(cl.out , '' )
# should be able to log warnings again
logging.set_verbosity_warning()
with CaptureLogger(_a ) as cl:
logger.warning(_a )
self.assertEqual(cl.out , msg + '\n' )
# restore to the original level
logging.set_verbosity(_a )
@mockenv(TRANSFORMERS_VERBOSITY='error' )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
# reset for the env var to take effect, next time some logger call is made
transformers.utils.logging._reset_library_root_logger()
# this action activates the env var
__a =logging.get_logger('transformers.models.bart.tokenization_bart' )
__a =os.getenv('TRANSFORMERS_VERBOSITY' , _a )
__a =logging.log_levels[env_level_str]
__a =logging.get_verbosity()
self.assertEqual(
_a , _a , f'TRANSFORMERS_VERBOSITY={env_level_str}/{env_level}, but internal verbosity is {current_level}' , )
# restore to the original level
__a =''''''
transformers.utils.logging._reset_library_root_logger()
@mockenv(TRANSFORMERS_VERBOSITY='super-error' )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
# reset for the env var to take effect, next time some logger call is made
transformers.utils.logging._reset_library_root_logger()
__a =logging.logging.getLogger()
with CaptureLogger(_a ) as cl:
# this action activates the env var
logging.get_logger('transformers.models.bart.tokenization_bart' )
self.assertIn('Unknown option TRANSFORMERS_VERBOSITY=super-error' , cl.out )
# no need to restore as nothing was changed
def __magic_name__ ( self ) -> int:
'''simple docstring'''
# testing `logger.warning_advice()`
transformers.utils.logging._reset_library_root_logger()
__a =logging.get_logger('transformers.models.bart.tokenization_bart' )
__a ='''Testing 1, 2, 3'''
with mockenv_context(TRANSFORMERS_NO_ADVISORY_WARNINGS='1' ):
# nothing should be logged as env var disables this method
with CaptureLogger(_a ) as cl:
logger.warning_advice(_a )
self.assertEqual(cl.out , '' )
with mockenv_context(TRANSFORMERS_NO_ADVISORY_WARNINGS='' ):
# should log normally as TRANSFORMERS_NO_ADVISORY_WARNINGS is unset
with CaptureLogger(_a ) as cl:
logger.warning_advice(_a )
self.assertEqual(cl.out , msg + '\n' )
def UpperCamelCase_( ):
"""simple docstring"""
disable_progress_bar()
assert are_progress_bars_disabled()
enable_progress_bar()
assert not are_progress_bars_disabled()
| 365
|
import json
import os
import unittest
from transformers.models.blenderbot_small.tokenization_blenderbot_small import (
VOCAB_FILES_NAMES,
BlenderbotSmallTokenizer,
)
from ...test_tokenization_common import TokenizerTesterMixin
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = BlenderbotSmallTokenizer
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
super().setUp()
__a =['__start__', 'adapt', 'act', 'ap@@', 'te', '__end__', '__unk__']
__a =dict(zip(__snake_case , range(len(__snake_case ) ) ) )
__a =['#version: 0.2', 'a p', 't e</w>', 'ap t</w>', 'a d', 'ad apt</w>', 'a c', 'ac t</w>', '']
__a ={'unk_token': '__unk__', 'bos_token': '__start__', 'eos_token': '__end__'}
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(__snake_case ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(__snake_case ) )
def __magic_name__ ( self , **__snake_case ) -> Any:
'''simple docstring'''
kwargs.update(self.special_tokens_map )
return BlenderbotSmallTokenizer.from_pretrained(self.tmpdirname , **__snake_case )
def __magic_name__ ( self , __snake_case ) -> List[Any]:
'''simple docstring'''
__a ='adapt act apte'
__a ='adapt act apte'
return input_text, output_text
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
__a ='adapt act apte'
__a =['adapt', 'act', 'ap@@', 'te']
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
__a =[tokenizer.bos_token] + tokens + [tokenizer.eos_token]
__a =[0, 1, 2, 3, 4, 5]
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
assert tok('sam' ).input_ids == [1384]
__a ='I am a small frog.'
__a =tok([src_text] , padding=__snake_case , truncation=__snake_case )['input_ids']
__a =tok.batch_decode(__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case )[0]
assert src_text != decoded # I wish it did!
assert decoded == "i am a small frog ."
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
__a ='I am a small frog .'
__a ='.'
__a =tok(__snake_case )['input_ids']
__a =tok(__snake_case )['input_ids']
assert encoded[-1] == encoded_dot[0]
| 308
| 0
|
import datasets
_lowerCAmelCase : Optional[Any] = "\\n@InProceedings{conneau2018xnli,\n author = \"Conneau, Alexis\n and Rinott, Ruty\n and Lample, Guillaume\n and Williams, Adina\n and Bowman, Samuel R.\n and Schwenk, Holger\n and Stoyanov, Veselin\",\n title = \"XNLI: Evaluating Cross-lingual Sentence Representations\",\n booktitle = \"Proceedings of the 2018 Conference on Empirical Methods\n in Natural Language Processing\",\n year = \"2018\",\n publisher = \"Association for Computational Linguistics\",\n location = \"Brussels, Belgium\",\n}\n"
_lowerCAmelCase : Dict = "\\nXNLI is a subset of a few thousand examples from MNLI which has been translated\ninto a 14 different languages (some low-ish resource). As with MNLI, the goal is\nto predict textual entailment (does sentence A imply/contradict/neither sentence\nB) and is a classification task (given two sentences, predict one of three\nlabels).\n"
_lowerCAmelCase : Optional[Any] = "\nComputes XNLI score which is just simple accuracy.\nArgs:\n predictions: Predicted labels.\n references: Ground truth labels.\nReturns:\n \'accuracy\': accuracy\nExamples:\n\n >>> predictions = [0, 1]\n >>> references = [0, 1]\n >>> xnli_metric = datasets.load_metric(\"xnli\")\n >>> results = xnli_metric.compute(predictions=predictions, references=references)\n >>> print(results)\n {\'accuracy\': 1.0}\n"
def UpperCamelCase_( _snake_case : Union[str, Any] , _snake_case : Any ):
"""simple docstring"""
return (preds == labels).mean()
@datasets.utils.file_utils.add_start_docstrings(_DESCRIPTION , _KWARGS_DESCRIPTION )
class __magic_name__ ( datasets.Metric ):
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
return datasets.MetricInfo(
description=_DESCRIPTION , citation=_CITATION , inputs_description=_KWARGS_DESCRIPTION , features=datasets.Features(
{
'predictions': datasets.Value('int64' if self.config_name != 'sts-b' else 'float32' ),
'references': datasets.Value('int64' if self.config_name != 'sts-b' else 'float32' ),
} ) , codebase_urls=[] , reference_urls=[] , format='numpy' , )
def __magic_name__ ( self , __snake_case , __snake_case ) -> Optional[Any]:
'''simple docstring'''
return {"accuracy": simple_accuracy(__UpperCAmelCase , __UpperCAmelCase )}
| 366
|
import unittest
from transformers import load_tool
from transformers.utils import is_torch_available
if is_torch_available():
import torch
from transformers.testing_utils import require_torch
from .test_tools_common import ToolTesterMixin
@require_torch
class __magic_name__ ( unittest.TestCase , lowerCAmelCase_ ):
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =load_tool('text-to-speech' )
self.tool.setup()
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
| 308
| 0
|
import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers.testing_utils import require_vision
from transformers.utils import is_vision_available
if is_vision_available():
from PIL import Image
from transformers import AutoProcessor, BertTokenizer, BlipImageProcessor, BlipProcessor, PreTrainedTokenizerFast
@require_vision
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =tempfile.mkdtemp()
__a =BlipImageProcessor()
__a =BertTokenizer.from_pretrained('hf-internal-testing/tiny-random-BertModel' )
__a =BlipProcessor(lowerCAmelCase__ , lowerCAmelCase__ )
processor.save_pretrained(self.tmpdirname )
def __magic_name__ ( self , **__snake_case ) -> Union[str, Any]:
'''simple docstring'''
return AutoProcessor.from_pretrained(self.tmpdirname , **lowerCAmelCase__ ).tokenizer
def __magic_name__ ( self , **__snake_case ) -> List[Any]:
'''simple docstring'''
return AutoProcessor.from_pretrained(self.tmpdirname , **lowerCAmelCase__ ).image_processor
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
shutil.rmtree(self.tmpdirname )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =[np.random.randint(255 , size=(3, 30, 400) , dtype=np.uinta )]
__a =[Image.fromarray(np.moveaxis(lowerCAmelCase__ , 0 , -1 ) ) for x in image_inputs]
return image_inputs
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =BlipProcessor(tokenizer=self.get_tokenizer() , image_processor=self.get_image_processor() )
processor.save_pretrained(self.tmpdirname )
__a =self.get_tokenizer(bos_token='(BOS)' , eos_token='(EOS)' )
__a =self.get_image_processor(do_normalize=lowerCAmelCase__ , padding_value=1.0 )
__a =BlipProcessor.from_pretrained(
self.tmpdirname , bos_token='(BOS)' , eos_token='(EOS)' , do_normalize=lowerCAmelCase__ , padding_value=1.0 )
self.assertEqual(processor.tokenizer.get_vocab() , tokenizer_add_kwargs.get_vocab() )
self.assertIsInstance(processor.tokenizer , lowerCAmelCase__ )
self.assertEqual(processor.image_processor.to_json_string() , image_processor_add_kwargs.to_json_string() )
self.assertIsInstance(processor.image_processor , lowerCAmelCase__ )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.get_image_processor()
__a =self.get_tokenizer()
__a =BlipProcessor(tokenizer=lowerCAmelCase__ , image_processor=lowerCAmelCase__ )
__a =self.prepare_image_inputs()
__a =image_processor(lowerCAmelCase__ , return_tensors='np' )
__a =processor(images=lowerCAmelCase__ , return_tensors='np' )
for key in input_feat_extract.keys():
self.assertAlmostEqual(input_feat_extract[key].sum() , input_processor[key].sum() , delta=1e-2 )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.get_image_processor()
__a =self.get_tokenizer()
__a =BlipProcessor(tokenizer=lowerCAmelCase__ , image_processor=lowerCAmelCase__ )
__a ="lower newer"
__a =processor(text=lowerCAmelCase__ )
__a =tokenizer(lowerCAmelCase__ , return_token_type_ids=lowerCAmelCase__ )
for key in encoded_tok.keys():
self.assertListEqual(encoded_tok[key] , encoded_processor[key] )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =self.get_image_processor()
__a =self.get_tokenizer()
__a =BlipProcessor(tokenizer=lowerCAmelCase__ , image_processor=lowerCAmelCase__ )
__a ="lower newer"
__a =self.prepare_image_inputs()
__a =processor(text=lowerCAmelCase__ , images=lowerCAmelCase__ )
self.assertListEqual(list(inputs.keys() ) , ['pixel_values', 'input_ids', 'attention_mask'] )
# test if it raises when no input is passed
with pytest.raises(lowerCAmelCase__ ):
processor()
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =self.get_image_processor()
__a =self.get_tokenizer()
__a =BlipProcessor(tokenizer=lowerCAmelCase__ , image_processor=lowerCAmelCase__ )
__a =[[1, 4, 5, 8, 1, 0, 8], [3, 4, 3, 1, 1, 8, 9]]
__a =processor.batch_decode(lowerCAmelCase__ )
__a =tokenizer.batch_decode(lowerCAmelCase__ )
self.assertListEqual(lowerCAmelCase__ , lowerCAmelCase__ )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.get_image_processor()
__a =self.get_tokenizer()
__a =BlipProcessor(tokenizer=lowerCAmelCase__ , image_processor=lowerCAmelCase__ )
__a ="lower newer"
__a =self.prepare_image_inputs()
__a =processor(text=lowerCAmelCase__ , images=lowerCAmelCase__ )
# For now the processor supports only ['pixel_values', 'input_ids', 'attention_mask']
self.assertListEqual(list(inputs.keys() ) , ['pixel_values', 'input_ids', 'attention_mask'] )
| 367
|
import flax.linen as nn
import jax.numpy as jnp
from .attention_flax import FlaxTransformeraDModel
from .resnet_flax import FlaxDownsampleaD, FlaxResnetBlockaD, FlaxUpsampleaD
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> Optional[Any]:
'''simple docstring'''
__a =()
for resnet, attn in zip(self.resnets , self.attentions ):
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case=True ) -> Optional[int]:
'''simple docstring'''
__a =()
for resnet in self.resnets:
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet, attn in zip(self.resnets , self.attentions ):
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet in self.resnets:
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# there is always at least one resnet
__a =[
FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
]
__a =[]
for _ in range(self.num_layers ):
__a =FlaxTransformeraDModel(
in_channels=self.in_channels , n_heads=self.num_attention_heads , d_head=self.in_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
__a =attentions
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[str]:
'''simple docstring'''
__a =self.resnets[0](__snake_case , __snake_case )
for attn, resnet in zip(self.attentions , self.resnets[1:] ):
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
return hidden_states
| 308
| 0
|
import contextlib
import copy
import random
from typing import Any, Dict, Iterable, Optional, Union
import numpy as np
import torch
from .utils import deprecate, is_transformers_available
if is_transformers_available():
import transformers
def UpperCamelCase_( _snake_case : int ):
"""simple docstring"""
random.seed(_UpperCamelCase )
np.random.seed(_UpperCamelCase )
torch.manual_seed(_UpperCamelCase )
torch.cuda.manual_seed_all(_UpperCamelCase )
# ^^ safe to call this function even if cuda is not available
class __magic_name__ :
def __init__( self , __snake_case , __snake_case = 0.9999 , __snake_case = 0.0 , __snake_case = 0 , __snake_case = False , __snake_case = 1.0 , __snake_case = 2 / 3 , __snake_case = None , __snake_case = None , **__snake_case , ) -> Optional[Any]:
'''simple docstring'''
if isinstance(_SCREAMING_SNAKE_CASE , torch.nn.Module ):
__a =(
'Passing a `torch.nn.Module` to `ExponentialMovingAverage` is deprecated. '
'Please pass the parameters of the module instead.'
)
deprecate(
'passing a `torch.nn.Module` to `ExponentialMovingAverage`' , '1.0.0' , _SCREAMING_SNAKE_CASE , standard_warn=_SCREAMING_SNAKE_CASE , )
__a =parameters.parameters()
# set use_ema_warmup to True if a torch.nn.Module is passed for backwards compatibility
__a =True
if kwargs.get('max_value' , _SCREAMING_SNAKE_CASE ) is not None:
__a ='The `max_value` argument is deprecated. Please use `decay` instead.'
deprecate('max_value' , '1.0.0' , _SCREAMING_SNAKE_CASE , standard_warn=_SCREAMING_SNAKE_CASE )
__a =kwargs['max_value']
if kwargs.get('min_value' , _SCREAMING_SNAKE_CASE ) is not None:
__a ='The `min_value` argument is deprecated. Please use `min_decay` instead.'
deprecate('min_value' , '1.0.0' , _SCREAMING_SNAKE_CASE , standard_warn=_SCREAMING_SNAKE_CASE )
__a =kwargs['min_value']
__a =list(_SCREAMING_SNAKE_CASE )
__a =[p.clone().detach() for p in parameters]
if kwargs.get('device' , _SCREAMING_SNAKE_CASE ) is not None:
__a ='The `device` argument is deprecated. Please use `to` instead.'
deprecate('device' , '1.0.0' , _SCREAMING_SNAKE_CASE , standard_warn=_SCREAMING_SNAKE_CASE )
self.to(device=kwargs['device'] )
__a =None
__a =decay
__a =min_decay
__a =update_after_step
__a =use_ema_warmup
__a =inv_gamma
__a =power
__a =0
__a =None # set in `step()`
__a =model_cls
__a =model_config
@classmethod
def __magic_name__ ( cls , __snake_case , __snake_case ) -> "EMAModel":
'''simple docstring'''
__a , __a =model_cls.load_config(_SCREAMING_SNAKE_CASE , return_unused_kwargs=_SCREAMING_SNAKE_CASE )
__a =model_cls.from_pretrained(_SCREAMING_SNAKE_CASE )
__a =cls(model.parameters() , model_cls=_SCREAMING_SNAKE_CASE , model_config=model.config )
ema_model.load_state_dict(_SCREAMING_SNAKE_CASE )
return ema_model
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
if self.model_cls is None:
raise ValueError('`save_pretrained` can only be used if `model_cls` was defined at __init__.' )
if self.model_config is None:
raise ValueError('`save_pretrained` can only be used if `model_config` was defined at __init__.' )
__a =self.model_cls.from_config(self.model_config )
__a =self.state_dict()
state_dict.pop('shadow_params' , _SCREAMING_SNAKE_CASE )
model.register_to_config(**_SCREAMING_SNAKE_CASE )
self.copy_to(model.parameters() )
model.save_pretrained(_SCREAMING_SNAKE_CASE )
def __magic_name__ ( self , __snake_case ) -> float:
'''simple docstring'''
__a =max(0 , optimization_step - self.update_after_step - 1 )
if step <= 0:
return 0.0
if self.use_ema_warmup:
__a =1 - (1 + step / self.inv_gamma) ** -self.power
else:
__a =(1 + step) / (10 + step)
__a =min(_SCREAMING_SNAKE_CASE , self.decay )
# make sure decay is not smaller than min_decay
__a =max(_SCREAMING_SNAKE_CASE , self.min_decay )
return cur_decay_value
@torch.no_grad()
def __magic_name__ ( self , __snake_case ) -> Tuple:
'''simple docstring'''
if isinstance(_SCREAMING_SNAKE_CASE , torch.nn.Module ):
__a =(
'Passing a `torch.nn.Module` to `ExponentialMovingAverage.step` is deprecated. '
'Please pass the parameters of the module instead.'
)
deprecate(
'passing a `torch.nn.Module` to `ExponentialMovingAverage.step`' , '1.0.0' , _SCREAMING_SNAKE_CASE , standard_warn=_SCREAMING_SNAKE_CASE , )
__a =parameters.parameters()
__a =list(_SCREAMING_SNAKE_CASE )
self.optimization_step += 1
# Compute the decay factor for the exponential moving average.
__a =self.get_decay(self.optimization_step )
__a =decay
__a =1 - decay
__a =contextlib.nullcontext
if is_transformers_available() and transformers.deepspeed.is_deepspeed_zeroa_enabled():
import deepspeed
for s_param, param in zip(self.shadow_params , _SCREAMING_SNAKE_CASE ):
if is_transformers_available() and transformers.deepspeed.is_deepspeed_zeroa_enabled():
__a =deepspeed.zero.GatheredParameters(_SCREAMING_SNAKE_CASE , modifier_rank=_SCREAMING_SNAKE_CASE )
with context_manager():
if param.requires_grad:
s_param.sub_(one_minus_decay * (s_param - param) )
else:
s_param.copy_(_SCREAMING_SNAKE_CASE )
def __magic_name__ ( self , __snake_case ) -> None:
'''simple docstring'''
__a =list(_SCREAMING_SNAKE_CASE )
for s_param, param in zip(self.shadow_params , _SCREAMING_SNAKE_CASE ):
param.data.copy_(s_param.to(param.device ).data )
def __magic_name__ ( self , __snake_case=None , __snake_case=None ) -> None:
'''simple docstring'''
__a =[
p.to(device=_SCREAMING_SNAKE_CASE , dtype=_SCREAMING_SNAKE_CASE ) if p.is_floating_point() else p.to(device=_SCREAMING_SNAKE_CASE )
for p in self.shadow_params
]
def __magic_name__ ( self ) -> dict:
'''simple docstring'''
return {
"decay": self.decay,
"min_decay": self.min_decay,
"optimization_step": self.optimization_step,
"update_after_step": self.update_after_step,
"use_ema_warmup": self.use_ema_warmup,
"inv_gamma": self.inv_gamma,
"power": self.power,
"shadow_params": self.shadow_params,
}
def __magic_name__ ( self , __snake_case ) -> None:
'''simple docstring'''
__a =[param.detach().cpu().clone() for param in parameters]
def __magic_name__ ( self , __snake_case ) -> None:
'''simple docstring'''
if self.temp_stored_params is None:
raise RuntimeError('This ExponentialMovingAverage has no `store()`ed weights ' 'to `restore()`' )
for c_param, param in zip(self.temp_stored_params , _SCREAMING_SNAKE_CASE ):
param.data.copy_(c_param.data )
# Better memory-wise.
__a =None
def __magic_name__ ( self , __snake_case ) -> None:
'''simple docstring'''
__a =copy.deepcopy(_SCREAMING_SNAKE_CASE )
__a =state_dict.get('decay' , self.decay )
if self.decay < 0.0 or self.decay > 1.0:
raise ValueError('Decay must be between 0 and 1' )
__a =state_dict.get('min_decay' , self.min_decay )
if not isinstance(self.min_decay , _SCREAMING_SNAKE_CASE ):
raise ValueError('Invalid min_decay' )
__a =state_dict.get('optimization_step' , self.optimization_step )
if not isinstance(self.optimization_step , _SCREAMING_SNAKE_CASE ):
raise ValueError('Invalid optimization_step' )
__a =state_dict.get('update_after_step' , self.update_after_step )
if not isinstance(self.update_after_step , _SCREAMING_SNAKE_CASE ):
raise ValueError('Invalid update_after_step' )
__a =state_dict.get('use_ema_warmup' , self.use_ema_warmup )
if not isinstance(self.use_ema_warmup , _SCREAMING_SNAKE_CASE ):
raise ValueError('Invalid use_ema_warmup' )
__a =state_dict.get('inv_gamma' , self.inv_gamma )
if not isinstance(self.inv_gamma , (float, int) ):
raise ValueError('Invalid inv_gamma' )
__a =state_dict.get('power' , self.power )
if not isinstance(self.power , (float, int) ):
raise ValueError('Invalid power' )
__a =state_dict.get('shadow_params' , _SCREAMING_SNAKE_CASE )
if shadow_params is not None:
__a =shadow_params
if not isinstance(self.shadow_params , _SCREAMING_SNAKE_CASE ):
raise ValueError('shadow_params must be a list' )
if not all(isinstance(_SCREAMING_SNAKE_CASE , torch.Tensor ) for p in self.shadow_params ):
raise ValueError('shadow_params must all be Tensors' )
| 368
|
import json
import os
import subprocess
import unittest
from ast import literal_eval
import pytest
from parameterized import parameterized, parameterized_class
from . import is_sagemaker_available
if is_sagemaker_available():
from sagemaker import Session, TrainingJobAnalytics
from sagemaker.huggingface import HuggingFace
@pytest.mark.skipif(
literal_eval(os.getenv('TEST_SAGEMAKER' , 'False' ) ) is not True , reason='Skipping test because should only be run when releasing minor transformers version' , )
@pytest.mark.usefixtures('sm_env' )
@parameterized_class(
[
{
'framework': 'pytorch',
'script': 'run_glue.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_5_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'pytorch',
'script': 'run_ddp.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'tensorflow',
'script': 'run_tf_dist.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.6, 'eval_loss': 0.7},
},
] )
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.framework == "pytorch":
subprocess.run(
f'cp ./examples/pytorch/text-classification/run_glue.py {self.env.test_path}/run_glue.py'.split() , encoding='utf-8' , check=__snake_case , )
assert hasattr(self , 'env' )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
__a =f'{self.env.base_job_name}-{instance_count}-{"ddp" if "ddp" in self.script else "smd"}'
# distributed data settings
__a ={'smdistributed': {'dataparallel': {'enabled': True}}} if self.script != 'run_ddp.py' else None
# creates estimator
return HuggingFace(
entry_point=self.script , source_dir=self.env.test_path , role=self.env.role , image_uri=self.env.image_uri , base_job_name=__snake_case , instance_count=__snake_case , instance_type=self.instance_type , debugger_hook_config=__snake_case , hyperparameters={**self.env.distributed_hyperparameters, 'model_name_or_path': self.model_name_or_path} , metric_definitions=self.env.metric_definitions , distribution=__snake_case , py_version='py36' , )
def __magic_name__ ( self , __snake_case ) -> Optional[Any]:
'''simple docstring'''
TrainingJobAnalytics(__snake_case ).export_csv(f'{self.env.test_path}/{job_name}_metrics.csv' )
@parameterized.expand([(2,)] )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
# create estimator
__a =self.create_estimator(__snake_case )
# run training
estimator.fit()
# result dataframe
__a =TrainingJobAnalytics(estimator.latest_training_job.name ).dataframe()
# extract kpis
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_accuracy']['value'] )
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_loss']['value'] )
# get train time from SageMaker job, this includes starting, preprocessing, stopping
__a =(
Session().describe_training_job(estimator.latest_training_job.name ).get('TrainingTimeInSeconds' , 99_9999 )
)
# assert kpis
assert train_runtime <= self.results["train_runtime"]
assert all(t >= self.results['eval_accuracy'] for t in eval_accuracy )
assert all(t <= self.results['eval_loss'] for t in eval_loss )
# dump tests result into json file to share in PR
with open(f'{estimator.latest_training_job.name}.json' , 'w' ) as outfile:
json.dump({'train_time': train_runtime, 'eval_accuracy': eval_accuracy, 'eval_loss': eval_loss} , __snake_case )
| 308
| 0
|
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
_lowerCAmelCase : Tuple = '''▁'''
_lowerCAmelCase : List[Any] = {'''vocab_file''': '''spiece.model'''}
_lowerCAmelCase : Optional[Any] = {
'''vocab_file''': {'''google/pegasus-xsum''': '''https://huggingface.co/google/pegasus-xsum/resolve/main/spiece.model'''}
}
_lowerCAmelCase : str = {
'''google/pegasus-xsum''': 512,
}
_lowerCAmelCase : str = logging.get_logger(__name__)
class __magic_name__ ( a__ ):
SCREAMING_SNAKE_CASE = VOCAB_FILES_NAMES
SCREAMING_SNAKE_CASE = VOCAB_FILES_NAMES
SCREAMING_SNAKE_CASE = PRETRAINED_VOCAB_FILES_MAP
SCREAMING_SNAKE_CASE = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
SCREAMING_SNAKE_CASE = ['input_ids', 'attention_mask']
def __init__( self , __snake_case , __snake_case="<pad>" , __snake_case="</s>" , __snake_case="<unk>" , __snake_case="<mask_2>" , __snake_case="<mask_1>" , __snake_case=None , __snake_case=103 , __snake_case = None , **__snake_case , ) -> None:
'''simple docstring'''
__a =offset
if additional_special_tokens is not None:
if not isinstance(_lowerCamelCase , _lowerCamelCase ):
raise TypeError(
f'additional_special_tokens should be of type {type(_lowerCamelCase )}, but is'
f' {type(_lowerCamelCase )}' )
__a =(
([mask_token_sent] + additional_special_tokens)
if mask_token_sent not in additional_special_tokens and mask_token_sent is not None
else additional_special_tokens
)
# fill additional tokens with ..., <unk_token_102> in case not all additional tokens are already taken
additional_special_tokens_extended += [
f'<unk_{i}>' for i in range(len(_lowerCamelCase ) , self.offset - 1 )
]
if len(set(_lowerCamelCase ) ) != len(_lowerCamelCase ):
raise ValueError(
'Please make sure that the provided additional_special_tokens do not contain an incorrectly'
f' shifted list of <unk_x> tokens. Found {additional_special_tokens_extended}.' )
__a =additional_special_tokens_extended
else:
__a =[mask_token_sent] if mask_token_sent is not None else []
additional_special_tokens += [f'<unk_{i}>' for i in range(2 , self.offset )]
__a ={} if sp_model_kwargs is None else sp_model_kwargs
super().__init__(
eos_token=_lowerCamelCase , unk_token=_lowerCamelCase , mask_token=_lowerCamelCase , pad_token=_lowerCamelCase , mask_token_sent=_lowerCamelCase , offset=_lowerCamelCase , additional_special_tokens=_lowerCamelCase , sp_model_kwargs=self.sp_model_kwargs , **_lowerCamelCase , )
__a =mask_token_sent
__a =vocab_file
__a =spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(_lowerCamelCase )
# add special tokens to encoder dict
__a ={
0: self.pad_token,
1: self.eos_token,
}
if self.mask_token_sent is not None:
self.encoder.update(
{
2: self.mask_token_sent,
3: self.mask_token,
} )
if self.offset > 0:
# entries 2-104 are only used for pretraining and called <mask_1>, <mask_2>, unk_2, ...unk_102
# mask_token_sent is already added to list -> so start at 1
self.encoder.update({i + 3: additional_special_tokens[i] for i in range(1 , self.offset - 1 )} )
__a ={v: k for k, v in self.encoder.items()}
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return len(self.sp_model ) + self.offset
def __magic_name__ ( self ) -> Dict[str, int]:
'''simple docstring'''
__a ={self.convert_ids_to_tokens(_lowerCamelCase ): i for i in range(self.vocab_size )}
vocab.update(self.added_tokens_encoder )
return vocab
def __getstate__( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.__dict__.copy()
__a =None
return state
def __setstate__( self , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =d
# for backward compatibility
if not hasattr(self , 'sp_model_kwargs' ):
__a ={}
__a =spm.SentencePieceProcessor(**self.sp_model_kwargs )
self.sp_model.Load(self.vocab_file )
def __magic_name__ ( self , __snake_case ) -> List[str]:
'''simple docstring'''
return self.sp_model.encode(_lowerCamelCase , out_type=_lowerCamelCase )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
if token in self.decoder:
return self.decoder[token]
elif token in self.added_tokens_decoder:
return self.added_tokens_decoder[token]
__a =self.sp_model.piece_to_id(_lowerCamelCase )
return sp_id + self.offset
def __magic_name__ ( self , __snake_case ) -> str:
'''simple docstring'''
if index in self.encoder:
return self.encoder[index]
elif index in self.added_tokens_encoder:
return self.added_tokens_encoder[index]
else:
__a =self.sp_model.IdToPiece(index - self.offset )
return token
def __magic_name__ ( self , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =[]
__a =''''''
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(_lowerCamelCase ) + token
__a =[]
else:
current_sub_tokens.append(_lowerCamelCase )
out_string += self.sp_model.decode(_lowerCamelCase )
return out_string.strip()
def __magic_name__ ( self , __snake_case=False ) -> str:
'''simple docstring'''
return 1
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
__a =set(self.all_special_ids ) # call it once instead of inside list comp
all_special_ids.remove(self.unk_token_id ) # <unk> is only sometimes special
return [1 if x in all_special_ids else 0 for x in seq]
def __magic_name__ ( self , __snake_case , __snake_case = None , __snake_case = False ) -> List[int]:
'''simple docstring'''
if already_has_special_tokens:
return self._special_token_mask(_lowerCamelCase )
elif token_ids_a is None:
return self._special_token_mask(_lowerCamelCase ) + [1]
else:
return self._special_token_mask(token_ids_a + token_ids_a ) + [1]
def __magic_name__ ( self , __snake_case , __snake_case=None ) -> List[int]:
'''simple docstring'''
if token_ids_a is None:
return token_ids_a + [self.eos_token_id]
# We don't expect to process pairs, but leave the pair logic for API consistency
return token_ids_a + token_ids_a + [self.eos_token_id]
def __magic_name__ ( self , __snake_case , __snake_case = None ) -> Tuple[str]:
'''simple docstring'''
if not os.path.isdir(_lowerCamelCase ):
logger.error(f'Vocabulary path ({save_directory}) should be a directory' )
return
__a =os.path.join(
_lowerCamelCase , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
if os.path.abspath(self.vocab_file ) != os.path.abspath(_lowerCamelCase ) and os.path.isfile(self.vocab_file ):
copyfile(self.vocab_file , _lowerCamelCase )
elif not os.path.isfile(self.vocab_file ):
with open(_lowerCamelCase , 'wb' ) as fi:
__a =self.sp_model.serialized_model_proto()
fi.write(_lowerCamelCase )
return (out_vocab_file,)
| 369
|
import argparse
import logging
from collections import namedtuple
import torch
from model_bertabs import BertAbsSummarizer
from models.model_builder import AbsSummarizer # The authors' implementation
from transformers import BertTokenizer
logging.basicConfig(level=logging.INFO)
_lowerCAmelCase : List[Any] = logging.getLogger(__name__)
_lowerCAmelCase : Optional[Any] = "Hello world! cécé herlolip"
_lowerCAmelCase : str = namedtuple(
"BertAbsConfig",
[
"temp_dir",
"large",
"use_bert_emb",
"finetune_bert",
"encoder",
"share_emb",
"max_pos",
"enc_layers",
"enc_hidden_size",
"enc_heads",
"enc_ff_size",
"enc_dropout",
"dec_layers",
"dec_hidden_size",
"dec_heads",
"dec_ff_size",
"dec_dropout",
],
)
def UpperCamelCase_( _snake_case : str , _snake_case : List[Any] ):
"""simple docstring"""
__a =BertAbsConfig(
temp_dir='.' , finetune_bert=_snake_case , large=_snake_case , share_emb=_snake_case , use_bert_emb=_snake_case , encoder='bert' , max_pos=512 , enc_layers=6 , enc_hidden_size=512 , enc_heads=8 , enc_ff_size=512 , enc_dropout=0.2 , dec_layers=6 , dec_hidden_size=768 , dec_heads=8 , dec_ff_size=2048 , dec_dropout=0.2 , )
__a =torch.load(_snake_case , lambda _snake_case , _snake_case : storage )
__a =AbsSummarizer(_snake_case , torch.device('cpu' ) , _snake_case )
original.eval()
__a =BertAbsSummarizer(_snake_case , torch.device('cpu' ) )
new_model.eval()
# -------------------
# Convert the weights
# -------------------
logging.info('convert the model' )
new_model.bert.load_state_dict(original.bert.state_dict() )
new_model.decoder.load_state_dict(original.decoder.state_dict() )
new_model.generator.load_state_dict(original.generator.state_dict() )
# ----------------------------------
# Make sure the outpus are identical
# ----------------------------------
logging.info('Make sure that the models\' outputs are identical' )
__a =BertTokenizer.from_pretrained('bert-base-uncased' )
# prepare the model inputs
__a =tokenizer.encode('This is sample éàalj\'-.' )
encoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
__a =tokenizer.encode('This is sample 3 éàalj\'-.' )
decoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
# failsafe to make sure the weights reset does not affect the
# loaded weights.
assert torch.max(torch.abs(original.generator[0].weight - new_model.generator[0].weight ) ) == 0
# forward pass
__a =encoder_input_ids
__a =decoder_input_ids
__a =__a =None
__a =None
__a =__a =None
__a =__a =None
__a =None
# The original model does not apply the geneator layer immediatly but rather in
# the beam search (where it combines softmax + linear layer). Since we already
# apply the softmax in our generation process we only apply the linear layer here.
# We make sure that the outputs of the full stack are identical
__a =original(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =original.generator(_snake_case )
__a =new_model(
_snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =new_model.generator(_snake_case )
__a =torch.max(torch.abs(output_converted_model - output_original_model ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.max(torch.abs(output_converted_generator - output_original_generator ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.allclose(_snake_case , _snake_case , atol=1e-3 )
if are_identical:
logging.info('all weights are equal up to 1e-3' )
else:
raise ValueError('the weights are different. The new model is likely different from the original one.' )
# The model has been saved with torch.save(model) and this is bound to the exact
# directory structure. We save the state_dict instead.
logging.info('saving the model\'s state dictionary' )
torch.save(
new_model.state_dict() , './bertabs-finetuned-cnndm-extractive-abstractive-summarization/pytorch_model.bin' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
parser.add_argument(
"--bertabs_checkpoint_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch dump.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
type=str,
required=True,
help="Path to the output PyTorch model.",
)
_lowerCAmelCase : Optional[Any] = parser.parse_args()
convert_bertabs_checkpoints(
args.bertabs_checkpoint_path,
args.pytorch_dump_folder_path,
)
| 308
| 0
|
from collections import defaultdict
from pathlib import Path
import pandas as pd
from rouge_cli import calculate_rouge_path
from utils import calculate_rouge
_lowerCAmelCase : Optional[Any] = [
"Prosecutor: \"No videos were used in the crash investigation\" German papers say they saw a cell phone video of the"
" final seconds on board Flight 9525. The Germanwings co-pilot says he had a \"previous episode of severe"
" depression\" German airline confirms it knew of Andreas Lubitz's depression years before he took control.",
"The Palestinian Authority officially becomes the 123rd member of the International Criminal Court. The formal"
" accession was marked with a ceremony at The Hague, in the Netherlands. The Palestinians signed the ICC's"
" founding Rome Statute in January. Israel and the United States opposed the Palestinians' efforts to join the"
" body.",
"Amnesty International releases its annual report on the death penalty. The report catalogs the use of"
" state-sanctioned killing as a punitive measure across the globe. At least 607 people were executed around the"
" world in 2014, compared to 778 in 2013. The U.S. remains one of the worst offenders for imposing capital"
" punishment.",
]
_lowerCAmelCase : List[Any] = [
"Marseille prosecutor says \"so far no videos were used in the crash investigation\" despite media reports ."
" Journalists at Bild and Paris Match are \"very confident\" the video clip is real, an editor says . Andreas Lubitz"
" had informed his Lufthansa training school of an episode of severe depression, airline says .",
"Membership gives the ICC jurisdiction over alleged crimes committed in Palestinian territories since last June ."
" Israel and the United States opposed the move, which could open the door to war crimes investigations against"
" Israelis .",
"Amnesty's annual death penalty report catalogs encouraging signs, but setbacks in numbers of those sentenced to"
" death . Organization claims that governments around the world are using the threat of terrorism to advance"
" executions . The number of executions worldwide has gone down by almost 22% compared with 2013, but death"
" sentences up by 28% .",
]
def UpperCamelCase_( ):
"""simple docstring"""
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , bootstrap_aggregation=__SCREAMING_SNAKE_CASE , rouge_keys=['rouge2', 'rougeL'] )
assert isinstance(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE )
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , bootstrap_aggregation=__SCREAMING_SNAKE_CASE , rouge_keys=['rouge2'] )
assert (
pd.DataFrame(no_aggregation['rouge2'] ).fmeasure.mean()
== pd.DataFrame(no_aggregation_just_ra['rouge2'] ).fmeasure.mean()
)
def UpperCamelCase_( ):
"""simple docstring"""
__a ="""rougeLsum"""
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , newline_sep=__SCREAMING_SNAKE_CASE , rouge_keys=[k] )[k]
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , newline_sep=__SCREAMING_SNAKE_CASE , rouge_keys=[k] )[k]
assert score > score_no_sep
def UpperCamelCase_( ):
"""simple docstring"""
__a =["""rouge1""", """rouge2""", """rougeL"""]
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , newline_sep=__SCREAMING_SNAKE_CASE , rouge_keys=__SCREAMING_SNAKE_CASE )
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , newline_sep=__SCREAMING_SNAKE_CASE , rouge_keys=__SCREAMING_SNAKE_CASE )
assert score_sep == score_no_sep
def UpperCamelCase_( ):
"""simple docstring"""
__a =[
"""Her older sister, Margot Frank, died in 1945, a month earlier than previously thought.""",
"""Marseille prosecutor says \"so far no videos were used in the crash investigation\" despite media reports .""",
]
__a =[
"""Margot Frank, died in 1945, a month earlier than previously thought.""",
"""Prosecutor: \"No videos were used in the crash investigation\" German papers say they saw a cell phone video of"""
""" the final seconds on board Flight 9525.""",
]
assert calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , newline_sep=__SCREAMING_SNAKE_CASE ) == calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , newline_sep=__SCREAMING_SNAKE_CASE )
def UpperCamelCase_( ):
"""simple docstring"""
__a =[
"""\" \"a person who has such a video needs to immediately give it to the investigators,\" prosecutor says .<n> \"it is a very disturbing scene,\" editor-in-chief of bild online tells \"erin burnett: outfront\" """
]
__a =[
""" Marseille prosecutor says \"so far no videos were used in the crash investigation\" despite media reports . Journalists at Bild and Paris Match are \"very confident\" the video clip is real, an editor says . Andreas Lubitz had informed his Lufthansa training school of an episode of severe depression, airline says ."""
]
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , rouge_keys=['rougeLsum'] , newline_sep=__SCREAMING_SNAKE_CASE )["""rougeLsum"""]
__a =calculate_rouge(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , rouge_keys=['rougeLsum'] )["""rougeLsum"""]
assert new_score > prev_score
def UpperCamelCase_( ):
"""simple docstring"""
__a =Path('examples/seq2seq/test_data/wmt_en_ro' )
__a =calculate_rouge_path(data_dir.joinpath('test.source' ) , data_dir.joinpath('test.target' ) )
assert isinstance(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE )
__a =calculate_rouge_path(
data_dir.joinpath('test.source' ) , data_dir.joinpath('test.target' ) , bootstrap_aggregation=__SCREAMING_SNAKE_CASE )
assert isinstance(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE )
| 370
|
import unittest
from transformers import XLMConfig, is_torch_available
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 (
XLMForMultipleChoice,
XLMForQuestionAnswering,
XLMForQuestionAnsweringSimple,
XLMForSequenceClassification,
XLMForTokenClassification,
XLMModel,
XLMWithLMHeadModel,
)
from transformers.models.xlm.modeling_xlm import XLM_PRETRAINED_MODEL_ARCHIVE_LIST
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=13 , __snake_case=7 , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=False , __snake_case=False , __snake_case=False , __snake_case=2 , __snake_case=99 , __snake_case=0 , __snake_case=32 , __snake_case=5 , __snake_case=4 , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=2 , __snake_case=0.02 , __snake_case=2 , __snake_case=4 , __snake_case="last" , __snake_case=True , __snake_case=None , __snake_case=0 , ) -> Optional[Any]:
'''simple docstring'''
__a =parent
__a =batch_size
__a =seq_length
__a =is_training
__a =use_input_lengths
__a =use_token_type_ids
__a =use_labels
__a =gelu_activation
__a =sinusoidal_embeddings
__a =causal
__a =asm
__a =n_langs
__a =vocab_size
__a =n_special
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_sequence_label_size
__a =initializer_range
__a =num_labels
__a =num_choices
__a =summary_type
__a =use_proj
__a =scope
__a =bos_token_id
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a =random_attention_mask([self.batch_size, self.seq_length] )
__a =None
if self.use_input_lengths:
__a =(
ids_tensor([self.batch_size] , vocab_size=2 ) + self.seq_length - 2
) # small variation of seq_length
__a =None
if self.use_token_type_ids:
__a =ids_tensor([self.batch_size, self.seq_length] , self.n_langs )
__a =None
__a =None
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a =ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a =ids_tensor([self.batch_size] , 2 ).float()
__a =ids_tensor([self.batch_size] , self.num_choices )
__a =self.get_config()
return (
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
)
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return XLMConfig(
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 , num_labels=self.num_labels , bos_token_id=self.bos_token_id , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMModel(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , lengths=__snake_case , langs=__snake_case )
__a =model(__snake_case , langs=__snake_case )
__a =model(__snake_case )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[int]:
'''simple docstring'''
__a =XLMWithLMHeadModel(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , token_type_ids=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Dict:
'''simple docstring'''
__a =XLMForQuestionAnsweringSimple(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
__a =outputs
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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMForQuestionAnswering(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , p_mask=__snake_case , )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , )
((__a) , ) =result_with_labels.to_tuple()
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
((__a) , ) =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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[Any]:
'''simple docstring'''
__a =XLMForSequenceClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Any:
'''simple docstring'''
__a =self.num_labels
__a =XLMForTokenClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , attention_mask=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Tuple:
'''simple docstring'''
__a =self.num_choices
__a =XLMForMultipleChoice(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =model(
__snake_case , attention_mask=__snake_case , token_type_ids=__snake_case , labels=__snake_case , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) =config_and_inputs
__a ={'input_ids': input_ids, 'token_type_ids': token_type_ids, 'lengths': input_lengths}
return config, inputs_dict
@require_torch
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ , lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
XLMModel,
XLMWithLMHeadModel,
XLMForQuestionAnswering,
XLMForSequenceClassification,
XLMForQuestionAnsweringSimple,
XLMForTokenClassification,
XLMForMultipleChoice,
)
if is_torch_available()
else ()
)
SCREAMING_SNAKE_CASE = (
(XLMWithLMHeadModel,) if is_torch_available() else ()
) # TODO (PVP): Check other models whether language generation is also applicable
SCREAMING_SNAKE_CASE = (
{
'feature-extraction': XLMModel,
'fill-mask': XLMWithLMHeadModel,
'question-answering': XLMForQuestionAnsweringSimple,
'text-classification': XLMForSequenceClassification,
'text-generation': XLMWithLMHeadModel,
'token-classification': XLMForTokenClassification,
'zero-shot': XLMForSequenceClassification,
}
if is_torch_available()
else {}
)
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> int:
'''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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case=False ) -> str:
'''simple docstring'''
__a =super()._prepare_for_class(__snake_case , __snake_case , return_labels=__snake_case )
if return_labels:
if model_class.__name__ == "XLMForQuestionAnswering":
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
return inputs_dict
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMModelTester(self )
__a =ConfigTester(self , config_class=__snake_case , emb_dim=37 )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.config_tester.run_common_tests()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_model(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_lm_head(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_simple_qa(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_qa(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_sequence_classif(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_token_classif(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_for_multiple_choice(*__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Optional[Any]:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_attentions in attentions] , [True] * len(__snake_case ) )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_attentions in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =min_length + idx + 1
__a =(
batch_size * num_beam_groups,
config.num_attention_heads,
tgt_len,
src_len,
)
# check attn size
self.assertListEqual(
[layer_attention.shape for layer_attention in iter_attentions] , [expected_shape] * len(__snake_case ) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Dict:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_hidden_states in hidden_states] , [True] * len(__snake_case ) , )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_hidden_states in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =(batch_size * num_beam_groups, seq_len, config.hidden_size)
# check hidden size
self.assertListEqual(
[layer_hidden_states.shape for layer_hidden_states in iter_hidden_states] , [expected_shape] * len(__snake_case ) , )
pass
@slow
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
for model_name in XLM_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =XLMModel.from_pretrained(__snake_case )
self.assertIsNotNone(__snake_case )
@require_torch
class __magic_name__ ( unittest.TestCase ):
@slow
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMWithLMHeadModel.from_pretrained('xlm-mlm-en-2048' )
model.to(__snake_case )
__a =torch.tensor([[14, 447]] , dtype=torch.long , device=__snake_case ) # the president
__a =[
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
] # the president the president the president the president the president the president the president the president the president the president
# TODO(PVP): this and other input_ids I tried for generation give pretty bad results. Not sure why. Model might just not be made for auto-regressive inference
__a =model.generate(__snake_case , do_sample=__snake_case )
self.assertListEqual(output_ids[0].cpu().numpy().tolist() , __snake_case )
| 308
| 0
|
"""simple docstring"""
import os
import pytest
from transformers.dynamic_module_utils import get_imports
_lowerCAmelCase : List[Any] = "\nimport os\n"
_lowerCAmelCase : List[str] = "\ndef foo():\n import os\n return False\n"
_lowerCAmelCase : List[str] = "\ndef foo():\n def bar():\n if True:\n import os\n return False\n return bar()\n"
_lowerCAmelCase : Any = "\nimport os\n\ntry:\n import bar\nexcept ImportError:\n raise ValueError()\n"
_lowerCAmelCase : str = "\nimport os\n\ndef foo():\n try:\n import bar\n except ImportError:\n raise ValueError()\n"
_lowerCAmelCase : List[Any] = "\nimport os\n\ntry:\n import bar\nexcept (ImportError, AttributeError):\n raise ValueError()\n"
_lowerCAmelCase : Optional[int] = "\nimport os\n\ntry:\n import bar\nexcept ImportError as e:\n raise ValueError()\n"
_lowerCAmelCase : Tuple = "\nimport os\n\ntry:\n import bar\nexcept:\n raise ValueError()\n"
_lowerCAmelCase : List[Any] = "\nimport os\n\ntry:\n import bar\n import baz\nexcept ImportError:\n raise ValueError()\n"
_lowerCAmelCase : Optional[int] = "\nimport os\n\ntry:\n import bar\n import baz\nexcept ImportError:\n x = 1\n raise ValueError()\n"
_lowerCAmelCase : Any = [
TOP_LEVEL_IMPORT,
IMPORT_IN_FUNCTION,
DEEPLY_NESTED_IMPORT,
TOP_LEVEL_TRY_IMPORT,
GENERIC_EXCEPT_IMPORT,
MULTILINE_TRY_IMPORT,
MULTILINE_BOTH_IMPORT,
MULTIPLE_EXCEPTS_IMPORT,
EXCEPT_AS_IMPORT,
TRY_IMPORT_IN_FUNCTION,
]
@pytest.mark.parametrize('case' , UpperCAmelCase__ )
def UpperCamelCase_( _snake_case : List[str] , _snake_case : Tuple ):
"""simple docstring"""
__a =os.path.join(UpperCAmelCase__ , 'test_file.py' )
with open(UpperCAmelCase__ , 'w' ) as _tmp_file:
_tmp_file.write(UpperCAmelCase__ )
__a =get_imports(UpperCAmelCase__ )
assert parsed_imports == ["os"]
| 371
|
from __future__ import annotations
import matplotlib.pyplot as plt # type: ignore
import numpy
# initial triangle of Koch snowflake
_lowerCAmelCase : Optional[Any] = numpy.array([0, 0])
_lowerCAmelCase : Dict = numpy.array([0.5, 0.8660254])
_lowerCAmelCase : Any = numpy.array([1, 0])
_lowerCAmelCase : int = [VECTOR_1, VECTOR_2, VECTOR_3, VECTOR_1]
def UpperCamelCase_( _snake_case : list[numpy.ndarray] , _snake_case : int ):
"""simple docstring"""
__a =initial_vectors
for _ in range(_snake_case ):
__a =iteration_step(_snake_case )
return vectors
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =[]
for i, start_vector in enumerate(vectors[:-1] ):
__a =vectors[i + 1]
new_vectors.append(_snake_case )
__a =end_vector - start_vector
new_vectors.append(start_vector + difference_vector / 3 )
new_vectors.append(
start_vector + difference_vector / 3 + rotate(difference_vector / 3 , 60 ) )
new_vectors.append(start_vector + difference_vector * 2 / 3 )
new_vectors.append(vectors[-1] )
return new_vectors
def UpperCamelCase_( _snake_case : numpy.ndarray , _snake_case : float ):
"""simple docstring"""
__a =numpy.radians(_snake_case )
__a , __a =numpy.cos(_snake_case ), numpy.sin(_snake_case )
__a =numpy.array(((c, -s), (s, c)) )
return numpy.dot(_snake_case , _snake_case )
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =plt.gca()
axes.set_aspect('equal' )
# matplotlib.pyplot.plot takes a list of all x-coordinates and a list of all
# y-coordinates as inputs, which are constructed from the vector-list using
# zip()
__a , __a =zip(*_snake_case )
plt.plot(_snake_case , _snake_case )
plt.show()
if __name__ == "__main__":
import doctest
doctest.testmod()
_lowerCAmelCase : List[Any] = iterate(INITIAL_VECTORS, 5)
plot(processed_vectors)
| 308
| 0
|
import math
def SCREAMING_SNAKE_CASE_( _snake_case : int ):
"""simple docstring"""
return math.sqrt(__a ) * math.sqrt(__a ) == num
def SCREAMING_SNAKE_CASE_( _snake_case : int ):
"""simple docstring"""
__a =0
__a =n
while left <= right:
__a =(left + right) // 2
if mid**2 == n:
return True
elif mid**2 > n:
__a =mid - 1
else:
__a =mid + 1
return False
if __name__ == "__main__":
import doctest
doctest.testmod()
| 350
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
_lowerCAmelCase : Tuple = {
"configuration_megatron_bert": ["MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MegatronBertConfig"],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : List[str] = [
"MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"MegatronBertForCausalLM",
"MegatronBertForMaskedLM",
"MegatronBertForMultipleChoice",
"MegatronBertForNextSentencePrediction",
"MegatronBertForPreTraining",
"MegatronBertForQuestionAnswering",
"MegatronBertForSequenceClassification",
"MegatronBertForTokenClassification",
"MegatronBertModel",
"MegatronBertPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_megatron_bert import MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MegatronBertConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_megatron_bert import (
MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
MegatronBertForCausalLM,
MegatronBertForMaskedLM,
MegatronBertForMultipleChoice,
MegatronBertForNextSentencePrediction,
MegatronBertForPreTraining,
MegatronBertForQuestionAnswering,
MegatronBertForSequenceClassification,
MegatronBertForTokenClassification,
MegatronBertModel,
MegatronBertPreTrainedModel,
)
else:
import sys
_lowerCAmelCase : Optional[int] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 0
|
import copy
import os
from typing import Union
from ...configuration_utils import PretrainedConfig
from ...utils import logging
_lowerCAmelCase : List[Any] = logging.get_logger(__name__)
_lowerCAmelCase : Optional[int] = {
"microsoft/git-base": "https://huggingface.co/microsoft/git-base/resolve/main/config.json",
}
class __magic_name__ ( lowercase__ ):
SCREAMING_SNAKE_CASE = """git_vision_model"""
def __init__( self , __snake_case=768 , __snake_case=3072 , __snake_case=12 , __snake_case=12 , __snake_case=3 , __snake_case=224 , __snake_case=16 , __snake_case="quick_gelu" , __snake_case=1e-5 , __snake_case=0.0 , __snake_case=0.02 , **__snake_case , ) -> Optional[int]:
'''simple docstring'''
super().__init__(**lowercase_ )
__a =hidden_size
__a =intermediate_size
__a =num_hidden_layers
__a =num_attention_heads
__a =num_channels
__a =patch_size
__a =image_size
__a =initializer_range
__a =attention_dropout
__a =layer_norm_eps
__a =hidden_act
@classmethod
def __magic_name__ ( cls , __snake_case , **__snake_case ) -> List[Any]:
'''simple docstring'''
cls._set_token_in_kwargs(lowercase_ )
__a =cls.get_config_dict(lowercase_ , **lowercase_ )
# get the vision config dict if we are loading from GITConfig
if config_dict.get('model_type' ) == "git":
__a =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(lowercase_ , **lowercase_ )
class __magic_name__ ( lowercase__ ):
SCREAMING_SNAKE_CASE = """git"""
def __init__( self , __snake_case=None , __snake_case=3_0522 , __snake_case=768 , __snake_case=6 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.1 , __snake_case=0.1 , __snake_case=1024 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=0 , __snake_case="absolute" , __snake_case=True , __snake_case=False , __snake_case=101 , __snake_case=102 , __snake_case=None , **__snake_case , ) -> int:
'''simple docstring'''
super().__init__(bos_token_id=lowercase_ , eos_token_id=lowercase_ , pad_token_id=lowercase_ , **lowercase_ )
if vision_config is None:
__a ={}
logger.info('vision_config is None. initializing the GitVisionConfig with default values.' )
__a =GitVisionConfig(**lowercase_ )
__a =vocab_size
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_act
__a =intermediate_size
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =initializer_range
__a =layer_norm_eps
__a =position_embedding_type
__a =use_cache
__a =tie_word_embeddings
__a =num_image_with_embedding
__a =bos_token_id
__a =eos_token_id
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =copy.deepcopy(self.__dict__ )
__a =self.vision_config.to_dict()
__a =self.__class__.model_type
return output
| 351
|
import hashlib
import unittest
from transformers import MODEL_FOR_DEPTH_ESTIMATION_MAPPING, is_torch_available, is_vision_available
from transformers.pipelines import DepthEstimationPipeline, pipeline
from transformers.testing_utils import (
is_pipeline_test,
nested_simplify,
require_tf,
require_timm,
require_torch,
require_vision,
slow,
)
from .test_pipelines_common import ANY
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
else:
class __magic_name__ :
@staticmethod
def __magic_name__ ( *__snake_case , **__snake_case ) -> List[str]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : Image ):
"""simple docstring"""
__a =hashlib.mda(image.tobytes() )
return m.hexdigest()
@is_pipeline_test
@require_vision
@require_timm
@require_torch
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = MODEL_FOR_DEPTH_ESTIMATION_MAPPING
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Dict:
'''simple docstring'''
__a =DepthEstimationPipeline(model=__snake_case , image_processor=__snake_case )
return depth_estimator, [
"./tests/fixtures/tests_samples/COCO/000000039769.png",
"./tests/fixtures/tests_samples/COCO/000000039769.png",
]
def __magic_name__ ( self , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
__a =depth_estimator('./tests/fixtures/tests_samples/COCO/000000039769.png' )
self.assertEqual({'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )} , __snake_case )
import datasets
__a =datasets.load_dataset('hf-internal-testing/fixtures_image_utils' , 'image' , split='test' )
__a =depth_estimator(
[
Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' ),
'http://images.cocodataset.org/val2017/000000039769.jpg',
# RGBA
dataset[0]['file'],
# LA
dataset[1]['file'],
# L
dataset[2]['file'],
] )
self.assertEqual(
[
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
] , __snake_case , )
@require_tf
@unittest.skip('Depth estimation is not implemented in TF' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
pass
@slow
@require_torch
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a ='Intel/dpt-large'
__a =pipeline('depth-estimation' , model=__snake_case )
__a =depth_estimator('http://images.cocodataset.org/val2017/000000039769.jpg' )
__a =hashimage(outputs['depth'] )
# This seems flaky.
# self.assertEqual(outputs["depth"], "1a39394e282e9f3b0741a90b9f108977")
self.assertEqual(nested_simplify(outputs['predicted_depth'].max().item() ) , 29.304 )
self.assertEqual(nested_simplify(outputs['predicted_depth'].min().item() ) , 2.662 )
@require_torch
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
# This is highly irregular to have no small tests.
self.skipTest('There is not hf-internal-testing tiny model for either GLPN nor DPT' )
| 308
| 0
|
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 ViTImageProcessor
class __magic_name__ ( unittest.TestCase ):
def __init__( self , __snake_case , __snake_case=13 , __snake_case=3 , __snake_case=224 , __snake_case=30 , __snake_case=400 , __snake_case=True , __snake_case=None , __snake_case=True , __snake_case=[0.5, 0.5, 0.5] , __snake_case=[0.5, 0.5, 0.5] , ) -> List[str]:
'''simple docstring'''
__a =size if size is not None else {'''height''': 18, '''width''': 18}
__a =parent
__a =batch_size
__a =num_channels
__a =image_size
__a =min_resolution
__a =max_resolution
__a =do_resize
__a =size
__a =do_normalize
__a =image_mean
__a =image_std
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
return {
"image_mean": self.image_mean,
"image_std": self.image_std,
"do_normalize": self.do_normalize,
"do_resize": self.do_resize,
"size": self.size,
}
@require_torch
@require_vision
class __magic_name__ ( __snake_case , unittest.TestCase ):
SCREAMING_SNAKE_CASE = ViTImageProcessor if is_vision_available() else None
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =EfficientFormerImageProcessorTester(self )
@property
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
return self.image_proc_tester.prepare_image_processor_dict()
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.image_processing_class(**self.image_processor_dict )
self.assertTrue(hasattr(a_ , 'image_mean' ) )
self.assertTrue(hasattr(a_ , 'image_std' ) )
self.assertTrue(hasattr(a_ , 'do_normalize' ) )
self.assertTrue(hasattr(a_ , 'do_resize' ) )
self.assertTrue(hasattr(a_ , 'size' ) )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.image_processing_class(**self.image_processor_dict )
# create random PIL images
__a =prepare_image_inputs(self.image_proc_tester , equal_resolution=a_ )
for image in image_inputs:
self.assertIsInstance(a_ , Image.Image )
# Test not batched input
__a =image_processor(image_inputs[0] , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_proc_tester.num_channels,
self.image_proc_tester.size['height'],
self.image_proc_tester.size['width'],
) , )
# Test batched
__a =image_processor(a_ , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_proc_tester.batch_size,
self.image_proc_tester.num_channels,
self.image_proc_tester.size['height'],
self.image_proc_tester.size['width'],
) , )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =self.image_processing_class(**self.image_processor_dict )
# create random numpy tensors
__a =prepare_image_inputs(self.image_proc_tester , equal_resolution=a_ , numpify=a_ )
for image in image_inputs:
self.assertIsInstance(a_ , np.ndarray )
# Test not batched input
__a =image_processor(image_inputs[0] , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_proc_tester.num_channels,
self.image_proc_tester.size['height'],
self.image_proc_tester.size['width'],
) , )
# Test batched
__a =image_processor(a_ , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_proc_tester.batch_size,
self.image_proc_tester.num_channels,
self.image_proc_tester.size['height'],
self.image_proc_tester.size['width'],
) , )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.image_processing_class(**self.image_processor_dict )
# create random PyTorch tensors
__a =prepare_image_inputs(self.image_proc_tester , equal_resolution=a_ , torchify=a_ )
for image in image_inputs:
self.assertIsInstance(a_ , torch.Tensor )
# Test not batched input
__a =image_processor(image_inputs[0] , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_proc_tester.num_channels,
self.image_proc_tester.size['height'],
self.image_proc_tester.size['width'],
) , )
# Test batched
__a =image_processor(a_ , return_tensors='pt' ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_proc_tester.batch_size,
self.image_proc_tester.num_channels,
self.image_proc_tester.size['height'],
self.image_proc_tester.size['width'],
) , )
| 352
|
import argparse
import dataclasses
import json
import logging
import os
import shutil
from typing import List, Optional
import datasets
from accelerate import Accelerator
from datasets import load_dataset
from finetuning import finetune
from tqdm.auto import tqdm
import transformers
from transformers import AutoConfig, set_seed
from transformers.trainer_utils import IntervalStrategy
_lowerCAmelCase : Optional[int] = logging.getLogger(__name__)
_lowerCAmelCase : Any = "pytorch_model.bin"
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co.'} , )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the data to predict on.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The name of the task to train on.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The list of labels for the task.'} )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'The output directory where the model predictions and checkpoints will be written.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='accuracy' , metadata={'help': 'The evaluation metric used for the task.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='no' , metadata={
'help': 'The evaluation strategy to adopt during training. Possible values are: ["no", "step", "epoch]'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={
'help': 'How much the specified evaluation metric must improve to satisfy early stopping conditions.'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the confidence score.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the validation performance.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to fine-tune on labeled data after pseudo training.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={'help': 'Confidence threshold for pseudo-labeled data filtering.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Random seed for initialization.'} , )
def UpperCamelCase_( _snake_case : int , _snake_case : str , _snake_case : Optional[int] , _snake_case : str , _snake_case : Union[str, Any] , _snake_case : List[Any] ):
"""simple docstring"""
__a =datasets.concatenate_datasets([infer_input, infer_output] , axis=1 )
if args.do_filter_by_confidence:
__a =dataset.filter(lambda _snake_case : example["probability"] > args.confidence_threshold )
if args.do_filter_by_val_performance:
assert eval_result >= 0.0 and eval_result <= 1.0
__a =int(eval_result * len(_snake_case ) )
print(_snake_case )
__a =dataset.sort('probability' , reverse=_snake_case )
__a =dataset.select(range(_snake_case ) )
__a =dataset.remove_columns(['label', 'probability'] )
__a =dataset.rename_column('prediction' , 'label' )
__a =dataset.map(lambda _snake_case : {"label": idalabel[example["label"]]} )
__a =dataset.shuffle(seed=args.seed )
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.data_file_extension == "csv":
dataset.to_csv(_snake_case , index=_snake_case )
else:
dataset.to_json(_snake_case )
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : str , _snake_case : int , _snake_case : Optional[int] , **_snake_case : List[str] ):
"""simple docstring"""
__a =Accelerator()
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , level=logging.INFO , )
logger.info(accelerator.state )
# Setup logging, we only want one process per machine to log things on the
# screen. accelerator.is_local_main_process is only True for one process per
# machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR )
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
__a =STModelArguments(model_name_or_path=_snake_case )
__a =STDataArguments(train_file=_snake_case , infer_file=_snake_case )
__a =STTrainingArguments(output_dir=_snake_case )
__a =argparse.Namespace()
for arg_class in (model_args, data_args, training_args):
for key, value in vars(_snake_case ).items():
setattr(_snake_case , _snake_case , _snake_case )
for key, value in kwargs.items():
if hasattr(_snake_case , _snake_case ):
setattr(_snake_case , _snake_case , _snake_case )
# Sanity checks
__a ={}
__a =None
# You need to provide the training data and the data to predict on
assert args.train_file is not None
assert args.infer_file is not None
__a =args.train_file
__a =args.infer_file
if args.evaluation_strategy != IntervalStrategy.NO.value:
assert args.eval_file is not None
__a =args.eval_file
for key in data_files:
__a =data_files[key].split('.' )[-1]
assert extension in ["csv", "json"], F'`{key}_file` should be a csv or a json file.'
if args.data_file_extension is None:
__a =extension
else:
assert extension == args.data_file_extension, F'`{key}_file` should be a {args.data_file_extension} file`.'
assert (
args.eval_metric in datasets.list_metrics()
), F'{args.eval_metric} not in the list of supported metrics {datasets.list_metrics()}.'
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed )
logger.info('Creating the initial data directory for self-training...' )
__a =F'{args.output_dir}/self-train_iter-{{}}'.format
__a =data_dir_format(0 )
if accelerator.is_main_process:
if args.output_dir is not None:
os.makedirs(args.output_dir , exist_ok=_snake_case )
os.makedirs(_snake_case , exist_ok=_snake_case )
accelerator.wait_for_everyone()
__a =None
__a =None
__a =0
__a =False
# Show the progress bar
__a =tqdm(range(args.max_selftrain_iterations ) , disable=not accelerator.is_local_main_process )
# Self-train
for iteration in range(0 , int(args.max_selftrain_iterations ) ):
__a =data_dir_format(_snake_case )
assert os.path.exists(_snake_case )
# Stage 1: initial fine-tuning for iteration = 0 or pseudo-training for
# iteration > 0
__a =os.path.join(_snake_case , 'stage-1' )
__a ={
'accelerator': accelerator,
'model_name_or_path': args.model_name_or_path,
'cache_dir': args.cache_dir,
'do_train': True,
'train_file': data_files['train'] if iteration == 0 else data_files['train_pseudo'],
'do_eval': True if args.eval_file is not None else False,
'eval_file': data_files['eval'],
'do_predict': True,
'infer_file': data_files['infer'],
'task_name': args.task_name,
'label_list': args.label_list,
'output_dir': current_output_dir,
'eval_metric': args.eval_metric,
'evaluation_strategy': args.evaluation_strategy,
'early_stopping_patience': args.early_stopping_patience,
'early_stopping_threshold': args.early_stopping_threshold,
'seed': args.seed,
}
# Add additional training arguments
for key, value in kwargs.items():
if key not in arguments_dict and not hasattr(_snake_case , _snake_case ):
arguments_dict.update({key: value} )
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 1.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 1 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 1.' , _snake_case )
if iteration > 0 and args.finetune_on_labeled_data:
# Stage 2 (optional): fine-tuning on the original labeled data
__a =os.path.join(_snake_case , 'best-checkpoint' )
__a =os.path.join(_snake_case , 'stage-2' )
# Update arguments_dict
__a =model_path
__a =data_files['train']
__a =current_output_dir
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 2.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 2 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 2.' , _snake_case )
__a =iteration
__a =data_dir_format(iteration + 1 )
__a =AutoConfig.from_pretrained(os.path.join(_snake_case , 'best-checkpoint' ) )
__a =config.idalabel
__a =os.path.join(_snake_case , 'eval_results_best-checkpoint.json' )
__a =os.path.join(_snake_case , 'test_results_best-checkpoint.json' )
assert os.path.exists(_snake_case )
with open(_snake_case , 'r' ) as f:
__a =float(json.load(_snake_case )[args.eval_metric] )
__a =os.path.join(_snake_case , 'infer_output_best-checkpoint.csv' )
assert os.path.exists(_snake_case )
# Loading the dataset from local csv or json files.
__a =load_dataset(args.data_file_extension , data_files={'data': data_files['infer']} )['data']
__a =load_dataset('csv' , data_files={'data': infer_output_file} )['data']
if accelerator.is_main_process:
os.makedirs(_snake_case , exist_ok=_snake_case )
shutil.copy(_snake_case , os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) )
if os.path.exists(_snake_case ):
shutil.copy(_snake_case , os.path.join(_snake_case , F'test_results_iter-{iteration}.json' ) )
create_pseudo_labeled_data(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )
accelerator.wait_for_everyone()
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.evaluation_strategy != IntervalStrategy.NO.value:
__a =eval_result
if best_iteration is None:
__a =new_iteration
__a =new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
__a =new_iteration
__a =new_eval_result
__a =0
else:
if new_eval_result == best_eval_result:
__a =new_iteration
__a =new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
__a =True
progress_bar.update(1 )
if should_training_stop:
break
if best_iteration is not None:
# Save the best iteration
logger.info('Best iteration: %d' , _snake_case )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
else:
# Assume that the last iteration is the best
logger.info('Best iteration: %d' , args.max_selftrain_iterations - 1 )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{args.max_selftrain_iterations - 1}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
| 308
| 0
|
import json
import os
from functools import lru_cache
from typing import TYPE_CHECKING, List, Optional, Tuple
import regex as re
from ...tokenization_utils import AddedToken, PreTrainedTokenizer
from ...utils import logging
if TYPE_CHECKING:
from transformers.pipelines.conversational import Conversation
_lowerCAmelCase : Union[str, Any] = logging.get_logger(__name__)
_lowerCAmelCase : str = {
"""vocab_file""": """vocab.json""",
"""merges_file""": """merges.txt""",
"""tokenizer_config_file""": """tokenizer_config.json""",
}
_lowerCAmelCase : Tuple = {
"""vocab_file""": {"""facebook/blenderbot-3B""": """https://huggingface.co/facebook/blenderbot-3B/resolve/main/vocab.json"""},
"""merges_file""": {"""facebook/blenderbot-3B""": """https://huggingface.co/facebook/blenderbot-3B/resolve/main/merges.txt"""},
"""tokenizer_config_file""": {
"""facebook/blenderbot-3B""": """https://huggingface.co/facebook/blenderbot-3B/resolve/main/tokenizer_config.json"""
},
}
_lowerCAmelCase : Dict = {"""facebook/blenderbot-3B""": 128}
@lru_cache()
# Copied from transformers.models.roberta.tokenization_roberta.bytes_to_unicode
def UpperCamelCase_( ):
"""simple docstring"""
__a =(
list(range(ord('!' ) , ord('~' ) + 1 ) ) + list(range(ord('¡' ) , ord('¬' ) + 1 ) ) + list(range(ord('®' ) , ord('ÿ' ) + 1 ) )
)
__a =bs[:]
__a =0
for b in range(2**8 ):
if b not in bs:
bs.append(_A )
cs.append(2**8 + n )
n += 1
__a =[chr(_A ) for n in cs]
return dict(zip(_A , _A ) )
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =set()
__a =word[0]
for char in word[1:]:
pairs.add((prev_char, char) )
__a =char
return pairs
class __magic_name__ ( SCREAMING_SNAKE_CASE_ ):
SCREAMING_SNAKE_CASE = VOCAB_FILES_NAMES
SCREAMING_SNAKE_CASE = PRETRAINED_VOCAB_FILES_MAP
SCREAMING_SNAKE_CASE = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
SCREAMING_SNAKE_CASE = ['input_ids', 'attention_mask']
def __init__( self , __snake_case , __snake_case , __snake_case="replace" , __snake_case="<s>" , __snake_case="</s>" , __snake_case="</s>" , __snake_case="<s>" , __snake_case="<unk>" , __snake_case="<pad>" , __snake_case="<mask>" , __snake_case=False , **__snake_case , ):
'''simple docstring'''
__a =AddedToken(snake_case__ , lstrip=snake_case__ , rstrip=snake_case__ ) if isinstance(snake_case__ , snake_case__ ) else bos_token
__a =AddedToken(snake_case__ , lstrip=snake_case__ , rstrip=snake_case__ ) if isinstance(snake_case__ , snake_case__ ) else eos_token
__a =AddedToken(snake_case__ , lstrip=snake_case__ , rstrip=snake_case__ ) if isinstance(snake_case__ , snake_case__ ) else sep_token
__a =AddedToken(snake_case__ , lstrip=snake_case__ , rstrip=snake_case__ ) if isinstance(snake_case__ , snake_case__ ) else cls_token
__a =AddedToken(snake_case__ , lstrip=snake_case__ , rstrip=snake_case__ ) if isinstance(snake_case__ , snake_case__ ) else unk_token
__a =AddedToken(snake_case__ , lstrip=snake_case__ , rstrip=snake_case__ ) if isinstance(snake_case__ , snake_case__ ) else pad_token
# Mask token behave like a normal word, i.e. include the space before it
__a =AddedToken(snake_case__ , lstrip=snake_case__ , rstrip=snake_case__ ) if isinstance(snake_case__ , snake_case__ ) else mask_token
super().__init__(
errors=snake_case__ , bos_token=snake_case__ , eos_token=snake_case__ , unk_token=snake_case__ , sep_token=snake_case__ , cls_token=snake_case__ , pad_token=snake_case__ , mask_token=snake_case__ , add_prefix_space=snake_case__ , **snake_case__ , )
with open(snake_case__ , encoding='utf-8' ) as vocab_handle:
__a =json.load(snake_case__ )
__a ={v: k for k, v in self.encoder.items()}
__a =errors # how to handle errors in decoding
__a =bytes_to_unicode()
__a ={v: k for k, v in self.byte_encoder.items()}
with open(snake_case__ , encoding='utf-8' ) as merges_handle:
__a =merges_handle.read().split('\n' )[1:-1]
__a =[tuple(merge.split() ) for merge in bpe_merges]
__a =dict(zip(snake_case__ , range(len(snake_case__ ) ) ) )
__a ={}
__a =add_prefix_space
# Should have added re.IGNORECASE so BPE merges can happen for capitalized versions of contractions
__a =re.compile(R'\'s|\'t|\'re|\'ve|\'m|\'ll|\'d| ?\p{L}+| ?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+' )
@property
# Copied from transformers.models.roberta.tokenization_roberta.RobertaTokenizer.vocab_size with Roberta->Blenderbot, RoBERTa->Blenderbot
def __magic_name__ ( self ):
'''simple docstring'''
return len(self.encoder )
def __magic_name__ ( self ):
'''simple docstring'''
return dict(self.encoder , **self.added_tokens_encoder )
def __magic_name__ ( self , __snake_case ):
'''simple docstring'''
if token in self.cache:
return self.cache[token]
__a =tuple(snake_case__ )
__a =get_pairs(snake_case__ )
if not pairs:
return token
while True:
__a =min(snake_case__ , key=lambda __snake_case : self.bpe_ranks.get(snake_case__ , float('inf' ) ) )
if bigram not in self.bpe_ranks:
break
__a =bigram
__a =[]
__a =0
while i < len(snake_case__ ):
try:
__a =word.index(snake_case__ , snake_case__ )
except ValueError:
new_word.extend(word[i:] )
break
else:
new_word.extend(word[i:j] )
__a =j
if word[i] == first and i < len(snake_case__ ) - 1 and word[i + 1] == second:
new_word.append(first + second )
i += 2
else:
new_word.append(word[i] )
i += 1
__a =tuple(snake_case__ )
__a =new_word
if len(snake_case__ ) == 1:
break
else:
__a =get_pairs(snake_case__ )
__a =' '.join(snake_case__ )
__a =word
return word
def __magic_name__ ( self , __snake_case ):
'''simple docstring'''
__a =[]
for token in re.findall(self.pat , snake_case__ ):
__a =''.join(
self.byte_encoder[b] for b in token.encode('utf-8' ) ) # Maps all our bytes to unicode strings, avoiding control tokens of the BPE (spaces in our case)
bpe_tokens.extend(bpe_token for bpe_token in self.bpe(snake_case__ ).split(' ' ) )
return bpe_tokens
def __magic_name__ ( self , __snake_case ):
'''simple docstring'''
return self.encoder.get(snake_case__ , self.encoder.get(self.unk_token ) )
def __magic_name__ ( self , __snake_case ):
'''simple docstring'''
return self.decoder.get(snake_case__ )
def __magic_name__ ( self , __snake_case ):
'''simple docstring'''
__a =''.join(snake_case__ )
__a =bytearray([self.byte_decoder[c] for c in text] ).decode('utf-8' , errors=self.errors )
return text
def __magic_name__ ( self , __snake_case , __snake_case = None ):
'''simple docstring'''
if not os.path.isdir(snake_case__ ):
logger.error(f'Vocabulary path ({save_directory}) should be a directory' )
return
__a =os.path.join(
snake_case__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(
snake_case__ , (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['merges_file'] )
with open(snake_case__ , 'w' , encoding='utf-8' ) as f:
f.write(json.dumps(self.encoder , indent=2 , sort_keys=snake_case__ , ensure_ascii=snake_case__ ) + '\n' )
__a =0
with open(snake_case__ , 'w' , encoding='utf-8' ) as writer:
writer.write('#version: 0.2\n' )
for bpe_tokens, token_index in sorted(self.bpe_ranks.items() , key=lambda __snake_case : kv[1] ):
if index != token_index:
logger.warning(
f'Saving vocabulary to {merge_file}: BPE merge indices are not consecutive.'
' Please check that the tokenizer is not corrupted!' )
__a =token_index
writer.write(' '.join(snake_case__ ) + '\n' )
index += 1
return vocab_file, merge_file
def __magic_name__ ( self , __snake_case , __snake_case = None , __snake_case = False ):
'''simple docstring'''
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=snake_case__ , token_ids_a=snake_case__ , already_has_special_tokens=snake_case__ )
if token_ids_a is None:
return [1] + ([0] * len(snake_case__ )) + [1]
return [1] + ([0] * len(snake_case__ )) + [1, 1] + ([0] * len(snake_case__ )) + [1]
def __magic_name__ ( self , __snake_case , __snake_case = None ):
'''simple docstring'''
__a =[self.sep_token_id]
__a =[self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
def __magic_name__ ( self , __snake_case , __snake_case=False , **__snake_case ):
'''simple docstring'''
__a =kwargs.pop('add_prefix_space' , self.add_prefix_space )
if (is_split_into_words or add_prefix_space) and (len(snake_case__ ) > 0 and not text[0].isspace()):
__a =' ' + text
return (text, kwargs)
def __magic_name__ ( self , __snake_case , __snake_case = None ):
'''simple docstring'''
return token_ids_a + [self.eos_token_id]
def __magic_name__ ( self , __snake_case ):
'''simple docstring'''
__a =[]
for is_user, text in conversation.iter_texts():
if is_user:
# We need to space prefix as it's being done within blenderbot
inputs.append(' ' + text )
else:
# Generated responses should contain them already.
inputs.append(snake_case__ )
__a =' '.join(snake_case__ )
__a =self.encode(snake_case__ )
if len(snake_case__ ) > self.model_max_length:
__a =input_ids[-self.model_max_length :]
logger.warning(f'Trimmed input from conversation as it was longer than {self.model_max_length} tokens.' )
return input_ids
| 353
|
import shutil
import tempfile
import unittest
from transformers import (
SPIECE_UNDERLINE,
AddedToken,
BatchEncoding,
NllbTokenizer,
NllbTokenizerFast,
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
_lowerCAmelCase : Any = get_tests_dir("fixtures/test_sentencepiece.model")
if is_torch_available():
from transformers.models.mam_aaa.modeling_mam_aaa import shift_tokens_right
_lowerCAmelCase : List[Any] = 256_047
_lowerCAmelCase : Dict = 256_145
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = NllbTokenizer
SCREAMING_SNAKE_CASE = NllbTokenizerFast
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = {}
def __magic_name__ ( self ) -> int:
'''simple docstring'''
super().setUp()
# We have a SentencePiece fixture for testing
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
tokenizer.save_pretrained(self.tmpdirname )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
__a =tokenizer.tokenize('This is a test' )
self.assertListEqual(__snake_case , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(__snake_case ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
__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 =tokenizer.convert_tokens_to_ids(__snake_case )
self.assertListEqual(
__snake_case , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
__a =tokenizer.convert_ids_to_tokens(__snake_case )
self.assertListEqual(
__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 __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =(self.rust_tokenizer_class, 'hf-internal-testing/tiny-random-nllb', {})
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =self.rust_tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =self.tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tuple(f for f in tokenizer_r_files if 'tokenizer.json' not in f )
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=True
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__snake_case )
# Checks it save with the same files
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=False
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
if not self.test_seqaseq:
return
__a =self.get_tokenizers()
for tokenizer in tokenizers:
with self.subTest(f'{tokenizer.__class__.__name__}' ):
# Longer text that will definitely require truncation.
__a =[
' 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.',
]
__a =[
'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.',
]
try:
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , tgt_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='ron_Latn' , )
except NotImplementedError:
return
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 10 )
# max_target_length will default to max_length if not specified
__a =tokenizer.prepare_seqaseq_batch(
__snake_case , tgt_texts=__snake_case , max_length=3 , return_tensors='pt' )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 3 )
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' )
self.assertEqual(batch_encoder_only.input_ids.shape[1] , 3 )
self.assertEqual(batch_encoder_only.attention_mask.shape[1] , 3 )
self.assertNotIn('decoder_input_ids' , __snake_case )
@unittest.skip('Unfortunately way too slow to build a BPE with SentencePiece.' )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =[AddedToken('<special>' , lstrip=__snake_case )]
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_r.encode('Hey this is a <special> token' )
__a =tokenizer_r.encode('<special>' , add_special_tokens=__snake_case )[0]
self.assertTrue(special_token_id in r_output )
if self.test_slow_tokenizer:
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case , )
__a =self.tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_p.encode('Hey this is a <special> token' )
__a =tokenizer_cr.encode('Hey this is a <special> token' )
self.assertEqual(__snake_case , __snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertTrue(special_token_id in p_output )
self.assertTrue(special_token_id in cr_output )
@require_torch
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = 'facebook/nllb-200-distilled-600M'
SCREAMING_SNAKE_CASE = [
' 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 = [
'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 = [
2_5_6_0_4_7,
1_6_2_9_7,
1_3_4_4_0_8,
8_1_6_5,
2_4_8_0_6_6,
1_4_7_3_4,
9_5_0,
1_1_3_5,
1_0_5_7_2_1,
3_5_7_3,
8_3,
2_7_3_5_2,
1_0_8,
4_9_4_8_6,
2,
]
@classmethod
def __magic_name__ ( cls ) -> Tuple:
'''simple docstring'''
__a =NllbTokenizer.from_pretrained(
cls.checkpoint_name , src_lang='eng_Latn' , tgt_lang='ron_Latn' )
__a =1
return cls
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Arab'] , 25_6001 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Latn'] , 25_6002 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['fra_Latn'] , 25_6057 )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer.batch_encode_plus(self.src_text ).input_ids[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
self.assertIn(__snake_case , self.tokenizer.all_special_ids )
# fmt: off
__a =[RO_CODE, 4254, 9_8068, 11_2923, 3_9072, 3909, 713, 10_2767, 26, 1_7314, 3_5642, 1_4683, 3_3118, 2022, 6_6987, 2, 25_6047]
# fmt: on
__a =self.tokenizer.decode(__snake_case , skip_special_tokens=__snake_case )
__a =self.tokenizer.decode(generated_ids[1:] , skip_special_tokens=__snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertNotIn(self.tokenizer.eos_token , __snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =['this is gunna be a long sentence ' * 20]
assert isinstance(src_text[0] , __snake_case )
__a =10
__a =self.tokenizer(__snake_case , max_length=__snake_case , truncation=__snake_case ).input_ids[0]
self.assertEqual(ids[-1] , 2 )
self.assertEqual(ids[0] , __snake_case )
self.assertEqual(len(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
self.assertListEqual(self.tokenizer.convert_tokens_to_ids(['<mask>', 'ar_AR'] ) , [25_6203, 3] )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =tempfile.mkdtemp()
__a =self.tokenizer.fairseq_tokens_to_ids
self.tokenizer.save_pretrained(__snake_case )
__a =NllbTokenizer.from_pretrained(__snake_case )
self.assertDictEqual(new_tok.fairseq_tokens_to_ids , __snake_case )
@require_torch
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.tokenizer(
self.src_text , text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=len(self.expected_src_tokens ) , return_tensors='pt' , )
__a =shift_tokens_right(
batch['labels'] , self.tokenizer.pad_token_id , self.tokenizer.lang_code_to_id['ron_Latn'] )
self.assertIsInstance(__snake_case , __snake_case )
self.assertEqual((2, 15) , batch.input_ids.shape )
self.assertEqual((2, 15) , batch.attention_mask.shape )
__a =batch.input_ids.tolist()[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
self.assertEqual(__snake_case , batch.decoder_input_ids[0, 0] ) # EOS
# Test that special tokens are reset
self.assertEqual(self.tokenizer.prefix_tokens , [EN_CODE] )
self.assertEqual(self.tokenizer.suffix_tokens , [self.tokenizer.eos_token_id] )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =self.tokenizer(self.src_text , padding=__snake_case , truncation=__snake_case , max_length=3 , return_tensors='pt' )
__a =self.tokenizer(
text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=10 , return_tensors='pt' )
__a =targets['input_ids']
__a =shift_tokens_right(
__snake_case , self.tokenizer.pad_token_id , decoder_start_token_id=self.tokenizer.lang_code_to_id[self.tokenizer.tgt_lang] , )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.decoder_input_ids.shape[1] , 10 )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.tokenizer._build_translation_inputs(
'A test' , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
nested_simplify(__snake_case ) , {
# A, test, EOS, en_XX
'input_ids': [[25_6047, 70, 7356, 2]],
'attention_mask': [[1, 1, 1, 1]],
# ar_AR
'forced_bos_token_id': 25_6057,
} , )
@require_torch
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =True
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2, 25_6047] )
__a =False
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [25_6047, 1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2] )
| 308
| 0
|
from __future__ import annotations
import typing
from collections.abc import Iterable
import numpy as np
_lowerCAmelCase : Optional[int] = typing.Union[Iterable[float], Iterable[int], np.ndarray] # noqa: UP007
_lowerCAmelCase : str = typing.Union[np.floataa, int, float] # noqa: UP007
def UpperCamelCase_( _snake_case : Vector , _snake_case : Vector ):
"""simple docstring"""
return np.sqrt(np.sum((np.asarray(_snake_case ) - np.asarray(_snake_case )) ** 2 ) )
def UpperCamelCase_( _snake_case : Vector , _snake_case : Vector ):
"""simple docstring"""
return sum((va - va) ** 2 for va, va in zip(_snake_case , _snake_case ) ) ** (1 / 2)
if __name__ == "__main__":
def UpperCamelCase_( ):
"""simple docstring"""
from timeit import timeit
print('Without Numpy' )
print(
timeit(
'euclidean_distance_no_np([1, 2, 3], [4, 5, 6])' , number=10000 , globals=globals() , ) )
print('With Numpy' )
print(
timeit(
'euclidean_distance([1, 2, 3], [4, 5, 6])' , number=10000 , globals=globals() , ) )
benchmark()
| 354
|
def UpperCamelCase_( _snake_case : str , _snake_case : int ):
"""simple docstring"""
return [sentence[i : i + ngram_size] for i in range(len(_snake_case ) - ngram_size + 1 )]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 308
| 0
|
import os
import unittest
from transformers import LayoutLMTokenizer, LayoutLMTokenizerFast
from transformers.models.layoutlm.tokenization_layoutlm import VOCAB_FILES_NAMES
from transformers.testing_utils import require_tokenizers
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class __magic_name__ ( _UpperCAmelCase , unittest.TestCase ):
SCREAMING_SNAKE_CASE = LayoutLMTokenizer
SCREAMING_SNAKE_CASE = LayoutLMTokenizerFast
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = True
def __magic_name__ ( self ) -> str:
'''simple docstring'''
super().setUp()
__a =[
"""[UNK]""",
"""[CLS]""",
"""[SEP]""",
"""want""",
"""##want""",
"""##ed""",
"""wa""",
"""un""",
"""runn""",
"""##ing""",
""",""",
"""low""",
"""lowest""",
]
__a =os.path.join(self.tmpdirname , 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] ) )
def __magic_name__ ( self , **__snake_case ) -> List[Any]:
'''simple docstring'''
return LayoutLMTokenizer.from_pretrained(self.tmpdirname , **SCREAMING_SNAKE_CASE_ )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
__a ="""UNwant\u00E9d,running"""
__a ="""unwanted, running"""
return input_text, output_text
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =self.tokenizer_class(self.vocab_file )
__a =tokenizer.tokenize('UNwant\u00E9d,running' )
self.assertListEqual(SCREAMING_SNAKE_CASE_ , ['un', '##want', '##ed', ',', 'runn', '##ing'] )
self.assertListEqual(tokenizer.convert_tokens_to_ids(SCREAMING_SNAKE_CASE_ ) , [7, 4, 5, 10, 8, 9] )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
pass
| 355
|
import argparse
import pytorch_lightning as pl
import torch
from torch import nn
from transformers import LongformerForQuestionAnswering, LongformerModel
class __magic_name__ ( pl.LightningModule ):
def __init__( self , __snake_case ) -> List[Any]:
'''simple docstring'''
super().__init__()
__a =model
__a =2
__a =nn.Linear(self.model.config.hidden_size , self.num_labels )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : str , _snake_case : str , _snake_case : str ):
"""simple docstring"""
__a =LongformerModel.from_pretrained(_snake_case )
__a =LightningModel(_snake_case )
__a =torch.load(_snake_case , map_location=torch.device('cpu' ) )
lightning_model.load_state_dict(ckpt['state_dict'] )
# init longformer question answering model
__a =LongformerForQuestionAnswering.from_pretrained(_snake_case )
# transfer weights
longformer_for_qa.longformer.load_state_dict(lightning_model.model.state_dict() )
longformer_for_qa.qa_outputs.load_state_dict(lightning_model.qa_outputs.state_dict() )
longformer_for_qa.eval()
# save model
longformer_for_qa.save_pretrained(_snake_case )
print(F'Conversion successful. Model saved under {pytorch_dump_folder_path}' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--longformer_model",
default=None,
type=str,
required=True,
help="model identifier of longformer. Should be either `longformer-base-4096` or `longformer-large-4096`.",
)
parser.add_argument(
"--longformer_question_answering_ckpt_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch Lightning Checkpoint.",
)
parser.add_argument(
"--pytorch_dump_folder_path", default=None, type=str, required=True, help="Path to the output PyTorch model."
)
_lowerCAmelCase : Optional[int] = parser.parse_args()
convert_longformer_qa_checkpoint_to_pytorch(
args.longformer_model, args.longformer_question_answering_ckpt_path, args.pytorch_dump_folder_path
)
| 308
| 0
|
import logging
import os
from typing import Dict, List, Optional, Union
import torch
import torch.nn as nn
from accelerate.utils.imports import (
is_abit_bnb_available,
is_abit_bnb_available,
is_bnb_available,
)
from ..big_modeling import dispatch_model, init_empty_weights
from .dataclasses import BnbQuantizationConfig
from .modeling import (
find_tied_parameters,
get_balanced_memory,
infer_auto_device_map,
load_checkpoint_in_model,
offload_weight,
set_module_tensor_to_device,
)
if is_bnb_available():
import bitsandbytes as bnb
from copy import deepcopy
_lowerCAmelCase : Union[str, Any] = logging.getLogger(__name__)
def UpperCamelCase_( _snake_case : torch.nn.Module , _snake_case : BnbQuantizationConfig , _snake_case : Union[str, os.PathLike] = None , _snake_case : Optional[Dict[str, Union[int, str, torch.device]]] = None , _snake_case : Optional[List[str]] = None , _snake_case : Optional[Dict[Union[int, str], Union[int, str]]] = None , _snake_case : Optional[Union[str, os.PathLike]] = None , _snake_case : bool = False , ):
"""simple docstring"""
__a =bnb_quantization_config.load_in_abit
__a =bnb_quantization_config.load_in_abit
if load_in_abit and not is_abit_bnb_available():
raise ImportError(
'You have a version of `bitsandbytes` that is not compatible with 8bit quantization,'
' make sure you have the latest version of `bitsandbytes` installed.' )
if load_in_abit and not is_abit_bnb_available():
raise ValueError(
'You have a version of `bitsandbytes` that is not compatible with 4bit quantization,'
'make sure you have the latest version of `bitsandbytes` installed.' )
__a =[]
# custom device map
if isinstance(lowercase__ , lowercase__ ) and len(device_map.keys() ) > 1:
__a =[key for key, value in device_map.items() if value in ["""disk""", """cpu"""]]
# We keep some modules such as the lm_head in their original dtype for numerical stability reasons
if bnb_quantization_config.skip_modules is None:
__a =get_keys_to_not_convert(lowercase__ )
# add cpu modules to skip modules only for 4-bit modules
if load_in_abit:
bnb_quantization_config.skip_modules.extend(lowercase__ )
__a =bnb_quantization_config.skip_modules
# We add the modules we want to keep in full precision
if bnb_quantization_config.keep_in_fpaa_modules is None:
__a =[]
__a =bnb_quantization_config.keep_in_fpaa_modules
modules_to_not_convert.extend(lowercase__ )
# compatibility with peft
__a =load_in_abit
__a =load_in_abit
__a =get_parameter_device(lowercase__ )
if model_device.type != "meta":
# quantization of an already loaded model
logger.warning(
'It is not recommended to quantize a loaded model. '
'The model should be instantiated under the `init_empty_weights` context manager.' )
__a =replace_with_bnb_layers(lowercase__ , lowercase__ , modules_to_not_convert=lowercase__ )
# convert param to the right dtype
__a =bnb_quantization_config.torch_dtype
for name, param in model.state_dict().items():
if any(module_to_keep_in_fpaa in name for module_to_keep_in_fpaa in keep_in_fpaa_modules ):
param.to(torch.floataa )
if param.dtype != torch.floataa:
__a =name.replace('.weight' , '' ).replace('.bias' , '' )
__a =getattr(lowercase__ , lowercase__ , lowercase__ )
if param is not None:
param.to(torch.floataa )
elif torch.is_floating_point(lowercase__ ):
param.to(lowercase__ )
if model_device.type == "cuda":
# move everything to cpu in the first place because we can't do quantization if the weights are already on cuda
model.cuda(torch.cuda.current_device() )
torch.cuda.empty_cache()
elif torch.cuda.is_available():
model.to(torch.cuda.current_device() )
else:
raise RuntimeError('No GPU found. A GPU is needed for quantization.' )
logger.info(
F'The model device type is {model_device.type}. However, cuda is needed for quantization.'
'We move the model to cuda.' )
return model
elif weights_location is None:
raise RuntimeError(
F'`weights_location` needs to be the folder path containing the weights of the model, but we found {weights_location} ' )
else:
with init_empty_weights():
__a =replace_with_bnb_layers(
lowercase__ , lowercase__ , modules_to_not_convert=lowercase__ )
__a =get_quantized_model_device_map(
lowercase__ , lowercase__ , lowercase__ , max_memory=lowercase__ , no_split_module_classes=lowercase__ , )
if offload_state_dict is None and device_map is not None and "disk" in device_map.values():
__a =True
__a =any(x in list(device_map.values() ) for x in ['cpu', 'disk'] )
load_checkpoint_in_model(
lowercase__ , lowercase__ , lowercase__ , dtype=bnb_quantization_config.torch_dtype , offload_folder=lowercase__ , offload_state_dict=lowercase__ , keep_in_fpaa_modules=bnb_quantization_config.keep_in_fpaa_modules , offload_abit_bnb=load_in_abit and offload , )
return dispatch_model(lowercase__ , device_map=lowercase__ , offload_dir=lowercase__ )
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : Tuple , _snake_case : int=None , _snake_case : int=None , _snake_case : Tuple=None ):
"""simple docstring"""
if device_map is None:
if torch.cuda.is_available():
__a ={"""""": torch.cuda.current_device()}
else:
raise RuntimeError('No GPU found. A GPU is needed for quantization.' )
logger.info('The device_map was not initialized.' 'Setting device_map to `{\'\':torch.cuda.current_device()}`.' )
if isinstance(lowercase__ , lowercase__ ):
if device_map not in ["auto", "balanced", "balanced_low_0", "sequential"]:
raise ValueError(
'If passing a string for `device_map`, please choose \'auto\', \'balanced\', \'balanced_low_0\' or '
'\'sequential\'.' )
__a ={}
special_dtypes.update(
{
name: bnb_quantization_config.torch_dtype
for name, _ in model.named_parameters()
if any(m in name for m in bnb_quantization_config.skip_modules )
} )
special_dtypes.update(
{
name: torch.floataa
for name, _ in model.named_parameters()
if any(m in name for m in bnb_quantization_config.keep_in_fpaa_modules )
} )
__a ={}
__a =special_dtypes
__a =no_split_module_classes
__a =bnb_quantization_config.target_dtype
# get max_memory for each device.
if device_map != "sequential":
__a =get_balanced_memory(
lowercase__ , low_zero=(device_map == 'balanced_low_0') , max_memory=lowercase__ , **lowercase__ , )
__a =max_memory
__a =infer_auto_device_map(lowercase__ , **lowercase__ )
if isinstance(lowercase__ , lowercase__ ):
# check if don't have any quantized module on the cpu
__a =bnb_quantization_config.skip_modules + bnb_quantization_config.keep_in_fpaa_modules
__a ={
key: device_map[key] for key in device_map.keys() if key not in modules_not_to_convert
}
for device in ["cpu", "disk"]:
if device in device_map_without_some_modules.values():
if bnb_quantization_config.load_in_abit:
raise ValueError(
'\n Some modules are dispatched on the CPU or the disk. Make sure you have enough GPU RAM to fit\n the quantized model. If you want to dispatch the model on the CPU or the disk while keeping\n these modules in `torch_dtype`, you need to pass a custom `device_map` to\n `load_and_quantize_model`. Check\n https://huggingface.co/docs/accelerate/main/en/usage_guides/quantization#offload-modules-to-cpu-and-disk\n for more details.\n ' )
else:
logger.info(
'Some modules are are offloaded to the CPU or the disk. Note that these modules will be converted to 8-bit' )
del device_map_without_some_modules
return device_map
def UpperCamelCase_( _snake_case : Optional[int] , _snake_case : Optional[Any] , _snake_case : str=None , _snake_case : Any=None ):
"""simple docstring"""
if modules_to_not_convert is None:
__a =[]
__a =_replace_with_bnb_layers(
lowercase__ , lowercase__ , lowercase__ , lowercase__ )
if not has_been_replaced:
logger.warning(
'You are loading your model in 8bit or 4bit but no linear modules were found in your model.'
' this can happen for some architectures such as gpt2 that uses Conv1D instead of Linear layers.'
' Please double check your model architecture, or submit an issue on github if you think this is'
' a bug.' )
return model
def UpperCamelCase_( _snake_case : Tuple , _snake_case : str , _snake_case : str=None , _snake_case : Union[str, Any]=None , ):
"""simple docstring"""
__a =False
for name, module in model.named_children():
if current_key_name is None:
__a =[]
current_key_name.append(lowercase__ )
if isinstance(lowercase__ , nn.Linear ) and name not in modules_to_not_convert:
# Check if the current key is not in the `modules_to_not_convert`
__a =""".""".join(lowercase__ )
__a =True
for key in modules_to_not_convert:
if (
(key in current_key_name_str) and (key + "." in current_key_name_str)
) or key == current_key_name_str:
__a =False
break
if proceed:
# Load bnb module with empty weight and replace ``nn.Linear` module
if bnb_quantization_config.load_in_abit:
__a =bnb.nn.LinearabitLt(
module.in_features , module.out_features , module.bias is not None , has_fpaa_weights=lowercase__ , threshold=bnb_quantization_config.llm_inta_threshold , )
elif bnb_quantization_config.load_in_abit:
__a =bnb.nn.Linearabit(
module.in_features , module.out_features , module.bias is not None , bnb_quantization_config.bnb_abit_compute_dtype , compress_statistics=bnb_quantization_config.bnb_abit_use_double_quant , quant_type=bnb_quantization_config.bnb_abit_quant_type , )
else:
raise ValueError('load_in_8bit and load_in_4bit can\'t be both False' )
__a =module.weight.data
if module.bias is not None:
__a =module.bias.data
bnb_module.requires_grad_(lowercase__ )
setattr(lowercase__ , lowercase__ , lowercase__ )
__a =True
if len(list(module.children() ) ) > 0:
__a =_replace_with_bnb_layers(
lowercase__ , lowercase__ , lowercase__ , lowercase__ )
__a =has_been_replaced | _has_been_replaced
# Remove the last key for recursion
current_key_name.pop(-1 )
return model, has_been_replaced
def UpperCamelCase_( _snake_case : Union[str, Any] ):
"""simple docstring"""
with init_empty_weights():
__a =deepcopy(lowercase__ ) # this has 0 cost since it is done inside `init_empty_weights` context manager`
__a =find_tied_parameters(lowercase__ )
# For compatibility with Accelerate < 0.18
if isinstance(lowercase__ , lowercase__ ):
__a =sum(list(tied_params.values() ) , [] ) + list(tied_params.keys() )
else:
__a =sum(lowercase__ , [] )
__a =len(lowercase__ ) > 0
# Check if it is a base model
__a =False
if hasattr(lowercase__ , 'base_model_prefix' ):
__a =not hasattr(lowercase__ , model.base_model_prefix )
# Ignore this for base models (BertModel, GPT2Model, etc.)
if (not has_tied_params) and is_base_model:
return []
# otherwise they have an attached head
__a =list(model.named_children() )
__a =[list_modules[-1][0]]
# add last module together with tied weights
__a =set(lowercase__ ) - set(lowercase__ )
__a =list(set(lowercase__ ) ) + list(lowercase__ )
# remove ".weight" from the keys
__a =[""".weight""", """.bias"""]
__a =[]
for name in list_untouched:
for name_to_remove in names_to_remove:
if name_to_remove in name:
__a =name.replace(lowercase__ , '' )
filtered_module_names.append(lowercase__ )
return filtered_module_names
def UpperCamelCase_( _snake_case : List[str] ):
"""simple docstring"""
for m in model.modules():
if isinstance(lowercase__ , bnb.nn.Linearabit ):
return True
return False
def UpperCamelCase_( _snake_case : nn.Module ):
"""simple docstring"""
return next(parameter.parameters() ).device
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Tuple , _snake_case : str , _snake_case : int , _snake_case : Tuple , _snake_case : List[str] , _snake_case : List[str] ):
"""simple docstring"""
if fpaa_statistics is None:
set_module_tensor_to_device(lowercase__ , lowercase__ , 0 , dtype=lowercase__ , value=lowercase__ )
__a =param_name
__a =model
if "." in tensor_name:
__a =tensor_name.split('.' )
for split in splits[:-1]:
__a =getattr(lowercase__ , lowercase__ )
if new_module is None:
raise ValueError(F'{module} has no attribute {split}.' )
__a =new_module
__a =splits[-1]
# offload weights
__a =False
offload_weight(module._parameters[tensor_name] , lowercase__ , lowercase__ , index=lowercase__ )
if hasattr(module._parameters[tensor_name] , 'SCB' ):
offload_weight(
module._parameters[tensor_name].SCB , param_name.replace('weight' , 'SCB' ) , lowercase__ , index=lowercase__ , )
else:
offload_weight(lowercase__ , lowercase__ , lowercase__ , index=lowercase__ )
offload_weight(lowercase__ , param_name.replace('weight' , 'SCB' ) , lowercase__ , index=lowercase__ )
set_module_tensor_to_device(lowercase__ , lowercase__ , 'meta' , dtype=lowercase__ , value=torch.empty(*param.size() ) )
| 356
|
import logging
import os
import random
import sys
from dataclasses import dataclass, field
from typing import Optional
import datasets
import numpy as np
import pandas as pd
from datasets import load_dataset
import transformers
from transformers import (
AutoConfig,
BartForSequenceClassification,
DataCollatorWithPadding,
EvalPrediction,
HfArgumentParser,
TapexTokenizer,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
_lowerCAmelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The name of the dataset to use (via the datasets library).'} )
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} , )
SCREAMING_SNAKE_CASE = field(
default=1_0_2_4 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Overwrite the cached preprocessed datasets or not.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Whether to pad all samples to `max_seq_length`. '
'If False, will pad the samples dynamically when batching to the maximum length in the batch.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of training examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of evaluation examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of prediction examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = field(default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the test data.'} )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
if self.dataset_name is not None:
pass
elif self.train_file is None or self.validation_file is None:
raise ValueError('Need either a GLUE task, a training/validation file or a dataset name.' )
else:
__a =self.train_file.split('.' )[-1]
assert train_extension in ["csv", "json"], "`train_file` should be a csv or a json file."
__a =self.validation_file.split('.' )[-1]
assert (
validation_extension == train_extension
), "`validation_file` should have the same extension (csv or json) as `train_file`."
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to use one of the fast tokenizer (backed by the tokenizers library) or not.'} , )
SCREAMING_SNAKE_CASE = field(
default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Will use the token generated when running `huggingface-cli login` (necessary to use this script '
'with private models).'
)
} , )
def UpperCamelCase_( ):
"""simple docstring"""
__a =HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments) )
if len(sys.argv ) == 2 and sys.argv[1].endswith('.json' ):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
__a , __a , __a =parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) )
else:
__a , __a , __a =parser.parse_args_into_dataclasses()
# Setup logging
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , handlers=[logging.StreamHandler(sys.stdout )] , )
__a =training_args.get_process_log_level()
logger.setLevel(_snake_case )
datasets.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Log on each process the small summary:
logger.warning(
F'Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}'
+ F'distributed training: {bool(training_args.local_rank != -1 )}, 16-bits training: {training_args.fpaa}' )
logger.info(F'Training/evaluation parameters {training_args}' )
# Detecting last checkpoint.
__a =None
if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir:
__a =get_last_checkpoint(training_args.output_dir )
if last_checkpoint is None and len(os.listdir(training_args.output_dir ) ) > 0:
raise ValueError(
F'Output directory ({training_args.output_dir}) already exists and is not empty. '
'Use --overwrite_output_dir to overcome.' )
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
F'Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change '
'the `--output_dir` or add `--overwrite_output_dir` to train from scratch.' )
# Set seed before initializing model.
set_seed(training_args.seed )
# Get the datasets: you can either provide your own CSV/JSON training and evaluation files (see below)
# or specify a GLUE benchmark task (the dataset will be downloaded automatically from the datasets Hub).
#
# For JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# If the CSVs/JSONs contain only one non-label column, the script does single sentence classification on this
# single column. You can easily tweak this behavior (see below)
#
# In distributed training, the load_dataset function guarantee that only one local process can concurrently
# download the dataset.
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
__a =load_dataset(
data_args.dataset_name , data_args.dataset_config_name , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from your local files.
# CSV/JSON training and evaluation files are needed.
__a ={'train': data_args.train_file, 'validation': data_args.validation_file}
# Get the test dataset: you can provide your own CSV/JSON test file (see below)
# when you use `do_predict` without specifying a GLUE benchmark task.
if training_args.do_predict:
if data_args.test_file is not None:
__a =data_args.train_file.split('.' )[-1]
__a =data_args.test_file.split('.' )[-1]
assert (
test_extension == train_extension
), "`test_file` should have the same extension (csv or json) as `train_file`."
__a =data_args.test_file
else:
raise ValueError('Need either a GLUE task or a test file for `do_predict`.' )
for key in data_files.keys():
logger.info(F'load a local file for {key}: {data_files[key]}' )
if data_args.train_file.endswith('.csv' ):
# Loading a dataset from local csv files
__a =load_dataset('csv' , data_files=_snake_case , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from local json files
__a =load_dataset('json' , data_files=_snake_case , cache_dir=model_args.cache_dir )
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.html.
# Labels
__a =raw_datasets['train'].features['label'].names
__a =len(_snake_case )
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
__a =AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# load tapex tokenizer
__a =TapexTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , use_fast=model_args.use_fast_tokenizer , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , add_prefix_space=_snake_case , )
__a =BartForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_tf=bool('.ckpt' in model_args.model_name_or_path ) , config=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# Padding strategy
if data_args.pad_to_max_length:
__a ='max_length'
else:
# We will pad later, dynamically at batch creation, to the max sequence length in each batch
__a =False
# Some models have set the order of the labels to use, so let's make sure we do use it.
__a ={'Refused': 0, 'Entailed': 1}
__a ={0: 'Refused', 1: 'Entailed'}
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warning(
F'The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the'
F'model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}.' )
__a =min(data_args.max_seq_length , tokenizer.model_max_length )
def preprocess_tabfact_function(_snake_case : Tuple ):
# Tokenize the texts
def _convert_table_text_to_pandas(_snake_case : Optional[Any] ):
__a =[_table_row.split('#' ) for _table_row in _table_text.strip('\n' ).split('\n' )]
__a =pd.DataFrame.from_records(_table_content[1:] , columns=_table_content[0] )
return _table_pd
__a =examples['statement']
__a =list(map(_convert_table_text_to_pandas , examples['table_text'] ) )
__a =tokenizer(_snake_case , _snake_case , padding=_snake_case , max_length=_snake_case , truncation=_snake_case )
__a =examples['label']
return result
with training_args.main_process_first(desc='dataset map pre-processing' ):
__a =raw_datasets.map(
_snake_case , batched=_snake_case , load_from_cache_file=not data_args.overwrite_cache , desc='Running tokenizer on dataset' , )
if training_args.do_train:
if "train" not in raw_datasets:
raise ValueError('--do_train requires a train dataset' )
__a =raw_datasets['train']
if data_args.max_train_samples is not None:
__a =train_dataset.select(range(data_args.max_train_samples ) )
if training_args.do_eval:
if "validation" not in raw_datasets and "validation_matched" not in raw_datasets:
raise ValueError('--do_eval requires a validation dataset' )
__a =raw_datasets['validation']
if data_args.max_eval_samples is not None:
__a =eval_dataset.select(range(data_args.max_eval_samples ) )
if training_args.do_predict or data_args.test_file is not None:
if "test" not in raw_datasets and "test_matched" not in raw_datasets:
raise ValueError('--do_predict requires a test dataset' )
__a =raw_datasets['test']
if data_args.max_predict_samples is not None:
__a =predict_dataset.select(range(data_args.max_predict_samples ) )
# Log a few random samples from the training set:
if training_args.do_train:
for index in random.sample(range(len(_snake_case ) ) , 3 ):
logger.info(F'Sample {index} of the training set: {train_dataset[index]}.' )
# You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
# predictions and label_ids field) and has to return a dictionary string to float.
def compute_metrics(_snake_case : EvalPrediction ):
__a =p.predictions[0] if isinstance(p.predictions , _snake_case ) else p.predictions
__a =np.argmax(_snake_case , axis=1 )
return {"accuracy": (preds == p.label_ids).astype(np.floataa ).mean().item()}
# Data collator will default to DataCollatorWithPadding, so we change it if we already did the padding.
if data_args.pad_to_max_length:
__a =default_data_collator
elif training_args.fpaa:
__a =DataCollatorWithPadding(_snake_case , pad_to_multiple_of=8 )
else:
__a =None
# Initialize our Trainer
__a =Trainer(
model=_snake_case , args=_snake_case , train_dataset=train_dataset if training_args.do_train else None , eval_dataset=eval_dataset if training_args.do_eval else None , compute_metrics=_snake_case , tokenizer=_snake_case , data_collator=_snake_case , )
# Training
if training_args.do_train:
__a =None
if training_args.resume_from_checkpoint is not None:
__a =training_args.resume_from_checkpoint
elif last_checkpoint is not None:
__a =last_checkpoint
__a =trainer.train(resume_from_checkpoint=_snake_case )
__a =train_result.metrics
__a =(
data_args.max_train_samples if data_args.max_train_samples is not None else len(_snake_case )
)
__a =min(_snake_case , len(_snake_case ) )
trainer.save_model() # Saves the tokenizer too for easy upload
trainer.log_metrics('train' , _snake_case )
trainer.save_metrics('train' , _snake_case )
trainer.save_state()
# Evaluation
if training_args.do_eval:
logger.info('*** Evaluate ***' )
__a =trainer.evaluate(eval_dataset=_snake_case )
__a =data_args.max_eval_samples if data_args.max_eval_samples is not None else len(_snake_case )
__a =min(_snake_case , len(_snake_case ) )
trainer.log_metrics('eval' , _snake_case )
trainer.save_metrics('eval' , _snake_case )
if training_args.do_predict:
logger.info('*** Predict ***' )
# Removing the `label` columns because it contains -1 and Trainer won't like that.
__a =predict_dataset.remove_columns('label' )
__a =trainer.predict(_snake_case , metric_key_prefix='predict' ).predictions
__a =np.argmax(_snake_case , axis=1 )
__a =os.path.join(training_args.output_dir , 'predict_results_tabfact.txt' )
if trainer.is_world_process_zero():
with open(_snake_case , 'w' ) as writer:
logger.info('***** Predict Results *****' )
writer.write('index\tprediction\n' )
for index, item in enumerate(_snake_case ):
__a =label_list[item]
writer.write(F'{index}\t{item}\n' )
__a ={'finetuned_from': model_args.model_name_or_path, 'tasks': 'text-classification'}
if training_args.push_to_hub:
trainer.push_to_hub(**_snake_case )
else:
trainer.create_model_card(**_snake_case )
def UpperCamelCase_( _snake_case : Union[str, Any] ):
"""simple docstring"""
main()
if __name__ == "__main__":
main()
| 308
| 0
|
import random
import unittest
import numpy as np
import torch
from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer
from diffusers import (
AutoencoderKL,
DDIMScheduler,
UNetaDConditionModel,
VideoToVideoSDPipeline,
)
from diffusers.utils import floats_tensor, is_xformers_available, skip_mps
from diffusers.utils.testing_utils import enable_full_determinism, slow, torch_device
from ..pipeline_params import (
TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS,
TEXT_GUIDED_IMAGE_VARIATION_PARAMS,
)
from ..test_pipelines_common import PipelineTesterMixin
enable_full_determinism()
@skip_mps
class __magic_name__ ( __lowerCamelCase , unittest.TestCase ):
SCREAMING_SNAKE_CASE = VideoToVideoSDPipeline
SCREAMING_SNAKE_CASE = TEXT_GUIDED_IMAGE_VARIATION_PARAMS.union({'video'} ) - {'image', 'width', 'height'}
SCREAMING_SNAKE_CASE = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS.union({'video'} ) - {'image'}
SCREAMING_SNAKE_CASE = PipelineTesterMixin.required_optional_params - {'latents'}
SCREAMING_SNAKE_CASE = False
# No `output_type`.
SCREAMING_SNAKE_CASE = frozenset(
[
'num_inference_steps',
'generator',
'latents',
'return_dict',
'callback',
'callback_steps',
] )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
torch.manual_seed(0 )
__a =UNetaDConditionModel(
block_out_channels=(32, 64, 64, 64) , layers_per_block=2 , sample_size=32 , in_channels=4 , out_channels=4 , down_block_types=('CrossAttnDownBlock3D', 'CrossAttnDownBlock3D', 'CrossAttnDownBlock3D', 'DownBlock3D') , up_block_types=('UpBlock3D', 'CrossAttnUpBlock3D', 'CrossAttnUpBlock3D', 'CrossAttnUpBlock3D') , cross_attention_dim=32 , attention_head_dim=4 , )
__a =DDIMScheduler(
beta_start=0.0_0085 , beta_end=0.012 , beta_schedule='scaled_linear' , clip_sample=UpperCamelCase_ , set_alpha_to_one=UpperCamelCase_ , )
torch.manual_seed(0 )
__a =AutoencoderKL(
block_out_channels=[32, 64] , in_channels=3 , out_channels=3 , down_block_types=['DownEncoderBlock2D', 'DownEncoderBlock2D'] , up_block_types=['UpDecoderBlock2D', 'UpDecoderBlock2D'] , latent_channels=4 , sample_size=128 , )
torch.manual_seed(0 )
__a =CLIPTextConfig(
bos_token_id=0 , eos_token_id=2 , hidden_size=32 , intermediate_size=37 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , hidden_act='gelu' , projection_dim=512 , )
__a =CLIPTextModel(UpperCamelCase_ )
__a =CLIPTokenizer.from_pretrained('hf-internal-testing/tiny-random-clip' )
__a ={
'unet': unet,
'scheduler': scheduler,
'vae': vae,
'text_encoder': text_encoder,
'tokenizer': tokenizer,
}
return components
def __magic_name__ ( self , __snake_case , __snake_case=0 ) -> str:
'''simple docstring'''
# 3 frames
__a =floats_tensor((1, 3, 3, 32, 32) , rng=random.Random(UpperCamelCase_ ) ).to(UpperCamelCase_ )
if str(UpperCamelCase_ ).startswith('mps' ):
__a =torch.manual_seed(UpperCamelCase_ )
else:
__a =torch.Generator(device=UpperCamelCase_ ).manual_seed(UpperCamelCase_ )
__a ={
'prompt': 'A painting of a squirrel eating a burger',
'video': video,
'generator': generator,
'num_inference_steps': 2,
'guidance_scale': 6.0,
'output_type': 'pt',
}
return inputs
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a ='cpu' # ensure determinism for the device-dependent torch.Generator
__a =self.get_dummy_components()
__a =VideoToVideoSDPipeline(**UpperCamelCase_ )
__a =sd_pipe.to(UpperCamelCase_ )
sd_pipe.set_progress_bar_config(disable=UpperCamelCase_ )
__a =self.get_dummy_inputs(UpperCamelCase_ )
__a ='np'
__a =sd_pipe(**UpperCamelCase_ ).frames
__a =frames[0][-3:, -3:, -1]
assert frames[0].shape == (32, 32, 3)
__a =np.array([106, 117, 113, 174, 137, 112, 148, 151, 131] )
assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2
@unittest.skipIf(
torch_device != 'cuda' or not is_xformers_available() , reason='XFormers attention is only available with CUDA and `xformers` installed' , )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
self._test_xformers_attention_forwardGenerator_pass(test_mean_pixel_difference=UpperCamelCase_ , expected_max_diff=5e-3 )
@unittest.skip(reason='Batching needs to be properly figured out first for this pipeline.' )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
pass
@unittest.skip(reason='Batching needs to be properly figured out first for this pipeline.' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
pass
@unittest.skip(reason='`num_images_per_prompt` argument is not supported for this pipeline.' )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
return super().test_progress_bar()
@slow
@skip_mps
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =VideoToVideoSDPipeline.from_pretrained('cerspense/zeroscope_v2_XL' , torch_dtype=torch.floataa )
pipe.enable_model_cpu_offload()
# 10 frames
__a =torch.Generator(device='cpu' ).manual_seed(0 )
__a =torch.randn((1, 10, 3, 1024, 576) , generator=UpperCamelCase_ )
__a =video.to('cuda' )
__a ='Spiderman is surfing'
__a =pipe(UpperCamelCase_ , video=UpperCamelCase_ , generator=UpperCamelCase_ , num_inference_steps=3 , output_type='pt' ).frames
__a =np.array([-1.045_8984, -1.127_9297, -0.966_3086, -0.9150_3906, -0.7509_7656] )
assert np.abs(video_frames.cpu().numpy()[0, 0, 0, 0, -5:] - expected_array ).sum() < 1e-2
| 357
|
from __future__ import annotations
import time
import numpy as np
_lowerCAmelCase : List[str] = [8, 5, 9, 7]
_lowerCAmelCase : List[str] = [
[2, 0, 1, 1],
[0, 1, 2, 1],
[4, 0, 0, 3],
[0, 2, 1, 0],
[1, 0, 3, 0],
]
_lowerCAmelCase : List[Any] = [
[3, 2, 1, 4],
[0, 2, 5, 2],
[5, 1, 0, 5],
[1, 5, 3, 0],
[3, 0, 3, 3],
]
class __magic_name__ :
def __init__( self , __snake_case , __snake_case , __snake_case , ) -> None:
'''simple docstring'''
__a =claim_vector
__a =allocated_resources_table
__a =maximum_claim_table
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return [
sum(p_item[i] for p_item in self.__allocated_resources_table )
for i in range(len(self.__allocated_resources_table[0] ) )
]
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return np.array(self.__claim_vector ) - np.array(
self.__processes_resource_summation() )
def __magic_name__ ( self ) -> list[list[int]]:
'''simple docstring'''
return [
list(np.array(self.__maximum_claim_table[i] ) - np.array(__snake_case ) )
for i, allocated_resource in enumerate(self.__allocated_resources_table )
]
def __magic_name__ ( self ) -> dict[int, list[int]]:
'''simple docstring'''
return {self.__need().index(__snake_case ): i for i in self.__need()}
def __magic_name__ ( self , **__snake_case ) -> None:
'''simple docstring'''
__a =self.__need()
__a =self.__allocated_resources_table
__a =self.__available_resources()
__a =self.__need_index_manager()
for kw, val in kwargs.items():
if kw and val is True:
self.__pretty_data()
print('_' * 50 + '\n' )
while need_list:
__a =False
for each_need in need_list:
__a =True
for index, need in enumerate(__snake_case ):
if need > available_resources[index]:
__a =False
break
if execution:
__a =True
# get the original index of the process from ind_ctrl db
for original_need_index, need_clone in need_index_manager.items():
if each_need == need_clone:
__a =original_need_index
print(f'Process {process_number + 1} is executing.' )
# remove the process run from stack
need_list.remove(__snake_case )
# update available/freed resources stack
__a =np.array(__snake_case ) + np.array(
alloc_resources_table[process_number] )
print(
'Updated available resource stack for processes: '
+ ' '.join([str(__snake_case ) for x in available_resources] ) )
break
if safe:
print('The process is in a safe state.\n' )
else:
print('System in unsafe state. Aborting...\n' )
break
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
print(' ' * 9 + 'Allocated Resource Table' )
for item in self.__allocated_resources_table:
print(
f'P{self.__allocated_resources_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(' ' * 9 + 'System Resource Table' )
for item in self.__maximum_claim_table:
print(
f'P{self.__maximum_claim_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(
'Current Usage by Active Processes: '
+ ' '.join(str(__snake_case ) for x in self.__claim_vector ) )
print(
'Initial Available Resources: '
+ ' '.join(str(__snake_case ) for x in self.__available_resources() ) )
time.sleep(1 )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 0
|
"""simple docstring"""
import math
def UpperCamelCase_( _snake_case : List[Any] ):
"""simple docstring"""
__a =[True] * n
__a =False
__a =False
__a =True
for i in range(3 , int(n**0.5 + 1 ) , 2 ):
__a =i * 2
while index < n:
__a =False
__a =index + i
__a =[2]
for i in range(3 , _snake_case , 2 ):
if is_prime[i]:
primes.append(_snake_case )
return primes
def UpperCamelCase_( _snake_case : Union[str, Any] = 999966663333 ):
"""simple docstring"""
__a =math.floor(math.sqrt(_snake_case ) ) + 100
__a =prime_sieve(_snake_case )
__a =0
__a =0
__a =primes[prime_index]
while (last_prime**2) <= limit:
__a =primes[prime_index + 1]
__a =last_prime**2
__a =next_prime**2
# Get numbers divisible by lps(current)
__a =lower_bound + last_prime
while upper_bound > current <= limit:
matches_sum += current
current += last_prime
# Reset the upper_bound
while (upper_bound - next_prime) > limit:
upper_bound -= next_prime
# Add the numbers divisible by ups(current)
__a =upper_bound - next_prime
while current > lower_bound:
matches_sum += current
current -= next_prime
# Remove the numbers divisible by both ups and lps
__a =0
while upper_bound > current <= limit:
if current <= lower_bound:
# Increment the current number
current += last_prime * next_prime
continue
if current > limit:
break
# Remove twice since it was added by both ups and lps
matches_sum -= current * 2
# Increment the current number
current += last_prime * next_prime
# Setup for next pair
__a =next_prime
prime_index += 1
return matches_sum
if __name__ == "__main__":
print(solution())
| 358
|
import string
# frequency taken from https://en.wikipedia.org/wiki/Letter_frequency
_lowerCAmelCase : Tuple = {
"E": 12.70,
"T": 9.06,
"A": 8.17,
"O": 7.51,
"I": 6.97,
"N": 6.75,
"S": 6.33,
"H": 6.09,
"R": 5.99,
"D": 4.25,
"L": 4.03,
"C": 2.78,
"U": 2.76,
"M": 2.41,
"W": 2.36,
"F": 2.23,
"G": 2.02,
"Y": 1.97,
"P": 1.93,
"B": 1.29,
"V": 0.98,
"K": 0.77,
"J": 0.15,
"X": 0.15,
"Q": 0.10,
"Z": 0.07,
}
_lowerCAmelCase : Optional[int] = "ETAOINSHRDLCUMWFGYPBVKJXQZ"
_lowerCAmelCase : Any = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a ={letter: 0 for letter in string.ascii_uppercase}
for letter in message.upper():
if letter in LETTERS:
letter_count[letter] += 1
return letter_count
def UpperCamelCase_( _snake_case : tuple ):
"""simple docstring"""
return x[0]
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_letter_count(_snake_case )
__a ={
freq: [] for letter, freq in letter_to_freq.items()
}
for letter in LETTERS:
freq_to_letter[letter_to_freq[letter]].append(_snake_case )
__a ={}
for freq in freq_to_letter:
freq_to_letter[freq].sort(key=ETAOIN.find , reverse=_snake_case )
__a =''.join(freq_to_letter[freq] )
__a =list(freq_to_letter_str.items() )
freq_pairs.sort(key=_snake_case , reverse=_snake_case )
__a =[freq_pair[1] for freq_pair in freq_pairs]
return "".join(_snake_case )
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_frequency_order(_snake_case )
__a =0
for common_letter in ETAOIN[:6]:
if common_letter in freq_order[:6]:
match_score += 1
for uncommon_letter in ETAOIN[-6:]:
if uncommon_letter in freq_order[-6:]:
match_score += 1
return match_score
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 0
|
from __future__ import annotations
import string
from itertools import cycle, product
from pathlib import Path
_lowerCAmelCase : str = (
string.ascii_letters + string.digits + string.punctuation + string.whitespace
)
_lowerCAmelCase : list[int] = [ord(letter) for letter in string.ascii_lowercase]
_lowerCAmelCase : set[int] = {ord(char) for char in VALID_CHARS}
_lowerCAmelCase : list[str] = ["the", "be", "to", "of", "and", "in", "that", "have"]
def UpperCamelCase_( _snake_case : list[int] , _snake_case : tuple[int, ...] ):
"""simple docstring"""
__a =''
__a =42
__a =42
__a =42
for keychar, cipherchar in zip(cycle(_UpperCamelCase ) , _UpperCamelCase ):
__a =cipherchar ^ keychar
if decodedchar not in VALID_INTS:
return None
decoded += chr(_UpperCamelCase )
return decoded
def UpperCamelCase_( _snake_case : list[int] ):
"""simple docstring"""
__a =[]
for key in product(_UpperCamelCase , repeat=3 ):
__a =try_key(_UpperCamelCase , _UpperCamelCase )
if encoded is not None:
possibles.append(_UpperCamelCase )
return possibles
def UpperCamelCase_( _snake_case : list[str] , _snake_case : str ):
"""simple docstring"""
return [possible for possible in possibles if common_word in possible.lower()]
def UpperCamelCase_( _snake_case : str = "p059_cipher.txt" ):
"""simple docstring"""
__a =42
__a =42
__a =42
__a =42
__a =Path(_UpperCamelCase ).parent.joinpath(_UpperCamelCase ).read_text(encoding='utf-8' )
__a =[int(_UpperCamelCase ) for number in data.strip().split(',' )]
__a =filter_valid_chars(_UpperCamelCase )
for common_word in COMMON_WORDS:
__a =filter_common_word(_UpperCamelCase , _UpperCamelCase )
if len(_UpperCamelCase ) == 1:
break
__a =possibles[0]
return sum(ord(_UpperCamelCase ) for char in decoded_text )
if __name__ == "__main__":
print(f'''{solution() = }''')
| 359
|
from ...configuration_utils import PretrainedConfig
from ...utils import logging
_lowerCAmelCase : Any = logging.get_logger(__name__)
_lowerCAmelCase : int = {
"caidas/swin2sr-classicalsr-x2-64": (
"https://huggingface.co/caidas/swin2sr-classicalsr-x2-64/resolve/main/config.json"
),
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'swin2sr'
SCREAMING_SNAKE_CASE = {
'hidden_size': 'embed_dim',
'num_attention_heads': 'num_heads',
'num_hidden_layers': 'num_layers',
}
def __init__( self , __snake_case=64 , __snake_case=1 , __snake_case=3 , __snake_case=180 , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=8 , __snake_case=2.0 , __snake_case=True , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.1 , __snake_case="gelu" , __snake_case=False , __snake_case=0.02 , __snake_case=1e-5 , __snake_case=2 , __snake_case=1.0 , __snake_case="1conv" , __snake_case="pixelshuffle" , **__snake_case , ) -> Dict:
'''simple docstring'''
super().__init__(**__snake_case )
__a =image_size
__a =patch_size
__a =num_channels
__a =embed_dim
__a =depths
__a =len(__snake_case )
__a =num_heads
__a =window_size
__a =mlp_ratio
__a =qkv_bias
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =drop_path_rate
__a =hidden_act
__a =use_absolute_embeddings
__a =layer_norm_eps
__a =initializer_range
__a =upscale
__a =img_range
__a =resi_connection
__a =upsampler
| 308
| 0
|
import itertools
import string
from collections.abc import Generator, Iterable
def UpperCamelCase_( _snake_case : Iterable[str] , _snake_case : int ):
"""simple docstring"""
__a =iter(lowerCAmelCase__ )
while True:
__a =tuple(itertools.islice(lowerCAmelCase__ , lowerCAmelCase__ ) )
if not chunk:
return
yield chunk
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =''.join([c.upper() for c in dirty if c in string.ascii_letters] )
__a =''
if len(lowerCAmelCase__ ) < 2:
return dirty
for i in range(len(lowerCAmelCase__ ) - 1 ):
clean += dirty[i]
if dirty[i] == dirty[i + 1]:
clean += "X"
clean += dirty[-1]
if len(lowerCAmelCase__ ) & 1:
clean += "X"
return clean
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a ='ABCDEFGHIKLMNOPQRSTUVWXYZ'
# we're using a list instead of a '2d' array because it makes the math
# for setting up the table and doing the actual encoding/decoding simpler
__a =[]
# copy key chars into the table if they are in `alphabet` ignoring duplicates
for char in key.upper():
if char not in table and char in alphabet:
table.append(lowerCAmelCase__ )
# fill the rest of the table in with the remaining alphabet chars
for char in alphabet:
if char not in table:
table.append(lowerCAmelCase__ )
return table
def UpperCamelCase_( _snake_case : str , _snake_case : str ):
"""simple docstring"""
__a =generate_table(lowerCAmelCase__ )
__a =prepare_input(lowerCAmelCase__ )
__a =''
# https://en.wikipedia.org/wiki/Playfair_cipher#Description
for chara, chara in chunker(lowerCAmelCase__ , 2 ):
__a , __a =divmod(table.index(lowerCAmelCase__ ) , 5 )
__a , __a =divmod(table.index(lowerCAmelCase__ ) , 5 )
if rowa == rowa:
ciphertext += table[rowa * 5 + (cola + 1) % 5]
ciphertext += table[rowa * 5 + (cola + 1) % 5]
elif cola == cola:
ciphertext += table[((rowa + 1) % 5) * 5 + cola]
ciphertext += table[((rowa + 1) % 5) * 5 + cola]
else: # rectangle
ciphertext += table[rowa * 5 + cola]
ciphertext += table[rowa * 5 + cola]
return ciphertext
def UpperCamelCase_( _snake_case : str , _snake_case : str ):
"""simple docstring"""
__a =generate_table(lowerCAmelCase__ )
__a =''
# https://en.wikipedia.org/wiki/Playfair_cipher#Description
for chara, chara in chunker(lowerCAmelCase__ , 2 ):
__a , __a =divmod(table.index(lowerCAmelCase__ ) , 5 )
__a , __a =divmod(table.index(lowerCAmelCase__ ) , 5 )
if rowa == rowa:
plaintext += table[rowa * 5 + (cola - 1) % 5]
plaintext += table[rowa * 5 + (cola - 1) % 5]
elif cola == cola:
plaintext += table[((rowa - 1) % 5) * 5 + cola]
plaintext += table[((rowa - 1) % 5) * 5 + cola]
else: # rectangle
plaintext += table[rowa * 5 + cola]
plaintext += table[rowa * 5 + cola]
return plaintext
| 360
|
import os
def UpperCamelCase_( _snake_case : str = "input.txt" ):
"""simple docstring"""
with open(os.path.join(os.path.dirname(_snake_case ) , _snake_case ) ) as input_file:
__a =[
[int(_snake_case ) for element in line.split(',' )]
for line in input_file.readlines()
]
__a =len(_snake_case )
__a =len(matrix[0] )
__a =[[-1 for _ in range(_snake_case )] for _ in range(_snake_case )]
for i in range(_snake_case ):
__a =matrix[i][0]
for j in range(1 , _snake_case ):
for i in range(_snake_case ):
__a =minimal_path_sums[i][j - 1] + matrix[i][j]
for i in range(1 , _snake_case ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i - 1][j] + matrix[i][j] )
for i in range(rows - 2 , -1 , -1 ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i + 1][j] + matrix[i][j] )
return min(minimal_path_sums_row[-1] for minimal_path_sums_row in minimal_path_sums )
if __name__ == "__main__":
print(f'''{solution() = }''')
| 308
| 0
|
import os
# Precomputes a list of the 100 first triangular numbers
_lowerCAmelCase : List[Any] = [int(0.5 * n * (n + 1)) for n in range(1, 101)]
def UpperCamelCase_( ):
"""simple docstring"""
__a =os.path.dirname(os.path.realpath(a__ ) )
__a =os.path.join(a__ , 'words.txt' )
__a =''
with open(a__ ) as f:
__a =f.readline()
__a =[word.strip('"' ) for word in words.strip('\r\n' ).split(',' )]
__a =[
word
for word in [sum(ord(a__ ) - 64 for x in word ) for word in words]
if word in TRIANGULAR_NUMBERS
]
return len(a__ )
if __name__ == "__main__":
print(solution())
| 361
|
import warnings
from ...utils import logging
from .image_processing_mobilevit import MobileViTImageProcessor
_lowerCAmelCase : Any = logging.get_logger(__name__)
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , *__snake_case , **__snake_case ) -> None:
'''simple docstring'''
warnings.warn(
'The class MobileViTFeatureExtractor is deprecated and will be removed in version 5 of Transformers.'
' Please use MobileViTImageProcessor instead.' , __snake_case , )
super().__init__(*__snake_case , **__snake_case )
| 308
| 0
|
import numpy as np
import pandas as pd
from sklearn.preprocessing import MinMaxScaler
from tensorflow.keras.layers import LSTM, Dense
from tensorflow.keras.models import Sequential
if __name__ == "__main__":
_lowerCAmelCase : Optional[Any] = pd.read_csv("sample_data.csv", header=None)
_lowerCAmelCase : Optional[Any] = df.shape[:1][0]
# If you're using some other dataset input the target column
_lowerCAmelCase : Any = df.iloc[:, 1:2]
_lowerCAmelCase : Union[str, Any] = actual_data.values.reshape(len_data, 1)
_lowerCAmelCase : Optional[Any] = MinMaxScaler().fit_transform(actual_data)
_lowerCAmelCase : Tuple = 10
_lowerCAmelCase : List[str] = 5
_lowerCAmelCase : int = 20
_lowerCAmelCase : Union[str, Any] = len_data - periods * look_back
_lowerCAmelCase : int = actual_data[:division]
_lowerCAmelCase : str = actual_data[division - look_back :]
_lowerCAmelCase , _lowerCAmelCase : Dict = [], []
_lowerCAmelCase , _lowerCAmelCase : Any = [], []
for i in range(0, len(train_data) - forward_days - look_back + 1):
train_x.append(train_data[i : i + look_back])
train_y.append(train_data[i + look_back : i + look_back + forward_days])
for i in range(0, len(test_data) - forward_days - look_back + 1):
test_x.append(test_data[i : i + look_back])
test_y.append(test_data[i + look_back : i + look_back + forward_days])
_lowerCAmelCase : List[str] = np.array(train_x)
_lowerCAmelCase : str = np.array(test_x)
_lowerCAmelCase : str = np.array([list(i.ravel()) for i in train_y])
_lowerCAmelCase : str = np.array([list(i.ravel()) for i in test_y])
_lowerCAmelCase : List[Any] = Sequential()
model.add(LSTM(128, input_shape=(look_back, 1), return_sequences=True))
model.add(LSTM(64, input_shape=(128, 1)))
model.add(Dense(forward_days))
model.compile(loss="mean_squared_error", optimizer="adam")
_lowerCAmelCase : List[str] = model.fit(
x_train, y_train, epochs=150, verbose=1, shuffle=True, batch_size=4
)
_lowerCAmelCase : Dict = model.predict(x_test)
| 362
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_sentencepiece_available
_lowerCAmelCase : int = {}
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Dict = ["BartphoTokenizer"]
if TYPE_CHECKING:
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_bartpho import BartphoTokenizer
else:
import sys
_lowerCAmelCase : Optional[Any] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 0
|
from __future__ import annotations
import inspect
import unittest
import numpy as np
from transformers import ResNetConfig
from transformers.testing_utils import require_tf, require_vision, slow
from transformers.utils import cached_property, is_tf_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_tf_common import TFModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_tf_available():
import tensorflow as tf
from transformers import TFResNetForImageClassification, TFResNetModel
from transformers.models.resnet.modeling_tf_resnet import TF_RESNET_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
from PIL import Image
from transformers import AutoImageProcessor
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=3 , __snake_case=32 , __snake_case=3 , __snake_case=10 , __snake_case=[10, 20, 30, 40] , __snake_case=[1, 1, 2, 1] , __snake_case=True , __snake_case=True , __snake_case="relu" , __snake_case=3 , __snake_case=None , ) -> int:
'''simple docstring'''
__a =parent
__a =batch_size
__a =image_size
__a =num_channels
__a =embeddings_size
__a =hidden_sizes
__a =depths
__a =is_training
__a =use_labels
__a =hidden_act
__a =num_labels
__a =scope
__a =len(_a )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.num_labels )
__a =self.get_config()
return config, pixel_values, labels
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
return ResNetConfig(
num_channels=self.num_channels , embeddings_size=self.embeddings_size , hidden_sizes=self.hidden_sizes , depths=self.depths , hidden_act=self.hidden_act , num_labels=self.num_labels , image_size=self.image_size , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> List[str]:
'''simple docstring'''
__a =TFResNetModel(config=_a )
__a =model(_a )
# expected last hidden states: B, C, H // 32, W // 32
self.parent.assertEqual(
result.last_hidden_state.shape , (self.batch_size, self.hidden_sizes[-1], self.image_size // 32, self.image_size // 32) , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =self.num_labels
__a =TFResNetForImageClassification(_a )
__a =model(_a , labels=_a )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
__a , __a , __a =config_and_inputs
__a ={'pixel_values': pixel_values}
return config, inputs_dict
@require_tf
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (TFResNetModel, TFResNetForImageClassification) if is_tf_available() else ()
SCREAMING_SNAKE_CASE = (
{'''feature-extraction''': TFResNetModel, '''image-classification''': TFResNetForImageClassification}
if is_tf_available()
else {}
)
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =TFResNetModelTester(self )
__a =ConfigTester(self , config_class=_a , has_text_modality=_a )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
self.create_and_test_config_common_properties()
self.config_tester.create_and_test_config_to_json_string()
self.config_tester.create_and_test_config_to_json_file()
self.config_tester.create_and_test_config_from_and_save_pretrained()
self.config_tester.create_and_test_config_with_num_labels()
self.config_tester.check_config_can_be_init_without_params()
self.config_tester.check_config_arguments_init()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
return
@unittest.skip(reason='ResNet does not use inputs_embeds' )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
pass
@unittest.skip(reason='ResNet does not support input and output embeddings' )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a , __a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
__a =model_class(_a )
__a =inspect.signature(model.call )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
__a =[*signature.parameters.keys()]
__a =['pixel_values']
self.assertListEqual(arg_names[:1] , _a )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*_a )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
def check_hidden_states_output(__snake_case , __snake_case , __snake_case ):
__a =model_class(_a )
__a =model(**self._prepare_for_class(_a , _a ) )
__a =outputs.encoder_hidden_states if config.is_encoder_decoder else outputs.hidden_states
__a =self.model_tester.num_stages
self.assertEqual(len(_a ) , expected_num_stages + 1 )
# ResNet's feature maps are of shape (batch_size, num_channels, height, width)
self.assertListEqual(
list(hidden_states[0].shape[-2:] ) , [self.model_tester.image_size // 4, self.model_tester.image_size // 4] , )
__a , __a =self.model_tester.prepare_config_and_inputs_for_common()
__a =['basic', 'bottleneck']
for model_class in self.all_model_classes:
for layer_type in layers_type:
__a =layer_type
__a =True
check_hidden_states_output(_a , _a , _a )
# check that output_hidden_states also work using config
del inputs_dict["output_hidden_states"]
__a =True
check_hidden_states_output(_a , _a , _a )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*_a )
@slow
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
for model_name in TF_RESNET_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =TFResNetModel.from_pretrained(_a )
self.assertIsNotNone(_a )
def UpperCamelCase_( ):
"""simple docstring"""
__a =Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' )
return image
@require_tf
@require_vision
class __magic_name__ ( unittest.TestCase ):
@cached_property
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return (
AutoImageProcessor.from_pretrained(TF_RESNET_PRETRAINED_MODEL_ARCHIVE_LIST[0] )
if is_vision_available()
else None
)
@slow
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =TFResNetForImageClassification.from_pretrained(TF_RESNET_PRETRAINED_MODEL_ARCHIVE_LIST[0] )
__a =self.default_image_processor
__a =prepare_img()
__a =image_processor(images=_a , return_tensors='tf' )
# forward pass
__a =model(**_a )
# verify the logits
__a =tf.TensorShape((1, 1000) )
self.assertEqual(outputs.logits.shape , _a )
__a =tf.constant([-11.1069, -9.7877, -8.3777] )
self.assertTrue(np.allclose(outputs.logits[0, :3].numpy() , _a , atol=1e-4 ) )
| 363
|
from collections import OrderedDict
from typing import Mapping
from packaging import version
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
_lowerCAmelCase : List[str] = logging.get_logger(__name__)
_lowerCAmelCase : List[Any] = {
"hustvl/yolos-small": "https://huggingface.co/hustvl/yolos-small/resolve/main/config.json",
# See all YOLOS models at https://huggingface.co/models?filter=yolos
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'yolos'
def __init__( self , __snake_case=768 , __snake_case=12 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=[512, 864] , __snake_case=16 , __snake_case=3 , __snake_case=True , __snake_case=100 , __snake_case=True , __snake_case=False , __snake_case=1 , __snake_case=5 , __snake_case=2 , __snake_case=5 , __snake_case=2 , __snake_case=0.1 , **__snake_case , ) -> str:
'''simple docstring'''
super().__init__(**__snake_case )
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =initializer_range
__a =layer_norm_eps
__a =image_size
__a =patch_size
__a =num_channels
__a =qkv_bias
__a =num_detection_tokens
__a =use_mid_position_embeddings
__a =auxiliary_loss
# Hungarian matcher
__a =class_cost
__a =bbox_cost
__a =giou_cost
# Loss coefficients
__a =bbox_loss_coefficient
__a =giou_loss_coefficient
__a =eos_coefficient
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = version.parse('1.11' )
@property
def __magic_name__ ( self ) -> Mapping[str, Mapping[int, str]]:
'''simple docstring'''
return OrderedDict(
[
('pixel_values', {0: 'batch', 1: 'num_channels', 2: 'height', 3: 'width'}),
] )
@property
def __magic_name__ ( self ) -> float:
'''simple docstring'''
return 1e-4
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return 12
| 308
| 0
|
from __future__ import annotations
import requests
_lowerCAmelCase : Any = set(
"approved_at_utc approved_by author_flair_background_color\nauthor_flair_css_class author_flair_richtext author_flair_template_id author_fullname\nauthor_premium can_mod_post category clicked content_categories created_utc downs\nedited gilded gildings hidden hide_score is_created_from_ads_ui is_meta\nis_original_content is_reddit_media_domain is_video link_flair_css_class\nlink_flair_richtext link_flair_text link_flair_text_color media_embed mod_reason_title\nname permalink pwls quarantine saved score secure_media secure_media_embed selftext\nsubreddit subreddit_name_prefixed subreddit_type thumbnail title top_awarded_type\ntotal_awards_received ups upvote_ratio url user_reports".split()
)
def UpperCamelCase_( _snake_case : str , _snake_case : int = 1 , _snake_case : str = "new" , _snake_case : list | None = None ):
"""simple docstring"""
__a =wanted_data or []
if invalid_search_terms := ", ".join(sorted(set(_a ) - valid_terms ) ):
__a =F'Invalid search term: {invalid_search_terms}'
raise ValueError(_a )
__a =requests.get(
F'https://reddit.com/r/{subreddit}/{age}.json?limit={limit}' , headers={'User-agent': 'A random string'} , )
if response.status_code == 429:
raise requests.HTTPError
__a =response.json()
if not wanted_data:
return {id_: data["data"]["children"][id_] for id_ in range(_a )}
__a ={}
for id_ in range(_a ):
__a ={
item: data["data"]["children"][id_]["data"][item] for item in wanted_data
}
return data_dict
if __name__ == "__main__":
# If you get Error 429, that means you are rate limited.Try after some time
print(get_subreddit_data("learnpython", wanted_data=["title", "url", "selftext"]))
| 364
|
import argparse
import torch
from datasets import load_dataset
from donut import DonutModel
from transformers import (
DonutImageProcessor,
DonutProcessor,
DonutSwinConfig,
DonutSwinModel,
MBartConfig,
MBartForCausalLM,
VisionEncoderDecoderModel,
XLMRobertaTokenizerFast,
)
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =model.config
__a =DonutSwinConfig(
image_size=original_config.input_size , patch_size=4 , depths=original_config.encoder_layer , num_heads=[4, 8, 16, 32] , window_size=original_config.window_size , embed_dim=128 , )
__a =MBartConfig(
is_decoder=_snake_case , is_encoder_decoder=_snake_case , add_cross_attention=_snake_case , decoder_layers=original_config.decoder_layer , max_position_embeddings=original_config.max_position_embeddings , vocab_size=len(
model.decoder.tokenizer ) , scale_embedding=_snake_case , add_final_layer_norm=_snake_case , )
return encoder_config, decoder_config
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
if "encoder.model" in name:
__a =name.replace('encoder.model' , 'encoder' )
if "decoder.model" in name:
__a =name.replace('decoder.model' , 'decoder' )
if "patch_embed.proj" in name:
__a =name.replace('patch_embed.proj' , 'embeddings.patch_embeddings.projection' )
if "patch_embed.norm" in name:
__a =name.replace('patch_embed.norm' , 'embeddings.norm' )
if name.startswith('encoder' ):
if "layers" in name:
__a ='encoder.' + name
if "attn.proj" in name:
__a =name.replace('attn.proj' , 'attention.output.dense' )
if "attn" in name and "mask" not in name:
__a =name.replace('attn' , 'attention.self' )
if "norm1" in name:
__a =name.replace('norm1' , 'layernorm_before' )
if "norm2" in name:
__a =name.replace('norm2' , 'layernorm_after' )
if "mlp.fc1" in name:
__a =name.replace('mlp.fc1' , 'intermediate.dense' )
if "mlp.fc2" in name:
__a =name.replace('mlp.fc2' , 'output.dense' )
if name == "encoder.norm.weight":
__a ='encoder.layernorm.weight'
if name == "encoder.norm.bias":
__a ='encoder.layernorm.bias'
return name
def UpperCamelCase_( _snake_case : Tuple , _snake_case : str ):
"""simple docstring"""
for key in orig_state_dict.copy().keys():
__a =orig_state_dict.pop(_snake_case )
if "qkv" in key:
__a =key.split('.' )
__a =int(key_split[3] )
__a =int(key_split[5] )
__a =model.encoder.encoder.layers[layer_num].blocks[block_num].attention.self.all_head_size
if "weight" in key:
__a =val[:dim, :]
__a =val[dim : dim * 2, :]
__a =val[-dim:, :]
else:
__a =val[:dim]
__a =val[dim : dim * 2]
__a =val[-dim:]
elif "attn_mask" in key or key in ["encoder.model.norm.weight", "encoder.model.norm.bias"]:
# HuggingFace implementation doesn't use attn_mask buffer
# and model doesn't use final LayerNorms for the encoder
pass
else:
__a =val
return orig_state_dict
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Union[str, Any]=None , _snake_case : List[Any]=False ):
"""simple docstring"""
__a =DonutModel.from_pretrained(_snake_case ).eval()
# load HuggingFace model
__a , __a =get_configs(_snake_case )
__a =DonutSwinModel(_snake_case )
__a =MBartForCausalLM(_snake_case )
__a =VisionEncoderDecoderModel(encoder=_snake_case , decoder=_snake_case )
model.eval()
__a =original_model.state_dict()
__a =convert_state_dict(_snake_case , _snake_case )
model.load_state_dict(_snake_case )
# verify results on scanned document
__a =load_dataset('hf-internal-testing/example-documents' )
__a =dataset['test'][0]['image'].convert('RGB' )
__a =XLMRobertaTokenizerFast.from_pretrained(_snake_case , from_slow=_snake_case )
__a =DonutImageProcessor(
do_align_long_axis=original_model.config.align_long_axis , size=original_model.config.input_size[::-1] )
__a =DonutProcessor(_snake_case , _snake_case )
__a =processor(_snake_case , return_tensors='pt' ).pixel_values
if model_name == "naver-clova-ix/donut-base-finetuned-docvqa":
__a ='<s_docvqa><s_question>{user_input}</s_question><s_answer>'
__a ='When is the coffee break?'
__a =task_prompt.replace('{user_input}' , _snake_case )
elif model_name == "naver-clova-ix/donut-base-finetuned-rvlcdip":
__a ='<s_rvlcdip>'
elif model_name in [
"naver-clova-ix/donut-base-finetuned-cord-v1",
"naver-clova-ix/donut-base-finetuned-cord-v1-2560",
]:
__a ='<s_cord>'
elif model_name == "naver-clova-ix/donut-base-finetuned-cord-v2":
__a ='s_cord-v2>'
elif model_name == "naver-clova-ix/donut-base-finetuned-zhtrainticket":
__a ='<s_zhtrainticket>'
elif model_name in ["naver-clova-ix/donut-proto", "naver-clova-ix/donut-base"]:
# use a random prompt
__a ='hello world'
else:
raise ValueError('Model name not supported' )
__a =original_model.decoder.tokenizer(_snake_case , add_special_tokens=_snake_case , return_tensors='pt' )[
'input_ids'
]
__a =original_model.encoder.model.patch_embed(_snake_case )
__a , __a =model.encoder.embeddings(_snake_case )
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
# verify encoder hidden states
__a =original_model.encoder(_snake_case )
__a =model.encoder(_snake_case ).last_hidden_state
assert torch.allclose(_snake_case , _snake_case , atol=1e-2 )
# verify decoder hidden states
__a =original_model(_snake_case , _snake_case , _snake_case ).logits
__a =model(_snake_case , decoder_input_ids=_snake_case ).logits
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
print('Looks ok!' )
if pytorch_dump_folder_path is not None:
print(F'Saving model and processor to {pytorch_dump_folder_path}' )
model.save_pretrained(_snake_case )
processor.save_pretrained(_snake_case )
if push_to_hub:
model.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
processor.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
if __name__ == "__main__":
_lowerCAmelCase : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_name",
default="naver-clova-ix/donut-base-finetuned-docvqa",
required=False,
type=str,
help="Name of the original model you'd like to convert.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
required=False,
type=str,
help="Path to the output PyTorch model directory.",
)
parser.add_argument(
"--push_to_hub",
action="store_true",
help="Whether or not to push the converted model and processor to the 🤗 hub.",
)
_lowerCAmelCase : List[Any] = parser.parse_args()
convert_donut_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 308
| 0
|
from typing import Callable, List, Optional, Tuple, Union
import torch
from transformers import CLIPTextModel, CLIPTokenizer
from ...configuration_utils import ConfigMixin, register_to_config
from ...models import ModelMixin, TransformeraDModel, VQModel
from ...schedulers import VQDiffusionScheduler
from ...utils import logging
from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
_lowerCAmelCase : List[Any] = logging.get_logger(__name__) # pylint: disable=invalid-name
class __magic_name__ ( snake_case__ , snake_case__ ):
@register_to_config
def __init__( self , __snake_case , __snake_case = None , __snake_case = None ) -> Optional[int]:
'''simple docstring'''
super().__init__()
__a =learnable
if self.learnable:
assert hidden_size is not None, "learnable=True requires `hidden_size` to be set"
assert length is not None, "learnable=True requires `length` to be set"
__a =torch.zeros(UpperCAmelCase_ , UpperCAmelCase_ )
else:
__a =None
__a =torch.nn.Parameter(UpperCAmelCase_ )
class __magic_name__ ( snake_case__ ):
SCREAMING_SNAKE_CASE = 4_2
SCREAMING_SNAKE_CASE = 4_2
SCREAMING_SNAKE_CASE = 4_2
SCREAMING_SNAKE_CASE = 4_2
SCREAMING_SNAKE_CASE = 4_2
SCREAMING_SNAKE_CASE = 4_2
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[Any]:
'''simple docstring'''
super().__init__()
self.register_modules(
vqvae=UpperCAmelCase_ , transformer=UpperCAmelCase_ , text_encoder=UpperCAmelCase_ , tokenizer=UpperCAmelCase_ , scheduler=UpperCAmelCase_ , learned_classifier_free_sampling_embeddings=UpperCAmelCase_ , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> List[Any]:
'''simple docstring'''
__a =len(UpperCAmelCase_ ) if isinstance(UpperCAmelCase_ , UpperCAmelCase_ ) else 1
# get prompt text embeddings
__a =self.tokenizer(
UpperCAmelCase_ , padding='max_length' , max_length=self.tokenizer.model_max_length , return_tensors='pt' , )
__a =text_inputs.input_ids
if text_input_ids.shape[-1] > self.tokenizer.model_max_length:
__a =self.tokenizer.batch_decode(text_input_ids[:, self.tokenizer.model_max_length :] )
logger.warning(
'The following part of your input was truncated because CLIP can only handle sequences up to'
f' {self.tokenizer.model_max_length} tokens: {removed_text}' )
__a =text_input_ids[:, : self.tokenizer.model_max_length]
__a =self.text_encoder(text_input_ids.to(self.device ) )[0]
# NOTE: This additional step of normalizing the text embeddings is from VQ-Diffusion.
# While CLIP does normalize the pooled output of the text transformer when combining
# the image and text embeddings, CLIP does not directly normalize the last hidden state.
#
# CLIP normalizing the pooled output.
# https://github.com/huggingface/transformers/blob/d92e22d1f28324f513f3080e5c47c071a3916721/src/transformers/models/clip/modeling_clip.py#L1052-L1053
__a =prompt_embeds / prompt_embeds.norm(dim=-1 , keepdim=UpperCAmelCase_ )
# duplicate text embeddings for each generation per prompt
__a =prompt_embeds.repeat_interleave(UpperCAmelCase_ , dim=0 )
if do_classifier_free_guidance:
if self.learned_classifier_free_sampling_embeddings.learnable:
__a =self.learned_classifier_free_sampling_embeddings.embeddings
__a =negative_prompt_embeds.unsqueeze(0 ).repeat(UpperCAmelCase_ , 1 , 1 )
else:
__a =[""] * batch_size
__a =text_input_ids.shape[-1]
__a =self.tokenizer(
UpperCAmelCase_ , padding='max_length' , max_length=UpperCAmelCase_ , truncation=UpperCAmelCase_ , return_tensors='pt' , )
__a =self.text_encoder(uncond_input.input_ids.to(self.device ) )[0]
# See comment for normalizing text embeddings
__a =negative_prompt_embeds / negative_prompt_embeds.norm(dim=-1 , keepdim=UpperCAmelCase_ )
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
__a =negative_prompt_embeds.shape[1]
__a =negative_prompt_embeds.repeat(1 , UpperCAmelCase_ , 1 )
__a =negative_prompt_embeds.view(batch_size * num_images_per_prompt , UpperCAmelCase_ , -1 )
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch
# to avoid doing two forward passes
__a =torch.cat([negative_prompt_embeds, prompt_embeds] )
return prompt_embeds
@torch.no_grad()
def __call__( self , __snake_case , __snake_case = 100 , __snake_case = 5.0 , __snake_case = 1.0 , __snake_case = 1 , __snake_case = None , __snake_case = None , __snake_case = "pil" , __snake_case = True , __snake_case = None , __snake_case = 1 , ) -> List[Any]:
'''simple docstring'''
if isinstance(UpperCAmelCase_ , UpperCAmelCase_ ):
__a =1
elif isinstance(UpperCAmelCase_ , UpperCAmelCase_ ):
__a =len(UpperCAmelCase_ )
else:
raise ValueError(f'`prompt` has to be of type `str` or `list` but is {type(UpperCAmelCase_ )}' )
__a =batch_size * num_images_per_prompt
__a =guidance_scale > 1.0
__a =self._encode_prompt(UpperCAmelCase_ , UpperCAmelCase_ , UpperCAmelCase_ )
if (callback_steps is None) or (
callback_steps is not None and (not isinstance(UpperCAmelCase_ , UpperCAmelCase_ ) or callback_steps <= 0)
):
raise ValueError(
f'`callback_steps` has to be a positive integer but is {callback_steps} of type'
f' {type(UpperCAmelCase_ )}.' )
# get the initial completely masked latents unless the user supplied it
__a =(batch_size, self.transformer.num_latent_pixels)
if latents is None:
__a =self.transformer.num_vector_embeds - 1
__a =torch.full(UpperCAmelCase_ , UpperCAmelCase_ ).to(self.device )
else:
if latents.shape != latents_shape:
raise ValueError(f'Unexpected latents shape, got {latents.shape}, expected {latents_shape}' )
if (latents < 0).any() or (latents >= self.transformer.num_vector_embeds).any():
raise ValueError(
'Unexpected latents value(s). All latents be valid embedding indices i.e. in the range 0,'
f' {self.transformer.num_vector_embeds - 1} (inclusive).' )
__a =latents.to(self.device )
# set timesteps
self.scheduler.set_timesteps(UpperCAmelCase_ , device=self.device )
__a =self.scheduler.timesteps.to(self.device )
__a =latents
for i, t in enumerate(self.progress_bar(UpperCAmelCase_ ) ):
# expand the sample if we are doing classifier free guidance
__a =torch.cat([sample] * 2 ) if do_classifier_free_guidance else sample
# predict the un-noised image
# model_output == `log_p_x_0`
__a =self.transformer(UpperCAmelCase_ , encoder_hidden_states=UpperCAmelCase_ , timestep=UpperCAmelCase_ ).sample
if do_classifier_free_guidance:
__a =model_output.chunk(2 )
__a =model_output_uncond + guidance_scale * (model_output_text - model_output_uncond)
model_output -= torch.logsumexp(UpperCAmelCase_ , dim=1 , keepdim=UpperCAmelCase_ )
__a =self.truncate(UpperCAmelCase_ , UpperCAmelCase_ )
# remove `log(0)`'s (`-inf`s)
__a =model_output.clamp(-70 )
# compute the previous noisy sample x_t -> x_t-1
__a =self.scheduler.step(UpperCAmelCase_ , timestep=UpperCAmelCase_ , sample=UpperCAmelCase_ , generator=UpperCAmelCase_ ).prev_sample
# call the callback, if provided
if callback is not None and i % callback_steps == 0:
callback(UpperCAmelCase_ , UpperCAmelCase_ , UpperCAmelCase_ )
__a =self.vqvae.config.vq_embed_dim
__a =(batch_size, self.transformer.height, self.transformer.width, embedding_channels)
__a =self.vqvae.quantize.get_codebook_entry(UpperCAmelCase_ , shape=UpperCAmelCase_ )
__a =self.vqvae.decode(UpperCAmelCase_ , force_not_quantize=UpperCAmelCase_ ).sample
__a =(image / 2 + 0.5).clamp(0 , 1 )
__a =image.cpu().permute(0 , 2 , 3 , 1 ).numpy()
if output_type == "pil":
__a =self.numpy_to_pil(UpperCAmelCase_ )
if not return_dict:
return (image,)
return ImagePipelineOutput(images=UpperCAmelCase_ )
def __magic_name__ ( self , __snake_case , __snake_case ) -> List[str]:
'''simple docstring'''
__a =torch.sort(UpperCAmelCase_ , 1 , descending=UpperCAmelCase_ )
__a =torch.exp(UpperCAmelCase_ )
__a =sorted_p_x_0.cumsum(dim=1 ) < truncation_rate
# Ensure that at least the largest probability is not zeroed out
__a =torch.full_like(keep_mask[:, 0:1, :] , UpperCAmelCase_ )
__a =torch.cat((all_true, keep_mask) , dim=1 )
__a =keep_mask[:, :-1, :]
__a =keep_mask.gather(1 , indices.argsort(1 ) )
__a =log_p_x_0.clone()
__a =-torch.inf # -inf = log(0)
return rv
| 365
|
import json
import os
import unittest
from transformers.models.blenderbot_small.tokenization_blenderbot_small import (
VOCAB_FILES_NAMES,
BlenderbotSmallTokenizer,
)
from ...test_tokenization_common import TokenizerTesterMixin
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = BlenderbotSmallTokenizer
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
super().setUp()
__a =['__start__', 'adapt', 'act', 'ap@@', 'te', '__end__', '__unk__']
__a =dict(zip(__snake_case , range(len(__snake_case ) ) ) )
__a =['#version: 0.2', 'a p', 't e</w>', 'ap t</w>', 'a d', 'ad apt</w>', 'a c', 'ac t</w>', '']
__a ={'unk_token': '__unk__', 'bos_token': '__start__', 'eos_token': '__end__'}
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(__snake_case ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(__snake_case ) )
def __magic_name__ ( self , **__snake_case ) -> Any:
'''simple docstring'''
kwargs.update(self.special_tokens_map )
return BlenderbotSmallTokenizer.from_pretrained(self.tmpdirname , **__snake_case )
def __magic_name__ ( self , __snake_case ) -> List[Any]:
'''simple docstring'''
__a ='adapt act apte'
__a ='adapt act apte'
return input_text, output_text
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
__a ='adapt act apte'
__a =['adapt', 'act', 'ap@@', 'te']
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
__a =[tokenizer.bos_token] + tokens + [tokenizer.eos_token]
__a =[0, 1, 2, 3, 4, 5]
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
assert tok('sam' ).input_ids == [1384]
__a ='I am a small frog.'
__a =tok([src_text] , padding=__snake_case , truncation=__snake_case )['input_ids']
__a =tok.batch_decode(__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case )[0]
assert src_text != decoded # I wish it did!
assert decoded == "i am a small frog ."
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
__a ='I am a small frog .'
__a ='.'
__a =tok(__snake_case )['input_ids']
__a =tok(__snake_case )['input_ids']
assert encoded[-1] == encoded_dot[0]
| 308
| 0
|
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =[int(SCREAMING_SNAKE_CASE_ ) for i in ip_va_address.split('.' ) if i.isdigit()]
return len(SCREAMING_SNAKE_CASE_ ) == 4 and all(0 <= int(SCREAMING_SNAKE_CASE_ ) <= 254 for octet in octets )
if __name__ == "__main__":
_lowerCAmelCase : int = input().strip()
_lowerCAmelCase : int = "valid" if is_ip_va_address_valid(ip) else "invalid"
print(f'''{ip} is a {valid_or_invalid} IP v4 address.''')
| 366
|
import unittest
from transformers import load_tool
from transformers.utils import is_torch_available
if is_torch_available():
import torch
from transformers.testing_utils import require_torch
from .test_tools_common import ToolTesterMixin
@require_torch
class __magic_name__ ( unittest.TestCase , lowerCAmelCase_ ):
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =load_tool('text-to-speech' )
self.tool.setup()
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
| 308
| 0
|
import copy
import re
class __magic_name__ :
SCREAMING_SNAKE_CASE = 'hp'
SCREAMING_SNAKE_CASE = {}
SCREAMING_SNAKE_CASE = None
@classmethod
def __magic_name__ ( cls , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
__a =prefix
__a =defaults
cls.build_naming_info()
@staticmethod
def __magic_name__ ( __snake_case , __snake_case ) -> Optional[Any]:
'''simple docstring'''
if len(__lowerCamelCase ) == 0:
return ""
__a =None
if any(char.isdigit() for char in word ):
raise Exception(f'Parameters should not contain numbers: \'{word}\' contains a number' )
if word in info["short_word"]:
return info["short_word"][word]
for prefix_len in range(1 , len(__lowerCamelCase ) + 1 ):
__a =word[:prefix_len]
if prefix in info["reverse_short_word"]:
continue
else:
__a =prefix
break
if short_word is None:
# Paranoid fallback
def int_to_alphabetic(__snake_case ):
__a =''
while integer != 0:
__a =chr(ord('A' ) + integer % 10 ) + s
integer //= 10
return s
__a =0
while True:
__a =word + '#' + int_to_alphabetic(__lowerCamelCase )
if sword in info["reverse_short_word"]:
continue
else:
__a =sword
break
__a =short_word
__a =word
return short_word
@staticmethod
def __magic_name__ ( __snake_case , __snake_case ) -> List[str]:
'''simple docstring'''
__a =param_name.split('_' )
__a =[TrialShortNamer.shortname_for_word(__lowerCamelCase , __lowerCamelCase ) for word in words]
# We try to create a separatorless short name, but if there is a collision we have to fallback
# to a separated short name
__a =['', '_']
for separator in separators:
__a =separator.join(__lowerCamelCase )
if shortname not in info["reverse_short_param"]:
__a =shortname
__a =param_name
return shortname
return param_name
@staticmethod
def __magic_name__ ( __snake_case , __snake_case ) -> Dict:
'''simple docstring'''
__a =TrialShortNamer.shortname_for_key(__lowerCamelCase , __lowerCamelCase )
__a =short_name
__a =param_name
@classmethod
def __magic_name__ ( cls ) -> Tuple:
'''simple docstring'''
if cls.NAMING_INFO is not None:
return
__a ={
'short_word': {},
'reverse_short_word': {},
'short_param': {},
'reverse_short_param': {},
}
__a =list(cls.DEFAULTS.keys() )
for k in field_keys:
cls.add_new_param_name(__lowerCamelCase , __lowerCamelCase )
__a =info
@classmethod
def __magic_name__ ( cls , __snake_case ) -> List[Any]:
'''simple docstring'''
cls.build_naming_info()
assert cls.PREFIX is not None
__a =[copy.copy(cls.PREFIX )]
for k, v in params.items():
if k not in cls.DEFAULTS:
raise Exception(f'You should provide a default value for the param name {k} with value {v}' )
if v == cls.DEFAULTS[k]:
# The default value is not added to the name
continue
__a =cls.NAMING_INFO['short_param'][k]
if isinstance(__lowerCamelCase , __lowerCamelCase ):
__a =1 if v else 0
__a ='' if isinstance(__lowerCamelCase , (int, float) ) else '-'
__a =f'{key}{sep}{v}'
name.append(__lowerCamelCase )
return "_".join(__lowerCamelCase )
@classmethod
def __magic_name__ ( cls , __snake_case ) -> Optional[int]:
'''simple docstring'''
__a =repr[len(cls.PREFIX ) + 1 :]
if repr == "":
__a =[]
else:
__a =repr.split('_' )
__a ={}
for value in values:
if "-" in value:
__a , __a =value.split('-' )
else:
__a =re.sub('[0-9.]' , '' , __lowerCamelCase )
__a =float(re.sub('[^0-9.]' , '' , __lowerCamelCase ) )
__a =cls.NAMING_INFO['reverse_short_param'][p_k]
__a =p_v
for k in cls.DEFAULTS:
if k not in parameters:
__a =cls.DEFAULTS[k]
return parameters
| 367
|
import flax.linen as nn
import jax.numpy as jnp
from .attention_flax import FlaxTransformeraDModel
from .resnet_flax import FlaxDownsampleaD, FlaxResnetBlockaD, FlaxUpsampleaD
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> Optional[Any]:
'''simple docstring'''
__a =()
for resnet, attn in zip(self.resnets , self.attentions ):
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case=True ) -> Optional[int]:
'''simple docstring'''
__a =()
for resnet in self.resnets:
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet, attn in zip(self.resnets , self.attentions ):
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet in self.resnets:
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# there is always at least one resnet
__a =[
FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
]
__a =[]
for _ in range(self.num_layers ):
__a =FlaxTransformeraDModel(
in_channels=self.in_channels , n_heads=self.num_attention_heads , d_head=self.in_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
__a =attentions
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[str]:
'''simple docstring'''
__a =self.resnets[0](__snake_case , __snake_case )
for attn, resnet in zip(self.attentions , self.resnets[1:] ):
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
return hidden_states
| 308
| 0
|
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,
convert_to_rgb,
get_resize_output_image_size,
normalize,
rescale,
resize,
to_channel_dimension_format,
)
from ...image_utils import (
OPENAI_CLIP_MEAN,
OPENAI_CLIP_STD,
ChannelDimension,
ImageInput,
PILImageResampling,
make_list_of_images,
to_numpy_array,
valid_images,
)
from ...utils import TensorType, is_vision_available, logging
_lowerCAmelCase : Any = logging.get_logger(__name__)
if is_vision_available():
import PIL
class __magic_name__ ( A_ ):
SCREAMING_SNAKE_CASE = ['pixel_values']
def __init__( self , __snake_case = True , __snake_case = None , __snake_case = PILImageResampling.BICUBIC , __snake_case = True , __snake_case = None , __snake_case = True , __snake_case = 1 / 255 , __snake_case = True , __snake_case = None , __snake_case = None , __snake_case = True , **__snake_case , ) -> List[str]:
'''simple docstring'''
super().__init__(**_lowerCamelCase )
__a =size if size is not None else {'shortest_edge': 224}
__a =get_size_dict(_lowerCamelCase , default_to_square=_lowerCamelCase )
__a =crop_size if crop_size is not None else {'height': 224, 'width': 224}
__a =get_size_dict(_lowerCamelCase , default_to_square=_lowerCamelCase , param_name='crop_size' )
__a =do_resize
__a =size
__a =resample
__a =do_center_crop
__a =crop_size
__a =do_rescale
__a =rescale_factor
__a =do_normalize
__a =image_mean if image_mean is not None else OPENAI_CLIP_MEAN
__a =image_std if image_std is not None else OPENAI_CLIP_STD
__a =do_convert_rgb
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case = PILImageResampling.BICUBIC , __snake_case = None , **__snake_case , ) -> Optional[int]:
'''simple docstring'''
__a =get_size_dict(_lowerCamelCase , default_to_square=_lowerCamelCase )
if "shortest_edge" not in size:
raise ValueError(f'The `size` parameter must contain the key `shortest_edge`. Got {size.keys()}' )
__a =get_resize_output_image_size(_lowerCamelCase , size=size['shortest_edge'] , default_to_square=_lowerCamelCase )
return resize(_lowerCamelCase , size=_lowerCamelCase , resample=_lowerCamelCase , data_format=_lowerCamelCase , **_lowerCamelCase )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case = None , **__snake_case , ) -> Any:
'''simple docstring'''
__a =get_size_dict(_lowerCamelCase )
if "height" not in size or "width" not in size:
raise ValueError(f'The `size` parameter must contain the keys (height, width). Got {size.keys()}' )
return center_crop(_lowerCamelCase , size=(size['height'], size['width']) , data_format=_lowerCamelCase , **_lowerCamelCase )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case = None , **__snake_case , ) -> Optional[Any]:
'''simple docstring'''
return rescale(_lowerCamelCase , scale=_lowerCamelCase , data_format=_lowerCamelCase , **_lowerCamelCase )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case = None , **__snake_case , ) -> Any:
'''simple docstring'''
return normalize(_lowerCamelCase , mean=_lowerCamelCase , std=_lowerCamelCase , data_format=_lowerCamelCase , **_lowerCamelCase )
def __magic_name__ ( self , __snake_case , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = None , __snake_case = ChannelDimension.FIRST , **__snake_case , ) -> List[Any]:
'''simple docstring'''
__a =do_resize if do_resize is not None else self.do_resize
__a =size if size is not None else self.size
__a =get_size_dict(_lowerCamelCase , param_name='size' , default_to_square=_lowerCamelCase )
__a =resample if resample is not None else self.resample
__a =do_center_crop if do_center_crop is not None else self.do_center_crop
__a =crop_size if crop_size is not None else self.crop_size
__a =get_size_dict(_lowerCamelCase , param_name='crop_size' , default_to_square=_lowerCamelCase )
__a =do_rescale if do_rescale is not None else self.do_rescale
__a =rescale_factor if rescale_factor is not None else self.rescale_factor
__a =do_normalize if do_normalize is not None else self.do_normalize
__a =image_mean if image_mean is not None else self.image_mean
__a =image_std if image_std is not None else self.image_std
__a =do_convert_rgb if do_convert_rgb is not None else self.do_convert_rgb
__a =make_list_of_images(_lowerCamelCase )
if not valid_images(_lowerCamelCase ):
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.' )
# PIL RGBA images are converted to RGB
if do_convert_rgb:
__a =[convert_to_rgb(_lowerCamelCase ) for image in images]
# All transformations expect numpy arrays.
__a =[to_numpy_array(_lowerCamelCase ) for image in images]
if do_resize:
__a =[self.resize(image=_lowerCamelCase , size=_lowerCamelCase , resample=_lowerCamelCase ) for image in images]
if do_center_crop:
__a =[self.center_crop(image=_lowerCamelCase , size=_lowerCamelCase ) for image in images]
if do_rescale:
__a =[self.rescale(image=_lowerCamelCase , scale=_lowerCamelCase ) for image in images]
if do_normalize:
__a =[self.normalize(image=_lowerCamelCase , mean=_lowerCamelCase , std=_lowerCamelCase ) for image in images]
__a =[to_channel_dimension_format(_lowerCamelCase , _lowerCamelCase ) for image in images]
__a ={'pixel_values': images}
return BatchFeature(data=_lowerCamelCase , tensor_type=_lowerCamelCase )
| 368
|
import json
import os
import subprocess
import unittest
from ast import literal_eval
import pytest
from parameterized import parameterized, parameterized_class
from . import is_sagemaker_available
if is_sagemaker_available():
from sagemaker import Session, TrainingJobAnalytics
from sagemaker.huggingface import HuggingFace
@pytest.mark.skipif(
literal_eval(os.getenv('TEST_SAGEMAKER' , 'False' ) ) is not True , reason='Skipping test because should only be run when releasing minor transformers version' , )
@pytest.mark.usefixtures('sm_env' )
@parameterized_class(
[
{
'framework': 'pytorch',
'script': 'run_glue.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_5_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'pytorch',
'script': 'run_ddp.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'tensorflow',
'script': 'run_tf_dist.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.6, 'eval_loss': 0.7},
},
] )
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.framework == "pytorch":
subprocess.run(
f'cp ./examples/pytorch/text-classification/run_glue.py {self.env.test_path}/run_glue.py'.split() , encoding='utf-8' , check=__snake_case , )
assert hasattr(self , 'env' )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
__a =f'{self.env.base_job_name}-{instance_count}-{"ddp" if "ddp" in self.script else "smd"}'
# distributed data settings
__a ={'smdistributed': {'dataparallel': {'enabled': True}}} if self.script != 'run_ddp.py' else None
# creates estimator
return HuggingFace(
entry_point=self.script , source_dir=self.env.test_path , role=self.env.role , image_uri=self.env.image_uri , base_job_name=__snake_case , instance_count=__snake_case , instance_type=self.instance_type , debugger_hook_config=__snake_case , hyperparameters={**self.env.distributed_hyperparameters, 'model_name_or_path': self.model_name_or_path} , metric_definitions=self.env.metric_definitions , distribution=__snake_case , py_version='py36' , )
def __magic_name__ ( self , __snake_case ) -> Optional[Any]:
'''simple docstring'''
TrainingJobAnalytics(__snake_case ).export_csv(f'{self.env.test_path}/{job_name}_metrics.csv' )
@parameterized.expand([(2,)] )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
# create estimator
__a =self.create_estimator(__snake_case )
# run training
estimator.fit()
# result dataframe
__a =TrainingJobAnalytics(estimator.latest_training_job.name ).dataframe()
# extract kpis
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_accuracy']['value'] )
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_loss']['value'] )
# get train time from SageMaker job, this includes starting, preprocessing, stopping
__a =(
Session().describe_training_job(estimator.latest_training_job.name ).get('TrainingTimeInSeconds' , 99_9999 )
)
# assert kpis
assert train_runtime <= self.results["train_runtime"]
assert all(t >= self.results['eval_accuracy'] for t in eval_accuracy )
assert all(t <= self.results['eval_loss'] for t in eval_loss )
# dump tests result into json file to share in PR
with open(f'{estimator.latest_training_job.name}.json' , 'w' ) as outfile:
json.dump({'train_time': train_runtime, 'eval_accuracy': eval_accuracy, 'eval_loss': eval_loss} , __snake_case )
| 308
| 0
|
def UpperCamelCase_( _snake_case : int = 1000 ):
"""simple docstring"""
__a =-1
__a =0
for a in range(1 , n // 3 ):
# Solving the two equations a**2+b**2=c**2 and a+b+c=N eliminating c
__a =(n * n - 2 * a * n) // (2 * n - 2 * a)
__a =n - a - b
if c * c == (a * a + b * b):
__a =a * b * c
if candidate >= product:
__a =candidate
return product
if __name__ == "__main__":
print(f'''{solution() = }''')
| 369
|
import argparse
import logging
from collections import namedtuple
import torch
from model_bertabs import BertAbsSummarizer
from models.model_builder import AbsSummarizer # The authors' implementation
from transformers import BertTokenizer
logging.basicConfig(level=logging.INFO)
_lowerCAmelCase : List[Any] = logging.getLogger(__name__)
_lowerCAmelCase : Optional[Any] = "Hello world! cécé herlolip"
_lowerCAmelCase : str = namedtuple(
"BertAbsConfig",
[
"temp_dir",
"large",
"use_bert_emb",
"finetune_bert",
"encoder",
"share_emb",
"max_pos",
"enc_layers",
"enc_hidden_size",
"enc_heads",
"enc_ff_size",
"enc_dropout",
"dec_layers",
"dec_hidden_size",
"dec_heads",
"dec_ff_size",
"dec_dropout",
],
)
def UpperCamelCase_( _snake_case : str , _snake_case : List[Any] ):
"""simple docstring"""
__a =BertAbsConfig(
temp_dir='.' , finetune_bert=_snake_case , large=_snake_case , share_emb=_snake_case , use_bert_emb=_snake_case , encoder='bert' , max_pos=512 , enc_layers=6 , enc_hidden_size=512 , enc_heads=8 , enc_ff_size=512 , enc_dropout=0.2 , dec_layers=6 , dec_hidden_size=768 , dec_heads=8 , dec_ff_size=2048 , dec_dropout=0.2 , )
__a =torch.load(_snake_case , lambda _snake_case , _snake_case : storage )
__a =AbsSummarizer(_snake_case , torch.device('cpu' ) , _snake_case )
original.eval()
__a =BertAbsSummarizer(_snake_case , torch.device('cpu' ) )
new_model.eval()
# -------------------
# Convert the weights
# -------------------
logging.info('convert the model' )
new_model.bert.load_state_dict(original.bert.state_dict() )
new_model.decoder.load_state_dict(original.decoder.state_dict() )
new_model.generator.load_state_dict(original.generator.state_dict() )
# ----------------------------------
# Make sure the outpus are identical
# ----------------------------------
logging.info('Make sure that the models\' outputs are identical' )
__a =BertTokenizer.from_pretrained('bert-base-uncased' )
# prepare the model inputs
__a =tokenizer.encode('This is sample éàalj\'-.' )
encoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
__a =tokenizer.encode('This is sample 3 éàalj\'-.' )
decoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
# failsafe to make sure the weights reset does not affect the
# loaded weights.
assert torch.max(torch.abs(original.generator[0].weight - new_model.generator[0].weight ) ) == 0
# forward pass
__a =encoder_input_ids
__a =decoder_input_ids
__a =__a =None
__a =None
__a =__a =None
__a =__a =None
__a =None
# The original model does not apply the geneator layer immediatly but rather in
# the beam search (where it combines softmax + linear layer). Since we already
# apply the softmax in our generation process we only apply the linear layer here.
# We make sure that the outputs of the full stack are identical
__a =original(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =original.generator(_snake_case )
__a =new_model(
_snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =new_model.generator(_snake_case )
__a =torch.max(torch.abs(output_converted_model - output_original_model ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.max(torch.abs(output_converted_generator - output_original_generator ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.allclose(_snake_case , _snake_case , atol=1e-3 )
if are_identical:
logging.info('all weights are equal up to 1e-3' )
else:
raise ValueError('the weights are different. The new model is likely different from the original one.' )
# The model has been saved with torch.save(model) and this is bound to the exact
# directory structure. We save the state_dict instead.
logging.info('saving the model\'s state dictionary' )
torch.save(
new_model.state_dict() , './bertabs-finetuned-cnndm-extractive-abstractive-summarization/pytorch_model.bin' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
parser.add_argument(
"--bertabs_checkpoint_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch dump.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
type=str,
required=True,
help="Path to the output PyTorch model.",
)
_lowerCAmelCase : Optional[Any] = parser.parse_args()
convert_bertabs_checkpoints(
args.bertabs_checkpoint_path,
args.pytorch_dump_folder_path,
)
| 308
| 0
|
import os
import shutil
import sys
import tempfile
import unittest
from pathlib import Path
import pytest
import transformers
from transformers import (
BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP,
AutoTokenizer,
BertConfig,
BertTokenizer,
BertTokenizerFast,
CTRLTokenizer,
GPTaTokenizer,
GPTaTokenizerFast,
PreTrainedTokenizerFast,
RobertaTokenizer,
RobertaTokenizerFast,
is_tokenizers_available,
)
from transformers.models.auto.configuration_auto import CONFIG_MAPPING, AutoConfig
from transformers.models.auto.tokenization_auto import (
TOKENIZER_MAPPING,
get_tokenizer_config,
tokenizer_class_from_name,
)
from transformers.models.roberta.configuration_roberta import RobertaConfig
from transformers.testing_utils import (
DUMMY_DIFF_TOKENIZER_IDENTIFIER,
DUMMY_UNKNOWN_IDENTIFIER,
SMALL_MODEL_IDENTIFIER,
RequestCounter,
require_tokenizers,
slow,
)
sys.path.append(str(Path(__file__).parent.parent.parent.parent / "utils"))
from test_module.custom_configuration import CustomConfig # noqa E402
from test_module.custom_tokenization import CustomTokenizer # noqa E402
if is_tokenizers_available():
from test_module.custom_tokenization_fast import CustomTokenizerFast
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =0
@slow
def __magic_name__ ( self ) -> int:
'''simple docstring'''
for model_name in (x for x in BERT_PRETRAINED_CONFIG_ARCHIVE_MAP.keys() if "japanese" not in x):
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsNotNone(_a )
self.assertIsInstance(_a , (BertTokenizer, BertTokenizerFast) )
self.assertGreater(len(_a ) , 0 )
for model_name in GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP.keys():
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsNotNone(_a )
self.assertIsInstance(_a , (GPTaTokenizer, GPTaTokenizerFast) )
self.assertGreater(len(_a ) , 0 )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsInstance(_a , (BertTokenizer, BertTokenizerFast) )
self.assertEqual(tokenizer.vocab_size , 12 )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsInstance(_a , (RobertaTokenizer, RobertaTokenizerFast) )
self.assertEqual(tokenizer.vocab_size , 20 )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =AutoConfig.from_pretrained(_a )
self.assertIsInstance(_a , _a )
# Check that tokenizer_type ≠ model_type
__a =AutoTokenizer.from_pretrained(_a , config=_a )
self.assertIsInstance(_a , (BertTokenizer, BertTokenizerFast) )
self.assertEqual(tokenizer.vocab_size , 12 )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
with tempfile.TemporaryDirectory() as tmp_dir:
shutil.copy('./tests/fixtures/vocab.txt' , os.path.join(_a , 'vocab.txt' ) )
__a =AutoTokenizer.from_pretrained(_a , tokenizer_type='bert' , use_fast=_a )
self.assertIsInstance(_a , _a )
with tempfile.TemporaryDirectory() as tmp_dir:
shutil.copy('./tests/fixtures/vocab.json' , os.path.join(_a , 'vocab.json' ) )
shutil.copy('./tests/fixtures/merges.txt' , os.path.join(_a , 'merges.txt' ) )
__a =AutoTokenizer.from_pretrained(_a , tokenizer_type='gpt2' , use_fast=_a )
self.assertIsInstance(_a , _a )
@require_tokenizers
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
with tempfile.TemporaryDirectory() as tmp_dir:
shutil.copy('./tests/fixtures/vocab.txt' , os.path.join(_a , 'vocab.txt' ) )
__a =AutoTokenizer.from_pretrained(_a , tokenizer_type='bert' )
self.assertIsInstance(_a , _a )
with tempfile.TemporaryDirectory() as tmp_dir:
shutil.copy('./tests/fixtures/vocab.json' , os.path.join(_a , 'vocab.json' ) )
shutil.copy('./tests/fixtures/merges.txt' , os.path.join(_a , 'merges.txt' ) )
__a =AutoTokenizer.from_pretrained(_a , tokenizer_type='gpt2' )
self.assertIsInstance(_a , _a )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
with pytest.raises(_a ):
AutoTokenizer.from_pretrained('./' , tokenizer_type='xxx' )
@require_tokenizers
def __magic_name__ ( self ) -> str:
'''simple docstring'''
for tokenizer_class in [BertTokenizer, BertTokenizerFast, AutoTokenizer]:
__a =tokenizer_class.from_pretrained('wietsedv/bert-base-dutch-cased' )
self.assertIsInstance(_a , (BertTokenizer, BertTokenizerFast) )
if isinstance(_a , _a ):
self.assertEqual(tokenizer.basic_tokenizer.do_lower_case , _a )
else:
self.assertEqual(tokenizer.do_lower_case , _a )
self.assertEqual(tokenizer.model_max_length , 512 )
@require_tokenizers
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
for tokenizer_class in [BertTokenizer, BertTokenizerFast, AutoTokenizer]:
with self.assertRaisesRegex(
_a , 'julien-c/herlolip-not-exists is not a local folder and is not a valid model identifier' , ):
__a =tokenizer_class.from_pretrained('julien-c/herlolip-not-exists' )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
# tests: https://github.com/huggingface/transformers/pull/13251
# 1. models with `-`, e.g. xlm-roberta -> xlm_roberta
# 2. models that don't remap 1-1 from model-name to model file, e.g., openai-gpt -> openai
__a =TOKENIZER_MAPPING.values()
__a =[]
for slow_tok, fast_tok in tokenizers:
if slow_tok is not None:
tokenizer_names.append(slow_tok.__name__ )
if fast_tok is not None:
tokenizer_names.append(fast_tok.__name__ )
for tokenizer_name in tokenizer_names:
# must find the right class
tokenizer_class_from_name(_a )
@require_tokenizers
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.assertIsInstance(AutoTokenizer.from_pretrained('bert-base-cased' , use_fast=_a ) , _a )
self.assertIsInstance(AutoTokenizer.from_pretrained('bert-base-cased' ) , _a )
@require_tokenizers
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =AutoTokenizer.from_pretrained('distilbert-base-uncased' , do_lower_case=_a )
__a ="Hello, world. How are you?"
__a =tokenizer.tokenize(_a )
self.assertEqual('[UNK]' , tokens[0] )
__a =AutoTokenizer.from_pretrained('microsoft/mpnet-base' , do_lower_case=_a )
__a =tokenizer.tokenize(_a )
self.assertEqual('[UNK]' , tokens[0] )
@require_tokenizers
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =AutoTokenizer.from_pretrained('robot-test/dummy-tokenizer-fast-with-model-config' )
self.assertEqual(type(_a ) , _a )
self.assertEqual(tokenizer.model_max_length , 512 )
self.assertEqual(tokenizer.vocab_size , 3_0000 )
self.assertEqual(tokenizer.unk_token , '[UNK]' )
self.assertEqual(tokenizer.padding_side , 'right' )
self.assertEqual(tokenizer.truncation_side , 'right' )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsInstance(_a , (BertTokenizer, BertTokenizerFast) )
with tempfile.TemporaryDirectory() as tmp_dir:
tokenizer.save_pretrained(_a )
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsInstance(_a , tokenizer.__class__ )
self.assertEqual(tokenizera.vocab_size , 12 )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =AutoTokenizer.from_pretrained('ctrl' )
# There is no fast CTRL so this always gives us a slow tokenizer.
self.assertIsInstance(_a , _a )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
# Check we can load the tokenizer config of an online model.
__a =get_tokenizer_config('bert-base-cased' )
__a =config.pop('_commit_hash' , _a )
# If we ever update bert-base-cased tokenizer config, this dict here will need to be updated.
self.assertEqual(_a , {'do_lower_case': False} )
# This model does not have a tokenizer_config so we get back an empty dict.
__a =get_tokenizer_config(_a )
self.assertDictEqual(_a , {} )
# A tokenizer saved with `save_pretrained` always creates a tokenizer config.
__a =AutoTokenizer.from_pretrained(_a )
with tempfile.TemporaryDirectory() as tmp_dir:
tokenizer.save_pretrained(_a )
__a =get_tokenizer_config(_a )
# Check the class of the tokenizer was properly saved (note that it always saves the slow class).
self.assertEqual(config['tokenizer_class'] , 'BertTokenizer' )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
try:
AutoConfig.register('custom' , _a )
AutoTokenizer.register(_a , slow_tokenizer_class=_a )
# Trying to register something existing in the Transformers library will raise an error
with self.assertRaises(_a ):
AutoTokenizer.register(_a , slow_tokenizer_class=_a )
__a =CustomTokenizer.from_pretrained(_a )
with tempfile.TemporaryDirectory() as tmp_dir:
tokenizer.save_pretrained(_a )
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsInstance(_a , _a )
finally:
if "custom" in CONFIG_MAPPING._extra_content:
del CONFIG_MAPPING._extra_content["custom"]
if CustomConfig in TOKENIZER_MAPPING._extra_content:
del TOKENIZER_MAPPING._extra_content[CustomConfig]
@require_tokenizers
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
try:
AutoConfig.register('custom' , _a )
# Can register in two steps
AutoTokenizer.register(_a , slow_tokenizer_class=_a )
self.assertEqual(TOKENIZER_MAPPING[CustomConfig] , (CustomTokenizer, None) )
AutoTokenizer.register(_a , fast_tokenizer_class=_a )
self.assertEqual(TOKENIZER_MAPPING[CustomConfig] , (CustomTokenizer, CustomTokenizerFast) )
del TOKENIZER_MAPPING._extra_content[CustomConfig]
# Can register in one step
AutoTokenizer.register(
_a , slow_tokenizer_class=_a , fast_tokenizer_class=_a )
self.assertEqual(TOKENIZER_MAPPING[CustomConfig] , (CustomTokenizer, CustomTokenizerFast) )
# Trying to register something existing in the Transformers library will raise an error
with self.assertRaises(_a ):
AutoTokenizer.register(_a , fast_tokenizer_class=_a )
# We pass through a bert tokenizer fast cause there is no converter slow to fast for our new toknizer
# and that model does not have a tokenizer.json
with tempfile.TemporaryDirectory() as tmp_dir:
__a =BertTokenizerFast.from_pretrained(_a )
bert_tokenizer.save_pretrained(_a )
__a =CustomTokenizerFast.from_pretrained(_a )
with tempfile.TemporaryDirectory() as tmp_dir:
tokenizer.save_pretrained(_a )
__a =AutoTokenizer.from_pretrained(_a )
self.assertIsInstance(_a , _a )
__a =AutoTokenizer.from_pretrained(_a , use_fast=_a )
self.assertIsInstance(_a , _a )
finally:
if "custom" in CONFIG_MAPPING._extra_content:
del CONFIG_MAPPING._extra_content["custom"]
if CustomConfig in TOKENIZER_MAPPING._extra_content:
del TOKENIZER_MAPPING._extra_content[CustomConfig]
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
# If remote code is not set, we will time out when asking whether to load the model.
with self.assertRaises(_a ):
__a =AutoTokenizer.from_pretrained('hf-internal-testing/test_dynamic_tokenizer' )
# If remote code is disabled, we can't load this config.
with self.assertRaises(_a ):
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer' , trust_remote_code=_a )
__a =AutoTokenizer.from_pretrained('hf-internal-testing/test_dynamic_tokenizer' , trust_remote_code=_a )
self.assertTrue(tokenizer.special_attribute_present )
# Test tokenizer can be reloaded.
with tempfile.TemporaryDirectory() as tmp_dir:
tokenizer.save_pretrained(_a )
__a =AutoTokenizer.from_pretrained(_a , trust_remote_code=_a )
self.assertTrue(reloaded_tokenizer.special_attribute_present )
if is_tokenizers_available():
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizerFast' )
self.assertEqual(reloaded_tokenizer.__class__.__name__ , 'NewTokenizerFast' )
# Test we can also load the slow version
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer' , trust_remote_code=_a , use_fast=_a )
self.assertTrue(tokenizer.special_attribute_present )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizer' )
# Test tokenizer can be reloaded.
with tempfile.TemporaryDirectory() as tmp_dir:
tokenizer.save_pretrained(_a )
__a =AutoTokenizer.from_pretrained(_a , trust_remote_code=_a , use_fast=_a )
self.assertEqual(reloaded_tokenizer.__class__.__name__ , 'NewTokenizer' )
self.assertTrue(reloaded_tokenizer.special_attribute_present )
else:
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizer' )
self.assertEqual(reloaded_tokenizer.__class__.__name__ , 'NewTokenizer' )
@require_tokenizers
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
class __magic_name__ ( _a ):
SCREAMING_SNAKE_CASE = False
class __magic_name__ ( _a ):
SCREAMING_SNAKE_CASE = NewTokenizer
SCREAMING_SNAKE_CASE = False
try:
AutoConfig.register('custom' , _a )
AutoTokenizer.register(_a , slow_tokenizer_class=_a )
AutoTokenizer.register(_a , fast_tokenizer_class=_a )
# If remote code is not set, the default is to use local
__a =AutoTokenizer.from_pretrained('hf-internal-testing/test_dynamic_tokenizer' )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizerFast' )
self.assertFalse(tokenizer.special_attribute_present )
__a =AutoTokenizer.from_pretrained('hf-internal-testing/test_dynamic_tokenizer' , use_fast=_a )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizer' )
self.assertFalse(tokenizer.special_attribute_present )
# If remote code is disabled, we load the local one.
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer' , trust_remote_code=_a )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizerFast' )
self.assertFalse(tokenizer.special_attribute_present )
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer' , trust_remote_code=_a , use_fast=_a )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizer' )
self.assertFalse(tokenizer.special_attribute_present )
# If remote is enabled, we load from the Hub
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer' , trust_remote_code=_a )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizerFast' )
self.assertTrue(tokenizer.special_attribute_present )
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer' , trust_remote_code=_a , use_fast=_a )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizer' )
self.assertTrue(tokenizer.special_attribute_present )
finally:
if "custom" in CONFIG_MAPPING._extra_content:
del CONFIG_MAPPING._extra_content["custom"]
if CustomConfig in TOKENIZER_MAPPING._extra_content:
del TOKENIZER_MAPPING._extra_content[CustomConfig]
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer_legacy' , trust_remote_code=_a )
self.assertTrue(tokenizer.special_attribute_present )
if is_tokenizers_available():
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizerFast' )
# Test we can also load the slow version
__a =AutoTokenizer.from_pretrained(
'hf-internal-testing/test_dynamic_tokenizer_legacy' , trust_remote_code=_a , use_fast=_a )
self.assertTrue(tokenizer.special_attribute_present )
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizer' )
else:
self.assertEqual(tokenizer.__class__.__name__ , 'NewTokenizer' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
with self.assertRaisesRegex(
_a , 'bert-base is not a local folder and is not a valid model identifier' ):
__a =AutoTokenizer.from_pretrained('bert-base' )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
with self.assertRaisesRegex(
_a , R'aaaaaa is not a valid git identifier \(branch name, tag name or commit id\)' ):
__a =AutoTokenizer.from_pretrained(_a , revision='aaaaaa' )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
# Make sure we have cached the tokenizer.
__a =AutoTokenizer.from_pretrained('hf-internal-testing/tiny-random-bert' )
with RequestCounter() as counter:
__a =AutoTokenizer.from_pretrained('hf-internal-testing/tiny-random-bert' )
self.assertEqual(counter.get_request_count , 0 )
self.assertEqual(counter.head_request_count , 1 )
self.assertEqual(counter.other_request_count , 0 )
| 370
|
import unittest
from transformers import XLMConfig, is_torch_available
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 (
XLMForMultipleChoice,
XLMForQuestionAnswering,
XLMForQuestionAnsweringSimple,
XLMForSequenceClassification,
XLMForTokenClassification,
XLMModel,
XLMWithLMHeadModel,
)
from transformers.models.xlm.modeling_xlm import XLM_PRETRAINED_MODEL_ARCHIVE_LIST
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=13 , __snake_case=7 , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=False , __snake_case=False , __snake_case=False , __snake_case=2 , __snake_case=99 , __snake_case=0 , __snake_case=32 , __snake_case=5 , __snake_case=4 , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=2 , __snake_case=0.02 , __snake_case=2 , __snake_case=4 , __snake_case="last" , __snake_case=True , __snake_case=None , __snake_case=0 , ) -> Optional[Any]:
'''simple docstring'''
__a =parent
__a =batch_size
__a =seq_length
__a =is_training
__a =use_input_lengths
__a =use_token_type_ids
__a =use_labels
__a =gelu_activation
__a =sinusoidal_embeddings
__a =causal
__a =asm
__a =n_langs
__a =vocab_size
__a =n_special
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_sequence_label_size
__a =initializer_range
__a =num_labels
__a =num_choices
__a =summary_type
__a =use_proj
__a =scope
__a =bos_token_id
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a =random_attention_mask([self.batch_size, self.seq_length] )
__a =None
if self.use_input_lengths:
__a =(
ids_tensor([self.batch_size] , vocab_size=2 ) + self.seq_length - 2
) # small variation of seq_length
__a =None
if self.use_token_type_ids:
__a =ids_tensor([self.batch_size, self.seq_length] , self.n_langs )
__a =None
__a =None
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a =ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a =ids_tensor([self.batch_size] , 2 ).float()
__a =ids_tensor([self.batch_size] , self.num_choices )
__a =self.get_config()
return (
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
)
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return XLMConfig(
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 , num_labels=self.num_labels , bos_token_id=self.bos_token_id , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMModel(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , lengths=__snake_case , langs=__snake_case )
__a =model(__snake_case , langs=__snake_case )
__a =model(__snake_case )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[int]:
'''simple docstring'''
__a =XLMWithLMHeadModel(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , token_type_ids=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Dict:
'''simple docstring'''
__a =XLMForQuestionAnsweringSimple(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
__a =outputs
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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMForQuestionAnswering(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , p_mask=__snake_case , )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , )
((__a) , ) =result_with_labels.to_tuple()
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
((__a) , ) =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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[Any]:
'''simple docstring'''
__a =XLMForSequenceClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Any:
'''simple docstring'''
__a =self.num_labels
__a =XLMForTokenClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , attention_mask=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Tuple:
'''simple docstring'''
__a =self.num_choices
__a =XLMForMultipleChoice(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =model(
__snake_case , attention_mask=__snake_case , token_type_ids=__snake_case , labels=__snake_case , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) =config_and_inputs
__a ={'input_ids': input_ids, 'token_type_ids': token_type_ids, 'lengths': input_lengths}
return config, inputs_dict
@require_torch
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ , lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
XLMModel,
XLMWithLMHeadModel,
XLMForQuestionAnswering,
XLMForSequenceClassification,
XLMForQuestionAnsweringSimple,
XLMForTokenClassification,
XLMForMultipleChoice,
)
if is_torch_available()
else ()
)
SCREAMING_SNAKE_CASE = (
(XLMWithLMHeadModel,) if is_torch_available() else ()
) # TODO (PVP): Check other models whether language generation is also applicable
SCREAMING_SNAKE_CASE = (
{
'feature-extraction': XLMModel,
'fill-mask': XLMWithLMHeadModel,
'question-answering': XLMForQuestionAnsweringSimple,
'text-classification': XLMForSequenceClassification,
'text-generation': XLMWithLMHeadModel,
'token-classification': XLMForTokenClassification,
'zero-shot': XLMForSequenceClassification,
}
if is_torch_available()
else {}
)
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> int:
'''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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case=False ) -> str:
'''simple docstring'''
__a =super()._prepare_for_class(__snake_case , __snake_case , return_labels=__snake_case )
if return_labels:
if model_class.__name__ == "XLMForQuestionAnswering":
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
return inputs_dict
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMModelTester(self )
__a =ConfigTester(self , config_class=__snake_case , emb_dim=37 )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.config_tester.run_common_tests()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_model(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_lm_head(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_simple_qa(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_qa(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_sequence_classif(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_token_classif(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_for_multiple_choice(*__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Optional[Any]:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_attentions in attentions] , [True] * len(__snake_case ) )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_attentions in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =min_length + idx + 1
__a =(
batch_size * num_beam_groups,
config.num_attention_heads,
tgt_len,
src_len,
)
# check attn size
self.assertListEqual(
[layer_attention.shape for layer_attention in iter_attentions] , [expected_shape] * len(__snake_case ) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Dict:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_hidden_states in hidden_states] , [True] * len(__snake_case ) , )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_hidden_states in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =(batch_size * num_beam_groups, seq_len, config.hidden_size)
# check hidden size
self.assertListEqual(
[layer_hidden_states.shape for layer_hidden_states in iter_hidden_states] , [expected_shape] * len(__snake_case ) , )
pass
@slow
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
for model_name in XLM_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =XLMModel.from_pretrained(__snake_case )
self.assertIsNotNone(__snake_case )
@require_torch
class __magic_name__ ( unittest.TestCase ):
@slow
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMWithLMHeadModel.from_pretrained('xlm-mlm-en-2048' )
model.to(__snake_case )
__a =torch.tensor([[14, 447]] , dtype=torch.long , device=__snake_case ) # the president
__a =[
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
] # the president the president the president the president the president the president the president the president the president the president
# TODO(PVP): this and other input_ids I tried for generation give pretty bad results. Not sure why. Model might just not be made for auto-regressive inference
__a =model.generate(__snake_case , do_sample=__snake_case )
self.assertListEqual(output_ids[0].cpu().numpy().tolist() , __snake_case )
| 308
| 0
|
"""simple docstring"""
import os
from collections import deque
import torch
from torch.utils.data import Dataset
class __magic_name__ ( _a ):
def __init__( self , __snake_case="" , __snake_case="train" ) -> Optional[int]:
'''simple docstring'''
assert os.path.isdir(snake_case_ )
__a =[]
__a =os.listdir(snake_case_ )
for story_filename in story_filenames_list:
if "summary" in story_filename:
continue
__a =os.path.join(snake_case_ , snake_case_ )
if not os.path.isfile(snake_case_ ):
continue
self.documents.append(snake_case_ )
def __len__( self ) -> Union[str, Any]:
'''simple docstring'''
return len(self.documents )
def __getitem__( self , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =self.documents[idx]
__a =document_path.split('/' )[-1]
with open(snake_case_ , encoding='utf-8' ) as source:
__a =source.read()
__a =process_story(snake_case_ )
return document_name, story_lines, summary_lines
def UpperCamelCase_( _snake_case : Any ):
"""simple docstring"""
__a =list(filter(lambda _snake_case : len(_lowerCAmelCase ) != 0 , [line.strip() for line in raw_story.split('\n' )] ) )
# for some unknown reason some lines miss a period, add it
__a =[_add_missing_period(_lowerCAmelCase ) for line in nonempty_lines]
# gather article lines
__a =[]
__a =deque(_lowerCAmelCase )
while True:
try:
__a =lines.popleft()
if element.startswith('@highlight' ):
break
story_lines.append(_lowerCAmelCase )
except IndexError:
# if "@highlight" is absent from the file we pop
# all elements until there is None, raising an exception.
return story_lines, []
# gather summary lines
__a =list(filter(lambda _snake_case : not t.startswith('@highlight' ) , _lowerCAmelCase ) )
return story_lines, summary_lines
def UpperCamelCase_( _snake_case : Any ):
"""simple docstring"""
__a =[""".""", """!""", """?""", """...""", """'""", """`""", """\"""", """\u2019""", """\u2019""", """)"""]
if line.startswith('@highlight' ):
return line
if line[-1] in END_TOKENS:
return line
return line + "."
def UpperCamelCase_( _snake_case : str , _snake_case : int , _snake_case : str ):
"""simple docstring"""
if len(_lowerCAmelCase ) > block_size:
return sequence[:block_size]
else:
sequence.extend([pad_token_id] * (block_size - len(_lowerCAmelCase )) )
return sequence
def UpperCamelCase_( _snake_case : int , _snake_case : Union[str, Any] ):
"""simple docstring"""
__a =torch.ones_like(_lowerCAmelCase )
__a =sequence == pad_token_id
__a =0
return mask
def UpperCamelCase_( _snake_case : Dict , _snake_case : Tuple , _snake_case : Tuple ):
"""simple docstring"""
__a =[tokenizer.encode(_lowerCAmelCase ) for line in story_lines]
__a =[token for sentence in story_lines_token_ids for token in sentence]
__a =[tokenizer.encode(_lowerCAmelCase ) for line in summary_lines]
__a =[token for sentence in summary_lines_token_ids for token in sentence]
return story_token_ids, summary_token_ids
def UpperCamelCase_( _snake_case : List[str] , _snake_case : List[str] ):
"""simple docstring"""
__a =[]
for sequence in batch:
__a =-1
__a =[]
for s in sequence:
if s == separator_token_id:
sentence_num += 1
embeddings.append(sentence_num % 2 )
batch_embeddings.append(_lowerCAmelCase )
return torch.tensor(_lowerCAmelCase )
| 371
|
from __future__ import annotations
import matplotlib.pyplot as plt # type: ignore
import numpy
# initial triangle of Koch snowflake
_lowerCAmelCase : Optional[Any] = numpy.array([0, 0])
_lowerCAmelCase : Dict = numpy.array([0.5, 0.8660254])
_lowerCAmelCase : Any = numpy.array([1, 0])
_lowerCAmelCase : int = [VECTOR_1, VECTOR_2, VECTOR_3, VECTOR_1]
def UpperCamelCase_( _snake_case : list[numpy.ndarray] , _snake_case : int ):
"""simple docstring"""
__a =initial_vectors
for _ in range(_snake_case ):
__a =iteration_step(_snake_case )
return vectors
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =[]
for i, start_vector in enumerate(vectors[:-1] ):
__a =vectors[i + 1]
new_vectors.append(_snake_case )
__a =end_vector - start_vector
new_vectors.append(start_vector + difference_vector / 3 )
new_vectors.append(
start_vector + difference_vector / 3 + rotate(difference_vector / 3 , 60 ) )
new_vectors.append(start_vector + difference_vector * 2 / 3 )
new_vectors.append(vectors[-1] )
return new_vectors
def UpperCamelCase_( _snake_case : numpy.ndarray , _snake_case : float ):
"""simple docstring"""
__a =numpy.radians(_snake_case )
__a , __a =numpy.cos(_snake_case ), numpy.sin(_snake_case )
__a =numpy.array(((c, -s), (s, c)) )
return numpy.dot(_snake_case , _snake_case )
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =plt.gca()
axes.set_aspect('equal' )
# matplotlib.pyplot.plot takes a list of all x-coordinates and a list of all
# y-coordinates as inputs, which are constructed from the vector-list using
# zip()
__a , __a =zip(*_snake_case )
plt.plot(_snake_case , _snake_case )
plt.show()
if __name__ == "__main__":
import doctest
doctest.testmod()
_lowerCAmelCase : List[Any] = iterate(INITIAL_VECTORS, 5)
plot(processed_vectors)
| 308
| 0
|
from sklearn.metrics import mean_squared_error
import datasets
_lowerCAmelCase : Union[str, Any] = "\\n@article{scikit-learn,\n title={Scikit-learn: Machine Learning in {P}ython},\n author={Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V.\n and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P.\n and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and\n Cournapeau, D. and Brucher, M. and Perrot, M. and Duchesnay, E.},\n journal={Journal of Machine Learning Research},\n volume={12},\n pages={2825--2830},\n year={2011}\n}\n"
_lowerCAmelCase : str = "\\nMean Squared Error(MSE) is the average of the square of difference between the predicted\nand actual values.\n"
_lowerCAmelCase : Union[str, Any] = "\nArgs:\n predictions: array-like of shape (n_samples,) or (n_samples, n_outputs)\n Estimated target values.\n references: array-like of shape (n_samples,) or (n_samples, n_outputs)\n Ground truth (correct) target values.\n sample_weight: array-like of shape (n_samples,), default=None\n Sample weights.\n multioutput: {\"raw_values\", \"uniform_average\"} or array-like of shape (n_outputs,), default=\"uniform_average\"\n Defines aggregating of multiple output values. Array-like value defines weights used to average errors.\n\n \"raw_values\" : Returns a full set of errors in case of multioutput input.\n\n \"uniform_average\" : Errors of all outputs are averaged with uniform weight.\n\n squared : bool, default=True\n If True returns MSE value, if False returns RMSE (Root Mean Squared Error) value.\n\nReturns:\n mse : mean squared error.\nExamples:\n\n >>> mse_metric = datasets.load_metric(\"mse\")\n >>> predictions = [2.5, 0.0, 2, 8]\n >>> references = [3, -0.5, 2, 7]\n >>> results = mse_metric.compute(predictions=predictions, references=references)\n >>> print(results)\n {\'mse\': 0.375}\n >>> rmse_result = mse_metric.compute(predictions=predictions, references=references, squared=False)\n >>> print(rmse_result)\n {\'mse\': 0.6123724356957945}\n\n If you\'re using multi-dimensional lists, then set the config as follows :\n\n >>> mse_metric = datasets.load_metric(\"mse\", \"multilist\")\n >>> predictions = [[0.5, 1], [-1, 1], [7, -6]]\n >>> references = [[0, 2], [-1, 2], [8, -5]]\n >>> results = mse_metric.compute(predictions=predictions, references=references)\n >>> print(results)\n {\'mse\': 0.7083333333333334}\n >>> results = mse_metric.compute(predictions=predictions, references=references, multioutput=\'raw_values\')\n >>> print(results) # doctest: +NORMALIZE_WHITESPACE\n {\'mse\': array([0.41666667, 1. ])}\n"
@datasets.utils.file_utils.add_start_docstrings(_DESCRIPTION , _KWARGS_DESCRIPTION )
class __magic_name__ ( datasets.Metric ):
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
return datasets.MetricInfo(
description=_DESCRIPTION , citation=_CITATION , inputs_description=_KWARGS_DESCRIPTION , features=datasets.Features(self._get_feature_types() ) , reference_urls=[
'https://scikit-learn.org/stable/modules/generated/sklearn.metrics.mean_squared_error.html'
] , )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.config_name == "multilist":
return {
"predictions": datasets.Sequence(datasets.Value('float' ) ),
"references": datasets.Sequence(datasets.Value('float' ) ),
}
else:
return {
"predictions": datasets.Value('float' ),
"references": datasets.Value('float' ),
}
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case=None , __snake_case="uniform_average" , __snake_case=True ) -> Optional[int]:
'''simple docstring'''
__a =mean_squared_error(
__lowercase , __lowercase , sample_weight=__lowercase , multioutput=__lowercase , squared=__lowercase )
return {"mse": mse}
| 350
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
_lowerCAmelCase : Tuple = {
"configuration_megatron_bert": ["MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MegatronBertConfig"],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : List[str] = [
"MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"MegatronBertForCausalLM",
"MegatronBertForMaskedLM",
"MegatronBertForMultipleChoice",
"MegatronBertForNextSentencePrediction",
"MegatronBertForPreTraining",
"MegatronBertForQuestionAnswering",
"MegatronBertForSequenceClassification",
"MegatronBertForTokenClassification",
"MegatronBertModel",
"MegatronBertPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_megatron_bert import MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MegatronBertConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_megatron_bert import (
MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
MegatronBertForCausalLM,
MegatronBertForMaskedLM,
MegatronBertForMultipleChoice,
MegatronBertForNextSentencePrediction,
MegatronBertForPreTraining,
MegatronBertForQuestionAnswering,
MegatronBertForSequenceClassification,
MegatronBertForTokenClassification,
MegatronBertModel,
MegatronBertPreTrainedModel,
)
else:
import sys
_lowerCAmelCase : Optional[int] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 0
|
import inspect
import unittest
from huggingface_hub import hf_hub_download
from transformers import ASTConfig
from transformers.testing_utils import require_torch, require_torchaudio, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_torchaudio_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import ASTForAudioClassification, ASTModel
from transformers.models.audio_spectrogram_transformer.modeling_audio_spectrogram_transformer import (
AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
)
if is_torchaudio_available():
import torchaudio
from transformers import ASTFeatureExtractor
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=13 , __snake_case=2 , __snake_case=24 , __snake_case=16 , __snake_case=True , __snake_case=True , __snake_case=32 , __snake_case=5 , __snake_case=4 , __snake_case=37 , __snake_case="gelu" , __snake_case=0.1 , __snake_case=0.1 , __snake_case=10 , __snake_case=0.02 , __snake_case=None , __snake_case=2 , __snake_case=2 , ) -> Any:
'''simple docstring'''
__a =parent
__a =batch_size
__a =patch_size
__a =max_length
__a =num_mel_bins
__a =is_training
__a =use_labels
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =type_sequence_label_size
__a =initializer_range
__a =scope
__a =frequency_stride
__a =time_stride
# in AST, the seq length equals the number of patches + 2 (we add 2 for the [CLS] and distillation tokens)
__a =(self.num_mel_bins - self.patch_size) // self.frequency_stride + 1
__a =(self.max_length - self.patch_size) // self.time_stride + 1
__a =frequency_out_dimension * time_out_dimension
__a =num_patches + 2
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =floats_tensor([self.batch_size, self.max_length, self.num_mel_bins] )
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a =self.get_config()
return config, input_values, labels
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
return ASTConfig(
patch_size=self.patch_size , max_length=self.max_length , num_mel_bins=self.num_mel_bins , 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 , is_decoder=__snake_case , initializer_range=self.initializer_range , frequency_stride=self.frequency_stride , time_stride=self.time_stride , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =ASTModel(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
(
__a
) =config_and_inputs
__a ={'input_values': input_values}
return config, inputs_dict
@require_torch
class __magic_name__ ( __UpperCamelCase , __UpperCamelCase , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
ASTModel,
ASTForAudioClassification,
)
if is_torch_available()
else ()
)
SCREAMING_SNAKE_CASE = (
{'audio-classification': ASTForAudioClassification, 'feature-extraction': ASTModel}
if is_torch_available()
else {}
)
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> Optional[Any]:
'''simple docstring'''
if pipeline_test_casse_name == "AudioClassificationPipelineTests":
return True
return False
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =ASTModelTester(self )
__a =ConfigTester(self , config_class=__snake_case , has_text_modality=__snake_case , hidden_size=37 )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.config_tester.run_common_tests()
@unittest.skip(reason='AST does not use inputs_embeds' )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
__a =model_class(__snake_case )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
__a =model.get_output_embeddings()
self.assertTrue(x is None or isinstance(__snake_case , nn.Linear ) )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
__a =model_class(__snake_case )
__a =inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
__a =[*signature.parameters.keys()]
__a =['input_values']
self.assertListEqual(arg_names[:1] , __snake_case )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*__snake_case )
@slow
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
for model_name in AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =ASTModel.from_pretrained(__snake_case )
self.assertIsNotNone(__snake_case )
def UpperCamelCase_( ):
"""simple docstring"""
__a =hf_hub_download(
repo_id='nielsr/audio-spectogram-transformer-checkpoint' , filename='sample_audio.flac' , repo_type='dataset' )
__a =torchaudio.load(_UpperCAmelCase )
return audio, sampling_rate
@require_torch
@require_torchaudio
class __magic_name__ ( unittest.TestCase ):
@cached_property
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
return (
ASTFeatureExtractor.from_pretrained('MIT/ast-finetuned-audioset-10-10-0.4593' )
if is_torchaudio_available()
else None
)
@slow
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.default_feature_extractor
__a =ASTForAudioClassification.from_pretrained('MIT/ast-finetuned-audioset-10-10-0.4593' ).to(__snake_case )
__a =self.default_feature_extractor
__a =prepare_audio()
__a =audio.squeeze().numpy()
__a =feature_extractor(__snake_case , sampling_rate=__snake_case , return_tensors='pt' ).to(__snake_case )
# forward pass
with torch.no_grad():
__a =model(**__snake_case )
# verify the logits
__a =torch.Size((1, 527) )
self.assertEqual(outputs.logits.shape , __snake_case )
__a =torch.tensor([-0.8760, -7.0042, -8.6602] ).to(__snake_case )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , __snake_case , atol=1e-4 ) )
| 351
|
import hashlib
import unittest
from transformers import MODEL_FOR_DEPTH_ESTIMATION_MAPPING, is_torch_available, is_vision_available
from transformers.pipelines import DepthEstimationPipeline, pipeline
from transformers.testing_utils import (
is_pipeline_test,
nested_simplify,
require_tf,
require_timm,
require_torch,
require_vision,
slow,
)
from .test_pipelines_common import ANY
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
else:
class __magic_name__ :
@staticmethod
def __magic_name__ ( *__snake_case , **__snake_case ) -> List[str]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : Image ):
"""simple docstring"""
__a =hashlib.mda(image.tobytes() )
return m.hexdigest()
@is_pipeline_test
@require_vision
@require_timm
@require_torch
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = MODEL_FOR_DEPTH_ESTIMATION_MAPPING
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Dict:
'''simple docstring'''
__a =DepthEstimationPipeline(model=__snake_case , image_processor=__snake_case )
return depth_estimator, [
"./tests/fixtures/tests_samples/COCO/000000039769.png",
"./tests/fixtures/tests_samples/COCO/000000039769.png",
]
def __magic_name__ ( self , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
__a =depth_estimator('./tests/fixtures/tests_samples/COCO/000000039769.png' )
self.assertEqual({'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )} , __snake_case )
import datasets
__a =datasets.load_dataset('hf-internal-testing/fixtures_image_utils' , 'image' , split='test' )
__a =depth_estimator(
[
Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' ),
'http://images.cocodataset.org/val2017/000000039769.jpg',
# RGBA
dataset[0]['file'],
# LA
dataset[1]['file'],
# L
dataset[2]['file'],
] )
self.assertEqual(
[
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
] , __snake_case , )
@require_tf
@unittest.skip('Depth estimation is not implemented in TF' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
pass
@slow
@require_torch
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a ='Intel/dpt-large'
__a =pipeline('depth-estimation' , model=__snake_case )
__a =depth_estimator('http://images.cocodataset.org/val2017/000000039769.jpg' )
__a =hashimage(outputs['depth'] )
# This seems flaky.
# self.assertEqual(outputs["depth"], "1a39394e282e9f3b0741a90b9f108977")
self.assertEqual(nested_simplify(outputs['predicted_depth'].max().item() ) , 29.304 )
self.assertEqual(nested_simplify(outputs['predicted_depth'].min().item() ) , 2.662 )
@require_torch
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
# This is highly irregular to have no small tests.
self.skipTest('There is not hf-internal-testing tiny model for either GLPN nor DPT' )
| 308
| 0
|
import os # noqa: this is just for tests
import os as renamed_os # noqa: this is just for tests
from os import path # noqa: this is just for tests
from os import path as renamed_path # noqa: this is just for tests
from os.path import join # noqa: this is just for tests
from os.path import join as renamed_join # noqa: this is just for tests
_lowerCAmelCase : Any = open # noqa: we just need to have a builtin inside this module to test it properly
| 352
|
import argparse
import dataclasses
import json
import logging
import os
import shutil
from typing import List, Optional
import datasets
from accelerate import Accelerator
from datasets import load_dataset
from finetuning import finetune
from tqdm.auto import tqdm
import transformers
from transformers import AutoConfig, set_seed
from transformers.trainer_utils import IntervalStrategy
_lowerCAmelCase : Optional[int] = logging.getLogger(__name__)
_lowerCAmelCase : Any = "pytorch_model.bin"
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co.'} , )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the data to predict on.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The name of the task to train on.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The list of labels for the task.'} )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'The output directory where the model predictions and checkpoints will be written.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='accuracy' , metadata={'help': 'The evaluation metric used for the task.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='no' , metadata={
'help': 'The evaluation strategy to adopt during training. Possible values are: ["no", "step", "epoch]'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={
'help': 'How much the specified evaluation metric must improve to satisfy early stopping conditions.'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the confidence score.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the validation performance.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to fine-tune on labeled data after pseudo training.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={'help': 'Confidence threshold for pseudo-labeled data filtering.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Random seed for initialization.'} , )
def UpperCamelCase_( _snake_case : int , _snake_case : str , _snake_case : Optional[int] , _snake_case : str , _snake_case : Union[str, Any] , _snake_case : List[Any] ):
"""simple docstring"""
__a =datasets.concatenate_datasets([infer_input, infer_output] , axis=1 )
if args.do_filter_by_confidence:
__a =dataset.filter(lambda _snake_case : example["probability"] > args.confidence_threshold )
if args.do_filter_by_val_performance:
assert eval_result >= 0.0 and eval_result <= 1.0
__a =int(eval_result * len(_snake_case ) )
print(_snake_case )
__a =dataset.sort('probability' , reverse=_snake_case )
__a =dataset.select(range(_snake_case ) )
__a =dataset.remove_columns(['label', 'probability'] )
__a =dataset.rename_column('prediction' , 'label' )
__a =dataset.map(lambda _snake_case : {"label": idalabel[example["label"]]} )
__a =dataset.shuffle(seed=args.seed )
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.data_file_extension == "csv":
dataset.to_csv(_snake_case , index=_snake_case )
else:
dataset.to_json(_snake_case )
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : str , _snake_case : int , _snake_case : Optional[int] , **_snake_case : List[str] ):
"""simple docstring"""
__a =Accelerator()
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , level=logging.INFO , )
logger.info(accelerator.state )
# Setup logging, we only want one process per machine to log things on the
# screen. accelerator.is_local_main_process is only True for one process per
# machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR )
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
__a =STModelArguments(model_name_or_path=_snake_case )
__a =STDataArguments(train_file=_snake_case , infer_file=_snake_case )
__a =STTrainingArguments(output_dir=_snake_case )
__a =argparse.Namespace()
for arg_class in (model_args, data_args, training_args):
for key, value in vars(_snake_case ).items():
setattr(_snake_case , _snake_case , _snake_case )
for key, value in kwargs.items():
if hasattr(_snake_case , _snake_case ):
setattr(_snake_case , _snake_case , _snake_case )
# Sanity checks
__a ={}
__a =None
# You need to provide the training data and the data to predict on
assert args.train_file is not None
assert args.infer_file is not None
__a =args.train_file
__a =args.infer_file
if args.evaluation_strategy != IntervalStrategy.NO.value:
assert args.eval_file is not None
__a =args.eval_file
for key in data_files:
__a =data_files[key].split('.' )[-1]
assert extension in ["csv", "json"], F'`{key}_file` should be a csv or a json file.'
if args.data_file_extension is None:
__a =extension
else:
assert extension == args.data_file_extension, F'`{key}_file` should be a {args.data_file_extension} file`.'
assert (
args.eval_metric in datasets.list_metrics()
), F'{args.eval_metric} not in the list of supported metrics {datasets.list_metrics()}.'
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed )
logger.info('Creating the initial data directory for self-training...' )
__a =F'{args.output_dir}/self-train_iter-{{}}'.format
__a =data_dir_format(0 )
if accelerator.is_main_process:
if args.output_dir is not None:
os.makedirs(args.output_dir , exist_ok=_snake_case )
os.makedirs(_snake_case , exist_ok=_snake_case )
accelerator.wait_for_everyone()
__a =None
__a =None
__a =0
__a =False
# Show the progress bar
__a =tqdm(range(args.max_selftrain_iterations ) , disable=not accelerator.is_local_main_process )
# Self-train
for iteration in range(0 , int(args.max_selftrain_iterations ) ):
__a =data_dir_format(_snake_case )
assert os.path.exists(_snake_case )
# Stage 1: initial fine-tuning for iteration = 0 or pseudo-training for
# iteration > 0
__a =os.path.join(_snake_case , 'stage-1' )
__a ={
'accelerator': accelerator,
'model_name_or_path': args.model_name_or_path,
'cache_dir': args.cache_dir,
'do_train': True,
'train_file': data_files['train'] if iteration == 0 else data_files['train_pseudo'],
'do_eval': True if args.eval_file is not None else False,
'eval_file': data_files['eval'],
'do_predict': True,
'infer_file': data_files['infer'],
'task_name': args.task_name,
'label_list': args.label_list,
'output_dir': current_output_dir,
'eval_metric': args.eval_metric,
'evaluation_strategy': args.evaluation_strategy,
'early_stopping_patience': args.early_stopping_patience,
'early_stopping_threshold': args.early_stopping_threshold,
'seed': args.seed,
}
# Add additional training arguments
for key, value in kwargs.items():
if key not in arguments_dict and not hasattr(_snake_case , _snake_case ):
arguments_dict.update({key: value} )
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 1.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 1 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 1.' , _snake_case )
if iteration > 0 and args.finetune_on_labeled_data:
# Stage 2 (optional): fine-tuning on the original labeled data
__a =os.path.join(_snake_case , 'best-checkpoint' )
__a =os.path.join(_snake_case , 'stage-2' )
# Update arguments_dict
__a =model_path
__a =data_files['train']
__a =current_output_dir
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 2.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 2 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 2.' , _snake_case )
__a =iteration
__a =data_dir_format(iteration + 1 )
__a =AutoConfig.from_pretrained(os.path.join(_snake_case , 'best-checkpoint' ) )
__a =config.idalabel
__a =os.path.join(_snake_case , 'eval_results_best-checkpoint.json' )
__a =os.path.join(_snake_case , 'test_results_best-checkpoint.json' )
assert os.path.exists(_snake_case )
with open(_snake_case , 'r' ) as f:
__a =float(json.load(_snake_case )[args.eval_metric] )
__a =os.path.join(_snake_case , 'infer_output_best-checkpoint.csv' )
assert os.path.exists(_snake_case )
# Loading the dataset from local csv or json files.
__a =load_dataset(args.data_file_extension , data_files={'data': data_files['infer']} )['data']
__a =load_dataset('csv' , data_files={'data': infer_output_file} )['data']
if accelerator.is_main_process:
os.makedirs(_snake_case , exist_ok=_snake_case )
shutil.copy(_snake_case , os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) )
if os.path.exists(_snake_case ):
shutil.copy(_snake_case , os.path.join(_snake_case , F'test_results_iter-{iteration}.json' ) )
create_pseudo_labeled_data(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )
accelerator.wait_for_everyone()
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.evaluation_strategy != IntervalStrategy.NO.value:
__a =eval_result
if best_iteration is None:
__a =new_iteration
__a =new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
__a =new_iteration
__a =new_eval_result
__a =0
else:
if new_eval_result == best_eval_result:
__a =new_iteration
__a =new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
__a =True
progress_bar.update(1 )
if should_training_stop:
break
if best_iteration is not None:
# Save the best iteration
logger.info('Best iteration: %d' , _snake_case )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
else:
# Assume that the last iteration is the best
logger.info('Best iteration: %d' , args.max_selftrain_iterations - 1 )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{args.max_selftrain_iterations - 1}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
| 308
| 0
|
import os
import sys
import unittest
_lowerCAmelCase : Tuple = os.path.abspath(os.path.dirname(os.path.dirname(os.path.dirname(__file__))))
sys.path.append(os.path.join(git_repo_path, "utils"))
import get_test_info # noqa: E402
from get_test_info import ( # noqa: E402
get_model_to_test_mapping,
get_model_to_tester_mapping,
get_test_to_tester_mapping,
)
_lowerCAmelCase : Union[str, Any] = os.path.join("tests", "models", "bert", "test_modeling_bert.py")
_lowerCAmelCase : str = os.path.join("tests", "models", "blip", "test_modeling_blip.py")
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ):
'''simple docstring'''
__a =get_test_to_tester_mapping(__snake_case )
__a =get_test_to_tester_mapping(__snake_case )
__a ={'BertModelTest': 'BertModelTester'}
__a ={
'BlipModelTest': 'BlipModelTester',
'BlipTextImageModelTest': 'BlipTextImageModelsModelTester',
'BlipTextModelTest': 'BlipTextModelTester',
'BlipTextRetrievalModelTest': 'BlipTextRetrievalModelTester',
'BlipVQAModelTest': 'BlipVQAModelTester',
'BlipVisionModelTest': 'BlipVisionModelTester',
}
self.assertEqual(get_test_info.to_json(__snake_case ) , __snake_case )
self.assertEqual(get_test_info.to_json(__snake_case ) , __snake_case )
def __magic_name__ ( self ):
'''simple docstring'''
__a =get_model_to_test_mapping(__snake_case )
__a =get_model_to_test_mapping(__snake_case )
__a ={
'BertForMaskedLM': ['BertModelTest'],
'BertForMultipleChoice': ['BertModelTest'],
'BertForNextSentencePrediction': ['BertModelTest'],
'BertForPreTraining': ['BertModelTest'],
'BertForQuestionAnswering': ['BertModelTest'],
'BertForSequenceClassification': ['BertModelTest'],
'BertForTokenClassification': ['BertModelTest'],
'BertLMHeadModel': ['BertModelTest'],
'BertModel': ['BertModelTest'],
}
__a ={
'BlipForConditionalGeneration': ['BlipTextImageModelTest'],
'BlipForImageTextRetrieval': ['BlipTextRetrievalModelTest'],
'BlipForQuestionAnswering': ['BlipVQAModelTest'],
'BlipModel': ['BlipModelTest'],
'BlipTextModel': ['BlipTextModelTest'],
'BlipVisionModel': ['BlipVisionModelTest'],
}
self.assertEqual(get_test_info.to_json(__snake_case ) , __snake_case )
self.assertEqual(get_test_info.to_json(__snake_case ) , __snake_case )
def __magic_name__ ( self ):
'''simple docstring'''
__a =get_model_to_tester_mapping(__snake_case )
__a =get_model_to_tester_mapping(__snake_case )
__a ={
'BertForMaskedLM': ['BertModelTester'],
'BertForMultipleChoice': ['BertModelTester'],
'BertForNextSentencePrediction': ['BertModelTester'],
'BertForPreTraining': ['BertModelTester'],
'BertForQuestionAnswering': ['BertModelTester'],
'BertForSequenceClassification': ['BertModelTester'],
'BertForTokenClassification': ['BertModelTester'],
'BertLMHeadModel': ['BertModelTester'],
'BertModel': ['BertModelTester'],
}
__a ={
'BlipForConditionalGeneration': ['BlipTextImageModelsModelTester'],
'BlipForImageTextRetrieval': ['BlipTextRetrievalModelTester'],
'BlipForQuestionAnswering': ['BlipVQAModelTester'],
'BlipModel': ['BlipModelTester'],
'BlipTextModel': ['BlipTextModelTester'],
'BlipVisionModel': ['BlipVisionModelTester'],
}
self.assertEqual(get_test_info.to_json(__snake_case ) , __snake_case )
self.assertEqual(get_test_info.to_json(__snake_case ) , __snake_case )
| 353
|
import shutil
import tempfile
import unittest
from transformers import (
SPIECE_UNDERLINE,
AddedToken,
BatchEncoding,
NllbTokenizer,
NllbTokenizerFast,
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
_lowerCAmelCase : Any = get_tests_dir("fixtures/test_sentencepiece.model")
if is_torch_available():
from transformers.models.mam_aaa.modeling_mam_aaa import shift_tokens_right
_lowerCAmelCase : List[Any] = 256_047
_lowerCAmelCase : Dict = 256_145
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = NllbTokenizer
SCREAMING_SNAKE_CASE = NllbTokenizerFast
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = {}
def __magic_name__ ( self ) -> int:
'''simple docstring'''
super().setUp()
# We have a SentencePiece fixture for testing
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
tokenizer.save_pretrained(self.tmpdirname )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =NllbTokenizer(__snake_case , keep_accents=__snake_case )
__a =tokenizer.tokenize('This is a test' )
self.assertListEqual(__snake_case , ['▁This', '▁is', '▁a', '▁t', 'est'] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(__snake_case ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
__a =tokenizer.tokenize('I was born in 92000, and this is falsé.' )
self.assertListEqual(
__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 =tokenizer.convert_tokens_to_ids(__snake_case )
self.assertListEqual(
__snake_case , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
__a =tokenizer.convert_ids_to_tokens(__snake_case )
self.assertListEqual(
__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 __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =(self.rust_tokenizer_class, 'hf-internal-testing/tiny-random-nllb', {})
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =self.rust_tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =self.tokenizer_class.from_pretrained(__snake_case , **__snake_case )
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tuple(f for f in tokenizer_r_files if 'tokenizer.json' not in f )
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=True
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__snake_case )
# Checks it save with the same files
self.assertSequenceEqual(__snake_case , __snake_case )
# Checks everything loads correctly in the same way
__a =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
# Save tokenizer rust, legacy_format=False
__a =tempfile.mkdtemp()
__a =tokenizer_r.save_pretrained(__snake_case , legacy_format=__snake_case )
__a =tokenizer_p.save_pretrained(__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 =tokenizer_r.from_pretrained(__snake_case )
__a =tokenizer_p.from_pretrained(__snake_case )
# Check special tokens are set accordingly on Rust and Python
for key in tokenizer_pp.special_tokens_map:
self.assertTrue(hasattr(__snake_case , __snake_case ) )
shutil.rmtree(__snake_case )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
if not self.test_seqaseq:
return
__a =self.get_tokenizers()
for tokenizer in tokenizers:
with self.subTest(f'{tokenizer.__class__.__name__}' ):
# Longer text that will definitely require truncation.
__a =[
' 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.',
]
__a =[
'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.',
]
try:
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , tgt_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='ron_Latn' , )
except NotImplementedError:
return
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 10 )
# max_target_length will default to max_length if not specified
__a =tokenizer.prepare_seqaseq_batch(
__snake_case , tgt_texts=__snake_case , max_length=3 , return_tensors='pt' )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.labels.shape[1] , 3 )
__a =tokenizer.prepare_seqaseq_batch(
src_texts=__snake_case , max_length=3 , max_target_length=10 , return_tensors='pt' )
self.assertEqual(batch_encoder_only.input_ids.shape[1] , 3 )
self.assertEqual(batch_encoder_only.attention_mask.shape[1] , 3 )
self.assertNotIn('decoder_input_ids' , __snake_case )
@unittest.skip('Unfortunately way too slow to build a BPE with SentencePiece.' )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'{tokenizer.__class__.__name__} ({pretrained_name})' ):
__a =[AddedToken('<special>' , lstrip=__snake_case )]
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_r.encode('Hey this is a <special> token' )
__a =tokenizer_r.encode('<special>' , add_special_tokens=__snake_case )[0]
self.assertTrue(special_token_id in r_output )
if self.test_slow_tokenizer:
__a =self.rust_tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case , )
__a =self.tokenizer_class.from_pretrained(
__snake_case , additional_special_tokens=__snake_case , **__snake_case )
__a =tokenizer_p.encode('Hey this is a <special> token' )
__a =tokenizer_cr.encode('Hey this is a <special> token' )
self.assertEqual(__snake_case , __snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertTrue(special_token_id in p_output )
self.assertTrue(special_token_id in cr_output )
@require_torch
@require_sentencepiece
@require_tokenizers
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = 'facebook/nllb-200-distilled-600M'
SCREAMING_SNAKE_CASE = [
' 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 = [
'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 = [
2_5_6_0_4_7,
1_6_2_9_7,
1_3_4_4_0_8,
8_1_6_5,
2_4_8_0_6_6,
1_4_7_3_4,
9_5_0,
1_1_3_5,
1_0_5_7_2_1,
3_5_7_3,
8_3,
2_7_3_5_2,
1_0_8,
4_9_4_8_6,
2,
]
@classmethod
def __magic_name__ ( cls ) -> Tuple:
'''simple docstring'''
__a =NllbTokenizer.from_pretrained(
cls.checkpoint_name , src_lang='eng_Latn' , tgt_lang='ron_Latn' )
__a =1
return cls
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Arab'] , 25_6001 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['ace_Latn'] , 25_6002 )
self.assertEqual(self.tokenizer.fairseq_tokens_to_ids['fra_Latn'] , 25_6057 )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.tokenizer.batch_encode_plus(self.src_text ).input_ids[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
self.assertIn(__snake_case , self.tokenizer.all_special_ids )
# fmt: off
__a =[RO_CODE, 4254, 9_8068, 11_2923, 3_9072, 3909, 713, 10_2767, 26, 1_7314, 3_5642, 1_4683, 3_3118, 2022, 6_6987, 2, 25_6047]
# fmt: on
__a =self.tokenizer.decode(__snake_case , skip_special_tokens=__snake_case )
__a =self.tokenizer.decode(generated_ids[1:] , skip_special_tokens=__snake_case )
self.assertEqual(__snake_case , __snake_case )
self.assertNotIn(self.tokenizer.eos_token , __snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =['this is gunna be a long sentence ' * 20]
assert isinstance(src_text[0] , __snake_case )
__a =10
__a =self.tokenizer(__snake_case , max_length=__snake_case , truncation=__snake_case ).input_ids[0]
self.assertEqual(ids[-1] , 2 )
self.assertEqual(ids[0] , __snake_case )
self.assertEqual(len(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
self.assertListEqual(self.tokenizer.convert_tokens_to_ids(['<mask>', 'ar_AR'] ) , [25_6203, 3] )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =tempfile.mkdtemp()
__a =self.tokenizer.fairseq_tokens_to_ids
self.tokenizer.save_pretrained(__snake_case )
__a =NllbTokenizer.from_pretrained(__snake_case )
self.assertDictEqual(new_tok.fairseq_tokens_to_ids , __snake_case )
@require_torch
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.tokenizer(
self.src_text , text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=len(self.expected_src_tokens ) , return_tensors='pt' , )
__a =shift_tokens_right(
batch['labels'] , self.tokenizer.pad_token_id , self.tokenizer.lang_code_to_id['ron_Latn'] )
self.assertIsInstance(__snake_case , __snake_case )
self.assertEqual((2, 15) , batch.input_ids.shape )
self.assertEqual((2, 15) , batch.attention_mask.shape )
__a =batch.input_ids.tolist()[0]
self.assertListEqual(self.expected_src_tokens , __snake_case )
self.assertEqual(__snake_case , batch.decoder_input_ids[0, 0] ) # EOS
# Test that special tokens are reset
self.assertEqual(self.tokenizer.prefix_tokens , [EN_CODE] )
self.assertEqual(self.tokenizer.suffix_tokens , [self.tokenizer.eos_token_id] )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =self.tokenizer(self.src_text , padding=__snake_case , truncation=__snake_case , max_length=3 , return_tensors='pt' )
__a =self.tokenizer(
text_target=self.tgt_text , padding=__snake_case , truncation=__snake_case , max_length=10 , return_tensors='pt' )
__a =targets['input_ids']
__a =shift_tokens_right(
__snake_case , self.tokenizer.pad_token_id , decoder_start_token_id=self.tokenizer.lang_code_to_id[self.tokenizer.tgt_lang] , )
self.assertEqual(batch.input_ids.shape[1] , 3 )
self.assertEqual(batch.decoder_input_ids.shape[1] , 10 )
@require_torch
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.tokenizer._build_translation_inputs(
'A test' , return_tensors='pt' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
nested_simplify(__snake_case ) , {
# A, test, EOS, en_XX
'input_ids': [[25_6047, 70, 7356, 2]],
'attention_mask': [[1, 1, 1, 1]],
# ar_AR
'forced_bos_token_id': 25_6057,
} , )
@require_torch
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =True
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2, 25_6047] )
__a =False
__a =self.tokenizer(
'UN Chief says there is no military solution in Syria' , src_lang='eng_Latn' , tgt_lang='fra_Latn' )
self.assertEqual(
inputs.input_ids , [25_6047, 1_6297, 13_4408, 2_5653, 6370, 248, 254, 10_3929, 9_4995, 108, 4_9486, 2] )
| 308
| 0
|
import warnings
from ...processing_utils import ProcessorMixin
from ...tokenization_utils_base import BatchEncoding
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = ['image_processor', 'tokenizer']
SCREAMING_SNAKE_CASE = 'CLIPImageProcessor'
SCREAMING_SNAKE_CASE = ('CLIPTokenizer', 'CLIPTokenizerFast')
def __init__( self , __snake_case=None , __snake_case=None , **__snake_case ) -> Dict:
'''simple docstring'''
__a =None
if "feature_extractor" in kwargs:
warnings.warn(
'The `feature_extractor` argument is deprecated and will be removed in v5, use `image_processor`'
' instead.' , _lowerCAmelCase , )
__a =kwargs.pop('feature_extractor' )
__a =image_processor if image_processor is not None else feature_extractor
if image_processor is None:
raise ValueError('You need to specify an `image_processor`.' )
if tokenizer is None:
raise ValueError('You need to specify a `tokenizer`.' )
super().__init__(_lowerCAmelCase , _lowerCAmelCase )
def __call__( self , __snake_case=None , __snake_case=None , __snake_case=None , **__snake_case ) -> str:
'''simple docstring'''
if text is None and images is None:
raise ValueError('You have to specify either text or images. Both cannot be none.' )
if text is not None:
__a =self.tokenizer(_lowerCAmelCase , return_tensors=_lowerCAmelCase , **_lowerCAmelCase )
if images is not None:
__a =self.image_processor(_lowerCAmelCase , return_tensors=_lowerCAmelCase , **_lowerCAmelCase )
if text is not None and images is not None:
__a =image_features.pixel_values
return encoding
elif text is not None:
return encoding
else:
return BatchEncoding(data=dict(**_lowerCAmelCase ) , tensor_type=_lowerCAmelCase )
def __magic_name__ ( self , *__snake_case , **__snake_case ) -> int:
'''simple docstring'''
return self.tokenizer.batch_decode(*_lowerCAmelCase , **_lowerCAmelCase )
def __magic_name__ ( self , *__snake_case , **__snake_case ) -> Union[str, Any]:
'''simple docstring'''
return self.tokenizer.decode(*_lowerCAmelCase , **_lowerCAmelCase )
@property
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.tokenizer.model_input_names
__a =self.image_processor.model_input_names
return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names ) )
@property
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
warnings.warn(
'`feature_extractor_class` is deprecated and will be removed in v5. Use `image_processor_class` instead.' , _lowerCAmelCase , )
return self.image_processor_class
@property
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
warnings.warn(
'`feature_extractor` is deprecated and will be removed in v5. Use `image_processor` instead.' , _lowerCAmelCase , )
return self.image_processor
| 354
|
def UpperCamelCase_( _snake_case : str , _snake_case : int ):
"""simple docstring"""
return [sentence[i : i + ngram_size] for i in range(len(_snake_case ) - ngram_size + 1 )]
if __name__ == "__main__":
from doctest import testmod
testmod()
| 308
| 0
|
import inspect
import unittest
import numpy as np
from transformers import BeitConfig
from transformers.testing_utils import require_flax, require_vision, slow
from transformers.utils import cached_property, is_flax_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_flax_common import FlaxModelTesterMixin, floats_tensor, ids_tensor
if is_flax_available():
import jax
from transformers import FlaxBeitForImageClassification, FlaxBeitForMaskedImageModeling, FlaxBeitModel
if is_vision_available():
from PIL import Image
from transformers import BeitImageProcessor
class __magic_name__ ( unittest.TestCase ):
def __init__( self , __snake_case , __snake_case=100 , __snake_case=13 , __snake_case=30 , __snake_case=2 , __snake_case=3 , __snake_case=True , __snake_case=True , __snake_case=32 , __snake_case=5 , __snake_case=4 , __snake_case=37 , __snake_case="gelu" , __snake_case=0.1 , __snake_case=0.1 , __snake_case=10 , __snake_case=0.02 , __snake_case=3 , ) -> int:
'''simple docstring'''
__a =parent
__a =vocab_size
__a =batch_size
__a =image_size
__a =patch_size
__a =num_channels
__a =is_training
__a =use_labels
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =type_sequence_label_size
__a =initializer_range
# in BeiT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token)
__a =(image_size // patch_size) ** 2
__a =num_patches + 1
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a =BeitConfig(
vocab_size=self.vocab_size , image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , 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 , is_decoder=_SCREAMING_SNAKE_CASE , initializer_range=self.initializer_range , )
return config, pixel_values, labels
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =FlaxBeitModel(config=_SCREAMING_SNAKE_CASE )
__a =model(_SCREAMING_SNAKE_CASE )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> int:
'''simple docstring'''
__a =FlaxBeitForMaskedImageModeling(config=_SCREAMING_SNAKE_CASE )
__a =model(_SCREAMING_SNAKE_CASE )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length - 1, self.vocab_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Union[str, Any]:
'''simple docstring'''
__a =self.type_sequence_label_size
__a =FlaxBeitForImageClassification(config=_SCREAMING_SNAKE_CASE )
__a =model(_SCREAMING_SNAKE_CASE )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
# test greyscale images
__a =1
__a =FlaxBeitForImageClassification(_SCREAMING_SNAKE_CASE )
__a =floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
__a =model(_SCREAMING_SNAKE_CASE )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) ,
) =config_and_inputs
__a ={'pixel_values': pixel_values}
return config, inputs_dict
@require_flax
class __magic_name__ ( lowerCAmelCase__ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(FlaxBeitModel, FlaxBeitForImageClassification, FlaxBeitForMaskedImageModeling) if is_flax_available() else ()
)
def __magic_name__ ( self ) -> None:
'''simple docstring'''
__a =FlaxBeitModelTester(self )
__a =ConfigTester(self , config_class=_SCREAMING_SNAKE_CASE , has_text_modality=_SCREAMING_SNAKE_CASE , hidden_size=37 )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
self.config_tester.run_common_tests()
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a , __a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
__a =model_class(_SCREAMING_SNAKE_CASE )
__a =inspect.signature(model.__call__ )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
__a =[*signature.parameters.keys()]
__a =['pixel_values']
self.assertListEqual(arg_names[:1] , _SCREAMING_SNAKE_CASE )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a , __a =self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
with self.subTest(model_class.__name__ ):
__a =self._prepare_for_class(_SCREAMING_SNAKE_CASE , _SCREAMING_SNAKE_CASE )
__a =model_class(_SCREAMING_SNAKE_CASE )
@jax.jit
def model_jitted(__snake_case , **__snake_case ):
return model(pixel_values=_SCREAMING_SNAKE_CASE , **_SCREAMING_SNAKE_CASE )
with self.subTest('JIT Enabled' ):
__a =model_jitted(**_SCREAMING_SNAKE_CASE ).to_tuple()
with self.subTest('JIT Disabled' ):
with jax.disable_jit():
__a =model_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 __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*_SCREAMING_SNAKE_CASE )
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*_SCREAMING_SNAKE_CASE )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*_SCREAMING_SNAKE_CASE )
@slow
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
for model_class_name in self.all_model_classes:
__a =model_class_name.from_pretrained('microsoft/beit-base-patch16-224' )
__a =model(np.ones((1, 3, 224, 224) ) )
self.assertIsNotNone(_SCREAMING_SNAKE_CASE )
def UpperCamelCase_( ):
"""simple docstring"""
__a =Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' )
return image
@require_vision
@require_flax
class __magic_name__ ( unittest.TestCase ):
@cached_property
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
return BeitImageProcessor.from_pretrained('microsoft/beit-base-patch16-224' ) if is_vision_available() else None
@slow
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =FlaxBeitForMaskedImageModeling.from_pretrained('microsoft/beit-base-patch16-224-pt22k' )
__a =self.default_image_processor
__a =prepare_img()
__a =image_processor(images=_SCREAMING_SNAKE_CASE , return_tensors='np' ).pixel_values
# prepare bool_masked_pos
__a =np.ones((1, 196) , dtype=_SCREAMING_SNAKE_CASE )
# forward pass
__a =model(pixel_values=_SCREAMING_SNAKE_CASE , bool_masked_pos=_SCREAMING_SNAKE_CASE )
__a =outputs.logits
# verify the logits
__a =(1, 196, 8192)
self.assertEqual(logits.shape , _SCREAMING_SNAKE_CASE )
__a =np.array(
[[-3.2437, 0.5072, -13.9174], [-3.2456, 0.4948, -13.9401], [-3.2033, 0.5121, -13.8550]] )
self.assertTrue(np.allclose(logits[bool_masked_pos][:3, :3] , _SCREAMING_SNAKE_CASE , atol=1e-2 ) )
@slow
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =FlaxBeitForImageClassification.from_pretrained('microsoft/beit-base-patch16-224' )
__a =self.default_image_processor
__a =prepare_img()
__a =image_processor(images=_SCREAMING_SNAKE_CASE , return_tensors='np' )
# forward pass
__a =model(**_SCREAMING_SNAKE_CASE )
__a =outputs.logits
# verify the logits
__a =(1, 1000)
self.assertEqual(logits.shape , _SCREAMING_SNAKE_CASE )
__a =np.array([-1.2385, -1.0987, -1.0108] )
self.assertTrue(np.allclose(logits[0, :3] , _SCREAMING_SNAKE_CASE , atol=1e-4 ) )
__a =281
self.assertEqual(logits.argmax(-1 ).item() , _SCREAMING_SNAKE_CASE )
@slow
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =FlaxBeitForImageClassification.from_pretrained('microsoft/beit-large-patch16-224-pt22k-ft22k' )
__a =self.default_image_processor
__a =prepare_img()
__a =image_processor(images=_SCREAMING_SNAKE_CASE , return_tensors='np' )
# forward pass
__a =model(**_SCREAMING_SNAKE_CASE )
__a =outputs.logits
# verify the logits
__a =(1, 2_1841)
self.assertEqual(logits.shape , _SCREAMING_SNAKE_CASE )
__a =np.array([1.6881, -0.2787, 0.5901] )
self.assertTrue(np.allclose(logits[0, :3] , _SCREAMING_SNAKE_CASE , atol=1e-4 ) )
__a =2396
self.assertEqual(logits.argmax(-1 ).item() , _SCREAMING_SNAKE_CASE )
| 355
|
import argparse
import pytorch_lightning as pl
import torch
from torch import nn
from transformers import LongformerForQuestionAnswering, LongformerModel
class __magic_name__ ( pl.LightningModule ):
def __init__( self , __snake_case ) -> List[Any]:
'''simple docstring'''
super().__init__()
__a =model
__a =2
__a =nn.Linear(self.model.config.hidden_size , self.num_labels )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : str , _snake_case : str , _snake_case : str ):
"""simple docstring"""
__a =LongformerModel.from_pretrained(_snake_case )
__a =LightningModel(_snake_case )
__a =torch.load(_snake_case , map_location=torch.device('cpu' ) )
lightning_model.load_state_dict(ckpt['state_dict'] )
# init longformer question answering model
__a =LongformerForQuestionAnswering.from_pretrained(_snake_case )
# transfer weights
longformer_for_qa.longformer.load_state_dict(lightning_model.model.state_dict() )
longformer_for_qa.qa_outputs.load_state_dict(lightning_model.qa_outputs.state_dict() )
longformer_for_qa.eval()
# save model
longformer_for_qa.save_pretrained(_snake_case )
print(F'Conversion successful. Model saved under {pytorch_dump_folder_path}' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--longformer_model",
default=None,
type=str,
required=True,
help="model identifier of longformer. Should be either `longformer-base-4096` or `longformer-large-4096`.",
)
parser.add_argument(
"--longformer_question_answering_ckpt_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch Lightning Checkpoint.",
)
parser.add_argument(
"--pytorch_dump_folder_path", default=None, type=str, required=True, help="Path to the output PyTorch model."
)
_lowerCAmelCase : Optional[int] = parser.parse_args()
convert_longformer_qa_checkpoint_to_pytorch(
args.longformer_model, args.longformer_question_answering_ckpt_path, args.pytorch_dump_folder_path
)
| 308
| 0
|
from typing import TYPE_CHECKING
from ...utils import (
OptionalDependencyNotAvailable,
_LazyModule,
is_sentencepiece_available,
is_torch_available,
)
_lowerCAmelCase : Dict = {
"configuration_speecht5": [
"SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP",
"SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP",
"SpeechT5Config",
"SpeechT5HifiGanConfig",
],
"feature_extraction_speecht5": ["SpeechT5FeatureExtractor"],
"processing_speecht5": ["SpeechT5Processor"],
}
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Optional[int] = ["SpeechT5Tokenizer"]
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Tuple = [
"SPEECHT5_PRETRAINED_MODEL_ARCHIVE_LIST",
"SpeechT5ForSpeechToText",
"SpeechT5ForSpeechToSpeech",
"SpeechT5ForTextToSpeech",
"SpeechT5Model",
"SpeechT5PreTrainedModel",
"SpeechT5HifiGan",
]
if TYPE_CHECKING:
from .configuration_speechta import (
SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP,
SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP,
SpeechTaConfig,
SpeechTaHifiGanConfig,
)
from .feature_extraction_speechta import SpeechTaFeatureExtractor
from .processing_speechta import SpeechTaProcessor
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_speechta import SpeechTaTokenizer
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_speechta import (
SPEECHT5_PRETRAINED_MODEL_ARCHIVE_LIST,
SpeechTaForSpeechToSpeech,
SpeechTaForSpeechToText,
SpeechTaForTextToSpeech,
SpeechTaHifiGan,
SpeechTaModel,
SpeechTaPreTrainedModel,
)
else:
import sys
_lowerCAmelCase : List[Any] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 356
|
import logging
import os
import random
import sys
from dataclasses import dataclass, field
from typing import Optional
import datasets
import numpy as np
import pandas as pd
from datasets import load_dataset
import transformers
from transformers import (
AutoConfig,
BartForSequenceClassification,
DataCollatorWithPadding,
EvalPrediction,
HfArgumentParser,
TapexTokenizer,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
_lowerCAmelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The name of the dataset to use (via the datasets library).'} )
SCREAMING_SNAKE_CASE = field(
default='tab_fact' , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} , )
SCREAMING_SNAKE_CASE = field(
default=1_0_2_4 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Overwrite the cached preprocessed datasets or not.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Whether to pad all samples to `max_seq_length`. '
'If False, will pad the samples dynamically when batching to the maximum length in the batch.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of training examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of evaluation examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of prediction examples to this '
'value if set.'
)
} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = field(default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the test data.'} )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
if self.dataset_name is not None:
pass
elif self.train_file is None or self.validation_file is None:
raise ValueError('Need either a GLUE task, a training/validation file or a dataset name.' )
else:
__a =self.train_file.split('.' )[-1]
assert train_extension in ["csv", "json"], "`train_file` should be a csv or a json file."
__a =self.validation_file.split('.' )[-1]
assert (
validation_extension == train_extension
), "`validation_file` should have the same extension (csv or json) as `train_file`."
@dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to use one of the fast tokenizer (backed by the tokenizers library) or not.'} , )
SCREAMING_SNAKE_CASE = field(
default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , )
SCREAMING_SNAKE_CASE = field(
default=lowerCAmelCase_ , metadata={
'help': (
'Will use the token generated when running `huggingface-cli login` (necessary to use this script '
'with private models).'
)
} , )
def UpperCamelCase_( ):
"""simple docstring"""
__a =HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments) )
if len(sys.argv ) == 2 and sys.argv[1].endswith('.json' ):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
__a , __a , __a =parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) )
else:
__a , __a , __a =parser.parse_args_into_dataclasses()
# Setup logging
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , handlers=[logging.StreamHandler(sys.stdout )] , )
__a =training_args.get_process_log_level()
logger.setLevel(_snake_case )
datasets.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.set_verbosity(_snake_case )
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Log on each process the small summary:
logger.warning(
F'Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}'
+ F'distributed training: {bool(training_args.local_rank != -1 )}, 16-bits training: {training_args.fpaa}' )
logger.info(F'Training/evaluation parameters {training_args}' )
# Detecting last checkpoint.
__a =None
if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir:
__a =get_last_checkpoint(training_args.output_dir )
if last_checkpoint is None and len(os.listdir(training_args.output_dir ) ) > 0:
raise ValueError(
F'Output directory ({training_args.output_dir}) already exists and is not empty. '
'Use --overwrite_output_dir to overcome.' )
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
F'Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change '
'the `--output_dir` or add `--overwrite_output_dir` to train from scratch.' )
# Set seed before initializing model.
set_seed(training_args.seed )
# Get the datasets: you can either provide your own CSV/JSON training and evaluation files (see below)
# or specify a GLUE benchmark task (the dataset will be downloaded automatically from the datasets Hub).
#
# For JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# If the CSVs/JSONs contain only one non-label column, the script does single sentence classification on this
# single column. You can easily tweak this behavior (see below)
#
# In distributed training, the load_dataset function guarantee that only one local process can concurrently
# download the dataset.
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
__a =load_dataset(
data_args.dataset_name , data_args.dataset_config_name , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from your local files.
# CSV/JSON training and evaluation files are needed.
__a ={'train': data_args.train_file, 'validation': data_args.validation_file}
# Get the test dataset: you can provide your own CSV/JSON test file (see below)
# when you use `do_predict` without specifying a GLUE benchmark task.
if training_args.do_predict:
if data_args.test_file is not None:
__a =data_args.train_file.split('.' )[-1]
__a =data_args.test_file.split('.' )[-1]
assert (
test_extension == train_extension
), "`test_file` should have the same extension (csv or json) as `train_file`."
__a =data_args.test_file
else:
raise ValueError('Need either a GLUE task or a test file for `do_predict`.' )
for key in data_files.keys():
logger.info(F'load a local file for {key}: {data_files[key]}' )
if data_args.train_file.endswith('.csv' ):
# Loading a dataset from local csv files
__a =load_dataset('csv' , data_files=_snake_case , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from local json files
__a =load_dataset('json' , data_files=_snake_case , cache_dir=model_args.cache_dir )
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.html.
# Labels
__a =raw_datasets['train'].features['label'].names
__a =len(_snake_case )
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
__a =AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# load tapex tokenizer
__a =TapexTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , use_fast=model_args.use_fast_tokenizer , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , add_prefix_space=_snake_case , )
__a =BartForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_tf=bool('.ckpt' in model_args.model_name_or_path ) , config=_snake_case , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# Padding strategy
if data_args.pad_to_max_length:
__a ='max_length'
else:
# We will pad later, dynamically at batch creation, to the max sequence length in each batch
__a =False
# Some models have set the order of the labels to use, so let's make sure we do use it.
__a ={'Refused': 0, 'Entailed': 1}
__a ={0: 'Refused', 1: 'Entailed'}
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warning(
F'The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the'
F'model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}.' )
__a =min(data_args.max_seq_length , tokenizer.model_max_length )
def preprocess_tabfact_function(_snake_case : Tuple ):
# Tokenize the texts
def _convert_table_text_to_pandas(_snake_case : Optional[Any] ):
__a =[_table_row.split('#' ) for _table_row in _table_text.strip('\n' ).split('\n' )]
__a =pd.DataFrame.from_records(_table_content[1:] , columns=_table_content[0] )
return _table_pd
__a =examples['statement']
__a =list(map(_convert_table_text_to_pandas , examples['table_text'] ) )
__a =tokenizer(_snake_case , _snake_case , padding=_snake_case , max_length=_snake_case , truncation=_snake_case )
__a =examples['label']
return result
with training_args.main_process_first(desc='dataset map pre-processing' ):
__a =raw_datasets.map(
_snake_case , batched=_snake_case , load_from_cache_file=not data_args.overwrite_cache , desc='Running tokenizer on dataset' , )
if training_args.do_train:
if "train" not in raw_datasets:
raise ValueError('--do_train requires a train dataset' )
__a =raw_datasets['train']
if data_args.max_train_samples is not None:
__a =train_dataset.select(range(data_args.max_train_samples ) )
if training_args.do_eval:
if "validation" not in raw_datasets and "validation_matched" not in raw_datasets:
raise ValueError('--do_eval requires a validation dataset' )
__a =raw_datasets['validation']
if data_args.max_eval_samples is not None:
__a =eval_dataset.select(range(data_args.max_eval_samples ) )
if training_args.do_predict or data_args.test_file is not None:
if "test" not in raw_datasets and "test_matched" not in raw_datasets:
raise ValueError('--do_predict requires a test dataset' )
__a =raw_datasets['test']
if data_args.max_predict_samples is not None:
__a =predict_dataset.select(range(data_args.max_predict_samples ) )
# Log a few random samples from the training set:
if training_args.do_train:
for index in random.sample(range(len(_snake_case ) ) , 3 ):
logger.info(F'Sample {index} of the training set: {train_dataset[index]}.' )
# You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
# predictions and label_ids field) and has to return a dictionary string to float.
def compute_metrics(_snake_case : EvalPrediction ):
__a =p.predictions[0] if isinstance(p.predictions , _snake_case ) else p.predictions
__a =np.argmax(_snake_case , axis=1 )
return {"accuracy": (preds == p.label_ids).astype(np.floataa ).mean().item()}
# Data collator will default to DataCollatorWithPadding, so we change it if we already did the padding.
if data_args.pad_to_max_length:
__a =default_data_collator
elif training_args.fpaa:
__a =DataCollatorWithPadding(_snake_case , pad_to_multiple_of=8 )
else:
__a =None
# Initialize our Trainer
__a =Trainer(
model=_snake_case , args=_snake_case , train_dataset=train_dataset if training_args.do_train else None , eval_dataset=eval_dataset if training_args.do_eval else None , compute_metrics=_snake_case , tokenizer=_snake_case , data_collator=_snake_case , )
# Training
if training_args.do_train:
__a =None
if training_args.resume_from_checkpoint is not None:
__a =training_args.resume_from_checkpoint
elif last_checkpoint is not None:
__a =last_checkpoint
__a =trainer.train(resume_from_checkpoint=_snake_case )
__a =train_result.metrics
__a =(
data_args.max_train_samples if data_args.max_train_samples is not None else len(_snake_case )
)
__a =min(_snake_case , len(_snake_case ) )
trainer.save_model() # Saves the tokenizer too for easy upload
trainer.log_metrics('train' , _snake_case )
trainer.save_metrics('train' , _snake_case )
trainer.save_state()
# Evaluation
if training_args.do_eval:
logger.info('*** Evaluate ***' )
__a =trainer.evaluate(eval_dataset=_snake_case )
__a =data_args.max_eval_samples if data_args.max_eval_samples is not None else len(_snake_case )
__a =min(_snake_case , len(_snake_case ) )
trainer.log_metrics('eval' , _snake_case )
trainer.save_metrics('eval' , _snake_case )
if training_args.do_predict:
logger.info('*** Predict ***' )
# Removing the `label` columns because it contains -1 and Trainer won't like that.
__a =predict_dataset.remove_columns('label' )
__a =trainer.predict(_snake_case , metric_key_prefix='predict' ).predictions
__a =np.argmax(_snake_case , axis=1 )
__a =os.path.join(training_args.output_dir , 'predict_results_tabfact.txt' )
if trainer.is_world_process_zero():
with open(_snake_case , 'w' ) as writer:
logger.info('***** Predict Results *****' )
writer.write('index\tprediction\n' )
for index, item in enumerate(_snake_case ):
__a =label_list[item]
writer.write(F'{index}\t{item}\n' )
__a ={'finetuned_from': model_args.model_name_or_path, 'tasks': 'text-classification'}
if training_args.push_to_hub:
trainer.push_to_hub(**_snake_case )
else:
trainer.create_model_card(**_snake_case )
def UpperCamelCase_( _snake_case : Union[str, Any] ):
"""simple docstring"""
main()
if __name__ == "__main__":
main()
| 308
| 0
|
# Copyright 2022 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.
import argparse
import os
import subprocess
from packaging.version import Version, parse
from accelerate.commands.config.config_args import default_config_file, load_config_from_file
_lowerCAmelCase : int = "Run commands across TPU VMs for initial setup before running `accelerate launch`."
def UpperCamelCase_( _snake_case : List[str]=None ):
"""simple docstring"""
if subparsers is not None:
__a =subparsers.add_parser('tpu-config' , description=_description )
else:
__a =argparse.ArgumentParser('Accelerate tpu-config command' , description=_description )
# Core arguments
__a =parser.add_argument_group(
'Config Arguments' , 'Arguments that can be configured through `accelerate config`.' )
config_args.add_argument(
'--config_file' , type=_snake_case , default=_snake_case , help='Path to the config file to use for accelerate.' , )
config_args.add_argument(
'--tpu_name' , default=_snake_case , help='The name of the TPU to use. If not specified, will use the TPU specified in the config file.' , )
config_args.add_argument(
'--tpu_zone' , default=_snake_case , help='The zone of the TPU to use. If not specified, will use the zone specified in the config file.' , )
__a =parser.add_argument_group('TPU Arguments' , 'Arguments for options ran inside the TPU.' )
pod_args.add_argument(
'--use_alpha' , action='store_true' , help='Whether to use `gcloud alpha` when running the TPU training script instead of `gcloud`.' , )
pod_args.add_argument(
'--command_file' , default=_snake_case , help='The path to the file containing the commands to run on the pod on startup.' , )
pod_args.add_argument(
'--command' , action='append' , nargs='+' , help='A command to run on the pod. Can be passed multiple times.' , )
pod_args.add_argument(
'--install_accelerate' , action='store_true' , help='Whether to install accelerate on the pod. Defaults to False.' , )
pod_args.add_argument(
'--accelerate_version' , default='latest' , help='The version of accelerate to install on the pod. If not specified, will use the latest pypi version. Specify \'dev\' to install from GitHub.' , )
pod_args.add_argument(
'--debug' , action='store_true' , help='If set, will print the command that would be run instead of running it.' )
if subparsers is not None:
parser.set_defaults(func=_snake_case )
return parser
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =None
# Get the default from the config file if it exists.
if args.config_file is not None or os.path.isfile(_snake_case ):
__a =load_config_from_file(args.config_file )
if not args.command_file and defaults.command_file is not None and not args.command:
__a =defaults.command_file
if not args.command and defaults.commands is not None:
__a =defaults.commands
if not args.tpu_name:
__a =defaults.tpu_name
if not args.tpu_zone:
__a =defaults.tpu_zone
if args.accelerate_version == "dev":
__a ='''git+https://github.com/huggingface/accelerate.git'''
elif args.accelerate_version == "latest":
__a ='''accelerate -U'''
elif isinstance(parse(args.accelerate_version ) , _snake_case ):
__a =F'accelerate=={args.accelerate_version}'
if not args.command_file and not args.command:
raise ValueError('You must specify either a command file or a command to run on the pod.' )
if args.command_file:
with open(args.command_file , 'r' ) as f:
__a =[f.read().splitlines()]
# To turn list of lists into list of strings
if isinstance(args.command[0] , _snake_case ):
__a =[line for cmd in args.command for line in cmd]
# Default to the shared folder and install accelerate
__a =['''cd /usr/share''']
if args.install_accelerate:
new_cmd += [F'pip install {args.accelerate_version}']
new_cmd += args.command
__a ='''; '''.join(_snake_case )
# Then send it to gcloud
# Eventually try to use google-api-core to do this instead of subprocess
__a =['''gcloud''']
if args.use_alpha:
cmd += ["alpha"]
cmd += [
"compute",
"tpus",
"tpu-vm",
"ssh",
args.tpu_name,
"--zone",
args.tpu_zone,
"--command",
args.command,
"--worker",
"all",
]
if args.debug:
print(F'Running {" ".join(_snake_case )}' )
return
subprocess.run(_snake_case )
print('Successfully setup pod.' )
def UpperCamelCase_( ):
"""simple docstring"""
__a =tpu_command_parser()
__a =parser.parse_args()
tpu_command_launcher(_snake_case )
| 357
|
from __future__ import annotations
import time
import numpy as np
_lowerCAmelCase : List[str] = [8, 5, 9, 7]
_lowerCAmelCase : List[str] = [
[2, 0, 1, 1],
[0, 1, 2, 1],
[4, 0, 0, 3],
[0, 2, 1, 0],
[1, 0, 3, 0],
]
_lowerCAmelCase : List[Any] = [
[3, 2, 1, 4],
[0, 2, 5, 2],
[5, 1, 0, 5],
[1, 5, 3, 0],
[3, 0, 3, 3],
]
class __magic_name__ :
def __init__( self , __snake_case , __snake_case , __snake_case , ) -> None:
'''simple docstring'''
__a =claim_vector
__a =allocated_resources_table
__a =maximum_claim_table
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return [
sum(p_item[i] for p_item in self.__allocated_resources_table )
for i in range(len(self.__allocated_resources_table[0] ) )
]
def __magic_name__ ( self ) -> list[int]:
'''simple docstring'''
return np.array(self.__claim_vector ) - np.array(
self.__processes_resource_summation() )
def __magic_name__ ( self ) -> list[list[int]]:
'''simple docstring'''
return [
list(np.array(self.__maximum_claim_table[i] ) - np.array(__snake_case ) )
for i, allocated_resource in enumerate(self.__allocated_resources_table )
]
def __magic_name__ ( self ) -> dict[int, list[int]]:
'''simple docstring'''
return {self.__need().index(__snake_case ): i for i in self.__need()}
def __magic_name__ ( self , **__snake_case ) -> None:
'''simple docstring'''
__a =self.__need()
__a =self.__allocated_resources_table
__a =self.__available_resources()
__a =self.__need_index_manager()
for kw, val in kwargs.items():
if kw and val is True:
self.__pretty_data()
print('_' * 50 + '\n' )
while need_list:
__a =False
for each_need in need_list:
__a =True
for index, need in enumerate(__snake_case ):
if need > available_resources[index]:
__a =False
break
if execution:
__a =True
# get the original index of the process from ind_ctrl db
for original_need_index, need_clone in need_index_manager.items():
if each_need == need_clone:
__a =original_need_index
print(f'Process {process_number + 1} is executing.' )
# remove the process run from stack
need_list.remove(__snake_case )
# update available/freed resources stack
__a =np.array(__snake_case ) + np.array(
alloc_resources_table[process_number] )
print(
'Updated available resource stack for processes: '
+ ' '.join([str(__snake_case ) for x in available_resources] ) )
break
if safe:
print('The process is in a safe state.\n' )
else:
print('System in unsafe state. Aborting...\n' )
break
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
print(' ' * 9 + 'Allocated Resource Table' )
for item in self.__allocated_resources_table:
print(
f'P{self.__allocated_resources_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(' ' * 9 + 'System Resource Table' )
for item in self.__maximum_claim_table:
print(
f'P{self.__maximum_claim_table.index(__snake_case ) + 1}'
+ ' '.join(f'{it:>8}' for it in item )
+ '\n' )
print(
'Current Usage by Active Processes: '
+ ' '.join(str(__snake_case ) for x in self.__claim_vector ) )
print(
'Initial Available Resources: '
+ ' '.join(str(__snake_case ) for x in self.__available_resources() ) )
time.sleep(1 )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 0
|
"""simple docstring"""
from dataclasses import dataclass
from typing import List, Optional, Union
import numpy as np
import PIL
import torch
from transformers import CLIPImageProcessor, CLIPVisionModel
from ...models import PriorTransformer
from ...pipelines import DiffusionPipeline
from ...schedulers import HeunDiscreteScheduler
from ...utils import (
BaseOutput,
is_accelerate_available,
logging,
randn_tensor,
replace_example_docstring,
)
from .renderer import ShapERenderer
_lowerCAmelCase : Optional[Any] = logging.get_logger(__name__) # pylint: disable=invalid-name
_lowerCAmelCase : str = "\n Examples:\n ```py\n >>> from PIL import Image\n >>> import torch\n >>> from diffusers import DiffusionPipeline\n >>> from diffusers.utils import export_to_gif, load_image\n\n >>> device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n\n >>> repo = \"openai/shap-e-img2img\"\n >>> pipe = DiffusionPipeline.from_pretrained(repo, torch_dtype=torch.float16)\n >>> pipe = pipe.to(device)\n\n >>> guidance_scale = 3.0\n >>> image_url = \"https://hf.co/datasets/diffusers/docs-images/resolve/main/shap-e/corgi.png\"\n >>> image = load_image(image_url).convert(\"RGB\")\n\n >>> images = pipe(\n ... image,\n ... guidance_scale=guidance_scale,\n ... num_inference_steps=64,\n ... frame_size=256,\n ... ).images\n\n >>> gif_path = export_to_gif(images[0], \"corgi_3d.gif\")\n ```\n"
@dataclass
class __magic_name__ ( lowerCamelCase__ ):
SCREAMING_SNAKE_CASE = 42
class __magic_name__ ( lowerCamelCase__ ):
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> int:
'''simple docstring'''
super().__init__()
self.register_modules(
prior=lowercase__ , image_encoder=lowercase__ , image_processor=lowercase__ , scheduler=lowercase__ , renderer=lowercase__ , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> Dict:
'''simple docstring'''
if latents is None:
__a =randn_tensor(lowercase__ , generator=lowercase__ , device=lowercase__ , dtype=lowercase__ )
else:
if latents.shape != shape:
raise ValueError(f'Unexpected latents shape, got {latents.shape}, expected {shape}' )
__a =latents.to(lowercase__ )
__a =latents * scheduler.init_noise_sigma
return latents
def __magic_name__ ( self , __snake_case=0 ) -> Dict:
'''simple docstring'''
if is_accelerate_available():
from accelerate import cpu_offload
else:
raise ImportError('Please install accelerate via `pip install accelerate`' )
__a =torch.device(f'cuda:{gpu_id}' )
__a =[self.image_encoder, self.prior]
for cpu_offloaded_model in models:
if cpu_offloaded_model is not None:
cpu_offload(lowercase__ , lowercase__ )
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.device != torch.device('meta' ) or not hasattr(self.image_encoder , '_hf_hook' ):
return self.device
for module in self.image_encoder.modules():
if (
hasattr(lowercase__ , '_hf_hook' )
and hasattr(module._hf_hook , 'execution_device' )
and module._hf_hook.execution_device is not None
):
return torch.device(module._hf_hook.execution_device )
return self.device
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Union[str, Any]:
'''simple docstring'''
if isinstance(lowercase__ , lowercase__ ) and isinstance(image[0] , torch.Tensor ):
__a =torch.cat(lowercase__ , axis=0 ) if image[0].ndim == 4 else torch.stack(lowercase__ , axis=0 )
if not isinstance(lowercase__ , torch.Tensor ):
__a =self.image_processor(lowercase__ , return_tensors='pt' ).pixel_values[0].unsqueeze(0 )
__a =image.to(dtype=self.image_encoder.dtype , device=lowercase__ )
__a =self.image_encoder(lowercase__ )['last_hidden_state']
__a =image_embeds[:, 1:, :].contiguous() # batch_size, dim, 256
__a =image_embeds.repeat_interleave(lowercase__ , dim=0 )
if do_classifier_free_guidance:
__a =torch.zeros_like(lowercase__ )
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch
# to avoid doing two forward passes
__a =torch.cat([negative_image_embeds, image_embeds] )
return image_embeds
@torch.no_grad()
@replace_example_docstring(lowercase__ )
def __call__( self , __snake_case , __snake_case = 1 , __snake_case = 25 , __snake_case = None , __snake_case = None , __snake_case = 4.0 , __snake_case = 64 , __snake_case = "pil" , __snake_case = True , ) -> Any:
'''simple docstring'''
if isinstance(lowercase__ , PIL.Image.Image ):
__a =1
elif isinstance(lowercase__ , torch.Tensor ):
__a =image.shape[0]
elif isinstance(lowercase__ , lowercase__ ) and isinstance(image[0] , (torch.Tensor, PIL.Image.Image) ):
__a =len(lowercase__ )
else:
raise ValueError(
f'`image` has to be of type `PIL.Image.Image`, `torch.Tensor`, `List[PIL.Image.Image]` or `List[torch.Tensor]` but is {type(lowercase__ )}' )
__a =self._execution_device
__a =batch_size * num_images_per_prompt
__a =guidance_scale > 1.0
__a =self._encode_image(lowercase__ , lowercase__ , lowercase__ , lowercase__ )
# prior
self.scheduler.set_timesteps(lowercase__ , device=lowercase__ )
__a =self.scheduler.timesteps
__a =self.prior.config.num_embeddings
__a =self.prior.config.embedding_dim
__a =self.prepare_latents(
(batch_size, num_embeddings * embedding_dim) , image_embeds.dtype , lowercase__ , lowercase__ , lowercase__ , self.scheduler , )
# YiYi notes: for testing only to match ldm, we can directly create a latents with desired shape: batch_size, num_embeddings, embedding_dim
__a =latents.reshape(latents.shape[0] , lowercase__ , lowercase__ )
for i, t in enumerate(self.progress_bar(lowercase__ ) ):
# expand the latents if we are doing classifier free guidance
__a =torch.cat([latents] * 2 ) if do_classifier_free_guidance else latents
__a =self.scheduler.scale_model_input(lowercase__ , lowercase__ )
__a =self.prior(
lowercase__ , timestep=lowercase__ , proj_embedding=lowercase__ , ).predicted_image_embedding
# remove the variance
__a , __a =noise_pred.split(
scaled_model_input.shape[2] , dim=2 ) # batch_size, num_embeddings, embedding_dim
if do_classifier_free_guidance is not None:
__a , __a =noise_pred.chunk(2 )
__a =noise_pred_uncond + guidance_scale * (noise_pred - noise_pred_uncond)
__a =self.scheduler.step(
lowercase__ , timestep=lowercase__ , sample=lowercase__ , ).prev_sample
if output_type == "latent":
return ShapEPipelineOutput(images=lowercase__ )
__a =[]
for i, latent in enumerate(lowercase__ ):
print()
__a =self.renderer.decode(
latent[None, :] , lowercase__ , size=lowercase__ , ray_batch_size=4096 , n_coarse_samples=64 , n_fine_samples=128 , )
images.append(lowercase__ )
__a =torch.stack(lowercase__ )
if output_type not in ["np", "pil"]:
raise ValueError(f'Only the output types `pil` and `np` are supported not output_type={output_type}' )
__a =images.cpu().numpy()
if output_type == "pil":
__a =[self.numpy_to_pil(lowercase__ ) for image in images]
# Offload last model to CPU
if hasattr(self , 'final_offload_hook' ) and self.final_offload_hook is not None:
self.final_offload_hook.offload()
if not return_dict:
return (images,)
return ShapEPipelineOutput(images=lowercase__ )
| 358
|
import string
# frequency taken from https://en.wikipedia.org/wiki/Letter_frequency
_lowerCAmelCase : Tuple = {
"E": 12.70,
"T": 9.06,
"A": 8.17,
"O": 7.51,
"I": 6.97,
"N": 6.75,
"S": 6.33,
"H": 6.09,
"R": 5.99,
"D": 4.25,
"L": 4.03,
"C": 2.78,
"U": 2.76,
"M": 2.41,
"W": 2.36,
"F": 2.23,
"G": 2.02,
"Y": 1.97,
"P": 1.93,
"B": 1.29,
"V": 0.98,
"K": 0.77,
"J": 0.15,
"X": 0.15,
"Q": 0.10,
"Z": 0.07,
}
_lowerCAmelCase : Optional[int] = "ETAOINSHRDLCUMWFGYPBVKJXQZ"
_lowerCAmelCase : Any = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a ={letter: 0 for letter in string.ascii_uppercase}
for letter in message.upper():
if letter in LETTERS:
letter_count[letter] += 1
return letter_count
def UpperCamelCase_( _snake_case : tuple ):
"""simple docstring"""
return x[0]
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_letter_count(_snake_case )
__a ={
freq: [] for letter, freq in letter_to_freq.items()
}
for letter in LETTERS:
freq_to_letter[letter_to_freq[letter]].append(_snake_case )
__a ={}
for freq in freq_to_letter:
freq_to_letter[freq].sort(key=ETAOIN.find , reverse=_snake_case )
__a =''.join(freq_to_letter[freq] )
__a =list(freq_to_letter_str.items() )
freq_pairs.sort(key=_snake_case , reverse=_snake_case )
__a =[freq_pair[1] for freq_pair in freq_pairs]
return "".join(_snake_case )
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =get_frequency_order(_snake_case )
__a =0
for common_letter in ETAOIN[:6]:
if common_letter in freq_order[:6]:
match_score += 1
for uncommon_letter in ETAOIN[-6:]:
if uncommon_letter in freq_order[-6:]:
match_score += 1
return match_score
if __name__ == "__main__":
import doctest
doctest.testmod()
| 308
| 0
|
import json
import os
import pickle
import shutil
import tempfile
from unittest import TestCase
from unittest.mock import patch
import numpy as np
from datasets import Dataset
from transformers import is_faiss_available
from transformers.models.bart.configuration_bart import BartConfig
from transformers.models.bart.tokenization_bart import BartTokenizer
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.dpr.tokenization_dpr import DPRContextEncoderTokenizer, DPRQuestionEncoderTokenizer
from transformers.models.rag.configuration_rag import RagConfig
from transformers.models.rag.retrieval_rag import CustomHFIndex, RagRetriever
from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
from transformers.testing_utils import require_faiss, require_sentencepiece, require_tokenizers, require_torch
if is_faiss_available():
import faiss
@require_faiss
class __magic_name__ ( SCREAMING_SNAKE_CASE__ ):
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =tempfile.mkdtemp()
__a =8
# DPR tok
__a =[
'[UNK]',
'[CLS]',
'[SEP]',
'[PAD]',
'[MASK]',
'want',
'##want',
'##ed',
'wa',
'un',
'runn',
'##ing',
',',
'low',
'lowest',
]
__a =os.path.join(self.tmpdirname , 'dpr_tokenizer' )
os.makedirs(__snake_case , exist_ok=__snake_case )
__a =os.path.join(__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 =[
'l',
'o',
'w',
'e',
'r',
's',
't',
'i',
'd',
'n',
'\u0120',
'\u0120l',
'\u0120n',
'\u0120lo',
'\u0120low',
'er',
'\u0120lowest',
'\u0120newer',
'\u0120wider',
'<unk>',
]
__a =dict(zip(__snake_case , range(len(__snake_case ) ) ) )
__a =['#version: 0.2', '\u0120 l', '\u0120l o', '\u0120lo w', 'e r', '']
__a ={'unk_token': '<unk>'}
__a =os.path.join(self.tmpdirname , 'bart_tokenizer' )
os.makedirs(__snake_case , exist_ok=__snake_case )
__a =os.path.join(__snake_case , BART_VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(__snake_case , BART_VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(__snake_case ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(__snake_case ) )
def __magic_name__ ( self ) -> DPRQuestionEncoderTokenizer:
'''simple docstring'''
return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'dpr_tokenizer' ) )
def __magic_name__ ( self ) -> DPRContextEncoderTokenizer:
'''simple docstring'''
return DPRContextEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'dpr_tokenizer' ) )
def __magic_name__ ( self ) -> BartTokenizer:
'''simple docstring'''
return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , 'bart_tokenizer' ) )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
shutil.rmtree(self.tmpdirname )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =Dataset.from_dict(
{
'id': ['0', '1'],
'text': ['foo', 'bar'],
'title': ['Foo', 'Bar'],
'embeddings': [np.ones(self.retrieval_vector_size ), 2 * np.ones(self.retrieval_vector_size )],
} )
dataset.add_faiss_index('embeddings' , string_factory='Flat' , metric_type=faiss.METRIC_INNER_PRODUCT )
return dataset
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =self.get_dummy_dataset()
__a =RagConfig(
retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , )
with patch('transformers.models.rag.retrieval_rag.load_dataset' ) as mock_load_dataset:
__a =dataset
__a =RagRetriever(
__snake_case , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , )
return retriever
def __magic_name__ ( self , __snake_case ) -> List[str]:
'''simple docstring'''
__a =self.get_dummy_dataset()
__a =RagConfig(
retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , index_name='custom' , )
if from_disk:
__a =os.path.join(self.tmpdirname , 'dataset' )
__a =os.path.join(self.tmpdirname , 'index.faiss' )
dataset.get_index('embeddings' ).save(os.path.join(self.tmpdirname , 'index.faiss' ) )
dataset.drop_index('embeddings' )
dataset.save_to_disk(os.path.join(self.tmpdirname , 'dataset' ) )
del dataset
__a =RagRetriever(
__snake_case , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , )
else:
__a =RagRetriever(
__snake_case , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , index=CustomHFIndex(config.retrieval_vector_size , __snake_case ) , )
return retriever
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =Dataset.from_dict(
{
'id': ['0', '1'],
'text': ['foo', 'bar'],
'title': ['Foo', 'Bar'],
'embeddings': [np.ones(self.retrieval_vector_size + 1 ), 2 * np.ones(self.retrieval_vector_size + 1 )],
} )
dataset.add_faiss_index('embeddings' , string_factory='Flat' , metric_type=faiss.METRIC_INNER_PRODUCT )
__a =os.path.join(self.tmpdirname , 'hf_bert_base.hnswSQ8_correct_phi_128.c_index' )
dataset.save_faiss_index('embeddings' , index_file_name + '.index.dpr' )
pickle.dump(dataset['id'] , open(index_file_name + '.index_meta.dpr' , 'wb' ) )
__a =os.path.join(self.tmpdirname , 'psgs_w100.tsv.pkl' )
__a ={sample['id']: [sample['text'], sample['title']] for sample in dataset}
pickle.dump(__snake_case , open(__snake_case , 'wb' ) )
__a =RagConfig(
retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , index_name='legacy' , index_path=self.tmpdirname , )
__a =RagRetriever(
__snake_case , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() )
return retriever
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =1
__a =self.get_dummy_canonical_hf_index_retriever()
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=__snake_case )
self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) )
self.assertEqual(len(__snake_case ) , 2 )
self.assertEqual(sorted(doc_dicts[0] ) , ['embeddings', 'id', 'text', 'title'] )
self.assertEqual(len(doc_dicts[0]['id'] ) , __snake_case )
self.assertEqual(doc_dicts[0]['id'][0] , '1' ) # max inner product is reached with second doc
self.assertEqual(doc_dicts[1]['id'][0] , '0' ) # max inner product is reached with first doc
self.assertListEqual(doc_ids.tolist() , [[1], [0]] )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =self.get_dummy_canonical_hf_index_retriever()
with tempfile.TemporaryDirectory() as tmp_dirname:
with patch('transformers.models.rag.retrieval_rag.load_dataset' ) as mock_load_dataset:
__a =self.get_dummy_dataset()
retriever.save_pretrained(__snake_case )
__a =RagRetriever.from_pretrained(__snake_case )
self.assertIsInstance(__snake_case , __snake_case )
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=1 )
self.assertTrue(out is not None )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =1
__a =self.get_dummy_custom_hf_index_retriever(from_disk=__snake_case )
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=__snake_case )
self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) )
self.assertEqual(len(__snake_case ) , 2 )
self.assertEqual(sorted(doc_dicts[0] ) , ['embeddings', 'id', 'text', 'title'] )
self.assertEqual(len(doc_dicts[0]['id'] ) , __snake_case )
self.assertEqual(doc_dicts[0]['id'][0] , '1' ) # max inner product is reached with second doc
self.assertEqual(doc_dicts[1]['id'][0] , '0' ) # max inner product is reached with first doc
self.assertListEqual(doc_ids.tolist() , [[1], [0]] )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =self.get_dummy_custom_hf_index_retriever(from_disk=__snake_case )
with tempfile.TemporaryDirectory() as tmp_dirname:
retriever.save_pretrained(__snake_case )
__a =RagRetriever.from_pretrained(__snake_case )
self.assertIsInstance(__snake_case , __snake_case )
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=1 )
self.assertTrue(out is not None )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =1
__a =self.get_dummy_custom_hf_index_retriever(from_disk=__snake_case )
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=__snake_case )
self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) )
self.assertEqual(len(__snake_case ) , 2 )
self.assertEqual(sorted(doc_dicts[0] ) , ['embeddings', 'id', 'text', 'title'] )
self.assertEqual(len(doc_dicts[0]['id'] ) , __snake_case )
self.assertEqual(doc_dicts[0]['id'][0] , '1' ) # max inner product is reached with second doc
self.assertEqual(doc_dicts[1]['id'][0] , '0' ) # max inner product is reached with first doc
self.assertListEqual(doc_ids.tolist() , [[1], [0]] )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.get_dummy_custom_hf_index_retriever(from_disk=__snake_case )
with tempfile.TemporaryDirectory() as tmp_dirname:
retriever.save_pretrained(__snake_case )
__a =RagRetriever.from_pretrained(__snake_case )
self.assertIsInstance(__snake_case , __snake_case )
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=1 )
self.assertTrue(out is not None )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =1
__a =self.get_dummy_legacy_index_retriever()
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=__snake_case )
self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) )
self.assertEqual(len(__snake_case ) , 2 )
self.assertEqual(sorted(doc_dicts[0] ) , ['text', 'title'] )
self.assertEqual(len(doc_dicts[0]['text'] ) , __snake_case )
self.assertEqual(doc_dicts[0]['text'][0] , 'bar' ) # max inner product is reached with second doc
self.assertEqual(doc_dicts[1]['text'][0] , 'foo' ) # max inner product is reached with first doc
self.assertListEqual(doc_ids.tolist() , [[1], [0]] )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
__a =self.get_dummy_legacy_index_retriever()
with tempfile.TemporaryDirectory() as tmp_dirname:
retriever.save_pretrained(__snake_case )
__a =RagRetriever.from_pretrained(__snake_case )
self.assertIsInstance(__snake_case , __snake_case )
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever.retrieve(__snake_case , n_docs=1 )
self.assertTrue(out is not None )
@require_torch
@require_tokenizers
@require_sentencepiece
def __magic_name__ ( self ) -> int:
'''simple docstring'''
import torch
__a =1
__a =self.get_dummy_canonical_hf_index_retriever()
__a =[[5, 7], [10, 11]]
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever(__snake_case , __snake_case , prefix=retriever.config.generator.prefix , n_docs=__snake_case )
__a =(
out['context_input_ids'],
out['context_attention_mask'],
out['retrieved_doc_embeds'],
)
self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) )
self.assertIsInstance(__snake_case , __snake_case )
self.assertIsInstance(__snake_case , __snake_case )
self.assertIsInstance(__snake_case , np.ndarray )
__a =retriever(
__snake_case , __snake_case , prefix=retriever.config.generator.prefix , n_docs=__snake_case , return_tensors='pt' , )
__a =( # noqa: F841
out['context_input_ids'],
out['context_attention_mask'],
out['retrieved_doc_embeds'],
out['doc_ids'],
)
self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) )
self.assertIsInstance(__snake_case , torch.Tensor )
self.assertIsInstance(__snake_case , torch.Tensor )
self.assertIsInstance(__snake_case , torch.Tensor )
@require_torch
@require_tokenizers
@require_sentencepiece
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.get_dpr_ctx_encoder_tokenizer()
__a =1
__a =self.get_dummy_custom_hf_index_retriever(from_disk=__snake_case )
retriever.set_ctx_encoder_tokenizer(__snake_case )
__a =[[5, 7], [10, 11]]
__a =np.array(
[np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa )
__a =retriever(__snake_case , __snake_case , prefix=retriever.config.generator.prefix , n_docs=__snake_case )
self.assertEqual(
len(__snake_case ) , 6 ) # check whether the retriever output consist of 6 attributes including tokenized docs
self.assertEqual(
all(k in out for k in ('tokenized_doc_ids', 'tokenized_doc_attention_mask') ) , __snake_case ) # check for doc token related keys in dictionary.
| 359
|
from ...configuration_utils import PretrainedConfig
from ...utils import logging
_lowerCAmelCase : Any = logging.get_logger(__name__)
_lowerCAmelCase : int = {
"caidas/swin2sr-classicalsr-x2-64": (
"https://huggingface.co/caidas/swin2sr-classicalsr-x2-64/resolve/main/config.json"
),
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'swin2sr'
SCREAMING_SNAKE_CASE = {
'hidden_size': 'embed_dim',
'num_attention_heads': 'num_heads',
'num_hidden_layers': 'num_layers',
}
def __init__( self , __snake_case=64 , __snake_case=1 , __snake_case=3 , __snake_case=180 , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=[6, 6, 6, 6, 6, 6] , __snake_case=8 , __snake_case=2.0 , __snake_case=True , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.1 , __snake_case="gelu" , __snake_case=False , __snake_case=0.02 , __snake_case=1e-5 , __snake_case=2 , __snake_case=1.0 , __snake_case="1conv" , __snake_case="pixelshuffle" , **__snake_case , ) -> Dict:
'''simple docstring'''
super().__init__(**__snake_case )
__a =image_size
__a =patch_size
__a =num_channels
__a =embed_dim
__a =depths
__a =len(__snake_case )
__a =num_heads
__a =window_size
__a =mlp_ratio
__a =qkv_bias
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =drop_path_rate
__a =hidden_act
__a =use_absolute_embeddings
__a =layer_norm_eps
__a =initializer_range
__a =upscale
__a =img_range
__a =resi_connection
__a =upsampler
| 308
| 0
|
import os
import re
import sys
import traceback
import warnings
from pathlib import Path
from typing import Dict, Optional, Union
from uuid import uuida
from huggingface_hub import HfFolder, ModelCard, ModelCardData, hf_hub_download, whoami
from huggingface_hub.file_download import REGEX_COMMIT_HASH
from huggingface_hub.utils import (
EntryNotFoundError,
RepositoryNotFoundError,
RevisionNotFoundError,
is_jinja_available,
)
from packaging import version
from requests import HTTPError
from .. import __version__
from .constants import (
DEPRECATED_REVISION_ARGS,
DIFFUSERS_CACHE,
HUGGINGFACE_CO_RESOLVE_ENDPOINT,
SAFETENSORS_WEIGHTS_NAME,
WEIGHTS_NAME,
)
from .import_utils import (
ENV_VARS_TRUE_VALUES,
_flax_version,
_jax_version,
_onnxruntime_version,
_torch_version,
is_flax_available,
is_onnx_available,
is_torch_available,
)
from .logging import get_logger
_lowerCAmelCase : Any = get_logger(__name__)
_lowerCAmelCase : Tuple = Path(__file__).parent / "model_card_template.md"
_lowerCAmelCase : Any = uuida().hex
_lowerCAmelCase : Tuple = os.getenv("HF_HUB_OFFLINE", "").upper() in ENV_VARS_TRUE_VALUES
_lowerCAmelCase : List[Any] = os.getenv("DISABLE_TELEMETRY", "").upper() in ENV_VARS_TRUE_VALUES
_lowerCAmelCase : Tuple = HUGGINGFACE_CO_RESOLVE_ENDPOINT + "/api/telemetry/"
def UpperCamelCase_( _snake_case : Union[Dict, str, None] = None ):
"""simple docstring"""
__a =F'diffusers/{__version__}; python/{sys.version.split()[0]}; session_id/{SESSION_ID}'
if DISABLE_TELEMETRY or HF_HUB_OFFLINE:
return ua + "; telemetry/off"
if is_torch_available():
ua += F'; torch/{_torch_version}'
if is_flax_available():
ua += F'; jax/{_jax_version}'
ua += F'; flax/{_flax_version}'
if is_onnx_available():
ua += F'; onnxruntime/{_onnxruntime_version}'
# CI will set this value to True
if os.environ.get('DIFFUSERS_IS_CI' , '' ).upper() in ENV_VARS_TRUE_VALUES:
ua += "; is_ci/true"
if isinstance(a__ , a__ ):
ua += "; " + "; ".join(F'{k}/{v}' for k, v in user_agent.items() )
elif isinstance(a__ , a__ ):
ua += "; " + user_agent
return ua
def UpperCamelCase_( _snake_case : str , _snake_case : Optional[str] = None , _snake_case : Optional[str] = None ):
"""simple docstring"""
if token is None:
__a =HfFolder.get_token()
if organization is None:
__a =whoami(a__ )['name']
return F'{username}/{model_id}'
else:
return F'{organization}/{model_id}'
def UpperCamelCase_( _snake_case : str , _snake_case : Optional[int] ):
"""simple docstring"""
if not is_jinja_available():
raise ValueError(
'Modelcard rendering is based on Jinja templates.'
' Please make sure to have `jinja` installed before using `create_model_card`.'
' To install it, please run `pip install Jinja2`.' )
if hasattr(a__ , 'local_rank' ) and args.local_rank not in [-1, 0]:
return
__a =args.hub_token if hasattr(a__ , 'hub_token' ) else None
__a =get_full_repo_name(a__ , token=a__ )
__a =ModelCard.from_template(
card_data=ModelCardData( # Card metadata object that will be converted to YAML block
language='en' , license='apache-2.0' , library_name='diffusers' , tags=[] , datasets=args.dataset_name , metrics=[] , ) , template_path=a__ , model_name=a__ , repo_name=a__ , dataset_name=args.dataset_name if hasattr(a__ , 'dataset_name' ) else None , learning_rate=args.learning_rate , train_batch_size=args.train_batch_size , eval_batch_size=args.eval_batch_size , gradient_accumulation_steps=(
args.gradient_accumulation_steps if hasattr(a__ , 'gradient_accumulation_steps' ) else None
) , adam_betaa=args.adam_betaa if hasattr(a__ , 'adam_beta1' ) else None , adam_betaa=args.adam_betaa if hasattr(a__ , 'adam_beta2' ) else None , adam_weight_decay=args.adam_weight_decay if hasattr(a__ , 'adam_weight_decay' ) else None , adam_epsilon=args.adam_epsilon if hasattr(a__ , 'adam_epsilon' ) else None , lr_scheduler=args.lr_scheduler if hasattr(a__ , 'lr_scheduler' ) else None , lr_warmup_steps=args.lr_warmup_steps if hasattr(a__ , 'lr_warmup_steps' ) else None , ema_inv_gamma=args.ema_inv_gamma if hasattr(a__ , 'ema_inv_gamma' ) else None , ema_power=args.ema_power if hasattr(a__ , 'ema_power' ) else None , ema_max_decay=args.ema_max_decay if hasattr(a__ , 'ema_max_decay' ) else None , mixed_precision=args.mixed_precision , )
__a =os.path.join(args.output_dir , 'README.md' )
model_card.save(a__ )
def UpperCamelCase_( _snake_case : Optional[str] , _snake_case : Optional[str] = None ):
"""simple docstring"""
if resolved_file is None or commit_hash is not None:
return commit_hash
__a =str(Path(a__ ).as_posix() )
__a =re.search(r'snapshots/([^/]+)/' , a__ )
if search is None:
return None
__a =search.groups()[0]
return commit_hash if REGEX_COMMIT_HASH.match(a__ ) else None
# Old default cache path, potentially to be migrated.
# This logic was more or less taken from `transformers`, with the following differences:
# - Diffusers doesn't use custom environment variables to specify the cache path.
# - There is no need to migrate the cache format, just move the files to the new location.
_lowerCAmelCase : Tuple = os.path.expanduser(
os.getenv("HF_HOME", os.path.join(os.getenv("XDG_CACHE_HOME", "~/.cache"), "huggingface"))
)
_lowerCAmelCase : List[str] = os.path.join(hf_cache_home, "diffusers")
def UpperCamelCase_( _snake_case : Optional[str] = None , _snake_case : Optional[str] = None ):
"""simple docstring"""
if new_cache_dir is None:
__a =DIFFUSERS_CACHE
if old_cache_dir is None:
__a =old_diffusers_cache
__a =Path(a__ ).expanduser()
__a =Path(a__ ).expanduser()
for old_blob_path in old_cache_dir.glob('**/blobs/*' ):
if old_blob_path.is_file() and not old_blob_path.is_symlink():
__a =new_cache_dir / old_blob_path.relative_to(a__ )
new_blob_path.parent.mkdir(parents=a__ , exist_ok=a__ )
os.replace(a__ , a__ )
try:
os.symlink(a__ , a__ )
except OSError:
logger.warning(
'Could not create symlink between old cache and new cache. If you use an older version of diffusers again, files will be re-downloaded.' )
# At this point, old_cache_dir contains symlinks to the new cache (it can still be used).
_lowerCAmelCase : Dict = os.path.join(DIFFUSERS_CACHE, "version_diffusers_cache.txt")
if not os.path.isfile(cache_version_file):
_lowerCAmelCase : Optional[Any] = 0
else:
with open(cache_version_file) as f:
try:
_lowerCAmelCase : List[str] = int(f.read())
except ValueError:
_lowerCAmelCase : Optional[int] = 0
if cache_version < 1:
_lowerCAmelCase : Optional[int] = os.path.isdir(old_diffusers_cache) and len(os.listdir(old_diffusers_cache)) > 0
if old_cache_is_not_empty:
logger.warning(
"The cache for model files in Diffusers v0.14.0 has moved to a new location. Moving your "
"existing cached models. This is a one-time operation, you can interrupt it or run it "
"later by calling `diffusers.utils.hub_utils.move_cache()`."
)
try:
move_cache()
except Exception as e:
_lowerCAmelCase : str = "\n".join(traceback.format_tb(e.__traceback__))
logger.error(
f'''There was a problem when trying to move your cache:\n\n{trace}\n{e.__class__.__name__}: {e}\n\nPlease '''
"file an issue at https://github.com/huggingface/diffusers/issues/new/choose, copy paste this whole "
"message and we will do our best to help."
)
if cache_version < 1:
try:
os.makedirs(DIFFUSERS_CACHE, exist_ok=True)
with open(cache_version_file, "w") as f:
f.write("1")
except Exception:
logger.warning(
f'''There was a problem when trying to write in your cache folder ({DIFFUSERS_CACHE}). Please, ensure '''
"the directory exists and can be written to."
)
def UpperCamelCase_( _snake_case : str , _snake_case : Optional[str] = None ):
"""simple docstring"""
if variant is not None:
__a =weights_name.split('.' )
__a =splits[:-1] + [variant] + splits[-1:]
__a ='.'.join(a__ )
return weights_name
def UpperCamelCase_( _snake_case : Union[str, Any] , *,
_snake_case : str , _snake_case : Any , _snake_case : Dict , _snake_case : List[Any] , _snake_case : Optional[int] , _snake_case : List[Any] , _snake_case : Union[str, Any] , _snake_case : Union[str, Any] , _snake_case : List[str] , _snake_case : Union[str, Any] , _snake_case : List[Any]=None , ):
"""simple docstring"""
__a =str(a__ )
if os.path.isfile(a__ ):
return pretrained_model_name_or_path
elif os.path.isdir(a__ ):
if os.path.isfile(os.path.join(a__ , a__ ) ):
# Load from a PyTorch checkpoint
__a =os.path.join(a__ , a__ )
return model_file
elif subfolder is not None and os.path.isfile(
os.path.join(a__ , a__ , a__ ) ):
__a =os.path.join(a__ , a__ , a__ )
return model_file
else:
raise EnvironmentError(
F'Error no file named {weights_name} found in directory {pretrained_model_name_or_path}.' )
else:
# 1. First check if deprecated way of loading from branches is used
if (
revision in DEPRECATED_REVISION_ARGS
and (weights_name == WEIGHTS_NAME or weights_name == SAFETENSORS_WEIGHTS_NAME)
and version.parse(version.parse(a__ ).base_version ) >= version.parse('0.20.0' )
):
try:
__a =hf_hub_download(
a__ , filename=_add_variant(a__ , a__ ) , cache_dir=a__ , force_download=a__ , proxies=a__ , resume_download=a__ , local_files_only=a__ , use_auth_token=a__ , user_agent=a__ , subfolder=a__ , revision=revision or commit_hash , )
warnings.warn(
F'Loading the variant {revision} from {pretrained_model_name_or_path} via `revision=\'{revision}\'` is deprecated. Loading instead from `revision=\'main\'` with `variant={revision}`. Loading model variants via `revision=\'{revision}\'` will be removed in diffusers v1. Please use `variant=\'{revision}\'` instead.' , a__ , )
return model_file
except: # noqa: E722
warnings.warn(
F'You are loading the variant {revision} from {pretrained_model_name_or_path} via `revision=\'{revision}\'`. This behavior is deprecated and will be removed in diffusers v1. One should use `variant=\'{revision}\'` instead. However, it appears that {pretrained_model_name_or_path} currently does not have a {_add_variant(a__ , a__ )} file in the \'main\' branch of {pretrained_model_name_or_path}. \n The Diffusers team and community would be very grateful if you could open an issue: https://github.com/huggingface/diffusers/issues/new with the title \'{pretrained_model_name_or_path} is missing {_add_variant(a__ , a__ )}\' so that the correct variant file can be added.' , a__ , )
try:
# 2. Load model file as usual
__a =hf_hub_download(
a__ , filename=a__ , cache_dir=a__ , force_download=a__ , proxies=a__ , resume_download=a__ , local_files_only=a__ , use_auth_token=a__ , user_agent=a__ , subfolder=a__ , revision=revision or commit_hash , )
return model_file
except RepositoryNotFoundError:
raise EnvironmentError(
F'{pretrained_model_name_or_path} is not a local folder and is not a valid model identifier '
'listed on \'https://huggingface.co/models\'\nIf this is a private repository, make sure to pass a '
'token having permission to this repo with `use_auth_token` or log in with `huggingface-cli '
'login`.' )
except RevisionNotFoundError:
raise EnvironmentError(
F'{revision} is not a valid git identifier (branch name, tag name or commit id) that exists for '
'this model name. Check the model page at '
F'\'https://huggingface.co/{pretrained_model_name_or_path}\' for available revisions.' )
except EntryNotFoundError:
raise EnvironmentError(
F'{pretrained_model_name_or_path} does not appear to have a file named {weights_name}.' )
except HTTPError as err:
raise EnvironmentError(
F'There was a specific connection error when trying to load {pretrained_model_name_or_path}:\n{err}' )
except ValueError:
raise EnvironmentError(
F'We couldn\'t connect to \'{HUGGINGFACE_CO_RESOLVE_ENDPOINT}\' to load this model, couldn\'t find it'
F' in the cached files and it looks like {pretrained_model_name_or_path} is not the path to a'
F' directory containing a file named {weights_name} or'
' \nCheckout your internet connection or see how to run the library in'
' offline mode at \'https://huggingface.co/docs/diffusers/installation#offline-mode\'.' )
except EnvironmentError:
raise EnvironmentError(
F'Can\'t load the model for \'{pretrained_model_name_or_path}\'. If you were trying to load it from '
'\'https://huggingface.co/models\', make sure you don\'t have a local directory with the same name. '
F'Otherwise, make sure \'{pretrained_model_name_or_path}\' is the correct path to a directory '
F'containing a file named {weights_name}' )
| 360
|
import os
def UpperCamelCase_( _snake_case : str = "input.txt" ):
"""simple docstring"""
with open(os.path.join(os.path.dirname(_snake_case ) , _snake_case ) ) as input_file:
__a =[
[int(_snake_case ) for element in line.split(',' )]
for line in input_file.readlines()
]
__a =len(_snake_case )
__a =len(matrix[0] )
__a =[[-1 for _ in range(_snake_case )] for _ in range(_snake_case )]
for i in range(_snake_case ):
__a =matrix[i][0]
for j in range(1 , _snake_case ):
for i in range(_snake_case ):
__a =minimal_path_sums[i][j - 1] + matrix[i][j]
for i in range(1 , _snake_case ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i - 1][j] + matrix[i][j] )
for i in range(rows - 2 , -1 , -1 ):
__a =min(
minimal_path_sums[i][j] , minimal_path_sums[i + 1][j] + matrix[i][j] )
return min(minimal_path_sums_row[-1] for minimal_path_sums_row in minimal_path_sums )
if __name__ == "__main__":
print(f'''{solution() = }''')
| 308
| 0
|
from __future__ import annotations
_lowerCAmelCase : Union[str, Any] = list[list[int]]
# assigning initial values to the grid
_lowerCAmelCase : List[Any] = [
[3, 0, 6, 5, 0, 8, 4, 0, 0],
[5, 2, 0, 0, 0, 0, 0, 0, 0],
[0, 8, 7, 0, 0, 0, 0, 3, 1],
[0, 0, 3, 0, 1, 0, 0, 8, 0],
[9, 0, 0, 8, 6, 3, 0, 0, 5],
[0, 5, 0, 0, 9, 0, 6, 0, 0],
[1, 3, 0, 0, 0, 0, 2, 5, 0],
[0, 0, 0, 0, 0, 0, 0, 7, 4],
[0, 0, 5, 2, 0, 6, 3, 0, 0],
]
# a grid with no solution
_lowerCAmelCase : Any = [
[5, 0, 6, 5, 0, 8, 4, 0, 3],
[5, 2, 0, 0, 0, 0, 0, 0, 2],
[1, 8, 7, 0, 0, 0, 0, 3, 1],
[0, 0, 3, 0, 1, 0, 0, 8, 0],
[9, 0, 0, 8, 6, 3, 0, 0, 5],
[0, 5, 0, 0, 9, 0, 6, 0, 0],
[1, 3, 0, 0, 0, 0, 2, 5, 0],
[0, 0, 0, 0, 0, 0, 0, 7, 4],
[0, 0, 5, 2, 0, 6, 3, 0, 0],
]
def UpperCamelCase_( _snake_case : Matrix , _snake_case : int , _snake_case : int , _snake_case : int ):
"""simple docstring"""
for i in range(9 ):
if grid[row][i] == n or grid[i][column] == n:
return False
for i in range(3 ):
for j in range(3 ):
if grid[(row - row % 3) + i][(column - column % 3) + j] == n:
return False
return True
def UpperCamelCase_( _snake_case : Matrix ):
"""simple docstring"""
for i in range(9 ):
for j in range(9 ):
if grid[i][j] == 0:
return i, j
return None
def UpperCamelCase_( _snake_case : Matrix ):
"""simple docstring"""
if location := find_empty_location(_snake_case ):
__a , __a =location
else:
# If the location is ``None``, then the grid is solved.
return grid
for digit in range(1 , 10 ):
if is_safe(_snake_case , _snake_case , _snake_case , _snake_case ):
__a =digit
if sudoku(_snake_case ) is not None:
return grid
__a =0
return None
def UpperCamelCase_( _snake_case : Matrix ):
"""simple docstring"""
for row in grid:
for cell in row:
print(_snake_case , end=' ' )
print()
if __name__ == "__main__":
# make a copy of grid so that you can compare with the unmodified grid
for example_grid in (initial_grid, no_solution):
print("\nExample grid:\n" + "=" * 20)
print_solution(example_grid)
print("\nExample grid solution:")
_lowerCAmelCase : int = sudoku(example_grid)
if solution is not None:
print_solution(solution)
else:
print("Cannot find a solution.")
| 361
|
import warnings
from ...utils import logging
from .image_processing_mobilevit import MobileViTImageProcessor
_lowerCAmelCase : Any = logging.get_logger(__name__)
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , *__snake_case , **__snake_case ) -> None:
'''simple docstring'''
warnings.warn(
'The class MobileViTFeatureExtractor is deprecated and will be removed in version 5 of Transformers.'
' Please use MobileViTImageProcessor instead.' , __snake_case , )
super().__init__(*__snake_case , **__snake_case )
| 308
| 0
|
from __future__ import annotations
def UpperCamelCase_( _snake_case : str , _snake_case : str , _snake_case : Optional[Any] , _snake_case : Optional[Any] ):
"""simple docstring"""
__a =[]
__a =input_list[low:mid], input_list[mid : high + 1]
while left and right:
result.append((left if left[0] <= right[0] else right).pop(0 ) )
__a =result + left + right
return input_list
def UpperCamelCase_( _snake_case : Optional[int] ):
"""simple docstring"""
if len(__lowerCAmelCase ) <= 1:
return input_list
__a =list(__lowerCAmelCase )
# iteration for two-way merging
__a =2
while p <= len(__lowerCAmelCase ):
# getting low, high and middle value for merge-sort of single list
for i in range(0 , len(__lowerCAmelCase ) , __lowerCAmelCase ):
__a =i
__a =i + p - 1
__a =(low + high + 1) // 2
__a =merge(__lowerCAmelCase , __lowerCAmelCase , __lowerCAmelCase , __lowerCAmelCase )
# final merge of last two parts
if p * 2 >= len(__lowerCAmelCase ):
__a =i
__a =merge(__lowerCAmelCase , 0 , __lowerCAmelCase , len(__lowerCAmelCase ) - 1 )
break
p *= 2
return input_list
if __name__ == "__main__":
_lowerCAmelCase : Union[str, Any] = input("Enter numbers separated by a comma:\n").strip()
if user_input == "":
_lowerCAmelCase : str = []
else:
_lowerCAmelCase : List[str] = [int(item.strip()) for item in user_input.split(",")]
print(iter_merge_sort(unsorted))
| 362
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_sentencepiece_available
_lowerCAmelCase : int = {}
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : Dict = ["BartphoTokenizer"]
if TYPE_CHECKING:
try:
if not is_sentencepiece_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .tokenization_bartpho import BartphoTokenizer
else:
import sys
_lowerCAmelCase : Optional[Any] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 0
|
from typing import Callable, List, Optional, Union
import PIL
import torch
from transformers import (
CLIPImageProcessor,
CLIPSegForImageSegmentation,
CLIPSegProcessor,
CLIPTextModel,
CLIPTokenizer,
)
from diffusers import DiffusionPipeline
from diffusers.configuration_utils import FrozenDict
from diffusers.models import AutoencoderKL, UNetaDConditionModel
from diffusers.pipelines.stable_diffusion import StableDiffusionInpaintPipeline
from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
from diffusers.utils import deprecate, is_accelerate_available, logging
_lowerCAmelCase : str = logging.get_logger(__name__) # pylint: disable=invalid-name
class __magic_name__ ( _A ):
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[int]:
'''simple docstring'''
super().__init__()
if hasattr(scheduler.config , 'steps_offset' ) and scheduler.config.steps_offset != 1:
__a =(
f'The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`'
f' should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure '
'to update the config accordingly as leaving `steps_offset` might led to incorrect results'
' in future versions. If you have downloaded this checkpoint from the Hugging Face Hub,'
' it would be very nice if you could open a Pull request for the `scheduler/scheduler_config.json`'
' file'
)
deprecate('steps_offset!=1' , '1.0.0' , __SCREAMING_SNAKE_CASE , standard_warn=__SCREAMING_SNAKE_CASE )
__a =dict(scheduler.config )
__a =1
__a =FrozenDict(__SCREAMING_SNAKE_CASE )
if hasattr(scheduler.config , 'skip_prk_steps' ) and scheduler.config.skip_prk_steps is False:
__a =(
f'The configuration file of this scheduler: {scheduler} has not set the configuration'
' `skip_prk_steps`. `skip_prk_steps` should be set to True in the configuration file. Please make'
' sure to update the config accordingly as not setting `skip_prk_steps` in the config might lead to'
' incorrect results in future versions. If you have downloaded this checkpoint from the Hugging Face'
' Hub, it would be very nice if you could open a Pull request for the'
' `scheduler/scheduler_config.json` file'
)
deprecate('skip_prk_steps not set' , '1.0.0' , __SCREAMING_SNAKE_CASE , standard_warn=__SCREAMING_SNAKE_CASE )
__a =dict(scheduler.config )
__a =True
__a =FrozenDict(__SCREAMING_SNAKE_CASE )
if safety_checker is None:
logger.warning(
f'You have disabled the safety checker for {self.__class__} by passing `safety_checker=None`. Ensure'
' that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered'
' results in services or applications open to the public. Both the diffusers team and Hugging Face'
' strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling'
' it only for use-cases that involve analyzing network behavior or auditing its results. For more'
' information, please have a look at https://github.com/huggingface/diffusers/pull/254 .' )
self.register_modules(
segmentation_model=__SCREAMING_SNAKE_CASE , segmentation_processor=__SCREAMING_SNAKE_CASE , vae=__SCREAMING_SNAKE_CASE , text_encoder=__SCREAMING_SNAKE_CASE , tokenizer=__SCREAMING_SNAKE_CASE , unet=__SCREAMING_SNAKE_CASE , scheduler=__SCREAMING_SNAKE_CASE , safety_checker=__SCREAMING_SNAKE_CASE , feature_extractor=__SCREAMING_SNAKE_CASE , )
def __magic_name__ ( self , __snake_case = "auto" ) -> Dict:
'''simple docstring'''
if slice_size == "auto":
# half the attention head size is usually a good trade-off between
# speed and memory
__a =self.unet.config.attention_head_dim // 2
self.unet.set_attention_slice(__SCREAMING_SNAKE_CASE )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
self.enable_attention_slicing(__SCREAMING_SNAKE_CASE )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
if is_accelerate_available():
from accelerate import cpu_offload
else:
raise ImportError('Please install accelerate via `pip install accelerate`' )
__a =torch.device('cuda' )
for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae, self.safety_checker]:
if cpu_offloaded_model is not None:
cpu_offload(__SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE )
@property
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
if self.device != torch.device('meta' ) or not hasattr(self.unet , '_hf_hook' ):
return self.device
for module in self.unet.modules():
if (
hasattr(__SCREAMING_SNAKE_CASE , '_hf_hook' )
and hasattr(module._hf_hook , 'execution_device' )
and module._hf_hook.execution_device is not None
):
return torch.device(module._hf_hook.execution_device )
return self.device
@torch.no_grad()
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case = 512 , __snake_case = 512 , __snake_case = 50 , __snake_case = 7.5 , __snake_case = None , __snake_case = 1 , __snake_case = 0.0 , __snake_case = None , __snake_case = None , __snake_case = "pil" , __snake_case = True , __snake_case = None , __snake_case = 1 , **__snake_case , ) -> str:
'''simple docstring'''
__a =self.segmentation_processor(
text=[text] , images=[image] , padding='max_length' , return_tensors='pt' ).to(self.device )
__a =self.segmentation_model(**__SCREAMING_SNAKE_CASE )
__a =torch.sigmoid(outputs.logits ).cpu().detach().unsqueeze(-1 ).numpy()
__a =self.numpy_to_pil(__SCREAMING_SNAKE_CASE )[0].resize(image.size )
# Run inpainting pipeline with the generated mask
__a =StableDiffusionInpaintPipeline(
vae=self.vae , text_encoder=self.text_encoder , tokenizer=self.tokenizer , unet=self.unet , scheduler=self.scheduler , safety_checker=self.safety_checker , feature_extractor=self.feature_extractor , )
return inpainting_pipeline(
prompt=__SCREAMING_SNAKE_CASE , image=__SCREAMING_SNAKE_CASE , mask_image=__SCREAMING_SNAKE_CASE , height=__SCREAMING_SNAKE_CASE , width=__SCREAMING_SNAKE_CASE , num_inference_steps=__SCREAMING_SNAKE_CASE , guidance_scale=__SCREAMING_SNAKE_CASE , negative_prompt=__SCREAMING_SNAKE_CASE , num_images_per_prompt=__SCREAMING_SNAKE_CASE , eta=__SCREAMING_SNAKE_CASE , generator=__SCREAMING_SNAKE_CASE , latents=__SCREAMING_SNAKE_CASE , output_type=__SCREAMING_SNAKE_CASE , return_dict=__SCREAMING_SNAKE_CASE , callback=__SCREAMING_SNAKE_CASE , callback_steps=__SCREAMING_SNAKE_CASE , )
| 363
|
from collections import OrderedDict
from typing import Mapping
from packaging import version
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
_lowerCAmelCase : List[str] = logging.get_logger(__name__)
_lowerCAmelCase : List[Any] = {
"hustvl/yolos-small": "https://huggingface.co/hustvl/yolos-small/resolve/main/config.json",
# See all YOLOS models at https://huggingface.co/models?filter=yolos
}
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = 'yolos'
def __init__( self , __snake_case=768 , __snake_case=12 , __snake_case=12 , __snake_case=3072 , __snake_case="gelu" , __snake_case=0.0 , __snake_case=0.0 , __snake_case=0.02 , __snake_case=1e-12 , __snake_case=[512, 864] , __snake_case=16 , __snake_case=3 , __snake_case=True , __snake_case=100 , __snake_case=True , __snake_case=False , __snake_case=1 , __snake_case=5 , __snake_case=2 , __snake_case=5 , __snake_case=2 , __snake_case=0.1 , **__snake_case , ) -> str:
'''simple docstring'''
super().__init__(**__snake_case )
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =intermediate_size
__a =hidden_act
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =initializer_range
__a =layer_norm_eps
__a =image_size
__a =patch_size
__a =num_channels
__a =qkv_bias
__a =num_detection_tokens
__a =use_mid_position_embeddings
__a =auxiliary_loss
# Hungarian matcher
__a =class_cost
__a =bbox_cost
__a =giou_cost
# Loss coefficients
__a =bbox_loss_coefficient
__a =giou_loss_coefficient
__a =eos_coefficient
class __magic_name__ ( lowerCAmelCase_ ):
SCREAMING_SNAKE_CASE = version.parse('1.11' )
@property
def __magic_name__ ( self ) -> Mapping[str, Mapping[int, str]]:
'''simple docstring'''
return OrderedDict(
[
('pixel_values', {0: 'batch', 1: 'num_channels', 2: 'height', 3: 'width'}),
] )
@property
def __magic_name__ ( self ) -> float:
'''simple docstring'''
return 1e-4
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return 12
| 308
| 0
|
from __future__ import annotations
import math
_lowerCAmelCase : int = '2020.9.26'
_lowerCAmelCase : int = 'xcodz-dot, cclaus, dhruvmanila'
def UpperCamelCase_( _snake_case : Tuple , _snake_case : int , _snake_case : Optional[int] , _snake_case : int , _snake_case : str ):
"""simple docstring"""
if not all(isinstance(__snake_case , (float, int) ) for val in locals().values() ):
__a =F'Input values must either be float or int: {list(locals().values() )}'
raise TypeError(__snake_case )
__a =((x * distance) / (z + distance)) * scale
__a =((y * distance) / (z + distance)) * scale
return projected_x, projected_y
def UpperCamelCase_( _snake_case : Optional[Any] , _snake_case : Any , _snake_case : str , _snake_case : Tuple , _snake_case : str ):
"""simple docstring"""
if not isinstance(__snake_case , __snake_case ):
raise TypeError('Axis must be a str' )
__a =locals()
del input_variables["axis"]
if not all(isinstance(__snake_case , (float, int) ) for val in input_variables.values() ):
__a =(
"Input values except axis must either be float or int: "
F'{list(input_variables.values() )}'
)
raise TypeError(__snake_case )
__a =(angle % 360) / 450 * 180 / math.pi
if axis == "z":
__a =x * math.cos(__snake_case ) - y * math.sin(__snake_case )
__a =y * math.cos(__snake_case ) + x * math.sin(__snake_case )
__a =z
elif axis == "x":
__a =y * math.cos(__snake_case ) - z * math.sin(__snake_case )
__a =z * math.cos(__snake_case ) + y * math.sin(__snake_case )
__a =x
elif axis == "y":
__a =x * math.cos(__snake_case ) - z * math.sin(__snake_case )
__a =z * math.cos(__snake_case ) + x * math.sin(__snake_case )
__a =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, 10.0, 10.0) = }''')
print(f'''{rotate(1.0, 2.0, 3.0, "y", 90.0) = }''')
| 364
|
import argparse
import torch
from datasets import load_dataset
from donut import DonutModel
from transformers import (
DonutImageProcessor,
DonutProcessor,
DonutSwinConfig,
DonutSwinModel,
MBartConfig,
MBartForCausalLM,
VisionEncoderDecoderModel,
XLMRobertaTokenizerFast,
)
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
__a =model.config
__a =DonutSwinConfig(
image_size=original_config.input_size , patch_size=4 , depths=original_config.encoder_layer , num_heads=[4, 8, 16, 32] , window_size=original_config.window_size , embed_dim=128 , )
__a =MBartConfig(
is_decoder=_snake_case , is_encoder_decoder=_snake_case , add_cross_attention=_snake_case , decoder_layers=original_config.decoder_layer , max_position_embeddings=original_config.max_position_embeddings , vocab_size=len(
model.decoder.tokenizer ) , scale_embedding=_snake_case , add_final_layer_norm=_snake_case , )
return encoder_config, decoder_config
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
if "encoder.model" in name:
__a =name.replace('encoder.model' , 'encoder' )
if "decoder.model" in name:
__a =name.replace('decoder.model' , 'decoder' )
if "patch_embed.proj" in name:
__a =name.replace('patch_embed.proj' , 'embeddings.patch_embeddings.projection' )
if "patch_embed.norm" in name:
__a =name.replace('patch_embed.norm' , 'embeddings.norm' )
if name.startswith('encoder' ):
if "layers" in name:
__a ='encoder.' + name
if "attn.proj" in name:
__a =name.replace('attn.proj' , 'attention.output.dense' )
if "attn" in name and "mask" not in name:
__a =name.replace('attn' , 'attention.self' )
if "norm1" in name:
__a =name.replace('norm1' , 'layernorm_before' )
if "norm2" in name:
__a =name.replace('norm2' , 'layernorm_after' )
if "mlp.fc1" in name:
__a =name.replace('mlp.fc1' , 'intermediate.dense' )
if "mlp.fc2" in name:
__a =name.replace('mlp.fc2' , 'output.dense' )
if name == "encoder.norm.weight":
__a ='encoder.layernorm.weight'
if name == "encoder.norm.bias":
__a ='encoder.layernorm.bias'
return name
def UpperCamelCase_( _snake_case : Tuple , _snake_case : str ):
"""simple docstring"""
for key in orig_state_dict.copy().keys():
__a =orig_state_dict.pop(_snake_case )
if "qkv" in key:
__a =key.split('.' )
__a =int(key_split[3] )
__a =int(key_split[5] )
__a =model.encoder.encoder.layers[layer_num].blocks[block_num].attention.self.all_head_size
if "weight" in key:
__a =val[:dim, :]
__a =val[dim : dim * 2, :]
__a =val[-dim:, :]
else:
__a =val[:dim]
__a =val[dim : dim * 2]
__a =val[-dim:]
elif "attn_mask" in key or key in ["encoder.model.norm.weight", "encoder.model.norm.bias"]:
# HuggingFace implementation doesn't use attn_mask buffer
# and model doesn't use final LayerNorms for the encoder
pass
else:
__a =val
return orig_state_dict
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Union[str, Any]=None , _snake_case : List[Any]=False ):
"""simple docstring"""
__a =DonutModel.from_pretrained(_snake_case ).eval()
# load HuggingFace model
__a , __a =get_configs(_snake_case )
__a =DonutSwinModel(_snake_case )
__a =MBartForCausalLM(_snake_case )
__a =VisionEncoderDecoderModel(encoder=_snake_case , decoder=_snake_case )
model.eval()
__a =original_model.state_dict()
__a =convert_state_dict(_snake_case , _snake_case )
model.load_state_dict(_snake_case )
# verify results on scanned document
__a =load_dataset('hf-internal-testing/example-documents' )
__a =dataset['test'][0]['image'].convert('RGB' )
__a =XLMRobertaTokenizerFast.from_pretrained(_snake_case , from_slow=_snake_case )
__a =DonutImageProcessor(
do_align_long_axis=original_model.config.align_long_axis , size=original_model.config.input_size[::-1] )
__a =DonutProcessor(_snake_case , _snake_case )
__a =processor(_snake_case , return_tensors='pt' ).pixel_values
if model_name == "naver-clova-ix/donut-base-finetuned-docvqa":
__a ='<s_docvqa><s_question>{user_input}</s_question><s_answer>'
__a ='When is the coffee break?'
__a =task_prompt.replace('{user_input}' , _snake_case )
elif model_name == "naver-clova-ix/donut-base-finetuned-rvlcdip":
__a ='<s_rvlcdip>'
elif model_name in [
"naver-clova-ix/donut-base-finetuned-cord-v1",
"naver-clova-ix/donut-base-finetuned-cord-v1-2560",
]:
__a ='<s_cord>'
elif model_name == "naver-clova-ix/donut-base-finetuned-cord-v2":
__a ='s_cord-v2>'
elif model_name == "naver-clova-ix/donut-base-finetuned-zhtrainticket":
__a ='<s_zhtrainticket>'
elif model_name in ["naver-clova-ix/donut-proto", "naver-clova-ix/donut-base"]:
# use a random prompt
__a ='hello world'
else:
raise ValueError('Model name not supported' )
__a =original_model.decoder.tokenizer(_snake_case , add_special_tokens=_snake_case , return_tensors='pt' )[
'input_ids'
]
__a =original_model.encoder.model.patch_embed(_snake_case )
__a , __a =model.encoder.embeddings(_snake_case )
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
# verify encoder hidden states
__a =original_model.encoder(_snake_case )
__a =model.encoder(_snake_case ).last_hidden_state
assert torch.allclose(_snake_case , _snake_case , atol=1e-2 )
# verify decoder hidden states
__a =original_model(_snake_case , _snake_case , _snake_case ).logits
__a =model(_snake_case , decoder_input_ids=_snake_case ).logits
assert torch.allclose(_snake_case , _snake_case , atol=1e-3 )
print('Looks ok!' )
if pytorch_dump_folder_path is not None:
print(F'Saving model and processor to {pytorch_dump_folder_path}' )
model.save_pretrained(_snake_case )
processor.save_pretrained(_snake_case )
if push_to_hub:
model.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
processor.push_to_hub('nielsr/' + model_name.split('/' )[-1] , commit_message='Update model' )
if __name__ == "__main__":
_lowerCAmelCase : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_name",
default="naver-clova-ix/donut-base-finetuned-docvqa",
required=False,
type=str,
help="Name of the original model you'd like to convert.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
required=False,
type=str,
help="Path to the output PyTorch model directory.",
)
parser.add_argument(
"--push_to_hub",
action="store_true",
help="Whether or not to push the converted model and processor to the 🤗 hub.",
)
_lowerCAmelCase : List[Any] = parser.parse_args()
convert_donut_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 308
| 0
|
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 MobileNetVaImageProcessor
class __magic_name__ ( unittest.TestCase ):
def __init__( self , __snake_case , __snake_case=7 , __snake_case=3 , __snake_case=18 , __snake_case=30 , __snake_case=400 , __snake_case=True , __snake_case=None , __snake_case=True , __snake_case=None , ) -> Any:
'''simple docstring'''
__a =size if size is not None else {"""shortest_edge""": 20}
__a =crop_size if crop_size is not None else {"""height""": 18, """width""": 18}
__a =parent
__a =batch_size
__a =num_channels
__a =image_size
__a =min_resolution
__a =max_resolution
__a =do_resize
__a =size
__a =do_center_crop
__a =crop_size
def __magic_name__ ( self ) -> str:
'''simple docstring'''
return {
"do_resize": self.do_resize,
"size": self.size,
"do_center_crop": self.do_center_crop,
"crop_size": self.crop_size,
}
@require_torch
@require_vision
class __magic_name__ ( __snake_case , unittest.TestCase ):
SCREAMING_SNAKE_CASE = MobileNetVaImageProcessor if is_vision_available() else None
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =MobileNetVaImageProcessingTester(self )
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return self.image_processor_tester.prepare_image_processor_dict()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.image_processing_class(**self.image_processor_dict )
self.assertTrue(hasattr(lowerCamelCase_ , 'do_resize' ) )
self.assertTrue(hasattr(lowerCamelCase_ , 'size' ) )
self.assertTrue(hasattr(lowerCamelCase_ , 'do_center_crop' ) )
self.assertTrue(hasattr(lowerCamelCase_ , 'crop_size' ) )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.image_processing_class.from_dict(self.image_processor_dict )
self.assertEqual(image_processor.size , {'shortest_edge': 20} )
self.assertEqual(image_processor.crop_size , {'height': 18, 'width': 18} )
__a =self.image_processing_class.from_dict(self.image_processor_dict , size=42 , crop_size=84 )
self.assertEqual(image_processor.size , {'shortest_edge': 42} )
self.assertEqual(image_processor.crop_size , {'height': 84, 'width': 84} )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
pass
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
# Initialize image_processing
__a =self.image_processing_class(**self.image_processor_dict )
# create random PIL images
__a =prepare_image_inputs(self.image_processor_tester , equal_resolution=lowerCamelCase_ )
for image in image_inputs:
self.assertIsInstance(lowerCamelCase_ , Image.Image )
# Test not batched input
__a =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 =image_processing(lowerCamelCase_ , 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 __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# Initialize image_processing
__a =self.image_processing_class(**self.image_processor_dict )
# create random numpy tensors
__a =prepare_image_inputs(self.image_processor_tester , equal_resolution=lowerCamelCase_ , numpify=lowerCamelCase_ )
for image in image_inputs:
self.assertIsInstance(lowerCamelCase_ , np.ndarray )
# Test not batched input
__a =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 =image_processing(lowerCamelCase_ , 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 __magic_name__ ( self ) -> int:
'''simple docstring'''
# Initialize image_processing
__a =self.image_processing_class(**self.image_processor_dict )
# create random PyTorch tensors
__a =prepare_image_inputs(self.image_processor_tester , equal_resolution=lowerCamelCase_ , torchify=lowerCamelCase_ )
for image in image_inputs:
self.assertIsInstance(lowerCamelCase_ , torch.Tensor )
# Test not batched input
__a =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 =image_processing(lowerCamelCase_ , 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'],
) , )
| 365
|
import json
import os
import unittest
from transformers.models.blenderbot_small.tokenization_blenderbot_small import (
VOCAB_FILES_NAMES,
BlenderbotSmallTokenizer,
)
from ...test_tokenization_common import TokenizerTesterMixin
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = BlenderbotSmallTokenizer
SCREAMING_SNAKE_CASE = False
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
super().setUp()
__a =['__start__', 'adapt', 'act', 'ap@@', 'te', '__end__', '__unk__']
__a =dict(zip(__snake_case , range(len(__snake_case ) ) ) )
__a =['#version: 0.2', 'a p', 't e</w>', 'ap t</w>', 'a d', 'ad apt</w>', 'a c', 'ac t</w>', '']
__a ={'unk_token': '__unk__', 'bos_token': '__start__', 'eos_token': '__end__'}
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['vocab_file'] )
__a =os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['merges_file'] )
with open(self.vocab_file , 'w' , encoding='utf-8' ) as fp:
fp.write(json.dumps(__snake_case ) + '\n' )
with open(self.merges_file , 'w' , encoding='utf-8' ) as fp:
fp.write('\n'.join(__snake_case ) )
def __magic_name__ ( self , **__snake_case ) -> Any:
'''simple docstring'''
kwargs.update(self.special_tokens_map )
return BlenderbotSmallTokenizer.from_pretrained(self.tmpdirname , **__snake_case )
def __magic_name__ ( self , __snake_case ) -> List[Any]:
'''simple docstring'''
__a ='adapt act apte'
__a ='adapt act apte'
return input_text, output_text
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
__a ='adapt act apte'
__a =['adapt', 'act', 'ap@@', 'te']
__a =tokenizer.tokenize(__snake_case )
self.assertListEqual(__snake_case , __snake_case )
__a =[tokenizer.bos_token] + tokens + [tokenizer.eos_token]
__a =[0, 1, 2, 3, 4, 5]
self.assertListEqual(tokenizer.convert_tokens_to_ids(__snake_case ) , __snake_case )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
assert tok('sam' ).input_ids == [1384]
__a ='I am a small frog.'
__a =tok([src_text] , padding=__snake_case , truncation=__snake_case )['input_ids']
__a =tok.batch_decode(__snake_case , skip_special_tokens=__snake_case , clean_up_tokenization_spaces=__snake_case )[0]
assert src_text != decoded # I wish it did!
assert decoded == "i am a small frog ."
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =BlenderbotSmallTokenizer.from_pretrained('facebook/blenderbot-90M' )
__a ='I am a small frog .'
__a ='.'
__a =tok(__snake_case )['input_ids']
__a =tok(__snake_case )['input_ids']
assert encoded[-1] == encoded_dot[0]
| 308
| 0
|
from timeit import timeit
_lowerCAmelCase : Union[str, Any] = {
"""MALAYALAM""": True,
"""String""": False,
"""rotor""": True,
"""level""": True,
"""A""": True,
"""BB""": True,
"""ABC""": False,
"""amanaplanacanalpanama""": True, # "a man a plan a canal panama"
}
# Ensure our test data is valid
assert all((key == key[::-1]) is value for key, value in test_data.items())
def UpperCamelCase_( _snake_case : Any ):
"""simple docstring"""
__a =0
__a =len(SCREAMING_SNAKE_CASE_ ) - 1
while start_i < end_i:
if s[start_i] == s[end_i]:
start_i += 1
end_i -= 1
else:
return False
return True
def UpperCamelCase_( _snake_case : Any ):
"""simple docstring"""
__a =len(SCREAMING_SNAKE_CASE_ ) // 2
__a =len(SCREAMING_SNAKE_CASE_ )
# We need to traverse till half of the length of string
# as we can get access of the i'th last element from
# i'th index.
# eg: [0,1,2,3,4,5] => 4th index can be accessed
# with the help of 1st index (i==n-i-1)
# where n is length of string
return all(s[i] == s[n - i - 1] for i in range(SCREAMING_SNAKE_CASE_ ) )
def UpperCamelCase_( _snake_case : Any ):
"""simple docstring"""
if len(SCREAMING_SNAKE_CASE_ ) <= 2:
return True
if s[0] == s[len(SCREAMING_SNAKE_CASE_ ) - 1]:
return is_palindrome_recursive(s[1:-1] )
else:
return False
def UpperCamelCase_( _snake_case : Tuple ):
"""simple docstring"""
return s == s[::-1]
def UpperCamelCase_( _snake_case : Any ):
"""simple docstring"""
__a =F'all({name}(key) is value for key, value in test_data.items())'
__a =F'from __main__ import test_data, {name}'
__a =500000
__a =timeit(stmt=SCREAMING_SNAKE_CASE_ , setup=SCREAMING_SNAKE_CASE_ , number=SCREAMING_SNAKE_CASE_ )
print(F'{name:<35} finished {number:,} runs in {result:.5f} seconds' )
if __name__ == "__main__":
for key, value in test_data.items():
assert is_palindrome(key) is is_palindrome_recursive(key)
assert is_palindrome(key) is is_palindrome_slice(key)
print(f'''{key:21} {value}''')
print("a man a plan a canal panama")
# finished 500,000 runs in 0.46793 seconds
benchmark_function("is_palindrome_slice")
# finished 500,000 runs in 0.85234 seconds
benchmark_function("is_palindrome")
# finished 500,000 runs in 1.32028 seconds
benchmark_function("is_palindrome_recursive")
# finished 500,000 runs in 2.08679 seconds
benchmark_function("is_palindrome_traversal")
| 366
|
import unittest
from transformers import load_tool
from transformers.utils import is_torch_available
if is_torch_available():
import torch
from transformers.testing_utils import require_torch
from .test_tools_common import ToolTesterMixin
@require_torch
class __magic_name__ ( unittest.TestCase , lowerCAmelCase_ ):
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =load_tool('text-to-speech' )
self.tool.setup()
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# SpeechT5 isn't deterministic
torch.manual_seed(0 )
__a =self.tool('hey' )
__a =result.to_raw()
self.assertTrue(
torch.allclose(
resulting_tensor[:3] , torch.tensor([-0.000_5966_6688_3211_5829, -0.000_3657_6401_9079_5064, -0.0001_3439_5027_9988_3485] ) , ) )
| 308
| 0
|
import argparse
import json
import pickle
from pathlib import Path
import requests
import torch
from huggingface_hub import hf_hub_download
from PIL import Image
from transformers import MaskFormerConfig, MaskFormerForInstanceSegmentation, MaskFormerImageProcessor, SwinConfig
from transformers.utils import logging
logging.set_verbosity_info()
_lowerCAmelCase : Optional[Any] = logging.get_logger(__name__)
def UpperCamelCase_( _snake_case : str ):
"""simple docstring"""
__a =SwinConfig.from_pretrained(
'microsoft/swin-tiny-patch4-window7-224' , out_features=['stage1', 'stage2', 'stage3', 'stage4'] )
__a =MaskFormerConfig(backbone_config=SCREAMING_SNAKE_CASE__ )
__a ='huggingface/label-files'
if "ade20k-full" in model_name:
# this should be ok
__a =847
__a ='maskformer-ade20k-full-id2label.json'
elif "ade" in model_name:
# this should be ok
__a =150
__a ='ade20k-id2label.json'
elif "coco-stuff" in model_name:
# this should be ok
__a =171
__a ='maskformer-coco-stuff-id2label.json'
elif "coco" in model_name:
# TODO
__a =133
__a ='coco-panoptic-id2label.json'
elif "cityscapes" in model_name:
# this should be ok
__a =19
__a ='cityscapes-id2label.json'
elif "vistas" in model_name:
# this should be ok
__a =65
__a ='mapillary-vistas-id2label.json'
__a =json.load(open(hf_hub_download(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , repo_type='dataset' ) , 'r' ) )
__a ={int(SCREAMING_SNAKE_CASE__ ): v for k, v in idalabel.items()}
return config
def UpperCamelCase_( _snake_case : Optional[int] ):
"""simple docstring"""
__a =[]
# stem
# fmt: off
rename_keys.append(('backbone.patch_embed.proj.weight', 'model.pixel_level_module.encoder.model.embeddings.patch_embeddings.projection.weight') )
rename_keys.append(('backbone.patch_embed.proj.bias', 'model.pixel_level_module.encoder.model.embeddings.patch_embeddings.projection.bias') )
rename_keys.append(('backbone.patch_embed.norm.weight', 'model.pixel_level_module.encoder.model.embeddings.norm.weight') )
rename_keys.append(('backbone.patch_embed.norm.bias', 'model.pixel_level_module.encoder.model.embeddings.norm.bias') )
# stages
for i in range(len(config.backbone_config.depths ) ):
for j in range(config.backbone_config.depths[i] ):
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.norm1.weight', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.layernorm_before.weight') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.norm1.bias', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.layernorm_before.bias') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.attn.relative_position_bias_table', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.attention.self.relative_position_bias_table') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.attn.relative_position_index', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.attention.self.relative_position_index') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.attn.proj.weight', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.attention.output.dense.weight') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.attn.proj.bias', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.attention.output.dense.bias') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.norm2.weight', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.layernorm_after.weight') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.norm2.bias', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.layernorm_after.bias') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.mlp.fc1.weight', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.intermediate.dense.weight') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.mlp.fc1.bias', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.intermediate.dense.bias') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.mlp.fc2.weight', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.output.dense.weight') )
rename_keys.append((F'backbone.layers.{i}.blocks.{j}.mlp.fc2.bias', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.blocks.{j}.output.dense.bias') )
if i < 3:
rename_keys.append((F'backbone.layers.{i}.downsample.reduction.weight', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.downsample.reduction.weight') )
rename_keys.append((F'backbone.layers.{i}.downsample.norm.weight', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.downsample.norm.weight') )
rename_keys.append((F'backbone.layers.{i}.downsample.norm.bias', F'model.pixel_level_module.encoder.model.encoder.layers.{i}.downsample.norm.bias') )
rename_keys.append((F'backbone.norm{i}.weight', F'model.pixel_level_module.encoder.hidden_states_norms.{i}.weight') )
rename_keys.append((F'backbone.norm{i}.bias', F'model.pixel_level_module.encoder.hidden_states_norms.{i}.bias') )
# FPN
rename_keys.append(('sem_seg_head.layer_4.weight', 'model.pixel_level_module.decoder.fpn.stem.0.weight') )
rename_keys.append(('sem_seg_head.layer_4.norm.weight', 'model.pixel_level_module.decoder.fpn.stem.1.weight') )
rename_keys.append(('sem_seg_head.layer_4.norm.bias', 'model.pixel_level_module.decoder.fpn.stem.1.bias') )
for source_index, target_index in zip(range(3 , 0 , -1 ) , range(0 , 3 ) ):
rename_keys.append((F'sem_seg_head.adapter_{source_index}.weight', F'model.pixel_level_module.decoder.fpn.layers.{target_index}.proj.0.weight') )
rename_keys.append((F'sem_seg_head.adapter_{source_index}.norm.weight', F'model.pixel_level_module.decoder.fpn.layers.{target_index}.proj.1.weight') )
rename_keys.append((F'sem_seg_head.adapter_{source_index}.norm.bias', F'model.pixel_level_module.decoder.fpn.layers.{target_index}.proj.1.bias') )
rename_keys.append((F'sem_seg_head.layer_{source_index}.weight', F'model.pixel_level_module.decoder.fpn.layers.{target_index}.block.0.weight') )
rename_keys.append((F'sem_seg_head.layer_{source_index}.norm.weight', F'model.pixel_level_module.decoder.fpn.layers.{target_index}.block.1.weight') )
rename_keys.append((F'sem_seg_head.layer_{source_index}.norm.bias', F'model.pixel_level_module.decoder.fpn.layers.{target_index}.block.1.bias') )
rename_keys.append(('sem_seg_head.mask_features.weight', 'model.pixel_level_module.decoder.mask_projection.weight') )
rename_keys.append(('sem_seg_head.mask_features.bias', 'model.pixel_level_module.decoder.mask_projection.bias') )
# Transformer decoder
for idx in range(config.decoder_config.decoder_layers ):
# self-attention out projection
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.self_attn.out_proj.weight', F'model.transformer_module.decoder.layers.{idx}.self_attn.out_proj.weight') )
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.self_attn.out_proj.bias', F'model.transformer_module.decoder.layers.{idx}.self_attn.out_proj.bias') )
# cross-attention out projection
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.multihead_attn.out_proj.weight', F'model.transformer_module.decoder.layers.{idx}.encoder_attn.out_proj.weight') )
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.multihead_attn.out_proj.bias', F'model.transformer_module.decoder.layers.{idx}.encoder_attn.out_proj.bias') )
# MLP 1
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.linear1.weight', F'model.transformer_module.decoder.layers.{idx}.fc1.weight') )
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.linear1.bias', F'model.transformer_module.decoder.layers.{idx}.fc1.bias') )
# MLP 2
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.linear2.weight', F'model.transformer_module.decoder.layers.{idx}.fc2.weight') )
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.linear2.bias', F'model.transformer_module.decoder.layers.{idx}.fc2.bias') )
# layernorm 1 (self-attention layernorm)
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.norm1.weight', F'model.transformer_module.decoder.layers.{idx}.self_attn_layer_norm.weight') )
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.norm1.bias', F'model.transformer_module.decoder.layers.{idx}.self_attn_layer_norm.bias') )
# layernorm 2 (cross-attention layernorm)
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.norm2.weight', F'model.transformer_module.decoder.layers.{idx}.encoder_attn_layer_norm.weight') )
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.norm2.bias', F'model.transformer_module.decoder.layers.{idx}.encoder_attn_layer_norm.bias') )
# layernorm 3 (final layernorm)
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.norm3.weight', F'model.transformer_module.decoder.layers.{idx}.final_layer_norm.weight') )
rename_keys.append((F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.norm3.bias', F'model.transformer_module.decoder.layers.{idx}.final_layer_norm.bias') )
rename_keys.append(('sem_seg_head.predictor.transformer.decoder.norm.weight', 'model.transformer_module.decoder.layernorm.weight') )
rename_keys.append(('sem_seg_head.predictor.transformer.decoder.norm.bias', 'model.transformer_module.decoder.layernorm.bias') )
# heads on top
rename_keys.append(('sem_seg_head.predictor.query_embed.weight', 'model.transformer_module.queries_embedder.weight') )
rename_keys.append(('sem_seg_head.predictor.input_proj.weight', 'model.transformer_module.input_projection.weight') )
rename_keys.append(('sem_seg_head.predictor.input_proj.bias', 'model.transformer_module.input_projection.bias') )
rename_keys.append(('sem_seg_head.predictor.class_embed.weight', 'class_predictor.weight') )
rename_keys.append(('sem_seg_head.predictor.class_embed.bias', 'class_predictor.bias') )
for i in range(3 ):
rename_keys.append((F'sem_seg_head.predictor.mask_embed.layers.{i}.weight', F'mask_embedder.{i}.0.weight') )
rename_keys.append((F'sem_seg_head.predictor.mask_embed.layers.{i}.bias', F'mask_embedder.{i}.0.bias') )
# fmt: on
return rename_keys
def UpperCamelCase_( _snake_case : Any , _snake_case : int , _snake_case : Dict ):
"""simple docstring"""
__a =dct.pop(SCREAMING_SNAKE_CASE__ )
__a =val
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : List[Any] ):
"""simple docstring"""
__a =[int(backbone_config.embed_dim * 2**i ) for i in range(len(backbone_config.depths ) )]
for i in range(len(backbone_config.depths ) ):
__a =num_features[i]
for j in range(backbone_config.depths[i] ):
# fmt: off
# read in weights + bias of input projection layer (in original implementation, this is a single matrix + bias)
__a =state_dict.pop(F'backbone.layers.{i}.blocks.{j}.attn.qkv.weight' )
__a =state_dict.pop(F'backbone.layers.{i}.blocks.{j}.attn.qkv.bias' )
# next, add query, keys and values (in that order) to the state dict
__a =in_proj_weight[:dim, :]
__a =in_proj_bias[: dim]
__a =in_proj_weight[
dim : dim * 2, :
]
__a =in_proj_bias[
dim : dim * 2
]
__a =in_proj_weight[
-dim :, :
]
__a =in_proj_bias[-dim :]
# fmt: on
def UpperCamelCase_( _snake_case : List[str] , _snake_case : List[Any] ):
"""simple docstring"""
__a =config.decoder_config.hidden_size
for idx in range(config.decoder_config.decoder_layers ):
# read in weights + bias of self-attention input projection layer (in the original implementation, this is a single matrix + bias)
__a =state_dict.pop(F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.self_attn.in_proj_weight' )
__a =state_dict.pop(F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.self_attn.in_proj_bias' )
# next, add query, keys and values (in that order) to the state dict
__a =in_proj_weight[: hidden_size, :]
__a =in_proj_bias[:config.hidden_size]
__a =in_proj_weight[hidden_size : hidden_size * 2, :]
__a =in_proj_bias[hidden_size : hidden_size * 2]
__a =in_proj_weight[-hidden_size :, :]
__a =in_proj_bias[-hidden_size :]
# read in weights + bias of cross-attention input projection layer (in the original implementation, this is a single matrix + bias)
__a =state_dict.pop(F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.multihead_attn.in_proj_weight' )
__a =state_dict.pop(F'sem_seg_head.predictor.transformer.decoder.layers.{idx}.multihead_attn.in_proj_bias' )
# next, add query, keys and values (in that order) to the state dict
__a =in_proj_weight[: hidden_size, :]
__a =in_proj_bias[:config.hidden_size]
__a =in_proj_weight[hidden_size : hidden_size * 2, :]
__a =in_proj_bias[hidden_size : hidden_size * 2]
__a =in_proj_weight[-hidden_size :, :]
__a =in_proj_bias[-hidden_size :]
# fmt: on
def UpperCamelCase_( ):
"""simple docstring"""
__a ='http://images.cocodataset.org/val2017/000000039769.jpg'
__a =Image.open(requests.get(SCREAMING_SNAKE_CASE__ , stream=SCREAMING_SNAKE_CASE__ ).raw )
return im
@torch.no_grad()
def UpperCamelCase_( _snake_case : str , _snake_case : str , _snake_case : str , _snake_case : bool = False ):
"""simple docstring"""
__a =get_maskformer_config(SCREAMING_SNAKE_CASE__ )
# load original state_dict
with open(SCREAMING_SNAKE_CASE__ , 'rb' ) as f:
__a =pickle.load(SCREAMING_SNAKE_CASE__ )
__a =data['model']
# for name, param in state_dict.items():
# print(name, param.shape)
# rename keys
__a =create_rename_keys(SCREAMING_SNAKE_CASE__ )
for src, dest in rename_keys:
rename_key(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
read_in_swin_q_k_v(SCREAMING_SNAKE_CASE__ , config.backbone_config )
read_in_decoder_q_k_v(SCREAMING_SNAKE_CASE__ , SCREAMING_SNAKE_CASE__ )
# update to torch tensors
for key, value in state_dict.items():
__a =torch.from_numpy(SCREAMING_SNAKE_CASE__ )
# load 🤗 model
__a =MaskFormerForInstanceSegmentation(SCREAMING_SNAKE_CASE__ )
model.eval()
for name, param in model.named_parameters():
print(SCREAMING_SNAKE_CASE__ , param.shape )
__a , __a =model.load_state_dict(SCREAMING_SNAKE_CASE__ , strict=SCREAMING_SNAKE_CASE__ )
assert missing_keys == [
"model.pixel_level_module.encoder.model.layernorm.weight",
"model.pixel_level_module.encoder.model.layernorm.bias",
]
assert len(SCREAMING_SNAKE_CASE__ ) == 0, F'Unexpected keys: {unexpected_keys}'
# verify results
__a =prepare_img()
if "vistas" in model_name:
__a =65
elif "cityscapes" in model_name:
__a =65535
else:
__a =255
__a =True if 'ade' in model_name else False
__a =MaskFormerImageProcessor(ignore_index=SCREAMING_SNAKE_CASE__ , reduce_labels=SCREAMING_SNAKE_CASE__ )
__a =image_processor(SCREAMING_SNAKE_CASE__ , return_tensors='pt' )
__a =model(**SCREAMING_SNAKE_CASE__ )
print('Logits:' , outputs.class_queries_logits[0, :3, :3] )
if model_name == "maskformer-swin-tiny-ade":
__a =torch.tensor(
[[3.6_353, -4.4_770, -2.6_065], [0.5_081, -4.2_394, -3.5_343], [2.1_909, -5.0_353, -1.9_323]] )
assert torch.allclose(outputs.class_queries_logits[0, :3, :3] , SCREAMING_SNAKE_CASE__ , atol=1e-4 )
print('Looks ok!' )
if pytorch_dump_folder_path is not None:
print(F'Saving model and image processor to {pytorch_dump_folder_path}' )
Path(SCREAMING_SNAKE_CASE__ ).mkdir(exist_ok=SCREAMING_SNAKE_CASE__ )
model.save_pretrained(SCREAMING_SNAKE_CASE__ )
image_processor.save_pretrained(SCREAMING_SNAKE_CASE__ )
if push_to_hub:
print('Pushing model and image processor to the hub...' )
model.push_to_hub(F'nielsr/{model_name}' )
image_processor.push_to_hub(F'nielsr/{model_name}' )
if __name__ == "__main__":
_lowerCAmelCase : str = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_name",
default="maskformer-swin-tiny-ade",
type=str,
help=("Name of the MaskFormer model you\'d like to convert",),
)
parser.add_argument(
"--checkpoint_path",
default="/Users/nielsrogge/Documents/MaskFormer_checkpoints/MaskFormer-Swin-tiny-ADE20k/model.pkl",
type=str,
help="Path to the original state dict (.pth file).",
)
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 or not to push the converted model to the 🤗 hub."
)
_lowerCAmelCase : Dict = parser.parse_args()
convert_maskformer_checkpoint(
args.model_name, args.checkpoint_path, args.pytorch_dump_folder_path, args.push_to_hub
)
| 367
|
import flax.linen as nn
import jax.numpy as jnp
from .attention_flax import FlaxTransformeraDModel
from .resnet_flax import FlaxDownsampleaD, FlaxResnetBlockaD, FlaxUpsampleaD
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> Optional[Any]:
'''simple docstring'''
__a =()
for resnet, attn in zip(self.resnets , self.attentions ):
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=__snake_case , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_downsample:
__a =FlaxDownsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case=True ) -> Optional[int]:
'''simple docstring'''
__a =()
for resnet in self.resnets:
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
output_states += (hidden_states,)
if self.add_downsample:
__a =self.downsamplers_a(__snake_case )
output_states += (hidden_states,)
return hidden_states, output_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
__a =[]
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =FlaxTransformeraDModel(
in_channels=self.out_channels , n_heads=self.num_attention_heads , d_head=self.out_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , only_cross_attention=self.only_cross_attention , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =resnets
__a =attentions
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet, attn in zip(self.resnets , self.attentions ):
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = True
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =[]
for i in range(self.num_layers ):
__a =self.in_channels if (i == self.num_layers - 1) else self.out_channels
__a =self.prev_output_channel if i == 0 else self.out_channels
__a =FlaxResnetBlockaD(
in_channels=resnet_in_channels + res_skip_channels , out_channels=self.out_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
if self.add_upsample:
__a =FlaxUpsampleaD(self.out_channels , dtype=self.dtype )
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[Any]:
'''simple docstring'''
for resnet in self.resnets:
# pop res hidden states
__a =res_hidden_states_tuple[-1]
__a =res_hidden_states_tuple[:-1]
__a =jnp.concatenate((hidden_states, res_hidden_states) , axis=-1 )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
if self.add_upsample:
__a =self.upsamplers_a(__snake_case )
return hidden_states
class __magic_name__ ( nn.Module ):
SCREAMING_SNAKE_CASE = 42
SCREAMING_SNAKE_CASE = 0.0
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = 1
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = False
SCREAMING_SNAKE_CASE = jnp.floataa
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
# there is always at least one resnet
__a =[
FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
]
__a =[]
for _ in range(self.num_layers ):
__a =FlaxTransformeraDModel(
in_channels=self.in_channels , n_heads=self.num_attention_heads , d_head=self.in_channels // self.num_attention_heads , depth=1 , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , )
attentions.append(__snake_case )
__a =FlaxResnetBlockaD(
in_channels=self.in_channels , out_channels=self.in_channels , dropout_prob=self.dropout , dtype=self.dtype , )
resnets.append(__snake_case )
__a =resnets
__a =attentions
def __call__( self , __snake_case , __snake_case , __snake_case , __snake_case=True ) -> List[str]:
'''simple docstring'''
__a =self.resnets[0](__snake_case , __snake_case )
for attn, resnet in zip(self.attentions , self.resnets[1:] ):
__a =attn(__snake_case , __snake_case , deterministic=__snake_case )
__a =resnet(__snake_case , __snake_case , deterministic=__snake_case )
return hidden_states
| 308
| 0
|
from __future__ import annotations
def UpperCamelCase_( _snake_case : int , _snake_case : int ):
"""simple docstring"""
__a =[]
create_all_state(1 , _snake_case , _snake_case , [] , _snake_case )
return result
def UpperCamelCase_( _snake_case : int , _snake_case : int , _snake_case : int , _snake_case : list[int] , _snake_case : list[list[int]] , ):
"""simple docstring"""
if level == 0:
total_list.append(current_list[:] )
return
for i in range(_snake_case , total_number - level + 2 ):
current_list.append(_snake_case )
create_all_state(i + 1 , _snake_case , level - 1 , _snake_case , _snake_case )
current_list.pop()
def UpperCamelCase_( _snake_case : list[list[int]] ):
"""simple docstring"""
for i in total_list:
print(*_snake_case )
if __name__ == "__main__":
_lowerCAmelCase : Optional[Any] = 4
_lowerCAmelCase : Tuple = 2
_lowerCAmelCase : Optional[Any] = generate_all_combinations(n, k)
print_all_state(total_list)
| 368
|
import json
import os
import subprocess
import unittest
from ast import literal_eval
import pytest
from parameterized import parameterized, parameterized_class
from . import is_sagemaker_available
if is_sagemaker_available():
from sagemaker import Session, TrainingJobAnalytics
from sagemaker.huggingface import HuggingFace
@pytest.mark.skipif(
literal_eval(os.getenv('TEST_SAGEMAKER' , 'False' ) ) is not True , reason='Skipping test because should only be run when releasing minor transformers version' , )
@pytest.mark.usefixtures('sm_env' )
@parameterized_class(
[
{
'framework': 'pytorch',
'script': 'run_glue.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_5_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'pytorch',
'script': 'run_ddp.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.7, 'eval_loss': 0.6},
},
{
'framework': 'tensorflow',
'script': 'run_tf_dist.py',
'model_name_or_path': 'distilbert-base-cased',
'instance_type': 'ml.p3.16xlarge',
'results': {'train_runtime': 6_0_0, 'eval_accuracy': 0.6, 'eval_loss': 0.7},
},
] )
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> int:
'''simple docstring'''
if self.framework == "pytorch":
subprocess.run(
f'cp ./examples/pytorch/text-classification/run_glue.py {self.env.test_path}/run_glue.py'.split() , encoding='utf-8' , check=__snake_case , )
assert hasattr(self , 'env' )
def __magic_name__ ( self , __snake_case ) -> int:
'''simple docstring'''
__a =f'{self.env.base_job_name}-{instance_count}-{"ddp" if "ddp" in self.script else "smd"}'
# distributed data settings
__a ={'smdistributed': {'dataparallel': {'enabled': True}}} if self.script != 'run_ddp.py' else None
# creates estimator
return HuggingFace(
entry_point=self.script , source_dir=self.env.test_path , role=self.env.role , image_uri=self.env.image_uri , base_job_name=__snake_case , instance_count=__snake_case , instance_type=self.instance_type , debugger_hook_config=__snake_case , hyperparameters={**self.env.distributed_hyperparameters, 'model_name_or_path': self.model_name_or_path} , metric_definitions=self.env.metric_definitions , distribution=__snake_case , py_version='py36' , )
def __magic_name__ ( self , __snake_case ) -> Optional[Any]:
'''simple docstring'''
TrainingJobAnalytics(__snake_case ).export_csv(f'{self.env.test_path}/{job_name}_metrics.csv' )
@parameterized.expand([(2,)] )
def __magic_name__ ( self , __snake_case ) -> Optional[int]:
'''simple docstring'''
# create estimator
__a =self.create_estimator(__snake_case )
# run training
estimator.fit()
# result dataframe
__a =TrainingJobAnalytics(estimator.latest_training_job.name ).dataframe()
# extract kpis
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_accuracy']['value'] )
__a =list(result_metrics_df[result_metrics_df.metric_name == 'eval_loss']['value'] )
# get train time from SageMaker job, this includes starting, preprocessing, stopping
__a =(
Session().describe_training_job(estimator.latest_training_job.name ).get('TrainingTimeInSeconds' , 99_9999 )
)
# assert kpis
assert train_runtime <= self.results["train_runtime"]
assert all(t >= self.results['eval_accuracy'] for t in eval_accuracy )
assert all(t <= self.results['eval_loss'] for t in eval_loss )
# dump tests result into json file to share in PR
with open(f'{estimator.latest_training_job.name}.json' , 'w' ) as outfile:
json.dump({'train_time': train_runtime, 'eval_accuracy': eval_accuracy, 'eval_loss': eval_loss} , __snake_case )
| 308
| 0
|
import gc
import unittest
import numpy as np
import torch
from transformers import CLIPTextConfig, CLIPTextModelWithProjection, CLIPTokenizer
from diffusers import HeunDiscreteScheduler, PriorTransformer, ShapEPipeline
from diffusers.pipelines.shap_e import ShapERenderer
from diffusers.utils import load_numpy, slow
from diffusers.utils.testing_utils import require_torch_gpu, torch_device
from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference
class __magic_name__ ( lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = ShapEPipeline
SCREAMING_SNAKE_CASE = ['prompt']
SCREAMING_SNAKE_CASE = ['prompt']
SCREAMING_SNAKE_CASE = [
'num_images_per_prompt',
'num_inference_steps',
'generator',
'latents',
'guidance_scale',
'frame_size',
'output_type',
'return_dict',
]
SCREAMING_SNAKE_CASE = False
@property
def __magic_name__ ( self ) -> int:
'''simple docstring'''
return 32
@property
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
return 32
@property
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
return self.time_input_dim * 4
@property
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
return 8
@property
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =CLIPTokenizer.from_pretrained('hf-internal-testing/tiny-random-clip' )
return tokenizer
@property
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
torch.manual_seed(0 )
__a =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=37 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , )
return CLIPTextModelWithProjection(lowercase_ )
@property
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
torch.manual_seed(0 )
__a ={
'num_attention_heads': 2,
'attention_head_dim': 16,
'embedding_dim': self.time_input_dim,
'num_embeddings': 32,
'embedding_proj_dim': self.text_embedder_hidden_size,
'time_embed_dim': self.time_embed_dim,
'num_layers': 1,
'clip_embed_dim': self.time_input_dim * 2,
'additional_embeddings': 0,
'time_embed_act_fn': 'gelu',
'norm_in_type': 'layer',
'encoder_hid_proj_type': None,
'added_emb_type': None,
}
__a =PriorTransformer(**lowercase_ )
return model
@property
def __magic_name__ ( self ) -> str:
'''simple docstring'''
torch.manual_seed(0 )
__a ={
'param_shapes': (
(self.renderer_dim, 93),
(self.renderer_dim, 8),
(self.renderer_dim, 8),
(self.renderer_dim, 8),
),
'd_latent': self.time_input_dim,
'd_hidden': self.renderer_dim,
'n_output': 12,
'background': (
0.1,
0.1,
0.1,
),
}
__a =ShapERenderer(**lowercase_ )
return model
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
__a =self.dummy_prior
__a =self.dummy_text_encoder
__a =self.dummy_tokenizer
__a =self.dummy_renderer
__a =HeunDiscreteScheduler(
beta_schedule='exp' , num_train_timesteps=1024 , prediction_type='sample' , use_karras_sigmas=lowercase_ , clip_sample=lowercase_ , clip_sample_range=1.0 , )
__a ={
'prior': prior,
'text_encoder': text_encoder,
'tokenizer': tokenizer,
'renderer': renderer,
'scheduler': scheduler,
}
return components
def __magic_name__ ( self , __snake_case , __snake_case=0 ) -> Any:
'''simple docstring'''
if str(lowercase_ ).startswith('mps' ):
__a =torch.manual_seed(lowercase_ )
else:
__a =torch.Generator(device=lowercase_ ).manual_seed(lowercase_ )
__a ={
'prompt': 'horse',
'generator': generator,
'num_inference_steps': 1,
'frame_size': 32,
'output_type': 'np',
}
return inputs
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a ='cpu'
__a =self.get_dummy_components()
__a =self.pipeline_class(**lowercase_ )
__a =pipe.to(lowercase_ )
pipe.set_progress_bar_config(disable=lowercase_ )
__a =pipe(**self.get_dummy_inputs(lowercase_ ) )
__a =output.images[0]
__a =image[0, -3:, -3:, -1]
assert image.shape == (20, 32, 32, 3)
__a =np.array(
[
0.0003_9216,
0.0003_9216,
0.0003_9216,
0.0003_9216,
0.0003_9216,
0.0003_9216,
0.0003_9216,
0.0003_9216,
0.0003_9216,
] )
assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
self._test_inference_batch_consistent(batch_sizes=[1, 2] )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
__a =torch_device == 'cpu'
__a =True
self._test_inference_batch_single_identical(
batch_size=2 , test_max_difference=lowercase_ , relax_max_difference=lowercase_ , )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.get_dummy_components()
__a =self.pipeline_class(**lowercase_ )
__a =pipe.to(lowercase_ )
pipe.set_progress_bar_config(disable=lowercase_ )
__a =1
__a =2
__a =self.get_dummy_inputs(lowercase_ )
for key in inputs.keys():
if key in self.batch_params:
__a =batch_size * [inputs[key]]
__a =pipe(**lowercase_ , num_images_per_prompt=lowercase_ )[0]
assert images.shape[0] == batch_size * num_images_per_prompt
@slow
@require_torch_gpu
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
super().tearDown()
gc.collect()
torch.cuda.empty_cache()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main'
'/shap_e/test_shap_e_np_out.npy' )
__a =ShapEPipeline.from_pretrained('openai/shap-e' )
__a =pipe.to(lowercase_ )
pipe.set_progress_bar_config(disable=lowercase_ )
__a =torch.Generator(device=lowercase_ ).manual_seed(0 )
__a =pipe(
'a shark' , generator=lowercase_ , guidance_scale=15.0 , num_inference_steps=64 , frame_size=64 , output_type='np' , ).images[0]
assert images.shape == (20, 64, 64, 3)
assert_mean_pixel_difference(lowercase_ , lowercase_ )
| 369
|
import argparse
import logging
from collections import namedtuple
import torch
from model_bertabs import BertAbsSummarizer
from models.model_builder import AbsSummarizer # The authors' implementation
from transformers import BertTokenizer
logging.basicConfig(level=logging.INFO)
_lowerCAmelCase : List[Any] = logging.getLogger(__name__)
_lowerCAmelCase : Optional[Any] = "Hello world! cécé herlolip"
_lowerCAmelCase : str = namedtuple(
"BertAbsConfig",
[
"temp_dir",
"large",
"use_bert_emb",
"finetune_bert",
"encoder",
"share_emb",
"max_pos",
"enc_layers",
"enc_hidden_size",
"enc_heads",
"enc_ff_size",
"enc_dropout",
"dec_layers",
"dec_hidden_size",
"dec_heads",
"dec_ff_size",
"dec_dropout",
],
)
def UpperCamelCase_( _snake_case : str , _snake_case : List[Any] ):
"""simple docstring"""
__a =BertAbsConfig(
temp_dir='.' , finetune_bert=_snake_case , large=_snake_case , share_emb=_snake_case , use_bert_emb=_snake_case , encoder='bert' , max_pos=512 , enc_layers=6 , enc_hidden_size=512 , enc_heads=8 , enc_ff_size=512 , enc_dropout=0.2 , dec_layers=6 , dec_hidden_size=768 , dec_heads=8 , dec_ff_size=2048 , dec_dropout=0.2 , )
__a =torch.load(_snake_case , lambda _snake_case , _snake_case : storage )
__a =AbsSummarizer(_snake_case , torch.device('cpu' ) , _snake_case )
original.eval()
__a =BertAbsSummarizer(_snake_case , torch.device('cpu' ) )
new_model.eval()
# -------------------
# Convert the weights
# -------------------
logging.info('convert the model' )
new_model.bert.load_state_dict(original.bert.state_dict() )
new_model.decoder.load_state_dict(original.decoder.state_dict() )
new_model.generator.load_state_dict(original.generator.state_dict() )
# ----------------------------------
# Make sure the outpus are identical
# ----------------------------------
logging.info('Make sure that the models\' outputs are identical' )
__a =BertTokenizer.from_pretrained('bert-base-uncased' )
# prepare the model inputs
__a =tokenizer.encode('This is sample éàalj\'-.' )
encoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
__a =tokenizer.encode('This is sample 3 éàalj\'-.' )
decoder_input_ids.extend([tokenizer.pad_token_id] * (512 - len(_snake_case )) )
__a =torch.tensor(_snake_case ).unsqueeze(0 )
# failsafe to make sure the weights reset does not affect the
# loaded weights.
assert torch.max(torch.abs(original.generator[0].weight - new_model.generator[0].weight ) ) == 0
# forward pass
__a =encoder_input_ids
__a =decoder_input_ids
__a =__a =None
__a =None
__a =__a =None
__a =__a =None
__a =None
# The original model does not apply the geneator layer immediatly but rather in
# the beam search (where it combines softmax + linear layer). Since we already
# apply the softmax in our generation process we only apply the linear layer here.
# We make sure that the outputs of the full stack are identical
__a =original(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =original.generator(_snake_case )
__a =new_model(
_snake_case , _snake_case , _snake_case , _snake_case , _snake_case )[0]
__a =new_model.generator(_snake_case )
__a =torch.max(torch.abs(output_converted_model - output_original_model ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.max(torch.abs(output_converted_generator - output_original_generator ) ).item()
print('Maximum absolute difference beween weights: {:.2f}'.format(_snake_case ) )
__a =torch.allclose(_snake_case , _snake_case , atol=1e-3 )
if are_identical:
logging.info('all weights are equal up to 1e-3' )
else:
raise ValueError('the weights are different. The new model is likely different from the original one.' )
# The model has been saved with torch.save(model) and this is bound to the exact
# directory structure. We save the state_dict instead.
logging.info('saving the model\'s state dictionary' )
torch.save(
new_model.state_dict() , './bertabs-finetuned-cnndm-extractive-abstractive-summarization/pytorch_model.bin' )
if __name__ == "__main__":
_lowerCAmelCase : Optional[int] = argparse.ArgumentParser()
parser.add_argument(
"--bertabs_checkpoint_path",
default=None,
type=str,
required=True,
help="Path the official PyTorch dump.",
)
parser.add_argument(
"--pytorch_dump_folder_path",
default=None,
type=str,
required=True,
help="Path to the output PyTorch model.",
)
_lowerCAmelCase : Optional[Any] = parser.parse_args()
convert_bertabs_checkpoints(
args.bertabs_checkpoint_path,
args.pytorch_dump_folder_path,
)
| 308
| 0
|
from typing import TYPE_CHECKING
from ...utils import _LazyModule
_lowerCAmelCase : Union[str, Any] = {"processing_wav2vec2_with_lm": ["Wav2Vec2ProcessorWithLM"]}
if TYPE_CHECKING:
from .processing_wavaveca_with_lm import WavaVecaProcessorWithLM
else:
import sys
_lowerCAmelCase : Dict = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 370
|
import unittest
from transformers import XLMConfig, is_torch_available
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 (
XLMForMultipleChoice,
XLMForQuestionAnswering,
XLMForQuestionAnsweringSimple,
XLMForSequenceClassification,
XLMForTokenClassification,
XLMModel,
XLMWithLMHeadModel,
)
from transformers.models.xlm.modeling_xlm import XLM_PRETRAINED_MODEL_ARCHIVE_LIST
class __magic_name__ :
def __init__( self , __snake_case , __snake_case=13 , __snake_case=7 , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=True , __snake_case=False , __snake_case=False , __snake_case=False , __snake_case=2 , __snake_case=99 , __snake_case=0 , __snake_case=32 , __snake_case=5 , __snake_case=4 , __snake_case=0.1 , __snake_case=0.1 , __snake_case=512 , __snake_case=2 , __snake_case=0.02 , __snake_case=2 , __snake_case=4 , __snake_case="last" , __snake_case=True , __snake_case=None , __snake_case=0 , ) -> Optional[Any]:
'''simple docstring'''
__a =parent
__a =batch_size
__a =seq_length
__a =is_training
__a =use_input_lengths
__a =use_token_type_ids
__a =use_labels
__a =gelu_activation
__a =sinusoidal_embeddings
__a =causal
__a =asm
__a =n_langs
__a =vocab_size
__a =n_special
__a =hidden_size
__a =num_hidden_layers
__a =num_attention_heads
__a =hidden_dropout_prob
__a =attention_probs_dropout_prob
__a =max_position_embeddings
__a =type_sequence_label_size
__a =initializer_range
__a =num_labels
__a =num_choices
__a =summary_type
__a =use_proj
__a =scope
__a =bos_token_id
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
__a =random_attention_mask([self.batch_size, self.seq_length] )
__a =None
if self.use_input_lengths:
__a =(
ids_tensor([self.batch_size] , vocab_size=2 ) + self.seq_length - 2
) # small variation of seq_length
__a =None
if self.use_token_type_ids:
__a =ids_tensor([self.batch_size, self.seq_length] , self.n_langs )
__a =None
__a =None
__a =None
if self.use_labels:
__a =ids_tensor([self.batch_size] , self.type_sequence_label_size )
__a =ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
__a =ids_tensor([self.batch_size] , 2 ).float()
__a =ids_tensor([self.batch_size] , self.num_choices )
__a =self.get_config()
return (
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
)
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return XLMConfig(
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 , num_labels=self.num_labels , bos_token_id=self.bos_token_id , )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMModel(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , lengths=__snake_case , langs=__snake_case )
__a =model(__snake_case , langs=__snake_case )
__a =model(__snake_case )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[int]:
'''simple docstring'''
__a =XLMWithLMHeadModel(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , token_type_ids=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Dict:
'''simple docstring'''
__a =XLMForQuestionAnsweringSimple(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
__a =outputs
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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> List[Any]:
'''simple docstring'''
__a =XLMForQuestionAnswering(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , p_mask=__snake_case , )
__a =model(
__snake_case , start_positions=__snake_case , end_positions=__snake_case , cls_index=__snake_case , is_impossible=__snake_case , )
((__a) , ) =result_with_labels.to_tuple()
__a =model(__snake_case , start_positions=__snake_case , end_positions=__snake_case )
((__a) , ) =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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Optional[Any]:
'''simple docstring'''
__a =XLMForSequenceClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case )
__a =model(__snake_case , labels=__snake_case )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Any:
'''simple docstring'''
__a =self.num_labels
__a =XLMForTokenClassification(__snake_case )
model.to(__snake_case )
model.eval()
__a =model(__snake_case , attention_mask=__snake_case , labels=__snake_case )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , ) -> Tuple:
'''simple docstring'''
__a =self.num_choices
__a =XLMForMultipleChoice(config=__snake_case )
model.to(__snake_case )
model.eval()
__a =input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
__a =model(
__snake_case , attention_mask=__snake_case , token_type_ids=__snake_case , labels=__snake_case , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =self.prepare_config_and_inputs()
(
(
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) , (
__a
) ,
) =config_and_inputs
__a ={'input_ids': input_ids, 'token_type_ids': token_type_ids, 'lengths': input_lengths}
return config, inputs_dict
@require_torch
class __magic_name__ ( lowerCAmelCase_ , lowerCAmelCase_ , lowerCAmelCase_ , unittest.TestCase ):
SCREAMING_SNAKE_CASE = (
(
XLMModel,
XLMWithLMHeadModel,
XLMForQuestionAnswering,
XLMForSequenceClassification,
XLMForQuestionAnsweringSimple,
XLMForTokenClassification,
XLMForMultipleChoice,
)
if is_torch_available()
else ()
)
SCREAMING_SNAKE_CASE = (
(XLMWithLMHeadModel,) if is_torch_available() else ()
) # TODO (PVP): Check other models whether language generation is also applicable
SCREAMING_SNAKE_CASE = (
{
'feature-extraction': XLMModel,
'fill-mask': XLMWithLMHeadModel,
'question-answering': XLMForQuestionAnsweringSimple,
'text-classification': XLMForSequenceClassification,
'text-generation': XLMWithLMHeadModel,
'token-classification': XLMForTokenClassification,
'zero-shot': XLMForSequenceClassification,
}
if is_torch_available()
else {}
)
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case ) -> int:
'''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 __magic_name__ ( self , __snake_case , __snake_case , __snake_case=False ) -> str:
'''simple docstring'''
__a =super()._prepare_for_class(__snake_case , __snake_case , return_labels=__snake_case )
if return_labels:
if model_class.__name__ == "XLMForQuestionAnswering":
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
__a =torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=__snake_case )
return inputs_dict
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMModelTester(self )
__a =ConfigTester(self , config_class=__snake_case , emb_dim=37 )
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self.config_tester.run_common_tests()
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_model(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_lm_head(*__snake_case )
def __magic_name__ ( self ) -> List[str]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_simple_qa(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_qa(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_sequence_classif(*__snake_case )
def __magic_name__ ( self ) -> Optional[int]:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_token_classif(*__snake_case )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
__a =self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_for_multiple_choice(*__snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Optional[Any]:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_attentions in attentions] , [True] * len(__snake_case ) )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_attentions in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =min_length + idx + 1
__a =(
batch_size * num_beam_groups,
config.num_attention_heads,
tgt_len,
src_len,
)
# check attn size
self.assertListEqual(
[layer_attention.shape for layer_attention in iter_attentions] , [expected_shape] * len(__snake_case ) )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case , __snake_case=False , __snake_case=1 ) -> Dict:
'''simple docstring'''
self.assertIsInstance(__snake_case , __snake_case )
self.assertListEqual(
[isinstance(__snake_case , __snake_case ) for iter_hidden_states in hidden_states] , [True] * len(__snake_case ) , )
self.assertEqual(len(__snake_case ) , (max_length - min_length) * num_beam_groups )
for idx, iter_hidden_states in enumerate(__snake_case ):
# adds PAD dummy token
__a =min_length + idx + 1
__a =(batch_size * num_beam_groups, seq_len, config.hidden_size)
# check hidden size
self.assertListEqual(
[layer_hidden_states.shape for layer_hidden_states in iter_hidden_states] , [expected_shape] * len(__snake_case ) , )
pass
@slow
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
for model_name in XLM_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
__a =XLMModel.from_pretrained(__snake_case )
self.assertIsNotNone(__snake_case )
@require_torch
class __magic_name__ ( unittest.TestCase ):
@slow
def __magic_name__ ( self ) -> Tuple:
'''simple docstring'''
__a =XLMWithLMHeadModel.from_pretrained('xlm-mlm-en-2048' )
model.to(__snake_case )
__a =torch.tensor([[14, 447]] , dtype=torch.long , device=__snake_case ) # the president
__a =[
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
] # the president the president the president the president the president the president the president the president the president the president
# TODO(PVP): this and other input_ids I tried for generation give pretty bad results. Not sure why. Model might just not be made for auto-regressive inference
__a =model.generate(__snake_case , do_sample=__snake_case )
self.assertListEqual(output_ids[0].cpu().numpy().tolist() , __snake_case )
| 308
| 0
|
"""simple docstring"""
import math
def UpperCamelCase_( ):
"""simple docstring"""
__a =input('Enter message: ' )
__a =int(input(F'Enter key [2-{len(_snake_case ) - 1}]: ' ) )
__a =input('Encryption/Decryption [e/d]: ' )
if mode.lower().startswith('e' ):
__a =encrypt_message(_snake_case , _snake_case )
elif mode.lower().startswith('d' ):
__a =decrypt_message(_snake_case , _snake_case )
# Append pipe symbol (vertical bar) to identify spaces at the end.
print(F'Output:\n{text + "|"}' )
def UpperCamelCase_( _snake_case : Tuple , _snake_case : List[Any] ):
"""simple docstring"""
__a =[''''''] * key
for col in range(_snake_case ):
__a =col
while pointer < len(_snake_case ):
cipher_text[col] += message[pointer]
pointer += key
return "".join(_snake_case )
def UpperCamelCase_( _snake_case : Tuple , _snake_case : Any ):
"""simple docstring"""
__a =math.ceil(len(_snake_case ) / key )
__a =key
__a =(num_cols * num_rows) - len(_snake_case )
__a =[''''''] * num_cols
__a =0
__a =0
for symbol in message:
plain_text[col] += symbol
col += 1
if (
(col == num_cols)
or (col == num_cols - 1)
and (row >= num_rows - num_shaded_boxes)
):
__a =0
row += 1
return "".join(_snake_case )
if __name__ == "__main__":
import doctest
doctest.testmod()
main()
| 371
|
from __future__ import annotations
import matplotlib.pyplot as plt # type: ignore
import numpy
# initial triangle of Koch snowflake
_lowerCAmelCase : Optional[Any] = numpy.array([0, 0])
_lowerCAmelCase : Dict = numpy.array([0.5, 0.8660254])
_lowerCAmelCase : Any = numpy.array([1, 0])
_lowerCAmelCase : int = [VECTOR_1, VECTOR_2, VECTOR_3, VECTOR_1]
def UpperCamelCase_( _snake_case : list[numpy.ndarray] , _snake_case : int ):
"""simple docstring"""
__a =initial_vectors
for _ in range(_snake_case ):
__a =iteration_step(_snake_case )
return vectors
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =[]
for i, start_vector in enumerate(vectors[:-1] ):
__a =vectors[i + 1]
new_vectors.append(_snake_case )
__a =end_vector - start_vector
new_vectors.append(start_vector + difference_vector / 3 )
new_vectors.append(
start_vector + difference_vector / 3 + rotate(difference_vector / 3 , 60 ) )
new_vectors.append(start_vector + difference_vector * 2 / 3 )
new_vectors.append(vectors[-1] )
return new_vectors
def UpperCamelCase_( _snake_case : numpy.ndarray , _snake_case : float ):
"""simple docstring"""
__a =numpy.radians(_snake_case )
__a , __a =numpy.cos(_snake_case ), numpy.sin(_snake_case )
__a =numpy.array(((c, -s), (s, c)) )
return numpy.dot(_snake_case , _snake_case )
def UpperCamelCase_( _snake_case : list[numpy.ndarray] ):
"""simple docstring"""
__a =plt.gca()
axes.set_aspect('equal' )
# matplotlib.pyplot.plot takes a list of all x-coordinates and a list of all
# y-coordinates as inputs, which are constructed from the vector-list using
# zip()
__a , __a =zip(*_snake_case )
plt.plot(_snake_case , _snake_case )
plt.show()
if __name__ == "__main__":
import doctest
doctest.testmod()
_lowerCAmelCase : List[Any] = iterate(INITIAL_VECTORS, 5)
plot(processed_vectors)
| 308
| 0
|
_lowerCAmelCase : Tuple = "Tobias Carryer"
from time import time
class __magic_name__ :
def __init__( self , __snake_case , __snake_case , __snake_case , __snake_case=int(time() ) ) -> Optional[int]: # noqa: B008
'''simple docstring'''
__a =multiplier
__a =increment
__a =modulo
__a =seed
def __magic_name__ ( self ) -> Union[str, Any]:
'''simple docstring'''
__a =(self.multiplier * self.seed + self.increment) % self.modulo
return self.seed
if __name__ == "__main__":
# Show the LCG in action.
_lowerCAmelCase : List[Any] = LinearCongruentialGenerator(1_664_525, 1_013_904_223, 2 << 31)
while True:
print(lcg.next_number())
| 350
|
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
_lowerCAmelCase : Tuple = {
"configuration_megatron_bert": ["MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MegatronBertConfig"],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
_lowerCAmelCase : List[str] = [
"MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"MegatronBertForCausalLM",
"MegatronBertForMaskedLM",
"MegatronBertForMultipleChoice",
"MegatronBertForNextSentencePrediction",
"MegatronBertForPreTraining",
"MegatronBertForQuestionAnswering",
"MegatronBertForSequenceClassification",
"MegatronBertForTokenClassification",
"MegatronBertModel",
"MegatronBertPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_megatron_bert import MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MegatronBertConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_megatron_bert import (
MEGATRON_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
MegatronBertForCausalLM,
MegatronBertForMaskedLM,
MegatronBertForMultipleChoice,
MegatronBertForNextSentencePrediction,
MegatronBertForPreTraining,
MegatronBertForQuestionAnswering,
MegatronBertForSequenceClassification,
MegatronBertForTokenClassification,
MegatronBertModel,
MegatronBertPreTrainedModel,
)
else:
import sys
_lowerCAmelCase : Optional[int] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
| 308
| 0
|
import gc
import random
import unittest
import torch
from diffusers import (
IFImgaImgPipeline,
IFImgaImgSuperResolutionPipeline,
IFInpaintingPipeline,
IFInpaintingSuperResolutionPipeline,
IFPipeline,
IFSuperResolutionPipeline,
)
from diffusers.models.attention_processor import AttnAddedKVProcessor
from diffusers.utils.import_utils import is_xformers_available
from diffusers.utils.testing_utils import floats_tensor, load_numpy, require_torch_gpu, skip_mps, slow, torch_device
from ..pipeline_params import TEXT_TO_IMAGE_BATCH_PARAMS, TEXT_TO_IMAGE_PARAMS
from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference
from . import IFPipelineTesterMixin
@skip_mps
class __magic_name__ ( __SCREAMING_SNAKE_CASE , __SCREAMING_SNAKE_CASE , unittest.TestCase ):
SCREAMING_SNAKE_CASE = IFPipeline
SCREAMING_SNAKE_CASE = TEXT_TO_IMAGE_PARAMS - {"width", "height", "latents"}
SCREAMING_SNAKE_CASE = TEXT_TO_IMAGE_BATCH_PARAMS
SCREAMING_SNAKE_CASE = PipelineTesterMixin.required_optional_params - {"latents"}
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
return self._get_dummy_components()
def __magic_name__ ( self , __snake_case , __snake_case=0 ) -> str:
'''simple docstring'''
if str(_snake_case ).startswith('mps' ):
__a =torch.manual_seed(_snake_case )
else:
__a =torch.Generator(device=_snake_case ).manual_seed(_snake_case )
__a ={
"prompt": "A painting of a squirrel eating a burger",
"generator": generator,
"num_inference_steps": 2,
"output_type": "numpy",
}
return inputs
def __magic_name__ ( self ) -> str:
'''simple docstring'''
self._test_save_load_optional_components()
@unittest.skipIf(torch_device != 'cuda' , reason='float16 requires CUDA' )
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
# Due to non-determinism in save load of the hf-internal-testing/tiny-random-t5 text encoder
super().test_save_load_floataa(expected_max_diff=1e-1 )
def __magic_name__ ( self ) -> int:
'''simple docstring'''
self._test_attention_slicing_forward_pass(expected_max_diff=1e-2 )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
self._test_save_load_local()
def __magic_name__ ( self ) -> List[Any]:
'''simple docstring'''
self._test_inference_batch_single_identical(
expected_max_diff=1e-2 , )
@unittest.skipIf(
torch_device != 'cuda' or not is_xformers_available() , reason='XFormers attention is only available with CUDA and `xformers` installed' , )
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
self._test_xformers_attention_forwardGenerator_pass(expected_max_diff=1e-3 )
@slow
@require_torch_gpu
class __magic_name__ ( unittest.TestCase ):
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
# clean up the VRAM after each test
super().tearDown()
gc.collect()
torch.cuda.empty_cache()
def __magic_name__ ( self ) -> Dict:
'''simple docstring'''
# if
__a =IFPipeline.from_pretrained('DeepFloyd/IF-I-XL-v1.0' , variant='fp16' , torch_dtype=torch.floataa )
__a =IFSuperResolutionPipeline.from_pretrained(
'DeepFloyd/IF-II-L-v1.0' , variant='fp16' , torch_dtype=torch.floataa , text_encoder=_snake_case , tokenizer=_snake_case )
# pre compute text embeddings and remove T5 to save memory
pipe_a.text_encoder.to('cuda' )
__a =pipe_a.encode_prompt('anime turtle' , device='cuda' )
del pipe_a.tokenizer
del pipe_a.text_encoder
gc.collect()
__a =None
__a =None
pipe_a.enable_model_cpu_offload()
pipe_a.enable_model_cpu_offload()
pipe_a.unet.set_attn_processor(AttnAddedKVProcessor() )
pipe_a.unet.set_attn_processor(AttnAddedKVProcessor() )
self._test_if(_snake_case , _snake_case , _snake_case , _snake_case )
pipe_a.remove_all_hooks()
pipe_a.remove_all_hooks()
# img2img
__a =IFImgaImgPipeline(**pipe_a.components )
__a =IFImgaImgSuperResolutionPipeline(**pipe_a.components )
pipe_a.enable_model_cpu_offload()
pipe_a.enable_model_cpu_offload()
pipe_a.unet.set_attn_processor(AttnAddedKVProcessor() )
pipe_a.unet.set_attn_processor(AttnAddedKVProcessor() )
self._test_if_imgaimg(_snake_case , _snake_case , _snake_case , _snake_case )
pipe_a.remove_all_hooks()
pipe_a.remove_all_hooks()
# inpainting
__a =IFInpaintingPipeline(**pipe_a.components )
__a =IFInpaintingSuperResolutionPipeline(**pipe_a.components )
pipe_a.enable_model_cpu_offload()
pipe_a.enable_model_cpu_offload()
pipe_a.unet.set_attn_processor(AttnAddedKVProcessor() )
pipe_a.unet.set_attn_processor(AttnAddedKVProcessor() )
self._test_if_inpainting(_snake_case , _snake_case , _snake_case , _snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case ) -> int:
'''simple docstring'''
# pipeline 1
_start_torch_memory_measurement()
__a =torch.Generator(device='cpu' ).manual_seed(0 )
__a =pipe_a(
prompt_embeds=_snake_case , negative_prompt_embeds=_snake_case , num_inference_steps=2 , generator=_snake_case , output_type='np' , )
__a =output.images[0]
assert image.shape == (64, 64, 3)
__a =torch.cuda.max_memory_allocated()
assert mem_bytes < 13 * 10**9
__a =load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/if/test_if.npy' )
assert_mean_pixel_difference(_snake_case , _snake_case )
# pipeline 2
_start_torch_memory_measurement()
__a =torch.Generator(device='cpu' ).manual_seed(0 )
__a =floats_tensor((1, 3, 64, 64) , rng=random.Random(0 ) ).to(_snake_case )
__a =pipe_a(
prompt_embeds=_snake_case , negative_prompt_embeds=_snake_case , image=_snake_case , generator=_snake_case , num_inference_steps=2 , output_type='np' , )
__a =output.images[0]
assert image.shape == (256, 256, 3)
__a =torch.cuda.max_memory_allocated()
assert mem_bytes < 4 * 10**9
__a =load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/if/test_if_superresolution_stage_II.npy' )
assert_mean_pixel_difference(_snake_case , _snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case ) -> Optional[int]:
'''simple docstring'''
# pipeline 1
_start_torch_memory_measurement()
__a =floats_tensor((1, 3, 64, 64) , rng=random.Random(0 ) ).to(_snake_case )
__a =torch.Generator(device='cpu' ).manual_seed(0 )
__a =pipe_a(
prompt_embeds=_snake_case , negative_prompt_embeds=_snake_case , image=_snake_case , num_inference_steps=2 , generator=_snake_case , output_type='np' , )
__a =output.images[0]
assert image.shape == (64, 64, 3)
__a =torch.cuda.max_memory_allocated()
assert mem_bytes < 10 * 10**9
__a =load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/if/test_if_img2img.npy' )
assert_mean_pixel_difference(_snake_case , _snake_case )
# pipeline 2
_start_torch_memory_measurement()
__a =torch.Generator(device='cpu' ).manual_seed(0 )
__a =floats_tensor((1, 3, 256, 256) , rng=random.Random(0 ) ).to(_snake_case )
__a =floats_tensor((1, 3, 64, 64) , rng=random.Random(0 ) ).to(_snake_case )
__a =pipe_a(
prompt_embeds=_snake_case , negative_prompt_embeds=_snake_case , image=_snake_case , original_image=_snake_case , generator=_snake_case , num_inference_steps=2 , output_type='np' , )
__a =output.images[0]
assert image.shape == (256, 256, 3)
__a =torch.cuda.max_memory_allocated()
assert mem_bytes < 4 * 10**9
__a =load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/if/test_if_img2img_superresolution_stage_II.npy' )
assert_mean_pixel_difference(_snake_case , _snake_case )
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case , __snake_case ) -> Optional[Any]:
'''simple docstring'''
# pipeline 1
_start_torch_memory_measurement()
__a =floats_tensor((1, 3, 64, 64) , rng=random.Random(0 ) ).to(_snake_case )
__a =floats_tensor((1, 3, 64, 64) , rng=random.Random(1 ) ).to(_snake_case )
__a =torch.Generator(device='cpu' ).manual_seed(0 )
__a =pipe_a(
prompt_embeds=_snake_case , negative_prompt_embeds=_snake_case , image=_snake_case , mask_image=_snake_case , num_inference_steps=2 , generator=_snake_case , output_type='np' , )
__a =output.images[0]
assert image.shape == (64, 64, 3)
__a =torch.cuda.max_memory_allocated()
assert mem_bytes < 10 * 10**9
__a =load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/if/test_if_inpainting.npy' )
assert_mean_pixel_difference(_snake_case , _snake_case )
# pipeline 2
_start_torch_memory_measurement()
__a =torch.Generator(device='cpu' ).manual_seed(0 )
__a =floats_tensor((1, 3, 64, 64) , rng=random.Random(0 ) ).to(_snake_case )
__a =floats_tensor((1, 3, 256, 256) , rng=random.Random(0 ) ).to(_snake_case )
__a =floats_tensor((1, 3, 256, 256) , rng=random.Random(1 ) ).to(_snake_case )
__a =pipe_a(
prompt_embeds=_snake_case , negative_prompt_embeds=_snake_case , image=_snake_case , mask_image=_snake_case , original_image=_snake_case , generator=_snake_case , num_inference_steps=2 , output_type='np' , )
__a =output.images[0]
assert image.shape == (256, 256, 3)
__a =torch.cuda.max_memory_allocated()
assert mem_bytes < 4 * 10**9
__a =load_numpy(
'https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/if/test_if_inpainting_superresolution_stage_II.npy' )
assert_mean_pixel_difference(_snake_case , _snake_case )
def UpperCamelCase_( ):
"""simple docstring"""
torch.cuda.empty_cache()
torch.cuda.reset_max_memory_allocated()
torch.cuda.reset_peak_memory_stats()
| 351
|
import hashlib
import unittest
from transformers import MODEL_FOR_DEPTH_ESTIMATION_MAPPING, is_torch_available, is_vision_available
from transformers.pipelines import DepthEstimationPipeline, pipeline
from transformers.testing_utils import (
is_pipeline_test,
nested_simplify,
require_tf,
require_timm,
require_torch,
require_vision,
slow,
)
from .test_pipelines_common import ANY
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
else:
class __magic_name__ :
@staticmethod
def __magic_name__ ( *__snake_case , **__snake_case ) -> List[str]:
'''simple docstring'''
pass
def UpperCamelCase_( _snake_case : Image ):
"""simple docstring"""
__a =hashlib.mda(image.tobytes() )
return m.hexdigest()
@is_pipeline_test
@require_vision
@require_timm
@require_torch
class __magic_name__ ( unittest.TestCase ):
SCREAMING_SNAKE_CASE = MODEL_FOR_DEPTH_ESTIMATION_MAPPING
def __magic_name__ ( self , __snake_case , __snake_case , __snake_case ) -> Dict:
'''simple docstring'''
__a =DepthEstimationPipeline(model=__snake_case , image_processor=__snake_case )
return depth_estimator, [
"./tests/fixtures/tests_samples/COCO/000000039769.png",
"./tests/fixtures/tests_samples/COCO/000000039769.png",
]
def __magic_name__ ( self , __snake_case , __snake_case ) -> Tuple:
'''simple docstring'''
__a =depth_estimator('./tests/fixtures/tests_samples/COCO/000000039769.png' )
self.assertEqual({'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )} , __snake_case )
import datasets
__a =datasets.load_dataset('hf-internal-testing/fixtures_image_utils' , 'image' , split='test' )
__a =depth_estimator(
[
Image.open('./tests/fixtures/tests_samples/COCO/000000039769.png' ),
'http://images.cocodataset.org/val2017/000000039769.jpg',
# RGBA
dataset[0]['file'],
# LA
dataset[1]['file'],
# L
dataset[2]['file'],
] )
self.assertEqual(
[
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
{'predicted_depth': ANY(torch.Tensor ), 'depth': ANY(Image.Image )},
] , __snake_case , )
@require_tf
@unittest.skip('Depth estimation is not implemented in TF' )
def __magic_name__ ( self ) -> Optional[Any]:
'''simple docstring'''
pass
@slow
@require_torch
def __magic_name__ ( self ) -> int:
'''simple docstring'''
__a ='Intel/dpt-large'
__a =pipeline('depth-estimation' , model=__snake_case )
__a =depth_estimator('http://images.cocodataset.org/val2017/000000039769.jpg' )
__a =hashimage(outputs['depth'] )
# This seems flaky.
# self.assertEqual(outputs["depth"], "1a39394e282e9f3b0741a90b9f108977")
self.assertEqual(nested_simplify(outputs['predicted_depth'].max().item() ) , 29.304 )
self.assertEqual(nested_simplify(outputs['predicted_depth'].min().item() ) , 2.662 )
@require_torch
def __magic_name__ ( self ) -> Any:
'''simple docstring'''
# This is highly irregular to have no small tests.
self.skipTest('There is not hf-internal-testing tiny model for either GLPN nor DPT' )
| 308
| 0
|
import logging
import torch
from accelerate import Accelerator
from arguments import EvaluationArguments
from datasets import load_dataset
from torch.utils.data import IterableDataset
from torch.utils.data.dataloader import DataLoader
from transformers import AutoModelForCausalLM, AutoTokenizer, HfArgumentParser, set_seed
class __magic_name__ ( lowerCAmelCase_ ):
def __init__( self , __snake_case , __snake_case , __snake_case=1024 , __snake_case=1024 , __snake_case=3.6 ) -> Optional[Any]:
'''simple docstring'''
__a =tokenizer
__a =tokenizer.bos_token_id
__a =dataset
__a =seq_length
__a =seq_length * chars_per_token * num_of_sequences
def __iter__( self ) -> Tuple:
'''simple docstring'''
__a =iter(self.dataset )
__a =True
while more_examples:
__a =[], 0
while True:
if buffer_len >= self.input_characters:
break
try:
buffer.append(next(__snake_case )['content'] )
buffer_len += len(buffer[-1] )
except StopIteration:
__a =False
break
__a =tokenizer(__snake_case , truncation=__snake_case )['''input_ids''']
__a =[]
for tokenized_input in tokenized_inputs:
all_token_ids.extend(tokenized_input + [self.concat_token_id] )
for i in range(0 , len(__snake_case ) , self.seq_length ):
__a =all_token_ids[i : i + self.seq_length]
if len(__snake_case ) == self.seq_length:
yield torch.tensor(__snake_case )
def UpperCamelCase_( _snake_case : Dict ):
"""simple docstring"""
__a ={'''streaming''': True}
__a =load_dataset(args.dataset_name , split='train' , **UpperCamelCase__ )
__a =ConstantLengthDataset(UpperCamelCase__ , UpperCamelCase__ , seq_length=args.seq_length )
__a =DataLoader(UpperCamelCase__ , batch_size=args.batch_size )
return eval_dataloader
def UpperCamelCase_( _snake_case : Optional[Any] ):
"""simple docstring"""
model.eval()
__a =[]
for step, batch in enumerate(UpperCamelCase__ ):
with torch.no_grad():
__a =model(UpperCamelCase__ , labels=UpperCamelCase__ )
__a =outputs.loss.repeat(args.batch_size )
losses.append(accelerator.gather(UpperCamelCase__ ) )
if args.max_eval_steps > 0 and step >= args.max_eval_steps:
break
__a =torch.mean(torch.cat(UpperCamelCase__ ) )
try:
__a =torch.exp(UpperCamelCase__ )
except OverflowError:
__a =float('inf' )
return loss.item(), perplexity.item()
# Setup Accelerator
_lowerCAmelCase : str = Accelerator()
# Parse configuration
_lowerCAmelCase : List[Any] = HfArgumentParser(EvaluationArguments)
_lowerCAmelCase : Any = parser.parse_args()
set_seed(args.seed)
# Logging
_lowerCAmelCase : Union[str, Any] = logging.getLogger(__name__)
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO
)
# Load model and tokenizer
_lowerCAmelCase : Any = AutoModelForCausalLM.from_pretrained(args.model_ckpt)
_lowerCAmelCase : List[str] = AutoTokenizer.from_pretrained(args.model_ckpt)
# Load dataset and dataloader
_lowerCAmelCase : Optional[Any] = create_dataloader(args)
# Prepare everything with our `accelerator`.
_lowerCAmelCase : Any = accelerator.prepare(model, eval_dataloader)
# Evaluate and save the last checkpoint
logger.info("Evaluating and saving model after training")
_lowerCAmelCase : List[Any] = evaluate(args)
logger.info(f'''loss/eval: {eval_loss}, perplexity: {perplexity}''')
| 352
|
import argparse
import dataclasses
import json
import logging
import os
import shutil
from typing import List, Optional
import datasets
from accelerate import Accelerator
from datasets import load_dataset
from finetuning import finetune
from tqdm.auto import tqdm
import transformers
from transformers import AutoConfig, set_seed
from transformers.trainer_utils import IntervalStrategy
_lowerCAmelCase : Optional[int] = logging.getLogger(__name__)
_lowerCAmelCase : Any = "pytorch_model.bin"
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co.'} , )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the training data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(metadata={'help': 'A csv or a json file containing the data to predict on.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'A csv or a json file containing the validation data.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The name of the task to train on.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'The list of labels for the task.'} )
@dataclasses.dataclass
class __magic_name__ :
SCREAMING_SNAKE_CASE = dataclasses.field(
metadata={'help': 'The output directory where the model predictions and checkpoints will be written.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='accuracy' , metadata={'help': 'The evaluation metric used for the task.'} )
SCREAMING_SNAKE_CASE = dataclasses.field(
default='no' , metadata={
'help': 'The evaluation strategy to adopt during training. Possible values are: ["no", "step", "epoch]'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={
'help': 'How much the specified evaluation metric must improve to satisfy early stopping conditions.'
} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the confidence score.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to filter the pseudo-labeled data based on the validation performance.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Whether to fine-tune on labeled data after pseudo training.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=0.0 , metadata={'help': 'Confidence threshold for pseudo-labeled data filtering.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=1_0_0 , metadata={'help': 'Number of evaluation calls with no improvement after which training will be stopped.'} , )
SCREAMING_SNAKE_CASE = dataclasses.field(
default=lowerCAmelCase_ , metadata={'help': 'Random seed for initialization.'} , )
def UpperCamelCase_( _snake_case : int , _snake_case : str , _snake_case : Optional[int] , _snake_case : str , _snake_case : Union[str, Any] , _snake_case : List[Any] ):
"""simple docstring"""
__a =datasets.concatenate_datasets([infer_input, infer_output] , axis=1 )
if args.do_filter_by_confidence:
__a =dataset.filter(lambda _snake_case : example["probability"] > args.confidence_threshold )
if args.do_filter_by_val_performance:
assert eval_result >= 0.0 and eval_result <= 1.0
__a =int(eval_result * len(_snake_case ) )
print(_snake_case )
__a =dataset.sort('probability' , reverse=_snake_case )
__a =dataset.select(range(_snake_case ) )
__a =dataset.remove_columns(['label', 'probability'] )
__a =dataset.rename_column('prediction' , 'label' )
__a =dataset.map(lambda _snake_case : {"label": idalabel[example["label"]]} )
__a =dataset.shuffle(seed=args.seed )
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.data_file_extension == "csv":
dataset.to_csv(_snake_case , index=_snake_case )
else:
dataset.to_json(_snake_case )
def UpperCamelCase_( _snake_case : List[Any] , _snake_case : str , _snake_case : int , _snake_case : Optional[int] , **_snake_case : List[str] ):
"""simple docstring"""
__a =Accelerator()
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s' , datefmt='%m/%d/%Y %H:%M:%S' , level=logging.INFO , )
logger.info(accelerator.state )
# Setup logging, we only want one process per machine to log things on the
# screen. accelerator.is_local_main_process is only True for one process per
# machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR )
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
__a =STModelArguments(model_name_or_path=_snake_case )
__a =STDataArguments(train_file=_snake_case , infer_file=_snake_case )
__a =STTrainingArguments(output_dir=_snake_case )
__a =argparse.Namespace()
for arg_class in (model_args, data_args, training_args):
for key, value in vars(_snake_case ).items():
setattr(_snake_case , _snake_case , _snake_case )
for key, value in kwargs.items():
if hasattr(_snake_case , _snake_case ):
setattr(_snake_case , _snake_case , _snake_case )
# Sanity checks
__a ={}
__a =None
# You need to provide the training data and the data to predict on
assert args.train_file is not None
assert args.infer_file is not None
__a =args.train_file
__a =args.infer_file
if args.evaluation_strategy != IntervalStrategy.NO.value:
assert args.eval_file is not None
__a =args.eval_file
for key in data_files:
__a =data_files[key].split('.' )[-1]
assert extension in ["csv", "json"], F'`{key}_file` should be a csv or a json file.'
if args.data_file_extension is None:
__a =extension
else:
assert extension == args.data_file_extension, F'`{key}_file` should be a {args.data_file_extension} file`.'
assert (
args.eval_metric in datasets.list_metrics()
), F'{args.eval_metric} not in the list of supported metrics {datasets.list_metrics()}.'
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed )
logger.info('Creating the initial data directory for self-training...' )
__a =F'{args.output_dir}/self-train_iter-{{}}'.format
__a =data_dir_format(0 )
if accelerator.is_main_process:
if args.output_dir is not None:
os.makedirs(args.output_dir , exist_ok=_snake_case )
os.makedirs(_snake_case , exist_ok=_snake_case )
accelerator.wait_for_everyone()
__a =None
__a =None
__a =0
__a =False
# Show the progress bar
__a =tqdm(range(args.max_selftrain_iterations ) , disable=not accelerator.is_local_main_process )
# Self-train
for iteration in range(0 , int(args.max_selftrain_iterations ) ):
__a =data_dir_format(_snake_case )
assert os.path.exists(_snake_case )
# Stage 1: initial fine-tuning for iteration = 0 or pseudo-training for
# iteration > 0
__a =os.path.join(_snake_case , 'stage-1' )
__a ={
'accelerator': accelerator,
'model_name_or_path': args.model_name_or_path,
'cache_dir': args.cache_dir,
'do_train': True,
'train_file': data_files['train'] if iteration == 0 else data_files['train_pseudo'],
'do_eval': True if args.eval_file is not None else False,
'eval_file': data_files['eval'],
'do_predict': True,
'infer_file': data_files['infer'],
'task_name': args.task_name,
'label_list': args.label_list,
'output_dir': current_output_dir,
'eval_metric': args.eval_metric,
'evaluation_strategy': args.evaluation_strategy,
'early_stopping_patience': args.early_stopping_patience,
'early_stopping_threshold': args.early_stopping_threshold,
'seed': args.seed,
}
# Add additional training arguments
for key, value in kwargs.items():
if key not in arguments_dict and not hasattr(_snake_case , _snake_case ):
arguments_dict.update({key: value} )
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 1.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 1 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 1.' , _snake_case )
if iteration > 0 and args.finetune_on_labeled_data:
# Stage 2 (optional): fine-tuning on the original labeled data
__a =os.path.join(_snake_case , 'best-checkpoint' )
__a =os.path.join(_snake_case , 'stage-2' )
# Update arguments_dict
__a =model_path
__a =data_files['train']
__a =current_output_dir
__a =os.path.join(_snake_case , 'best-checkpoint' , _snake_case )
if os.path.exists(_snake_case ):
logger.info(
'Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 2.' , _snake_case , _snake_case , )
else:
logger.info('***** Running self-training: iteration: %d, stage: 2 *****' , _snake_case )
finetune(**_snake_case )
accelerator.wait_for_everyone()
assert os.path.exists(_snake_case )
logger.info('Self-training job completed: iteration: %d, stage: 2.' , _snake_case )
__a =iteration
__a =data_dir_format(iteration + 1 )
__a =AutoConfig.from_pretrained(os.path.join(_snake_case , 'best-checkpoint' ) )
__a =config.idalabel
__a =os.path.join(_snake_case , 'eval_results_best-checkpoint.json' )
__a =os.path.join(_snake_case , 'test_results_best-checkpoint.json' )
assert os.path.exists(_snake_case )
with open(_snake_case , 'r' ) as f:
__a =float(json.load(_snake_case )[args.eval_metric] )
__a =os.path.join(_snake_case , 'infer_output_best-checkpoint.csv' )
assert os.path.exists(_snake_case )
# Loading the dataset from local csv or json files.
__a =load_dataset(args.data_file_extension , data_files={'data': data_files['infer']} )['data']
__a =load_dataset('csv' , data_files={'data': infer_output_file} )['data']
if accelerator.is_main_process:
os.makedirs(_snake_case , exist_ok=_snake_case )
shutil.copy(_snake_case , os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) )
if os.path.exists(_snake_case ):
shutil.copy(_snake_case , os.path.join(_snake_case , F'test_results_iter-{iteration}.json' ) )
create_pseudo_labeled_data(_snake_case , _snake_case , _snake_case , _snake_case , _snake_case , _snake_case )
accelerator.wait_for_everyone()
__a =os.path.join(_snake_case , F'train_pseudo.{args.data_file_extension}' )
if args.evaluation_strategy != IntervalStrategy.NO.value:
__a =eval_result
if best_iteration is None:
__a =new_iteration
__a =new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
__a =new_iteration
__a =new_eval_result
__a =0
else:
if new_eval_result == best_eval_result:
__a =new_iteration
__a =new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
__a =True
progress_bar.update(1 )
if should_training_stop:
break
if best_iteration is not None:
# Save the best iteration
logger.info('Best iteration: %d' , _snake_case )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{iteration}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
else:
# Assume that the last iteration is the best
logger.info('Best iteration: %d' , args.max_selftrain_iterations - 1 )
logger.info('Best evaluation result: %s = %f' , args.eval_metric , _snake_case )
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(_snake_case , F'eval_results_iter-{args.max_selftrain_iterations - 1}.json' ) , os.path.join(_snake_case , 'eval_results_best-iteration.json' ) , )
| 308
| 0
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.