code
stringlengths
87
55.2k
code_codestyle
int64
0
349
style_context
stringlengths
135
49.1k
style_context_codestyle
int64
0
349
label
int64
0
1
import warnings from ...utils import is_sklearn_available, requires_backends if is_sklearn_available(): from scipy.stats import pearsonr, spearmanr from sklearn.metrics import fa_score, matthews_corrcoef a__ : Optional[int] = ( '''This metric will be removed from the library soon, metrics should be handled with the 🤗 Evaluate ''' '''library. You can have a look at this example script for pointers: ''' '''https://github.com/huggingface/transformers/blob/main/examples/pytorch/text-classification/run_glue.py''' ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" warnings.warn(a__ , a__ ) requires_backends(a__ , '''sklearn''' ) return (preds == labels).mean() def UpperCAmelCase_( a__ , a__ ): """simple docstring""" warnings.warn(a__ , a__ ) requires_backends(a__ , '''sklearn''' ) SCREAMING_SNAKE_CASE : Tuple = simple_accuracy(a__ , a__ ) SCREAMING_SNAKE_CASE : Tuple = fa_score(y_true=a__ , y_pred=a__ ) return { "acc": acc, "f1": fa, "acc_and_f1": (acc + fa) / 2, } def UpperCAmelCase_( a__ , a__ ): """simple docstring""" warnings.warn(a__ , a__ ) requires_backends(a__ , '''sklearn''' ) SCREAMING_SNAKE_CASE : str = pearsonr(a__ , a__ )[0] SCREAMING_SNAKE_CASE : List[str] = spearmanr(a__ , a__ )[0] return { "pearson": pearson_corr, "spearmanr": spearman_corr, "corr": (pearson_corr + spearman_corr) / 2, } def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" warnings.warn(a__ , a__ ) requires_backends(a__ , '''sklearn''' ) assert len(a__ ) == len(a__ ), F"""Predictions and labels have mismatched lengths {len(a__ )} and {len(a__ )}""" if task_name == "cola": return {"mcc": matthews_corrcoef(a__ , a__ )} elif task_name == "sst-2": return {"acc": simple_accuracy(a__ , a__ )} elif task_name == "mrpc": return acc_and_fa(a__ , a__ ) elif task_name == "sts-b": return pearson_and_spearman(a__ , a__ ) elif task_name == "qqp": return acc_and_fa(a__ , a__ ) elif task_name == "mnli": return {"mnli/acc": simple_accuracy(a__ , a__ )} elif task_name == "mnli-mm": return {"mnli-mm/acc": simple_accuracy(a__ , a__ )} elif task_name == "qnli": return {"acc": simple_accuracy(a__ , a__ )} elif task_name == "rte": return {"acc": simple_accuracy(a__ , a__ )} elif task_name == "wnli": return {"acc": simple_accuracy(a__ , a__ )} elif task_name == "hans": return {"acc": simple_accuracy(a__ , a__ )} else: raise KeyError(a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" warnings.warn(a__ , a__ ) requires_backends(a__ , '''sklearn''' ) if len(a__ ) != len(a__ ): raise ValueError(F"""Predictions and labels have mismatched lengths {len(a__ )} and {len(a__ )}""" ) if task_name == "xnli": return {"acc": simple_accuracy(a__ , a__ )} else: raise KeyError(a__ )
313
from __future__ import annotations import math def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if len(a__ ) != 2 or len(a[0] ) != 2 or len(a__ ) != 2 or len(b[0] ) != 2: raise Exception('''Matrices are not 2x2''' ) SCREAMING_SNAKE_CASE : Dict = [ [a[0][0] * b[0][0] + a[0][1] * b[1][0], a[0][0] * b[0][1] + a[0][1] * b[1][1]], [a[1][0] * b[0][0] + a[1][1] * b[1][0], a[1][0] * b[0][1] + a[1][1] * b[1][1]], ] return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] + matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] - matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) % 2 != 0 or len(a[0] ) % 2 != 0: raise Exception('''Odd matrices are not supported!''' ) SCREAMING_SNAKE_CASE : str = len(a__ ) SCREAMING_SNAKE_CASE : Any = matrix_length // 2 SCREAMING_SNAKE_CASE : Tuple = [[a[i][j] for j in range(a__ , a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : Optional[int] = [ [a[i][j] for j in range(a__ , a__ )] for i in range(a__ , a__ ) ] SCREAMING_SNAKE_CASE : Optional[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : List[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ , a__ )] return top_left, top_right, bot_left, bot_right def UpperCAmelCase_( a__ ): """simple docstring""" return len(a__ ), len(matrix[0] ) def UpperCAmelCase_( a__ ): """simple docstring""" print('''\n'''.join(str(a__ ) for line in matrix ) ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ ) == (2, 2): return default_matrix_multiplication(a__ , a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE : Dict = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : int = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Any = actual_strassen(matrix_addition(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = matrix_addition(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = matrix_subtraction(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) # construct the new matrix from our 4 quadrants SCREAMING_SNAKE_CASE : Optional[Any] = [] for i in range(len(a__ ) ): new_matrix.append(top_left[i] + top_right[i] ) for i in range(len(a__ ) ): new_matrix.append(bot_left[i] + bot_right[i] ) return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ )[1] != matrix_dimensions(a__ )[0]: SCREAMING_SNAKE_CASE : Any = ( '''Unable to multiply these matrices, please check the dimensions.\n''' F"""Matrix A: {matrixa}\n""" F"""Matrix B: {matrixa}""" ) raise Exception(a__ ) SCREAMING_SNAKE_CASE : str = matrix_dimensions(a__ ) SCREAMING_SNAKE_CASE : Tuple = matrix_dimensions(a__ ) if dimensiona[0] == dimensiona[1] and dimensiona[0] == dimensiona[1]: return [matrixa, matrixa] SCREAMING_SNAKE_CASE : str = max(*a__ , *a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = int(math.pow(2 , math.ceil(math.loga(a__ ) ) ) ) SCREAMING_SNAKE_CASE : Optional[int] = matrixa SCREAMING_SNAKE_CASE : Tuple = matrixa # Adding zeros to the matrices so that the arrays dimensions are the same and also # power of 2 for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) SCREAMING_SNAKE_CASE : Optional[Any] = actual_strassen(a__ , a__ ) # Removing the additional zeros for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): final_matrix[i].pop() else: final_matrix.pop() return final_matrix if __name__ == "__main__": a__ : Dict = [ [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 2, 3, 1], ] a__ : Union[str, Any] = [[0, 2, 1, 1], [16, 2, 3, 3], [2, 2, 7, 7], [13, 11, 22, 4]] print(strassen(matrixa, matrixa))
313
1
import torch from transformers import PreTrainedModel, XLMRobertaConfig, XLMRobertaModel class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = 'M-CLIP' def __init__( self , _lowerCamelCase=1024 , _lowerCamelCase=768 , **_lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = transformerDimSize SCREAMING_SNAKE_CASE : Dict = imageDimSize super().__init__(**_lowerCamelCase ) class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Dict = MCLIPConfig def __init__( self , _lowerCamelCase , *_lowerCamelCase , **_lowerCamelCase ) ->Dict: super().__init__(_lowerCamelCase , *_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = XLMRobertaModel(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = torch.nn.Linear( in_features=config.transformerDimensions , out_features=config.numDims ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Dict = self.transformer(input_ids=_lowerCamelCase , attention_mask=_lowerCamelCase )[0] SCREAMING_SNAKE_CASE : str = (embs * attention_mask.unsqueeze(2 )).sum(dim=1 ) / attention_mask.sum(dim=1 )[:, None] return self.LinearTransformation(_lowerCamelCase ), embs
313
# We ignore warnings about stepping the scheduler since we step it ourselves during gradient accumulation import warnings from .state import AcceleratorState, GradientState warnings.filterwarnings('''ignore''', category=UserWarning, module='''torch.optim.lr_scheduler''') class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , _lowerCamelCase = False ) ->Any: SCREAMING_SNAKE_CASE : str = scheduler SCREAMING_SNAKE_CASE : List[str] = optimizers if isinstance(_lowerCamelCase , (list, tuple) ) else [optimizers] SCREAMING_SNAKE_CASE : Union[str, Any] = split_batches SCREAMING_SNAKE_CASE : List[Any] = step_with_optimizer SCREAMING_SNAKE_CASE : List[str] = GradientState() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->Optional[Any]: if not self.step_with_optimizer: # No link between scheduler and optimizer -> just step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) return # Otherwise, first make sure the optimizer was stepped. if not self.gradient_state.sync_gradients: if self.gradient_state.adjust_scheduler: self.scheduler._step_count += 1 return for opt in self.optimizers: if opt.step_was_skipped: return if self.split_batches: # Split batches -> the training dataloader batch size is not changed so one step per training step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: # Otherwise the training dataloader batch size was multiplied by `num_processes`, so we need to do # num_processes steps per training step SCREAMING_SNAKE_CASE : List[str] = AcceleratorState().num_processes for _ in range(_lowerCamelCase ): # Special case when using OneCycle and `drop_last` was not used if hasattr(self.scheduler , '''total_steps''' ): if self.scheduler._step_count <= self.scheduler.total_steps: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Union[str, Any]: return self.scheduler.get_last_lr() def __lowerCAmelCase ( self ) ->List[str]: return self.scheduler.state_dict() def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: self.scheduler.load_state_dict(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: return self.scheduler.get_lr() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->List[str]: return self.scheduler.print_lr(*_lowerCamelCase , **_lowerCamelCase )
313
1
from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_tokenizers_available, is_torch_available, is_vision_available, ) a__ : int = { '''configuration_layoutlmv2''': ['''LAYOUTLMV2_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''LayoutLMv2Config'''], '''processing_layoutlmv2''': ['''LayoutLMv2Processor'''], '''tokenization_layoutlmv2''': ['''LayoutLMv2Tokenizer'''], } try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : List[Any] = ['''LayoutLMv2TokenizerFast'''] try: if not is_vision_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Any = ['''LayoutLMv2FeatureExtractor'''] a__ : Optional[Any] = ['''LayoutLMv2ImageProcessor'''] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : str = [ '''LAYOUTLMV2_PRETRAINED_MODEL_ARCHIVE_LIST''', '''LayoutLMv2ForQuestionAnswering''', '''LayoutLMv2ForSequenceClassification''', '''LayoutLMv2ForTokenClassification''', '''LayoutLMv2Layer''', '''LayoutLMv2Model''', '''LayoutLMv2PreTrainedModel''', ] if TYPE_CHECKING: from .configuration_layoutlmva import LAYOUTLMV2_PRETRAINED_CONFIG_ARCHIVE_MAP, LayoutLMvaConfig from .processing_layoutlmva import LayoutLMvaProcessor from .tokenization_layoutlmva import LayoutLMvaTokenizer try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_layoutlmva_fast import LayoutLMvaTokenizerFast try: if not is_vision_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .feature_extraction_layoutlmva import LayoutLMvaFeatureExtractor, LayoutLMvaImageProcessor try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_layoutlmva import ( LAYOUTLMV2_PRETRAINED_MODEL_ARCHIVE_LIST, LayoutLMvaForQuestionAnswering, LayoutLMvaForSequenceClassification, LayoutLMvaForTokenClassification, LayoutLMvaLayer, LayoutLMvaModel, LayoutLMvaPreTrainedModel, ) else: import sys a__ : List[str] = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
import argparse import os from pathlib import Path from typing import Dict import tensorflow as tf import torch from tqdm import tqdm from transformers import PegasusConfig, PegasusForConditionalGeneration, PegasusTokenizer from transformers.models.pegasus.configuration_pegasus import DEFAULTS, task_specific_params a__ : Optional[Any] = [ # replace left string with right string to get the relevant state_dict key (identical state dict to bart) ['''memory_attention''', '''encoder_attn'''], ['''attention''', '''attn'''], ['''/''', '''.'''], ['''.LayerNorm.gamma''', '''_layer_norm.weight'''], ['''.LayerNorm.beta''', '''_layer_norm.bias'''], ['''r.layer_''', '''r.layers.'''], ['''output_proj''', '''out_proj'''], ['''ffn.dense_1.''', '''fc2.'''], ['''ffn.dense.''', '''fc1.'''], ['''ffn_layer_norm''', '''final_layer_norm'''], ['''kernel''', '''weight'''], ['''encoder_layer_norm.''', '''encoder.layer_norm.'''], ['''decoder_layer_norm.''', '''decoder.layer_norm.'''], ['''embeddings.weights''', '''shared.weight'''], ] def UpperCAmelCase_( a__ ): """simple docstring""" for pegasus_name, hf_name in PATTERNS: SCREAMING_SNAKE_CASE : Union[str, Any] = k.replace(a__ , a__ ) return k def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = DEFAULTS.copy() cfg_kwargs.update(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = PegasusConfig(**a__ ) SCREAMING_SNAKE_CASE : Optional[int] = PegasusForConditionalGeneration(a__ ) SCREAMING_SNAKE_CASE : Dict = torch_model.model.state_dict() SCREAMING_SNAKE_CASE : List[str] = {} for k, v in tf_weights.items(): SCREAMING_SNAKE_CASE : int = rename_state_dict_key(a__ ) if new_k not in sd: raise ValueError(F"""could not find new key {new_k} in state dict. (converted from {k})""" ) if "dense" in k or "proj" in new_k: SCREAMING_SNAKE_CASE : Dict = v.T SCREAMING_SNAKE_CASE : Tuple = torch.tensor(a__ , dtype=sd[new_k].dtype ) assert v.shape == sd[new_k].shape, F"""{new_k}, {k}, {v.shape}, {sd[new_k].shape}""" # make sure embedding.padding_idx is respected SCREAMING_SNAKE_CASE : Tuple = torch.zeros_like(mapping['''shared.weight'''][cfg.pad_token_id + 1] ) SCREAMING_SNAKE_CASE : int = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Union[str, Any] = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Optional[Any] = {k: torch.zeros_like(a__ ) for k, v in sd.items() if k.endswith('''bias''' ) and k not in mapping} mapping.update(**a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = torch_model.model.load_state_dict(a__ , strict=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = [ k for k in missing if k not in ['''encoder.embed_positions.weight''', '''decoder.embed_positions.weight'''] ] assert unexpected_missing == [], F"""no matches found for the following torch keys {unexpected_missing}""" assert extra == [], F"""no matches found for the following tf keys {extra}""" return torch_model def UpperCAmelCase_( a__="./ckpt/aeslc/model.ckpt-32000" ): """simple docstring""" SCREAMING_SNAKE_CASE : str = tf.train.list_variables(a__ ) SCREAMING_SNAKE_CASE : str = {} SCREAMING_SNAKE_CASE : List[Any] = ['''Adafactor''', '''global_step'''] for name, shape in tqdm(a__ , desc='''converting tf checkpoint to dict''' ): SCREAMING_SNAKE_CASE : Union[str, Any] = any(pat in name for pat in ignore_name ) if skip_key: continue SCREAMING_SNAKE_CASE : Dict = tf.train.load_variable(a__ , a__ ) SCREAMING_SNAKE_CASE : Any = array return tf_weights def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = Path(a__ ).parent.name SCREAMING_SNAKE_CASE : Union[str, Any] = task_specific_params[F"""summarization_{dataset}"""]['''max_position_embeddings'''] SCREAMING_SNAKE_CASE : Dict = PegasusTokenizer.from_pretrained('''sshleifer/pegasus''' , model_max_length=a__ ) assert tok.model_max_length == desired_max_model_length tok.save_pretrained(a__ ) # convert model SCREAMING_SNAKE_CASE : Any = get_tf_weights_as_numpy(a__ ) SCREAMING_SNAKE_CASE : List[str] = task_specific_params[F"""summarization_{dataset}"""] if dataset == "large": SCREAMING_SNAKE_CASE : int = task_specific_params SCREAMING_SNAKE_CASE : List[str] = convert_pegasus(a__ , a__ ) torch_model.save_pretrained(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch_model.state_dict() sd.pop('''model.decoder.embed_positions.weight''' ) sd.pop('''model.encoder.embed_positions.weight''' ) torch.save(a__ , Path(a__ ) / '''pytorch_model.bin''' ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument('''tf_ckpt_path''', type=str, help='''passed to tf.train.list_variables''') parser.add_argument('''save_dir''', default=None, type=str, help='''Path to the output PyTorch model.''') a__ : List[str] = parser.parse_args() if args.save_dir is None: a__ : Any = Path(args.tf_ckpt_path).parent.name a__ : int = os.path.join('''pegasus''', dataset) convert_pegasus_ckpt_to_pytorch(args.tf_ckpt_path, args.save_dir)
313
1
import csv import tweepy # Twitter API credentials a__ : Union[str, Any] = '''''' a__ : List[str] = '''''' a__ : Any = '''''' a__ : List[str] = '''''' def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = tweepy.OAuthHandler(a__ , a__ ) auth.set_access_token(a__ , a__ ) SCREAMING_SNAKE_CASE : List[str] = tweepy.API(a__ ) # initialize a list to hold all the tweepy Tweets SCREAMING_SNAKE_CASE : Any = [] # make initial request for most recent tweets (200 is the maximum allowed count) SCREAMING_SNAKE_CASE : List[Any] = api.user_timeline(screen_name=a__ , count=200 ) # save most recent tweets alltweets.extend(a__ ) # save the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Tuple = alltweets[-1].id - 1 # keep grabbing tweets until there are no tweets left to grab while len(a__ ) > 0: print(F"""getting tweets before {oldest}""" ) # all subsequent requests use the max_id param to prevent duplicates SCREAMING_SNAKE_CASE : Any = api.user_timeline( screen_name=a__ , count=200 , max_id=a__ ) # save most recent tweets alltweets.extend(a__ ) # update the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Dict = alltweets[-1].id - 1 print(F"""...{len(a__ )} tweets downloaded so far""" ) # transform the tweepy tweets into a 2D array that will populate the csv SCREAMING_SNAKE_CASE : Optional[Any] = [[tweet.id_str, tweet.created_at, tweet.text] for tweet in alltweets] # write the csv with open(F"""new_{screen_name}_tweets.csv""" , '''w''' ) as f: SCREAMING_SNAKE_CASE : List[Any] = csv.writer(a__ ) writer.writerow(['''id''', '''created_at''', '''text'''] ) writer.writerows(a__ ) if __name__ == "__main__": # pass in the username of the account you want to download get_all_tweets('''FirePing32''')
313
import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import XLMRobertaTokenizerFast from diffusers import DDIMScheduler, KandinskyImgaImgPipeline, KandinskyPriorPipeline, UNetaDConditionModel, VQModel from diffusers.pipelines.kandinsky.text_encoder import MCLIPConfig, MultilingualCLIP from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = KandinskyImgaImgPipeline __SCREAMING_SNAKE_CASE : str = ['prompt', 'image_embeds', 'negative_image_embeds', 'image'] __SCREAMING_SNAKE_CASE : int = [ 'prompt', 'negative_prompt', 'image_embeds', 'negative_image_embeds', 'image', ] __SCREAMING_SNAKE_CASE : int = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'negative_prompt', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->int: return 32 @property def __lowerCAmelCase ( self ) ->List[str]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Tuple: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 100 @property def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = XLMRobertaTokenizerFast.from_pretrained('''YiYiXu/tiny-random-mclip-base''' ) return tokenizer @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[Any] = MCLIPConfig( numDims=self.cross_attention_dim , transformerDimensions=self.text_embedder_hidden_size , hidden_size=self.text_embedder_hidden_size , intermediate_size=37 , num_attention_heads=4 , num_hidden_layers=5 , vocab_size=1005 , ) SCREAMING_SNAKE_CASE : Dict = MultilingualCLIP(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = text_encoder.eval() return text_encoder @property def __lowerCAmelCase ( self ) ->Union[str, Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = { '''in_channels''': 4, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''text_image''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''text_image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : Tuple = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->List[str]: return { "block_out_channels": [32, 64], "down_block_types": ["DownEncoderBlock2D", "AttnDownEncoderBlock2D"], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": [ "AttnUpDecoderBlock2D", "UpDecoderBlock2D", ], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Optional[Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.dummy_text_encoder SCREAMING_SNAKE_CASE : Any = self.dummy_tokenizer SCREAMING_SNAKE_CASE : List[Any] = self.dummy_unet SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_movq SCREAMING_SNAKE_CASE : Optional[Any] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : Optional[Any] = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = { '''text_encoder''': text_encoder, '''tokenizer''': tokenizer, '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->str: SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(seed + 1 ) ).to(_lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : str = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : str = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : str = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = { '''prompt''': '''horse''', '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : List[str] = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : List[Any] = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : List[Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Optional[int] = np.array( [0.6_1_4_7_4_9_4_3, 0.6_0_7_3_5_3_9, 0.4_3_3_0_8_5_4_4, 0.5_9_2_8_2_6_9, 0.4_7_4_9_3_5_9_5, 0.4_6_7_5_5_9_7_3, 0.4_6_1_3_8_3_8, 0.4_5_3_6_8_7_9_7, 0.5_0_1_1_9_2_3_3] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/kandinsky_img2img_frog.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : str = '''A red cartoon frog, 4k''' SCREAMING_SNAKE_CASE : Any = KandinskyPriorPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = KandinskyImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : str = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=5 , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : Dict = pipeline( _lowerCamelCase , image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=768 , width=768 , strength=0.2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Tuple = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_tensorflow_text_available, is_tf_available, is_tokenizers_available, is_torch_available, ) a__ : List[str] = { '''configuration_bert''': ['''BERT_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''BertConfig''', '''BertOnnxConfig'''], '''tokenization_bert''': ['''BasicTokenizer''', '''BertTokenizer''', '''WordpieceTokenizer'''], } try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Optional[Any] = ['''BertTokenizerFast'''] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Dict = [ '''BERT_PRETRAINED_MODEL_ARCHIVE_LIST''', '''BertForMaskedLM''', '''BertForMultipleChoice''', '''BertForNextSentencePrediction''', '''BertForPreTraining''', '''BertForQuestionAnswering''', '''BertForSequenceClassification''', '''BertForTokenClassification''', '''BertLayer''', '''BertLMHeadModel''', '''BertModel''', '''BertPreTrainedModel''', '''load_tf_weights_in_bert''', ] try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Any = [ '''TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST''', '''TFBertEmbeddings''', '''TFBertForMaskedLM''', '''TFBertForMultipleChoice''', '''TFBertForNextSentencePrediction''', '''TFBertForPreTraining''', '''TFBertForQuestionAnswering''', '''TFBertForSequenceClassification''', '''TFBertForTokenClassification''', '''TFBertLMHeadModel''', '''TFBertMainLayer''', '''TFBertModel''', '''TFBertPreTrainedModel''', ] try: if not is_tensorflow_text_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : int = ['''TFBertTokenizer'''] try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Optional[int] = [ '''FlaxBertForCausalLM''', '''FlaxBertForMaskedLM''', '''FlaxBertForMultipleChoice''', '''FlaxBertForNextSentencePrediction''', '''FlaxBertForPreTraining''', '''FlaxBertForQuestionAnswering''', '''FlaxBertForSequenceClassification''', '''FlaxBertForTokenClassification''', '''FlaxBertModel''', '''FlaxBertPreTrainedModel''', ] if TYPE_CHECKING: from .configuration_bert import BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, BertConfig, BertOnnxConfig from .tokenization_bert import BasicTokenizer, BertTokenizer, WordpieceTokenizer try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_bert_fast import BertTokenizerFast try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_bert import ( BERT_PRETRAINED_MODEL_ARCHIVE_LIST, BertForMaskedLM, BertForMultipleChoice, BertForNextSentencePrediction, BertForPreTraining, BertForQuestionAnswering, BertForSequenceClassification, BertForTokenClassification, BertLayer, BertLMHeadModel, BertModel, BertPreTrainedModel, load_tf_weights_in_bert, ) try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_tf_bert import ( TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST, TFBertEmbeddings, TFBertForMaskedLM, TFBertForMultipleChoice, TFBertForNextSentencePrediction, TFBertForPreTraining, TFBertForQuestionAnswering, TFBertForSequenceClassification, TFBertForTokenClassification, TFBertLMHeadModel, TFBertMainLayer, TFBertModel, TFBertPreTrainedModel, ) try: if not is_tensorflow_text_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_bert_tf import TFBertTokenizer try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_flax_bert import ( FlaxBertForCausalLM, FlaxBertForMaskedLM, FlaxBertForMultipleChoice, FlaxBertForNextSentencePrediction, FlaxBertForPreTraining, FlaxBertForQuestionAnswering, FlaxBertForSequenceClassification, FlaxBertForTokenClassification, FlaxBertModel, FlaxBertPreTrainedModel, ) else: import sys a__ : Tuple = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
import argparse import json import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ViTImageProcessor, ViTMSNConfig, ViTMSNModel from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD torch.set_grad_enabled(False) def UpperCAmelCase_( a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [] for i in range(config.num_hidden_layers ): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append((F"""module.blocks.{i}.norm1.weight""", F"""vit.encoder.layer.{i}.layernorm_before.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm1.bias""", F"""vit.encoder.layer.{i}.layernorm_before.bias""") ) rename_keys.append( (F"""module.blocks.{i}.attn.proj.weight""", F"""vit.encoder.layer.{i}.attention.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.attn.proj.bias""", F"""vit.encoder.layer.{i}.attention.output.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.norm2.weight""", F"""vit.encoder.layer.{i}.layernorm_after.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm2.bias""", F"""vit.encoder.layer.{i}.layernorm_after.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.weight""", F"""vit.encoder.layer.{i}.intermediate.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.bias""", F"""vit.encoder.layer.{i}.intermediate.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.weight""", F"""vit.encoder.layer.{i}.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.bias""", F"""vit.encoder.layer.{i}.output.dense.bias""") ) # projection layer + position embeddings rename_keys.extend( [ ('''module.cls_token''', '''vit.embeddings.cls_token'''), ('''module.patch_embed.proj.weight''', '''vit.embeddings.patch_embeddings.projection.weight'''), ('''module.patch_embed.proj.bias''', '''vit.embeddings.patch_embeddings.projection.bias'''), ('''module.pos_embed''', '''vit.embeddings.position_embeddings'''), ] ) if base_model: # layernorm + pooler rename_keys.extend( [ ('''module.norm.weight''', '''layernorm.weight'''), ('''module.norm.bias''', '''layernorm.bias'''), ] ) # if just the base model, we should remove "vit" from all keys that start with "vit" SCREAMING_SNAKE_CASE : Any = [(pair[0], pair[1][4:]) if pair[1].startswith('''vit''' ) else pair for pair in rename_keys] else: # layernorm + classification head rename_keys.extend( [ ('''norm.weight''', '''vit.layernorm.weight'''), ('''norm.bias''', '''vit.layernorm.bias'''), ('''head.weight''', '''classifier.weight'''), ('''head.bias''', '''classifier.bias'''), ] ) return rename_keys def UpperCAmelCase_( a__ , a__ , a__=False ): """simple docstring""" for i in range(config.num_hidden_layers ): if base_model: SCREAMING_SNAKE_CASE : Any = '''''' else: SCREAMING_SNAKE_CASE : Optional[int] = '''vit.''' # read in weights + bias of input projection layer (in timm, this is a single matrix + bias) SCREAMING_SNAKE_CASE : Optional[Any] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.weight""" ) SCREAMING_SNAKE_CASE : List[str] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.bias""" ) # next, add query, keys and values (in that order) to the state dict SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ : config.hidden_size, : ] SCREAMING_SNAKE_CASE : Any = in_proj_bias[: config.hidden_size] SCREAMING_SNAKE_CASE : int = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] SCREAMING_SNAKE_CASE : List[Any] = in_proj_bias[ config.hidden_size : config.hidden_size * 2 ] SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ -config.hidden_size :, : ] SCREAMING_SNAKE_CASE : Optional[Any] = in_proj_bias[-config.hidden_size :] def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = ['''head.weight''', '''head.bias'''] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = [ '''module.fc.fc1.weight''', '''module.fc.fc1.bias''', '''module.fc.bn1.weight''', '''module.fc.bn1.bias''', '''module.fc.bn1.running_mean''', '''module.fc.bn1.running_var''', '''module.fc.bn1.num_batches_tracked''', '''module.fc.fc2.weight''', '''module.fc.fc2.bias''', '''module.fc.bn2.weight''', '''module.fc.bn2.bias''', '''module.fc.bn2.running_mean''', '''module.fc.bn2.running_var''', '''module.fc.bn2.num_batches_tracked''', '''module.fc.fc3.weight''', '''module.fc.fc3.bias''', ] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = dct.pop(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = val def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = ViTMSNConfig() SCREAMING_SNAKE_CASE : Optional[int] = 1_000 SCREAMING_SNAKE_CASE : str = '''datasets/huggingface/label-files''' SCREAMING_SNAKE_CASE : List[str] = '''imagenet-1k-id2label.json''' SCREAMING_SNAKE_CASE : Union[str, Any] = json.load(open(hf_hub_download(a__ , a__ ) , '''r''' ) ) SCREAMING_SNAKE_CASE : List[Any] = {int(a__ ): v for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : str = idalabel SCREAMING_SNAKE_CASE : int = {v: k for k, v in idalabel.items()} if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = 384 SCREAMING_SNAKE_CASE : Any = 1_536 SCREAMING_SNAKE_CASE : List[str] = 6 elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Optional[int] = 1_024 SCREAMING_SNAKE_CASE : Optional[int] = 4_096 SCREAMING_SNAKE_CASE : Tuple = 24 SCREAMING_SNAKE_CASE : Union[str, Any] = 16 SCREAMING_SNAKE_CASE : Dict = 0.1 elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = 4 elif "l7" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = 7 SCREAMING_SNAKE_CASE : Union[str, Any] = 1_024 SCREAMING_SNAKE_CASE : List[Any] = 4_096 SCREAMING_SNAKE_CASE : List[Any] = 24 SCREAMING_SNAKE_CASE : Tuple = 16 SCREAMING_SNAKE_CASE : Union[str, Any] = 0.1 SCREAMING_SNAKE_CASE : Union[str, Any] = ViTMSNModel(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = torch.hub.load_state_dict_from_url(a__ , map_location='''cpu''' )['''target_encoder'''] SCREAMING_SNAKE_CASE : Any = ViTImageProcessor(size=config.image_size ) remove_projection_head(a__ ) SCREAMING_SNAKE_CASE : Any = create_rename_keys(a__ , base_model=a__ ) for src, dest in rename_keys: rename_key(a__ , a__ , a__ ) read_in_q_k_v(a__ , a__ , base_model=a__ ) model.load_state_dict(a__ ) model.eval() SCREAMING_SNAKE_CASE : Any = '''http://images.cocodataset.org/val2017/000000039769.jpg''' SCREAMING_SNAKE_CASE : Dict = Image.open(requests.get(a__ , stream=a__ ).raw ) SCREAMING_SNAKE_CASE : Optional[int] = ViTImageProcessor( size=config.image_size , image_mean=a__ , image_std=a__ ) SCREAMING_SNAKE_CASE : int = image_processor(images=a__ , return_tensors='''pt''' ) # forward pass torch.manual_seed(2 ) SCREAMING_SNAKE_CASE : Tuple = model(**a__ ) SCREAMING_SNAKE_CASE : str = outputs.last_hidden_state # The following Colab Notebook was used to generate these outputs: # https://colab.research.google.com/gist/sayakpaul/3672419a04f5997827503fd84079bdd1/scratchpad.ipynb if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[-1.0_915, -1.4_876, -1.1_809]] ) elif "b16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[14.2_889, -18.9_045, 11.7_281]] ) elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.tensor([[41.5_028, -22.8_681, 45.6_475]] ) elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = torch.tensor([[-4.3_868, 5.2_932, -0.4_137]] ) else: SCREAMING_SNAKE_CASE : List[str] = torch.tensor([[-0.1_792, -0.6_465, 2.4_263]] ) # verify logits assert torch.allclose(last_hidden_state[:, 0, :3] , a__ , atol=1e-4 ) print(F"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(a__ ) print(F"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(a__ ) if __name__ == "__main__": a__ : Dict = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--checkpoint_url''', default='''https://dl.fbaipublicfiles.com/msn/vits16_800ep.pth.tar''', type=str, help='''URL of the checkpoint you\'d like to convert.''', ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.''' ) a__ : Any = parser.parse_args() convert_vit_msn_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
313
1
import argparse from pathlib import Path from typing import Dict, OrderedDict, Tuple import torch from audiocraft.models import MusicGen from transformers import ( AutoFeatureExtractor, AutoTokenizer, EncodecModel, MusicgenDecoderConfig, MusicgenForConditionalGeneration, MusicgenProcessor, TaEncoderModel, ) from transformers.models.musicgen.modeling_musicgen import MusicgenForCausalLM from transformers.utils import logging logging.set_verbosity_info() a__ : Union[str, Any] = logging.get_logger(__name__) a__ : int = ['''model.decoder.embed_positions.weights'''] def UpperCAmelCase_( a__ ): """simple docstring""" if "emb" in name: SCREAMING_SNAKE_CASE : Dict = name.replace('''emb''' , '''model.decoder.embed_tokens''' ) if "transformer" in name: SCREAMING_SNAKE_CASE : int = name.replace('''transformer''' , '''model.decoder''' ) if "cross_attention" in name: SCREAMING_SNAKE_CASE : Optional[int] = name.replace('''cross_attention''' , '''encoder_attn''' ) if "linear1" in name: SCREAMING_SNAKE_CASE : Any = name.replace('''linear1''' , '''fc1''' ) if "linear2" in name: SCREAMING_SNAKE_CASE : Optional[int] = name.replace('''linear2''' , '''fc2''' ) if "norm1" in name: SCREAMING_SNAKE_CASE : List[str] = name.replace('''norm1''' , '''self_attn_layer_norm''' ) if "norm_cross" in name: SCREAMING_SNAKE_CASE : str = name.replace('''norm_cross''' , '''encoder_attn_layer_norm''' ) if "norm2" in name: SCREAMING_SNAKE_CASE : Tuple = name.replace('''norm2''' , '''final_layer_norm''' ) if "out_norm" in name: SCREAMING_SNAKE_CASE : Tuple = name.replace('''out_norm''' , '''model.decoder.layer_norm''' ) if "linears" in name: SCREAMING_SNAKE_CASE : Dict = name.replace('''linears''' , '''lm_heads''' ) if "condition_provider.conditioners.description.output_proj" in name: SCREAMING_SNAKE_CASE : Any = name.replace('''condition_provider.conditioners.description.output_proj''' , '''enc_to_dec_proj''' ) return name def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = list(state_dict.keys() ) SCREAMING_SNAKE_CASE : Any = {} for key in keys: SCREAMING_SNAKE_CASE : List[Any] = state_dict.pop(a__ ) SCREAMING_SNAKE_CASE : Tuple = rename_keys(a__ ) if "in_proj_weight" in key: # split fused qkv proj SCREAMING_SNAKE_CASE : List[Any] = val[:hidden_size, :] SCREAMING_SNAKE_CASE : List[str] = val[hidden_size : 2 * hidden_size, :] SCREAMING_SNAKE_CASE : str = val[-hidden_size:, :] elif "enc_to_dec_proj" in key: SCREAMING_SNAKE_CASE : int = val else: SCREAMING_SNAKE_CASE : str = val return state_dict, enc_dec_proj_state_dict def UpperCAmelCase_( a__ ): """simple docstring""" if checkpoint == "small": # default config values SCREAMING_SNAKE_CASE : Union[str, Any] = 1_024 SCREAMING_SNAKE_CASE : Dict = 24 SCREAMING_SNAKE_CASE : Dict = 16 elif checkpoint == "medium": SCREAMING_SNAKE_CASE : List[Any] = 1_536 SCREAMING_SNAKE_CASE : Union[str, Any] = 48 SCREAMING_SNAKE_CASE : List[Any] = 24 elif checkpoint == "large": SCREAMING_SNAKE_CASE : Dict = 2_048 SCREAMING_SNAKE_CASE : List[str] = 48 SCREAMING_SNAKE_CASE : Optional[Any] = 32 else: raise ValueError(F"""Checkpoint should be one of `['small', 'medium', 'large']`, got {checkpoint}.""" ) SCREAMING_SNAKE_CASE : Tuple = MusicgenDecoderConfig( hidden_size=a__ , ffn_dim=hidden_size * 4 , num_hidden_layers=a__ , num_attention_heads=a__ , ) return config @torch.no_grad() def UpperCAmelCase_( a__ , a__=None , a__=None , a__="cpu" ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = MusicGen.get_pretrained(a__ , device=a__ ) SCREAMING_SNAKE_CASE : str = decoder_config_from_checkpoint(a__ ) SCREAMING_SNAKE_CASE : List[str] = fairseq_model.lm.state_dict() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = rename_state_dict( a__ , hidden_size=decoder_config.hidden_size ) SCREAMING_SNAKE_CASE : int = TaEncoderModel.from_pretrained('''t5-base''' ) SCREAMING_SNAKE_CASE : Tuple = EncodecModel.from_pretrained('''facebook/encodec_32khz''' ) SCREAMING_SNAKE_CASE : str = MusicgenForCausalLM(a__ ).eval() # load all decoder weights - expect that we'll be missing embeddings and enc-dec projection SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = decoder.load_state_dict(a__ , strict=a__ ) for key in missing_keys.copy(): if key.startswith(('''text_encoder''', '''audio_encoder''') ) or key in EXPECTED_MISSING_KEYS: missing_keys.remove(a__ ) if len(a__ ) > 0: raise ValueError(F"""Missing key(s) in state_dict: {missing_keys}""" ) if len(a__ ) > 0: raise ValueError(F"""Unexpected key(s) in state_dict: {unexpected_keys}""" ) # init the composite model SCREAMING_SNAKE_CASE : str = MusicgenForConditionalGeneration(text_encoder=a__ , audio_encoder=a__ , decoder=a__ ) # load the pre-trained enc-dec projection (from the decoder state dict) model.enc_to_dec_proj.load_state_dict(a__ ) # check we can do a forward pass SCREAMING_SNAKE_CASE : Any = torch.arange(0 , 8 , dtype=torch.long ).reshape(2 , -1 ) SCREAMING_SNAKE_CASE : Optional[int] = input_ids.reshape(2 * 4 , -1 ) with torch.no_grad(): SCREAMING_SNAKE_CASE : int = model(input_ids=a__ , decoder_input_ids=a__ ).logits if logits.shape != (8, 1, 2_048): raise ValueError('''Incorrect shape for logits''' ) # now construct the processor SCREAMING_SNAKE_CASE : str = AutoTokenizer.from_pretrained('''t5-base''' ) SCREAMING_SNAKE_CASE : int = AutoFeatureExtractor.from_pretrained('''facebook/encodec_32khz''' , padding_side='''left''' ) SCREAMING_SNAKE_CASE : Tuple = MusicgenProcessor(feature_extractor=a__ , tokenizer=a__ ) # set the appropriate bos/pad token ids SCREAMING_SNAKE_CASE : List[str] = 2_048 SCREAMING_SNAKE_CASE : str = 2_048 # set other default generation config params SCREAMING_SNAKE_CASE : Any = int(30 * audio_encoder.config.frame_rate ) SCREAMING_SNAKE_CASE : Dict = True SCREAMING_SNAKE_CASE : List[Any] = 3.0 if pytorch_dump_folder is not None: Path(a__ ).mkdir(exist_ok=a__ ) logger.info(F"""Saving model {checkpoint} to {pytorch_dump_folder}""" ) model.save_pretrained(a__ ) processor.save_pretrained(a__ ) if repo_id: logger.info(F"""Pushing model {checkpoint} to {repo_id}""" ) model.push_to_hub(a__ ) processor.push_to_hub(a__ ) if __name__ == "__main__": a__ : Tuple = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--checkpoint''', default='''small''', type=str, help='''Checkpoint size of the MusicGen model you\'d like to convert. Can be one of: `[\'small\', \'medium\', \'large\']`.''', ) parser.add_argument( '''--pytorch_dump_folder''', required=True, default=None, type=str, help='''Path to the output PyTorch model directory.''', ) parser.add_argument( '''--push_to_hub''', default=None, type=str, help='''Where to upload the converted model on the 🤗 hub.''' ) parser.add_argument( '''--device''', default='''cpu''', type=str, help='''Torch device to run the conversion, either cpu or cuda.''' ) a__ : int = parser.parse_args() convert_musicgen_checkpoint(args.checkpoint, args.pytorch_dump_folder, args.push_to_hub)
313
import csv import tweepy # Twitter API credentials a__ : Union[str, Any] = '''''' a__ : List[str] = '''''' a__ : Any = '''''' a__ : List[str] = '''''' def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = tweepy.OAuthHandler(a__ , a__ ) auth.set_access_token(a__ , a__ ) SCREAMING_SNAKE_CASE : List[str] = tweepy.API(a__ ) # initialize a list to hold all the tweepy Tweets SCREAMING_SNAKE_CASE : Any = [] # make initial request for most recent tweets (200 is the maximum allowed count) SCREAMING_SNAKE_CASE : List[Any] = api.user_timeline(screen_name=a__ , count=200 ) # save most recent tweets alltweets.extend(a__ ) # save the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Tuple = alltweets[-1].id - 1 # keep grabbing tweets until there are no tweets left to grab while len(a__ ) > 0: print(F"""getting tweets before {oldest}""" ) # all subsequent requests use the max_id param to prevent duplicates SCREAMING_SNAKE_CASE : Any = api.user_timeline( screen_name=a__ , count=200 , max_id=a__ ) # save most recent tweets alltweets.extend(a__ ) # update the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Dict = alltweets[-1].id - 1 print(F"""...{len(a__ )} tweets downloaded so far""" ) # transform the tweepy tweets into a 2D array that will populate the csv SCREAMING_SNAKE_CASE : Optional[Any] = [[tweet.id_str, tweet.created_at, tweet.text] for tweet in alltweets] # write the csv with open(F"""new_{screen_name}_tweets.csv""" , '''w''' ) as f: SCREAMING_SNAKE_CASE : List[Any] = csv.writer(a__ ) writer.writerow(['''id''', '''created_at''', '''text'''] ) writer.writerows(a__ ) if __name__ == "__main__": # pass in the username of the account you want to download get_all_tweets('''FirePing32''')
313
1
from ...configuration_utils import PretrainedConfig from ...utils import logging from ...utils.backbone_utils import BackboneConfigMixin, get_aligned_output_features_output_indices a__ : List[Any] = logging.get_logger(__name__) a__ : List[Any] = { '''google/bit-50''': '''https://huggingface.co/google/bit-50/resolve/main/config.json''', } class a_ ( a__ , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Any = 'bit' __SCREAMING_SNAKE_CASE : Tuple = ['preactivation', 'bottleneck'] __SCREAMING_SNAKE_CASE : Tuple = ['SAME', 'VALID'] def __init__( self , _lowerCamelCase=3 , _lowerCamelCase=64 , _lowerCamelCase=[256, 512, 1024, 2048] , _lowerCamelCase=[3, 4, 6, 3] , _lowerCamelCase="preactivation" , _lowerCamelCase="relu" , _lowerCamelCase=None , _lowerCamelCase=32 , _lowerCamelCase=0.0 , _lowerCamelCase=False , _lowerCamelCase=32 , _lowerCamelCase=1 , _lowerCamelCase=None , _lowerCamelCase=None , **_lowerCamelCase , ) ->Union[str, Any]: super().__init__(**_lowerCamelCase ) if layer_type not in self.layer_types: raise ValueError(F"""layer_type={layer_type} is not one of {",".join(self.layer_types )}""" ) if global_padding is not None: if global_padding.upper() in self.supported_padding: SCREAMING_SNAKE_CASE : Any = global_padding.upper() else: raise ValueError(F"""Padding strategy {global_padding} not supported""" ) SCREAMING_SNAKE_CASE : List[Any] = num_channels SCREAMING_SNAKE_CASE : Optional[Any] = embedding_size SCREAMING_SNAKE_CASE : Union[str, Any] = hidden_sizes SCREAMING_SNAKE_CASE : Any = depths SCREAMING_SNAKE_CASE : Dict = layer_type SCREAMING_SNAKE_CASE : int = hidden_act SCREAMING_SNAKE_CASE : List[Any] = global_padding SCREAMING_SNAKE_CASE : Optional[Any] = num_groups SCREAMING_SNAKE_CASE : Optional[Any] = drop_path_rate SCREAMING_SNAKE_CASE : Any = embedding_dynamic_padding SCREAMING_SNAKE_CASE : List[Any] = output_stride SCREAMING_SNAKE_CASE : int = width_factor SCREAMING_SNAKE_CASE : Optional[int] = ['''stem'''] + [F"""stage{idx}""" for idx in range(1 , len(_lowerCamelCase ) + 1 )] SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = get_aligned_output_features_output_indices( out_features=_lowerCamelCase , out_indices=_lowerCamelCase , stage_names=self.stage_names )
313
from collections import OrderedDict from typing import Mapping from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a__ : Optional[Any] = logging.get_logger(__name__) a__ : List[str] = { '''kssteven/ibert-roberta-base''': '''https://huggingface.co/kssteven/ibert-roberta-base/resolve/main/config.json''', '''kssteven/ibert-roberta-large''': '''https://huggingface.co/kssteven/ibert-roberta-large/resolve/main/config.json''', '''kssteven/ibert-roberta-large-mnli''': ( '''https://huggingface.co/kssteven/ibert-roberta-large-mnli/resolve/main/config.json''' ), } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'ibert' def __init__( self , _lowerCamelCase=3_0522 , _lowerCamelCase=768 , _lowerCamelCase=12 , _lowerCamelCase=12 , _lowerCamelCase=3072 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1e-12 , _lowerCamelCase=1 , _lowerCamelCase=0 , _lowerCamelCase=2 , _lowerCamelCase="absolute" , _lowerCamelCase=False , _lowerCamelCase="none" , **_lowerCamelCase , ) ->Any: super().__init__(pad_token_id=_lowerCamelCase , bos_token_id=_lowerCamelCase , eos_token_id=_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = vocab_size SCREAMING_SNAKE_CASE : str = hidden_size SCREAMING_SNAKE_CASE : Tuple = num_hidden_layers SCREAMING_SNAKE_CASE : List[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Dict = hidden_act SCREAMING_SNAKE_CASE : Optional[int] = intermediate_size SCREAMING_SNAKE_CASE : Optional[Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Tuple = type_vocab_size SCREAMING_SNAKE_CASE : Optional[int] = initializer_range SCREAMING_SNAKE_CASE : Union[str, Any] = layer_norm_eps SCREAMING_SNAKE_CASE : str = position_embedding_type SCREAMING_SNAKE_CASE : Optional[int] = quant_mode SCREAMING_SNAKE_CASE : Dict = force_dequant class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: if self.task == "multiple-choice": SCREAMING_SNAKE_CASE : Dict = {0: '''batch''', 1: '''choice''', 2: '''sequence'''} else: SCREAMING_SNAKE_CASE : List[Any] = {0: '''batch''', 1: '''sequence'''} return OrderedDict( [ ('''input_ids''', dynamic_axis), ('''attention_mask''', dynamic_axis), ] )
313
1
import json import pathlib import unittest import numpy as np from transformers.testing_utils import require_torch, require_vision, slow 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 YolosImageProcessor class a_ ( unittest.TestCase ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=7 , _lowerCamelCase=3 , _lowerCamelCase=30 , _lowerCamelCase=400 , _lowerCamelCase=True , _lowerCamelCase=None , _lowerCamelCase=True , _lowerCamelCase=[0.5, 0.5, 0.5] , _lowerCamelCase=[0.5, 0.5, 0.5] , _lowerCamelCase=True , _lowerCamelCase=1 / 255 , _lowerCamelCase=True , ) ->List[str]: # by setting size["longest_edge"] > max_resolution we're effectively not testing this :p SCREAMING_SNAKE_CASE : int = size if size is not None else {'''shortest_edge''': 18, '''longest_edge''': 1333} SCREAMING_SNAKE_CASE : List[Any] = parent SCREAMING_SNAKE_CASE : List[Any] = batch_size SCREAMING_SNAKE_CASE : str = num_channels SCREAMING_SNAKE_CASE : Optional[int] = min_resolution SCREAMING_SNAKE_CASE : Any = max_resolution SCREAMING_SNAKE_CASE : List[str] = do_resize SCREAMING_SNAKE_CASE : List[Any] = size SCREAMING_SNAKE_CASE : Union[str, Any] = do_normalize SCREAMING_SNAKE_CASE : Union[str, Any] = image_mean SCREAMING_SNAKE_CASE : Optional[Any] = image_std SCREAMING_SNAKE_CASE : Optional[int] = do_rescale SCREAMING_SNAKE_CASE : Any = rescale_factor SCREAMING_SNAKE_CASE : Optional[Any] = do_pad def __lowerCAmelCase ( self ) ->List[Any]: return { "do_resize": self.do_resize, "size": self.size, "do_normalize": self.do_normalize, "image_mean": self.image_mean, "image_std": self.image_std, "do_rescale": self.do_rescale, "rescale_factor": self.rescale_factor, "do_pad": self.do_pad, } def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=False ) ->Optional[int]: if not batched: SCREAMING_SNAKE_CASE : Optional[Any] = image_inputs[0] if isinstance(_lowerCamelCase , Image.Image ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = image.size else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = image.shape[1], image.shape[2] if w < h: SCREAMING_SNAKE_CASE : Optional[Any] = int(self.size['''shortest_edge'''] * h / w ) SCREAMING_SNAKE_CASE : str = self.size['''shortest_edge'''] elif w > h: SCREAMING_SNAKE_CASE : Any = self.size['''shortest_edge'''] SCREAMING_SNAKE_CASE : str = int(self.size['''shortest_edge'''] * w / h ) else: SCREAMING_SNAKE_CASE : Tuple = self.size['''shortest_edge'''] SCREAMING_SNAKE_CASE : Tuple = self.size['''shortest_edge'''] else: SCREAMING_SNAKE_CASE : str = [] for image in image_inputs: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.get_expected_values([image] ) expected_values.append((expected_height, expected_width) ) SCREAMING_SNAKE_CASE : str = max(_lowerCamelCase , key=lambda _lowerCamelCase : item[0] )[0] SCREAMING_SNAKE_CASE : List[str] = max(_lowerCamelCase , key=lambda _lowerCamelCase : item[1] )[1] return expected_height, expected_width @require_torch @require_vision class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[Any] = YolosImageProcessor if is_vision_available() else None def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Tuple = YolosImageProcessingTester(self ) @property def __lowerCAmelCase ( self ) ->int: return self.image_processor_tester.prepare_image_processor_dict() def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Optional[Any] = self.image_processing_class(**self.image_processor_dict ) self.assertTrue(hasattr(_lowerCamelCase , '''image_mean''' ) ) self.assertTrue(hasattr(_lowerCamelCase , '''image_std''' ) ) self.assertTrue(hasattr(_lowerCamelCase , '''do_normalize''' ) ) self.assertTrue(hasattr(_lowerCamelCase , '''do_resize''' ) ) self.assertTrue(hasattr(_lowerCamelCase , '''size''' ) ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Tuple = self.image_processing_class.from_dict(self.image_processor_dict ) self.assertEqual(image_processor.size , {'''shortest_edge''': 18, '''longest_edge''': 1333} ) self.assertEqual(image_processor.do_pad , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = self.image_processing_class.from_dict( self.image_processor_dict , size=42 , max_size=84 , pad_and_return_pixel_mask=_lowerCamelCase ) self.assertEqual(image_processor.size , {'''shortest_edge''': 42, '''longest_edge''': 84} ) self.assertEqual(image_processor.do_pad , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: pass def __lowerCAmelCase ( self ) ->Optional[Any]: # Initialize image_processing SCREAMING_SNAKE_CASE : Optional[Any] = self.image_processing_class(**self.image_processor_dict ) # create random PIL images SCREAMING_SNAKE_CASE : List[str] = prepare_image_inputs(self.image_processor_tester , equal_resolution=_lowerCamelCase ) for image in image_inputs: self.assertIsInstance(_lowerCamelCase , Image.Image ) # Test not batched input SCREAMING_SNAKE_CASE : List[Any] = image_processing(image_inputs[0] , return_tensors='''pt''' ).pixel_values SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = self.image_processor_tester.get_expected_values(_lowerCamelCase ) self.assertEqual( encoded_images.shape , (1, self.image_processor_tester.num_channels, expected_height, expected_width) , ) # Test batched SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = self.image_processor_tester.get_expected_values(_lowerCamelCase , batched=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = 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, expected_height, expected_width, ) , ) def __lowerCAmelCase ( self ) ->Dict: # Initialize image_processing SCREAMING_SNAKE_CASE : Optional[Any] = self.image_processing_class(**self.image_processor_dict ) # create random numpy tensors SCREAMING_SNAKE_CASE : Union[str, Any] = 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 SCREAMING_SNAKE_CASE : Dict = image_processing(image_inputs[0] , return_tensors='''pt''' ).pixel_values SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.image_processor_tester.get_expected_values(_lowerCamelCase ) self.assertEqual( encoded_images.shape , (1, self.image_processor_tester.num_channels, expected_height, expected_width) , ) # Test batched SCREAMING_SNAKE_CASE : Optional[int] = image_processing(_lowerCamelCase , return_tensors='''pt''' ).pixel_values SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.image_processor_tester.get_expected_values(_lowerCamelCase , batched=_lowerCamelCase ) self.assertEqual( encoded_images.shape , ( self.image_processor_tester.batch_size, self.image_processor_tester.num_channels, expected_height, expected_width, ) , ) def __lowerCAmelCase ( self ) ->Optional[int]: # Initialize image_processing SCREAMING_SNAKE_CASE : Any = self.image_processing_class(**self.image_processor_dict ) # create random PyTorch tensors SCREAMING_SNAKE_CASE : List[Any] = 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 SCREAMING_SNAKE_CASE : Tuple = image_processing(image_inputs[0] , return_tensors='''pt''' ).pixel_values SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.image_processor_tester.get_expected_values(_lowerCamelCase ) self.assertEqual( encoded_images.shape , (1, self.image_processor_tester.num_channels, expected_height, expected_width) , ) # Test batched SCREAMING_SNAKE_CASE : int = image_processing(_lowerCamelCase , return_tensors='''pt''' ).pixel_values SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = self.image_processor_tester.get_expected_values(_lowerCamelCase , batched=_lowerCamelCase ) self.assertEqual( encoded_images.shape , ( self.image_processor_tester.batch_size, self.image_processor_tester.num_channels, expected_height, expected_width, ) , ) def __lowerCAmelCase ( self ) ->Optional[Any]: # Initialize image_processings SCREAMING_SNAKE_CASE : Dict = self.image_processing_class(**self.image_processor_dict ) SCREAMING_SNAKE_CASE : int = self.image_processing_class(do_resize=_lowerCamelCase , do_normalize=_lowerCamelCase , do_rescale=_lowerCamelCase ) # create random PyTorch tensors SCREAMING_SNAKE_CASE : List[Any] = prepare_image_inputs(self.image_processor_tester , equal_resolution=_lowerCamelCase , torchify=_lowerCamelCase ) for image in image_inputs: self.assertIsInstance(_lowerCamelCase , torch.Tensor ) # Test whether the method "pad" and calling the image processor return the same tensors SCREAMING_SNAKE_CASE : List[str] = image_processing_a.pad(_lowerCamelCase , return_tensors='''pt''' ) SCREAMING_SNAKE_CASE : Union[str, Any] = image_processing_a(_lowerCamelCase , return_tensors='''pt''' ) self.assertTrue( torch.allclose(encoded_images_with_method['''pixel_values'''] , encoded_images['''pixel_values'''] , atol=1e-4 ) ) @slow def __lowerCAmelCase ( self ) ->Dict: # prepare image and target SCREAMING_SNAKE_CASE : Union[str, Any] = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) with open('''./tests/fixtures/tests_samples/COCO/coco_annotations.txt''' , '''r''' ) as f: SCREAMING_SNAKE_CASE : Union[str, Any] = json.loads(f.read() ) SCREAMING_SNAKE_CASE : Union[str, Any] = {'''image_id''': 3_9769, '''annotations''': target} # encode them SCREAMING_SNAKE_CASE : Optional[Any] = YolosImageProcessor.from_pretrained('''hustvl/yolos-small''' ) SCREAMING_SNAKE_CASE : List[Any] = image_processing(images=_lowerCamelCase , annotations=_lowerCamelCase , return_tensors='''pt''' ) # verify pixel values SCREAMING_SNAKE_CASE : List[str] = torch.Size([1, 3, 800, 1066] ) self.assertEqual(encoding['''pixel_values'''].shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.tensor([0.2_7_9_6, 0.3_1_3_8, 0.3_4_8_1] ) self.assertTrue(torch.allclose(encoding['''pixel_values'''][0, 0, 0, :3] , _lowerCamelCase , atol=1e-4 ) ) # verify area SCREAMING_SNAKE_CASE : List[Any] = torch.tensor([5_8_8_7.9_6_0_0, 1_1_2_5_0.2_0_6_1, 4_8_9_3_5_3.8_4_3_8, 8_3_7_1_2_2.7_5_0_0, 1_4_7_9_6_7.5_1_5_6, 1_6_5_7_3_2.3_4_3_8] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''area'''] , _lowerCamelCase ) ) # verify boxes SCREAMING_SNAKE_CASE : List[str] = torch.Size([6, 4] ) self.assertEqual(encoding['''labels'''][0]['''boxes'''].shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = torch.tensor([0.5_5_0_3, 0.2_7_6_5, 0.0_6_0_4, 0.2_2_1_5] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''boxes'''][0] , _lowerCamelCase , atol=1e-3 ) ) # verify image_id SCREAMING_SNAKE_CASE : List[str] = torch.tensor([3_9769] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''image_id'''] , _lowerCamelCase ) ) # verify is_crowd SCREAMING_SNAKE_CASE : Optional[Any] = torch.tensor([0, 0, 0, 0, 0, 0] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''iscrowd'''] , _lowerCamelCase ) ) # verify class_labels SCREAMING_SNAKE_CASE : Dict = torch.tensor([75, 75, 63, 65, 17, 17] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''class_labels'''] , _lowerCamelCase ) ) # verify orig_size SCREAMING_SNAKE_CASE : Optional[int] = torch.tensor([480, 640] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''orig_size'''] , _lowerCamelCase ) ) # verify size SCREAMING_SNAKE_CASE : str = torch.tensor([800, 1066] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''size'''] , _lowerCamelCase ) ) @slow def __lowerCAmelCase ( self ) ->List[Any]: # prepare image, target and masks_path SCREAMING_SNAKE_CASE : List[str] = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) with open('''./tests/fixtures/tests_samples/COCO/coco_panoptic_annotations.txt''' , '''r''' ) as f: SCREAMING_SNAKE_CASE : Union[str, Any] = json.loads(f.read() ) SCREAMING_SNAKE_CASE : List[str] = {'''file_name''': '''000000039769.png''', '''image_id''': 3_9769, '''segments_info''': target} SCREAMING_SNAKE_CASE : Union[str, Any] = pathlib.Path('''./tests/fixtures/tests_samples/COCO/coco_panoptic''' ) # encode them SCREAMING_SNAKE_CASE : Optional[Any] = YolosImageProcessor(format='''coco_panoptic''' ) SCREAMING_SNAKE_CASE : int = image_processing(images=_lowerCamelCase , annotations=_lowerCamelCase , masks_path=_lowerCamelCase , return_tensors='''pt''' ) # verify pixel values SCREAMING_SNAKE_CASE : Any = torch.Size([1, 3, 800, 1066] ) self.assertEqual(encoding['''pixel_values'''].shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = torch.tensor([0.2_7_9_6, 0.3_1_3_8, 0.3_4_8_1] ) self.assertTrue(torch.allclose(encoding['''pixel_values'''][0, 0, 0, :3] , _lowerCamelCase , atol=1e-4 ) ) # verify area SCREAMING_SNAKE_CASE : List[str] = torch.tensor([1_4_7_9_7_9.6_8_7_5, 1_6_5_5_2_7.0_4_6_9, 4_8_4_6_3_8.5_9_3_8, 1_1_2_9_2.9_3_7_5, 5_8_7_9.6_5_6_2, 7_6_3_4.1_1_4_7] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''area'''] , _lowerCamelCase ) ) # verify boxes SCREAMING_SNAKE_CASE : Dict = torch.Size([6, 4] ) self.assertEqual(encoding['''labels'''][0]['''boxes'''].shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : str = torch.tensor([0.2_6_2_5, 0.5_4_3_7, 0.4_6_8_8, 0.8_6_2_5] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''boxes'''][0] , _lowerCamelCase , atol=1e-3 ) ) # verify image_id SCREAMING_SNAKE_CASE : Dict = torch.tensor([3_9769] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''image_id'''] , _lowerCamelCase ) ) # verify is_crowd SCREAMING_SNAKE_CASE : int = torch.tensor([0, 0, 0, 0, 0, 0] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''iscrowd'''] , _lowerCamelCase ) ) # verify class_labels SCREAMING_SNAKE_CASE : List[Any] = torch.tensor([17, 17, 63, 75, 75, 93] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''class_labels'''] , _lowerCamelCase ) ) # verify masks SCREAMING_SNAKE_CASE : Optional[Any] = 82_2873 self.assertEqual(encoding['''labels'''][0]['''masks'''].sum().item() , _lowerCamelCase ) # verify orig_size SCREAMING_SNAKE_CASE : int = torch.tensor([480, 640] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''orig_size'''] , _lowerCamelCase ) ) # verify size SCREAMING_SNAKE_CASE : List[Any] = torch.tensor([800, 1066] ) self.assertTrue(torch.allclose(encoding['''labels'''][0]['''size'''] , _lowerCamelCase ) )
313
from collections import OrderedDict from typing import TYPE_CHECKING, Any, Mapping, Optional from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging if TYPE_CHECKING: from ... import FeatureExtractionMixin, TensorType a__ : Any = logging.get_logger(__name__) a__ : Dict = { '''openai/imagegpt-small''': '''''', '''openai/imagegpt-medium''': '''''', '''openai/imagegpt-large''': '''''', } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 'imagegpt' __SCREAMING_SNAKE_CASE : Optional[Any] = ['past_key_values'] __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'n_embd', 'max_position_embeddings': 'n_positions', 'num_attention_heads': 'n_head', 'num_hidden_layers': 'n_layer', } def __init__( self , _lowerCamelCase=512 + 1 , _lowerCamelCase=32 * 32 , _lowerCamelCase=512 , _lowerCamelCase=24 , _lowerCamelCase=8 , _lowerCamelCase=None , _lowerCamelCase="quick_gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=1e-5 , _lowerCamelCase=0.0_2 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=False , _lowerCamelCase=False , **_lowerCamelCase , ) ->str: SCREAMING_SNAKE_CASE : Any = vocab_size SCREAMING_SNAKE_CASE : Optional[int] = n_positions SCREAMING_SNAKE_CASE : Optional[int] = n_embd SCREAMING_SNAKE_CASE : List[Any] = n_layer SCREAMING_SNAKE_CASE : List[Any] = n_head SCREAMING_SNAKE_CASE : int = n_inner SCREAMING_SNAKE_CASE : Dict = activation_function SCREAMING_SNAKE_CASE : Union[str, Any] = resid_pdrop SCREAMING_SNAKE_CASE : Dict = embd_pdrop SCREAMING_SNAKE_CASE : List[str] = attn_pdrop SCREAMING_SNAKE_CASE : Optional[int] = layer_norm_epsilon SCREAMING_SNAKE_CASE : Tuple = initializer_range SCREAMING_SNAKE_CASE : int = scale_attn_weights SCREAMING_SNAKE_CASE : Optional[int] = use_cache SCREAMING_SNAKE_CASE : Optional[Any] = scale_attn_by_inverse_layer_idx SCREAMING_SNAKE_CASE : str = reorder_and_upcast_attn SCREAMING_SNAKE_CASE : List[str] = tie_word_embeddings super().__init__(tie_word_embeddings=_lowerCamelCase , **_lowerCamelCase ) class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: return OrderedDict( [ ('''input_ids''', {0: '''batch''', 1: '''sequence'''}), ] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 1 , _lowerCamelCase = -1 , _lowerCamelCase = False , _lowerCamelCase = None , _lowerCamelCase = 3 , _lowerCamelCase = 32 , _lowerCamelCase = 32 , ) ->Mapping[str, Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = self._generate_dummy_images(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = dict(preprocessor(images=_lowerCamelCase , return_tensors=_lowerCamelCase ) ) return inputs
313
1
import argparse import json import os import pickle import shutil import numpy as np import torch from distiller import Distiller from lm_seqs_dataset import LmSeqsDataset from transformers import ( BertConfig, BertForMaskedLM, BertTokenizer, DistilBertConfig, DistilBertForMaskedLM, DistilBertTokenizer, GPTaConfig, GPTaLMHeadModel, GPTaTokenizer, RobertaConfig, RobertaForMaskedLM, RobertaTokenizer, ) from utils import git_log, init_gpu_params, logger, set_seed a__ : List[Any] = { '''distilbert''': (DistilBertConfig, DistilBertForMaskedLM, DistilBertTokenizer), '''roberta''': (RobertaConfig, RobertaForMaskedLM, RobertaTokenizer), '''bert''': (BertConfig, BertForMaskedLM, BertTokenizer), '''gpt2''': (GPTaConfig, GPTaLMHeadModel, GPTaTokenizer), } def UpperCAmelCase_( a__ ): """simple docstring""" assert (args.mlm and args.alpha_mlm > 0.0) or (not args.mlm and args.alpha_mlm == 0.0) assert (args.alpha_mlm > 0.0 and args.alpha_clm == 0.0) or (args.alpha_mlm == 0.0 and args.alpha_clm > 0.0) if args.mlm: assert os.path.isfile(args.token_counts ) assert (args.student_type in ["roberta", "distilbert"]) and (args.teacher_type in ["roberta", "bert"]) else: assert (args.student_type in ["gpt2"]) and (args.teacher_type in ["gpt2"]) assert args.teacher_type == args.student_type or ( args.student_type == "distilbert" and args.teacher_type == "bert" ) assert os.path.isfile(args.student_config ) if args.student_pretrained_weights is not None: assert os.path.isfile(args.student_pretrained_weights ) if args.freeze_token_type_embds: assert args.student_type in ["roberta"] assert args.alpha_ce >= 0.0 assert args.alpha_mlm >= 0.0 assert args.alpha_clm >= 0.0 assert args.alpha_mse >= 0.0 assert args.alpha_cos >= 0.0 assert args.alpha_ce + args.alpha_mlm + args.alpha_clm + args.alpha_mse + args.alpha_cos > 0.0 def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if args.student_type == "roberta": SCREAMING_SNAKE_CASE : Union[str, Any] = False elif args.student_type == "gpt2": SCREAMING_SNAKE_CASE : Optional[Any] = False def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if args.student_type == "roberta": SCREAMING_SNAKE_CASE : Any = False def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = argparse.ArgumentParser(description='''Training''' ) parser.add_argument('''--force''' , action='''store_true''' , help='''Overwrite dump_path if it already exists.''' ) parser.add_argument( '''--dump_path''' , type=a__ , required=a__ , help='''The output directory (log, checkpoints, parameters, etc.)''' ) parser.add_argument( '''--data_file''' , type=a__ , required=a__ , help='''The binarized file (tokenized + tokens_to_ids) and grouped by sequence.''' , ) parser.add_argument( '''--student_type''' , type=a__ , choices=['''distilbert''', '''roberta''', '''gpt2'''] , required=a__ , help='''The student type (DistilBERT, RoBERTa).''' , ) parser.add_argument('''--student_config''' , type=a__ , required=a__ , help='''Path to the student configuration.''' ) parser.add_argument( '''--student_pretrained_weights''' , default=a__ , type=a__ , help='''Load student initialization checkpoint.''' ) parser.add_argument( '''--teacher_type''' , choices=['''bert''', '''roberta''', '''gpt2'''] , required=a__ , help='''Teacher type (BERT, RoBERTa).''' ) parser.add_argument('''--teacher_name''' , type=a__ , required=a__ , help='''The teacher model.''' ) parser.add_argument('''--temperature''' , default=2.0 , type=a__ , help='''Temperature for the softmax temperature.''' ) parser.add_argument( '''--alpha_ce''' , default=0.5 , type=a__ , help='''Linear weight for the distillation loss. Must be >=0.''' ) parser.add_argument( '''--alpha_mlm''' , default=0.0 , type=a__ , help='''Linear weight for the MLM loss. Must be >=0. Should be used in conjunction with `mlm` flag.''' , ) parser.add_argument('''--alpha_clm''' , default=0.5 , type=a__ , help='''Linear weight for the CLM loss. Must be >=0.''' ) parser.add_argument('''--alpha_mse''' , default=0.0 , type=a__ , help='''Linear weight of the MSE loss. Must be >=0.''' ) parser.add_argument( '''--alpha_cos''' , default=0.0 , type=a__ , help='''Linear weight of the cosine embedding loss. Must be >=0.''' ) parser.add_argument( '''--mlm''' , action='''store_true''' , help='''The LM step: MLM or CLM. If `mlm` is True, the MLM is used over CLM.''' ) parser.add_argument( '''--mlm_mask_prop''' , default=0.15 , type=a__ , help='''Proportion of tokens for which we need to make a prediction.''' , ) parser.add_argument('''--word_mask''' , default=0.8 , type=a__ , help='''Proportion of tokens to mask out.''' ) parser.add_argument('''--word_keep''' , default=0.1 , type=a__ , help='''Proportion of tokens to keep.''' ) parser.add_argument('''--word_rand''' , default=0.1 , type=a__ , help='''Proportion of tokens to randomly replace.''' ) parser.add_argument( '''--mlm_smoothing''' , default=0.7 , type=a__ , help='''Smoothing parameter to emphasize more rare tokens (see XLM, similar to word2vec).''' , ) parser.add_argument('''--token_counts''' , type=a__ , help='''The token counts in the data_file for MLM.''' ) parser.add_argument( '''--restrict_ce_to_mask''' , action='''store_true''' , help='''If true, compute the distillation loss only the [MLM] prediction distribution.''' , ) parser.add_argument( '''--freeze_pos_embs''' , action='''store_true''' , help='''Freeze positional embeddings during distillation. For student_type in [\'roberta\', \'gpt2\'] only.''' , ) parser.add_argument( '''--freeze_token_type_embds''' , action='''store_true''' , help='''Freeze token type embeddings during distillation if existent. For student_type in [\'roberta\'] only.''' , ) parser.add_argument('''--n_epoch''' , type=a__ , default=3 , help='''Number of pass on the whole dataset.''' ) parser.add_argument('''--batch_size''' , type=a__ , default=5 , help='''Batch size (for each process).''' ) parser.add_argument( '''--group_by_size''' , action='''store_false''' , help='''If true, group sequences that have similar length into the same batch. Default is true.''' , ) parser.add_argument( '''--gradient_accumulation_steps''' , type=a__ , default=50 , help='''Gradient accumulation for larger training batches.''' , ) parser.add_argument('''--warmup_prop''' , default=0.05 , type=a__ , help='''Linear warmup proportion.''' ) parser.add_argument('''--weight_decay''' , default=0.0 , type=a__ , help='''Weight decay if we apply some.''' ) parser.add_argument('''--learning_rate''' , default=5e-4 , type=a__ , help='''The initial learning rate for Adam.''' ) parser.add_argument('''--adam_epsilon''' , default=1e-6 , type=a__ , help='''Epsilon for Adam optimizer.''' ) parser.add_argument('''--max_grad_norm''' , default=5.0 , type=a__ , help='''Max gradient norm.''' ) parser.add_argument('''--initializer_range''' , default=0.02 , type=a__ , help='''Random initialization range.''' ) parser.add_argument( '''--fp16''' , action='''store_true''' , help='''Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit''' , ) parser.add_argument( '''--fp16_opt_level''' , type=a__ , default='''O1''' , help=( '''For fp16: Apex AMP optimization level selected in [\'O0\', \'O1\', \'O2\', and \'O3\'].''' '''See details at https://nvidia.github.io/apex/amp.html''' ) , ) parser.add_argument('''--n_gpu''' , type=a__ , default=1 , help='''Number of GPUs in the node.''' ) parser.add_argument('''--local_rank''' , type=a__ , default=-1 , help='''Distributed training - Local rank''' ) parser.add_argument('''--seed''' , type=a__ , default=56 , help='''Random seed''' ) parser.add_argument('''--log_interval''' , type=a__ , default=500 , help='''Tensorboard logging interval.''' ) parser.add_argument('''--checkpoint_interval''' , type=a__ , default=4_000 , help='''Checkpoint interval.''' ) SCREAMING_SNAKE_CASE : Optional[Any] = parser.parse_args() sanity_checks(a__ ) # ARGS # init_gpu_params(a__ ) set_seed(a__ ) if args.is_master: if os.path.exists(args.dump_path ): if not args.force: raise ValueError( F"""Serialization dir {args.dump_path} already exists, but you have not precised wheter to overwrite""" ''' itUse `--force` if you want to overwrite it''' ) else: shutil.rmtree(args.dump_path ) if not os.path.exists(args.dump_path ): os.makedirs(args.dump_path ) logger.info(F"""Experiment will be dumped and logged in {args.dump_path}""" ) # SAVE PARAMS # logger.info(F"""Param: {args}""" ) with open(os.path.join(args.dump_path , '''parameters.json''' ) , '''w''' ) as f: json.dump(vars(a__ ) , a__ , indent=4 ) git_log(args.dump_path ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = MODEL_CLASSES[args.student_type] SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = MODEL_CLASSES[args.teacher_type] # TOKENIZER # SCREAMING_SNAKE_CASE : Optional[int] = teacher_tokenizer_class.from_pretrained(args.teacher_name ) SCREAMING_SNAKE_CASE : Optional[Any] = {} for tok_name, tok_symbol in tokenizer.special_tokens_map.items(): SCREAMING_SNAKE_CASE : int = tokenizer.all_special_tokens.index(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = tokenizer.all_special_ids[idx] logger.info(F"""Special tokens {special_tok_ids}""" ) SCREAMING_SNAKE_CASE : List[Any] = special_tok_ids SCREAMING_SNAKE_CASE : str = tokenizer.max_model_input_sizes[args.teacher_name] # DATA LOADER # logger.info(F"""Loading data from {args.data_file}""" ) with open(args.data_file , '''rb''' ) as fp: SCREAMING_SNAKE_CASE : Optional[Any] = pickle.load(a__ ) if args.mlm: logger.info(F"""Loading token counts from {args.token_counts} (already pre-computed)""" ) with open(args.token_counts , '''rb''' ) as fp: SCREAMING_SNAKE_CASE : str = pickle.load(a__ ) SCREAMING_SNAKE_CASE : List[str] = np.maximum(a__ , 1 ) ** -args.mlm_smoothing for idx in special_tok_ids.values(): SCREAMING_SNAKE_CASE : str = 0.0 # do not predict special tokens SCREAMING_SNAKE_CASE : List[Any] = torch.from_numpy(a__ ) else: SCREAMING_SNAKE_CASE : Optional[int] = None SCREAMING_SNAKE_CASE : int = LmSeqsDataset(params=a__ , data=a__ ) logger.info('''Data loader created.''' ) # STUDENT # logger.info(F"""Loading student config from {args.student_config}""" ) SCREAMING_SNAKE_CASE : Any = student_config_class.from_pretrained(args.student_config ) SCREAMING_SNAKE_CASE : int = True if args.student_pretrained_weights is not None: logger.info(F"""Loading pretrained weights from {args.student_pretrained_weights}""" ) SCREAMING_SNAKE_CASE : Any = student_model_class.from_pretrained(args.student_pretrained_weights , config=a__ ) else: SCREAMING_SNAKE_CASE : Tuple = student_model_class(a__ ) if args.n_gpu > 0: student.to(F"""cuda:{args.local_rank}""" ) logger.info('''Student loaded.''' ) # TEACHER # SCREAMING_SNAKE_CASE : Union[str, Any] = teacher_model_class.from_pretrained(args.teacher_name , output_hidden_states=a__ ) if args.n_gpu > 0: teacher.to(F"""cuda:{args.local_rank}""" ) logger.info(F"""Teacher loaded from {args.teacher_name}.""" ) # FREEZING # if args.freeze_pos_embs: freeze_pos_embeddings(a__ , a__ ) if args.freeze_token_type_embds: freeze_token_type_embeddings(a__ , a__ ) # SANITY CHECKS # assert student.config.vocab_size == teacher.config.vocab_size assert student.config.hidden_size == teacher.config.hidden_size assert student.config.max_position_embeddings == teacher.config.max_position_embeddings if args.mlm: assert token_probs.size(0 ) == stu_architecture_config.vocab_size # DISTILLER # torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Union[str, Any] = Distiller( params=a__ , dataset=a__ , token_probs=a__ , student=a__ , teacher=a__ ) distiller.train() logger.info('''Let\'s go get some drinks.''' ) if __name__ == "__main__": main()
313
from maths.prime_check import is_prime def UpperCAmelCase_( a__ ): """simple docstring""" if not isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = F"""Input value of [number={number}] must be an integer""" raise TypeError(a__ ) if is_prime(a__ ) and is_prime(number + 2 ): return number + 2 else: return -1 if __name__ == "__main__": import doctest doctest.testmod()
313
1
import json from typing import Dict, List, Optional, Tuple, Union from tokenizers import pre_tokenizers, processors from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput from ...tokenization_utils_fast import PreTrainedTokenizerFast from ...utils import PaddingStrategy, logging from .tokenization_led import LEDTokenizer a__ : Dict = logging.get_logger(__name__) a__ : Dict = {'''vocab_file''': '''vocab.json''', '''merges_file''': '''merges.txt''', '''tokenizer_file''': '''tokenizer.json'''} a__ : str = { '''vocab_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json''', }, '''merges_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt''', }, '''tokenizer_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json''', }, } a__ : Optional[int] = { '''allenai/led-base-16384''': 16_384, } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = VOCAB_FILES_NAMES __SCREAMING_SNAKE_CASE : Tuple = PRETRAINED_VOCAB_FILES_MAP __SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __SCREAMING_SNAKE_CASE : Union[str, Any] = LEDTokenizer __SCREAMING_SNAKE_CASE : Optional[int] = ['input_ids', 'attention_mask'] def __init__( self , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase="replace" , _lowerCamelCase="<s>" , _lowerCamelCase="</s>" , _lowerCamelCase="</s>" , _lowerCamelCase="<s>" , _lowerCamelCase="<unk>" , _lowerCamelCase="<pad>" , _lowerCamelCase="<mask>" , _lowerCamelCase=False , _lowerCamelCase=True , **_lowerCamelCase , ) ->Union[str, Any]: super().__init__( _lowerCamelCase , _lowerCamelCase , tokenizer_file=_lowerCamelCase , errors=_lowerCamelCase , bos_token=_lowerCamelCase , eos_token=_lowerCamelCase , sep_token=_lowerCamelCase , cls_token=_lowerCamelCase , unk_token=_lowerCamelCase , pad_token=_lowerCamelCase , mask_token=_lowerCamelCase , add_prefix_space=_lowerCamelCase , trim_offsets=_lowerCamelCase , **_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Dict = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() ) if pre_tok_state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : str = getattr(_lowerCamelCase , pre_tok_state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : Optional[int] = add_prefix_space SCREAMING_SNAKE_CASE : str = pre_tok_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = add_prefix_space # the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__` SCREAMING_SNAKE_CASE : List[Any] = '''post_processor''' SCREAMING_SNAKE_CASE : int = getattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) if tokenizer_component_instance: SCREAMING_SNAKE_CASE : Any = json.loads(tokenizer_component_instance.__getstate__() ) # The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class` if "sep" in state: SCREAMING_SNAKE_CASE : Optional[int] = tuple(state['''sep'''] ) if "cls" in state: SCREAMING_SNAKE_CASE : Optional[Any] = tuple(state['''cls'''] ) SCREAMING_SNAKE_CASE : Any = False if state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : Union[str, Any] = add_prefix_space SCREAMING_SNAKE_CASE : Union[str, Any] = True if state.get('''trim_offsets''' , _lowerCamelCase ) != trim_offsets: SCREAMING_SNAKE_CASE : List[Any] = trim_offsets SCREAMING_SNAKE_CASE : Union[str, Any] = True if changes_to_apply: SCREAMING_SNAKE_CASE : List[str] = getattr(_lowerCamelCase , state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : List[Any] = component_class(**_lowerCamelCase ) setattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) @property # Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED def __lowerCAmelCase ( self ) ->str: if self._mask_token is None: if self.verbose: logger.error('''Using mask_token, but it is not set yet.''' ) return None return str(self._mask_token ) @mask_token.setter def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : str = AddedToken(_lowerCamelCase , lstrip=_lowerCamelCase , rstrip=_lowerCamelCase ) if isinstance(_lowerCamelCase , _lowerCamelCase ) else value SCREAMING_SNAKE_CASE : List[Any] = value def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : Tuple = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._batch_encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : List[Any] = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->Tuple[str]: SCREAMING_SNAKE_CASE : Any = self._tokenizer.model.save(_lowerCamelCase , name=_lowerCamelCase ) return tuple(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=None ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id] if token_ids_a is None: return output return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id] def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->List[int]: SCREAMING_SNAKE_CASE : Any = [self.sep_token_id] SCREAMING_SNAKE_CASE : List[str] = [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 __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None , _lowerCamelCase = PaddingStrategy.DO_NOT_PAD , _lowerCamelCase = None , _lowerCamelCase = None , ) ->dict: SCREAMING_SNAKE_CASE : Tuple = super()._pad( encoded_inputs=_lowerCamelCase , max_length=_lowerCamelCase , padding_strategy=_lowerCamelCase , pad_to_multiple_of=_lowerCamelCase , return_attention_mask=_lowerCamelCase , ) # Load from model defaults if return_attention_mask is None: SCREAMING_SNAKE_CASE : Optional[Any] = '''attention_mask''' in self.model_input_names if return_attention_mask and "global_attention_mask" in encoded_inputs: SCREAMING_SNAKE_CASE : int = encoded_inputs[self.model_input_names[0]] # `global_attention_mask` need to have the same length as other (sequential) inputs. SCREAMING_SNAKE_CASE : Tuple = len(encoded_inputs['''global_attention_mask'''] ) != len(_lowerCamelCase ) if needs_to_be_padded: SCREAMING_SNAKE_CASE : int = len(_lowerCamelCase ) - len(encoded_inputs['''global_attention_mask'''] ) if self.padding_side == "right": # Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend` SCREAMING_SNAKE_CASE : str = ( encoded_inputs['''global_attention_mask'''] + [-1] * difference ) elif self.padding_side == "left": SCREAMING_SNAKE_CASE : Optional[Any] = [-1] * difference + encoded_inputs[ '''global_attention_mask''' ] else: raise ValueError('''Invalid padding strategy:''' + str(self.padding_side ) ) return encoded_inputs
313
import gc import random import unittest import numpy as np import torch from PIL import Image from diffusers import ( DDIMScheduler, KandinskyVaaControlnetImgaImgPipeline, KandinskyVaaPriorEmbaEmbPipeline, UNetaDConditionModel, VQModel, ) from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[Any] = KandinskyVaaControlnetImgaImgPipeline __SCREAMING_SNAKE_CASE : Optional[int] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[Any] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[str] = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->Optional[Any]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 32 @property def __lowerCAmelCase ( self ) ->str: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Dict: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Tuple: return 100 @property def __lowerCAmelCase ( self ) ->int: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Union[str, Any] = { '''in_channels''': 8, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''image_hint''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : List[str] = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->Any: return { "block_out_channels": [32, 32, 64, 64], "down_block_types": [ "DownEncoderBlock2D", "DownEncoderBlock2D", "DownEncoderBlock2D", "AttnDownEncoderBlock2D", ], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": ["AttnUpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D"], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : str = self.dummy_unet SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_movq SCREAMING_SNAKE_CASE : List[str] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : str = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = { '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->int: SCREAMING_SNAKE_CASE : str = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(seed + 1 ) ).to( _lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : Dict = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) # create hint SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : List[str] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = { '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''hint''': hint, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[int] = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : Any = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : Union[str, Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array( [0.5_4_9_8_5_0_3_4, 0.5_5_5_0_9_3_6_5, 0.5_2_5_6_1_5_0_4, 0.5_5_7_0_4_9_4, 0.5_5_9_3_8_1_8, 0.5_2_6_3_9_7_9, 0.5_0_2_8_5_6_4_3, 0.5_0_6_9_8_4_6, 0.5_1_1_9_6_7_3_6] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Union[str, Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/kandinskyv22_controlnet_img2img_robotcat_fp16.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = init_image.resize((512, 512) ) SCREAMING_SNAKE_CASE : Dict = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/hint_image_cat.png''' ) SCREAMING_SNAKE_CASE : List[Any] = torch.from_numpy(np.array(_lowerCamelCase ) ).float() / 2_5_5.0 SCREAMING_SNAKE_CASE : int = hint.permute(2 , 0 , 1 ).unsqueeze(0 ) SCREAMING_SNAKE_CASE : List[Any] = '''A robot, 4k photo''' SCREAMING_SNAKE_CASE : List[str] = KandinskyVaaPriorEmbaEmbPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = KandinskyVaaControlnetImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-controlnet-depth''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Any = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , image=_lowerCamelCase , strength=0.8_5 , generator=_lowerCamelCase , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : List[str] = pipeline( image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , hint=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=512 , width=512 , strength=0.5 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = output.images[0] assert image.shape == (512, 512, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
import string import numpy def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return b if a == 0 else greatest_common_divisor(b % a , a__ ) class a_ : """simple docstring""" __SCREAMING_SNAKE_CASE : Any = string.ascii_uppercase + string.digits # This cipher takes alphanumerics into account # i.e. a total of 36 characters # take x and return x % len(key_string) __SCREAMING_SNAKE_CASE : Dict = numpy.vectorize(lambda a__ : x % 36 ) __SCREAMING_SNAKE_CASE : Any = numpy.vectorize(a__ ) def __init__( self , _lowerCamelCase ) ->None: SCREAMING_SNAKE_CASE : Optional[int] = self.modulus(_lowerCamelCase ) # mod36 calc's on the encrypt key self.check_determinant() # validate the determinant of the encryption key SCREAMING_SNAKE_CASE : str = encrypt_key.shape[0] def __lowerCAmelCase ( self , _lowerCamelCase ) ->int: return self.key_string.index(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->str: return self.key_string[round(_lowerCamelCase )] def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Optional[int] = round(numpy.linalg.det(self.encrypt_key ) ) if det < 0: SCREAMING_SNAKE_CASE : Optional[int] = det % len(self.key_string ) SCREAMING_SNAKE_CASE : Dict = len(self.key_string ) if greatest_common_divisor(_lowerCamelCase , len(self.key_string ) ) != 1: SCREAMING_SNAKE_CASE : List[str] = ( F"""determinant modular {req_l} of encryption key({det}) """ F"""is not co prime w.r.t {req_l}.\nTry another key.""" ) raise ValueError(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->str: SCREAMING_SNAKE_CASE : Any = [char for char in text.upper() if char in self.key_string] SCREAMING_SNAKE_CASE : Union[str, Any] = chars[-1] while len(_lowerCamelCase ) % self.break_key != 0: chars.append(_lowerCamelCase ) return "".join(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->str: SCREAMING_SNAKE_CASE : int = self.process_text(text.upper() ) SCREAMING_SNAKE_CASE : Union[str, Any] = '''''' for i in range(0 , len(_lowerCamelCase ) - self.break_key + 1 , self.break_key ): SCREAMING_SNAKE_CASE : Any = text[i : i + self.break_key] SCREAMING_SNAKE_CASE : List[str] = [self.replace_letters(_lowerCamelCase ) for char in batch] SCREAMING_SNAKE_CASE : int = numpy.array([vec] ).T SCREAMING_SNAKE_CASE : str = self.modulus(self.encrypt_key.dot(_lowerCamelCase ) ).T.tolist()[ 0 ] SCREAMING_SNAKE_CASE : Tuple = ''''''.join( self.replace_digits(_lowerCamelCase ) for num in batch_encrypted ) encrypted += encrypted_batch return encrypted def __lowerCAmelCase ( self ) ->numpy.ndarray: SCREAMING_SNAKE_CASE : int = round(numpy.linalg.det(self.encrypt_key ) ) if det < 0: SCREAMING_SNAKE_CASE : Optional[Any] = det % len(self.key_string ) SCREAMING_SNAKE_CASE : Tuple = None for i in range(len(self.key_string ) ): if (det * i) % len(self.key_string ) == 1: SCREAMING_SNAKE_CASE : str = i break SCREAMING_SNAKE_CASE : str = ( det_inv * numpy.linalg.det(self.encrypt_key ) * numpy.linalg.inv(self.encrypt_key ) ) return self.to_int(self.modulus(_lowerCamelCase ) ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->str: SCREAMING_SNAKE_CASE : str = self.make_decrypt_key() SCREAMING_SNAKE_CASE : Union[str, Any] = self.process_text(text.upper() ) SCREAMING_SNAKE_CASE : Tuple = '''''' for i in range(0 , len(_lowerCamelCase ) - self.break_key + 1 , self.break_key ): SCREAMING_SNAKE_CASE : List[Any] = text[i : i + self.break_key] SCREAMING_SNAKE_CASE : Tuple = [self.replace_letters(_lowerCamelCase ) for char in batch] SCREAMING_SNAKE_CASE : Union[str, Any] = numpy.array([vec] ).T SCREAMING_SNAKE_CASE : Optional[Any] = self.modulus(decrypt_key.dot(_lowerCamelCase ) ).T.tolist()[0] SCREAMING_SNAKE_CASE : int = ''''''.join( self.replace_digits(_lowerCamelCase ) for num in batch_decrypted ) decrypted += decrypted_batch return decrypted def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = int(input('''Enter the order of the encryption key: ''' ) ) SCREAMING_SNAKE_CASE : Union[str, Any] = [] print('''Enter each row of the encryption key with space separated integers''' ) for _ in range(a__ ): SCREAMING_SNAKE_CASE : List[Any] = [int(a__ ) for x in input().split()] hill_matrix.append(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = HillCipher(numpy.array(a__ ) ) print('''Would you like to encrypt or decrypt some text? (1 or 2)''' ) SCREAMING_SNAKE_CASE : List[str] = input('''\n1. Encrypt\n2. Decrypt\n''' ) if option == "1": SCREAMING_SNAKE_CASE : Dict = input('''What text would you like to encrypt?: ''' ) print('''Your encrypted text is:''' ) print(hc.encrypt(a__ ) ) elif option == "2": SCREAMING_SNAKE_CASE : Any = input('''What text would you like to decrypt?: ''' ) print('''Your decrypted text is:''' ) print(hc.decrypt(a__ ) ) if __name__ == "__main__": import doctest doctest.testmod() main()
313
from typing import Any, Callable, Dict, List, Optional, Union import torch from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer from diffusers import ( AutoencoderKL, DDIMScheduler, DiffusionPipeline, LMSDiscreteScheduler, PNDMScheduler, StableDiffusionPipeline, UNetaDConditionModel, ) from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker a__ : List[str] = '''CompVis/stable-diffusion-v1-1''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-2''' a__ : Any = '''CompVis/stable-diffusion-v1-3''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-4''' class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , ) ->str: super()._init_() SCREAMING_SNAKE_CASE : Tuple = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = StableDiffusionPipeline( vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , unet=_lowerCamelCase , scheduler=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=_lowerCamelCase , requires_safety_checker=_lowerCamelCase , ) self.register_modules(pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea ) @property def __lowerCAmelCase ( self ) ->Dict[str, Any]: return {k: getattr(self , _lowerCamelCase ) for k in self.config.keys() if not k.startswith('''_''' )} def __lowerCAmelCase ( self , _lowerCamelCase = "auto" ) ->Optional[int]: if slice_size == "auto": # half the attention head size is usually a good trade-off between # speed and memory SCREAMING_SNAKE_CASE : Optional[int] = self.unet.config.attention_head_dim // 2 self.unet.set_attention_slice(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: self.enable_attention_slicing(_lowerCamelCase ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->str: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Tuple: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Optional[Any]: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: SCREAMING_SNAKE_CASE : Optional[Any] = '''cuda''' if torch.cuda.is_available() else '''cpu''' self.to(_lowerCamelCase ) # Checks if the height and width are divisible by 8 or not if height % 8 != 0 or width % 8 != 0: raise ValueError(F"""`height` and `width` must be divisible by 8 but are {height} and {width}.""" ) # Get first result from Stable Diffusion Checkpoint v1.1 SCREAMING_SNAKE_CASE : Optional[Any] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.2 SCREAMING_SNAKE_CASE : Any = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.3 SCREAMING_SNAKE_CASE : Optional[int] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.4 SCREAMING_SNAKE_CASE : str = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get all result images into a single list and pass it via StableDiffusionPipelineOutput for final result return StableDiffusionPipelineOutput([resa[0], resa[0], resa[0], resa[0]] )
313
1
import os from itertools import chain from random import randrange, shuffle import pytest from .sola import PokerHand a__ : int = ( '''4S 3H 2C 7S 5H''', '''9D 8H 2C 6S 7H''', '''2D 6D 9D TH 7D''', '''TC 8C 2S JH 6C''', '''JH 8S TH AH QH''', '''TS KS 5S 9S AC''', '''KD 6S 9D TH AD''', '''KS 8D 4D 9S 4S''', # pair '''8C 4S KH JS 4D''', # pair '''QH 8H KD JH 8S''', # pair '''KC 4H KS 2H 8D''', # pair '''KD 4S KC 3H 8S''', # pair '''AH 8S AS KC JH''', # pair '''3H 4C 4H 3S 2H''', # 2 pairs '''5S 5D 2C KH KH''', # 2 pairs '''3C KH 5D 5S KH''', # 2 pairs '''AS 3C KH AD KH''', # 2 pairs '''7C 7S 3S 7H 5S''', # 3 of a kind '''7C 7S KH 2H 7H''', # 3 of a kind '''AC KH QH AH AS''', # 3 of a kind '''2H 4D 3C AS 5S''', # straight (low ace) '''3C 5C 4C 2C 6H''', # straight '''6S 8S 7S 5H 9H''', # straight '''JS QS 9H TS KH''', # straight '''QC KH TS JS AH''', # straight (high ace) '''8C 9C 5C 3C TC''', # flush '''3S 8S 9S 5S KS''', # flush '''4C 5C 9C 8C KC''', # flush '''JH 8H AH KH QH''', # flush '''3D 2H 3H 2C 2D''', # full house '''2H 2C 3S 3H 3D''', # full house '''KH KC 3S 3H 3D''', # full house '''JC 6H JS JD JH''', # 4 of a kind '''JC 7H JS JD JH''', # 4 of a kind '''JC KH JS JD JH''', # 4 of a kind '''2S AS 4S 5S 3S''', # straight flush (low ace) '''2D 6D 3D 4D 5D''', # straight flush '''5C 6C 3C 7C 4C''', # straight flush '''JH 9H TH KH QH''', # straight flush '''JH AH TH KH QH''', # royal flush (high ace straight flush) ) a__ : str = ( ('''2H 3H 4H 5H 6H''', '''KS AS TS QS JS''', '''Loss'''), ('''2H 3H 4H 5H 6H''', '''AS AD AC AH JD''', '''Win'''), ('''AS AH 2H AD AC''', '''JS JD JC JH 3D''', '''Win'''), ('''2S AH 2H AS AC''', '''JS JD JC JH AD''', '''Loss'''), ('''2S AH 2H AS AC''', '''2H 3H 5H 6H 7H''', '''Win'''), ('''AS 3S 4S 8S 2S''', '''2H 3H 5H 6H 7H''', '''Win'''), ('''2H 3H 5H 6H 7H''', '''2S 3H 4H 5S 6C''', '''Win'''), ('''2S 3H 4H 5S 6C''', '''3D 4C 5H 6H 2S''', '''Tie'''), ('''2S 3H 4H 5S 6C''', '''AH AC 5H 6H AS''', '''Win'''), ('''2S 2H 4H 5S 4C''', '''AH AC 5H 6H AS''', '''Loss'''), ('''2S 2H 4H 5S 4C''', '''AH AC 5H 6H 7S''', '''Win'''), ('''6S AD 7H 4S AS''', '''AH AC 5H 6H 7S''', '''Loss'''), ('''2S AH 4H 5S KC''', '''AH AC 5H 6H 7S''', '''Loss'''), ('''2S 3H 6H 7S 9C''', '''7H 3C TH 6H 9S''', '''Loss'''), ('''4S 5H 6H TS AC''', '''3S 5H 6H TS AC''', '''Win'''), ('''2S AH 4H 5S 6C''', '''AD 4C 5H 6H 2C''', '''Tie'''), ('''AS AH 3H AD AC''', '''AS AH 2H AD AC''', '''Win'''), ('''AH AC 5H 5C QS''', '''AH AC 5H 5C KS''', '''Loss'''), ('''AH AC 5H 5C QS''', '''KH KC 5H 5C QS''', '''Win'''), ('''7C 7S KH 2H 7H''', '''3C 3S AH 2H 3H''', '''Win'''), ('''3C 3S AH 2H 3H''', '''7C 7S KH 2H 7H''', '''Loss'''), ('''6H 5H 4H 3H 2H''', '''5H 4H 3H 2H AH''', '''Win'''), ('''5H 4H 3H 2H AH''', '''5H 4H 3H 2H AH''', '''Tie'''), ('''5H 4H 3H 2H AH''', '''6H 5H 4H 3H 2H''', '''Loss'''), ('''AH AD KS KC AC''', '''AH KD KH AC KC''', '''Win'''), ('''2H 4D 3C AS 5S''', '''2H 4D 3C 6S 5S''', '''Loss'''), ('''2H 3S 3C 3H 2S''', '''3S 3C 2S 2H 2D''', '''Win'''), ('''4D 6D 5D 2D JH''', '''3S 8S 3H TC KH''', '''Loss'''), ('''4S 6C 8S 3S 7S''', '''AD KS 2D 7D 7C''', '''Loss'''), ('''6S 4C 7H 8C 3H''', '''5H JC AH 9D 9C''', '''Loss'''), ('''9D 9H JH TC QH''', '''3C 2S JS 5C 7H''', '''Win'''), ('''2H TC 8S AD 9S''', '''4H TS 7H 2C 5C''', '''Win'''), ('''9D 3S 2C 7S 7C''', '''JC TD 3C TC 9H''', '''Loss'''), ) a__ : List[Any] = ( ('''2H 3H 4H 5H 6H''', True), ('''AS AH 2H AD AC''', False), ('''2H 3H 5H 6H 7H''', True), ('''KS AS TS QS JS''', True), ('''8H 9H QS JS TH''', False), ('''AS 3S 4S 8S 2S''', True), ) a__ : Union[str, Any] = ( ('''2H 3H 4H 5H 6H''', True), ('''AS AH 2H AD AC''', False), ('''2H 3H 5H 6H 7H''', False), ('''KS AS TS QS JS''', True), ('''8H 9H QS JS TH''', True), ) a__ : Any = ( ('''2H 4D 3C AS 5S''', True, [5, 4, 3, 2, 14]), ('''2H 5D 3C AS 5S''', False, [14, 5, 5, 3, 2]), ('''JH QD KC AS TS''', False, [14, 13, 12, 11, 10]), ('''9D 3S 2C 7S 7C''', False, [9, 7, 7, 3, 2]), ) a__ : List[Any] = ( ('''JH AH TH KH QH''', 0), ('''JH 9H TH KH QH''', 0), ('''JC KH JS JD JH''', 7), ('''KH KC 3S 3H 3D''', 6), ('''8C 9C 5C 3C TC''', 0), ('''JS QS 9H TS KH''', 0), ('''7C 7S KH 2H 7H''', 3), ('''3C KH 5D 5S KH''', 2), ('''QH 8H KD JH 8S''', 1), ('''2D 6D 9D TH 7D''', 0), ) a__ : Optional[Any] = ( ('''JH AH TH KH QH''', 23), ('''JH 9H TH KH QH''', 22), ('''JC KH JS JD JH''', 21), ('''KH KC 3S 3H 3D''', 20), ('''8C 9C 5C 3C TC''', 19), ('''JS QS 9H TS KH''', 18), ('''7C 7S KH 2H 7H''', 17), ('''3C KH 5D 5S KH''', 16), ('''QH 8H KD JH 8S''', 15), ('''2D 6D 9D TH 7D''', 14), ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = randrange(len(a__ ) ), randrange(len(a__ ) ) SCREAMING_SNAKE_CASE : Optional[Any] = ['''Loss''', '''Tie''', '''Win'''][(play >= oppo) + (play > oppo)] SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = SORTED_HANDS[play], SORTED_HANDS[oppo] return hand, other, expected def UpperCAmelCase_( a__ = 100 ): """simple docstring""" return (generate_random_hand() for _ in range(a__ )) @pytest.mark.parametrize('''hand, expected''' , a__ ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" assert PokerHand(a__ )._is_flush() == expected @pytest.mark.parametrize('''hand, expected''' , a__ ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" assert PokerHand(a__ )._is_straight() == expected @pytest.mark.parametrize('''hand, expected, card_values''' , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = PokerHand(a__ ) assert player._is_five_high_straight() == expected assert player._card_values == card_values @pytest.mark.parametrize('''hand, expected''' , a__ ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" assert PokerHand(a__ )._is_same_kind() == expected @pytest.mark.parametrize('''hand, expected''' , a__ ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" assert PokerHand(a__ )._hand_type == expected @pytest.mark.parametrize('''hand, other, expected''' , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" assert PokerHand(a__ ).compare_with(PokerHand(a__ ) ) == expected @pytest.mark.parametrize('''hand, other, expected''' , generate_random_hands() ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" assert PokerHand(a__ ).compare_with(PokerHand(a__ ) ) == expected def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = [PokerHand(a__ ) for hand in SORTED_HANDS] SCREAMING_SNAKE_CASE : List[str] = poker_hands.copy() shuffle(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = chain(sorted(a__ ) ) for index, hand in enumerate(a__ ): assert hand == poker_hands[index] def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [PokerHand('''2D AC 3H 4H 5S''' ), PokerHand('''2S 3H 4H 5S 6C''' )] pokerhands.sort(reverse=a__ ) assert pokerhands[0].__str__() == "2S 3H 4H 5S 6C" def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : str = PokerHand('''2C 4S AS 3D 5C''' ) SCREAMING_SNAKE_CASE : str = True SCREAMING_SNAKE_CASE : str = [5, 4, 3, 2, 14] for _ in range(10 ): assert pokerhand._is_five_high_straight() == expected assert pokerhand._card_values == expected_card_values def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = 0 SCREAMING_SNAKE_CASE : str = os.path.abspath(os.path.dirname(a__ ) ) SCREAMING_SNAKE_CASE : List[str] = os.path.join(a__ , '''poker_hands.txt''' ) with open(a__ ) as file_hand: for line in file_hand: SCREAMING_SNAKE_CASE : Union[str, Any] = line[:14].strip() SCREAMING_SNAKE_CASE : Any = line[15:].strip() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = PokerHand(a__ ), PokerHand(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = player.compare_with(a__ ) if output == "Win": answer += 1 assert answer == 376
313
from typing import Optional, Tuple, Union import flax import flax.linen as nn import jax import jax.numpy as jnp from flax.core.frozen_dict import FrozenDict from ..configuration_utils import ConfigMixin, flax_register_to_config from ..utils import BaseOutput from .embeddings_flax import FlaxTimestepEmbedding, FlaxTimesteps from .modeling_flax_utils import FlaxModelMixin from .unet_ad_blocks_flax import ( FlaxCrossAttnDownBlockaD, FlaxCrossAttnUpBlockaD, FlaxDownBlockaD, FlaxUNetMidBlockaDCrossAttn, FlaxUpBlockaD, ) @flax.struct.dataclass class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : jnp.ndarray @flax_register_to_config class a_ ( nn.Module , a__ , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = 32 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : Tuple[str] = ( "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D", ) __SCREAMING_SNAKE_CASE : Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D") __SCREAMING_SNAKE_CASE : Union[bool, Tuple[bool]] = False __SCREAMING_SNAKE_CASE : Tuple[int] = (320, 640, 1280, 1280) __SCREAMING_SNAKE_CASE : int = 2 __SCREAMING_SNAKE_CASE : Union[int, Tuple[int]] = 8 __SCREAMING_SNAKE_CASE : Optional[Union[int, Tuple[int]]] = None __SCREAMING_SNAKE_CASE : int = 1280 __SCREAMING_SNAKE_CASE : float = 0.0 __SCREAMING_SNAKE_CASE : bool = False __SCREAMING_SNAKE_CASE : jnp.dtype = jnp.floataa __SCREAMING_SNAKE_CASE : bool = True __SCREAMING_SNAKE_CASE : int = 0 __SCREAMING_SNAKE_CASE : bool = False def __lowerCAmelCase ( self , _lowerCamelCase ) ->FrozenDict: # init input tensors SCREAMING_SNAKE_CASE : List[Any] = (1, self.in_channels, self.sample_size, self.sample_size) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros(_lowerCamelCase , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Tuple = jnp.ones((1,) , dtype=jnp.intaa ) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros((1, 1, self.cross_attention_dim) , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = jax.random.split(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = {'''params''': params_rng, '''dropout''': dropout_rng} return self.init(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase )["params"] def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : List[str] = self.block_out_channels SCREAMING_SNAKE_CASE : Optional[int] = block_out_channels[0] * 4 if self.num_attention_heads is not None: raise ValueError( '''At the moment it is not possible to define the number of attention heads via `num_attention_heads` because of a naming issue as described in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131. Passing `num_attention_heads` will only be supported in diffusers v0.19.''' ) # If `num_attention_heads` is not defined (which is the case for most models) # it will default to `attention_head_dim`. This looks weird upon first reading it and it is. # The reason for this behavior is to correct for incorrectly named variables that were introduced # when this library was created. The incorrect naming was only discovered much later in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131 # Changing `attention_head_dim` to `num_attention_heads` for 40,000+ configurations is too backwards breaking # which is why we correct for the naming here. SCREAMING_SNAKE_CASE : List[str] = self.num_attention_heads or self.attention_head_dim # input SCREAMING_SNAKE_CASE : Optional[int] = nn.Conv( block_out_channels[0] , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) # time SCREAMING_SNAKE_CASE : Tuple = FlaxTimesteps( block_out_channels[0] , flip_sin_to_cos=self.flip_sin_to_cos , freq_shift=self.config.freq_shift ) SCREAMING_SNAKE_CASE : Dict = FlaxTimestepEmbedding(_lowerCamelCase , dtype=self.dtype ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.only_cross_attention if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (only_cross_attention,) * len(self.down_block_types ) if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (num_attention_heads,) * len(self.down_block_types ) # down SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Optional[Any] = block_out_channels[0] for i, down_block_type in enumerate(self.down_block_types ): SCREAMING_SNAKE_CASE : str = output_channel SCREAMING_SNAKE_CASE : int = block_out_channels[i] SCREAMING_SNAKE_CASE : List[Any] = i == len(_lowerCamelCase ) - 1 if down_block_type == "CrossAttnDownBlock2D": SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxCrossAttnDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , num_attention_heads=num_attention_heads[i] , add_downsample=not is_final_block , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , add_downsample=not is_final_block , dtype=self.dtype , ) down_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = down_blocks # mid SCREAMING_SNAKE_CASE : int = FlaxUNetMidBlockaDCrossAttn( in_channels=block_out_channels[-1] , dropout=self.dropout , num_attention_heads=num_attention_heads[-1] , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) # up SCREAMING_SNAKE_CASE : str = [] SCREAMING_SNAKE_CASE : str = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = reversed_block_out_channels[0] for i, up_block_type in enumerate(self.up_block_types ): SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Union[str, Any] = reversed_block_out_channels[i] SCREAMING_SNAKE_CASE : Tuple = reversed_block_out_channels[min(i + 1 , len(_lowerCamelCase ) - 1 )] SCREAMING_SNAKE_CASE : Dict = i == len(_lowerCamelCase ) - 1 if up_block_type == "CrossAttnUpBlock2D": SCREAMING_SNAKE_CASE : str = FlaxCrossAttnUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , num_attention_heads=reversed_num_attention_heads[i] , add_upsample=not is_final_block , dropout=self.dropout , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Optional[int] = FlaxUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , add_upsample=not is_final_block , dropout=self.dropout , dtype=self.dtype , ) up_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Tuple = up_blocks # out SCREAMING_SNAKE_CASE : Any = nn.GroupNorm(num_groups=32 , epsilon=1e-5 ) SCREAMING_SNAKE_CASE : Any = nn.Conv( self.out_channels , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) def __call__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase = True , _lowerCamelCase = False , ) ->Union[FlaxUNetaDConditionOutput, Tuple]: # 1. time if not isinstance(_lowerCamelCase , jnp.ndarray ): SCREAMING_SNAKE_CASE : int = jnp.array([timesteps] , dtype=jnp.intaa ) elif isinstance(_lowerCamelCase , jnp.ndarray ) and len(timesteps.shape ) == 0: SCREAMING_SNAKE_CASE : List[str] = timesteps.astype(dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.expand_dims(_lowerCamelCase , 0 ) SCREAMING_SNAKE_CASE : List[str] = self.time_proj(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = self.time_embedding(_lowerCamelCase ) # 2. pre-process SCREAMING_SNAKE_CASE : int = jnp.transpose(_lowerCamelCase , (0, 2, 3, 1) ) SCREAMING_SNAKE_CASE : List[Any] = self.conv_in(_lowerCamelCase ) # 3. down SCREAMING_SNAKE_CASE : Optional[int] = (sample,) for down_block in self.down_blocks: if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , deterministic=not train ) down_block_res_samples += res_samples if down_block_additional_residuals is not None: SCREAMING_SNAKE_CASE : int = () for down_block_res_sample, down_block_additional_residual in zip( _lowerCamelCase , _lowerCamelCase ): down_block_res_sample += down_block_additional_residual new_down_block_res_samples += (down_block_res_sample,) SCREAMING_SNAKE_CASE : Dict = new_down_block_res_samples # 4. mid SCREAMING_SNAKE_CASE : Optional[Any] = self.mid_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) if mid_block_additional_residual is not None: sample += mid_block_additional_residual # 5. up for up_block in self.up_blocks: SCREAMING_SNAKE_CASE : Optional[Any] = down_block_res_samples[-(self.layers_per_block + 1) :] SCREAMING_SNAKE_CASE : Optional[int] = down_block_res_samples[: -(self.layers_per_block + 1)] if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = up_block( _lowerCamelCase , temb=_lowerCamelCase , encoder_hidden_states=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train , ) else: SCREAMING_SNAKE_CASE : Optional[int] = up_block(_lowerCamelCase , temb=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train ) # 6. post-process SCREAMING_SNAKE_CASE : Optional[int] = self.conv_norm_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.silu(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self.conv_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.transpose(_lowerCamelCase , (0, 3, 1, 2) ) if not return_dict: return (sample,) return FlaxUNetaDConditionOutput(sample=_lowerCamelCase )
313
1
def UpperCAmelCase_( a__ = 100 ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = set() SCREAMING_SNAKE_CASE : Optional[Any] = 0 SCREAMING_SNAKE_CASE : Optional[int] = n + 1 # maximum limit for a in range(2 , a__ ): for b in range(2 , a__ ): SCREAMING_SNAKE_CASE : str = a**b # calculates the current power collect_powers.add(a__ ) # adds the result to the set return len(a__ ) if __name__ == "__main__": print('''Number of terms ''', solution(int(str(input()).strip())))
313
import gc import random import unittest import numpy as np import torch from transformers import ( CLIPImageProcessor, CLIPTextConfig, CLIPTextModel, CLIPTokenizer, CLIPVisionConfig, CLIPVisionModelWithProjection, ) from diffusers import AutoencoderKL, DDIMScheduler, DDPMScheduler, StableUnCLIPImgaImgPipeline, UNetaDConditionModel from diffusers.pipelines.pipeline_utils import DiffusionPipeline from diffusers.pipelines.stable_diffusion.stable_unclip_image_normalizer import StableUnCLIPImageNormalizer from diffusers.utils.import_utils import is_xformers_available from diffusers.utils.testing_utils import ( enable_full_determinism, floats_tensor, load_image, load_numpy, require_torch_gpu, skip_mps, slow, torch_device, ) from ..pipeline_params import TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS from ..test_pipelines_common import ( PipelineKarrasSchedulerTesterMixin, PipelineLatentTesterMixin, PipelineTesterMixin, assert_mean_pixel_difference, ) enable_full_determinism() class a_ ( a__ , a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = StableUnCLIPImgaImgPipeline __SCREAMING_SNAKE_CASE : List[str] = TEXT_GUIDED_IMAGE_VARIATION_PARAMS __SCREAMING_SNAKE_CASE : List[Any] = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS __SCREAMING_SNAKE_CASE : Any = frozenset( [] ) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess __SCREAMING_SNAKE_CASE : Tuple = frozenset([] ) def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[Any] = 32 SCREAMING_SNAKE_CASE : Tuple = embedder_hidden_size # image encoding components SCREAMING_SNAKE_CASE : int = CLIPImageProcessor(crop_size=32 , size=32 ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = CLIPVisionModelWithProjection( CLIPVisionConfig( hidden_size=_lowerCamelCase , projection_dim=_lowerCamelCase , num_hidden_layers=5 , num_attention_heads=4 , image_size=32 , intermediate_size=37 , patch_size=1 , ) ) # regular denoising components torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = StableUnCLIPImageNormalizer(embedding_dim=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = DDPMScheduler(beta_schedule='''squaredcos_cap_v2''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = CLIPTextModel( CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=_lowerCamelCase , projection_dim=32 , intermediate_size=37 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , ) ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = UNetaDConditionModel( sample_size=32 , in_channels=4 , out_channels=4 , down_block_types=('''CrossAttnDownBlock2D''', '''DownBlock2D''') , up_block_types=('''UpBlock2D''', '''CrossAttnUpBlock2D''') , block_out_channels=(32, 64) , attention_head_dim=(2, 4) , class_embed_type='''projection''' , projection_class_embeddings_input_dim=embedder_projection_dim * 2 , cross_attention_dim=_lowerCamelCase , layers_per_block=1 , upcast_attention=_lowerCamelCase , use_linear_projection=_lowerCamelCase , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = DDIMScheduler( beta_schedule='''scaled_linear''' , beta_start=0.0_0_0_8_5 , beta_end=0.0_1_2 , prediction_type='''v_prediction''' , set_alpha_to_one=_lowerCamelCase , steps_offset=1 , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Dict = AutoencoderKL() SCREAMING_SNAKE_CASE : Optional[Any] = { # image encoding components '''feature_extractor''': feature_extractor, '''image_encoder''': image_encoder.eval(), # image noising components '''image_normalizer''': image_normalizer.eval(), '''image_noising_scheduler''': image_noising_scheduler, # regular denoising components '''tokenizer''': tokenizer, '''text_encoder''': text_encoder.eval(), '''unet''': unet.eval(), '''scheduler''': scheduler, '''vae''': vae.eval(), } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 , _lowerCamelCase=True ) ->Optional[int]: if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, 3, 32, 32) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if pil_image: SCREAMING_SNAKE_CASE : Any = input_image * 0.5 + 0.5 SCREAMING_SNAKE_CASE : int = input_image.clamp(0 , 1 ) SCREAMING_SNAKE_CASE : Union[str, Any] = input_image.cpu().permute(0 , 2 , 3 , 1 ).float().numpy() SCREAMING_SNAKE_CASE : List[str] = DiffusionPipeline.numpy_to_pil(_lowerCamelCase )[0] return { "prompt": "An anime racoon running a marathon", "image": input_image, "generator": generator, "num_inference_steps": 2, "output_type": "np", } @skip_mps def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Any = '''cpu''' # ensure determinism for the device-dependent torch.Generator SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = StableUnCLIPImgaImgPipeline(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = sd_pipe.to(_lowerCamelCase ) sd_pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = self.get_dummy_inputs(_lowerCamelCase ) inputs.update({'''image_embeds''': None} ) SCREAMING_SNAKE_CASE : Optional[int] = sd_pipe(**_lowerCamelCase ).images SCREAMING_SNAKE_CASE : Dict = image[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) SCREAMING_SNAKE_CASE : Tuple = np.array([0.3_8_7_2, 0.7_2_2_4, 0.5_6_0_1, 0.4_7_4_1, 0.6_8_7_2, 0.5_8_1_4, 0.4_6_3_6, 0.3_8_6_7, 0.5_0_7_8] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : str = torch_device in ['''cpu''', '''mps'''] self._test_attention_slicing_forward_pass(test_max_difference=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Tuple = torch_device in ['''cpu''', '''mps'''] self._test_inference_batch_single_identical(test_max_difference=_lowerCamelCase ) @unittest.skipIf( torch_device != '''cuda''' or not is_xformers_available() , reason='''XFormers attention is only available with CUDA and `xformers` installed''' , ) def __lowerCAmelCase ( self ) ->Optional[int]: self._test_xformers_attention_forwardGenerator_pass(test_max_difference=_lowerCamelCase ) @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->int: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : List[Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_l_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[Any] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-l-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : List[str] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[int] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_h_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[str] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) torch.cuda.empty_cache() torch.cuda.reset_max_memory_allocated() torch.cuda.reset_peak_memory_stats() SCREAMING_SNAKE_CASE : str = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Dict = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Dict = pipe( _lowerCamelCase , '''anime turtle''' , num_inference_steps=2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = torch.cuda.max_memory_allocated() # make sure that less than 7 GB is allocated assert mem_bytes < 7 * 10**9
313
1
import math def UpperCAmelCase_( a__ , a__ = 0 , a__ = 0 ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = end or len(a__ ) for i in range(a__ , a__ ): SCREAMING_SNAKE_CASE : int = i SCREAMING_SNAKE_CASE : str = array[i] while temp_index != start and temp_index_value < array[temp_index - 1]: SCREAMING_SNAKE_CASE : List[Any] = array[temp_index - 1] temp_index -= 1 SCREAMING_SNAKE_CASE : Any = temp_index_value return array def UpperCAmelCase_( a__ , a__ , a__ ): # Max Heap """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = index SCREAMING_SNAKE_CASE : Union[str, Any] = 2 * index + 1 # Left Node SCREAMING_SNAKE_CASE : int = 2 * index + 2 # Right Node if left_index < heap_size and array[largest] < array[left_index]: SCREAMING_SNAKE_CASE : Optional[int] = left_index if right_index < heap_size and array[largest] < array[right_index]: SCREAMING_SNAKE_CASE : str = right_index if largest != index: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = array[largest], array[index] heapify(a__ , a__ , a__ ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = len(a__ ) for i in range(n // 2 , -1 , -1 ): heapify(a__ , a__ , a__ ) for i in range(n - 1 , 0 , -1 ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = array[0], array[i] heapify(a__ , 0 , a__ ) return array def UpperCAmelCase_( a__ , a__ , a__ , a__ ): """simple docstring""" if (array[first_index] > array[middle_index]) != ( array[first_index] > array[last_index] ): return array[first_index] elif (array[middle_index] > array[first_index]) != ( array[middle_index] > array[last_index] ): return array[middle_index] else: return array[last_index] def UpperCAmelCase_( a__ , a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = low SCREAMING_SNAKE_CASE : Any = high while True: while array[i] < pivot: i += 1 j -= 1 while pivot < array[j]: j -= 1 if i >= j: return i SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = array[j], array[i] i += 1 def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) == 0: return array SCREAMING_SNAKE_CASE : Dict = 2 * math.ceil(math.loga(len(a__ ) ) ) SCREAMING_SNAKE_CASE : Any = 16 return intro_sort(a__ , 0 , len(a__ ) , a__ , a__ ) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ ): """simple docstring""" while end - start > size_threshold: if max_depth == 0: return heap_sort(a__ ) max_depth -= 1 SCREAMING_SNAKE_CASE : List[str] = median_of_a(a__ , a__ , start + ((end - start) // 2) + 1 , end - 1 ) SCREAMING_SNAKE_CASE : int = partition(a__ , a__ , a__ , a__ ) intro_sort(a__ , a__ , a__ , a__ , a__ ) SCREAMING_SNAKE_CASE : Any = p return insertion_sort(a__ , a__ , a__ ) if __name__ == "__main__": import doctest doctest.testmod() a__ : Optional[Any] = input('''Enter numbers separated by a comma : ''').strip() a__ : str = [float(item) for item in user_input.split(''',''')] print(sort(unsorted))
313
from abc import ABC, abstractmethod from typing import List, Optional class a_ ( a__ ): """simple docstring""" def __init__( self ) ->List[str]: # test for the above condition self.test() def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = False while not completed: if counter == 1: self.reset() SCREAMING_SNAKE_CASE : List[Any] = self.advance() if not self.does_advance(_lowerCamelCase ): raise Exception( '''Custom Constraint is not defined correctly. self.does_advance(self.advance()) must be true.''' ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.update(_lowerCamelCase ) counter += 1 if counter > 1_0000: raise Exception('''update() does not fulfill the constraint.''' ) if self.remaining() != 0: raise Exception('''Custom Constraint is not defined correctly.''' ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[int]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Union[str, Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Any: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->int: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`token_ids` has to be a non-empty list, but is {token_ids}.""" ) if any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ): raise ValueError(F"""Each list in `token_ids` has to be a list of positive integers, but is {token_ids}.""" ) SCREAMING_SNAKE_CASE : Optional[Any] = token_ids SCREAMING_SNAKE_CASE : Union[str, Any] = len(self.token_ids ) SCREAMING_SNAKE_CASE : Any = -1 # the index of the currently fulfilled step SCREAMING_SNAKE_CASE : Any = False def __lowerCAmelCase ( self ) ->List[Any]: if self.completed: return None return self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) if self.completed: return False return token_id == self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[str]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : str = False SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = False if self.does_advance(_lowerCamelCase ): self.fulfilled_idx += 1 SCREAMING_SNAKE_CASE : str = True if self.fulfilled_idx == (self.seqlen - 1): SCREAMING_SNAKE_CASE : Any = True SCREAMING_SNAKE_CASE : Union[str, Any] = completed else: # failed to make progress. SCREAMING_SNAKE_CASE : Dict = True self.reset() return stepped, completed, reset def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = 0 def __lowerCAmelCase ( self ) ->Any: return self.seqlen - (self.fulfilled_idx + 1) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Dict: SCREAMING_SNAKE_CASE : Any = PhrasalConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : Dict = self.seqlen SCREAMING_SNAKE_CASE : int = self.fulfilled_idx SCREAMING_SNAKE_CASE : Tuple = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=True ) ->Dict: SCREAMING_SNAKE_CASE : Any = max([len(_lowerCamelCase ) for one in nested_token_ids] ) SCREAMING_SNAKE_CASE : List[str] = {} for token_ids in nested_token_ids: SCREAMING_SNAKE_CASE : Optional[Any] = root for tidx, token_id in enumerate(_lowerCamelCase ): if token_id not in level: SCREAMING_SNAKE_CASE : Any = {} SCREAMING_SNAKE_CASE : Tuple = level[token_id] if no_subsets and self.has_subsets(_lowerCamelCase , _lowerCamelCase ): raise ValueError( '''Each list in `nested_token_ids` can\'t be a complete subset of another list, but is''' F""" {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = root def __lowerCAmelCase ( self , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = self.trie for current_token in current_seq: SCREAMING_SNAKE_CASE : int = start[current_token] SCREAMING_SNAKE_CASE : Optional[int] = list(start.keys() ) return next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.next_tokens(_lowerCamelCase ) return len(_lowerCamelCase ) == 0 def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = list(root.values() ) if len(_lowerCamelCase ) == 0: return 1 else: return sum([self.count_leaves(_lowerCamelCase ) for nn in next_nodes] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = self.count_leaves(_lowerCamelCase ) return len(_lowerCamelCase ) != leaf_count class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->str: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`nested_token_ids` has to be a non-empty list, but is {nested_token_ids}.""" ) if any(not isinstance(_lowerCamelCase , _lowerCamelCase ) for token_ids in nested_token_ids ): raise ValueError(F"""`nested_token_ids` has to be a list of lists, but is {nested_token_ids}.""" ) if any( any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ) for token_ids in nested_token_ids ): raise ValueError( F"""Each list in `nested_token_ids` has to be a list of positive integers, but is {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = DisjunctiveTrie(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = nested_token_ids SCREAMING_SNAKE_CASE : Optional[int] = self.trie.max_height SCREAMING_SNAKE_CASE : Union[str, Any] = [] SCREAMING_SNAKE_CASE : Optional[int] = False def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : str = self.trie.next_tokens(self.current_seq ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : List[str] = self.trie.next_tokens(self.current_seq ) return token_id in next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : int = False SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = False if self.does_advance(_lowerCamelCase ): self.current_seq.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = True else: SCREAMING_SNAKE_CASE : Dict = True self.reset() SCREAMING_SNAKE_CASE : Any = self.trie.reached_leaf(self.current_seq ) SCREAMING_SNAKE_CASE : List[Any] = completed return stepped, completed, reset def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = [] def __lowerCAmelCase ( self ) ->Optional[Any]: if self.completed: # since this can be completed without reaching max height return 0 else: return self.seqlen - len(self.current_seq ) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->List[str]: SCREAMING_SNAKE_CASE : str = DisjunctiveConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : str = self.seqlen SCREAMING_SNAKE_CASE : int = self.current_seq SCREAMING_SNAKE_CASE : Optional[int] = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[Any] = constraints # max # of steps required to fulfill a given constraint SCREAMING_SNAKE_CASE : str = max([c.seqlen for c in constraints] ) SCREAMING_SNAKE_CASE : List[str] = len(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = False self.init_state() def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Any = [] SCREAMING_SNAKE_CASE : List[Any] = None SCREAMING_SNAKE_CASE : Tuple = [constraint.copy(stateful=_lowerCamelCase ) for constraint in self.constraints] def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = 0 if self.inprogress_constraint: # extra points for having a constraint mid-fulfilled add += self.max_seqlen - self.inprogress_constraint.remaining() return (len(self.complete_constraints ) * self.max_seqlen) + add def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Tuple = [] if self.inprogress_constraint is None: for constraint in self.pending_constraints: # "pending" == "unfulfilled yet" SCREAMING_SNAKE_CASE : Optional[int] = constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = self.inprogress_constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: self.init_state() if token_ids is not None: for token in token_ids: # completes or steps **one** constraint SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.add(_lowerCamelCase ) # the entire list of constraints are fulfilled if self.completed: break def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` should be an `int`, but is `{token_id}`.""" ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = False, False if self.completed: SCREAMING_SNAKE_CASE : List[str] = True SCREAMING_SNAKE_CASE : Optional[int] = False return complete, stepped if self.inprogress_constraint is not None: # In the middle of fulfilling a constraint. If the `token_id` *does* makes an incremental progress to current # job, simply update the state SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.inprogress_constraint.update(_lowerCamelCase ) if reset: # 1. If the next token breaks the progress, then we must restart. # e.g. constraint = "I love pies" and sequence so far is "I love" but `token_id` == "books". # But that doesn't mean we self.init_state(), since we only reset the state for this particular # constraint, not the full list of constraints. self.pending_constraints.append(self.inprogress_constraint.copy(stateful=_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = None if complete: # 2. If the next token completes the constraint, move it to completed list, set # inprogress to None. If there are no pending constraints either, then this full list of constraints # is complete. self.complete_constraints.append(self.inprogress_constraint ) SCREAMING_SNAKE_CASE : str = None if len(self.pending_constraints ) == 0: # we're done! SCREAMING_SNAKE_CASE : Optional[Any] = True else: # Not in the middle of fulfilling a constraint. So does this `token_id` helps us step towards any of our list # of constraints? for cidx, pending_constraint in enumerate(self.pending_constraints ): if pending_constraint.does_advance(_lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = pending_constraint.update(_lowerCamelCase ) if not stepped: raise Exception( '''`constraint.update(token_id)` is not yielding incremental progress, ''' '''even though `constraint.does_advance(token_id)` is true.''' ) if complete: self.complete_constraints.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = None if not complete and stepped: SCREAMING_SNAKE_CASE : Optional[Any] = pending_constraint if complete or stepped: # If we made any progress at all, then it's at least not a "pending constraint". SCREAMING_SNAKE_CASE : Union[str, Any] = ( self.pending_constraints[:cidx] + self.pending_constraints[cidx + 1 :] ) if len(self.pending_constraints ) == 0 and self.inprogress_constraint is None: # If there's no longer any pending after this and no inprogress either, then we must be # complete. SCREAMING_SNAKE_CASE : str = True break # prevent accidentally stepping through multiple constraints with just one token. return complete, stepped def __lowerCAmelCase ( self , _lowerCamelCase=True ) ->str: SCREAMING_SNAKE_CASE : Dict = ConstraintListState(self.constraints ) # we actually never though self.constraints objects # throughout this process. So it's at initialization state. if stateful: SCREAMING_SNAKE_CASE : str = [ constraint.copy(stateful=_lowerCamelCase ) for constraint in self.complete_constraints ] if self.inprogress_constraint is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.inprogress_constraint.copy(stateful=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = [constraint.copy() for constraint in self.pending_constraints] return new_state
313
1
import math def UpperCAmelCase_( a__ ): """simple docstring""" if not isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : List[str] = F"""Input value of [number={number}] must be an integer""" raise TypeError(a__ ) if number < 1: SCREAMING_SNAKE_CASE : int = F"""Input value of [number={number}] must be > 0""" raise ValueError(a__ ) elif number == 1: return 3 elif number == 2: return 5 else: SCREAMING_SNAKE_CASE : List[str] = int(math.log(number // 3 , 2 ) ) + 2 SCREAMING_SNAKE_CASE : Optional[Any] = [3, 5] SCREAMING_SNAKE_CASE : List[Any] = 2 SCREAMING_SNAKE_CASE : Dict = 3 for block in range(1 , a__ ): for _ in range(a__ ): proth_list.append(2 ** (block + 1) + proth_list[proth_index - 1] ) proth_index += 1 increment *= 2 return proth_list[number - 1] if __name__ == "__main__": import doctest doctest.testmod() for number in range(11): a__ : Optional[int] = 0 try: a__ : Tuple = proth(number) except ValueError: print(F"ValueError: there is no {number}th Proth number") continue print(F"The {number}th Proth number: {value}")
313
import argparse import random import joblib import numpy as np import torch from igf.igf import ( SecondaryLearner, collect_objective_set, compute_perplexity, generate_datasets, load_gpta, recopy_gpta, set_seed, train_secondary_learner, ) from torch.utils.data import DataLoader, RandomSampler from transformers import GPTaLMHeadModel def UpperCAmelCase_( a__=32 , a__=10 , a__=100 , a__=1_026 , a__=True , a__="data/tokenized_stories_train_wikitext103.jbl" , a__="igf_context_pairs.jbl" , ): """simple docstring""" set_seed(3 ) # generate train_data and objective_set SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = generate_datasets( a__ , a__ , number=a__ , min_len=1_026 , trim=a__ ) # keeps model same across runs set_seed(4 ) # model, lm_optimizer, lm_scheduler = recopy_gpt2(model, device, max_steps) # store original model weights # can we train on GPU? SCREAMING_SNAKE_CASE : str = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) # load pretrained model SCREAMING_SNAKE_CASE : Dict = load_gpta('''gpt2''' ).to(a__ ) print('''computing perplexity on objective set''' ) SCREAMING_SNAKE_CASE : int = compute_perplexity(a__ , a__ , a__ ).item() print('''perplexity on objective set:''' , a__ ) # collect igf pairs and save to file demo.jbl collect_objective_set(a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ ) # clean up, delete model and data we don't need anymore del model, train_data, objective_set torch.cuda.empty_cache() def UpperCAmelCase_( a__ , a__=15 , a__=128 , a__=100 , a__="igf_model.pt" , ): """simple docstring""" set_seed(42 ) # Load pre-trained model SCREAMING_SNAKE_CASE : List[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) # Initialize secondary learner to use embedding weights of model SCREAMING_SNAKE_CASE : str = SecondaryLearner(a__ ) # Train secondary learner SCREAMING_SNAKE_CASE : Union[str, Any] = train_secondary_learner( a__ , a__ , max_epochs=a__ , batch_size=a__ , eval_freq=100 , igf_model_path=a__ , ) del model, secondary_learner_train_data torch.cuda.empty_cache() return secondary_learner def UpperCAmelCase_( a__ , a__ , a__ , a__=32 , a__=1_000 , a__=16 , a__=1.0 , a__=recopy_gpta , a__=None , a__=10 , a__="gpt2_finetuned.pt" , ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) SCREAMING_SNAKE_CASE : Optional[int] = RandomSampler(a__ ) SCREAMING_SNAKE_CASE : Dict = DataLoader(a__ , sampler=a__ ) SCREAMING_SNAKE_CASE : Tuple = max_steps // (len(a__ )) + 1 SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = torch.zeros((1, context_len) , dtype=torch.long , device=a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = recopy_model(a__ , a__ , a__ ) model.train() if secondary_learner is not None: secondary_learner.to(a__ ) secondary_learner.eval() SCREAMING_SNAKE_CASE : List[str] = [] SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Tuple = [] # Compute the performance of the transformer model at the beginning SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) for epoch in range(int(a__ ) ): for step, example in enumerate(a__ ): torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Union[str, Any] = random.randint(0 , example.size(2 ) - context_len - 1 ) SCREAMING_SNAKE_CASE : Optional[int] = example[0, 0, start : start + context_len] lm_optimizer.zero_grad() SCREAMING_SNAKE_CASE : Optional[Any] = model(a__ , labels=a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = True if secondary_learner is not None: SCREAMING_SNAKE_CASE : List[str] = secondary_learner.forward( torch.tensor(a__ , dtype=torch.long , device=a__ ).unsqueeze(0 ) )[0].item() observed_qs.append(float(a__ ) ) # Here we implement the simple non-constant threshold for the predicted IG(X) value # We will decay the selectivity of our secondary learner filter from # 1 standard deviation above average to 1 below average after 10 batches. if global_step == 10: SCREAMING_SNAKE_CASE : Dict = -1 if predicted_q < threshold: SCREAMING_SNAKE_CASE : str = False # If we passed the filter, add the context to the batch! if do_backprop: contexts.append(np.array(context.cpu() ) ) SCREAMING_SNAKE_CASE : List[str] = outputs[0] lm_loss.backward() examples += 1 del outputs # Once the batch is filled with enough contexts, backprop on the batch. if examples == batch_size: torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Any = 0 # Do LM backprop torch.nn.utils.clip_grad_norm_(model.parameters() , 3.0 ) lm_optimizer.step() lm_scheduler.step() # Update learning rate schedule global_step += 1 # Compute the performance of the transformer model at this batch if global_step % eval_interval == 0: SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) # Break out of the loop after 60 batches if max_steps > 0 and global_step > 60: break if max_steps > 0 and global_step > 60: break # save finetuned transformer model torch.save(model.state_dict() , a__ ) torch.cuda.empty_cache() # Do some cleaning up so we can reinitialize for the next run of this function del lm_optimizer del lm_scheduler return model def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = argparse.ArgumentParser(description='''Fine-tune a transformer model with IGF on a language modeling task''' ) # Required parameters parser.add_argument( '''--data_dir''' , default=a__ , type=a__ , required=a__ , help='''The input data dir. Should contain data files for WikiText.''' , ) parser.add_argument( '''--model_name_or_path''' , default=a__ , type=a__ , required=a__ , help='''Path to pretrained model or model identifier from huggingface.co/models''' , ) parser.add_argument( '''--data_file''' , type=a__ , default=a__ , help=( '''A jbl file containing tokenized data which can be split as objective dataset, ''' '''train_dataset and test_dataset.''' ) , ) parser.add_argument( '''--igf_data_file''' , type=a__ , default=a__ , help='''A jbl file containing the context and information gain pairs to train secondary learner.''' , ) parser.add_argument( '''--output_dir''' , default=a__ , type=a__ , required=a__ , help='''The output directory where the final fine-tuned model is stored.''' , ) parser.add_argument( '''--tokenizer_name''' , default=a__ , type=a__ , help='''Pretrained tokenizer name or path if not the same as model_name''' , ) parser.add_argument('''--seed''' , type=a__ , default=a__ , help='''A seed for reproducible training.''' ) parser.add_argument( '''--context_len''' , default=32 , type=a__ , help=( '''The maximum total input sequence length after tokenization. Sequences longer ''' '''than this will be truncated, sequences shorter will be padded.''' ) , ) parser.add_argument( '''--size_objective_set''' , default=100 , type=a__ , help='''number of articles that are long enough to be used as our objective set''' , ) parser.add_argument( '''--eval_freq''' , default=100 , type=a__ , help='''secondary model evaluation is triggered at eval_freq''' ) parser.add_argument('''--max_steps''' , default=1_000 , type=a__ , help='''To calculate training epochs''' ) parser.add_argument( '''--secondary_learner_batch_size''' , default=128 , type=a__ , help='''batch size of training data for secondary learner''' , ) parser.add_argument( '''--batch_size''' , default=16 , type=a__ , help='''batch size of training data of language model(gpt2) ''' ) parser.add_argument( '''--eval_interval''' , default=10 , type=a__ , help=( '''decay the selectivity of our secondary learner filter from''' '''1 standard deviation above average to 1 below average after 10 batches''' ) , ) parser.add_argument( '''--number''' , default=100 , type=a__ , help='''The number of examples split to be used as objective_set/test_data''' ) parser.add_argument( '''--min_len''' , default=1_026 , type=a__ , help='''The minimum length of the article to be used as objective set''' ) parser.add_argument( '''--secondary_learner_max_epochs''' , default=15 , type=a__ , help='''number of epochs to train secondary learner''' ) parser.add_argument('''--trim''' , default=a__ , type=a__ , help='''truncate the example if it exceeds context length''' ) parser.add_argument( '''--threshold''' , default=1.0 , type=a__ , help=( '''The threshold value used by secondary learner to filter the train_data and allow only''' ''' informative data as input to the model''' ) , ) parser.add_argument('''--finetuned_model_name''' , default='''gpt2_finetuned.pt''' , type=a__ , help='''finetuned_model_name''' ) parser.add_argument( '''--recopy_model''' , default=a__ , type=a__ , help='''Reset the model to the original pretrained GPT-2 weights after each iteration''' , ) # function calls # Collecting *n* pairs of context and information gain(X, IG(X)) for training the secondary learner generate_n_pairs( context_len=32 , max_steps=10 , size_objective_set=100 , min_len=1_026 , trim=a__ , data_file='''data/tokenized_stories_train_wikitext103.jbl''' , igf_data_file='''igf_context_pairs.jbl''' , ) # Load train data for secondary learner SCREAMING_SNAKE_CASE : List[Any] = joblib.load('''data/IGF_values.jbl''' ) # Train secondary learner SCREAMING_SNAKE_CASE : Tuple = training_secondary_learner( a__ , secondary_learner_max_epochs=15 , secondary_learner_batch_size=128 , eval_freq=100 , igf_model_path='''igf_model.pt''' , ) # load pretrained gpt2 model SCREAMING_SNAKE_CASE : Optional[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) set_seed(42 ) # Generate train and test data to train and evaluate gpt2 model SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = generate_datasets( context_len=32 , file='''data/tokenized_stories_train_wikitext103.jbl''' , number=100 , min_len=1_026 , trim=a__ ) # fine-tuning of the gpt2 model using igf (Information Gain Filtration) finetune( a__ , a__ , a__ , context_len=32 , max_steps=1_000 , batch_size=16 , threshold=1.0 , recopy_model=a__ , secondary_learner=a__ , eval_interval=10 , finetuned_model_name='''gpt2_finetuned.pt''' , ) if __name__ == "__main__": main()
313
1
from collections import OrderedDict from typing import Mapping from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a__ : Optional[Any] = logging.get_logger(__name__) a__ : List[str] = { '''kssteven/ibert-roberta-base''': '''https://huggingface.co/kssteven/ibert-roberta-base/resolve/main/config.json''', '''kssteven/ibert-roberta-large''': '''https://huggingface.co/kssteven/ibert-roberta-large/resolve/main/config.json''', '''kssteven/ibert-roberta-large-mnli''': ( '''https://huggingface.co/kssteven/ibert-roberta-large-mnli/resolve/main/config.json''' ), } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'ibert' def __init__( self , _lowerCamelCase=3_0522 , _lowerCamelCase=768 , _lowerCamelCase=12 , _lowerCamelCase=12 , _lowerCamelCase=3072 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1e-12 , _lowerCamelCase=1 , _lowerCamelCase=0 , _lowerCamelCase=2 , _lowerCamelCase="absolute" , _lowerCamelCase=False , _lowerCamelCase="none" , **_lowerCamelCase , ) ->Any: super().__init__(pad_token_id=_lowerCamelCase , bos_token_id=_lowerCamelCase , eos_token_id=_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = vocab_size SCREAMING_SNAKE_CASE : str = hidden_size SCREAMING_SNAKE_CASE : Tuple = num_hidden_layers SCREAMING_SNAKE_CASE : List[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Dict = hidden_act SCREAMING_SNAKE_CASE : Optional[int] = intermediate_size SCREAMING_SNAKE_CASE : Optional[Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Tuple = type_vocab_size SCREAMING_SNAKE_CASE : Optional[int] = initializer_range SCREAMING_SNAKE_CASE : Union[str, Any] = layer_norm_eps SCREAMING_SNAKE_CASE : str = position_embedding_type SCREAMING_SNAKE_CASE : Optional[int] = quant_mode SCREAMING_SNAKE_CASE : Dict = force_dequant class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: if self.task == "multiple-choice": SCREAMING_SNAKE_CASE : Dict = {0: '''batch''', 1: '''choice''', 2: '''sequence'''} else: SCREAMING_SNAKE_CASE : List[Any] = {0: '''batch''', 1: '''sequence'''} return OrderedDict( [ ('''input_ids''', dynamic_axis), ('''attention_mask''', dynamic_axis), ] )
313
import logging from pathlib import Path import numpy as np import pytorch_lightning as pl import torch from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint from pytorch_lightning.utilities import rank_zero_only from utils_rag import save_json def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = filter(lambda a__ : p.requires_grad , model.parameters() ) SCREAMING_SNAKE_CASE : List[Any] = sum([np.prod(p.size() ) for p in model_parameters] ) return params a__ : Any = logging.getLogger(__name__) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if metric == "rouge2": SCREAMING_SNAKE_CASE : str = '''{val_avg_rouge2:.4f}-{step_count}''' elif metric == "bleu": SCREAMING_SNAKE_CASE : List[Any] = '''{val_avg_bleu:.4f}-{step_count}''' elif metric == "em": SCREAMING_SNAKE_CASE : int = '''{val_avg_em:.4f}-{step_count}''' elif metric == "loss": SCREAMING_SNAKE_CASE : int = '''{val_avg_loss:.4f}-{step_count}''' else: raise NotImplementedError( F"""seq2seq callbacks only support rouge2 and bleu, got {metric}, You can make your own by adding to this""" ''' function.''' ) SCREAMING_SNAKE_CASE : Dict = ModelCheckpoint( dirpath=a__ , filename=a__ , monitor=F"""val_{metric}""" , mode='''max''' , save_top_k=1 , every_n_epochs=1 , ) return checkpoint_callback def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return EarlyStopping( monitor=F"""val_{metric}""" , mode='''min''' if '''loss''' in metric else '''max''' , patience=a__ , verbose=a__ , ) class a_ ( pl.Callback ): """simple docstring""" def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = {F"""lr_group_{i}""": param['''lr'''] for i, param in enumerate(pl_module.trainer.optimizers[0].param_groups )} pl_module.logger.log_metrics(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=True ) ->None: logger.info(F"""***** {type_path} results at step {trainer.global_step:05d} *****""" ) SCREAMING_SNAKE_CASE : Optional[int] = trainer.callback_metrics trainer.logger.log_metrics({k: v for k, v in metrics.items() if k not in ['''log''', '''progress_bar''', '''preds''']} ) # Log results SCREAMING_SNAKE_CASE : List[str] = Path(pl_module.hparams.output_dir ) if type_path == "test": SCREAMING_SNAKE_CASE : Any = od / '''test_results.txt''' SCREAMING_SNAKE_CASE : Optional[int] = od / '''test_generations.txt''' else: # this never gets hit. I prefer not to save intermediate generations, and results are in metrics.json # If people want this it will be easy enough to add back. SCREAMING_SNAKE_CASE : str = od / F"""{type_path}_results/{trainer.global_step:05d}.txt""" SCREAMING_SNAKE_CASE : Tuple = od / F"""{type_path}_generations/{trainer.global_step:05d}.txt""" results_file.parent.mkdir(exist_ok=_lowerCamelCase ) generations_file.parent.mkdir(exist_ok=_lowerCamelCase ) with open(_lowerCamelCase , '''a+''' ) as writer: for key in sorted(_lowerCamelCase ): if key in ["log", "progress_bar", "preds"]: continue SCREAMING_SNAKE_CASE : Tuple = metrics[key] if isinstance(_lowerCamelCase , torch.Tensor ): SCREAMING_SNAKE_CASE : List[Any] = val.item() SCREAMING_SNAKE_CASE : Tuple = F"""{key}: {val:.6f}\n""" writer.write(_lowerCamelCase ) if not save_generations: return if "preds" in metrics: SCREAMING_SNAKE_CASE : Optional[Any] = '''\n'''.join(metrics['''preds'''] ) generations_file.open('''w+''' ).write(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: try: SCREAMING_SNAKE_CASE : Any = pl_module.model.model.num_parameters() except AttributeError: SCREAMING_SNAKE_CASE : Optional[int] = pl_module.model.num_parameters() SCREAMING_SNAKE_CASE : int = count_trainable_parameters(_lowerCamelCase ) # mp stands for million parameters trainer.logger.log_metrics({'''n_params''': npars, '''mp''': npars / 1e6, '''grad_mp''': n_trainable_pars / 1e6} ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: save_json(pl_module.metrics , pl_module.metrics_save_path ) return self._write_logs(_lowerCamelCase , _lowerCamelCase , '''test''' ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Tuple: save_json(pl_module.metrics , pl_module.metrics_save_path ) # Uncommenting this will save val generations # return self._write_logs(trainer, pl_module, "valid")
313
1
a__ : List[Any] = {str(digit): digit**5 for digit in range(10)} def UpperCAmelCase_( a__ ): """simple docstring""" return sum(DIGITS_FIFTH_POWER[digit] for digit in str(a__ ) ) def UpperCAmelCase_( ): """simple docstring""" return sum( number for number in range(1_000 , 1_000_000 ) if number == digits_fifth_powers_sum(a__ ) ) if __name__ == "__main__": print(solution())
313
import argparse import json from typing import List from ltp import LTP from transformers import BertTokenizer def UpperCAmelCase_( a__ ): """simple docstring""" if ( (cp >= 0x4_E00 and cp <= 0x9_FFF) or (cp >= 0x3_400 and cp <= 0x4_DBF) # or (cp >= 0x20_000 and cp <= 0x2A_6DF) # or (cp >= 0x2A_700 and cp <= 0x2B_73F) # or (cp >= 0x2B_740 and cp <= 0x2B_81F) # or (cp >= 0x2B_820 and cp <= 0x2C_EAF) # or (cp >= 0xF_900 and cp <= 0xF_AFF) or (cp >= 0x2F_800 and cp <= 0x2F_A1F) # ): # return True return False def UpperCAmelCase_( a__ ): """simple docstring""" for char in word: SCREAMING_SNAKE_CASE : str = ord(a__ ) if not _is_chinese_char(a__ ): return 0 return 1 def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = set() for token in tokens: SCREAMING_SNAKE_CASE : str = len(a__ ) > 1 and is_chinese(a__ ) if chinese_word: word_set.add(a__ ) SCREAMING_SNAKE_CASE : str = list(a__ ) return word_list def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if not chinese_word_set: return bert_tokens SCREAMING_SNAKE_CASE : List[str] = max([len(a__ ) for w in chinese_word_set] ) SCREAMING_SNAKE_CASE : Tuple = bert_tokens SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = 0, len(a__ ) while start < end: SCREAMING_SNAKE_CASE : Dict = True if is_chinese(bert_word[start] ): SCREAMING_SNAKE_CASE : Optional[int] = min(end - start , a__ ) for i in range(a__ , 1 , -1 ): SCREAMING_SNAKE_CASE : Optional[int] = ''''''.join(bert_word[start : start + i] ) if whole_word in chinese_word_set: for j in range(start + 1 , start + i ): SCREAMING_SNAKE_CASE : Optional[int] = '''##''' + bert_word[j] SCREAMING_SNAKE_CASE : List[str] = start + i SCREAMING_SNAKE_CASE : Optional[Any] = False break if single_word: start += 1 return bert_word def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : Optional[Any] = ltp_tokenizer.seg(lines[i : i + 100] )[0] SCREAMING_SNAKE_CASE : Union[str, Any] = [get_chinese_word(a__ ) for r in res] ltp_res.extend(a__ ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : Any = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : int = bert_tokenizer(lines[i : i + 100] , add_special_tokens=a__ , truncation=a__ , max_length=512 ) bert_res.extend(res['''input_ids'''] ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : int = [] for input_ids, chinese_word in zip(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = [] for id in input_ids: SCREAMING_SNAKE_CASE : List[Any] = bert_tokenizer._convert_id_to_token(a__ ) input_tokens.append(a__ ) SCREAMING_SNAKE_CASE : List[str] = add_sub_symbol(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = [] # We only save pos of chinese subwords start with ##, which mean is part of a whole word. for i, token in enumerate(a__ ): if token[:2] == "##": SCREAMING_SNAKE_CASE : Optional[int] = token[2:] # save chinese tokens' pos if len(a__ ) == 1 and _is_chinese_char(ord(a__ ) ): ref_id.append(a__ ) ref_ids.append(a__ ) assert len(a__ ) == len(a__ ) return ref_ids def UpperCAmelCase_( a__ ): """simple docstring""" with open(args.file_name , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : List[str] = f.readlines() SCREAMING_SNAKE_CASE : Union[str, Any] = [line.strip() for line in data if len(a__ ) > 0 and not line.isspace()] # avoid delimiter like '\u2029' SCREAMING_SNAKE_CASE : List[str] = LTP(args.ltp ) # faster in GPU device SCREAMING_SNAKE_CASE : int = BertTokenizer.from_pretrained(args.bert ) SCREAMING_SNAKE_CASE : int = prepare_ref(a__ , a__ , a__ ) with open(args.save_path , '''w''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : Tuple = [json.dumps(a__ ) + '''\n''' for ref in ref_ids] f.writelines(a__ ) if __name__ == "__main__": a__ : int = argparse.ArgumentParser(description='''prepare_chinese_ref''') parser.add_argument( '''--file_name''', type=str, default='''./resources/chinese-demo.txt''', help='''file need process, same as training data in lm''', ) parser.add_argument( '''--ltp''', type=str, default='''./resources/ltp''', help='''resources for LTP tokenizer, usually a path''' ) parser.add_argument('''--bert''', type=str, default='''./resources/robert''', help='''resources for Bert tokenizer''') parser.add_argument('''--save_path''', type=str, default='''./resources/ref.txt''', help='''path to save res''') a__ : int = parser.parse_args() main(args)
313
1
from unittest.mock import patch import pyspark from datasets.packaged_modules.spark.spark import ( Spark, SparkExamplesIterable, _generate_iterable_examples, ) from ..utils import ( require_dill_gt_0_3_2, require_not_windows, ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = [] for part_id in partition_order: SCREAMING_SNAKE_CASE : Union[str, Any] = df.where(F"""SPARK_PARTITION_ID() = {part_id}""" ).collect() for row_idx, row in enumerate(a__ ): expected_row_ids_and_row_dicts.append((F"""{part_id}_{row_idx}""", row.asDict()) ) return expected_row_ids_and_row_dicts @require_not_windows @require_dill_gt_0_3_2 def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = pyspark.sql.SparkSession.builder.master('''local[*]''' ).appName('''pyspark''' ).getOrCreate() SCREAMING_SNAKE_CASE : Tuple = spark.range(100 ).repartition(1 ) SCREAMING_SNAKE_CASE : Optional[int] = Spark(a__ ) # The id ints will be converted to Pyarrow int64s, so each row will be 8 bytes. Setting a max_shard_size of 16 means # that each partition can hold 2 rows. spark_builder._repartition_df_if_needed(max_shard_size=16 ) # Given that the dataframe has 100 rows and each partition has 2 rows, we expect 50 partitions. assert spark_builder.df.rdd.getNumPartitions() == 50 @require_not_windows @require_dill_gt_0_3_2 def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = pyspark.sql.SparkSession.builder.master('''local[*]''' ).appName('''pyspark''' ).getOrCreate() SCREAMING_SNAKE_CASE : List[Any] = spark.range(10 ).repartition(2 ) SCREAMING_SNAKE_CASE : Dict = [1, 0] SCREAMING_SNAKE_CASE : Tuple = _generate_iterable_examples(a__ , a__ ) # Reverse the partitions. SCREAMING_SNAKE_CASE : str = _get_expected_row_ids_and_row_dicts_for_partition_order(a__ , a__ ) for i, (row_id, row_dict) in enumerate(generate_fn() ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = expected_row_ids_and_row_dicts[i] assert row_id == expected_row_id assert row_dict == expected_row_dict @require_not_windows @require_dill_gt_0_3_2 def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = pyspark.sql.SparkSession.builder.master('''local[*]''' ).appName('''pyspark''' ).getOrCreate() SCREAMING_SNAKE_CASE : List[str] = spark.range(10 ).repartition(1 ) SCREAMING_SNAKE_CASE : int = SparkExamplesIterable(a__ ) assert it.n_shards == 1 for i, (row_id, row_dict) in enumerate(a__ ): assert row_id == F"""0_{i}""" assert row_dict == {"id": i} @require_not_windows @require_dill_gt_0_3_2 def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : int = pyspark.sql.SparkSession.builder.master('''local[*]''' ).appName('''pyspark''' ).getOrCreate() SCREAMING_SNAKE_CASE : List[Any] = spark.range(30 ).repartition(3 ) # Mock the generator so that shuffle reverses the partition indices. with patch('''numpy.random.Generator''' ) as generator_mock: SCREAMING_SNAKE_CASE : List[str] = lambda a__ : x.reverse() SCREAMING_SNAKE_CASE : Union[str, Any] = _get_expected_row_ids_and_row_dicts_for_partition_order(a__ , [2, 1, 0] ) SCREAMING_SNAKE_CASE : Tuple = SparkExamplesIterable(a__ ).shuffle_data_sources(a__ ) assert shuffled_it.n_shards == 3 for i, (row_id, row_dict) in enumerate(a__ ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = expected_row_ids_and_row_dicts[i] assert row_id == expected_row_id assert row_dict == expected_row_dict @require_not_windows @require_dill_gt_0_3_2 def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = pyspark.sql.SparkSession.builder.master('''local[*]''' ).appName('''pyspark''' ).getOrCreate() SCREAMING_SNAKE_CASE : Tuple = spark.range(20 ).repartition(4 ) # Partitions 0 and 2 SCREAMING_SNAKE_CASE : Union[str, Any] = SparkExamplesIterable(a__ ).shard_data_sources(worker_id=0 , num_workers=2 ) assert shard_it_a.n_shards == 2 SCREAMING_SNAKE_CASE : Optional[Any] = _get_expected_row_ids_and_row_dicts_for_partition_order(a__ , [0, 2] ) for i, (row_id, row_dict) in enumerate(a__ ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = expected_row_ids_and_row_dicts_a[i] assert row_id == expected_row_id assert row_dict == expected_row_dict # Partitions 1 and 3 SCREAMING_SNAKE_CASE : List[Any] = SparkExamplesIterable(a__ ).shard_data_sources(worker_id=1 , num_workers=2 ) assert shard_it_a.n_shards == 2 SCREAMING_SNAKE_CASE : Optional[int] = _get_expected_row_ids_and_row_dicts_for_partition_order(a__ , [1, 3] ) for i, (row_id, row_dict) in enumerate(a__ ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = expected_row_ids_and_row_dicts_a[i] assert row_id == expected_row_id assert row_dict == expected_row_dict @require_not_windows @require_dill_gt_0_3_2 def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = pyspark.sql.SparkSession.builder.master('''local[*]''' ).appName('''pyspark''' ).getOrCreate() SCREAMING_SNAKE_CASE : Any = spark.range(100 ).repartition(1 ) SCREAMING_SNAKE_CASE : int = Spark(a__ ) # Choose a small max_shard_size for maximum partitioning. spark_builder._repartition_df_if_needed(max_shard_size=1 ) # The new number of partitions should not be greater than the number of rows. assert spark_builder.df.rdd.getNumPartitions() == 100
313
import datetime import platform import subprocess from typing import Optional, Tuple, Union import numpy as np def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Tuple = '''1''' SCREAMING_SNAKE_CASE : Union[str, Any] = '''f32le''' SCREAMING_SNAKE_CASE : List[Any] = [ '''ffmpeg''', '''-i''', '''pipe:0''', '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] try: with subprocess.Popen(a__ , stdin=subprocess.PIPE , stdout=subprocess.PIPE ) as ffmpeg_process: SCREAMING_SNAKE_CASE : Tuple = ffmpeg_process.communicate(a__ ) except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to load audio files from filename''' ) from error SCREAMING_SNAKE_CASE : Optional[Any] = output_stream[0] SCREAMING_SNAKE_CASE : Any = np.frombuffer(a__ , np.floataa ) if audio.shape[0] == 0: raise ValueError('''Malformed soundfile''' ) return audio def UpperCAmelCase_( a__ , a__ , a__ = "f32le" , ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Dict = '''1''' if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Dict = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = platform.system() if system == "Linux": SCREAMING_SNAKE_CASE : Dict = '''alsa''' SCREAMING_SNAKE_CASE : Any = '''default''' elif system == "Darwin": SCREAMING_SNAKE_CASE : Union[str, Any] = '''avfoundation''' SCREAMING_SNAKE_CASE : Optional[int] = ''':0''' elif system == "Windows": SCREAMING_SNAKE_CASE : int = '''dshow''' SCREAMING_SNAKE_CASE : Any = '''default''' SCREAMING_SNAKE_CASE : Union[str, Any] = [ '''ffmpeg''', '''-f''', format_, '''-i''', input_, '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-fflags''', '''nobuffer''', '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] SCREAMING_SNAKE_CASE : List[str] = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample SCREAMING_SNAKE_CASE : List[Any] = _ffmpeg_stream(a__ , a__ ) for item in iterator: yield item def UpperCAmelCase_( a__ , a__ , a__ = None , a__ = None , a__ = "f32le" , ): """simple docstring""" if stream_chunk_s is not None: SCREAMING_SNAKE_CASE : Tuple = stream_chunk_s else: SCREAMING_SNAKE_CASE : List[str] = chunk_length_s SCREAMING_SNAKE_CASE : Union[str, Any] = ffmpeg_microphone(a__ , a__ , format_for_conversion=a__ ) if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : Optional[int] = np.intaa SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Any = np.floataa SCREAMING_SNAKE_CASE : Union[str, Any] = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) if stride_length_s is None: SCREAMING_SNAKE_CASE : Optional[Any] = chunk_length_s / 6 SCREAMING_SNAKE_CASE : Dict = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample if isinstance(a__ , (int, float) ): SCREAMING_SNAKE_CASE : List[Any] = [stride_length_s, stride_length_s] SCREAMING_SNAKE_CASE : Any = int(round(sampling_rate * stride_length_s[0] ) ) * size_of_sample SCREAMING_SNAKE_CASE : int = int(round(sampling_rate * stride_length_s[1] ) ) * size_of_sample SCREAMING_SNAKE_CASE : Union[str, Any] = datetime.datetime.now() SCREAMING_SNAKE_CASE : Dict = datetime.timedelta(seconds=a__ ) for item in chunk_bytes_iter(a__ , a__ , stride=(stride_left, stride_right) , stream=a__ ): # Put everything back in numpy scale SCREAMING_SNAKE_CASE : Dict = np.frombuffer(item['''raw'''] , dtype=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = ( item['''stride'''][0] // size_of_sample, item['''stride'''][1] // size_of_sample, ) SCREAMING_SNAKE_CASE : Any = sampling_rate audio_time += delta if datetime.datetime.now() > audio_time + 10 * delta: # We're late !! SKIP continue yield item def UpperCAmelCase_( a__ , a__ , a__ , a__ = False ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = b'''''' SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = stride if stride_left + stride_right >= chunk_len: raise ValueError( F"""Stride needs to be strictly smaller than chunk_len: ({stride_left}, {stride_right}) vs {chunk_len}""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = 0 for raw in iterator: acc += raw if stream and len(a__ ) < chunk_len: SCREAMING_SNAKE_CASE : List[str] = (_stride_left, 0) yield {"raw": acc[:chunk_len], "stride": stride, "partial": True} else: while len(a__ ) >= chunk_len: # We are flushing the accumulator SCREAMING_SNAKE_CASE : str = (_stride_left, stride_right) SCREAMING_SNAKE_CASE : List[str] = {'''raw''': acc[:chunk_len], '''stride''': stride} if stream: SCREAMING_SNAKE_CASE : List[str] = False yield item SCREAMING_SNAKE_CASE : Dict = stride_left SCREAMING_SNAKE_CASE : int = acc[chunk_len - stride_left - stride_right :] # Last chunk if len(a__ ) > stride_left: SCREAMING_SNAKE_CASE : Optional[Any] = {'''raw''': acc, '''stride''': (_stride_left, 0)} if stream: SCREAMING_SNAKE_CASE : Union[str, Any] = False yield item def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : str = 2**24 # 16Mo try: with subprocess.Popen(a__ , stdout=subprocess.PIPE , bufsize=a__ ) as ffmpeg_process: while True: SCREAMING_SNAKE_CASE : str = ffmpeg_process.stdout.read(a__ ) if raw == b"": break yield raw except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to stream audio files from filename''' ) from error
313
1
import argparse import os from pathlib import Path from typing import Dict import tensorflow as tf import torch from tqdm import tqdm from transformers import PegasusConfig, PegasusForConditionalGeneration, PegasusTokenizer from transformers.models.pegasus.configuration_pegasus import DEFAULTS, task_specific_params a__ : Optional[Any] = [ # replace left string with right string to get the relevant state_dict key (identical state dict to bart) ['''memory_attention''', '''encoder_attn'''], ['''attention''', '''attn'''], ['''/''', '''.'''], ['''.LayerNorm.gamma''', '''_layer_norm.weight'''], ['''.LayerNorm.beta''', '''_layer_norm.bias'''], ['''r.layer_''', '''r.layers.'''], ['''output_proj''', '''out_proj'''], ['''ffn.dense_1.''', '''fc2.'''], ['''ffn.dense.''', '''fc1.'''], ['''ffn_layer_norm''', '''final_layer_norm'''], ['''kernel''', '''weight'''], ['''encoder_layer_norm.''', '''encoder.layer_norm.'''], ['''decoder_layer_norm.''', '''decoder.layer_norm.'''], ['''embeddings.weights''', '''shared.weight'''], ] def UpperCAmelCase_( a__ ): """simple docstring""" for pegasus_name, hf_name in PATTERNS: SCREAMING_SNAKE_CASE : Union[str, Any] = k.replace(a__ , a__ ) return k def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = DEFAULTS.copy() cfg_kwargs.update(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = PegasusConfig(**a__ ) SCREAMING_SNAKE_CASE : Optional[int] = PegasusForConditionalGeneration(a__ ) SCREAMING_SNAKE_CASE : Dict = torch_model.model.state_dict() SCREAMING_SNAKE_CASE : List[str] = {} for k, v in tf_weights.items(): SCREAMING_SNAKE_CASE : int = rename_state_dict_key(a__ ) if new_k not in sd: raise ValueError(F"""could not find new key {new_k} in state dict. (converted from {k})""" ) if "dense" in k or "proj" in new_k: SCREAMING_SNAKE_CASE : Dict = v.T SCREAMING_SNAKE_CASE : Tuple = torch.tensor(a__ , dtype=sd[new_k].dtype ) assert v.shape == sd[new_k].shape, F"""{new_k}, {k}, {v.shape}, {sd[new_k].shape}""" # make sure embedding.padding_idx is respected SCREAMING_SNAKE_CASE : Tuple = torch.zeros_like(mapping['''shared.weight'''][cfg.pad_token_id + 1] ) SCREAMING_SNAKE_CASE : int = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Union[str, Any] = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Optional[Any] = {k: torch.zeros_like(a__ ) for k, v in sd.items() if k.endswith('''bias''' ) and k not in mapping} mapping.update(**a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = torch_model.model.load_state_dict(a__ , strict=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = [ k for k in missing if k not in ['''encoder.embed_positions.weight''', '''decoder.embed_positions.weight'''] ] assert unexpected_missing == [], F"""no matches found for the following torch keys {unexpected_missing}""" assert extra == [], F"""no matches found for the following tf keys {extra}""" return torch_model def UpperCAmelCase_( a__="./ckpt/aeslc/model.ckpt-32000" ): """simple docstring""" SCREAMING_SNAKE_CASE : str = tf.train.list_variables(a__ ) SCREAMING_SNAKE_CASE : str = {} SCREAMING_SNAKE_CASE : List[Any] = ['''Adafactor''', '''global_step'''] for name, shape in tqdm(a__ , desc='''converting tf checkpoint to dict''' ): SCREAMING_SNAKE_CASE : Union[str, Any] = any(pat in name for pat in ignore_name ) if skip_key: continue SCREAMING_SNAKE_CASE : Dict = tf.train.load_variable(a__ , a__ ) SCREAMING_SNAKE_CASE : Any = array return tf_weights def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = Path(a__ ).parent.name SCREAMING_SNAKE_CASE : Union[str, Any] = task_specific_params[F"""summarization_{dataset}"""]['''max_position_embeddings'''] SCREAMING_SNAKE_CASE : Dict = PegasusTokenizer.from_pretrained('''sshleifer/pegasus''' , model_max_length=a__ ) assert tok.model_max_length == desired_max_model_length tok.save_pretrained(a__ ) # convert model SCREAMING_SNAKE_CASE : Any = get_tf_weights_as_numpy(a__ ) SCREAMING_SNAKE_CASE : List[str] = task_specific_params[F"""summarization_{dataset}"""] if dataset == "large": SCREAMING_SNAKE_CASE : int = task_specific_params SCREAMING_SNAKE_CASE : List[str] = convert_pegasus(a__ , a__ ) torch_model.save_pretrained(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch_model.state_dict() sd.pop('''model.decoder.embed_positions.weight''' ) sd.pop('''model.encoder.embed_positions.weight''' ) torch.save(a__ , Path(a__ ) / '''pytorch_model.bin''' ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument('''tf_ckpt_path''', type=str, help='''passed to tf.train.list_variables''') parser.add_argument('''save_dir''', default=None, type=str, help='''Path to the output PyTorch model.''') a__ : List[str] = parser.parse_args() if args.save_dir is None: a__ : Any = Path(args.tf_ckpt_path).parent.name a__ : int = os.path.join('''pegasus''', dataset) convert_pegasus_ckpt_to_pytorch(args.tf_ckpt_path, args.save_dir)
313
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a__ : Tuple = {'''configuration_wavlm''': ['''WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''WavLMConfig''']} try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Dict = [ '''WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST''', '''WavLMForAudioFrameClassification''', '''WavLMForCTC''', '''WavLMForSequenceClassification''', '''WavLMForXVector''', '''WavLMModel''', '''WavLMPreTrainedModel''', ] if TYPE_CHECKING: from .configuration_wavlm import WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP, WavLMConfig try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_wavlm import ( WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST, WavLMForAudioFrameClassification, WavLMForCTC, WavLMForSequenceClassification, WavLMForXVector, WavLMModel, WavLMPreTrainedModel, ) else: import sys a__ : Dict = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
1
def UpperCAmelCase_( a__ = 50 ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = [1] * (length + 1) for row_length in range(3 , length + 1 ): for block_length in range(3 , row_length + 1 ): for block_start in range(row_length - block_length ): ways_number[row_length] += ways_number[ row_length - block_start - block_length - 1 ] ways_number[row_length] += 1 return ways_number[length] if __name__ == "__main__": print(F"{solution() = }")
313
import copy from ...configuration_utils import PretrainedConfig from ...utils import logging from ..auto import CONFIG_MAPPING a__ : int = logging.get_logger(__name__) a__ : Optional[Any] = { '''SenseTime/deformable-detr''': '''https://huggingface.co/sensetime/deformable-detr/resolve/main/config.json''', # See all Deformable DETR models at https://huggingface.co/models?filter=deformable-detr } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'deformable_detr' __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'd_model', 'num_attention_heads': 'encoder_attention_heads', } def __init__( self , _lowerCamelCase=True , _lowerCamelCase=None , _lowerCamelCase=3 , _lowerCamelCase=300 , _lowerCamelCase=1024 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=0.0 , _lowerCamelCase=True , _lowerCamelCase="relu" , _lowerCamelCase=256 , _lowerCamelCase=0.1 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1.0 , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase="sine" , _lowerCamelCase="resnet50" , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=False , _lowerCamelCase=300 , _lowerCamelCase=False , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=1 , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=0.1 , _lowerCamelCase=0.2_5 , _lowerCamelCase=False , **_lowerCamelCase , ) ->Optional[Any]: if backbone_config is not None and use_timm_backbone: raise ValueError('''You can\'t specify both `backbone_config` and `use_timm_backbone`.''' ) if not use_timm_backbone: if backbone_config is None: logger.info('''`backbone_config` is `None`. Initializing the config with the default `ResNet` backbone.''' ) SCREAMING_SNAKE_CASE : Dict = CONFIG_MAPPING['''resnet'''](out_features=['''stage4'''] ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : List[Any] = backbone_config.get('''model_type''' ) SCREAMING_SNAKE_CASE : Optional[Any] = CONFIG_MAPPING[backbone_model_type] SCREAMING_SNAKE_CASE : int = config_class.from_dict(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = use_timm_backbone SCREAMING_SNAKE_CASE : Optional[int] = backbone_config SCREAMING_SNAKE_CASE : Union[str, Any] = num_channels SCREAMING_SNAKE_CASE : Optional[Any] = num_queries SCREAMING_SNAKE_CASE : Optional[int] = max_position_embeddings SCREAMING_SNAKE_CASE : Optional[int] = d_model SCREAMING_SNAKE_CASE : str = encoder_ffn_dim SCREAMING_SNAKE_CASE : str = encoder_layers SCREAMING_SNAKE_CASE : str = encoder_attention_heads SCREAMING_SNAKE_CASE : Optional[int] = decoder_ffn_dim SCREAMING_SNAKE_CASE : int = decoder_layers SCREAMING_SNAKE_CASE : Union[str, Any] = decoder_attention_heads SCREAMING_SNAKE_CASE : List[str] = dropout SCREAMING_SNAKE_CASE : Optional[int] = attention_dropout SCREAMING_SNAKE_CASE : str = activation_dropout SCREAMING_SNAKE_CASE : Optional[int] = activation_function SCREAMING_SNAKE_CASE : Optional[int] = init_std SCREAMING_SNAKE_CASE : List[str] = init_xavier_std SCREAMING_SNAKE_CASE : Optional[Any] = encoder_layerdrop SCREAMING_SNAKE_CASE : Union[str, Any] = auxiliary_loss SCREAMING_SNAKE_CASE : List[Any] = position_embedding_type SCREAMING_SNAKE_CASE : str = backbone SCREAMING_SNAKE_CASE : Dict = use_pretrained_backbone SCREAMING_SNAKE_CASE : Dict = dilation # deformable attributes SCREAMING_SNAKE_CASE : str = num_feature_levels SCREAMING_SNAKE_CASE : Optional[Any] = encoder_n_points SCREAMING_SNAKE_CASE : Any = decoder_n_points SCREAMING_SNAKE_CASE : str = two_stage SCREAMING_SNAKE_CASE : List[str] = two_stage_num_proposals SCREAMING_SNAKE_CASE : Dict = with_box_refine if two_stage is True and with_box_refine is False: raise ValueError('''If two_stage is True, with_box_refine must be True.''' ) # Hungarian matcher SCREAMING_SNAKE_CASE : int = class_cost SCREAMING_SNAKE_CASE : Union[str, Any] = bbox_cost SCREAMING_SNAKE_CASE : Optional[int] = giou_cost # Loss coefficients SCREAMING_SNAKE_CASE : Dict = mask_loss_coefficient SCREAMING_SNAKE_CASE : Union[str, Any] = dice_loss_coefficient SCREAMING_SNAKE_CASE : str = bbox_loss_coefficient SCREAMING_SNAKE_CASE : Tuple = giou_loss_coefficient SCREAMING_SNAKE_CASE : Optional[int] = eos_coefficient SCREAMING_SNAKE_CASE : Tuple = focal_alpha SCREAMING_SNAKE_CASE : Optional[int] = disable_custom_kernels super().__init__(is_encoder_decoder=_lowerCamelCase , **_lowerCamelCase ) @property def __lowerCAmelCase ( self ) ->int: return self.encoder_attention_heads @property def __lowerCAmelCase ( self ) ->int: return self.d_model def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : str = copy.deepcopy(self.__dict__ ) if self.backbone_config is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.backbone_config.to_dict() SCREAMING_SNAKE_CASE : Any = self.__class__.model_type return output
313
1
import unittest import torch from diffusers import VQModel from diffusers.utils import floats_tensor, torch_device from diffusers.utils.testing_utils import enable_full_determinism from .test_modeling_common import ModelTesterMixin, UNetTesterMixin enable_full_determinism() class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Any = VQModel __SCREAMING_SNAKE_CASE : Any = 'sample' @property def __lowerCAmelCase ( self , _lowerCamelCase=(32, 32) ) ->Optional[int]: SCREAMING_SNAKE_CASE : List[Any] = 4 SCREAMING_SNAKE_CASE : Optional[int] = 3 SCREAMING_SNAKE_CASE : str = floats_tensor((batch_size, num_channels) + sizes ).to(_lowerCamelCase ) return {"sample": image} @property def __lowerCAmelCase ( self ) ->List[str]: return (3, 32, 32) @property def __lowerCAmelCase ( self ) ->Optional[int]: return (3, 32, 32) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = { '''block_out_channels''': [32, 64], '''in_channels''': 3, '''out_channels''': 3, '''down_block_types''': ['''DownEncoderBlock2D''', '''DownEncoderBlock2D'''], '''up_block_types''': ['''UpDecoderBlock2D''', '''UpDecoderBlock2D'''], '''latent_channels''': 3, } SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_input return init_dict, inputs_dict def __lowerCAmelCase ( self ) ->Union[str, Any]: pass def __lowerCAmelCase ( self ) ->Union[str, Any]: pass def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = VQModel.from_pretrained('''fusing/vqgan-dummy''' , output_loading_info=_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) self.assertEqual(len(loading_info['''missing_keys'''] ) , 0 ) model.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = model(**self.dummy_input ) assert image is not None, "Make sure output is not None" def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Tuple = VQModel.from_pretrained('''fusing/vqgan-dummy''' ) model.to(_lowerCamelCase ).eval() torch.manual_seed(0 ) if torch.cuda.is_available(): torch.cuda.manual_seed_all(0 ) SCREAMING_SNAKE_CASE : Tuple = torch.randn(1 , model.config.in_channels , model.config.sample_size , model.config.sample_size ) SCREAMING_SNAKE_CASE : Optional[Any] = image.to(_lowerCamelCase ) with torch.no_grad(): SCREAMING_SNAKE_CASE : int = model(_lowerCamelCase ).sample SCREAMING_SNAKE_CASE : Any = output[0, -1, -3:, -3:].flatten().cpu() # fmt: off SCREAMING_SNAKE_CASE : Any = torch.tensor([-0.0_1_5_3, -0.4_0_4_4, -0.1_8_8_0, -0.5_1_6_1, -0.2_4_1_8, -0.4_0_7_2, -0.1_6_1_2, -0.0_6_3_3, -0.0_1_4_3] ) # fmt: on self.assertTrue(torch.allclose(_lowerCamelCase , _lowerCamelCase , atol=1e-3 ) )
313
import torch from diffusers import EulerDiscreteScheduler from diffusers.utils import torch_device from .test_schedulers import SchedulerCommonTest class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = (EulerDiscreteScheduler,) __SCREAMING_SNAKE_CASE : Optional[int] = 10 def __lowerCAmelCase ( self , **_lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = { '''num_train_timesteps''': 1100, '''beta_start''': 0.0_0_0_1, '''beta_end''': 0.0_2, '''beta_schedule''': '''linear''', } config.update(**_lowerCamelCase ) return config def __lowerCAmelCase ( self ) ->Tuple: for timesteps in [10, 50, 100, 1000]: self.check_over_configs(num_train_timesteps=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: for beta_start, beta_end in zip([0.0_0_0_0_1, 0.0_0_0_1, 0.0_0_1] , [0.0_0_0_2, 0.0_0_2, 0.0_2] ): self.check_over_configs(beta_start=_lowerCamelCase , beta_end=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: for schedule in ["linear", "scaled_linear"]: self.check_over_configs(beta_schedule=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: for prediction_type in ["epsilon", "v_prediction"]: self.check_over_configs(prediction_type=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : str = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : int = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Any = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = self.dummy_model() SCREAMING_SNAKE_CASE : int = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : Any = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[Any] = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_scheduler_config(prediction_type='''v_prediction''' ) SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = self.dummy_model() SCREAMING_SNAKE_CASE : List[str] = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : List[str] = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : str = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = output.prev_sample SCREAMING_SNAKE_CASE : str = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 0.0_0_0_2 ) < 1e-2 assert abs(result_mean.item() - 2.2676e-06 ) < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Dict = self.get_scheduler_config() SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[Any] = self.dummy_model() SCREAMING_SNAKE_CASE : str = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : Optional[Any] = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : Dict = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Dict = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Optional[Any] = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase , use_karras_sigmas=_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_model() SCREAMING_SNAKE_CASE : Dict = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : int = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : List[Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = output.prev_sample SCREAMING_SNAKE_CASE : Optional[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_2_4.5_2_2_9_9_4_9_9_5_1_1_7_1_9 ) < 1e-2 assert abs(result_mean.item() - 0.1_6_2_1_3_9_3_2_6_3_3_3_9_9_9_6_3 ) < 1e-3
313
1
from sympy import diff, lambdify, symbols from sympy.functions import * # noqa: F403 def UpperCAmelCase_( a__ , a__ , a__ = "x" , a__ = 10**-10 , a__ = 1 , ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = symbols(a__ ) SCREAMING_SNAKE_CASE : int = lambdify(a__ , a__ ) SCREAMING_SNAKE_CASE : str = lambdify(a__ , diff(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Dict = starting_point while True: if diff_function(a__ ) != 0: SCREAMING_SNAKE_CASE : List[str] = prev_guess - multiplicity * func(a__ ) / diff_function( a__ ) else: raise ZeroDivisionError('''Could not find root''' ) from None # Precision is checked by comparing the difference of consecutive guesses if abs(next_guess - prev_guess ) < precision: return next_guess SCREAMING_SNAKE_CASE : List[str] = next_guess # Let's Execute if __name__ == "__main__": # Find root of trigonometric function # Find value of pi print(F"The root of sin(x) = 0 is {newton_raphson('sin(x)', 2)}") # Find root of polynomial # Find fourth Root of 5 print(F"The root of x**4 - 5 = 0 is {newton_raphson('x**4 -5', 0.4 +5j)}") # Find value of e print( '''The root of log(y) - 1 = 0 is ''', F"{newton_raphson('log(y) - 1', 2, variable='y')}", ) # Exponential Roots print( '''The root of exp(x) - 1 = 0 is''', F"{newton_raphson('exp(x) - 1', 10, precision=0.0_05)}", ) # Find root of cos(x) print(F"The root of cos(x) = 0 is {newton_raphson('cos(x)', 0)}")
313
import json from typing import Dict, List, Optional, Tuple, Union from tokenizers import pre_tokenizers, processors from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput from ...tokenization_utils_fast import PreTrainedTokenizerFast from ...utils import PaddingStrategy, logging from .tokenization_led import LEDTokenizer a__ : Dict = logging.get_logger(__name__) a__ : Dict = {'''vocab_file''': '''vocab.json''', '''merges_file''': '''merges.txt''', '''tokenizer_file''': '''tokenizer.json'''} a__ : str = { '''vocab_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json''', }, '''merges_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt''', }, '''tokenizer_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json''', }, } a__ : Optional[int] = { '''allenai/led-base-16384''': 16_384, } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = VOCAB_FILES_NAMES __SCREAMING_SNAKE_CASE : Tuple = PRETRAINED_VOCAB_FILES_MAP __SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __SCREAMING_SNAKE_CASE : Union[str, Any] = LEDTokenizer __SCREAMING_SNAKE_CASE : Optional[int] = ['input_ids', 'attention_mask'] def __init__( self , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase="replace" , _lowerCamelCase="<s>" , _lowerCamelCase="</s>" , _lowerCamelCase="</s>" , _lowerCamelCase="<s>" , _lowerCamelCase="<unk>" , _lowerCamelCase="<pad>" , _lowerCamelCase="<mask>" , _lowerCamelCase=False , _lowerCamelCase=True , **_lowerCamelCase , ) ->Union[str, Any]: super().__init__( _lowerCamelCase , _lowerCamelCase , tokenizer_file=_lowerCamelCase , errors=_lowerCamelCase , bos_token=_lowerCamelCase , eos_token=_lowerCamelCase , sep_token=_lowerCamelCase , cls_token=_lowerCamelCase , unk_token=_lowerCamelCase , pad_token=_lowerCamelCase , mask_token=_lowerCamelCase , add_prefix_space=_lowerCamelCase , trim_offsets=_lowerCamelCase , **_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Dict = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() ) if pre_tok_state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : str = getattr(_lowerCamelCase , pre_tok_state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : Optional[int] = add_prefix_space SCREAMING_SNAKE_CASE : str = pre_tok_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = add_prefix_space # the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__` SCREAMING_SNAKE_CASE : List[Any] = '''post_processor''' SCREAMING_SNAKE_CASE : int = getattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) if tokenizer_component_instance: SCREAMING_SNAKE_CASE : Any = json.loads(tokenizer_component_instance.__getstate__() ) # The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class` if "sep" in state: SCREAMING_SNAKE_CASE : Optional[int] = tuple(state['''sep'''] ) if "cls" in state: SCREAMING_SNAKE_CASE : Optional[Any] = tuple(state['''cls'''] ) SCREAMING_SNAKE_CASE : Any = False if state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : Union[str, Any] = add_prefix_space SCREAMING_SNAKE_CASE : Union[str, Any] = True if state.get('''trim_offsets''' , _lowerCamelCase ) != trim_offsets: SCREAMING_SNAKE_CASE : List[Any] = trim_offsets SCREAMING_SNAKE_CASE : Union[str, Any] = True if changes_to_apply: SCREAMING_SNAKE_CASE : List[str] = getattr(_lowerCamelCase , state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : List[Any] = component_class(**_lowerCamelCase ) setattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) @property # Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED def __lowerCAmelCase ( self ) ->str: if self._mask_token is None: if self.verbose: logger.error('''Using mask_token, but it is not set yet.''' ) return None return str(self._mask_token ) @mask_token.setter def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : str = AddedToken(_lowerCamelCase , lstrip=_lowerCamelCase , rstrip=_lowerCamelCase ) if isinstance(_lowerCamelCase , _lowerCamelCase ) else value SCREAMING_SNAKE_CASE : List[Any] = value def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : Tuple = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._batch_encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : List[Any] = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->Tuple[str]: SCREAMING_SNAKE_CASE : Any = self._tokenizer.model.save(_lowerCamelCase , name=_lowerCamelCase ) return tuple(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=None ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id] if token_ids_a is None: return output return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id] def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->List[int]: SCREAMING_SNAKE_CASE : Any = [self.sep_token_id] SCREAMING_SNAKE_CASE : List[str] = [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 __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None , _lowerCamelCase = PaddingStrategy.DO_NOT_PAD , _lowerCamelCase = None , _lowerCamelCase = None , ) ->dict: SCREAMING_SNAKE_CASE : Tuple = super()._pad( encoded_inputs=_lowerCamelCase , max_length=_lowerCamelCase , padding_strategy=_lowerCamelCase , pad_to_multiple_of=_lowerCamelCase , return_attention_mask=_lowerCamelCase , ) # Load from model defaults if return_attention_mask is None: SCREAMING_SNAKE_CASE : Optional[Any] = '''attention_mask''' in self.model_input_names if return_attention_mask and "global_attention_mask" in encoded_inputs: SCREAMING_SNAKE_CASE : int = encoded_inputs[self.model_input_names[0]] # `global_attention_mask` need to have the same length as other (sequential) inputs. SCREAMING_SNAKE_CASE : Tuple = len(encoded_inputs['''global_attention_mask'''] ) != len(_lowerCamelCase ) if needs_to_be_padded: SCREAMING_SNAKE_CASE : int = len(_lowerCamelCase ) - len(encoded_inputs['''global_attention_mask'''] ) if self.padding_side == "right": # Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend` SCREAMING_SNAKE_CASE : str = ( encoded_inputs['''global_attention_mask'''] + [-1] * difference ) elif self.padding_side == "left": SCREAMING_SNAKE_CASE : Optional[Any] = [-1] * difference + encoded_inputs[ '''global_attention_mask''' ] else: raise ValueError('''Invalid padding strategy:''' + str(self.padding_side ) ) return encoded_inputs
313
1
import argparse import os import torch from transformers import ( XLNetConfig, XLNetForQuestionAnswering, XLNetForSequenceClassification, XLNetLMHeadModel, load_tf_weights_in_xlnet, ) from transformers.utils import CONFIG_NAME, WEIGHTS_NAME, logging a__ : Dict = { '''cola''': 2, '''mnli''': 3, '''mrpc''': 2, '''sst-2''': 2, '''sts-b''': 1, '''qqp''': 2, '''qnli''': 2, '''rte''': 2, '''wnli''': 2, } logging.set_verbosity_info() def UpperCAmelCase_( a__ , a__ , a__ , a__=None ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = XLNetConfig.from_json_file(a__ ) SCREAMING_SNAKE_CASE : List[str] = finetuning_task.lower() if finetuning_task is not None else '''''' if finetuning_task in GLUE_TASKS_NUM_LABELS: print(F"""Building PyTorch XLNetForSequenceClassification model from configuration: {config}""" ) SCREAMING_SNAKE_CASE : List[Any] = finetuning_task SCREAMING_SNAKE_CASE : Any = GLUE_TASKS_NUM_LABELS[finetuning_task] SCREAMING_SNAKE_CASE : Tuple = XLNetForSequenceClassification(a__ ) elif "squad" in finetuning_task: SCREAMING_SNAKE_CASE : Optional[Any] = finetuning_task SCREAMING_SNAKE_CASE : Tuple = XLNetForQuestionAnswering(a__ ) else: SCREAMING_SNAKE_CASE : int = XLNetLMHeadModel(a__ ) # Load weights from tf checkpoint load_tf_weights_in_xlnet(a__ , a__ , a__ ) # Save pytorch-model SCREAMING_SNAKE_CASE : Optional[Any] = os.path.join(a__ , a__ ) SCREAMING_SNAKE_CASE : str = os.path.join(a__ , a__ ) print(F"""Save PyTorch model to {os.path.abspath(a__ )}""" ) torch.save(model.state_dict() , a__ ) print(F"""Save configuration file to {os.path.abspath(a__ )}""" ) with open(a__ , '''w''' , encoding='''utf-8''' ) as f: f.write(config.to_json_string() ) if __name__ == "__main__": a__ : List[Any] = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--tf_checkpoint_path''', default=None, type=str, required=True, help='''Path to the TensorFlow checkpoint path.''' ) parser.add_argument( '''--xlnet_config_file''', default=None, type=str, required=True, help=( '''The config json file corresponding to the pre-trained XLNet model. \n''' '''This specifies the model architecture.''' ), ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the folder to store the PyTorch model or dataset/vocab.''', ) parser.add_argument( '''--finetuning_task''', default=None, type=str, help='''Name of a task on which the XLNet TensorFlow model was fine-tuned''', ) a__ : str = parser.parse_args() print(args) convert_xlnet_checkpoint_to_pytorch( args.tf_checkpoint_path, args.xlnet_config_file, args.pytorch_dump_folder_path, args.finetuning_task )
313
from __future__ import annotations import math def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if len(a__ ) != 2 or len(a[0] ) != 2 or len(a__ ) != 2 or len(b[0] ) != 2: raise Exception('''Matrices are not 2x2''' ) SCREAMING_SNAKE_CASE : Dict = [ [a[0][0] * b[0][0] + a[0][1] * b[1][0], a[0][0] * b[0][1] + a[0][1] * b[1][1]], [a[1][0] * b[0][0] + a[1][1] * b[1][0], a[1][0] * b[0][1] + a[1][1] * b[1][1]], ] return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] + matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] - matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) % 2 != 0 or len(a[0] ) % 2 != 0: raise Exception('''Odd matrices are not supported!''' ) SCREAMING_SNAKE_CASE : str = len(a__ ) SCREAMING_SNAKE_CASE : Any = matrix_length // 2 SCREAMING_SNAKE_CASE : Tuple = [[a[i][j] for j in range(a__ , a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : Optional[int] = [ [a[i][j] for j in range(a__ , a__ )] for i in range(a__ , a__ ) ] SCREAMING_SNAKE_CASE : Optional[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : List[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ , a__ )] return top_left, top_right, bot_left, bot_right def UpperCAmelCase_( a__ ): """simple docstring""" return len(a__ ), len(matrix[0] ) def UpperCAmelCase_( a__ ): """simple docstring""" print('''\n'''.join(str(a__ ) for line in matrix ) ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ ) == (2, 2): return default_matrix_multiplication(a__ , a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE : Dict = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : int = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Any = actual_strassen(matrix_addition(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = matrix_addition(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = matrix_subtraction(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) # construct the new matrix from our 4 quadrants SCREAMING_SNAKE_CASE : Optional[Any] = [] for i in range(len(a__ ) ): new_matrix.append(top_left[i] + top_right[i] ) for i in range(len(a__ ) ): new_matrix.append(bot_left[i] + bot_right[i] ) return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ )[1] != matrix_dimensions(a__ )[0]: SCREAMING_SNAKE_CASE : Any = ( '''Unable to multiply these matrices, please check the dimensions.\n''' F"""Matrix A: {matrixa}\n""" F"""Matrix B: {matrixa}""" ) raise Exception(a__ ) SCREAMING_SNAKE_CASE : str = matrix_dimensions(a__ ) SCREAMING_SNAKE_CASE : Tuple = matrix_dimensions(a__ ) if dimensiona[0] == dimensiona[1] and dimensiona[0] == dimensiona[1]: return [matrixa, matrixa] SCREAMING_SNAKE_CASE : str = max(*a__ , *a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = int(math.pow(2 , math.ceil(math.loga(a__ ) ) ) ) SCREAMING_SNAKE_CASE : Optional[int] = matrixa SCREAMING_SNAKE_CASE : Tuple = matrixa # Adding zeros to the matrices so that the arrays dimensions are the same and also # power of 2 for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) SCREAMING_SNAKE_CASE : Optional[Any] = actual_strassen(a__ , a__ ) # Removing the additional zeros for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): final_matrix[i].pop() else: final_matrix.pop() return final_matrix if __name__ == "__main__": a__ : Dict = [ [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 2, 3, 1], ] a__ : Union[str, Any] = [[0, 2, 1, 1], [16, 2, 3, 3], [2, 2, 7, 7], [13, 11, 22, 4]] print(strassen(matrixa, matrixa))
313
1
import torch import torch.nn as nn from transformers.modeling_utils import ModuleUtilsMixin from transformers.models.ta.modeling_ta import TaBlock, TaConfig, TaLayerNorm from ...configuration_utils import ConfigMixin, register_to_config from ...models import ModelMixin class a_ ( a__ , a__ , a__ ): """simple docstring""" @register_to_config def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = False , ) ->int: super().__init__() SCREAMING_SNAKE_CASE : List[Any] = nn.Embedding(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.Embedding(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : str = False SCREAMING_SNAKE_CASE : List[Any] = nn.Dropout(p=_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = TaConfig( vocab_size=_lowerCamelCase , d_model=_lowerCamelCase , num_heads=_lowerCamelCase , d_kv=_lowerCamelCase , d_ff=_lowerCamelCase , dropout_rate=_lowerCamelCase , feed_forward_proj=_lowerCamelCase , is_decoder=_lowerCamelCase , is_encoder_decoder=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Optional[int] = nn.ModuleList() for lyr_num in range(_lowerCamelCase ): SCREAMING_SNAKE_CASE : str = TaBlock(_lowerCamelCase ) self.encoders.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = TaLayerNorm(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.Dropout(p=_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->List[str]: SCREAMING_SNAKE_CASE : int = self.token_embedder(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = encoder_input_tokens.shape[1] SCREAMING_SNAKE_CASE : List[Any] = torch.arange(_lowerCamelCase , device=encoder_input_tokens.device ) x += self.position_encoding(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = self.dropout_pre(_lowerCamelCase ) # inverted the attention mask SCREAMING_SNAKE_CASE : List[str] = encoder_input_tokens.size() SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_extended_attention_mask(_lowerCamelCase , _lowerCamelCase ) for lyr in self.encoders: SCREAMING_SNAKE_CASE : Any = lyr(_lowerCamelCase , _lowerCamelCase )[0] SCREAMING_SNAKE_CASE : Optional[int] = self.layer_norm(_lowerCamelCase ) return self.dropout_post(_lowerCamelCase ), encoder_inputs_mask
313
# We ignore warnings about stepping the scheduler since we step it ourselves during gradient accumulation import warnings from .state import AcceleratorState, GradientState warnings.filterwarnings('''ignore''', category=UserWarning, module='''torch.optim.lr_scheduler''') class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , _lowerCamelCase = False ) ->Any: SCREAMING_SNAKE_CASE : str = scheduler SCREAMING_SNAKE_CASE : List[str] = optimizers if isinstance(_lowerCamelCase , (list, tuple) ) else [optimizers] SCREAMING_SNAKE_CASE : Union[str, Any] = split_batches SCREAMING_SNAKE_CASE : List[Any] = step_with_optimizer SCREAMING_SNAKE_CASE : List[str] = GradientState() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->Optional[Any]: if not self.step_with_optimizer: # No link between scheduler and optimizer -> just step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) return # Otherwise, first make sure the optimizer was stepped. if not self.gradient_state.sync_gradients: if self.gradient_state.adjust_scheduler: self.scheduler._step_count += 1 return for opt in self.optimizers: if opt.step_was_skipped: return if self.split_batches: # Split batches -> the training dataloader batch size is not changed so one step per training step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: # Otherwise the training dataloader batch size was multiplied by `num_processes`, so we need to do # num_processes steps per training step SCREAMING_SNAKE_CASE : List[str] = AcceleratorState().num_processes for _ in range(_lowerCamelCase ): # Special case when using OneCycle and `drop_last` was not used if hasattr(self.scheduler , '''total_steps''' ): if self.scheduler._step_count <= self.scheduler.total_steps: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Union[str, Any]: return self.scheduler.get_last_lr() def __lowerCAmelCase ( self ) ->List[str]: return self.scheduler.state_dict() def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: self.scheduler.load_state_dict(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: return self.scheduler.get_lr() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->List[str]: return self.scheduler.print_lr(*_lowerCamelCase , **_lowerCamelCase )
313
1
import gc import random import unittest import numpy as np import torch from transformers import XLMRobertaTokenizer from diffusers import ( AltDiffusionImgaImgPipeline, AutoencoderKL, PNDMScheduler, UNetaDConditionModel, ) from diffusers.image_processor import VaeImageProcessor from diffusers.pipelines.alt_diffusion.modeling_roberta_series import ( RobertaSeriesConfig, RobertaSeriesModelWithTransformation, ) from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu enable_full_determinism() class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Dict: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() @property def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Tuple = 1 SCREAMING_SNAKE_CASE : Any = 3 SCREAMING_SNAKE_CASE : Union[str, Any] = (32, 32) SCREAMING_SNAKE_CASE : Optional[Any] = floats_tensor((batch_size, num_channels) + sizes , rng=random.Random(0 ) ).to(_lowerCamelCase ) return image @property def __lowerCAmelCase ( self ) ->Any: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = UNetaDConditionModel( block_out_channels=(32, 64) , layers_per_block=2 , sample_size=32 , in_channels=4 , out_channels=4 , down_block_types=('''DownBlock2D''', '''CrossAttnDownBlock2D''') , up_block_types=('''CrossAttnUpBlock2D''', '''UpBlock2D''') , cross_attention_dim=32 , ) return model @property def __lowerCAmelCase ( self ) ->List[Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = 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 , ) return model @property def __lowerCAmelCase ( self ) ->List[Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = RobertaSeriesConfig( hidden_size=32 , project_dim=32 , intermediate_size=37 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=5006 , ) return RobertaSeriesModelWithTransformation(_lowerCamelCase ) @property def __lowerCAmelCase ( self ) ->Optional[Any]: def extract(*_lowerCamelCase , **_lowerCamelCase ): class a_ : """simple docstring""" def __init__( self ) ->Any: SCREAMING_SNAKE_CASE : List[Any] = torch.ones([0] ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: self.pixel_values.to(_lowerCamelCase ) return self return Out() return extract def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[int] = '''cpu''' # ensure determinism for the device-dependent torch.Generator SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_cond_unet SCREAMING_SNAKE_CASE : Optional[Any] = PNDMScheduler(skip_prk_steps=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_vae SCREAMING_SNAKE_CASE : Optional[Any] = self.dummy_text_encoder SCREAMING_SNAKE_CASE : Any = XLMRobertaTokenizer.from_pretrained('''hf-internal-testing/tiny-xlm-roberta''' ) SCREAMING_SNAKE_CASE : int = 77 SCREAMING_SNAKE_CASE : int = self.dummy_image.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = init_image / 2 + 0.5 # make sure here that pndm scheduler skips prk SCREAMING_SNAKE_CASE : Union[str, Any] = AltDiffusionImgaImgPipeline( unet=_lowerCamelCase , scheduler=_lowerCamelCase , vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=self.dummy_extractor , ) SCREAMING_SNAKE_CASE : Dict = VaeImageProcessor(vae_scale_factor=alt_pipe.vae_scale_factor , do_normalize=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = alt_pipe.to(_lowerCamelCase ) alt_pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = '''A painting of a squirrel eating a burger''' SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device=_lowerCamelCase ).manual_seed(0 ) SCREAMING_SNAKE_CASE : int = alt_pipe( [prompt] , generator=_lowerCamelCase , guidance_scale=6.0 , num_inference_steps=2 , output_type='''np''' , image=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Tuple = output.images SCREAMING_SNAKE_CASE : Tuple = torch.Generator(device=_lowerCamelCase ).manual_seed(0 ) SCREAMING_SNAKE_CASE : List[Any] = alt_pipe( [prompt] , generator=_lowerCamelCase , guidance_scale=6.0 , num_inference_steps=2 , output_type='''np''' , image=_lowerCamelCase , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : List[Any] = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : Optional[Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array([0.4_4_2_7, 0.3_7_3_1, 0.4_2_4_9, 0.4_9_4_1, 0.4_5_4_6, 0.4_1_4_8, 0.4_1_9_3, 0.4_6_6_6, 0.4_4_9_9] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 5e-3 assert np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 5e-3 @unittest.skipIf(torch_device != '''cuda''' , '''This test requires a GPU''' ) def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_cond_unet SCREAMING_SNAKE_CASE : Union[str, Any] = PNDMScheduler(skip_prk_steps=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_vae SCREAMING_SNAKE_CASE : int = self.dummy_text_encoder SCREAMING_SNAKE_CASE : Dict = XLMRobertaTokenizer.from_pretrained('''hf-internal-testing/tiny-xlm-roberta''' ) SCREAMING_SNAKE_CASE : Optional[int] = 77 SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_image.to(_lowerCamelCase ) # put models in fp16 SCREAMING_SNAKE_CASE : List[Any] = unet.half() SCREAMING_SNAKE_CASE : List[Any] = vae.half() SCREAMING_SNAKE_CASE : str = bert.half() # make sure here that pndm scheduler skips prk SCREAMING_SNAKE_CASE : Tuple = AltDiffusionImgaImgPipeline( unet=_lowerCamelCase , scheduler=_lowerCamelCase , vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=self.dummy_extractor , ) SCREAMING_SNAKE_CASE : List[Any] = VaeImageProcessor(vae_scale_factor=alt_pipe.vae_scale_factor , do_normalize=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = alt_pipe.to(_lowerCamelCase ) alt_pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = '''A painting of a squirrel eating a burger''' SCREAMING_SNAKE_CASE : str = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[Any] = alt_pipe( [prompt] , generator=_lowerCamelCase , num_inference_steps=2 , output_type='''np''' , image=_lowerCamelCase , ).images assert image.shape == (1, 32, 32, 3) @unittest.skipIf(torch_device != '''cuda''' , '''This test requires a GPU''' ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : str = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/img2img/sketch-mountains-input.jpg''' ) # resize to resolution that is divisible by 8 but not 16 or 32 SCREAMING_SNAKE_CASE : Union[str, Any] = init_image.resize((760, 504) ) SCREAMING_SNAKE_CASE : str = '''BAAI/AltDiffusion''' SCREAMING_SNAKE_CASE : List[Any] = AltDiffusionImgaImgPipeline.from_pretrained( _lowerCamelCase , safety_checker=_lowerCamelCase , ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) pipe.enable_attention_slicing() SCREAMING_SNAKE_CASE : Optional[int] = '''A fantasy landscape, trending on artstation''' SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Dict = pipe( prompt=_lowerCamelCase , image=_lowerCamelCase , strength=0.7_5 , guidance_scale=7.5 , generator=_lowerCamelCase , output_type='''np''' , ) SCREAMING_SNAKE_CASE : int = output.images[0] SCREAMING_SNAKE_CASE : str = image[255:258, 383:386, -1] assert image.shape == (504, 760, 3) SCREAMING_SNAKE_CASE : Optional[Any] = np.array([0.9_3_5_8, 0.9_3_9_7, 0.9_5_9_9, 0.9_9_0_1, 1.0_0_0_0, 1.0_0_0_0, 0.9_8_8_2, 1.0_0_0_0, 1.0_0_0_0] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->int: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/img2img/sketch-mountains-input.jpg''' ) SCREAMING_SNAKE_CASE : Union[str, Any] = init_image.resize((768, 512) ) SCREAMING_SNAKE_CASE : Any = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/img2img/fantasy_landscape_alt.npy''' ) SCREAMING_SNAKE_CASE : List[str] = '''BAAI/AltDiffusion''' SCREAMING_SNAKE_CASE : Tuple = AltDiffusionImgaImgPipeline.from_pretrained( _lowerCamelCase , safety_checker=_lowerCamelCase , ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) pipe.enable_attention_slicing() SCREAMING_SNAKE_CASE : str = '''A fantasy landscape, trending on artstation''' SCREAMING_SNAKE_CASE : Dict = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = pipe( prompt=_lowerCamelCase , image=_lowerCamelCase , strength=0.7_5 , guidance_scale=7.5 , generator=_lowerCamelCase , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Optional[Any] = output.images[0] assert image.shape == (512, 768, 3) # img2img is flaky across GPUs even in fp32, so using MAE here assert np.abs(expected_image - image ).max() < 1e-2
313
import argparse import os from pathlib import Path from typing import Dict import tensorflow as tf import torch from tqdm import tqdm from transformers import PegasusConfig, PegasusForConditionalGeneration, PegasusTokenizer from transformers.models.pegasus.configuration_pegasus import DEFAULTS, task_specific_params a__ : Optional[Any] = [ # replace left string with right string to get the relevant state_dict key (identical state dict to bart) ['''memory_attention''', '''encoder_attn'''], ['''attention''', '''attn'''], ['''/''', '''.'''], ['''.LayerNorm.gamma''', '''_layer_norm.weight'''], ['''.LayerNorm.beta''', '''_layer_norm.bias'''], ['''r.layer_''', '''r.layers.'''], ['''output_proj''', '''out_proj'''], ['''ffn.dense_1.''', '''fc2.'''], ['''ffn.dense.''', '''fc1.'''], ['''ffn_layer_norm''', '''final_layer_norm'''], ['''kernel''', '''weight'''], ['''encoder_layer_norm.''', '''encoder.layer_norm.'''], ['''decoder_layer_norm.''', '''decoder.layer_norm.'''], ['''embeddings.weights''', '''shared.weight'''], ] def UpperCAmelCase_( a__ ): """simple docstring""" for pegasus_name, hf_name in PATTERNS: SCREAMING_SNAKE_CASE : Union[str, Any] = k.replace(a__ , a__ ) return k def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = DEFAULTS.copy() cfg_kwargs.update(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = PegasusConfig(**a__ ) SCREAMING_SNAKE_CASE : Optional[int] = PegasusForConditionalGeneration(a__ ) SCREAMING_SNAKE_CASE : Dict = torch_model.model.state_dict() SCREAMING_SNAKE_CASE : List[str] = {} for k, v in tf_weights.items(): SCREAMING_SNAKE_CASE : int = rename_state_dict_key(a__ ) if new_k not in sd: raise ValueError(F"""could not find new key {new_k} in state dict. (converted from {k})""" ) if "dense" in k or "proj" in new_k: SCREAMING_SNAKE_CASE : Dict = v.T SCREAMING_SNAKE_CASE : Tuple = torch.tensor(a__ , dtype=sd[new_k].dtype ) assert v.shape == sd[new_k].shape, F"""{new_k}, {k}, {v.shape}, {sd[new_k].shape}""" # make sure embedding.padding_idx is respected SCREAMING_SNAKE_CASE : Tuple = torch.zeros_like(mapping['''shared.weight'''][cfg.pad_token_id + 1] ) SCREAMING_SNAKE_CASE : int = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Union[str, Any] = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Optional[Any] = {k: torch.zeros_like(a__ ) for k, v in sd.items() if k.endswith('''bias''' ) and k not in mapping} mapping.update(**a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = torch_model.model.load_state_dict(a__ , strict=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = [ k for k in missing if k not in ['''encoder.embed_positions.weight''', '''decoder.embed_positions.weight'''] ] assert unexpected_missing == [], F"""no matches found for the following torch keys {unexpected_missing}""" assert extra == [], F"""no matches found for the following tf keys {extra}""" return torch_model def UpperCAmelCase_( a__="./ckpt/aeslc/model.ckpt-32000" ): """simple docstring""" SCREAMING_SNAKE_CASE : str = tf.train.list_variables(a__ ) SCREAMING_SNAKE_CASE : str = {} SCREAMING_SNAKE_CASE : List[Any] = ['''Adafactor''', '''global_step'''] for name, shape in tqdm(a__ , desc='''converting tf checkpoint to dict''' ): SCREAMING_SNAKE_CASE : Union[str, Any] = any(pat in name for pat in ignore_name ) if skip_key: continue SCREAMING_SNAKE_CASE : Dict = tf.train.load_variable(a__ , a__ ) SCREAMING_SNAKE_CASE : Any = array return tf_weights def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = Path(a__ ).parent.name SCREAMING_SNAKE_CASE : Union[str, Any] = task_specific_params[F"""summarization_{dataset}"""]['''max_position_embeddings'''] SCREAMING_SNAKE_CASE : Dict = PegasusTokenizer.from_pretrained('''sshleifer/pegasus''' , model_max_length=a__ ) assert tok.model_max_length == desired_max_model_length tok.save_pretrained(a__ ) # convert model SCREAMING_SNAKE_CASE : Any = get_tf_weights_as_numpy(a__ ) SCREAMING_SNAKE_CASE : List[str] = task_specific_params[F"""summarization_{dataset}"""] if dataset == "large": SCREAMING_SNAKE_CASE : int = task_specific_params SCREAMING_SNAKE_CASE : List[str] = convert_pegasus(a__ , a__ ) torch_model.save_pretrained(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch_model.state_dict() sd.pop('''model.decoder.embed_positions.weight''' ) sd.pop('''model.encoder.embed_positions.weight''' ) torch.save(a__ , Path(a__ ) / '''pytorch_model.bin''' ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument('''tf_ckpt_path''', type=str, help='''passed to tf.train.list_variables''') parser.add_argument('''save_dir''', default=None, type=str, help='''Path to the output PyTorch model.''') a__ : List[str] = parser.parse_args() if args.save_dir is None: a__ : Any = Path(args.tf_ckpt_path).parent.name a__ : int = os.path.join('''pegasus''', dataset) convert_pegasus_ckpt_to_pytorch(args.tf_ckpt_path, args.save_dir)
313
1
import argparse import math import traceback import dateutil.parser as date_parser import requests def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = {} SCREAMING_SNAKE_CASE : int = job['''started_at'''] SCREAMING_SNAKE_CASE : int = job['''completed_at'''] SCREAMING_SNAKE_CASE : Optional[Any] = date_parser.parse(a__ ) SCREAMING_SNAKE_CASE : int = date_parser.parse(a__ ) SCREAMING_SNAKE_CASE : List[str] = round((end_datetime - start_datetime).total_seconds() / 60.0 ) SCREAMING_SNAKE_CASE : Union[str, Any] = start SCREAMING_SNAKE_CASE : Tuple = end SCREAMING_SNAKE_CASE : List[Any] = duration_in_min return job_info def UpperCAmelCase_( a__ , a__=None ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = None if token is not None: SCREAMING_SNAKE_CASE : Any = {'''Accept''': '''application/vnd.github+json''', '''Authorization''': F"""Bearer {token}"""} SCREAMING_SNAKE_CASE : List[Any] = F"""https://api.github.com/repos/huggingface/transformers/actions/runs/{workflow_run_id}/jobs?per_page=100""" SCREAMING_SNAKE_CASE : Any = requests.get(a__ , headers=a__ ).json() SCREAMING_SNAKE_CASE : Union[str, Any] = {} try: job_time.update({job['''name''']: extract_time_from_single_job(a__ ) for job in result['''jobs''']} ) SCREAMING_SNAKE_CASE : Union[str, Any] = math.ceil((result['''total_count'''] - 100) / 100 ) for i in range(a__ ): SCREAMING_SNAKE_CASE : Any = requests.get(url + F"""&page={i + 2}""" , headers=a__ ).json() job_time.update({job['''name''']: extract_time_from_single_job(a__ ) for job in result['''jobs''']} ) return job_time except Exception: print(F"""Unknown error, could not fetch links:\n{traceback.format_exc()}""" ) return {} if __name__ == "__main__": a__ : Any = argparse.ArgumentParser() # Required parameters parser.add_argument('''--workflow_run_id''', type=str, required=True, help='''A GitHub Actions workflow run id.''') a__ : Dict = parser.parse_args() a__ : List[Any] = get_job_time(args.workflow_run_id) a__ : str = dict(sorted(job_time.items(), key=lambda item: item[1]["duration"], reverse=True)) for k, v in job_time.items(): print(F"{k}: {v['duration']}")
313
import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import XLMRobertaTokenizerFast from diffusers import DDIMScheduler, KandinskyImgaImgPipeline, KandinskyPriorPipeline, UNetaDConditionModel, VQModel from diffusers.pipelines.kandinsky.text_encoder import MCLIPConfig, MultilingualCLIP from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = KandinskyImgaImgPipeline __SCREAMING_SNAKE_CASE : str = ['prompt', 'image_embeds', 'negative_image_embeds', 'image'] __SCREAMING_SNAKE_CASE : int = [ 'prompt', 'negative_prompt', 'image_embeds', 'negative_image_embeds', 'image', ] __SCREAMING_SNAKE_CASE : int = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'negative_prompt', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->int: return 32 @property def __lowerCAmelCase ( self ) ->List[str]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Tuple: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 100 @property def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = XLMRobertaTokenizerFast.from_pretrained('''YiYiXu/tiny-random-mclip-base''' ) return tokenizer @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[Any] = MCLIPConfig( numDims=self.cross_attention_dim , transformerDimensions=self.text_embedder_hidden_size , hidden_size=self.text_embedder_hidden_size , intermediate_size=37 , num_attention_heads=4 , num_hidden_layers=5 , vocab_size=1005 , ) SCREAMING_SNAKE_CASE : Dict = MultilingualCLIP(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = text_encoder.eval() return text_encoder @property def __lowerCAmelCase ( self ) ->Union[str, Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = { '''in_channels''': 4, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''text_image''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''text_image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : Tuple = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->List[str]: return { "block_out_channels": [32, 64], "down_block_types": ["DownEncoderBlock2D", "AttnDownEncoderBlock2D"], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": [ "AttnUpDecoderBlock2D", "UpDecoderBlock2D", ], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Optional[Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.dummy_text_encoder SCREAMING_SNAKE_CASE : Any = self.dummy_tokenizer SCREAMING_SNAKE_CASE : List[Any] = self.dummy_unet SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_movq SCREAMING_SNAKE_CASE : Optional[Any] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : Optional[Any] = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = { '''text_encoder''': text_encoder, '''tokenizer''': tokenizer, '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->str: SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(seed + 1 ) ).to(_lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : str = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : str = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : str = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = { '''prompt''': '''horse''', '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : List[str] = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : List[Any] = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : List[Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Optional[int] = np.array( [0.6_1_4_7_4_9_4_3, 0.6_0_7_3_5_3_9, 0.4_3_3_0_8_5_4_4, 0.5_9_2_8_2_6_9, 0.4_7_4_9_3_5_9_5, 0.4_6_7_5_5_9_7_3, 0.4_6_1_3_8_3_8, 0.4_5_3_6_8_7_9_7, 0.5_0_1_1_9_2_3_3] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/kandinsky_img2img_frog.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : str = '''A red cartoon frog, 4k''' SCREAMING_SNAKE_CASE : Any = KandinskyPriorPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = KandinskyImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : str = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=5 , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : Dict = pipeline( _lowerCamelCase , image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=768 , width=768 , strength=0.2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Tuple = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a__ : Union[str, Any] = { '''configuration_upernet''': ['''UperNetConfig'''], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : int = [ '''UperNetForSemanticSegmentation''', '''UperNetPreTrainedModel''', ] if TYPE_CHECKING: from .configuration_upernet import UperNetConfig try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_upernet import UperNetForSemanticSegmentation, UperNetPreTrainedModel else: import sys a__ : Dict = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
import argparse import json import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ViTImageProcessor, ViTMSNConfig, ViTMSNModel from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD torch.set_grad_enabled(False) def UpperCAmelCase_( a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [] for i in range(config.num_hidden_layers ): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append((F"""module.blocks.{i}.norm1.weight""", F"""vit.encoder.layer.{i}.layernorm_before.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm1.bias""", F"""vit.encoder.layer.{i}.layernorm_before.bias""") ) rename_keys.append( (F"""module.blocks.{i}.attn.proj.weight""", F"""vit.encoder.layer.{i}.attention.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.attn.proj.bias""", F"""vit.encoder.layer.{i}.attention.output.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.norm2.weight""", F"""vit.encoder.layer.{i}.layernorm_after.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm2.bias""", F"""vit.encoder.layer.{i}.layernorm_after.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.weight""", F"""vit.encoder.layer.{i}.intermediate.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.bias""", F"""vit.encoder.layer.{i}.intermediate.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.weight""", F"""vit.encoder.layer.{i}.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.bias""", F"""vit.encoder.layer.{i}.output.dense.bias""") ) # projection layer + position embeddings rename_keys.extend( [ ('''module.cls_token''', '''vit.embeddings.cls_token'''), ('''module.patch_embed.proj.weight''', '''vit.embeddings.patch_embeddings.projection.weight'''), ('''module.patch_embed.proj.bias''', '''vit.embeddings.patch_embeddings.projection.bias'''), ('''module.pos_embed''', '''vit.embeddings.position_embeddings'''), ] ) if base_model: # layernorm + pooler rename_keys.extend( [ ('''module.norm.weight''', '''layernorm.weight'''), ('''module.norm.bias''', '''layernorm.bias'''), ] ) # if just the base model, we should remove "vit" from all keys that start with "vit" SCREAMING_SNAKE_CASE : Any = [(pair[0], pair[1][4:]) if pair[1].startswith('''vit''' ) else pair for pair in rename_keys] else: # layernorm + classification head rename_keys.extend( [ ('''norm.weight''', '''vit.layernorm.weight'''), ('''norm.bias''', '''vit.layernorm.bias'''), ('''head.weight''', '''classifier.weight'''), ('''head.bias''', '''classifier.bias'''), ] ) return rename_keys def UpperCAmelCase_( a__ , a__ , a__=False ): """simple docstring""" for i in range(config.num_hidden_layers ): if base_model: SCREAMING_SNAKE_CASE : Any = '''''' else: SCREAMING_SNAKE_CASE : Optional[int] = '''vit.''' # read in weights + bias of input projection layer (in timm, this is a single matrix + bias) SCREAMING_SNAKE_CASE : Optional[Any] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.weight""" ) SCREAMING_SNAKE_CASE : List[str] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.bias""" ) # next, add query, keys and values (in that order) to the state dict SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ : config.hidden_size, : ] SCREAMING_SNAKE_CASE : Any = in_proj_bias[: config.hidden_size] SCREAMING_SNAKE_CASE : int = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] SCREAMING_SNAKE_CASE : List[Any] = in_proj_bias[ config.hidden_size : config.hidden_size * 2 ] SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ -config.hidden_size :, : ] SCREAMING_SNAKE_CASE : Optional[Any] = in_proj_bias[-config.hidden_size :] def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = ['''head.weight''', '''head.bias'''] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = [ '''module.fc.fc1.weight''', '''module.fc.fc1.bias''', '''module.fc.bn1.weight''', '''module.fc.bn1.bias''', '''module.fc.bn1.running_mean''', '''module.fc.bn1.running_var''', '''module.fc.bn1.num_batches_tracked''', '''module.fc.fc2.weight''', '''module.fc.fc2.bias''', '''module.fc.bn2.weight''', '''module.fc.bn2.bias''', '''module.fc.bn2.running_mean''', '''module.fc.bn2.running_var''', '''module.fc.bn2.num_batches_tracked''', '''module.fc.fc3.weight''', '''module.fc.fc3.bias''', ] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = dct.pop(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = val def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = ViTMSNConfig() SCREAMING_SNAKE_CASE : Optional[int] = 1_000 SCREAMING_SNAKE_CASE : str = '''datasets/huggingface/label-files''' SCREAMING_SNAKE_CASE : List[str] = '''imagenet-1k-id2label.json''' SCREAMING_SNAKE_CASE : Union[str, Any] = json.load(open(hf_hub_download(a__ , a__ ) , '''r''' ) ) SCREAMING_SNAKE_CASE : List[Any] = {int(a__ ): v for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : str = idalabel SCREAMING_SNAKE_CASE : int = {v: k for k, v in idalabel.items()} if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = 384 SCREAMING_SNAKE_CASE : Any = 1_536 SCREAMING_SNAKE_CASE : List[str] = 6 elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Optional[int] = 1_024 SCREAMING_SNAKE_CASE : Optional[int] = 4_096 SCREAMING_SNAKE_CASE : Tuple = 24 SCREAMING_SNAKE_CASE : Union[str, Any] = 16 SCREAMING_SNAKE_CASE : Dict = 0.1 elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = 4 elif "l7" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = 7 SCREAMING_SNAKE_CASE : Union[str, Any] = 1_024 SCREAMING_SNAKE_CASE : List[Any] = 4_096 SCREAMING_SNAKE_CASE : List[Any] = 24 SCREAMING_SNAKE_CASE : Tuple = 16 SCREAMING_SNAKE_CASE : Union[str, Any] = 0.1 SCREAMING_SNAKE_CASE : Union[str, Any] = ViTMSNModel(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = torch.hub.load_state_dict_from_url(a__ , map_location='''cpu''' )['''target_encoder'''] SCREAMING_SNAKE_CASE : Any = ViTImageProcessor(size=config.image_size ) remove_projection_head(a__ ) SCREAMING_SNAKE_CASE : Any = create_rename_keys(a__ , base_model=a__ ) for src, dest in rename_keys: rename_key(a__ , a__ , a__ ) read_in_q_k_v(a__ , a__ , base_model=a__ ) model.load_state_dict(a__ ) model.eval() SCREAMING_SNAKE_CASE : Any = '''http://images.cocodataset.org/val2017/000000039769.jpg''' SCREAMING_SNAKE_CASE : Dict = Image.open(requests.get(a__ , stream=a__ ).raw ) SCREAMING_SNAKE_CASE : Optional[int] = ViTImageProcessor( size=config.image_size , image_mean=a__ , image_std=a__ ) SCREAMING_SNAKE_CASE : int = image_processor(images=a__ , return_tensors='''pt''' ) # forward pass torch.manual_seed(2 ) SCREAMING_SNAKE_CASE : Tuple = model(**a__ ) SCREAMING_SNAKE_CASE : str = outputs.last_hidden_state # The following Colab Notebook was used to generate these outputs: # https://colab.research.google.com/gist/sayakpaul/3672419a04f5997827503fd84079bdd1/scratchpad.ipynb if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[-1.0_915, -1.4_876, -1.1_809]] ) elif "b16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[14.2_889, -18.9_045, 11.7_281]] ) elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.tensor([[41.5_028, -22.8_681, 45.6_475]] ) elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = torch.tensor([[-4.3_868, 5.2_932, -0.4_137]] ) else: SCREAMING_SNAKE_CASE : List[str] = torch.tensor([[-0.1_792, -0.6_465, 2.4_263]] ) # verify logits assert torch.allclose(last_hidden_state[:, 0, :3] , a__ , atol=1e-4 ) print(F"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(a__ ) print(F"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(a__ ) if __name__ == "__main__": a__ : Dict = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--checkpoint_url''', default='''https://dl.fbaipublicfiles.com/msn/vits16_800ep.pth.tar''', type=str, help='''URL of the checkpoint you\'d like to convert.''', ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.''' ) a__ : Any = parser.parse_args() convert_vit_msn_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
313
1
# Function to print upper half of diamond (pyramid) def UpperCAmelCase_( a__ ): """simple docstring""" for i in range(0 , a__ ): for _ in range(0 , n - i - 1 ): # printing spaces print(''' ''' , end='''''' ) for _ in range(0 , i + 1 ): # printing stars print('''* ''' , end='''''' ) print() def UpperCAmelCase_( a__ ): """simple docstring""" for i in range(a__ , 0 , -1 ): for _ in range(a__ , 0 , -1 ): # printing stars print('''* ''' , end='''''' ) print() for _ in range(n - i + 1 , 0 , -1 ): # printing spaces print(''' ''' , end='''''' ) def UpperCAmelCase_( a__ ): """simple docstring""" if n <= 0: print(''' ... .... nothing printing :(''' ) return floyd(a__ ) # upper half reverse_floyd(a__ ) # lower half if __name__ == "__main__": print(r'''| /\ | |- | |- |--| |\ /| |-''') print(r'''|/ \| |- |_ |_ |__| | \/ | |_''') a__ : Any = 1 while K: a__ : Optional[int] = int(input('''enter the number and , and see the magic : ''')) print() pretty_print(user_number) a__ : Optional[Any] = int(input('''press 0 to exit... and 1 to continue...''')) print('''Good Bye...''')
313
import csv import tweepy # Twitter API credentials a__ : Union[str, Any] = '''''' a__ : List[str] = '''''' a__ : Any = '''''' a__ : List[str] = '''''' def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = tweepy.OAuthHandler(a__ , a__ ) auth.set_access_token(a__ , a__ ) SCREAMING_SNAKE_CASE : List[str] = tweepy.API(a__ ) # initialize a list to hold all the tweepy Tweets SCREAMING_SNAKE_CASE : Any = [] # make initial request for most recent tweets (200 is the maximum allowed count) SCREAMING_SNAKE_CASE : List[Any] = api.user_timeline(screen_name=a__ , count=200 ) # save most recent tweets alltweets.extend(a__ ) # save the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Tuple = alltweets[-1].id - 1 # keep grabbing tweets until there are no tweets left to grab while len(a__ ) > 0: print(F"""getting tweets before {oldest}""" ) # all subsequent requests use the max_id param to prevent duplicates SCREAMING_SNAKE_CASE : Any = api.user_timeline( screen_name=a__ , count=200 , max_id=a__ ) # save most recent tweets alltweets.extend(a__ ) # update the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Dict = alltweets[-1].id - 1 print(F"""...{len(a__ )} tweets downloaded so far""" ) # transform the tweepy tweets into a 2D array that will populate the csv SCREAMING_SNAKE_CASE : Optional[Any] = [[tweet.id_str, tweet.created_at, tweet.text] for tweet in alltweets] # write the csv with open(F"""new_{screen_name}_tweets.csv""" , '''w''' ) as f: SCREAMING_SNAKE_CASE : List[Any] = csv.writer(a__ ) writer.writerow(['''id''', '''created_at''', '''text'''] ) writer.writerows(a__ ) if __name__ == "__main__": # pass in the username of the account you want to download get_all_tweets('''FirePing32''')
313
1
import inspect import unittest from transformers import MobileViTVaConfig from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device from transformers.utils import cached_property, is_torch_available, is_vision_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 transformers import MobileViTVaForImageClassification, MobileViTVaForSemanticSegmentation, MobileViTVaModel from transformers.models.mobilevitva.modeling_mobilevitva import ( MOBILEVITV2_PRETRAINED_MODEL_ARCHIVE_LIST, make_divisible, ) if is_vision_available(): from PIL import Image from transformers import MobileViTImageProcessor class a_ ( a__ ): """simple docstring""" def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : int = self.config_class(**self.inputs_dict ) self.parent.assertTrue(hasattr(_lowerCamelCase , '''width_multiplier''' ) ) class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=13 , _lowerCamelCase=64 , _lowerCamelCase=2 , _lowerCamelCase=3 , _lowerCamelCase="swish" , _lowerCamelCase=3 , _lowerCamelCase=32 , _lowerCamelCase=0.1 , _lowerCamelCase=0.0_2 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=10 , _lowerCamelCase=None , _lowerCamelCase=0.2_5 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0 , ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = parent SCREAMING_SNAKE_CASE : List[str] = batch_size SCREAMING_SNAKE_CASE : Dict = image_size SCREAMING_SNAKE_CASE : int = patch_size SCREAMING_SNAKE_CASE : Optional[int] = num_channels SCREAMING_SNAKE_CASE : int = make_divisible(512 * width_multiplier , divisor=8 ) SCREAMING_SNAKE_CASE : Union[str, Any] = hidden_act SCREAMING_SNAKE_CASE : int = conv_kernel_size SCREAMING_SNAKE_CASE : List[str] = output_stride SCREAMING_SNAKE_CASE : List[str] = classifier_dropout_prob SCREAMING_SNAKE_CASE : List[Any] = use_labels SCREAMING_SNAKE_CASE : Dict = is_training SCREAMING_SNAKE_CASE : Tuple = num_labels SCREAMING_SNAKE_CASE : Tuple = initializer_range SCREAMING_SNAKE_CASE : Any = scope SCREAMING_SNAKE_CASE : Union[str, Any] = width_multiplier SCREAMING_SNAKE_CASE : str = ffn_dropout SCREAMING_SNAKE_CASE : Optional[int] = attn_dropout def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : Dict = None SCREAMING_SNAKE_CASE : List[Any] = None if self.use_labels: SCREAMING_SNAKE_CASE : str = ids_tensor([self.batch_size] , self.num_labels ) SCREAMING_SNAKE_CASE : Union[str, Any] = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels ) SCREAMING_SNAKE_CASE : Tuple = self.get_config() return config, pixel_values, labels, pixel_labels def __lowerCAmelCase ( self ) ->Tuple: return MobileViTVaConfig( image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , hidden_act=self.hidden_act , conv_kernel_size=self.conv_kernel_size , output_stride=self.output_stride , classifier_dropout_prob=self.classifier_dropout_prob , initializer_range=self.initializer_range , width_multiplier=self.width_multiplier , ffn_dropout=self.ffn_dropout_prob , attn_dropout=self.attn_dropout_prob , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : Optional[Any] = MobileViTVaModel(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase ) self.parent.assertEqual( result.last_hidden_state.shape , ( self.batch_size, self.last_hidden_size, self.image_size // self.output_stride, self.image_size // self.output_stride, ) , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : List[str] = self.num_labels SCREAMING_SNAKE_CASE : str = MobileViTVaForImageClassification(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : Tuple = self.num_labels SCREAMING_SNAKE_CASE : Dict = MobileViTVaForSemanticSegmentation(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : int = model(_lowerCamelCase ) self.parent.assertEqual( result.logits.shape , ( self.batch_size, self.num_labels, self.image_size // self.output_stride, self.image_size // self.output_stride, ) , ) SCREAMING_SNAKE_CASE : List[str] = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual( result.logits.shape , ( self.batch_size, self.num_labels, self.image_size // self.output_stride, self.image_size // self.output_stride, ) , ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : int = self.prepare_config_and_inputs() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = config_and_inputs SCREAMING_SNAKE_CASE : Union[str, Any] = {'''pixel_values''': pixel_values} return config, inputs_dict @require_torch class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = ( (MobileViTVaModel, MobileViTVaForImageClassification, MobileViTVaForSemanticSegmentation) if is_torch_available() else () ) __SCREAMING_SNAKE_CASE : List[str] = ( { 'feature-extraction': MobileViTVaModel, 'image-classification': MobileViTVaForImageClassification, 'image-segmentation': MobileViTVaForSemanticSegmentation, } if is_torch_available() else {} ) __SCREAMING_SNAKE_CASE : Optional[Any] = False __SCREAMING_SNAKE_CASE : Any = False __SCREAMING_SNAKE_CASE : Any = False __SCREAMING_SNAKE_CASE : str = False def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = MobileViTVaModelTester(self ) SCREAMING_SNAKE_CASE : str = MobileViTVaConfigTester(self , config_class=_lowerCamelCase , has_text_modality=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: self.config_tester.run_common_tests() @unittest.skip(reason='''MobileViTV2 does not use inputs_embeds''' ) def __lowerCAmelCase ( self ) ->str: pass @unittest.skip(reason='''MobileViTV2 does not support input and output embeddings''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @unittest.skip(reason='''MobileViTV2 does not output attentions''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @require_torch_multi_gpu @unittest.skip(reason='''Got `CUDA error: misaligned address` for tests after this one being run.''' ) def __lowerCAmelCase ( self ) ->List[str]: pass @unittest.skip('''Will be fixed soon by reducing the size of the model used for common tests.''' ) def __lowerCAmelCase ( self ) ->Optional[int]: pass def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Tuple = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = inspect.signature(model.forward ) # signature.parameters is an OrderedDict => so arg_names order is deterministic SCREAMING_SNAKE_CASE : str = [*signature.parameters.keys()] SCREAMING_SNAKE_CASE : str = ['''pixel_values'''] self.assertListEqual(arg_names[:1] , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Optional[Any] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[int]: def check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : Optional[int] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : str = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = outputs.hidden_states SCREAMING_SNAKE_CASE : List[Any] = 5 self.assertEqual(len(_lowerCamelCase ) , _lowerCamelCase ) # MobileViTV2's feature maps are of shape (batch_size, num_channels, height, width) # with the width and height being successively divided by 2. SCREAMING_SNAKE_CASE : str = 2 for i in range(len(_lowerCamelCase ) ): self.assertListEqual( list(hidden_states[i].shape[-2:] ) , [self.model_tester.image_size // divisor, self.model_tester.image_size // divisor] , ) divisor *= 2 self.assertEqual(self.model_tester.output_stride , divisor // 2 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : int = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # check that output_hidden_states also work using config del inputs_dict["output_hidden_states"] SCREAMING_SNAKE_CASE : str = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : int = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_image_classification(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_semantic_segmentation(*_lowerCamelCase ) @slow def __lowerCAmelCase ( self ) ->Any: for model_name in MOBILEVITV2_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: SCREAMING_SNAKE_CASE : int = MobileViTVaModel.from_pretrained(_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) return image @require_torch @require_vision class a_ ( unittest.TestCase ): """simple docstring""" @cached_property def __lowerCAmelCase ( self ) ->str: return ( MobileViTImageProcessor.from_pretrained('''apple/mobilevitv2-1.0-imagenet1k-256''' ) if is_vision_available() else None ) @slow def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : List[Any] = MobileViTVaForImageClassification.from_pretrained('''apple/mobilevitv2-1.0-imagenet1k-256''' ).to( _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = self.default_image_processor SCREAMING_SNAKE_CASE : Optional[Any] = prepare_img() SCREAMING_SNAKE_CASE : List[Any] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : Optional[int] = model(**_lowerCamelCase ) # verify the logits SCREAMING_SNAKE_CASE : int = torch.Size((1, 1000) ) self.assertEqual(outputs.logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.tensor([-1.6336e00, -7.3204e-02, -5.1883e-01] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(outputs.logits[0, :3] , _lowerCamelCase , atol=1e-4 ) ) @slow def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Tuple = MobileViTVaForSemanticSegmentation.from_pretrained('''shehan97/mobilevitv2-1.0-voc-deeplabv3''' ) SCREAMING_SNAKE_CASE : Optional[Any] = model.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = MobileViTImageProcessor.from_pretrained('''shehan97/mobilevitv2-1.0-voc-deeplabv3''' ) SCREAMING_SNAKE_CASE : Any = prepare_img() SCREAMING_SNAKE_CASE : Union[str, Any] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : List[str] = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = outputs.logits # verify the logits SCREAMING_SNAKE_CASE : Tuple = torch.Size((1, 21, 32, 32) ) self.assertEqual(logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = torch.tensor( [ [[7.0_8_6_3, 7.1_5_2_5, 6.8_2_0_1], [6.6_9_3_1, 6.8_7_7_0, 6.8_9_3_3], [6.2_9_7_8, 7.0_3_6_6, 6.9_6_3_6]], [[-3.7_1_3_4, -3.6_7_1_2, -3.6_6_7_5], [-3.5_8_2_5, -3.3_5_4_9, -3.4_7_7_7], [-3.3_4_3_5, -3.3_9_7_9, -3.2_8_5_7]], [[-2.9_3_2_9, -2.8_0_0_3, -2.7_3_6_9], [-3.0_5_6_4, -2.4_7_8_0, -2.0_2_0_7], [-2.6_8_8_9, -1.9_2_9_8, -1.7_6_4_0]], ] , device=_lowerCamelCase , ) self.assertTrue(torch.allclose(logits[0, :3, :3, :3] , _lowerCamelCase , atol=1e-4 ) ) @slow def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Optional[int] = MobileViTVaForSemanticSegmentation.from_pretrained('''shehan97/mobilevitv2-1.0-voc-deeplabv3''' ) SCREAMING_SNAKE_CASE : Optional[int] = model.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = MobileViTImageProcessor.from_pretrained('''shehan97/mobilevitv2-1.0-voc-deeplabv3''' ) SCREAMING_SNAKE_CASE : str = prepare_img() SCREAMING_SNAKE_CASE : int = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : Dict = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = outputs.logits.detach().cpu() SCREAMING_SNAKE_CASE : str = image_processor.post_process_semantic_segmentation(outputs=_lowerCamelCase , target_sizes=[(50, 60)] ) SCREAMING_SNAKE_CASE : Tuple = torch.Size((50, 60) ) self.assertEqual(segmentation[0].shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = image_processor.post_process_semantic_segmentation(outputs=_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = torch.Size((32, 32) ) self.assertEqual(segmentation[0].shape , _lowerCamelCase )
313
from collections import OrderedDict from typing import Mapping from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a__ : Optional[Any] = logging.get_logger(__name__) a__ : List[str] = { '''kssteven/ibert-roberta-base''': '''https://huggingface.co/kssteven/ibert-roberta-base/resolve/main/config.json''', '''kssteven/ibert-roberta-large''': '''https://huggingface.co/kssteven/ibert-roberta-large/resolve/main/config.json''', '''kssteven/ibert-roberta-large-mnli''': ( '''https://huggingface.co/kssteven/ibert-roberta-large-mnli/resolve/main/config.json''' ), } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'ibert' def __init__( self , _lowerCamelCase=3_0522 , _lowerCamelCase=768 , _lowerCamelCase=12 , _lowerCamelCase=12 , _lowerCamelCase=3072 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1e-12 , _lowerCamelCase=1 , _lowerCamelCase=0 , _lowerCamelCase=2 , _lowerCamelCase="absolute" , _lowerCamelCase=False , _lowerCamelCase="none" , **_lowerCamelCase , ) ->Any: super().__init__(pad_token_id=_lowerCamelCase , bos_token_id=_lowerCamelCase , eos_token_id=_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = vocab_size SCREAMING_SNAKE_CASE : str = hidden_size SCREAMING_SNAKE_CASE : Tuple = num_hidden_layers SCREAMING_SNAKE_CASE : List[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Dict = hidden_act SCREAMING_SNAKE_CASE : Optional[int] = intermediate_size SCREAMING_SNAKE_CASE : Optional[Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Tuple = type_vocab_size SCREAMING_SNAKE_CASE : Optional[int] = initializer_range SCREAMING_SNAKE_CASE : Union[str, Any] = layer_norm_eps SCREAMING_SNAKE_CASE : str = position_embedding_type SCREAMING_SNAKE_CASE : Optional[int] = quant_mode SCREAMING_SNAKE_CASE : Dict = force_dequant class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: if self.task == "multiple-choice": SCREAMING_SNAKE_CASE : Dict = {0: '''batch''', 1: '''choice''', 2: '''sequence'''} else: SCREAMING_SNAKE_CASE : List[Any] = {0: '''batch''', 1: '''sequence'''} return OrderedDict( [ ('''input_ids''', dynamic_axis), ('''attention_mask''', dynamic_axis), ] )
313
1
import importlib import json import os from collections import OrderedDict from typing import Dict, Optional, Union # Build the list of all image processors from ...configuration_utils import PretrainedConfig from ...dynamic_module_utils import get_class_from_dynamic_module, resolve_trust_remote_code from ...image_processing_utils import ImageProcessingMixin from ...utils import CONFIG_NAME, IMAGE_PROCESSOR_NAME, get_file_from_repo, logging from .auto_factory import _LazyAutoMapping from .configuration_auto import ( CONFIG_MAPPING_NAMES, AutoConfig, model_type_to_module_name, replace_list_option_in_docstrings, ) a__ : int = logging.get_logger(__name__) a__ : List[str] = OrderedDict( [ ('''align''', '''EfficientNetImageProcessor'''), ('''beit''', '''BeitImageProcessor'''), ('''bit''', '''BitImageProcessor'''), ('''blip''', '''BlipImageProcessor'''), ('''blip-2''', '''BlipImageProcessor'''), ('''bridgetower''', '''BridgeTowerImageProcessor'''), ('''chinese_clip''', '''ChineseCLIPImageProcessor'''), ('''clip''', '''CLIPImageProcessor'''), ('''clipseg''', '''ViTImageProcessor'''), ('''conditional_detr''', '''ConditionalDetrImageProcessor'''), ('''convnext''', '''ConvNextImageProcessor'''), ('''convnextv2''', '''ConvNextImageProcessor'''), ('''cvt''', '''ConvNextImageProcessor'''), ('''data2vec-vision''', '''BeitImageProcessor'''), ('''deformable_detr''', '''DeformableDetrImageProcessor'''), ('''deit''', '''DeiTImageProcessor'''), ('''deta''', '''DetaImageProcessor'''), ('''detr''', '''DetrImageProcessor'''), ('''dinat''', '''ViTImageProcessor'''), ('''donut-swin''', '''DonutImageProcessor'''), ('''dpt''', '''DPTImageProcessor'''), ('''efficientformer''', '''EfficientFormerImageProcessor'''), ('''efficientnet''', '''EfficientNetImageProcessor'''), ('''flava''', '''FlavaImageProcessor'''), ('''focalnet''', '''BitImageProcessor'''), ('''git''', '''CLIPImageProcessor'''), ('''glpn''', '''GLPNImageProcessor'''), ('''groupvit''', '''CLIPImageProcessor'''), ('''imagegpt''', '''ImageGPTImageProcessor'''), ('''instructblip''', '''BlipImageProcessor'''), ('''layoutlmv2''', '''LayoutLMv2ImageProcessor'''), ('''layoutlmv3''', '''LayoutLMv3ImageProcessor'''), ('''levit''', '''LevitImageProcessor'''), ('''mask2former''', '''Mask2FormerImageProcessor'''), ('''maskformer''', '''MaskFormerImageProcessor'''), ('''mgp-str''', '''ViTImageProcessor'''), ('''mobilenet_v1''', '''MobileNetV1ImageProcessor'''), ('''mobilenet_v2''', '''MobileNetV2ImageProcessor'''), ('''mobilevit''', '''MobileViTImageProcessor'''), ('''mobilevit''', '''MobileViTImageProcessor'''), ('''mobilevitv2''', '''MobileViTImageProcessor'''), ('''nat''', '''ViTImageProcessor'''), ('''oneformer''', '''OneFormerImageProcessor'''), ('''owlvit''', '''OwlViTImageProcessor'''), ('''perceiver''', '''PerceiverImageProcessor'''), ('''pix2struct''', '''Pix2StructImageProcessor'''), ('''poolformer''', '''PoolFormerImageProcessor'''), ('''regnet''', '''ConvNextImageProcessor'''), ('''resnet''', '''ConvNextImageProcessor'''), ('''sam''', '''SamImageProcessor'''), ('''segformer''', '''SegformerImageProcessor'''), ('''swiftformer''', '''ViTImageProcessor'''), ('''swin''', '''ViTImageProcessor'''), ('''swin2sr''', '''Swin2SRImageProcessor'''), ('''swinv2''', '''ViTImageProcessor'''), ('''table-transformer''', '''DetrImageProcessor'''), ('''timesformer''', '''VideoMAEImageProcessor'''), ('''tvlt''', '''TvltImageProcessor'''), ('''upernet''', '''SegformerImageProcessor'''), ('''van''', '''ConvNextImageProcessor'''), ('''videomae''', '''VideoMAEImageProcessor'''), ('''vilt''', '''ViltImageProcessor'''), ('''vit''', '''ViTImageProcessor'''), ('''vit_hybrid''', '''ViTHybridImageProcessor'''), ('''vit_mae''', '''ViTImageProcessor'''), ('''vit_msn''', '''ViTImageProcessor'''), ('''xclip''', '''CLIPImageProcessor'''), ('''yolos''', '''YolosImageProcessor'''), ] ) a__ : List[str] = _LazyAutoMapping(CONFIG_MAPPING_NAMES, IMAGE_PROCESSOR_MAPPING_NAMES) def UpperCAmelCase_( a__ ): """simple docstring""" for module_name, extractors in IMAGE_PROCESSOR_MAPPING_NAMES.items(): if class_name in extractors: SCREAMING_SNAKE_CASE : Optional[int] = model_type_to_module_name(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = importlib.import_module(F""".{module_name}""" , '''transformers.models''' ) try: return getattr(a__ , a__ ) except AttributeError: continue for _, extractor in IMAGE_PROCESSOR_MAPPING._extra_content.items(): if getattr(a__ , '''__name__''' , a__ ) == class_name: return extractor # We did not fine the class, but maybe it's because a dep is missing. In that case, the class will be in the main # init and we return the proper dummy to get an appropriate error message. SCREAMING_SNAKE_CASE : Optional[int] = importlib.import_module('''transformers''' ) if hasattr(a__ , a__ ): return getattr(a__ , a__ ) return None def UpperCAmelCase_( a__ , a__ = None , a__ = False , a__ = False , a__ = None , a__ = None , a__ = None , a__ = False , **a__ , ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = get_file_from_repo( a__ , a__ , cache_dir=a__ , force_download=a__ , resume_download=a__ , proxies=a__ , use_auth_token=a__ , revision=a__ , local_files_only=a__ , ) if resolved_config_file is None: logger.info( '''Could not locate the image processor configuration file, will try to use the model config instead.''' ) return {} with open(a__ , encoding='''utf-8''' ) as reader: return json.load(a__ ) class a_ : """simple docstring""" def __init__( self ) ->Union[str, Any]: raise EnvironmentError( '''AutoImageProcessor is designed to be instantiated ''' '''using the `AutoImageProcessor.from_pretrained(pretrained_model_name_or_path)` method.''' ) @classmethod @replace_list_option_in_docstrings(_lowerCamelCase ) def __lowerCAmelCase ( cls , _lowerCamelCase , **_lowerCamelCase ) ->List[str]: SCREAMING_SNAKE_CASE : Tuple = kwargs.pop('''config''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = kwargs.pop('''trust_remote_code''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = True SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = ImageProcessingMixin.get_image_processor_dict(_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = config_dict.get('''image_processor_type''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = None if "AutoImageProcessor" in config_dict.get('''auto_map''' , {} ): SCREAMING_SNAKE_CASE : str = config_dict['''auto_map''']['''AutoImageProcessor'''] # If we still don't have the image processor class, check if we're loading from a previous feature extractor config # and if so, infer the image processor class from there. if image_processor_class is None and image_processor_auto_map is None: SCREAMING_SNAKE_CASE : Any = config_dict.pop('''feature_extractor_type''' , _lowerCamelCase ) if feature_extractor_class is not None: logger.warning( '''Could not find image processor class in the image processor config or the model config. Loading''' ''' based on pattern matching with the model\'s feature extractor configuration.''' ) SCREAMING_SNAKE_CASE : List[str] = feature_extractor_class.replace('''FeatureExtractor''' , '''ImageProcessor''' ) if "AutoFeatureExtractor" in config_dict.get('''auto_map''' , {} ): SCREAMING_SNAKE_CASE : Tuple = config_dict['''auto_map''']['''AutoFeatureExtractor'''] SCREAMING_SNAKE_CASE : List[str] = feature_extractor_auto_map.replace('''FeatureExtractor''' , '''ImageProcessor''' ) logger.warning( '''Could not find image processor auto map in the image processor config or the model config.''' ''' Loading based on pattern matching with the model\'s feature extractor configuration.''' ) # If we don't find the image processor class in the image processor config, let's try the model config. if image_processor_class is None and image_processor_auto_map is None: if not isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : int = AutoConfig.from_pretrained(_lowerCamelCase , **_lowerCamelCase ) # It could be in `config.image_processor_type`` SCREAMING_SNAKE_CASE : List[str] = getattr(_lowerCamelCase , '''image_processor_type''' , _lowerCamelCase ) if hasattr(_lowerCamelCase , '''auto_map''' ) and "AutoImageProcessor" in config.auto_map: SCREAMING_SNAKE_CASE : Tuple = config.auto_map['''AutoImageProcessor'''] if image_processor_class is not None: SCREAMING_SNAKE_CASE : Optional[int] = image_processor_class_from_name(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = image_processor_auto_map is not None SCREAMING_SNAKE_CASE : List[Any] = image_processor_class is not None or type(_lowerCamelCase ) in IMAGE_PROCESSOR_MAPPING SCREAMING_SNAKE_CASE : str = resolve_trust_remote_code( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) if has_remote_code and trust_remote_code: SCREAMING_SNAKE_CASE : List[Any] = get_class_from_dynamic_module( _lowerCamelCase , _lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = kwargs.pop('''code_revision''' , _lowerCamelCase ) if os.path.isdir(_lowerCamelCase ): image_processor_class.register_for_auto_class() return image_processor_class.from_dict(_lowerCamelCase , **_lowerCamelCase ) elif image_processor_class is not None: return image_processor_class.from_dict(_lowerCamelCase , **_lowerCamelCase ) # Last try: we use the IMAGE_PROCESSOR_MAPPING. elif type(_lowerCamelCase ) in IMAGE_PROCESSOR_MAPPING: SCREAMING_SNAKE_CASE : Any = IMAGE_PROCESSOR_MAPPING[type(_lowerCamelCase )] return image_processor_class.from_dict(_lowerCamelCase , **_lowerCamelCase ) raise ValueError( F"""Unrecognized image processor in {pretrained_model_name_or_path}. Should have a """ F"""`image_processor_type` key in its {IMAGE_PROCESSOR_NAME} of {CONFIG_NAME}, or one of the following """ F"""`model_type` keys in its {CONFIG_NAME}: {", ".join(c for c in IMAGE_PROCESSOR_MAPPING_NAMES.keys() )}""" ) @staticmethod def __lowerCAmelCase ( _lowerCamelCase , _lowerCamelCase ) ->Tuple: IMAGE_PROCESSOR_MAPPING.register(_lowerCamelCase , _lowerCamelCase )
313
from collections import OrderedDict from typing import TYPE_CHECKING, Any, Mapping, Optional from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging if TYPE_CHECKING: from ... import FeatureExtractionMixin, TensorType a__ : Any = logging.get_logger(__name__) a__ : Dict = { '''openai/imagegpt-small''': '''''', '''openai/imagegpt-medium''': '''''', '''openai/imagegpt-large''': '''''', } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 'imagegpt' __SCREAMING_SNAKE_CASE : Optional[Any] = ['past_key_values'] __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'n_embd', 'max_position_embeddings': 'n_positions', 'num_attention_heads': 'n_head', 'num_hidden_layers': 'n_layer', } def __init__( self , _lowerCamelCase=512 + 1 , _lowerCamelCase=32 * 32 , _lowerCamelCase=512 , _lowerCamelCase=24 , _lowerCamelCase=8 , _lowerCamelCase=None , _lowerCamelCase="quick_gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=1e-5 , _lowerCamelCase=0.0_2 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=False , _lowerCamelCase=False , **_lowerCamelCase , ) ->str: SCREAMING_SNAKE_CASE : Any = vocab_size SCREAMING_SNAKE_CASE : Optional[int] = n_positions SCREAMING_SNAKE_CASE : Optional[int] = n_embd SCREAMING_SNAKE_CASE : List[Any] = n_layer SCREAMING_SNAKE_CASE : List[Any] = n_head SCREAMING_SNAKE_CASE : int = n_inner SCREAMING_SNAKE_CASE : Dict = activation_function SCREAMING_SNAKE_CASE : Union[str, Any] = resid_pdrop SCREAMING_SNAKE_CASE : Dict = embd_pdrop SCREAMING_SNAKE_CASE : List[str] = attn_pdrop SCREAMING_SNAKE_CASE : Optional[int] = layer_norm_epsilon SCREAMING_SNAKE_CASE : Tuple = initializer_range SCREAMING_SNAKE_CASE : int = scale_attn_weights SCREAMING_SNAKE_CASE : Optional[int] = use_cache SCREAMING_SNAKE_CASE : Optional[Any] = scale_attn_by_inverse_layer_idx SCREAMING_SNAKE_CASE : str = reorder_and_upcast_attn SCREAMING_SNAKE_CASE : List[str] = tie_word_embeddings super().__init__(tie_word_embeddings=_lowerCamelCase , **_lowerCamelCase ) class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: return OrderedDict( [ ('''input_ids''', {0: '''batch''', 1: '''sequence'''}), ] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 1 , _lowerCamelCase = -1 , _lowerCamelCase = False , _lowerCamelCase = None , _lowerCamelCase = 3 , _lowerCamelCase = 32 , _lowerCamelCase = 32 , ) ->Mapping[str, Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = self._generate_dummy_images(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = dict(preprocessor(images=_lowerCamelCase , return_tensors=_lowerCamelCase ) ) return inputs
313
1
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 a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=3 , _lowerCamelCase=32 , _lowerCamelCase=3 , _lowerCamelCase=10 , _lowerCamelCase=[10, 20, 30, 40] , _lowerCamelCase=[1, 1, 2, 1] , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase="relu" , _lowerCamelCase=3 , _lowerCamelCase=None , ) ->Tuple: SCREAMING_SNAKE_CASE : Dict = parent SCREAMING_SNAKE_CASE : Optional[Any] = batch_size SCREAMING_SNAKE_CASE : Tuple = image_size SCREAMING_SNAKE_CASE : Optional[int] = num_channels SCREAMING_SNAKE_CASE : Dict = embeddings_size SCREAMING_SNAKE_CASE : Optional[Any] = hidden_sizes SCREAMING_SNAKE_CASE : str = depths SCREAMING_SNAKE_CASE : List[str] = is_training SCREAMING_SNAKE_CASE : str = use_labels SCREAMING_SNAKE_CASE : Any = hidden_act SCREAMING_SNAKE_CASE : Optional[Any] = num_labels SCREAMING_SNAKE_CASE : Any = scope SCREAMING_SNAKE_CASE : Union[str, Any] = len(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : str = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : str = None if self.use_labels: SCREAMING_SNAKE_CASE : int = ids_tensor([self.batch_size] , self.num_labels ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_config() return config, pixel_values, labels def __lowerCAmelCase ( self ) ->List[str]: 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 __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[str] = TFResNetModel(config=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = model(_lowerCamelCase ) # 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 __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[Any] = self.num_labels SCREAMING_SNAKE_CASE : Tuple = TFResNetForImageClassification(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Tuple = self.prepare_config_and_inputs() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = config_and_inputs SCREAMING_SNAKE_CASE : int = {'''pixel_values''': pixel_values} return config, inputs_dict @require_tf class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[int] = (TFResNetModel, TFResNetForImageClassification) if is_tf_available() else () __SCREAMING_SNAKE_CASE : List[Any] = ( {'feature-extraction': TFResNetModel, 'image-classification': TFResNetForImageClassification} if is_tf_available() else {} ) __SCREAMING_SNAKE_CASE : int = False __SCREAMING_SNAKE_CASE : str = False __SCREAMING_SNAKE_CASE : Tuple = False __SCREAMING_SNAKE_CASE : str = False __SCREAMING_SNAKE_CASE : Dict = False def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[Any] = TFResNetModelTester(self ) SCREAMING_SNAKE_CASE : Optional[Any] = ConfigTester(self , config_class=_lowerCamelCase , has_text_modality=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: 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 __lowerCAmelCase ( self ) ->str: return @unittest.skip(reason='''ResNet does not use inputs_embeds''' ) def __lowerCAmelCase ( self ) ->List[str]: pass @unittest.skip(reason='''ResNet does not support input and output embeddings''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : int = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = inspect.signature(model.call ) # signature.parameters is an OrderedDict => so arg_names order is deterministic SCREAMING_SNAKE_CASE : Dict = [*signature.parameters.keys()] SCREAMING_SNAKE_CASE : List[str] = ['''pixel_values'''] self.assertListEqual(arg_names[:1] , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Any = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: def check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : List[str] = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = outputs.encoder_hidden_states if config.is_encoder_decoder else outputs.hidden_states SCREAMING_SNAKE_CASE : Tuple = self.model_tester.num_stages self.assertEqual(len(_lowerCamelCase ) , 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] , ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : Optional[int] = ['''basic''', '''bottleneck'''] for model_class in self.all_model_classes: for layer_type in layers_type: SCREAMING_SNAKE_CASE : List[str] = layer_type SCREAMING_SNAKE_CASE : str = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # check that output_hidden_states also work using config del inputs_dict["output_hidden_states"] SCREAMING_SNAKE_CASE : Optional[int] = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_image_classification(*_lowerCamelCase ) @slow def __lowerCAmelCase ( self ) ->List[Any]: for model_name in TF_RESNET_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: SCREAMING_SNAKE_CASE : Optional[Any] = TFResNetModel.from_pretrained(_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) return image @require_tf @require_vision class a_ ( unittest.TestCase ): """simple docstring""" @cached_property def __lowerCAmelCase ( self ) ->Optional[int]: return ( AutoImageProcessor.from_pretrained(TF_RESNET_PRETRAINED_MODEL_ARCHIVE_LIST[0] ) if is_vision_available() else None ) @slow def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : int = TFResNetForImageClassification.from_pretrained(TF_RESNET_PRETRAINED_MODEL_ARCHIVE_LIST[0] ) SCREAMING_SNAKE_CASE : str = self.default_image_processor SCREAMING_SNAKE_CASE : Any = prepare_img() SCREAMING_SNAKE_CASE : Dict = image_processor(images=_lowerCamelCase , return_tensors='''tf''' ) # forward pass SCREAMING_SNAKE_CASE : Optional[int] = model(**_lowerCamelCase ) # verify the logits SCREAMING_SNAKE_CASE : int = tf.TensorShape((1, 1000) ) self.assertEqual(outputs.logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = tf.constant([-1_1.1_0_6_9, -9.7_8_7_7, -8.3_7_7_7] ) self.assertTrue(np.allclose(outputs.logits[0, :3].numpy() , _lowerCamelCase , atol=1e-4 ) )
313
from maths.prime_check import is_prime def UpperCAmelCase_( a__ ): """simple docstring""" if not isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = F"""Input value of [number={number}] must be an integer""" raise TypeError(a__ ) if is_prime(a__ ) and is_prime(number + 2 ): return number + 2 else: return -1 if __name__ == "__main__": import doctest doctest.testmod()
313
1
import argparse import json from pathlib import Path import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from timm import create_model from timm.data import resolve_data_config from timm.data.transforms_factory import create_transform from transformers import BitConfig, BitForImageClassification, BitImageProcessor from transformers.image_utils import PILImageResampling from transformers.utils import logging logging.set_verbosity_info() a__ : Optional[Any] = logging.get_logger(__name__) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = '''huggingface/label-files''' SCREAMING_SNAKE_CASE : Any = '''imagenet-1k-id2label.json''' SCREAMING_SNAKE_CASE : Tuple = json.load(open(hf_hub_download(a__ , a__ , repo_type='''dataset''' ) , '''r''' ) ) SCREAMING_SNAKE_CASE : int = {int(a__ ): v for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : Dict = {v: k for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : List[Any] = '''std_conv''' if '''bit''' in model_name else False # note that when using BiT as backbone for ViT-hybrid checkpoints, # one needs to additionally set config.layer_type = "bottleneck", config.stem_type = "same", # config.conv_layer = "std_conv_same" SCREAMING_SNAKE_CASE : Tuple = BitConfig( conv_layer=a__ , num_labels=1_000 , idalabel=a__ , labelaid=a__ , ) return config def UpperCAmelCase_( a__ ): """simple docstring""" if "stem.conv" in name: SCREAMING_SNAKE_CASE : int = name.replace('''stem.conv''' , '''bit.embedder.convolution''' ) if "blocks" in name: SCREAMING_SNAKE_CASE : List[str] = name.replace('''blocks''' , '''layers''' ) if "head.fc" in name: SCREAMING_SNAKE_CASE : List[Any] = name.replace('''head.fc''' , '''classifier.1''' ) if name.startswith('''norm''' ): SCREAMING_SNAKE_CASE : int = '''bit.''' + name if "bit" not in name and "classifier" not in name: SCREAMING_SNAKE_CASE : int = '''bit.encoder.''' + name return name def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = '''http://images.cocodataset.org/val2017/000000039769.jpg''' SCREAMING_SNAKE_CASE : Tuple = Image.open(requests.get(a__ , stream=a__ ).raw ) return im @torch.no_grad() def UpperCAmelCase_( a__ , a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = get_config(a__ ) # load original model from timm SCREAMING_SNAKE_CASE : List[Any] = create_model(a__ , pretrained=a__ ) timm_model.eval() # load state_dict of original model SCREAMING_SNAKE_CASE : Dict = timm_model.state_dict() for key in state_dict.copy().keys(): SCREAMING_SNAKE_CASE : Tuple = state_dict.pop(a__ ) SCREAMING_SNAKE_CASE : List[str] = val.squeeze() if '''head''' in key else val # load HuggingFace model SCREAMING_SNAKE_CASE : Optional[Any] = BitForImageClassification(a__ ) model.eval() model.load_state_dict(a__ ) # create image processor SCREAMING_SNAKE_CASE : Any = create_transform(**resolve_data_config({} , model=a__ ) ) SCREAMING_SNAKE_CASE : int = transform.transforms SCREAMING_SNAKE_CASE : Optional[int] = { '''bilinear''': PILImageResampling.BILINEAR, '''bicubic''': PILImageResampling.BICUBIC, '''nearest''': PILImageResampling.NEAREST, } SCREAMING_SNAKE_CASE : Tuple = BitImageProcessor( do_resize=a__ , size={'''shortest_edge''': timm_transforms[0].size} , resample=pillow_resamplings[timm_transforms[0].interpolation.value] , do_center_crop=a__ , crop_size={'''height''': timm_transforms[1].size[0], '''width''': timm_transforms[1].size[1]} , do_normalize=a__ , image_mean=timm_transforms[-1].mean.tolist() , image_std=timm_transforms[-1].std.tolist() , ) SCREAMING_SNAKE_CASE : Dict = prepare_img() SCREAMING_SNAKE_CASE : Optional[int] = transform(a__ ).unsqueeze(0 ) SCREAMING_SNAKE_CASE : Optional[int] = processor(a__ , return_tensors='''pt''' ).pixel_values # verify pixel values assert torch.allclose(a__ , a__ ) # verify logits with torch.no_grad(): SCREAMING_SNAKE_CASE : Any = model(a__ ) SCREAMING_SNAKE_CASE : int = outputs.logits print('''Logits:''' , logits[0, :3] ) print('''Predicted class:''' , model.config.idalabel[logits.argmax(-1 ).item()] ) SCREAMING_SNAKE_CASE : List[Any] = timm_model(a__ ) assert timm_logits.shape == outputs.logits.shape assert torch.allclose(a__ , outputs.logits , atol=1e-3 ) print('''Looks ok!''' ) if pytorch_dump_folder_path is not None: Path(a__ ).mkdir(exist_ok=a__ ) print(F"""Saving model {model_name} and processor to {pytorch_dump_folder_path}""" ) model.save_pretrained(a__ ) processor.save_pretrained(a__ ) if push_to_hub: print(F"""Pushing model {model_name} and processor to the hub""" ) model.push_to_hub(F"""ybelkada/{model_name}""" ) processor.push_to_hub(F"""ybelkada/{model_name}""" ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--model_name''', default='''resnetv2_50x1_bitm''', type=str, help='''Name of the BiT timm model you\'d like to convert.''', ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.''' ) parser.add_argument( '''--push_to_hub''', action='''store_true''', help='''Whether to push the model to the hub.''', ) a__ : Any = parser.parse_args() convert_bit_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
313
import gc import random import unittest import numpy as np import torch from PIL import Image from diffusers import ( DDIMScheduler, KandinskyVaaControlnetImgaImgPipeline, KandinskyVaaPriorEmbaEmbPipeline, UNetaDConditionModel, VQModel, ) from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[Any] = KandinskyVaaControlnetImgaImgPipeline __SCREAMING_SNAKE_CASE : Optional[int] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[Any] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[str] = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->Optional[Any]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 32 @property def __lowerCAmelCase ( self ) ->str: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Dict: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Tuple: return 100 @property def __lowerCAmelCase ( self ) ->int: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Union[str, Any] = { '''in_channels''': 8, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''image_hint''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : List[str] = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->Any: return { "block_out_channels": [32, 32, 64, 64], "down_block_types": [ "DownEncoderBlock2D", "DownEncoderBlock2D", "DownEncoderBlock2D", "AttnDownEncoderBlock2D", ], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": ["AttnUpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D"], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : str = self.dummy_unet SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_movq SCREAMING_SNAKE_CASE : List[str] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : str = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = { '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->int: SCREAMING_SNAKE_CASE : str = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(seed + 1 ) ).to( _lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : Dict = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) # create hint SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : List[str] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = { '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''hint''': hint, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[int] = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : Any = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : Union[str, Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array( [0.5_4_9_8_5_0_3_4, 0.5_5_5_0_9_3_6_5, 0.5_2_5_6_1_5_0_4, 0.5_5_7_0_4_9_4, 0.5_5_9_3_8_1_8, 0.5_2_6_3_9_7_9, 0.5_0_2_8_5_6_4_3, 0.5_0_6_9_8_4_6, 0.5_1_1_9_6_7_3_6] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Union[str, Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/kandinskyv22_controlnet_img2img_robotcat_fp16.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = init_image.resize((512, 512) ) SCREAMING_SNAKE_CASE : Dict = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/hint_image_cat.png''' ) SCREAMING_SNAKE_CASE : List[Any] = torch.from_numpy(np.array(_lowerCamelCase ) ).float() / 2_5_5.0 SCREAMING_SNAKE_CASE : int = hint.permute(2 , 0 , 1 ).unsqueeze(0 ) SCREAMING_SNAKE_CASE : List[Any] = '''A robot, 4k photo''' SCREAMING_SNAKE_CASE : List[str] = KandinskyVaaPriorEmbaEmbPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = KandinskyVaaControlnetImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-controlnet-depth''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Any = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , image=_lowerCamelCase , strength=0.8_5 , generator=_lowerCamelCase , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : List[str] = pipeline( image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , hint=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=512 , width=512 , strength=0.5 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = output.images[0] assert image.shape == (512, 512, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
import copy import inspect import unittest from transformers import AutoBackbone from transformers.configuration_utils import PretrainedConfig from transformers.testing_utils import require_timm, require_torch, torch_device from transformers.utils.import_utils import is_torch_available from ...test_backbone_common import BackboneTesterMixin from ...test_configuration_common import ConfigTester from ...test_modeling_common import ModelTesterMixin, floats_tensor if is_torch_available(): import torch from transformers import TimmBackbone, TimmBackboneConfig from ...test_pipeline_mixin import PipelineTesterMixin class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase="resnet50" , _lowerCamelCase=3 , _lowerCamelCase=32 , _lowerCamelCase=3 , _lowerCamelCase=True , _lowerCamelCase=True , ) ->Optional[int]: SCREAMING_SNAKE_CASE : List[Any] = parent SCREAMING_SNAKE_CASE : Tuple = out_indices if out_indices is not None else [4] SCREAMING_SNAKE_CASE : Dict = stage_names SCREAMING_SNAKE_CASE : Optional[Any] = out_features SCREAMING_SNAKE_CASE : Optional[Any] = backbone SCREAMING_SNAKE_CASE : Union[str, Any] = batch_size SCREAMING_SNAKE_CASE : Any = image_size SCREAMING_SNAKE_CASE : Any = num_channels SCREAMING_SNAKE_CASE : int = use_pretrained_backbone SCREAMING_SNAKE_CASE : Any = is_training def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[str] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : Optional[Any] = self.get_config() return config, pixel_values def __lowerCAmelCase ( self ) ->Union[str, Any]: return TimmBackboneConfig( image_size=self.image_size , num_channels=self.num_channels , out_features=self.out_features , out_indices=self.out_indices , stage_names=self.stage_names , use_pretrained_backbone=self.use_pretrained_backbone , backbone=self.backbone , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : Optional[int] = TimmBackbone(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : List[str] = model(_lowerCamelCase ) self.parent.assertEqual( result.feature_map[-1].shape , (self.batch_size, model.channels[-1], 14, 14) , ) def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[int] = self.prepare_config_and_inputs() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = config_and_inputs SCREAMING_SNAKE_CASE : List[Any] = {'''pixel_values''': pixel_values} return config, inputs_dict @require_torch @require_timm class a_ ( a__ , a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = (TimmBackbone,) if is_torch_available() else () __SCREAMING_SNAKE_CASE : str = {'feature-extraction': TimmBackbone} if is_torch_available() else {} __SCREAMING_SNAKE_CASE : Any = False __SCREAMING_SNAKE_CASE : Tuple = False __SCREAMING_SNAKE_CASE : Any = False __SCREAMING_SNAKE_CASE : Optional[Any] = False def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Tuple = TimmBackboneModelTester(self ) SCREAMING_SNAKE_CASE : Optional[int] = ConfigTester(self , config_class=_lowerCamelCase , has_text_modality=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Tuple: 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 __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Any = '''resnet18''' SCREAMING_SNAKE_CASE : Dict = '''microsoft/resnet-18''' SCREAMING_SNAKE_CASE : int = AutoBackbone.from_pretrained(_lowerCamelCase , use_timm_backbone=_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = AutoBackbone.from_pretrained(_lowerCamelCase ) self.assertEqual(len(timm_model.out_features ) , len(transformers_model.out_features ) ) self.assertEqual(len(timm_model.stage_names ) , len(transformers_model.stage_names ) ) self.assertEqual(timm_model.channels , transformers_model.channels ) # Out indices are set to the last layer by default. For timm models, we don't know # the number of layers in advance, so we set it to (-1,), whereas for transformers # models, we set it to [len(stage_names) - 1] (kept for backward compatibility). self.assertEqual(timm_model.out_indices , (-1,) ) self.assertEqual(transformers_model.out_indices , [len(timm_model.stage_names ) - 1] ) SCREAMING_SNAKE_CASE : Optional[Any] = AutoBackbone.from_pretrained(_lowerCamelCase , use_timm_backbone=_lowerCamelCase , out_indices=[1, 2, 3] ) SCREAMING_SNAKE_CASE : Dict = AutoBackbone.from_pretrained(_lowerCamelCase , out_indices=[1, 2, 3] ) self.assertEqual(timm_model.out_indices , transformers_model.out_indices ) self.assertEqual(len(timm_model.out_features ) , len(transformers_model.out_features ) ) self.assertEqual(timm_model.channels , transformers_model.channels ) @unittest.skip('''TimmBackbone doesn\'t support feed forward chunking''' ) def __lowerCAmelCase ( self ) ->Any: pass @unittest.skip('''TimmBackbone doesn\'t have num_hidden_layers attribute''' ) def __lowerCAmelCase ( self ) ->Any: pass @unittest.skip('''TimmBackbone initialization is managed on the timm side''' ) def __lowerCAmelCase ( self ) ->int: pass @unittest.skip('''TimmBackbone models doesn\'t have inputs_embeds''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass @unittest.skip('''TimmBackbone models doesn\'t have inputs_embeds''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @unittest.skip('''TimmBackbone model cannot be created without specifying a backbone checkpoint''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @unittest.skip('''Only checkpoints on timm can be loaded into TimmBackbone''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @unittest.skip('''model weights aren\'t tied in TimmBackbone.''' ) def __lowerCAmelCase ( self ) ->Tuple: pass @unittest.skip('''model weights aren\'t tied in TimmBackbone.''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass @unittest.skip('''Only checkpoints on timm can be loaded into TimmBackbone''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @unittest.skip('''Only checkpoints on timm can be loaded into TimmBackbone''' ) def __lowerCAmelCase ( self ) ->Tuple: pass @unittest.skip('''TimmBackbone doesn\'t have hidden size info in its configuration.''' ) def __lowerCAmelCase ( self ) ->str: pass @unittest.skip('''TimmBackbone doesn\'t support output_attentions.''' ) def __lowerCAmelCase ( self ) ->int: pass @unittest.skip('''Safetensors is not supported by timm.''' ) def __lowerCAmelCase ( self ) ->int: pass @unittest.skip('''Will be fixed soon by reducing the size of the model used for common tests.''' ) def __lowerCAmelCase ( self ) ->Tuple: pass def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Union[str, Any] = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = inspect.signature(model.forward ) # signature.parameters is an OrderedDict => so arg_names order is deterministic SCREAMING_SNAKE_CASE : str = [*signature.parameters.keys()] SCREAMING_SNAKE_CASE : int = ['''pixel_values'''] self.assertListEqual(arg_names[:1] , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : List[Any] = True SCREAMING_SNAKE_CASE : Dict = self.has_attentions # no need to test all models as different heads yield the same functionality SCREAMING_SNAKE_CASE : Optional[int] = self.all_model_classes[0] SCREAMING_SNAKE_CASE : Optional[Any] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = outputs[0][-1] # Encoder-/Decoder-only models SCREAMING_SNAKE_CASE : Tuple = outputs.hidden_states[0] hidden_states.retain_grad() if self.has_attentions: SCREAMING_SNAKE_CASE : List[str] = outputs.attentions[0] attentions.retain_grad() output.flatten()[0].backward(retain_graph=_lowerCamelCase ) self.assertIsNotNone(hidden_states.grad ) if self.has_attentions: self.assertIsNotNone(attentions.grad ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Dict = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Optional[int] = model(**_lowerCamelCase ) self.assertEqual(len(result.feature_maps ) , len(config.out_indices ) ) self.assertEqual(len(model.channels ) , len(config.out_indices ) ) # Check output of last stage is taken if out_features=None, out_indices=None SCREAMING_SNAKE_CASE : Optional[Any] = copy.deepcopy(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = None SCREAMING_SNAKE_CASE : int = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Optional[Any] = model(**_lowerCamelCase ) self.assertEqual(len(result.feature_maps ) , 1 ) self.assertEqual(len(model.channels ) , 1 ) # Check backbone can be initialized with fresh weights SCREAMING_SNAKE_CASE : Dict = copy.deepcopy(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = False SCREAMING_SNAKE_CASE : List[Any] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Union[str, Any] = model(**_lowerCamelCase )
313
from typing import Any, Callable, Dict, List, Optional, Union import torch from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer from diffusers import ( AutoencoderKL, DDIMScheduler, DiffusionPipeline, LMSDiscreteScheduler, PNDMScheduler, StableDiffusionPipeline, UNetaDConditionModel, ) from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker a__ : List[str] = '''CompVis/stable-diffusion-v1-1''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-2''' a__ : Any = '''CompVis/stable-diffusion-v1-3''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-4''' class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , ) ->str: super()._init_() SCREAMING_SNAKE_CASE : Tuple = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = StableDiffusionPipeline( vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , unet=_lowerCamelCase , scheduler=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=_lowerCamelCase , requires_safety_checker=_lowerCamelCase , ) self.register_modules(pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea ) @property def __lowerCAmelCase ( self ) ->Dict[str, Any]: return {k: getattr(self , _lowerCamelCase ) for k in self.config.keys() if not k.startswith('''_''' )} def __lowerCAmelCase ( self , _lowerCamelCase = "auto" ) ->Optional[int]: if slice_size == "auto": # half the attention head size is usually a good trade-off between # speed and memory SCREAMING_SNAKE_CASE : Optional[int] = self.unet.config.attention_head_dim // 2 self.unet.set_attention_slice(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: self.enable_attention_slicing(_lowerCamelCase ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->str: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Tuple: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Optional[Any]: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: SCREAMING_SNAKE_CASE : Optional[Any] = '''cuda''' if torch.cuda.is_available() else '''cpu''' self.to(_lowerCamelCase ) # Checks if the height and width are divisible by 8 or not if height % 8 != 0 or width % 8 != 0: raise ValueError(F"""`height` and `width` must be divisible by 8 but are {height} and {width}.""" ) # Get first result from Stable Diffusion Checkpoint v1.1 SCREAMING_SNAKE_CASE : Optional[Any] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.2 SCREAMING_SNAKE_CASE : Any = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.3 SCREAMING_SNAKE_CASE : Optional[int] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.4 SCREAMING_SNAKE_CASE : str = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get all result images into a single list and pass it via StableDiffusionPipelineOutput for final result return StableDiffusionPipelineOutput([resa[0], resa[0], resa[0], resa[0]] )
313
1
import math import os import unittest from transformers import MegatronBertConfig, is_torch_available from transformers.models.auto import get_values from transformers.testing_utils import require_sentencepiece, require_tokenizers, require_torch, slow, torch_device from ...test_configuration_common import ConfigTester from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask from ...test_pipeline_mixin import PipelineTesterMixin if is_torch_available(): import torch from transformers import ( MODEL_FOR_PRETRAINING_MAPPING, MegatronBertForCausalLM, MegatronBertForMaskedLM, MegatronBertForMultipleChoice, MegatronBertForNextSentencePrediction, MegatronBertForPreTraining, MegatronBertForQuestionAnswering, MegatronBertForSequenceClassification, MegatronBertForTokenClassification, MegatronBertModel, ) class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=13 , _lowerCamelCase=7 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=99 , _lowerCamelCase=64 , _lowerCamelCase=32 , _lowerCamelCase=5 , _lowerCamelCase=4 , _lowerCamelCase=37 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=16 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=3 , _lowerCamelCase=4 , _lowerCamelCase=None , ) ->str: SCREAMING_SNAKE_CASE : Optional[int] = parent SCREAMING_SNAKE_CASE : Dict = batch_size SCREAMING_SNAKE_CASE : int = seq_length SCREAMING_SNAKE_CASE : Dict = is_training SCREAMING_SNAKE_CASE : Union[str, Any] = use_input_mask SCREAMING_SNAKE_CASE : Union[str, Any] = use_token_type_ids SCREAMING_SNAKE_CASE : Optional[int] = use_labels SCREAMING_SNAKE_CASE : int = vocab_size SCREAMING_SNAKE_CASE : Dict = hidden_size SCREAMING_SNAKE_CASE : Any = embedding_size SCREAMING_SNAKE_CASE : List[str] = num_hidden_layers SCREAMING_SNAKE_CASE : Optional[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Tuple = intermediate_size SCREAMING_SNAKE_CASE : List[str] = hidden_act SCREAMING_SNAKE_CASE : Union[str, Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Union[str, Any] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Any = type_vocab_size SCREAMING_SNAKE_CASE : Any = type_sequence_label_size SCREAMING_SNAKE_CASE : int = initializer_range SCREAMING_SNAKE_CASE : Tuple = num_labels SCREAMING_SNAKE_CASE : Dict = num_choices SCREAMING_SNAKE_CASE : Union[str, Any] = scope def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[int] = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size ) SCREAMING_SNAKE_CASE : Optional[Any] = None if self.use_input_mask: SCREAMING_SNAKE_CASE : Tuple = random_attention_mask([self.batch_size, self.seq_length] ) SCREAMING_SNAKE_CASE : Any = None if self.use_token_type_ids: SCREAMING_SNAKE_CASE : int = ids_tensor([self.batch_size, self.seq_length] , self.type_vocab_size ) SCREAMING_SNAKE_CASE : Dict = None SCREAMING_SNAKE_CASE : int = None SCREAMING_SNAKE_CASE : Dict = None if self.use_labels: SCREAMING_SNAKE_CASE : List[Any] = ids_tensor([self.batch_size] , self.type_sequence_label_size ) SCREAMING_SNAKE_CASE : Dict = ids_tensor([self.batch_size, self.seq_length] , self.num_labels ) SCREAMING_SNAKE_CASE : List[str] = ids_tensor([self.batch_size] , self.num_choices ) SCREAMING_SNAKE_CASE : int = self.get_config() return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels def __lowerCAmelCase ( self ) ->Optional[int]: return MegatronBertConfig( 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 , embedding_size=self.embedding_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=_lowerCamelCase , initializer_range=self.initializer_range , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Any: SCREAMING_SNAKE_CASE : Optional[Any] = MegatronBertModel(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Tuple = model(_lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = model(_lowerCamelCase , token_type_ids=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = model(_lowerCamelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) ) self.parent.assertEqual(result.pooler_output.shape , (self.batch_size, self.hidden_size) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : str = MegatronBertForMaskedLM(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Optional[int] = model(_lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->List[str]: SCREAMING_SNAKE_CASE : List[Any] = MegatronBertForCausalLM(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : int = model(_lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : Tuple = MegatronBertForNextSentencePrediction(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Optional[int] = model( _lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , labels=_lowerCamelCase , ) self.parent.assertEqual(result.logits.shape , (self.batch_size, 2) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : Any = MegatronBertForPreTraining(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : List[str] = model( _lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , labels=_lowerCamelCase , next_sentence_label=_lowerCamelCase , ) self.parent.assertEqual(result.prediction_logits.shape , (self.batch_size, self.seq_length, self.vocab_size) ) self.parent.assertEqual(result.seq_relationship_logits.shape , (self.batch_size, 2) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : Optional[Any] = MegatronBertForQuestionAnswering(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Any = model( _lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , start_positions=_lowerCamelCase , end_positions=_lowerCamelCase , ) self.parent.assertEqual(result.start_logits.shape , (self.batch_size, self.seq_length) ) self.parent.assertEqual(result.end_logits.shape , (self.batch_size, self.seq_length) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->List[str]: SCREAMING_SNAKE_CASE : List[Any] = self.num_labels SCREAMING_SNAKE_CASE : str = MegatronBertForSequenceClassification(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : List[str] = model(_lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = self.num_labels SCREAMING_SNAKE_CASE : Any = MegatronBertForTokenClassification(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Any: SCREAMING_SNAKE_CASE : List[str] = self.num_choices SCREAMING_SNAKE_CASE : str = MegatronBertForMultipleChoice(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : int = input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous() SCREAMING_SNAKE_CASE : List[Any] = token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous() SCREAMING_SNAKE_CASE : List[Any] = input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous() SCREAMING_SNAKE_CASE : Tuple = model( _lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , labels=_lowerCamelCase , ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) ) def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Optional[int] = self.prepare_config_and_inputs() ( ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ) : Union[str, Any] = config_and_inputs SCREAMING_SNAKE_CASE : List[Any] = {'''input_ids''': input_ids, '''token_type_ids''': token_type_ids, '''attention_mask''': input_mask} return config, inputs_dict @require_torch class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = ( ( MegatronBertModel, MegatronBertForMaskedLM, MegatronBertForCausalLM, MegatronBertForMultipleChoice, MegatronBertForNextSentencePrediction, MegatronBertForPreTraining, MegatronBertForQuestionAnswering, MegatronBertForSequenceClassification, MegatronBertForTokenClassification, ) if is_torch_available() else () ) __SCREAMING_SNAKE_CASE : List[Any] = ( { 'feature-extraction': MegatronBertModel, 'fill-mask': MegatronBertForMaskedLM, 'question-answering': MegatronBertForQuestionAnswering, 'text-classification': MegatronBertForSequenceClassification, 'text-generation': MegatronBertForCausalLM, 'token-classification': MegatronBertForTokenClassification, 'zero-shot': MegatronBertForSequenceClassification, } if is_torch_available() else {} ) __SCREAMING_SNAKE_CASE : Any = True # test_resize_embeddings = False __SCREAMING_SNAKE_CASE : Dict = False def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=False ) ->Tuple: SCREAMING_SNAKE_CASE : Dict = super()._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) if return_labels: if model_class in get_values(_lowerCamelCase ): SCREAMING_SNAKE_CASE : Optional[Any] = torch.zeros( (self.model_tester.batch_size, self.model_tester.seq_length) , dtype=torch.long , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = torch.zeros( self.model_tester.batch_size , dtype=torch.long , device=_lowerCamelCase ) return inputs_dict def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : int = MegatronBertModelTester(self ) SCREAMING_SNAKE_CASE : Optional[Any] = ConfigTester(self , config_class=_lowerCamelCase , hidden_size=37 ) def __lowerCAmelCase ( self ) ->Tuple: self.config_tester.run_common_tests() def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : List[str] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_model(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_for_masked_lm(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Tuple = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_for_multiple_choice(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : List[Any] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_for_next_sequence_prediction(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Tuple = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_for_pretraining(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_for_question_answering(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Tuple = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_for_sequence_classification(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : str = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_megatron_bert_for_token_classification(*_lowerCamelCase ) def UpperCAmelCase_( a__ ): """simple docstring""" return torch.tensor( a__ , dtype=torch.long , device=a__ , ) a__ : Any = 1E-4 @require_torch @require_sentencepiece @require_tokenizers class a_ ( unittest.TestCase ): """simple docstring""" @slow @unittest.skip('''Model is not available.''' ) def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Union[str, Any] = '''nvidia/megatron-bert-uncased-345m''' if "MYDIR" in os.environ: SCREAMING_SNAKE_CASE : Union[str, Any] = os.path.join(os.environ['''MYDIR'''] , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = MegatronBertModel.from_pretrained(_lowerCamelCase ) model.to(_lowerCamelCase ) model.half() SCREAMING_SNAKE_CASE : Optional[int] = _long_tensor([[101, 7110, 1005, 1056, 2023, 1_1333, 1_7413, 1029, 102]] ) with torch.no_grad(): SCREAMING_SNAKE_CASE : str = model(_lowerCamelCase )[0] SCREAMING_SNAKE_CASE : List[str] = torch.Size((1, 9, 1024) ) self.assertEqual(output.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = [-0.6_0_4_0, -0.2_5_1_7, -0.1_0_2_5, 0.3_4_2_0, -0.6_7_5_8, -0.0_0_1_7, -0.1_0_8_9, -0.1_9_9_0, 0.5_7_2_8] for ii in range(3 ): for jj in range(3 ): SCREAMING_SNAKE_CASE : int = output[0, ii, jj] SCREAMING_SNAKE_CASE : Tuple = expected[3 * ii + jj] SCREAMING_SNAKE_CASE : Any = '''ii={} jj={} a={} b={}'''.format(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) self.assertTrue(math.isclose(_lowerCamelCase , _lowerCamelCase , rel_tol=_lowerCamelCase , abs_tol=_lowerCamelCase ) , msg=_lowerCamelCase )
313
from typing import Optional, Tuple, Union import flax import flax.linen as nn import jax import jax.numpy as jnp from flax.core.frozen_dict import FrozenDict from ..configuration_utils import ConfigMixin, flax_register_to_config from ..utils import BaseOutput from .embeddings_flax import FlaxTimestepEmbedding, FlaxTimesteps from .modeling_flax_utils import FlaxModelMixin from .unet_ad_blocks_flax import ( FlaxCrossAttnDownBlockaD, FlaxCrossAttnUpBlockaD, FlaxDownBlockaD, FlaxUNetMidBlockaDCrossAttn, FlaxUpBlockaD, ) @flax.struct.dataclass class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : jnp.ndarray @flax_register_to_config class a_ ( nn.Module , a__ , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = 32 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : Tuple[str] = ( "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D", ) __SCREAMING_SNAKE_CASE : Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D") __SCREAMING_SNAKE_CASE : Union[bool, Tuple[bool]] = False __SCREAMING_SNAKE_CASE : Tuple[int] = (320, 640, 1280, 1280) __SCREAMING_SNAKE_CASE : int = 2 __SCREAMING_SNAKE_CASE : Union[int, Tuple[int]] = 8 __SCREAMING_SNAKE_CASE : Optional[Union[int, Tuple[int]]] = None __SCREAMING_SNAKE_CASE : int = 1280 __SCREAMING_SNAKE_CASE : float = 0.0 __SCREAMING_SNAKE_CASE : bool = False __SCREAMING_SNAKE_CASE : jnp.dtype = jnp.floataa __SCREAMING_SNAKE_CASE : bool = True __SCREAMING_SNAKE_CASE : int = 0 __SCREAMING_SNAKE_CASE : bool = False def __lowerCAmelCase ( self , _lowerCamelCase ) ->FrozenDict: # init input tensors SCREAMING_SNAKE_CASE : List[Any] = (1, self.in_channels, self.sample_size, self.sample_size) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros(_lowerCamelCase , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Tuple = jnp.ones((1,) , dtype=jnp.intaa ) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros((1, 1, self.cross_attention_dim) , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = jax.random.split(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = {'''params''': params_rng, '''dropout''': dropout_rng} return self.init(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase )["params"] def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : List[str] = self.block_out_channels SCREAMING_SNAKE_CASE : Optional[int] = block_out_channels[0] * 4 if self.num_attention_heads is not None: raise ValueError( '''At the moment it is not possible to define the number of attention heads via `num_attention_heads` because of a naming issue as described in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131. Passing `num_attention_heads` will only be supported in diffusers v0.19.''' ) # If `num_attention_heads` is not defined (which is the case for most models) # it will default to `attention_head_dim`. This looks weird upon first reading it and it is. # The reason for this behavior is to correct for incorrectly named variables that were introduced # when this library was created. The incorrect naming was only discovered much later in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131 # Changing `attention_head_dim` to `num_attention_heads` for 40,000+ configurations is too backwards breaking # which is why we correct for the naming here. SCREAMING_SNAKE_CASE : List[str] = self.num_attention_heads or self.attention_head_dim # input SCREAMING_SNAKE_CASE : Optional[int] = nn.Conv( block_out_channels[0] , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) # time SCREAMING_SNAKE_CASE : Tuple = FlaxTimesteps( block_out_channels[0] , flip_sin_to_cos=self.flip_sin_to_cos , freq_shift=self.config.freq_shift ) SCREAMING_SNAKE_CASE : Dict = FlaxTimestepEmbedding(_lowerCamelCase , dtype=self.dtype ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.only_cross_attention if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (only_cross_attention,) * len(self.down_block_types ) if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (num_attention_heads,) * len(self.down_block_types ) # down SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Optional[Any] = block_out_channels[0] for i, down_block_type in enumerate(self.down_block_types ): SCREAMING_SNAKE_CASE : str = output_channel SCREAMING_SNAKE_CASE : int = block_out_channels[i] SCREAMING_SNAKE_CASE : List[Any] = i == len(_lowerCamelCase ) - 1 if down_block_type == "CrossAttnDownBlock2D": SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxCrossAttnDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , num_attention_heads=num_attention_heads[i] , add_downsample=not is_final_block , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , add_downsample=not is_final_block , dtype=self.dtype , ) down_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = down_blocks # mid SCREAMING_SNAKE_CASE : int = FlaxUNetMidBlockaDCrossAttn( in_channels=block_out_channels[-1] , dropout=self.dropout , num_attention_heads=num_attention_heads[-1] , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) # up SCREAMING_SNAKE_CASE : str = [] SCREAMING_SNAKE_CASE : str = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = reversed_block_out_channels[0] for i, up_block_type in enumerate(self.up_block_types ): SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Union[str, Any] = reversed_block_out_channels[i] SCREAMING_SNAKE_CASE : Tuple = reversed_block_out_channels[min(i + 1 , len(_lowerCamelCase ) - 1 )] SCREAMING_SNAKE_CASE : Dict = i == len(_lowerCamelCase ) - 1 if up_block_type == "CrossAttnUpBlock2D": SCREAMING_SNAKE_CASE : str = FlaxCrossAttnUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , num_attention_heads=reversed_num_attention_heads[i] , add_upsample=not is_final_block , dropout=self.dropout , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Optional[int] = FlaxUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , add_upsample=not is_final_block , dropout=self.dropout , dtype=self.dtype , ) up_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Tuple = up_blocks # out SCREAMING_SNAKE_CASE : Any = nn.GroupNorm(num_groups=32 , epsilon=1e-5 ) SCREAMING_SNAKE_CASE : Any = nn.Conv( self.out_channels , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) def __call__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase = True , _lowerCamelCase = False , ) ->Union[FlaxUNetaDConditionOutput, Tuple]: # 1. time if not isinstance(_lowerCamelCase , jnp.ndarray ): SCREAMING_SNAKE_CASE : int = jnp.array([timesteps] , dtype=jnp.intaa ) elif isinstance(_lowerCamelCase , jnp.ndarray ) and len(timesteps.shape ) == 0: SCREAMING_SNAKE_CASE : List[str] = timesteps.astype(dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.expand_dims(_lowerCamelCase , 0 ) SCREAMING_SNAKE_CASE : List[str] = self.time_proj(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = self.time_embedding(_lowerCamelCase ) # 2. pre-process SCREAMING_SNAKE_CASE : int = jnp.transpose(_lowerCamelCase , (0, 2, 3, 1) ) SCREAMING_SNAKE_CASE : List[Any] = self.conv_in(_lowerCamelCase ) # 3. down SCREAMING_SNAKE_CASE : Optional[int] = (sample,) for down_block in self.down_blocks: if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , deterministic=not train ) down_block_res_samples += res_samples if down_block_additional_residuals is not None: SCREAMING_SNAKE_CASE : int = () for down_block_res_sample, down_block_additional_residual in zip( _lowerCamelCase , _lowerCamelCase ): down_block_res_sample += down_block_additional_residual new_down_block_res_samples += (down_block_res_sample,) SCREAMING_SNAKE_CASE : Dict = new_down_block_res_samples # 4. mid SCREAMING_SNAKE_CASE : Optional[Any] = self.mid_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) if mid_block_additional_residual is not None: sample += mid_block_additional_residual # 5. up for up_block in self.up_blocks: SCREAMING_SNAKE_CASE : Optional[Any] = down_block_res_samples[-(self.layers_per_block + 1) :] SCREAMING_SNAKE_CASE : Optional[int] = down_block_res_samples[: -(self.layers_per_block + 1)] if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = up_block( _lowerCamelCase , temb=_lowerCamelCase , encoder_hidden_states=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train , ) else: SCREAMING_SNAKE_CASE : Optional[int] = up_block(_lowerCamelCase , temb=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train ) # 6. post-process SCREAMING_SNAKE_CASE : Optional[int] = self.conv_norm_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.silu(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self.conv_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.transpose(_lowerCamelCase , (0, 3, 1, 2) ) if not return_dict: return (sample,) return FlaxUNetaDConditionOutput(sample=_lowerCamelCase )
313
1
import numpy as np from matplotlib import pyplot as plt from sklearn.datasets import load_iris from sklearn.metrics import ConfusionMatrixDisplay from sklearn.model_selection import train_test_split from xgboost import XGBClassifier def UpperCAmelCase_( a__ ): """simple docstring""" return (data["data"], data["target"]) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = XGBClassifier() classifier.fit(a__ , a__ ) return classifier def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = load_iris() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = data_handling(a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = train_test_split( a__ , a__ , test_size=0.25 ) SCREAMING_SNAKE_CASE : int = iris['''target_names'''] # Create an XGBoost Classifier from the training data SCREAMING_SNAKE_CASE : Dict = xgboost(a__ , a__ ) # Display the confusion matrix of the classifier with both training and test sets ConfusionMatrixDisplay.from_estimator( a__ , a__ , a__ , display_labels=a__ , cmap='''Blues''' , normalize='''true''' , ) plt.title('''Normalized Confusion Matrix - IRIS Dataset''' ) plt.show() if __name__ == "__main__": import doctest doctest.testmod(verbose=True) main()
313
import gc import random import unittest import numpy as np import torch from transformers import ( CLIPImageProcessor, CLIPTextConfig, CLIPTextModel, CLIPTokenizer, CLIPVisionConfig, CLIPVisionModelWithProjection, ) from diffusers import AutoencoderKL, DDIMScheduler, DDPMScheduler, StableUnCLIPImgaImgPipeline, UNetaDConditionModel from diffusers.pipelines.pipeline_utils import DiffusionPipeline from diffusers.pipelines.stable_diffusion.stable_unclip_image_normalizer import StableUnCLIPImageNormalizer from diffusers.utils.import_utils import is_xformers_available from diffusers.utils.testing_utils import ( enable_full_determinism, floats_tensor, load_image, load_numpy, require_torch_gpu, skip_mps, slow, torch_device, ) from ..pipeline_params import TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS from ..test_pipelines_common import ( PipelineKarrasSchedulerTesterMixin, PipelineLatentTesterMixin, PipelineTesterMixin, assert_mean_pixel_difference, ) enable_full_determinism() class a_ ( a__ , a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = StableUnCLIPImgaImgPipeline __SCREAMING_SNAKE_CASE : List[str] = TEXT_GUIDED_IMAGE_VARIATION_PARAMS __SCREAMING_SNAKE_CASE : List[Any] = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS __SCREAMING_SNAKE_CASE : Any = frozenset( [] ) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess __SCREAMING_SNAKE_CASE : Tuple = frozenset([] ) def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[Any] = 32 SCREAMING_SNAKE_CASE : Tuple = embedder_hidden_size # image encoding components SCREAMING_SNAKE_CASE : int = CLIPImageProcessor(crop_size=32 , size=32 ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = CLIPVisionModelWithProjection( CLIPVisionConfig( hidden_size=_lowerCamelCase , projection_dim=_lowerCamelCase , num_hidden_layers=5 , num_attention_heads=4 , image_size=32 , intermediate_size=37 , patch_size=1 , ) ) # regular denoising components torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = StableUnCLIPImageNormalizer(embedding_dim=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = DDPMScheduler(beta_schedule='''squaredcos_cap_v2''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = CLIPTextModel( CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=_lowerCamelCase , projection_dim=32 , intermediate_size=37 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , ) ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = UNetaDConditionModel( sample_size=32 , in_channels=4 , out_channels=4 , down_block_types=('''CrossAttnDownBlock2D''', '''DownBlock2D''') , up_block_types=('''UpBlock2D''', '''CrossAttnUpBlock2D''') , block_out_channels=(32, 64) , attention_head_dim=(2, 4) , class_embed_type='''projection''' , projection_class_embeddings_input_dim=embedder_projection_dim * 2 , cross_attention_dim=_lowerCamelCase , layers_per_block=1 , upcast_attention=_lowerCamelCase , use_linear_projection=_lowerCamelCase , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = DDIMScheduler( beta_schedule='''scaled_linear''' , beta_start=0.0_0_0_8_5 , beta_end=0.0_1_2 , prediction_type='''v_prediction''' , set_alpha_to_one=_lowerCamelCase , steps_offset=1 , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Dict = AutoencoderKL() SCREAMING_SNAKE_CASE : Optional[Any] = { # image encoding components '''feature_extractor''': feature_extractor, '''image_encoder''': image_encoder.eval(), # image noising components '''image_normalizer''': image_normalizer.eval(), '''image_noising_scheduler''': image_noising_scheduler, # regular denoising components '''tokenizer''': tokenizer, '''text_encoder''': text_encoder.eval(), '''unet''': unet.eval(), '''scheduler''': scheduler, '''vae''': vae.eval(), } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 , _lowerCamelCase=True ) ->Optional[int]: if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, 3, 32, 32) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if pil_image: SCREAMING_SNAKE_CASE : Any = input_image * 0.5 + 0.5 SCREAMING_SNAKE_CASE : int = input_image.clamp(0 , 1 ) SCREAMING_SNAKE_CASE : Union[str, Any] = input_image.cpu().permute(0 , 2 , 3 , 1 ).float().numpy() SCREAMING_SNAKE_CASE : List[str] = DiffusionPipeline.numpy_to_pil(_lowerCamelCase )[0] return { "prompt": "An anime racoon running a marathon", "image": input_image, "generator": generator, "num_inference_steps": 2, "output_type": "np", } @skip_mps def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Any = '''cpu''' # ensure determinism for the device-dependent torch.Generator SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = StableUnCLIPImgaImgPipeline(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = sd_pipe.to(_lowerCamelCase ) sd_pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = self.get_dummy_inputs(_lowerCamelCase ) inputs.update({'''image_embeds''': None} ) SCREAMING_SNAKE_CASE : Optional[int] = sd_pipe(**_lowerCamelCase ).images SCREAMING_SNAKE_CASE : Dict = image[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) SCREAMING_SNAKE_CASE : Tuple = np.array([0.3_8_7_2, 0.7_2_2_4, 0.5_6_0_1, 0.4_7_4_1, 0.6_8_7_2, 0.5_8_1_4, 0.4_6_3_6, 0.3_8_6_7, 0.5_0_7_8] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : str = torch_device in ['''cpu''', '''mps'''] self._test_attention_slicing_forward_pass(test_max_difference=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Tuple = torch_device in ['''cpu''', '''mps'''] self._test_inference_batch_single_identical(test_max_difference=_lowerCamelCase ) @unittest.skipIf( torch_device != '''cuda''' or not is_xformers_available() , reason='''XFormers attention is only available with CUDA and `xformers` installed''' , ) def __lowerCAmelCase ( self ) ->Optional[int]: self._test_xformers_attention_forwardGenerator_pass(test_max_difference=_lowerCamelCase ) @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->int: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : List[Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_l_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[Any] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-l-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : List[str] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[int] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_h_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[str] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) torch.cuda.empty_cache() torch.cuda.reset_max_memory_allocated() torch.cuda.reset_peak_memory_stats() SCREAMING_SNAKE_CASE : str = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Dict = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Dict = pipe( _lowerCamelCase , '''anime turtle''' , num_inference_steps=2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = torch.cuda.max_memory_allocated() # make sure that less than 7 GB is allocated assert mem_bytes < 7 * 10**9
313
1
import copy from typing import Any, Dict, List, Optional, Union import numpy as np import torch from ...audio_utils import mel_filter_bank, spectrogram, window_function from ...feature_extraction_sequence_utils import SequenceFeatureExtractor from ...feature_extraction_utils import BatchFeature from ...utils import TensorType, logging a__ : Dict = logging.get_logger(__name__) class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = ['input_features', 'is_longer'] def __init__( self , _lowerCamelCase=64 , _lowerCamelCase=4_8000 , _lowerCamelCase=480 , _lowerCamelCase=10 , _lowerCamelCase=1024 , _lowerCamelCase=0.0 , _lowerCamelCase=False , _lowerCamelCase = 0 , _lowerCamelCase = 1_4000 , _lowerCamelCase = None , _lowerCamelCase = "fusion" , _lowerCamelCase = "repeatpad" , **_lowerCamelCase , ) ->int: super().__init__( feature_size=_lowerCamelCase , sampling_rate=_lowerCamelCase , padding_value=_lowerCamelCase , return_attention_mask=_lowerCamelCase , **_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Optional[Any] = top_db SCREAMING_SNAKE_CASE : Dict = truncation SCREAMING_SNAKE_CASE : str = padding SCREAMING_SNAKE_CASE : Any = fft_window_size SCREAMING_SNAKE_CASE : str = (fft_window_size >> 1) + 1 SCREAMING_SNAKE_CASE : str = hop_length SCREAMING_SNAKE_CASE : int = max_length_s SCREAMING_SNAKE_CASE : str = max_length_s * sampling_rate SCREAMING_SNAKE_CASE : List[Any] = sampling_rate SCREAMING_SNAKE_CASE : Optional[Any] = frequency_min SCREAMING_SNAKE_CASE : int = frequency_max SCREAMING_SNAKE_CASE : int = mel_filter_bank( num_frequency_bins=self.nb_frequency_bins , num_mel_filters=_lowerCamelCase , min_frequency=_lowerCamelCase , max_frequency=_lowerCamelCase , sampling_rate=_lowerCamelCase , norm=_lowerCamelCase , mel_scale='''htk''' , ) SCREAMING_SNAKE_CASE : int = mel_filter_bank( num_frequency_bins=self.nb_frequency_bins , num_mel_filters=_lowerCamelCase , min_frequency=_lowerCamelCase , max_frequency=_lowerCamelCase , sampling_rate=_lowerCamelCase , norm='''slaney''' , mel_scale='''slaney''' , ) def __lowerCAmelCase ( self ) ->Dict[str, Any]: SCREAMING_SNAKE_CASE : Dict = copy.deepcopy(self.__dict__ ) SCREAMING_SNAKE_CASE : Optional[Any] = self.__class__.__name__ if "mel_filters" in output: del output["mel_filters"] if "mel_filters_slaney" in output: del output["mel_filters_slaney"] return output def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->np.ndarray: SCREAMING_SNAKE_CASE : Dict = spectrogram( _lowerCamelCase , window_function(self.fft_window_size , '''hann''' ) , frame_length=self.fft_window_size , hop_length=self.hop_length , power=2.0 , mel_filters=_lowerCamelCase , log_mel='''dB''' , ) return log_mel_spectrogram.T def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : Optional[int] = np.array_split(list(range(0 , total_frames - chunk_frames + 1 ) ) , 3 ) if len(ranges[1] ) == 0: # if the audio is too short, we just use the first chunk SCREAMING_SNAKE_CASE : List[Any] = [0] if len(ranges[2] ) == 0: # if the audio is too short, we just use the first chunk SCREAMING_SNAKE_CASE : int = [0] # randomly choose index for each part SCREAMING_SNAKE_CASE : Dict = np.random.choice(ranges[0] ) SCREAMING_SNAKE_CASE : List[Any] = np.random.choice(ranges[1] ) SCREAMING_SNAKE_CASE : int = np.random.choice(ranges[2] ) SCREAMING_SNAKE_CASE : Tuple = mel[idx_front : idx_front + chunk_frames, :] SCREAMING_SNAKE_CASE : str = mel[idx_middle : idx_middle + chunk_frames, :] SCREAMING_SNAKE_CASE : Union[str, Any] = mel[idx_back : idx_back + chunk_frames, :] SCREAMING_SNAKE_CASE : Dict = torch.tensor(mel[None, None, :] ) SCREAMING_SNAKE_CASE : List[Any] = torch.nn.functional.interpolate( _lowerCamelCase , size=[chunk_frames, 64] , mode='''bilinear''' , align_corners=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = mel_shrink[0][0].numpy() SCREAMING_SNAKE_CASE : Dict = np.stack([mel_shrink, mel_chunk_front, mel_chunk_middle, mel_chunk_back] , axis=0 ) return mel_fusion def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->np.array: if waveform.shape[0] > max_length: if truncation == "rand_trunc": SCREAMING_SNAKE_CASE : Union[str, Any] = True # random crop to max_length (for compatibility) -> this should be handled by self.pad SCREAMING_SNAKE_CASE : List[Any] = len(_lowerCamelCase ) - max_length SCREAMING_SNAKE_CASE : Any = np.random.randint(0 , overflow + 1 ) SCREAMING_SNAKE_CASE : int = waveform[idx : idx + max_length] SCREAMING_SNAKE_CASE : Optional[Any] = self._np_extract_fbank_features(_lowerCamelCase , self.mel_filters_slaney )[None, :] elif truncation == "fusion": SCREAMING_SNAKE_CASE : Optional[int] = self._np_extract_fbank_features(_lowerCamelCase , self.mel_filters ) SCREAMING_SNAKE_CASE : Any = max_length // self.hop_length + 1 # the +1 related to how the spectrogram is computed SCREAMING_SNAKE_CASE : str = mel.shape[0] if chunk_frames == total_frames: # there is a corner case where the audio length is larger than max_length but smaller than max_length+hop_length. # In this case, we just use the whole audio. SCREAMING_SNAKE_CASE : int = np.stack([mel, mel, mel, mel] , axis=0 ) SCREAMING_SNAKE_CASE : Any = False else: SCREAMING_SNAKE_CASE : int = self._random_mel_fusion(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = True else: raise NotImplementedError(F"""data_truncating {truncation} not implemented""" ) else: SCREAMING_SNAKE_CASE : List[str] = False # only use repeat as a new possible value for padding. you repeat the audio before applying the usual max_length padding if waveform.shape[0] < max_length: if padding == "repeat": SCREAMING_SNAKE_CASE : str = int(max_length / len(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Tuple = np.stack(np.tile(_lowerCamelCase , n_repeat + 1 ) )[:max_length] if padding == "repeatpad": SCREAMING_SNAKE_CASE : List[Any] = int(max_length / len(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = np.stack(np.tile(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = np.pad(_lowerCamelCase , (0, max_length - waveform.shape[0]) , mode='''constant''' , constant_values=0 ) if truncation == "fusion": SCREAMING_SNAKE_CASE : Any = self._np_extract_fbank_features(_lowerCamelCase , self.mel_filters ) SCREAMING_SNAKE_CASE : Optional[Any] = np.stack([input_mel, input_mel, input_mel, input_mel] , axis=0 ) else: SCREAMING_SNAKE_CASE : Optional[int] = self._np_extract_fbank_features(_lowerCamelCase , self.mel_filters_slaney )[None, :] return input_mel, longer def __call__( self , _lowerCamelCase , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , **_lowerCamelCase , ) ->BatchFeature: SCREAMING_SNAKE_CASE : Union[str, Any] = truncation if truncation is not None else self.truncation SCREAMING_SNAKE_CASE : Tuple = padding if padding else self.padding if sampling_rate is not None: if sampling_rate != self.sampling_rate: raise ValueError( F"""The model corresponding to this feature extractor: {self.__class__.__name__} was trained using a""" F""" sampling rate of {self.sampling_rate}. Please make sure that the provided `raw_speech` input""" F""" was sampled with {self.sampling_rate} and not {sampling_rate}.""" ) else: logger.warning( '''It is strongly recommended to pass the `sampling_rate` argument to this function. ''' '''Failing to do so can result in silent errors that might be hard to debug.''' ) SCREAMING_SNAKE_CASE : Any = isinstance(_lowerCamelCase , np.ndarray ) and len(raw_speech.shape ) > 1 if is_batched_numpy and len(raw_speech.shape ) > 2: raise ValueError(F"""Only mono-channel audio is supported for input to {self}""" ) SCREAMING_SNAKE_CASE : Optional[Any] = is_batched_numpy or ( isinstance(_lowerCamelCase , (list, tuple) ) and (isinstance(raw_speech[0] , (np.ndarray, tuple, list) )) ) if is_batched: SCREAMING_SNAKE_CASE : List[str] = [np.asarray(_lowerCamelCase , dtype=np.floataa ) for speech in raw_speech] elif not is_batched and not isinstance(_lowerCamelCase , np.ndarray ): SCREAMING_SNAKE_CASE : Dict = np.asarray(_lowerCamelCase , dtype=np.floataa ) elif isinstance(_lowerCamelCase , np.ndarray ) and raw_speech.dtype is np.dtype(np.floataa ): SCREAMING_SNAKE_CASE : Union[str, Any] = raw_speech.astype(np.floataa ) # always return batch if not is_batched: SCREAMING_SNAKE_CASE : Any = [np.asarray(_lowerCamelCase )] # convert to mel spectrogram, truncate and pad if needed. SCREAMING_SNAKE_CASE : Union[str, Any] = [ self._get_input_mel(_lowerCamelCase , max_length if max_length else self.nb_max_samples , _lowerCamelCase , _lowerCamelCase ) for waveform in raw_speech ] SCREAMING_SNAKE_CASE : List[str] = [] SCREAMING_SNAKE_CASE : List[Any] = [] for mel, longer in padded_inputs: input_mel.append(_lowerCamelCase ) is_longer.append(_lowerCamelCase ) if truncation == "fusion" and sum(_lowerCamelCase ) == 0: # if no audio is longer than 10s, then randomly select one audio to be longer SCREAMING_SNAKE_CASE : str = np.random.randint(0 , len(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = True if isinstance(input_mel[0] , _lowerCamelCase ): SCREAMING_SNAKE_CASE : List[str] = [np.asarray(_lowerCamelCase , dtype=np.floataa ) for feature in input_mel] # is_longer is a list of bool SCREAMING_SNAKE_CASE : int = [[longer] for longer in is_longer] SCREAMING_SNAKE_CASE : Optional[Any] = {'''input_features''': input_mel, '''is_longer''': is_longer} SCREAMING_SNAKE_CASE : Union[str, Any] = BatchFeature(_lowerCamelCase ) if return_tensors is not None: SCREAMING_SNAKE_CASE : Optional[Any] = input_features.convert_to_tensors(_lowerCamelCase ) return input_features
313
from abc import ABC, abstractmethod from typing import List, Optional class a_ ( a__ ): """simple docstring""" def __init__( self ) ->List[str]: # test for the above condition self.test() def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = False while not completed: if counter == 1: self.reset() SCREAMING_SNAKE_CASE : List[Any] = self.advance() if not self.does_advance(_lowerCamelCase ): raise Exception( '''Custom Constraint is not defined correctly. self.does_advance(self.advance()) must be true.''' ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.update(_lowerCamelCase ) counter += 1 if counter > 1_0000: raise Exception('''update() does not fulfill the constraint.''' ) if self.remaining() != 0: raise Exception('''Custom Constraint is not defined correctly.''' ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[int]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Union[str, Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Any: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->int: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`token_ids` has to be a non-empty list, but is {token_ids}.""" ) if any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ): raise ValueError(F"""Each list in `token_ids` has to be a list of positive integers, but is {token_ids}.""" ) SCREAMING_SNAKE_CASE : Optional[Any] = token_ids SCREAMING_SNAKE_CASE : Union[str, Any] = len(self.token_ids ) SCREAMING_SNAKE_CASE : Any = -1 # the index of the currently fulfilled step SCREAMING_SNAKE_CASE : Any = False def __lowerCAmelCase ( self ) ->List[Any]: if self.completed: return None return self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) if self.completed: return False return token_id == self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[str]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : str = False SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = False if self.does_advance(_lowerCamelCase ): self.fulfilled_idx += 1 SCREAMING_SNAKE_CASE : str = True if self.fulfilled_idx == (self.seqlen - 1): SCREAMING_SNAKE_CASE : Any = True SCREAMING_SNAKE_CASE : Union[str, Any] = completed else: # failed to make progress. SCREAMING_SNAKE_CASE : Dict = True self.reset() return stepped, completed, reset def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = 0 def __lowerCAmelCase ( self ) ->Any: return self.seqlen - (self.fulfilled_idx + 1) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Dict: SCREAMING_SNAKE_CASE : Any = PhrasalConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : Dict = self.seqlen SCREAMING_SNAKE_CASE : int = self.fulfilled_idx SCREAMING_SNAKE_CASE : Tuple = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=True ) ->Dict: SCREAMING_SNAKE_CASE : Any = max([len(_lowerCamelCase ) for one in nested_token_ids] ) SCREAMING_SNAKE_CASE : List[str] = {} for token_ids in nested_token_ids: SCREAMING_SNAKE_CASE : Optional[Any] = root for tidx, token_id in enumerate(_lowerCamelCase ): if token_id not in level: SCREAMING_SNAKE_CASE : Any = {} SCREAMING_SNAKE_CASE : Tuple = level[token_id] if no_subsets and self.has_subsets(_lowerCamelCase , _lowerCamelCase ): raise ValueError( '''Each list in `nested_token_ids` can\'t be a complete subset of another list, but is''' F""" {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = root def __lowerCAmelCase ( self , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = self.trie for current_token in current_seq: SCREAMING_SNAKE_CASE : int = start[current_token] SCREAMING_SNAKE_CASE : Optional[int] = list(start.keys() ) return next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.next_tokens(_lowerCamelCase ) return len(_lowerCamelCase ) == 0 def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = list(root.values() ) if len(_lowerCamelCase ) == 0: return 1 else: return sum([self.count_leaves(_lowerCamelCase ) for nn in next_nodes] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = self.count_leaves(_lowerCamelCase ) return len(_lowerCamelCase ) != leaf_count class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->str: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`nested_token_ids` has to be a non-empty list, but is {nested_token_ids}.""" ) if any(not isinstance(_lowerCamelCase , _lowerCamelCase ) for token_ids in nested_token_ids ): raise ValueError(F"""`nested_token_ids` has to be a list of lists, but is {nested_token_ids}.""" ) if any( any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ) for token_ids in nested_token_ids ): raise ValueError( F"""Each list in `nested_token_ids` has to be a list of positive integers, but is {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = DisjunctiveTrie(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = nested_token_ids SCREAMING_SNAKE_CASE : Optional[int] = self.trie.max_height SCREAMING_SNAKE_CASE : Union[str, Any] = [] SCREAMING_SNAKE_CASE : Optional[int] = False def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : str = self.trie.next_tokens(self.current_seq ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : List[str] = self.trie.next_tokens(self.current_seq ) return token_id in next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : int = False SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = False if self.does_advance(_lowerCamelCase ): self.current_seq.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = True else: SCREAMING_SNAKE_CASE : Dict = True self.reset() SCREAMING_SNAKE_CASE : Any = self.trie.reached_leaf(self.current_seq ) SCREAMING_SNAKE_CASE : List[Any] = completed return stepped, completed, reset def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = [] def __lowerCAmelCase ( self ) ->Optional[Any]: if self.completed: # since this can be completed without reaching max height return 0 else: return self.seqlen - len(self.current_seq ) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->List[str]: SCREAMING_SNAKE_CASE : str = DisjunctiveConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : str = self.seqlen SCREAMING_SNAKE_CASE : int = self.current_seq SCREAMING_SNAKE_CASE : Optional[int] = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[Any] = constraints # max # of steps required to fulfill a given constraint SCREAMING_SNAKE_CASE : str = max([c.seqlen for c in constraints] ) SCREAMING_SNAKE_CASE : List[str] = len(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = False self.init_state() def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Any = [] SCREAMING_SNAKE_CASE : List[Any] = None SCREAMING_SNAKE_CASE : Tuple = [constraint.copy(stateful=_lowerCamelCase ) for constraint in self.constraints] def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = 0 if self.inprogress_constraint: # extra points for having a constraint mid-fulfilled add += self.max_seqlen - self.inprogress_constraint.remaining() return (len(self.complete_constraints ) * self.max_seqlen) + add def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Tuple = [] if self.inprogress_constraint is None: for constraint in self.pending_constraints: # "pending" == "unfulfilled yet" SCREAMING_SNAKE_CASE : Optional[int] = constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = self.inprogress_constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: self.init_state() if token_ids is not None: for token in token_ids: # completes or steps **one** constraint SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.add(_lowerCamelCase ) # the entire list of constraints are fulfilled if self.completed: break def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` should be an `int`, but is `{token_id}`.""" ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = False, False if self.completed: SCREAMING_SNAKE_CASE : List[str] = True SCREAMING_SNAKE_CASE : Optional[int] = False return complete, stepped if self.inprogress_constraint is not None: # In the middle of fulfilling a constraint. If the `token_id` *does* makes an incremental progress to current # job, simply update the state SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.inprogress_constraint.update(_lowerCamelCase ) if reset: # 1. If the next token breaks the progress, then we must restart. # e.g. constraint = "I love pies" and sequence so far is "I love" but `token_id` == "books". # But that doesn't mean we self.init_state(), since we only reset the state for this particular # constraint, not the full list of constraints. self.pending_constraints.append(self.inprogress_constraint.copy(stateful=_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = None if complete: # 2. If the next token completes the constraint, move it to completed list, set # inprogress to None. If there are no pending constraints either, then this full list of constraints # is complete. self.complete_constraints.append(self.inprogress_constraint ) SCREAMING_SNAKE_CASE : str = None if len(self.pending_constraints ) == 0: # we're done! SCREAMING_SNAKE_CASE : Optional[Any] = True else: # Not in the middle of fulfilling a constraint. So does this `token_id` helps us step towards any of our list # of constraints? for cidx, pending_constraint in enumerate(self.pending_constraints ): if pending_constraint.does_advance(_lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = pending_constraint.update(_lowerCamelCase ) if not stepped: raise Exception( '''`constraint.update(token_id)` is not yielding incremental progress, ''' '''even though `constraint.does_advance(token_id)` is true.''' ) if complete: self.complete_constraints.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = None if not complete and stepped: SCREAMING_SNAKE_CASE : Optional[Any] = pending_constraint if complete or stepped: # If we made any progress at all, then it's at least not a "pending constraint". SCREAMING_SNAKE_CASE : Union[str, Any] = ( self.pending_constraints[:cidx] + self.pending_constraints[cidx + 1 :] ) if len(self.pending_constraints ) == 0 and self.inprogress_constraint is None: # If there's no longer any pending after this and no inprogress either, then we must be # complete. SCREAMING_SNAKE_CASE : str = True break # prevent accidentally stepping through multiple constraints with just one token. return complete, stepped def __lowerCAmelCase ( self , _lowerCamelCase=True ) ->str: SCREAMING_SNAKE_CASE : Dict = ConstraintListState(self.constraints ) # we actually never though self.constraints objects # throughout this process. So it's at initialization state. if stateful: SCREAMING_SNAKE_CASE : str = [ constraint.copy(stateful=_lowerCamelCase ) for constraint in self.complete_constraints ] if self.inprogress_constraint is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.inprogress_constraint.copy(stateful=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = [constraint.copy() for constraint in self.pending_constraints] return new_state
313
1
import numpy as np import pandas as pd from sklearn.preprocessing import Normalizer from sklearn.svm import SVR from statsmodels.tsa.statespace.sarimax import SARIMAX def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = np.array([[1, item, train_mtch[i]] for i, item in enumerate(a__ )] ) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = np.dot(np.dot(np.linalg.inv(np.dot(x.transpose() , a__ ) ) , x.transpose() ) , a__ ) return abs(beta[0] + test_dt[0] * beta[1] + test_mtch[0] + beta[2] ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = (1, 2, 1) SCREAMING_SNAKE_CASE : int = (1, 1, 0, 7) SCREAMING_SNAKE_CASE : Dict = SARIMAX( a__ , exog=a__ , order=a__ , seasonal_order=a__ ) SCREAMING_SNAKE_CASE : str = model.fit(disp=a__ , maxiter=600 , method='''nm''' ) SCREAMING_SNAKE_CASE : Tuple = model_fit.predict(1 , len(a__ ) , exog=[test_match] ) return result[0] def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = SVR(kernel='''rbf''' , C=1 , gamma=0.1 , epsilon=0.1 ) regressor.fit(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[int] = regressor.predict(a__ ) return y_pred[0] def UpperCAmelCase_( a__ ): """simple docstring""" train_user.sort() SCREAMING_SNAKE_CASE : Any = np.percentile(a__ , 25 ) SCREAMING_SNAKE_CASE : List[str] = np.percentile(a__ , 75 ) SCREAMING_SNAKE_CASE : Dict = qa - qa SCREAMING_SNAKE_CASE : Union[str, Any] = qa - (iqr * 0.1) return low_lim def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Optional[int] = 0 for i in list_vote: if i > actual_result: SCREAMING_SNAKE_CASE : List[str] = not_safe + 1 else: if abs(abs(a__ ) - abs(a__ ) ) <= 0.1: safe += 1 else: not_safe += 1 return safe > not_safe if __name__ == "__main__": # data_input_df = pd.read_csv("ex_data.csv", header=None) a__ : Optional[Any] = [[18_231, 0.0, 1], [22_621, 1.0, 2], [15_675, 0.0, 3], [23_583, 1.0, 4]] a__ : Optional[int] = pd.DataFrame( data_input, columns=['''total_user''', '''total_even''', '''days'''] ) a__ : str = Normalizer().fit_transform(data_input_df.values) # split data a__ : int = normalize_df[:, 2].tolist() a__ : int = normalize_df[:, 0].tolist() a__ : Optional[Any] = normalize_df[:, 1].tolist() # for svr (input variable = total date and total match) a__ : Union[str, Any] = normalize_df[:, [1, 2]].tolist() a__ : Optional[int] = x[: len(x) - 1] a__ : List[str] = x[len(x) - 1 :] # for linear regression & sarimax a__ : str = total_date[: len(total_date) - 1] a__ : Optional[int] = total_user[: len(total_user) - 1] a__ : Optional[Any] = total_match[: len(total_match) - 1] a__ : int = total_date[len(total_date) - 1 :] a__ : List[str] = total_user[len(total_user) - 1 :] a__ : Optional[Any] = total_match[len(total_match) - 1 :] # voting system with forecasting a__ : Union[str, Any] = [ linear_regression_prediction( trn_date, trn_user, trn_match, tst_date, tst_match ), sarimax_predictor(trn_user, trn_match, tst_match), support_vector_regressor(x_train, x_test, trn_user), ] # check the safety of today's data a__ : Union[str, Any] = '''''' if data_safety_checker(res_vote, tst_user) else '''not ''' print('''Today\'s data is {not_str}safe.''')
313
import argparse import random import joblib import numpy as np import torch from igf.igf import ( SecondaryLearner, collect_objective_set, compute_perplexity, generate_datasets, load_gpta, recopy_gpta, set_seed, train_secondary_learner, ) from torch.utils.data import DataLoader, RandomSampler from transformers import GPTaLMHeadModel def UpperCAmelCase_( a__=32 , a__=10 , a__=100 , a__=1_026 , a__=True , a__="data/tokenized_stories_train_wikitext103.jbl" , a__="igf_context_pairs.jbl" , ): """simple docstring""" set_seed(3 ) # generate train_data and objective_set SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = generate_datasets( a__ , a__ , number=a__ , min_len=1_026 , trim=a__ ) # keeps model same across runs set_seed(4 ) # model, lm_optimizer, lm_scheduler = recopy_gpt2(model, device, max_steps) # store original model weights # can we train on GPU? SCREAMING_SNAKE_CASE : str = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) # load pretrained model SCREAMING_SNAKE_CASE : Dict = load_gpta('''gpt2''' ).to(a__ ) print('''computing perplexity on objective set''' ) SCREAMING_SNAKE_CASE : int = compute_perplexity(a__ , a__ , a__ ).item() print('''perplexity on objective set:''' , a__ ) # collect igf pairs and save to file demo.jbl collect_objective_set(a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ ) # clean up, delete model and data we don't need anymore del model, train_data, objective_set torch.cuda.empty_cache() def UpperCAmelCase_( a__ , a__=15 , a__=128 , a__=100 , a__="igf_model.pt" , ): """simple docstring""" set_seed(42 ) # Load pre-trained model SCREAMING_SNAKE_CASE : List[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) # Initialize secondary learner to use embedding weights of model SCREAMING_SNAKE_CASE : str = SecondaryLearner(a__ ) # Train secondary learner SCREAMING_SNAKE_CASE : Union[str, Any] = train_secondary_learner( a__ , a__ , max_epochs=a__ , batch_size=a__ , eval_freq=100 , igf_model_path=a__ , ) del model, secondary_learner_train_data torch.cuda.empty_cache() return secondary_learner def UpperCAmelCase_( a__ , a__ , a__ , a__=32 , a__=1_000 , a__=16 , a__=1.0 , a__=recopy_gpta , a__=None , a__=10 , a__="gpt2_finetuned.pt" , ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) SCREAMING_SNAKE_CASE : Optional[int] = RandomSampler(a__ ) SCREAMING_SNAKE_CASE : Dict = DataLoader(a__ , sampler=a__ ) SCREAMING_SNAKE_CASE : Tuple = max_steps // (len(a__ )) + 1 SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = torch.zeros((1, context_len) , dtype=torch.long , device=a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = recopy_model(a__ , a__ , a__ ) model.train() if secondary_learner is not None: secondary_learner.to(a__ ) secondary_learner.eval() SCREAMING_SNAKE_CASE : List[str] = [] SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Tuple = [] # Compute the performance of the transformer model at the beginning SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) for epoch in range(int(a__ ) ): for step, example in enumerate(a__ ): torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Union[str, Any] = random.randint(0 , example.size(2 ) - context_len - 1 ) SCREAMING_SNAKE_CASE : Optional[int] = example[0, 0, start : start + context_len] lm_optimizer.zero_grad() SCREAMING_SNAKE_CASE : Optional[Any] = model(a__ , labels=a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = True if secondary_learner is not None: SCREAMING_SNAKE_CASE : List[str] = secondary_learner.forward( torch.tensor(a__ , dtype=torch.long , device=a__ ).unsqueeze(0 ) )[0].item() observed_qs.append(float(a__ ) ) # Here we implement the simple non-constant threshold for the predicted IG(X) value # We will decay the selectivity of our secondary learner filter from # 1 standard deviation above average to 1 below average after 10 batches. if global_step == 10: SCREAMING_SNAKE_CASE : Dict = -1 if predicted_q < threshold: SCREAMING_SNAKE_CASE : str = False # If we passed the filter, add the context to the batch! if do_backprop: contexts.append(np.array(context.cpu() ) ) SCREAMING_SNAKE_CASE : List[str] = outputs[0] lm_loss.backward() examples += 1 del outputs # Once the batch is filled with enough contexts, backprop on the batch. if examples == batch_size: torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Any = 0 # Do LM backprop torch.nn.utils.clip_grad_norm_(model.parameters() , 3.0 ) lm_optimizer.step() lm_scheduler.step() # Update learning rate schedule global_step += 1 # Compute the performance of the transformer model at this batch if global_step % eval_interval == 0: SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) # Break out of the loop after 60 batches if max_steps > 0 and global_step > 60: break if max_steps > 0 and global_step > 60: break # save finetuned transformer model torch.save(model.state_dict() , a__ ) torch.cuda.empty_cache() # Do some cleaning up so we can reinitialize for the next run of this function del lm_optimizer del lm_scheduler return model def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = argparse.ArgumentParser(description='''Fine-tune a transformer model with IGF on a language modeling task''' ) # Required parameters parser.add_argument( '''--data_dir''' , default=a__ , type=a__ , required=a__ , help='''The input data dir. Should contain data files for WikiText.''' , ) parser.add_argument( '''--model_name_or_path''' , default=a__ , type=a__ , required=a__ , help='''Path to pretrained model or model identifier from huggingface.co/models''' , ) parser.add_argument( '''--data_file''' , type=a__ , default=a__ , help=( '''A jbl file containing tokenized data which can be split as objective dataset, ''' '''train_dataset and test_dataset.''' ) , ) parser.add_argument( '''--igf_data_file''' , type=a__ , default=a__ , help='''A jbl file containing the context and information gain pairs to train secondary learner.''' , ) parser.add_argument( '''--output_dir''' , default=a__ , type=a__ , required=a__ , help='''The output directory where the final fine-tuned model is stored.''' , ) parser.add_argument( '''--tokenizer_name''' , default=a__ , type=a__ , help='''Pretrained tokenizer name or path if not the same as model_name''' , ) parser.add_argument('''--seed''' , type=a__ , default=a__ , help='''A seed for reproducible training.''' ) parser.add_argument( '''--context_len''' , default=32 , type=a__ , help=( '''The maximum total input sequence length after tokenization. Sequences longer ''' '''than this will be truncated, sequences shorter will be padded.''' ) , ) parser.add_argument( '''--size_objective_set''' , default=100 , type=a__ , help='''number of articles that are long enough to be used as our objective set''' , ) parser.add_argument( '''--eval_freq''' , default=100 , type=a__ , help='''secondary model evaluation is triggered at eval_freq''' ) parser.add_argument('''--max_steps''' , default=1_000 , type=a__ , help='''To calculate training epochs''' ) parser.add_argument( '''--secondary_learner_batch_size''' , default=128 , type=a__ , help='''batch size of training data for secondary learner''' , ) parser.add_argument( '''--batch_size''' , default=16 , type=a__ , help='''batch size of training data of language model(gpt2) ''' ) parser.add_argument( '''--eval_interval''' , default=10 , type=a__ , help=( '''decay the selectivity of our secondary learner filter from''' '''1 standard deviation above average to 1 below average after 10 batches''' ) , ) parser.add_argument( '''--number''' , default=100 , type=a__ , help='''The number of examples split to be used as objective_set/test_data''' ) parser.add_argument( '''--min_len''' , default=1_026 , type=a__ , help='''The minimum length of the article to be used as objective set''' ) parser.add_argument( '''--secondary_learner_max_epochs''' , default=15 , type=a__ , help='''number of epochs to train secondary learner''' ) parser.add_argument('''--trim''' , default=a__ , type=a__ , help='''truncate the example if it exceeds context length''' ) parser.add_argument( '''--threshold''' , default=1.0 , type=a__ , help=( '''The threshold value used by secondary learner to filter the train_data and allow only''' ''' informative data as input to the model''' ) , ) parser.add_argument('''--finetuned_model_name''' , default='''gpt2_finetuned.pt''' , type=a__ , help='''finetuned_model_name''' ) parser.add_argument( '''--recopy_model''' , default=a__ , type=a__ , help='''Reset the model to the original pretrained GPT-2 weights after each iteration''' , ) # function calls # Collecting *n* pairs of context and information gain(X, IG(X)) for training the secondary learner generate_n_pairs( context_len=32 , max_steps=10 , size_objective_set=100 , min_len=1_026 , trim=a__ , data_file='''data/tokenized_stories_train_wikitext103.jbl''' , igf_data_file='''igf_context_pairs.jbl''' , ) # Load train data for secondary learner SCREAMING_SNAKE_CASE : List[Any] = joblib.load('''data/IGF_values.jbl''' ) # Train secondary learner SCREAMING_SNAKE_CASE : Tuple = training_secondary_learner( a__ , secondary_learner_max_epochs=15 , secondary_learner_batch_size=128 , eval_freq=100 , igf_model_path='''igf_model.pt''' , ) # load pretrained gpt2 model SCREAMING_SNAKE_CASE : Optional[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) set_seed(42 ) # Generate train and test data to train and evaluate gpt2 model SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = generate_datasets( context_len=32 , file='''data/tokenized_stories_train_wikitext103.jbl''' , number=100 , min_len=1_026 , trim=a__ ) # fine-tuning of the gpt2 model using igf (Information Gain Filtration) finetune( a__ , a__ , a__ , context_len=32 , max_steps=1_000 , batch_size=16 , threshold=1.0 , recopy_model=a__ , secondary_learner=a__ , eval_interval=10 , finetuned_model_name='''gpt2_finetuned.pt''' , ) if __name__ == "__main__": main()
313
1
import unittest from transformers import EsmConfig, is_torch_available from transformers.testing_utils import TestCasePlus, require_torch, slow, torch_device from ...test_configuration_common import ConfigTester from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask from ...test_pipeline_mixin import PipelineTesterMixin if is_torch_available(): import torch from transformers.models.esm.modeling_esmfold import EsmForProteinFolding class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=13 , _lowerCamelCase=7 , _lowerCamelCase=False , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=False , _lowerCamelCase=19 , _lowerCamelCase=32 , _lowerCamelCase=5 , _lowerCamelCase=4 , _lowerCamelCase=37 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=16 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=3 , _lowerCamelCase=4 , _lowerCamelCase=None , ) ->List[Any]: SCREAMING_SNAKE_CASE : int = parent SCREAMING_SNAKE_CASE : Any = batch_size SCREAMING_SNAKE_CASE : Dict = seq_length SCREAMING_SNAKE_CASE : Any = is_training SCREAMING_SNAKE_CASE : Any = use_input_mask SCREAMING_SNAKE_CASE : Tuple = use_token_type_ids SCREAMING_SNAKE_CASE : Tuple = use_labels SCREAMING_SNAKE_CASE : Dict = vocab_size SCREAMING_SNAKE_CASE : List[str] = hidden_size SCREAMING_SNAKE_CASE : List[str] = num_hidden_layers SCREAMING_SNAKE_CASE : Any = num_attention_heads SCREAMING_SNAKE_CASE : Optional[int] = intermediate_size SCREAMING_SNAKE_CASE : int = hidden_act SCREAMING_SNAKE_CASE : Any = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Union[str, Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Any = type_vocab_size SCREAMING_SNAKE_CASE : Optional[int] = type_sequence_label_size SCREAMING_SNAKE_CASE : Dict = initializer_range SCREAMING_SNAKE_CASE : Optional[Any] = num_labels SCREAMING_SNAKE_CASE : str = num_choices SCREAMING_SNAKE_CASE : Optional[Any] = scope def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Any = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size ) SCREAMING_SNAKE_CASE : Optional[int] = None if self.use_input_mask: SCREAMING_SNAKE_CASE : Dict = random_attention_mask([self.batch_size, self.seq_length] ) SCREAMING_SNAKE_CASE : Any = None SCREAMING_SNAKE_CASE : List[Any] = None SCREAMING_SNAKE_CASE : Tuple = None if self.use_labels: SCREAMING_SNAKE_CASE : Dict = ids_tensor([self.batch_size] , self.type_sequence_label_size ) SCREAMING_SNAKE_CASE : List[str] = ids_tensor([self.batch_size, self.seq_length] , self.num_labels ) SCREAMING_SNAKE_CASE : Any = ids_tensor([self.batch_size] , self.num_choices ) SCREAMING_SNAKE_CASE : List[Any] = self.get_config() return config, input_ids, input_mask, sequence_labels, token_labels, choice_labels def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = EsmConfig( vocab_size=33 , hidden_size=self.hidden_size , pad_token_id=1 , 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 , initializer_range=self.initializer_range , is_folding_model=_lowerCamelCase , esmfold_config={'''trunk''': {'''num_blocks''': 2}, '''fp16_esm''': False} , ) return config def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : List[Any] = EsmForProteinFolding(config=_lowerCamelCase ).float() model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Dict = model(_lowerCamelCase , attention_mask=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = model(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = model(_lowerCamelCase ) self.parent.assertEqual(result.positions.shape , (8, self.batch_size, self.seq_length, 14, 3) ) self.parent.assertEqual(result.angles.shape , (8, self.batch_size, self.seq_length, 7, 2) ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : List[str] = self.prepare_config_and_inputs() ( ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ) : Optional[Any] = config_and_inputs SCREAMING_SNAKE_CASE : Union[str, Any] = {'''input_ids''': input_ids, '''attention_mask''': input_mask} return config, inputs_dict @require_torch class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = False __SCREAMING_SNAKE_CASE : Union[str, Any] = (EsmForProteinFolding,) if is_torch_available() else () __SCREAMING_SNAKE_CASE : Optional[int] = () __SCREAMING_SNAKE_CASE : str = {} if is_torch_available() else {} __SCREAMING_SNAKE_CASE : Optional[int] = False def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Dict = EsmFoldModelTester(self ) SCREAMING_SNAKE_CASE : int = ConfigTester(self , config_class=_lowerCamelCase , hidden_size=37 ) def __lowerCAmelCase ( self ) ->Optional[Any]: self.config_tester.run_common_tests() def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*_lowerCamelCase ) @unittest.skip('''Does not support attention outputs''' ) def __lowerCAmelCase ( self ) ->int: pass @unittest.skip def __lowerCAmelCase ( self ) ->Dict: pass @unittest.skip('''Esm does not support embedding resizing''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @unittest.skip('''Esm does not support embedding resizing''' ) def __lowerCAmelCase ( self ) ->Dict: pass @unittest.skip('''ESMFold does not support passing input embeds!''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass @unittest.skip('''ESMFold does not support head pruning.''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass @unittest.skip('''ESMFold does not support head pruning.''' ) def __lowerCAmelCase ( self ) ->Optional[int]: pass @unittest.skip('''ESMFold does not support head pruning.''' ) def __lowerCAmelCase ( self ) ->int: pass @unittest.skip('''ESMFold does not support head pruning.''' ) def __lowerCAmelCase ( self ) ->List[str]: pass @unittest.skip('''ESMFold does not support head pruning.''' ) def __lowerCAmelCase ( self ) ->List[str]: pass @unittest.skip('''ESMFold does not output hidden states in the normal way.''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @unittest.skip('''ESMfold does not output hidden states in the normal way.''' ) def __lowerCAmelCase ( self ) ->str: pass @unittest.skip('''ESMFold only has one output format.''' ) def __lowerCAmelCase ( self ) ->Optional[int]: pass @unittest.skip('''This test doesn\'t work for ESMFold and doesn\'t test core functionality''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @unittest.skip('''ESMFold does not support input chunking.''' ) def __lowerCAmelCase ( self ) ->int: pass @unittest.skip('''ESMFold doesn\'t respect you and it certainly doesn\'t respect your initialization arguments.''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @unittest.skip('''ESMFold doesn\'t support torchscript compilation.''' ) def __lowerCAmelCase ( self ) ->Optional[int]: pass @unittest.skip('''ESMFold doesn\'t support torchscript compilation.''' ) def __lowerCAmelCase ( self ) ->Optional[int]: pass @unittest.skip('''ESMFold doesn\'t support torchscript compilation.''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass @unittest.skip('''ESMFold doesn\'t support data parallel.''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @unittest.skip('''Will be fixed soon by reducing the size of the model used for common tests.''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass @require_torch class a_ ( a__ ): """simple docstring""" @slow def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : int = EsmForProteinFolding.from_pretrained('''facebook/esmfold_v1''' ).float() model.eval() SCREAMING_SNAKE_CASE : int = torch.tensor([[0, 6, 4, 13, 5, 4, 16, 12, 11, 7, 2]] ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase )['''positions'''] SCREAMING_SNAKE_CASE : List[str] = torch.tensor([2.5_8_2_8, 0.7_9_9_3, -1_0.9_3_3_4] , dtype=torch.floataa ) self.assertTrue(torch.allclose(position_outputs[0, 0, 0, 0] , _lowerCamelCase , atol=1e-4 ) )
313
import logging from pathlib import Path import numpy as np import pytorch_lightning as pl import torch from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint from pytorch_lightning.utilities import rank_zero_only from utils_rag import save_json def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = filter(lambda a__ : p.requires_grad , model.parameters() ) SCREAMING_SNAKE_CASE : List[Any] = sum([np.prod(p.size() ) for p in model_parameters] ) return params a__ : Any = logging.getLogger(__name__) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if metric == "rouge2": SCREAMING_SNAKE_CASE : str = '''{val_avg_rouge2:.4f}-{step_count}''' elif metric == "bleu": SCREAMING_SNAKE_CASE : List[Any] = '''{val_avg_bleu:.4f}-{step_count}''' elif metric == "em": SCREAMING_SNAKE_CASE : int = '''{val_avg_em:.4f}-{step_count}''' elif metric == "loss": SCREAMING_SNAKE_CASE : int = '''{val_avg_loss:.4f}-{step_count}''' else: raise NotImplementedError( F"""seq2seq callbacks only support rouge2 and bleu, got {metric}, You can make your own by adding to this""" ''' function.''' ) SCREAMING_SNAKE_CASE : Dict = ModelCheckpoint( dirpath=a__ , filename=a__ , monitor=F"""val_{metric}""" , mode='''max''' , save_top_k=1 , every_n_epochs=1 , ) return checkpoint_callback def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return EarlyStopping( monitor=F"""val_{metric}""" , mode='''min''' if '''loss''' in metric else '''max''' , patience=a__ , verbose=a__ , ) class a_ ( pl.Callback ): """simple docstring""" def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = {F"""lr_group_{i}""": param['''lr'''] for i, param in enumerate(pl_module.trainer.optimizers[0].param_groups )} pl_module.logger.log_metrics(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=True ) ->None: logger.info(F"""***** {type_path} results at step {trainer.global_step:05d} *****""" ) SCREAMING_SNAKE_CASE : Optional[int] = trainer.callback_metrics trainer.logger.log_metrics({k: v for k, v in metrics.items() if k not in ['''log''', '''progress_bar''', '''preds''']} ) # Log results SCREAMING_SNAKE_CASE : List[str] = Path(pl_module.hparams.output_dir ) if type_path == "test": SCREAMING_SNAKE_CASE : Any = od / '''test_results.txt''' SCREAMING_SNAKE_CASE : Optional[int] = od / '''test_generations.txt''' else: # this never gets hit. I prefer not to save intermediate generations, and results are in metrics.json # If people want this it will be easy enough to add back. SCREAMING_SNAKE_CASE : str = od / F"""{type_path}_results/{trainer.global_step:05d}.txt""" SCREAMING_SNAKE_CASE : Tuple = od / F"""{type_path}_generations/{trainer.global_step:05d}.txt""" results_file.parent.mkdir(exist_ok=_lowerCamelCase ) generations_file.parent.mkdir(exist_ok=_lowerCamelCase ) with open(_lowerCamelCase , '''a+''' ) as writer: for key in sorted(_lowerCamelCase ): if key in ["log", "progress_bar", "preds"]: continue SCREAMING_SNAKE_CASE : Tuple = metrics[key] if isinstance(_lowerCamelCase , torch.Tensor ): SCREAMING_SNAKE_CASE : List[Any] = val.item() SCREAMING_SNAKE_CASE : Tuple = F"""{key}: {val:.6f}\n""" writer.write(_lowerCamelCase ) if not save_generations: return if "preds" in metrics: SCREAMING_SNAKE_CASE : Optional[Any] = '''\n'''.join(metrics['''preds'''] ) generations_file.open('''w+''' ).write(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: try: SCREAMING_SNAKE_CASE : Any = pl_module.model.model.num_parameters() except AttributeError: SCREAMING_SNAKE_CASE : Optional[int] = pl_module.model.num_parameters() SCREAMING_SNAKE_CASE : int = count_trainable_parameters(_lowerCamelCase ) # mp stands for million parameters trainer.logger.log_metrics({'''n_params''': npars, '''mp''': npars / 1e6, '''grad_mp''': n_trainable_pars / 1e6} ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: save_json(pl_module.metrics , pl_module.metrics_save_path ) return self._write_logs(_lowerCamelCase , _lowerCamelCase , '''test''' ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Tuple: save_json(pl_module.metrics , pl_module.metrics_save_path ) # Uncommenting this will save val generations # return self._write_logs(trainer, pl_module, "valid")
313
1
from random import shuffle import tensorflow as tf from numpy import array def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = int(a__ ) assert noofclusters < len(a__ ) # Find out the dimensionality SCREAMING_SNAKE_CASE : Dict = len(vectors[0] ) # Will help select random centroids from among the available vectors SCREAMING_SNAKE_CASE : Any = list(range(len(a__ ) ) ) shuffle(a__ ) # GRAPH OF COMPUTATION # We initialize a new graph and set it as the default during each run # of this algorithm. This ensures that as this function is called # multiple times, the default graph doesn't keep getting crowded with # unused ops and Variables from previous function calls. SCREAMING_SNAKE_CASE : Union[str, Any] = tf.Graph() with graph.as_default(): # SESSION OF COMPUTATION SCREAMING_SNAKE_CASE : Optional[Any] = tf.Session() ##CONSTRUCTING THE ELEMENTS OF COMPUTATION ##First lets ensure we have a Variable vector for each centroid, ##initialized to one of the vectors from the available data points SCREAMING_SNAKE_CASE : Union[str, Any] = [ tf.Variable(vectors[vector_indices[i]] ) for i in range(a__ ) ] ##These nodes will assign the centroid Variables the appropriate ##values SCREAMING_SNAKE_CASE : str = tf.placeholder('''float64''' , [dim] ) SCREAMING_SNAKE_CASE : Dict = [] for centroid in centroids: cent_assigns.append(tf.assign(a__ , a__ ) ) ##Variables for cluster assignments of individual vectors(initialized ##to 0 at first) SCREAMING_SNAKE_CASE : int = [tf.Variable(0 ) for i in range(len(a__ ) )] ##These nodes will assign an assignment Variable the appropriate ##value SCREAMING_SNAKE_CASE : str = tf.placeholder('''int32''' ) SCREAMING_SNAKE_CASE : int = [] for assignment in assignments: cluster_assigns.append(tf.assign(a__ , a__ ) ) ##Now lets construct the node that will compute the mean # The placeholder for the input SCREAMING_SNAKE_CASE : Optional[int] = tf.placeholder('''float''' , [None, dim] ) # The Node/op takes the input and computes a mean along the 0th # dimension, i.e. the list of input vectors SCREAMING_SNAKE_CASE : int = tf.reduce_mean(a__ , 0 ) ##Node for computing Euclidean distances # Placeholders for input SCREAMING_SNAKE_CASE : Union[str, Any] = tf.placeholder('''float''' , [dim] ) SCREAMING_SNAKE_CASE : str = tf.placeholder('''float''' , [dim] ) SCREAMING_SNAKE_CASE : Union[str, Any] = tf.sqrt(tf.reduce_sum(tf.pow(tf.sub(a__ , a__ ) , 2 ) ) ) ##This node will figure out which cluster to assign a vector to, ##based on Euclidean distances of the vector from the centroids. # Placeholder for input SCREAMING_SNAKE_CASE : int = tf.placeholder('''float''' , [noofclusters] ) SCREAMING_SNAKE_CASE : Union[str, Any] = tf.argmin(a__ , 0 ) ##INITIALIZING STATE VARIABLES ##This will help initialization of all Variables defined with respect ##to the graph. The Variable-initializer should be defined after ##all the Variables have been constructed, so that each of them ##will be included in the initialization. SCREAMING_SNAKE_CASE : int = tf.initialize_all_variables() # Initialize all variables sess.run(a__ ) ##CLUSTERING ITERATIONS # Now perform the Expectation-Maximization steps of K-Means clustering # iterations. To keep things simple, we will only do a set number of # iterations, instead of using a Stopping Criterion. SCREAMING_SNAKE_CASE : Tuple = 100 for _ in range(a__ ): ##EXPECTATION STEP ##Based on the centroid locations till last iteration, compute ##the _expected_ centroid assignments. # Iterate over each vector for vector_n in range(len(a__ ) ): SCREAMING_SNAKE_CASE : Union[str, Any] = vectors[vector_n] # Compute Euclidean distance between this vector and each # centroid. Remember that this list cannot be named #'centroid_distances', since that is the input to the # cluster assignment node. SCREAMING_SNAKE_CASE : Optional[int] = [ sess.run(a__ , feed_dict={va: vect, va: sess.run(a__ )} ) for centroid in centroids ] # Now use the cluster assignment node, with the distances # as the input SCREAMING_SNAKE_CASE : Optional[int] = sess.run( a__ , feed_dict={centroid_distances: distances} ) # Now assign the value to the appropriate state variable sess.run( cluster_assigns[vector_n] , feed_dict={assignment_value: assignment} ) ##MAXIMIZATION STEP # Based on the expected state computed from the Expectation Step, # compute the locations of the centroids so as to maximize the # overall objective of minimizing within-cluster Sum-of-Squares for cluster_n in range(a__ ): # Collect all the vectors assigned to this cluster SCREAMING_SNAKE_CASE : str = [ vectors[i] for i in range(len(a__ ) ) if sess.run(assignments[i] ) == cluster_n ] # Compute new centroid location SCREAMING_SNAKE_CASE : List[Any] = sess.run( a__ , feed_dict={mean_input: array(a__ )} ) # Assign value to appropriate variable sess.run( cent_assigns[cluster_n] , feed_dict={centroid_value: new_location} ) # Return centroids and assignments SCREAMING_SNAKE_CASE : Optional[Any] = sess.run(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = sess.run(a__ ) return centroids, assignments
313
import argparse import json from typing import List from ltp import LTP from transformers import BertTokenizer def UpperCAmelCase_( a__ ): """simple docstring""" if ( (cp >= 0x4_E00 and cp <= 0x9_FFF) or (cp >= 0x3_400 and cp <= 0x4_DBF) # or (cp >= 0x20_000 and cp <= 0x2A_6DF) # or (cp >= 0x2A_700 and cp <= 0x2B_73F) # or (cp >= 0x2B_740 and cp <= 0x2B_81F) # or (cp >= 0x2B_820 and cp <= 0x2C_EAF) # or (cp >= 0xF_900 and cp <= 0xF_AFF) or (cp >= 0x2F_800 and cp <= 0x2F_A1F) # ): # return True return False def UpperCAmelCase_( a__ ): """simple docstring""" for char in word: SCREAMING_SNAKE_CASE : str = ord(a__ ) if not _is_chinese_char(a__ ): return 0 return 1 def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = set() for token in tokens: SCREAMING_SNAKE_CASE : str = len(a__ ) > 1 and is_chinese(a__ ) if chinese_word: word_set.add(a__ ) SCREAMING_SNAKE_CASE : str = list(a__ ) return word_list def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if not chinese_word_set: return bert_tokens SCREAMING_SNAKE_CASE : List[str] = max([len(a__ ) for w in chinese_word_set] ) SCREAMING_SNAKE_CASE : Tuple = bert_tokens SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = 0, len(a__ ) while start < end: SCREAMING_SNAKE_CASE : Dict = True if is_chinese(bert_word[start] ): SCREAMING_SNAKE_CASE : Optional[int] = min(end - start , a__ ) for i in range(a__ , 1 , -1 ): SCREAMING_SNAKE_CASE : Optional[int] = ''''''.join(bert_word[start : start + i] ) if whole_word in chinese_word_set: for j in range(start + 1 , start + i ): SCREAMING_SNAKE_CASE : Optional[int] = '''##''' + bert_word[j] SCREAMING_SNAKE_CASE : List[str] = start + i SCREAMING_SNAKE_CASE : Optional[Any] = False break if single_word: start += 1 return bert_word def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : Optional[Any] = ltp_tokenizer.seg(lines[i : i + 100] )[0] SCREAMING_SNAKE_CASE : Union[str, Any] = [get_chinese_word(a__ ) for r in res] ltp_res.extend(a__ ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : Any = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : int = bert_tokenizer(lines[i : i + 100] , add_special_tokens=a__ , truncation=a__ , max_length=512 ) bert_res.extend(res['''input_ids'''] ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : int = [] for input_ids, chinese_word in zip(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = [] for id in input_ids: SCREAMING_SNAKE_CASE : List[Any] = bert_tokenizer._convert_id_to_token(a__ ) input_tokens.append(a__ ) SCREAMING_SNAKE_CASE : List[str] = add_sub_symbol(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = [] # We only save pos of chinese subwords start with ##, which mean is part of a whole word. for i, token in enumerate(a__ ): if token[:2] == "##": SCREAMING_SNAKE_CASE : Optional[int] = token[2:] # save chinese tokens' pos if len(a__ ) == 1 and _is_chinese_char(ord(a__ ) ): ref_id.append(a__ ) ref_ids.append(a__ ) assert len(a__ ) == len(a__ ) return ref_ids def UpperCAmelCase_( a__ ): """simple docstring""" with open(args.file_name , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : List[str] = f.readlines() SCREAMING_SNAKE_CASE : Union[str, Any] = [line.strip() for line in data if len(a__ ) > 0 and not line.isspace()] # avoid delimiter like '\u2029' SCREAMING_SNAKE_CASE : List[str] = LTP(args.ltp ) # faster in GPU device SCREAMING_SNAKE_CASE : int = BertTokenizer.from_pretrained(args.bert ) SCREAMING_SNAKE_CASE : int = prepare_ref(a__ , a__ , a__ ) with open(args.save_path , '''w''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : Tuple = [json.dumps(a__ ) + '''\n''' for ref in ref_ids] f.writelines(a__ ) if __name__ == "__main__": a__ : int = argparse.ArgumentParser(description='''prepare_chinese_ref''') parser.add_argument( '''--file_name''', type=str, default='''./resources/chinese-demo.txt''', help='''file need process, same as training data in lm''', ) parser.add_argument( '''--ltp''', type=str, default='''./resources/ltp''', help='''resources for LTP tokenizer, usually a path''' ) parser.add_argument('''--bert''', type=str, default='''./resources/robert''', help='''resources for Bert tokenizer''') parser.add_argument('''--save_path''', type=str, default='''./resources/ref.txt''', help='''path to save res''') a__ : int = parser.parse_args() main(args)
313
1
from __future__ import absolute_import, division, print_function, unicode_literals from torch import nn from torch.nn import CrossEntropyLoss, MSELoss from transformers import RobertaConfig from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward from transformers.models.roberta.modeling_roberta import ( ROBERTA_INPUTS_DOCSTRING, ROBERTA_START_DOCSTRING, RobertaEmbeddings, ) from .modeling_highway_bert import BertPreTrainedModel, DeeBertModel, HighwayException, entropy @add_start_docstrings( 'The RoBERTa Model transformer with early exiting (DeeRoBERTa). ' , a__ , ) class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = RobertaConfig __SCREAMING_SNAKE_CASE : Optional[int] = 'roberta' def __init__( self , _lowerCamelCase ) ->Tuple: super().__init__(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = RobertaEmbeddings(_lowerCamelCase ) self.init_weights() @add_start_docstrings( 'RoBERTa Model (with early exiting - DeeRoBERTa) with a classifier on top,\n also takes care of multi-layer training. ' , a__ , ) class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = RobertaConfig __SCREAMING_SNAKE_CASE : List[Any] = 'roberta' def __init__( self , _lowerCamelCase ) ->Union[str, Any]: super().__init__(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = config.num_labels SCREAMING_SNAKE_CASE : Any = config.num_hidden_layers SCREAMING_SNAKE_CASE : List[Any] = DeeRobertaModel(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.Dropout(config.hidden_dropout_prob ) SCREAMING_SNAKE_CASE : Dict = nn.Linear(config.hidden_size , self.config.num_labels ) @add_start_docstrings_to_model_forward(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=-1 , _lowerCamelCase=False , ) ->Tuple: SCREAMING_SNAKE_CASE : Any = self.num_layers try: SCREAMING_SNAKE_CASE : Optional[int] = self.roberta( _lowerCamelCase , attention_mask=_lowerCamelCase , token_type_ids=_lowerCamelCase , position_ids=_lowerCamelCase , head_mask=_lowerCamelCase , inputs_embeds=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : str = outputs[1] SCREAMING_SNAKE_CASE : Optional[Any] = self.dropout(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = self.classifier(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = (logits,) + outputs[2:] # add hidden states and attention if they are here except HighwayException as e: SCREAMING_SNAKE_CASE : Union[str, Any] = e.message SCREAMING_SNAKE_CASE : Optional[Any] = e.exit_layer SCREAMING_SNAKE_CASE : List[str] = outputs[0] if not self.training: SCREAMING_SNAKE_CASE : Optional[int] = entropy(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = [] SCREAMING_SNAKE_CASE : Optional[Any] = [] if labels is not None: if self.num_labels == 1: # We are doing regression SCREAMING_SNAKE_CASE : Dict = MSELoss() SCREAMING_SNAKE_CASE : int = loss_fct(logits.view(-1 ) , labels.view(-1 ) ) else: SCREAMING_SNAKE_CASE : str = CrossEntropyLoss() SCREAMING_SNAKE_CASE : str = loss_fct(logits.view(-1 , self.num_labels ) , labels.view(-1 ) ) # work with highway exits SCREAMING_SNAKE_CASE : Any = [] for highway_exit in outputs[-1]: SCREAMING_SNAKE_CASE : str = highway_exit[0] if not self.training: highway_logits_all.append(_lowerCamelCase ) highway_entropy.append(highway_exit[2] ) if self.num_labels == 1: # We are doing regression SCREAMING_SNAKE_CASE : List[Any] = MSELoss() SCREAMING_SNAKE_CASE : Dict = loss_fct(highway_logits.view(-1 ) , labels.view(-1 ) ) else: SCREAMING_SNAKE_CASE : Dict = CrossEntropyLoss() SCREAMING_SNAKE_CASE : Union[str, Any] = loss_fct(highway_logits.view(-1 , self.num_labels ) , labels.view(-1 ) ) highway_losses.append(_lowerCamelCase ) if train_highway: SCREAMING_SNAKE_CASE : str = (sum(highway_losses[:-1] ),) + outputs # exclude the final highway, of course else: SCREAMING_SNAKE_CASE : List[str] = (loss,) + outputs if not self.training: SCREAMING_SNAKE_CASE : str = outputs + ((original_entropy, highway_entropy), exit_layer) if output_layer >= 0: SCREAMING_SNAKE_CASE : Any = ( (outputs[0],) + (highway_logits_all[output_layer],) + outputs[2:] ) # use the highway of the last layer return outputs # (loss), logits, (hidden_states), (attentions), entropy
313
import datetime import platform import subprocess from typing import Optional, Tuple, Union import numpy as np def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Tuple = '''1''' SCREAMING_SNAKE_CASE : Union[str, Any] = '''f32le''' SCREAMING_SNAKE_CASE : List[Any] = [ '''ffmpeg''', '''-i''', '''pipe:0''', '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] try: with subprocess.Popen(a__ , stdin=subprocess.PIPE , stdout=subprocess.PIPE ) as ffmpeg_process: SCREAMING_SNAKE_CASE : Tuple = ffmpeg_process.communicate(a__ ) except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to load audio files from filename''' ) from error SCREAMING_SNAKE_CASE : Optional[Any] = output_stream[0] SCREAMING_SNAKE_CASE : Any = np.frombuffer(a__ , np.floataa ) if audio.shape[0] == 0: raise ValueError('''Malformed soundfile''' ) return audio def UpperCAmelCase_( a__ , a__ , a__ = "f32le" , ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Dict = '''1''' if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Dict = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = platform.system() if system == "Linux": SCREAMING_SNAKE_CASE : Dict = '''alsa''' SCREAMING_SNAKE_CASE : Any = '''default''' elif system == "Darwin": SCREAMING_SNAKE_CASE : Union[str, Any] = '''avfoundation''' SCREAMING_SNAKE_CASE : Optional[int] = ''':0''' elif system == "Windows": SCREAMING_SNAKE_CASE : int = '''dshow''' SCREAMING_SNAKE_CASE : Any = '''default''' SCREAMING_SNAKE_CASE : Union[str, Any] = [ '''ffmpeg''', '''-f''', format_, '''-i''', input_, '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-fflags''', '''nobuffer''', '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] SCREAMING_SNAKE_CASE : List[str] = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample SCREAMING_SNAKE_CASE : List[Any] = _ffmpeg_stream(a__ , a__ ) for item in iterator: yield item def UpperCAmelCase_( a__ , a__ , a__ = None , a__ = None , a__ = "f32le" , ): """simple docstring""" if stream_chunk_s is not None: SCREAMING_SNAKE_CASE : Tuple = stream_chunk_s else: SCREAMING_SNAKE_CASE : List[str] = chunk_length_s SCREAMING_SNAKE_CASE : Union[str, Any] = ffmpeg_microphone(a__ , a__ , format_for_conversion=a__ ) if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : Optional[int] = np.intaa SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Any = np.floataa SCREAMING_SNAKE_CASE : Union[str, Any] = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) if stride_length_s is None: SCREAMING_SNAKE_CASE : Optional[Any] = chunk_length_s / 6 SCREAMING_SNAKE_CASE : Dict = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample if isinstance(a__ , (int, float) ): SCREAMING_SNAKE_CASE : List[Any] = [stride_length_s, stride_length_s] SCREAMING_SNAKE_CASE : Any = int(round(sampling_rate * stride_length_s[0] ) ) * size_of_sample SCREAMING_SNAKE_CASE : int = int(round(sampling_rate * stride_length_s[1] ) ) * size_of_sample SCREAMING_SNAKE_CASE : Union[str, Any] = datetime.datetime.now() SCREAMING_SNAKE_CASE : Dict = datetime.timedelta(seconds=a__ ) for item in chunk_bytes_iter(a__ , a__ , stride=(stride_left, stride_right) , stream=a__ ): # Put everything back in numpy scale SCREAMING_SNAKE_CASE : Dict = np.frombuffer(item['''raw'''] , dtype=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = ( item['''stride'''][0] // size_of_sample, item['''stride'''][1] // size_of_sample, ) SCREAMING_SNAKE_CASE : Any = sampling_rate audio_time += delta if datetime.datetime.now() > audio_time + 10 * delta: # We're late !! SKIP continue yield item def UpperCAmelCase_( a__ , a__ , a__ , a__ = False ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = b'''''' SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = stride if stride_left + stride_right >= chunk_len: raise ValueError( F"""Stride needs to be strictly smaller than chunk_len: ({stride_left}, {stride_right}) vs {chunk_len}""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = 0 for raw in iterator: acc += raw if stream and len(a__ ) < chunk_len: SCREAMING_SNAKE_CASE : List[str] = (_stride_left, 0) yield {"raw": acc[:chunk_len], "stride": stride, "partial": True} else: while len(a__ ) >= chunk_len: # We are flushing the accumulator SCREAMING_SNAKE_CASE : str = (_stride_left, stride_right) SCREAMING_SNAKE_CASE : List[str] = {'''raw''': acc[:chunk_len], '''stride''': stride} if stream: SCREAMING_SNAKE_CASE : List[str] = False yield item SCREAMING_SNAKE_CASE : Dict = stride_left SCREAMING_SNAKE_CASE : int = acc[chunk_len - stride_left - stride_right :] # Last chunk if len(a__ ) > stride_left: SCREAMING_SNAKE_CASE : Optional[Any] = {'''raw''': acc, '''stride''': (_stride_left, 0)} if stream: SCREAMING_SNAKE_CASE : Union[str, Any] = False yield item def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : str = 2**24 # 16Mo try: with subprocess.Popen(a__ , stdout=subprocess.PIPE , bufsize=a__ ) as ffmpeg_process: while True: SCREAMING_SNAKE_CASE : str = ffmpeg_process.stdout.read(a__ ) if raw == b"": break yield raw except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to stream audio files from filename''' ) from error
313
1
import inspect import unittest from huggingface_hub import hf_hub_download from transformers import ConvNextConfig, UperNetConfig from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device from transformers.utils import is_torch_available, is_vision_available from ...test_configuration_common import ConfigTester from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor, ids_tensor from ...test_pipeline_mixin import PipelineTesterMixin if is_torch_available(): import torch from transformers import UperNetForSemanticSegmentation from transformers.models.upernet.modeling_upernet import UPERNET_PRETRAINED_MODEL_ARCHIVE_LIST if is_vision_available(): from PIL import Image from transformers import AutoImageProcessor class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=13 , _lowerCamelCase=32 , _lowerCamelCase=3 , _lowerCamelCase=4 , _lowerCamelCase=[10, 20, 30, 40] , _lowerCamelCase=[2, 2, 3, 2] , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=37 , _lowerCamelCase="gelu" , _lowerCamelCase=10 , _lowerCamelCase=0.0_2 , _lowerCamelCase=["stage2", "stage3", "stage4"] , _lowerCamelCase=3 , _lowerCamelCase=None , ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Dict = parent SCREAMING_SNAKE_CASE : Optional[Any] = batch_size SCREAMING_SNAKE_CASE : List[Any] = image_size SCREAMING_SNAKE_CASE : Union[str, Any] = num_channels SCREAMING_SNAKE_CASE : Optional[int] = num_stages SCREAMING_SNAKE_CASE : int = hidden_sizes SCREAMING_SNAKE_CASE : str = depths SCREAMING_SNAKE_CASE : Tuple = is_training SCREAMING_SNAKE_CASE : int = use_labels SCREAMING_SNAKE_CASE : int = intermediate_size SCREAMING_SNAKE_CASE : List[str] = hidden_act SCREAMING_SNAKE_CASE : List[Any] = type_sequence_label_size SCREAMING_SNAKE_CASE : Any = initializer_range SCREAMING_SNAKE_CASE : Any = out_features SCREAMING_SNAKE_CASE : Tuple = num_labels SCREAMING_SNAKE_CASE : int = scope SCREAMING_SNAKE_CASE : List[str] = num_stages def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Optional[Any] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : Dict = None if self.use_labels: SCREAMING_SNAKE_CASE : int = ids_tensor([self.batch_size] , self.type_sequence_label_size ) SCREAMING_SNAKE_CASE : Optional[int] = self.get_config() return config, pixel_values, labels def __lowerCAmelCase ( self ) ->Dict: return ConvNextConfig( num_channels=self.num_channels , num_stages=self.num_stages , hidden_sizes=self.hidden_sizes , depths=self.depths , is_training=self.is_training , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , out_features=self.out_features , ) def __lowerCAmelCase ( self ) ->List[str]: return UperNetConfig( backbone_config=self.get_backbone_config() , hidden_size=512 , pool_scales=[1, 2, 3, 6] , use_auxiliary_head=_lowerCamelCase , auxiliary_loss_weight=0.4 , auxiliary_in_channels=40 , auxiliary_channels=256 , auxiliary_num_convs=1 , auxiliary_concat_input=_lowerCamelCase , loss_ignore_index=255 , num_labels=self.num_labels , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = UperNetForSemanticSegmentation(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : int = model(_lowerCamelCase ) self.parent.assertEqual( result.logits.shape , (self.batch_size, self.num_labels, self.image_size, self.image_size) ) def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : List[Any] = self.prepare_config_and_inputs() ( ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ( SCREAMING_SNAKE_CASE ) , ) : Dict = config_and_inputs SCREAMING_SNAKE_CASE : int = {'''pixel_values''': pixel_values} return config, inputs_dict @require_torch class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Dict = (UperNetForSemanticSegmentation,) if is_torch_available() else () __SCREAMING_SNAKE_CASE : List[str] = {'image-segmentation': UperNetForSemanticSegmentation} if is_torch_available() else {} __SCREAMING_SNAKE_CASE : Union[str, Any] = False __SCREAMING_SNAKE_CASE : Dict = False __SCREAMING_SNAKE_CASE : Tuple = False __SCREAMING_SNAKE_CASE : Dict = False __SCREAMING_SNAKE_CASE : Any = False __SCREAMING_SNAKE_CASE : int = False def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : List[Any] = UperNetModelTester(self ) SCREAMING_SNAKE_CASE : List[str] = ConfigTester(self , config_class=_lowerCamelCase , has_text_modality=_lowerCamelCase , hidden_size=37 ) def __lowerCAmelCase ( self ) ->int: 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 __lowerCAmelCase ( self ) ->Optional[Any]: return def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : List[Any] = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = inspect.signature(model.forward ) # signature.parameters is an OrderedDict => so arg_names order is deterministic SCREAMING_SNAKE_CASE : List[str] = [*signature.parameters.keys()] SCREAMING_SNAKE_CASE : Dict = ['''pixel_values'''] self.assertListEqual(arg_names[:1] , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : int = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_semantic_segmentation(*_lowerCamelCase ) @unittest.skip(reason='''UperNet does not use inputs_embeds''' ) def __lowerCAmelCase ( self ) ->Tuple: pass @unittest.skip(reason='''UperNet does not support input and output embeddings''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @unittest.skip(reason='''UperNet does not have a base model''' ) def __lowerCAmelCase ( self ) ->int: pass @unittest.skip(reason='''UperNet does not have a base model''' ) def __lowerCAmelCase ( self ) ->Optional[Any]: pass @require_torch_multi_gpu @unittest.skip(reason='''UperNet has some layers using `add_module` which doesn\'t work well with `nn.DataParallel`''' ) def __lowerCAmelCase ( self ) ->str: pass @unittest.skip('''Will be fixed soon by reducing the size of the model used for common tests.''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: pass def __lowerCAmelCase ( self ) ->Union[str, Any]: def check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : Optional[Any] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : int = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = outputs.encoder_hidden_states if config.is_encoder_decoder else outputs.hidden_states SCREAMING_SNAKE_CASE : Union[str, Any] = self.model_tester.num_stages self.assertEqual(len(_lowerCamelCase ) , 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] , ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : List[Any] = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # check that output_hidden_states also work using config del inputs_dict["output_hidden_states"] SCREAMING_SNAKE_CASE : Union[str, Any] = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : Dict = _config_zero_init(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = _config_zero_init(configs_no_init.backbone_config ) for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Optional[Any] = model_class(config=_lowerCamelCase ) for name, param in model.named_parameters(): if param.requires_grad: self.assertIn( ((param.data.mean() * 1e9).round() / 1e9).item() , [0.0, 1.0] , msg=F"""Parameter {name} of model {model_class} seems not properly initialized""" , ) @unittest.skip(reason='''UperNet does not have tied weights''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @slow def __lowerCAmelCase ( self ) ->Union[str, Any]: for model_name in UPERNET_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: SCREAMING_SNAKE_CASE : Dict = UperNetForSemanticSegmentation.from_pretrained(_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = hf_hub_download( repo_id='''hf-internal-testing/fixtures_ade20k''' , repo_type='''dataset''' , filename='''ADE_val_00000001.jpg''' ) SCREAMING_SNAKE_CASE : Tuple = Image.open(a__ ).convert('''RGB''' ) return image @require_torch @require_vision @slow class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Optional[int] = AutoImageProcessor.from_pretrained('''openmmlab/upernet-swin-tiny''' ) SCREAMING_SNAKE_CASE : Optional[Any] = UperNetForSemanticSegmentation.from_pretrained('''openmmlab/upernet-swin-tiny''' ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = prepare_img() SCREAMING_SNAKE_CASE : Any = processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) with torch.no_grad(): SCREAMING_SNAKE_CASE : Dict = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = torch.Size((1, model.config.num_labels, 512, 512) ) self.assertEqual(outputs.logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = torch.tensor( [[-7.5_9_5_8, -7.5_9_5_8, -7.4_3_0_2], [-7.5_9_5_8, -7.5_9_5_8, -7.4_3_0_2], [-7.4_7_9_7, -7.4_7_9_7, -7.3_0_6_8]] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(outputs.logits[0, 0, :3, :3] , _lowerCamelCase , atol=1e-4 ) ) def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Tuple = AutoImageProcessor.from_pretrained('''openmmlab/upernet-convnext-tiny''' ) SCREAMING_SNAKE_CASE : int = UperNetForSemanticSegmentation.from_pretrained('''openmmlab/upernet-convnext-tiny''' ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = prepare_img() SCREAMING_SNAKE_CASE : Dict = processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) with torch.no_grad(): SCREAMING_SNAKE_CASE : List[str] = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = torch.Size((1, model.config.num_labels, 512, 512) ) self.assertEqual(outputs.logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.tensor( [[-8.8_1_1_0, -8.8_1_1_0, -8.6_5_2_1], [-8.8_1_1_0, -8.8_1_1_0, -8.6_5_2_1], [-8.7_7_4_6, -8.7_7_4_6, -8.6_1_3_0]] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(outputs.logits[0, 0, :3, :3] , _lowerCamelCase , atol=1e-4 ) )
313
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a__ : Tuple = {'''configuration_wavlm''': ['''WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''WavLMConfig''']} try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Dict = [ '''WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST''', '''WavLMForAudioFrameClassification''', '''WavLMForCTC''', '''WavLMForSequenceClassification''', '''WavLMForXVector''', '''WavLMModel''', '''WavLMPreTrainedModel''', ] if TYPE_CHECKING: from .configuration_wavlm import WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP, WavLMConfig try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_wavlm import ( WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST, WavLMForAudioFrameClassification, WavLMForCTC, WavLMForSequenceClassification, WavLMForXVector, WavLMModel, WavLMPreTrainedModel, ) else: import sys a__ : Dict = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
1
import argparse import json import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ViTImageProcessor, ViTMSNConfig, ViTMSNModel from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD torch.set_grad_enabled(False) def UpperCAmelCase_( a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [] for i in range(config.num_hidden_layers ): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append((F"""module.blocks.{i}.norm1.weight""", F"""vit.encoder.layer.{i}.layernorm_before.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm1.bias""", F"""vit.encoder.layer.{i}.layernorm_before.bias""") ) rename_keys.append( (F"""module.blocks.{i}.attn.proj.weight""", F"""vit.encoder.layer.{i}.attention.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.attn.proj.bias""", F"""vit.encoder.layer.{i}.attention.output.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.norm2.weight""", F"""vit.encoder.layer.{i}.layernorm_after.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm2.bias""", F"""vit.encoder.layer.{i}.layernorm_after.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.weight""", F"""vit.encoder.layer.{i}.intermediate.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.bias""", F"""vit.encoder.layer.{i}.intermediate.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.weight""", F"""vit.encoder.layer.{i}.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.bias""", F"""vit.encoder.layer.{i}.output.dense.bias""") ) # projection layer + position embeddings rename_keys.extend( [ ('''module.cls_token''', '''vit.embeddings.cls_token'''), ('''module.patch_embed.proj.weight''', '''vit.embeddings.patch_embeddings.projection.weight'''), ('''module.patch_embed.proj.bias''', '''vit.embeddings.patch_embeddings.projection.bias'''), ('''module.pos_embed''', '''vit.embeddings.position_embeddings'''), ] ) if base_model: # layernorm + pooler rename_keys.extend( [ ('''module.norm.weight''', '''layernorm.weight'''), ('''module.norm.bias''', '''layernorm.bias'''), ] ) # if just the base model, we should remove "vit" from all keys that start with "vit" SCREAMING_SNAKE_CASE : Any = [(pair[0], pair[1][4:]) if pair[1].startswith('''vit''' ) else pair for pair in rename_keys] else: # layernorm + classification head rename_keys.extend( [ ('''norm.weight''', '''vit.layernorm.weight'''), ('''norm.bias''', '''vit.layernorm.bias'''), ('''head.weight''', '''classifier.weight'''), ('''head.bias''', '''classifier.bias'''), ] ) return rename_keys def UpperCAmelCase_( a__ , a__ , a__=False ): """simple docstring""" for i in range(config.num_hidden_layers ): if base_model: SCREAMING_SNAKE_CASE : Any = '''''' else: SCREAMING_SNAKE_CASE : Optional[int] = '''vit.''' # read in weights + bias of input projection layer (in timm, this is a single matrix + bias) SCREAMING_SNAKE_CASE : Optional[Any] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.weight""" ) SCREAMING_SNAKE_CASE : List[str] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.bias""" ) # next, add query, keys and values (in that order) to the state dict SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ : config.hidden_size, : ] SCREAMING_SNAKE_CASE : Any = in_proj_bias[: config.hidden_size] SCREAMING_SNAKE_CASE : int = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] SCREAMING_SNAKE_CASE : List[Any] = in_proj_bias[ config.hidden_size : config.hidden_size * 2 ] SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ -config.hidden_size :, : ] SCREAMING_SNAKE_CASE : Optional[Any] = in_proj_bias[-config.hidden_size :] def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = ['''head.weight''', '''head.bias'''] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = [ '''module.fc.fc1.weight''', '''module.fc.fc1.bias''', '''module.fc.bn1.weight''', '''module.fc.bn1.bias''', '''module.fc.bn1.running_mean''', '''module.fc.bn1.running_var''', '''module.fc.bn1.num_batches_tracked''', '''module.fc.fc2.weight''', '''module.fc.fc2.bias''', '''module.fc.bn2.weight''', '''module.fc.bn2.bias''', '''module.fc.bn2.running_mean''', '''module.fc.bn2.running_var''', '''module.fc.bn2.num_batches_tracked''', '''module.fc.fc3.weight''', '''module.fc.fc3.bias''', ] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = dct.pop(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = val def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = ViTMSNConfig() SCREAMING_SNAKE_CASE : Optional[int] = 1_000 SCREAMING_SNAKE_CASE : str = '''datasets/huggingface/label-files''' SCREAMING_SNAKE_CASE : List[str] = '''imagenet-1k-id2label.json''' SCREAMING_SNAKE_CASE : Union[str, Any] = json.load(open(hf_hub_download(a__ , a__ ) , '''r''' ) ) SCREAMING_SNAKE_CASE : List[Any] = {int(a__ ): v for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : str = idalabel SCREAMING_SNAKE_CASE : int = {v: k for k, v in idalabel.items()} if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = 384 SCREAMING_SNAKE_CASE : Any = 1_536 SCREAMING_SNAKE_CASE : List[str] = 6 elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Optional[int] = 1_024 SCREAMING_SNAKE_CASE : Optional[int] = 4_096 SCREAMING_SNAKE_CASE : Tuple = 24 SCREAMING_SNAKE_CASE : Union[str, Any] = 16 SCREAMING_SNAKE_CASE : Dict = 0.1 elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = 4 elif "l7" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = 7 SCREAMING_SNAKE_CASE : Union[str, Any] = 1_024 SCREAMING_SNAKE_CASE : List[Any] = 4_096 SCREAMING_SNAKE_CASE : List[Any] = 24 SCREAMING_SNAKE_CASE : Tuple = 16 SCREAMING_SNAKE_CASE : Union[str, Any] = 0.1 SCREAMING_SNAKE_CASE : Union[str, Any] = ViTMSNModel(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = torch.hub.load_state_dict_from_url(a__ , map_location='''cpu''' )['''target_encoder'''] SCREAMING_SNAKE_CASE : Any = ViTImageProcessor(size=config.image_size ) remove_projection_head(a__ ) SCREAMING_SNAKE_CASE : Any = create_rename_keys(a__ , base_model=a__ ) for src, dest in rename_keys: rename_key(a__ , a__ , a__ ) read_in_q_k_v(a__ , a__ , base_model=a__ ) model.load_state_dict(a__ ) model.eval() SCREAMING_SNAKE_CASE : Any = '''http://images.cocodataset.org/val2017/000000039769.jpg''' SCREAMING_SNAKE_CASE : Dict = Image.open(requests.get(a__ , stream=a__ ).raw ) SCREAMING_SNAKE_CASE : Optional[int] = ViTImageProcessor( size=config.image_size , image_mean=a__ , image_std=a__ ) SCREAMING_SNAKE_CASE : int = image_processor(images=a__ , return_tensors='''pt''' ) # forward pass torch.manual_seed(2 ) SCREAMING_SNAKE_CASE : Tuple = model(**a__ ) SCREAMING_SNAKE_CASE : str = outputs.last_hidden_state # The following Colab Notebook was used to generate these outputs: # https://colab.research.google.com/gist/sayakpaul/3672419a04f5997827503fd84079bdd1/scratchpad.ipynb if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[-1.0_915, -1.4_876, -1.1_809]] ) elif "b16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[14.2_889, -18.9_045, 11.7_281]] ) elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.tensor([[41.5_028, -22.8_681, 45.6_475]] ) elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = torch.tensor([[-4.3_868, 5.2_932, -0.4_137]] ) else: SCREAMING_SNAKE_CASE : List[str] = torch.tensor([[-0.1_792, -0.6_465, 2.4_263]] ) # verify logits assert torch.allclose(last_hidden_state[:, 0, :3] , a__ , atol=1e-4 ) print(F"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(a__ ) print(F"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(a__ ) if __name__ == "__main__": a__ : Dict = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--checkpoint_url''', default='''https://dl.fbaipublicfiles.com/msn/vits16_800ep.pth.tar''', type=str, help='''URL of the checkpoint you\'d like to convert.''', ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.''' ) a__ : Any = parser.parse_args() convert_vit_msn_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
313
import copy from ...configuration_utils import PretrainedConfig from ...utils import logging from ..auto import CONFIG_MAPPING a__ : int = logging.get_logger(__name__) a__ : Optional[Any] = { '''SenseTime/deformable-detr''': '''https://huggingface.co/sensetime/deformable-detr/resolve/main/config.json''', # See all Deformable DETR models at https://huggingface.co/models?filter=deformable-detr } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'deformable_detr' __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'd_model', 'num_attention_heads': 'encoder_attention_heads', } def __init__( self , _lowerCamelCase=True , _lowerCamelCase=None , _lowerCamelCase=3 , _lowerCamelCase=300 , _lowerCamelCase=1024 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=0.0 , _lowerCamelCase=True , _lowerCamelCase="relu" , _lowerCamelCase=256 , _lowerCamelCase=0.1 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1.0 , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase="sine" , _lowerCamelCase="resnet50" , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=False , _lowerCamelCase=300 , _lowerCamelCase=False , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=1 , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=0.1 , _lowerCamelCase=0.2_5 , _lowerCamelCase=False , **_lowerCamelCase , ) ->Optional[Any]: if backbone_config is not None and use_timm_backbone: raise ValueError('''You can\'t specify both `backbone_config` and `use_timm_backbone`.''' ) if not use_timm_backbone: if backbone_config is None: logger.info('''`backbone_config` is `None`. Initializing the config with the default `ResNet` backbone.''' ) SCREAMING_SNAKE_CASE : Dict = CONFIG_MAPPING['''resnet'''](out_features=['''stage4'''] ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : List[Any] = backbone_config.get('''model_type''' ) SCREAMING_SNAKE_CASE : Optional[Any] = CONFIG_MAPPING[backbone_model_type] SCREAMING_SNAKE_CASE : int = config_class.from_dict(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = use_timm_backbone SCREAMING_SNAKE_CASE : Optional[int] = backbone_config SCREAMING_SNAKE_CASE : Union[str, Any] = num_channels SCREAMING_SNAKE_CASE : Optional[Any] = num_queries SCREAMING_SNAKE_CASE : Optional[int] = max_position_embeddings SCREAMING_SNAKE_CASE : Optional[int] = d_model SCREAMING_SNAKE_CASE : str = encoder_ffn_dim SCREAMING_SNAKE_CASE : str = encoder_layers SCREAMING_SNAKE_CASE : str = encoder_attention_heads SCREAMING_SNAKE_CASE : Optional[int] = decoder_ffn_dim SCREAMING_SNAKE_CASE : int = decoder_layers SCREAMING_SNAKE_CASE : Union[str, Any] = decoder_attention_heads SCREAMING_SNAKE_CASE : List[str] = dropout SCREAMING_SNAKE_CASE : Optional[int] = attention_dropout SCREAMING_SNAKE_CASE : str = activation_dropout SCREAMING_SNAKE_CASE : Optional[int] = activation_function SCREAMING_SNAKE_CASE : Optional[int] = init_std SCREAMING_SNAKE_CASE : List[str] = init_xavier_std SCREAMING_SNAKE_CASE : Optional[Any] = encoder_layerdrop SCREAMING_SNAKE_CASE : Union[str, Any] = auxiliary_loss SCREAMING_SNAKE_CASE : List[Any] = position_embedding_type SCREAMING_SNAKE_CASE : str = backbone SCREAMING_SNAKE_CASE : Dict = use_pretrained_backbone SCREAMING_SNAKE_CASE : Dict = dilation # deformable attributes SCREAMING_SNAKE_CASE : str = num_feature_levels SCREAMING_SNAKE_CASE : Optional[Any] = encoder_n_points SCREAMING_SNAKE_CASE : Any = decoder_n_points SCREAMING_SNAKE_CASE : str = two_stage SCREAMING_SNAKE_CASE : List[str] = two_stage_num_proposals SCREAMING_SNAKE_CASE : Dict = with_box_refine if two_stage is True and with_box_refine is False: raise ValueError('''If two_stage is True, with_box_refine must be True.''' ) # Hungarian matcher SCREAMING_SNAKE_CASE : int = class_cost SCREAMING_SNAKE_CASE : Union[str, Any] = bbox_cost SCREAMING_SNAKE_CASE : Optional[int] = giou_cost # Loss coefficients SCREAMING_SNAKE_CASE : Dict = mask_loss_coefficient SCREAMING_SNAKE_CASE : Union[str, Any] = dice_loss_coefficient SCREAMING_SNAKE_CASE : str = bbox_loss_coefficient SCREAMING_SNAKE_CASE : Tuple = giou_loss_coefficient SCREAMING_SNAKE_CASE : Optional[int] = eos_coefficient SCREAMING_SNAKE_CASE : Tuple = focal_alpha SCREAMING_SNAKE_CASE : Optional[int] = disable_custom_kernels super().__init__(is_encoder_decoder=_lowerCamelCase , **_lowerCamelCase ) @property def __lowerCAmelCase ( self ) ->int: return self.encoder_attention_heads @property def __lowerCAmelCase ( self ) ->int: return self.d_model def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : str = copy.deepcopy(self.__dict__ ) if self.backbone_config is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.backbone_config.to_dict() SCREAMING_SNAKE_CASE : Any = self.__class__.model_type return output
313
1
import numpy as np from scipy.spatial.distance import cdist from sklearn.metrics import fa_score import datasets a__ : Any = '''\ @inproceedings{kakwani2020indicnlpsuite, title={{IndicNLPSuite: Monolingual Corpora, Evaluation Benchmarks and Pre-trained Multilingual Language Models for Indian Languages}}, author={Divyanshu Kakwani and Anoop Kunchukuttan and Satish Golla and Gokul N.C. and Avik Bhattacharyya and Mitesh M. Khapra and Pratyush Kumar}, year={2020}, booktitle={Findings of EMNLP}, } ''' a__ : Tuple = '''\ IndicGLUE is a natural language understanding benchmark for Indian languages. It contains a wide variety of tasks and covers 11 major Indian languages - as, bn, gu, hi, kn, ml, mr, or, pa, ta, te. ''' a__ : Union[str, Any] = ''' Compute IndicGLUE evaluation metric associated to each IndicGLUE dataset. Args: predictions: list of predictions to score (as int64), except for \'cvit-mkb-clsr\' where each prediction is a vector (of float32). references: list of ground truth labels corresponding to the predictions (as int64), except for \'cvit-mkb-clsr\' where each reference is a vector (of float32). Returns: depending on the IndicGLUE subset, one or several of: "accuracy": Accuracy "f1": F1 score "precision": Precision@10 Examples: >>> indic_glue_metric = datasets.load_metric(\'indic_glue\', \'wnli\') # \'wnli\' or any of ["copa", "sna", "csqa", "wstp", "inltkh", "bbca", "iitp-mr", "iitp-pr", "actsa-sc", "md"] >>> references = [0, 1] >>> predictions = [0, 1] >>> results = indic_glue_metric.compute(predictions=predictions, references=references) >>> print(results) {\'accuracy\': 1.0} >>> indic_glue_metric = datasets.load_metric(\'indic_glue\', \'wiki-ner\') >>> references = [0, 1] >>> predictions = [0, 1] >>> results = indic_glue_metric.compute(predictions=predictions, references=references) >>> print(results) {\'accuracy\': 1.0, \'f1\': 1.0} >>> indic_glue_metric = datasets.load_metric(\'indic_glue\', \'cvit-mkb-clsr\') >>> references = [[0.5, 0.5, 0.5], [0.1, 0.2, 0.3]] >>> predictions = [[0.5, 0.5, 0.5], [0.1, 0.2, 0.3]] >>> results = indic_glue_metric.compute(predictions=predictions, references=references) >>> print(results) {\'precision@10\': 1.0} ''' def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return float((preds == labels).mean() ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = simple_accuracy(a__ , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = float(fa_score(y_true=a__ , y_pred=a__ ) ) return { "accuracy": acc, "f1": fa, } def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = np.array(a__ ) SCREAMING_SNAKE_CASE : int = np.array(a__ ) SCREAMING_SNAKE_CASE : int = en_sentvecs.shape[0] # mean centering SCREAMING_SNAKE_CASE : int = en_sentvecs - np.mean(a__ , axis=0 ) SCREAMING_SNAKE_CASE : List[Any] = in_sentvecs - np.mean(a__ , axis=0 ) SCREAMING_SNAKE_CASE : int = cdist(a__ , a__ , '''cosine''' ) SCREAMING_SNAKE_CASE : Optional[int] = np.array(range(a__ ) ) SCREAMING_SNAKE_CASE : Tuple = sim.argsort(axis=1 )[:, :10] SCREAMING_SNAKE_CASE : Optional[int] = np.any(preds == actual[:, None] , axis=1 ) return float(matches.mean() ) @datasets.utils.file_utils.add_start_docstrings(_DESCRIPTION , _KWARGS_DESCRIPTION ) class a_ ( datasets.Metric ): """simple docstring""" def __lowerCAmelCase ( self ) ->Dict: if self.config_name not in [ "wnli", "copa", "sna", "csqa", "wstp", "inltkh", "bbca", "cvit-mkb-clsr", "iitp-mr", "iitp-pr", "actsa-sc", "md", "wiki-ner", ]: raise KeyError( '''You should supply a configuration name selected in ''' '''["wnli", "copa", "sna", "csqa", "wstp", "inltkh", "bbca", ''' '''"cvit-mkb-clsr", "iitp-mr", "iitp-pr", "actsa-sc", "md", ''' '''"wiki-ner"]''' ) return datasets.MetricInfo( description=_DESCRIPTION , citation=_CITATION , inputs_description=_KWARGS_DESCRIPTION , features=datasets.Features( { '''predictions''': datasets.Value('''int64''' ) if self.config_name != '''cvit-mkb-clsr''' else datasets.Sequence(datasets.Value('''float32''' ) ), '''references''': datasets.Value('''int64''' ) if self.config_name != '''cvit-mkb-clsr''' else datasets.Sequence(datasets.Value('''float32''' ) ), } ) , codebase_urls=[] , reference_urls=[] , format='''numpy''' if self.config_name != '''cvit-mkb-clsr''' else None , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Tuple: if self.config_name == "cvit-mkb-clsr": return {"precision@10": precision_at_aa(_lowerCamelCase , _lowerCamelCase )} elif self.config_name in ["wiki-ner"]: return acc_and_fa(_lowerCamelCase , _lowerCamelCase ) elif self.config_name in [ "wnli", "copa", "sna", "csqa", "wstp", "inltkh", "bbca", "iitp-mr", "iitp-pr", "actsa-sc", "md", ]: return {"accuracy": simple_accuracy(_lowerCamelCase , _lowerCamelCase )} else: raise KeyError( '''You should supply a configuration name selected in ''' '''["wnli", "copa", "sna", "csqa", "wstp", "inltkh", "bbca", ''' '''"cvit-mkb-clsr", "iitp-mr", "iitp-pr", "actsa-sc", "md", ''' '''"wiki-ner"]''' )
313
import torch from diffusers import EulerDiscreteScheduler from diffusers.utils import torch_device from .test_schedulers import SchedulerCommonTest class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = (EulerDiscreteScheduler,) __SCREAMING_SNAKE_CASE : Optional[int] = 10 def __lowerCAmelCase ( self , **_lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = { '''num_train_timesteps''': 1100, '''beta_start''': 0.0_0_0_1, '''beta_end''': 0.0_2, '''beta_schedule''': '''linear''', } config.update(**_lowerCamelCase ) return config def __lowerCAmelCase ( self ) ->Tuple: for timesteps in [10, 50, 100, 1000]: self.check_over_configs(num_train_timesteps=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: for beta_start, beta_end in zip([0.0_0_0_0_1, 0.0_0_0_1, 0.0_0_1] , [0.0_0_0_2, 0.0_0_2, 0.0_2] ): self.check_over_configs(beta_start=_lowerCamelCase , beta_end=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: for schedule in ["linear", "scaled_linear"]: self.check_over_configs(beta_schedule=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: for prediction_type in ["epsilon", "v_prediction"]: self.check_over_configs(prediction_type=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : str = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : int = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Any = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = self.dummy_model() SCREAMING_SNAKE_CASE : int = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : Any = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[Any] = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_scheduler_config(prediction_type='''v_prediction''' ) SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = self.dummy_model() SCREAMING_SNAKE_CASE : List[str] = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : List[str] = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : str = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = output.prev_sample SCREAMING_SNAKE_CASE : str = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 0.0_0_0_2 ) < 1e-2 assert abs(result_mean.item() - 2.2676e-06 ) < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Dict = self.get_scheduler_config() SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[Any] = self.dummy_model() SCREAMING_SNAKE_CASE : str = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : Optional[Any] = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : Dict = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Dict = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Optional[Any] = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase , use_karras_sigmas=_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_model() SCREAMING_SNAKE_CASE : Dict = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : int = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : List[Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = output.prev_sample SCREAMING_SNAKE_CASE : Optional[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_2_4.5_2_2_9_9_4_9_9_5_1_1_7_1_9 ) < 1e-2 assert abs(result_mean.item() - 0.1_6_2_1_3_9_3_2_6_3_3_3_9_9_9_6_3 ) < 1e-3
313
1
import argparse import json from typing import List from ltp import LTP from transformers import BertTokenizer def UpperCAmelCase_( a__ ): """simple docstring""" if ( (cp >= 0x4_E00 and cp <= 0x9_FFF) or (cp >= 0x3_400 and cp <= 0x4_DBF) # or (cp >= 0x20_000 and cp <= 0x2A_6DF) # or (cp >= 0x2A_700 and cp <= 0x2B_73F) # or (cp >= 0x2B_740 and cp <= 0x2B_81F) # or (cp >= 0x2B_820 and cp <= 0x2C_EAF) # or (cp >= 0xF_900 and cp <= 0xF_AFF) or (cp >= 0x2F_800 and cp <= 0x2F_A1F) # ): # return True return False def UpperCAmelCase_( a__ ): """simple docstring""" for char in word: SCREAMING_SNAKE_CASE : str = ord(a__ ) if not _is_chinese_char(a__ ): return 0 return 1 def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = set() for token in tokens: SCREAMING_SNAKE_CASE : str = len(a__ ) > 1 and is_chinese(a__ ) if chinese_word: word_set.add(a__ ) SCREAMING_SNAKE_CASE : str = list(a__ ) return word_list def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if not chinese_word_set: return bert_tokens SCREAMING_SNAKE_CASE : List[str] = max([len(a__ ) for w in chinese_word_set] ) SCREAMING_SNAKE_CASE : Tuple = bert_tokens SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = 0, len(a__ ) while start < end: SCREAMING_SNAKE_CASE : Dict = True if is_chinese(bert_word[start] ): SCREAMING_SNAKE_CASE : Optional[int] = min(end - start , a__ ) for i in range(a__ , 1 , -1 ): SCREAMING_SNAKE_CASE : Optional[int] = ''''''.join(bert_word[start : start + i] ) if whole_word in chinese_word_set: for j in range(start + 1 , start + i ): SCREAMING_SNAKE_CASE : Optional[int] = '''##''' + bert_word[j] SCREAMING_SNAKE_CASE : List[str] = start + i SCREAMING_SNAKE_CASE : Optional[Any] = False break if single_word: start += 1 return bert_word def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : Optional[Any] = ltp_tokenizer.seg(lines[i : i + 100] )[0] SCREAMING_SNAKE_CASE : Union[str, Any] = [get_chinese_word(a__ ) for r in res] ltp_res.extend(a__ ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : Any = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : int = bert_tokenizer(lines[i : i + 100] , add_special_tokens=a__ , truncation=a__ , max_length=512 ) bert_res.extend(res['''input_ids'''] ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : int = [] for input_ids, chinese_word in zip(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = [] for id in input_ids: SCREAMING_SNAKE_CASE : List[Any] = bert_tokenizer._convert_id_to_token(a__ ) input_tokens.append(a__ ) SCREAMING_SNAKE_CASE : List[str] = add_sub_symbol(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = [] # We only save pos of chinese subwords start with ##, which mean is part of a whole word. for i, token in enumerate(a__ ): if token[:2] == "##": SCREAMING_SNAKE_CASE : Optional[int] = token[2:] # save chinese tokens' pos if len(a__ ) == 1 and _is_chinese_char(ord(a__ ) ): ref_id.append(a__ ) ref_ids.append(a__ ) assert len(a__ ) == len(a__ ) return ref_ids def UpperCAmelCase_( a__ ): """simple docstring""" with open(args.file_name , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : List[str] = f.readlines() SCREAMING_SNAKE_CASE : Union[str, Any] = [line.strip() for line in data if len(a__ ) > 0 and not line.isspace()] # avoid delimiter like '\u2029' SCREAMING_SNAKE_CASE : List[str] = LTP(args.ltp ) # faster in GPU device SCREAMING_SNAKE_CASE : int = BertTokenizer.from_pretrained(args.bert ) SCREAMING_SNAKE_CASE : int = prepare_ref(a__ , a__ , a__ ) with open(args.save_path , '''w''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : Tuple = [json.dumps(a__ ) + '''\n''' for ref in ref_ids] f.writelines(a__ ) if __name__ == "__main__": a__ : int = argparse.ArgumentParser(description='''prepare_chinese_ref''') parser.add_argument( '''--file_name''', type=str, default='''./resources/chinese-demo.txt''', help='''file need process, same as training data in lm''', ) parser.add_argument( '''--ltp''', type=str, default='''./resources/ltp''', help='''resources for LTP tokenizer, usually a path''' ) parser.add_argument('''--bert''', type=str, default='''./resources/robert''', help='''resources for Bert tokenizer''') parser.add_argument('''--save_path''', type=str, default='''./resources/ref.txt''', help='''path to save res''') a__ : int = parser.parse_args() main(args)
313
import json from typing import Dict, List, Optional, Tuple, Union from tokenizers import pre_tokenizers, processors from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput from ...tokenization_utils_fast import PreTrainedTokenizerFast from ...utils import PaddingStrategy, logging from .tokenization_led import LEDTokenizer a__ : Dict = logging.get_logger(__name__) a__ : Dict = {'''vocab_file''': '''vocab.json''', '''merges_file''': '''merges.txt''', '''tokenizer_file''': '''tokenizer.json'''} a__ : str = { '''vocab_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json''', }, '''merges_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt''', }, '''tokenizer_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json''', }, } a__ : Optional[int] = { '''allenai/led-base-16384''': 16_384, } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = VOCAB_FILES_NAMES __SCREAMING_SNAKE_CASE : Tuple = PRETRAINED_VOCAB_FILES_MAP __SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __SCREAMING_SNAKE_CASE : Union[str, Any] = LEDTokenizer __SCREAMING_SNAKE_CASE : Optional[int] = ['input_ids', 'attention_mask'] def __init__( self , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase="replace" , _lowerCamelCase="<s>" , _lowerCamelCase="</s>" , _lowerCamelCase="</s>" , _lowerCamelCase="<s>" , _lowerCamelCase="<unk>" , _lowerCamelCase="<pad>" , _lowerCamelCase="<mask>" , _lowerCamelCase=False , _lowerCamelCase=True , **_lowerCamelCase , ) ->Union[str, Any]: super().__init__( _lowerCamelCase , _lowerCamelCase , tokenizer_file=_lowerCamelCase , errors=_lowerCamelCase , bos_token=_lowerCamelCase , eos_token=_lowerCamelCase , sep_token=_lowerCamelCase , cls_token=_lowerCamelCase , unk_token=_lowerCamelCase , pad_token=_lowerCamelCase , mask_token=_lowerCamelCase , add_prefix_space=_lowerCamelCase , trim_offsets=_lowerCamelCase , **_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Dict = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() ) if pre_tok_state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : str = getattr(_lowerCamelCase , pre_tok_state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : Optional[int] = add_prefix_space SCREAMING_SNAKE_CASE : str = pre_tok_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = add_prefix_space # the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__` SCREAMING_SNAKE_CASE : List[Any] = '''post_processor''' SCREAMING_SNAKE_CASE : int = getattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) if tokenizer_component_instance: SCREAMING_SNAKE_CASE : Any = json.loads(tokenizer_component_instance.__getstate__() ) # The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class` if "sep" in state: SCREAMING_SNAKE_CASE : Optional[int] = tuple(state['''sep'''] ) if "cls" in state: SCREAMING_SNAKE_CASE : Optional[Any] = tuple(state['''cls'''] ) SCREAMING_SNAKE_CASE : Any = False if state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : Union[str, Any] = add_prefix_space SCREAMING_SNAKE_CASE : Union[str, Any] = True if state.get('''trim_offsets''' , _lowerCamelCase ) != trim_offsets: SCREAMING_SNAKE_CASE : List[Any] = trim_offsets SCREAMING_SNAKE_CASE : Union[str, Any] = True if changes_to_apply: SCREAMING_SNAKE_CASE : List[str] = getattr(_lowerCamelCase , state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : List[Any] = component_class(**_lowerCamelCase ) setattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) @property # Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED def __lowerCAmelCase ( self ) ->str: if self._mask_token is None: if self.verbose: logger.error('''Using mask_token, but it is not set yet.''' ) return None return str(self._mask_token ) @mask_token.setter def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : str = AddedToken(_lowerCamelCase , lstrip=_lowerCamelCase , rstrip=_lowerCamelCase ) if isinstance(_lowerCamelCase , _lowerCamelCase ) else value SCREAMING_SNAKE_CASE : List[Any] = value def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : Tuple = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._batch_encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : List[Any] = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->Tuple[str]: SCREAMING_SNAKE_CASE : Any = self._tokenizer.model.save(_lowerCamelCase , name=_lowerCamelCase ) return tuple(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=None ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id] if token_ids_a is None: return output return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id] def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->List[int]: SCREAMING_SNAKE_CASE : Any = [self.sep_token_id] SCREAMING_SNAKE_CASE : List[str] = [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 __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None , _lowerCamelCase = PaddingStrategy.DO_NOT_PAD , _lowerCamelCase = None , _lowerCamelCase = None , ) ->dict: SCREAMING_SNAKE_CASE : Tuple = super()._pad( encoded_inputs=_lowerCamelCase , max_length=_lowerCamelCase , padding_strategy=_lowerCamelCase , pad_to_multiple_of=_lowerCamelCase , return_attention_mask=_lowerCamelCase , ) # Load from model defaults if return_attention_mask is None: SCREAMING_SNAKE_CASE : Optional[Any] = '''attention_mask''' in self.model_input_names if return_attention_mask and "global_attention_mask" in encoded_inputs: SCREAMING_SNAKE_CASE : int = encoded_inputs[self.model_input_names[0]] # `global_attention_mask` need to have the same length as other (sequential) inputs. SCREAMING_SNAKE_CASE : Tuple = len(encoded_inputs['''global_attention_mask'''] ) != len(_lowerCamelCase ) if needs_to_be_padded: SCREAMING_SNAKE_CASE : int = len(_lowerCamelCase ) - len(encoded_inputs['''global_attention_mask'''] ) if self.padding_side == "right": # Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend` SCREAMING_SNAKE_CASE : str = ( encoded_inputs['''global_attention_mask'''] + [-1] * difference ) elif self.padding_side == "left": SCREAMING_SNAKE_CASE : Optional[Any] = [-1] * difference + encoded_inputs[ '''global_attention_mask''' ] else: raise ValueError('''Invalid padding strategy:''' + str(self.padding_side ) ) return encoded_inputs
313
1
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 a__ : Any = logging.get_logger(__name__) # pylint: disable=invalid-name class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , ) ->Optional[int]: super().__init__() if hasattr(scheduler.config , '''steps_offset''' ) and scheduler.config.steps_offset != 1: SCREAMING_SNAKE_CASE : Optional[Any] = ( 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''' , _lowerCamelCase , standard_warn=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = dict(scheduler.config ) SCREAMING_SNAKE_CASE : Any = 1 SCREAMING_SNAKE_CASE : Union[str, Any] = FrozenDict(_lowerCamelCase ) if hasattr(scheduler.config , '''skip_prk_steps''' ) and scheduler.config.skip_prk_steps is False: SCREAMING_SNAKE_CASE : str = ( 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''' , _lowerCamelCase , standard_warn=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = dict(scheduler.config ) SCREAMING_SNAKE_CASE : Union[str, Any] = True SCREAMING_SNAKE_CASE : Optional[int] = FrozenDict(_lowerCamelCase ) 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=_lowerCamelCase , segmentation_processor=_lowerCamelCase , vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , unet=_lowerCamelCase , scheduler=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=_lowerCamelCase , ) def __lowerCAmelCase ( self , _lowerCamelCase = "auto" ) ->Dict: if slice_size == "auto": # half the attention head size is usually a good trade-off between # speed and memory SCREAMING_SNAKE_CASE : Any = self.unet.config.attention_head_dim // 2 self.unet.set_attention_slice(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: self.enable_attention_slicing(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: if is_accelerate_available(): from accelerate import cpu_offload else: raise ImportError('''Please install accelerate via `pip install accelerate`''' ) SCREAMING_SNAKE_CASE : int = 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(_lowerCamelCase , _lowerCamelCase ) @property # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device def __lowerCAmelCase ( self ) ->List[str]: if self.device != torch.device('''meta''' ) or not hasattr(self.unet , '''_hf_hook''' ): return self.device for module in self.unet.modules(): if ( hasattr(_lowerCamelCase , '''_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 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->List[Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = self.segmentation_processor( text=[text] , images=[image] , padding='''max_length''' , return_tensors='''pt''' ).to(self.device ) SCREAMING_SNAKE_CASE : Any = self.segmentation_model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = torch.sigmoid(outputs.logits ).cpu().detach().unsqueeze(-1 ).numpy() SCREAMING_SNAKE_CASE : List[str] = self.numpy_to_pil(_lowerCamelCase )[0].resize(image.size ) # Run inpainting pipeline with the generated mask SCREAMING_SNAKE_CASE : str = 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=_lowerCamelCase , image=_lowerCamelCase , mask_image=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , )
313
from __future__ import annotations import math def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if len(a__ ) != 2 or len(a[0] ) != 2 or len(a__ ) != 2 or len(b[0] ) != 2: raise Exception('''Matrices are not 2x2''' ) SCREAMING_SNAKE_CASE : Dict = [ [a[0][0] * b[0][0] + a[0][1] * b[1][0], a[0][0] * b[0][1] + a[0][1] * b[1][1]], [a[1][0] * b[0][0] + a[1][1] * b[1][0], a[1][0] * b[0][1] + a[1][1] * b[1][1]], ] return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] + matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] - matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) % 2 != 0 or len(a[0] ) % 2 != 0: raise Exception('''Odd matrices are not supported!''' ) SCREAMING_SNAKE_CASE : str = len(a__ ) SCREAMING_SNAKE_CASE : Any = matrix_length // 2 SCREAMING_SNAKE_CASE : Tuple = [[a[i][j] for j in range(a__ , a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : Optional[int] = [ [a[i][j] for j in range(a__ , a__ )] for i in range(a__ , a__ ) ] SCREAMING_SNAKE_CASE : Optional[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : List[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ , a__ )] return top_left, top_right, bot_left, bot_right def UpperCAmelCase_( a__ ): """simple docstring""" return len(a__ ), len(matrix[0] ) def UpperCAmelCase_( a__ ): """simple docstring""" print('''\n'''.join(str(a__ ) for line in matrix ) ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ ) == (2, 2): return default_matrix_multiplication(a__ , a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE : Dict = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : int = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Any = actual_strassen(matrix_addition(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = matrix_addition(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = matrix_subtraction(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) # construct the new matrix from our 4 quadrants SCREAMING_SNAKE_CASE : Optional[Any] = [] for i in range(len(a__ ) ): new_matrix.append(top_left[i] + top_right[i] ) for i in range(len(a__ ) ): new_matrix.append(bot_left[i] + bot_right[i] ) return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ )[1] != matrix_dimensions(a__ )[0]: SCREAMING_SNAKE_CASE : Any = ( '''Unable to multiply these matrices, please check the dimensions.\n''' F"""Matrix A: {matrixa}\n""" F"""Matrix B: {matrixa}""" ) raise Exception(a__ ) SCREAMING_SNAKE_CASE : str = matrix_dimensions(a__ ) SCREAMING_SNAKE_CASE : Tuple = matrix_dimensions(a__ ) if dimensiona[0] == dimensiona[1] and dimensiona[0] == dimensiona[1]: return [matrixa, matrixa] SCREAMING_SNAKE_CASE : str = max(*a__ , *a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = int(math.pow(2 , math.ceil(math.loga(a__ ) ) ) ) SCREAMING_SNAKE_CASE : Optional[int] = matrixa SCREAMING_SNAKE_CASE : Tuple = matrixa # Adding zeros to the matrices so that the arrays dimensions are the same and also # power of 2 for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) SCREAMING_SNAKE_CASE : Optional[Any] = actual_strassen(a__ , a__ ) # Removing the additional zeros for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): final_matrix[i].pop() else: final_matrix.pop() return final_matrix if __name__ == "__main__": a__ : Dict = [ [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 2, 3, 1], ] a__ : Union[str, Any] = [[0, 2, 1, 1], [16, 2, 3, 3], [2, 2, 7, 7], [13, 11, 22, 4]] print(strassen(matrixa, matrixa))
313
1
from ...utils import ( OptionalDependencyNotAvailable, is_torch_available, is_transformers_available, is_transformers_version, ) try: if not (is_transformers_available() and is_torch_available() and is_transformers_version('''>=''', '''4.25.0''')): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: from ...utils.dummy_torch_and_transformers_objects import ( VersatileDiffusionDualGuidedPipeline, VersatileDiffusionImageVariationPipeline, VersatileDiffusionPipeline, VersatileDiffusionTextToImagePipeline, ) else: from .modeling_text_unet import UNetFlatConditionModel from .pipeline_versatile_diffusion import VersatileDiffusionPipeline from .pipeline_versatile_diffusion_dual_guided import VersatileDiffusionDualGuidedPipeline from .pipeline_versatile_diffusion_image_variation import VersatileDiffusionImageVariationPipeline from .pipeline_versatile_diffusion_text_to_image import VersatileDiffusionTextToImagePipeline
313
# We ignore warnings about stepping the scheduler since we step it ourselves during gradient accumulation import warnings from .state import AcceleratorState, GradientState warnings.filterwarnings('''ignore''', category=UserWarning, module='''torch.optim.lr_scheduler''') class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , _lowerCamelCase = False ) ->Any: SCREAMING_SNAKE_CASE : str = scheduler SCREAMING_SNAKE_CASE : List[str] = optimizers if isinstance(_lowerCamelCase , (list, tuple) ) else [optimizers] SCREAMING_SNAKE_CASE : Union[str, Any] = split_batches SCREAMING_SNAKE_CASE : List[Any] = step_with_optimizer SCREAMING_SNAKE_CASE : List[str] = GradientState() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->Optional[Any]: if not self.step_with_optimizer: # No link between scheduler and optimizer -> just step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) return # Otherwise, first make sure the optimizer was stepped. if not self.gradient_state.sync_gradients: if self.gradient_state.adjust_scheduler: self.scheduler._step_count += 1 return for opt in self.optimizers: if opt.step_was_skipped: return if self.split_batches: # Split batches -> the training dataloader batch size is not changed so one step per training step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: # Otherwise the training dataloader batch size was multiplied by `num_processes`, so we need to do # num_processes steps per training step SCREAMING_SNAKE_CASE : List[str] = AcceleratorState().num_processes for _ in range(_lowerCamelCase ): # Special case when using OneCycle and `drop_last` was not used if hasattr(self.scheduler , '''total_steps''' ): if self.scheduler._step_count <= self.scheduler.total_steps: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Union[str, Any]: return self.scheduler.get_last_lr() def __lowerCAmelCase ( self ) ->List[str]: return self.scheduler.state_dict() def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: self.scheduler.load_state_dict(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: return self.scheduler.get_lr() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->List[str]: return self.scheduler.print_lr(*_lowerCamelCase , **_lowerCamelCase )
313
1
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a__ : Tuple = {'''configuration_wavlm''': ['''WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''WavLMConfig''']} try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Dict = [ '''WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST''', '''WavLMForAudioFrameClassification''', '''WavLMForCTC''', '''WavLMForSequenceClassification''', '''WavLMForXVector''', '''WavLMModel''', '''WavLMPreTrainedModel''', ] if TYPE_CHECKING: from .configuration_wavlm import WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP, WavLMConfig try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_wavlm import ( WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST, WavLMForAudioFrameClassification, WavLMForCTC, WavLMForSequenceClassification, WavLMForXVector, WavLMModel, WavLMPreTrainedModel, ) else: import sys a__ : Dict = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
import argparse import os from pathlib import Path from typing import Dict import tensorflow as tf import torch from tqdm import tqdm from transformers import PegasusConfig, PegasusForConditionalGeneration, PegasusTokenizer from transformers.models.pegasus.configuration_pegasus import DEFAULTS, task_specific_params a__ : Optional[Any] = [ # replace left string with right string to get the relevant state_dict key (identical state dict to bart) ['''memory_attention''', '''encoder_attn'''], ['''attention''', '''attn'''], ['''/''', '''.'''], ['''.LayerNorm.gamma''', '''_layer_norm.weight'''], ['''.LayerNorm.beta''', '''_layer_norm.bias'''], ['''r.layer_''', '''r.layers.'''], ['''output_proj''', '''out_proj'''], ['''ffn.dense_1.''', '''fc2.'''], ['''ffn.dense.''', '''fc1.'''], ['''ffn_layer_norm''', '''final_layer_norm'''], ['''kernel''', '''weight'''], ['''encoder_layer_norm.''', '''encoder.layer_norm.'''], ['''decoder_layer_norm.''', '''decoder.layer_norm.'''], ['''embeddings.weights''', '''shared.weight'''], ] def UpperCAmelCase_( a__ ): """simple docstring""" for pegasus_name, hf_name in PATTERNS: SCREAMING_SNAKE_CASE : Union[str, Any] = k.replace(a__ , a__ ) return k def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = DEFAULTS.copy() cfg_kwargs.update(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = PegasusConfig(**a__ ) SCREAMING_SNAKE_CASE : Optional[int] = PegasusForConditionalGeneration(a__ ) SCREAMING_SNAKE_CASE : Dict = torch_model.model.state_dict() SCREAMING_SNAKE_CASE : List[str] = {} for k, v in tf_weights.items(): SCREAMING_SNAKE_CASE : int = rename_state_dict_key(a__ ) if new_k not in sd: raise ValueError(F"""could not find new key {new_k} in state dict. (converted from {k})""" ) if "dense" in k or "proj" in new_k: SCREAMING_SNAKE_CASE : Dict = v.T SCREAMING_SNAKE_CASE : Tuple = torch.tensor(a__ , dtype=sd[new_k].dtype ) assert v.shape == sd[new_k].shape, F"""{new_k}, {k}, {v.shape}, {sd[new_k].shape}""" # make sure embedding.padding_idx is respected SCREAMING_SNAKE_CASE : Tuple = torch.zeros_like(mapping['''shared.weight'''][cfg.pad_token_id + 1] ) SCREAMING_SNAKE_CASE : int = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Union[str, Any] = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Optional[Any] = {k: torch.zeros_like(a__ ) for k, v in sd.items() if k.endswith('''bias''' ) and k not in mapping} mapping.update(**a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = torch_model.model.load_state_dict(a__ , strict=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = [ k for k in missing if k not in ['''encoder.embed_positions.weight''', '''decoder.embed_positions.weight'''] ] assert unexpected_missing == [], F"""no matches found for the following torch keys {unexpected_missing}""" assert extra == [], F"""no matches found for the following tf keys {extra}""" return torch_model def UpperCAmelCase_( a__="./ckpt/aeslc/model.ckpt-32000" ): """simple docstring""" SCREAMING_SNAKE_CASE : str = tf.train.list_variables(a__ ) SCREAMING_SNAKE_CASE : str = {} SCREAMING_SNAKE_CASE : List[Any] = ['''Adafactor''', '''global_step'''] for name, shape in tqdm(a__ , desc='''converting tf checkpoint to dict''' ): SCREAMING_SNAKE_CASE : Union[str, Any] = any(pat in name for pat in ignore_name ) if skip_key: continue SCREAMING_SNAKE_CASE : Dict = tf.train.load_variable(a__ , a__ ) SCREAMING_SNAKE_CASE : Any = array return tf_weights def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = Path(a__ ).parent.name SCREAMING_SNAKE_CASE : Union[str, Any] = task_specific_params[F"""summarization_{dataset}"""]['''max_position_embeddings'''] SCREAMING_SNAKE_CASE : Dict = PegasusTokenizer.from_pretrained('''sshleifer/pegasus''' , model_max_length=a__ ) assert tok.model_max_length == desired_max_model_length tok.save_pretrained(a__ ) # convert model SCREAMING_SNAKE_CASE : Any = get_tf_weights_as_numpy(a__ ) SCREAMING_SNAKE_CASE : List[str] = task_specific_params[F"""summarization_{dataset}"""] if dataset == "large": SCREAMING_SNAKE_CASE : int = task_specific_params SCREAMING_SNAKE_CASE : List[str] = convert_pegasus(a__ , a__ ) torch_model.save_pretrained(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch_model.state_dict() sd.pop('''model.decoder.embed_positions.weight''' ) sd.pop('''model.encoder.embed_positions.weight''' ) torch.save(a__ , Path(a__ ) / '''pytorch_model.bin''' ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument('''tf_ckpt_path''', type=str, help='''passed to tf.train.list_variables''') parser.add_argument('''save_dir''', default=None, type=str, help='''Path to the output PyTorch model.''') a__ : List[str] = parser.parse_args() if args.save_dir is None: a__ : Any = Path(args.tf_ckpt_path).parent.name a__ : int = os.path.join('''pegasus''', dataset) convert_pegasus_ckpt_to_pytorch(args.tf_ckpt_path, args.save_dir)
313
1
import inspect import unittest from datasets import load_dataset from packaging import version from transformers import BeitConfig from transformers.models.auto import get_values from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device from transformers.utils import cached_property, is_torch_available, is_vision_available from ...test_configuration_common import ConfigTester from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor, ids_tensor from ...test_pipeline_mixin import PipelineTesterMixin if is_torch_available(): import torch from torch import nn from transformers import ( MODEL_MAPPING, BeitForImageClassification, BeitForMaskedImageModeling, BeitForSemanticSegmentation, BeitModel, ) from transformers.models.beit.modeling_beit import BEIT_PRETRAINED_MODEL_ARCHIVE_LIST if is_vision_available(): import PIL from PIL import Image from transformers import BeitImageProcessor class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=100 , _lowerCamelCase=13 , _lowerCamelCase=30 , _lowerCamelCase=2 , _lowerCamelCase=3 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=32 , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=37 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=10 , _lowerCamelCase=0.0_2 , _lowerCamelCase=3 , _lowerCamelCase=None , _lowerCamelCase=[0, 1, 2, 3] , ) ->Dict: SCREAMING_SNAKE_CASE : Union[str, Any] = parent SCREAMING_SNAKE_CASE : int = 100 SCREAMING_SNAKE_CASE : Dict = batch_size SCREAMING_SNAKE_CASE : Any = image_size SCREAMING_SNAKE_CASE : Any = patch_size SCREAMING_SNAKE_CASE : int = num_channels SCREAMING_SNAKE_CASE : Optional[Any] = is_training SCREAMING_SNAKE_CASE : Optional[int] = use_labels SCREAMING_SNAKE_CASE : Any = hidden_size SCREAMING_SNAKE_CASE : Union[str, Any] = num_hidden_layers SCREAMING_SNAKE_CASE : int = num_attention_heads SCREAMING_SNAKE_CASE : Union[str, Any] = intermediate_size SCREAMING_SNAKE_CASE : List[Any] = hidden_act SCREAMING_SNAKE_CASE : Tuple = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Union[str, Any] = type_sequence_label_size SCREAMING_SNAKE_CASE : Any = initializer_range SCREAMING_SNAKE_CASE : Any = scope SCREAMING_SNAKE_CASE : Any = out_indices SCREAMING_SNAKE_CASE : List[str] = num_labels # in BeiT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token) SCREAMING_SNAKE_CASE : Any = (image_size // patch_size) ** 2 SCREAMING_SNAKE_CASE : Optional[Any] = num_patches + 1 def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : List[str] = None SCREAMING_SNAKE_CASE : Tuple = None if self.use_labels: SCREAMING_SNAKE_CASE : Tuple = ids_tensor([self.batch_size] , self.type_sequence_label_size ) SCREAMING_SNAKE_CASE : int = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels ) SCREAMING_SNAKE_CASE : Dict = self.get_config() return config, pixel_values, labels, pixel_labels def __lowerCAmelCase ( self ) ->Union[str, Any]: return 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=_lowerCamelCase , initializer_range=self.initializer_range , out_indices=self.out_indices , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = BeitModel(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Tuple = model(_lowerCamelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->List[str]: SCREAMING_SNAKE_CASE : List[Any] = BeitForMaskedImageModeling(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : int = model(_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length - 1, self.vocab_size) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Any: SCREAMING_SNAKE_CASE : List[str] = self.type_sequence_label_size SCREAMING_SNAKE_CASE : List[Any] = BeitForImageClassification(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Optional[Any] = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) ) # test greyscale images SCREAMING_SNAKE_CASE : Dict = 1 SCREAMING_SNAKE_CASE : Union[str, Any] = BeitForImageClassification(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Tuple = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : str = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Any = self.num_labels SCREAMING_SNAKE_CASE : Union[str, Any] = BeitForSemanticSegmentation(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Any = model(_lowerCamelCase ) self.parent.assertEqual( result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual( result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Any = self.prepare_config_and_inputs() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = config_and_inputs SCREAMING_SNAKE_CASE : Union[str, Any] = {'''pixel_values''': pixel_values} return config, inputs_dict @require_torch class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = ( (BeitModel, BeitForImageClassification, BeitForMaskedImageModeling, BeitForSemanticSegmentation) if is_torch_available() else () ) __SCREAMING_SNAKE_CASE : Optional[int] = ( { 'feature-extraction': BeitModel, 'image-classification': BeitForImageClassification, 'image-segmentation': BeitForSemanticSegmentation, } if is_torch_available() else {} ) __SCREAMING_SNAKE_CASE : Dict = False __SCREAMING_SNAKE_CASE : List[Any] = False __SCREAMING_SNAKE_CASE : Union[str, Any] = False def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[Any] = BeitModelTester(self ) SCREAMING_SNAKE_CASE : Tuple = ConfigTester(self , config_class=_lowerCamelCase , has_text_modality=_lowerCamelCase , hidden_size=37 ) def __lowerCAmelCase ( self ) ->Dict: self.config_tester.run_common_tests() @unittest.skip(reason='''BEiT does not use inputs_embeds''' ) def __lowerCAmelCase ( self ) ->Any: pass @require_torch_multi_gpu @unittest.skip(reason='''BEiT has some layers using `add_module` which doesn\'t work well with `nn.DataParallel`''' ) def __lowerCAmelCase ( self ) ->Any: pass def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : str = model_class(_lowerCamelCase ) self.assertIsInstance(model.get_input_embeddings() , (nn.Module) ) SCREAMING_SNAKE_CASE : Tuple = model.get_output_embeddings() self.assertTrue(x is None or isinstance(_lowerCamelCase , nn.Linear ) ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : List[Any] = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = inspect.signature(model.forward ) # signature.parameters is an OrderedDict => so arg_names order is deterministic SCREAMING_SNAKE_CASE : Optional[int] = [*signature.parameters.keys()] SCREAMING_SNAKE_CASE : Any = ['''pixel_values'''] self.assertListEqual(arg_names[:1] , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_masked_lm(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : List[str] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_image_classification(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Union[str, Any] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_semantic_segmentation(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: if not self.model_tester.is_training: return SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : Any = True for model_class in self.all_model_classes: # we don't test BeitForMaskedImageModeling if model_class in [*get_values(_lowerCamelCase ), BeitForMaskedImageModeling]: continue SCREAMING_SNAKE_CASE : Any = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.train() SCREAMING_SNAKE_CASE : Optional[int] = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(**_lowerCamelCase ).loss loss.backward() def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs_for_common() if not self.model_tester.is_training: return SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : int = True for model_class in self.all_model_classes: # we don't test BeitForMaskedImageModeling if ( model_class in [*get_values(_lowerCamelCase ), BeitForMaskedImageModeling] or not model_class.supports_gradient_checkpointing ): continue SCREAMING_SNAKE_CASE : Union[str, Any] = model_class(_lowerCamelCase ) model.gradient_checkpointing_enable() model.to(_lowerCamelCase ) model.train() SCREAMING_SNAKE_CASE : int = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = model(**_lowerCamelCase ).loss loss.backward() def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : List[Any] = _config_zero_init(_lowerCamelCase ) for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Tuple = model_class(config=_lowerCamelCase ) for name, param in model.named_parameters(): # we skip lambda parameters as these require special initial values # determined by config.layer_scale_init_value if "lambda" in name: continue if param.requires_grad: self.assertIn( ((param.data.mean() * 1e9).round() / 1e9).item() , [0.0, 1.0] , msg=F"""Parameter {name} of model {model_class} seems not properly initialized""" , ) @slow def __lowerCAmelCase ( self ) ->int: for model_name in BEIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: SCREAMING_SNAKE_CASE : Any = BeitModel.from_pretrained(_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : str = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) return image @require_torch @require_vision class a_ ( unittest.TestCase ): """simple docstring""" @cached_property def __lowerCAmelCase ( self ) ->Union[str, Any]: return BeitImageProcessor.from_pretrained('''microsoft/beit-base-patch16-224''' ) if is_vision_available() else None @slow def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : List[Any] = BeitForMaskedImageModeling.from_pretrained('''microsoft/beit-base-patch16-224-pt22k''' ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = self.default_image_processor SCREAMING_SNAKE_CASE : Dict = prepare_img() SCREAMING_SNAKE_CASE : List[Any] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).pixel_values.to(_lowerCamelCase ) # prepare bool_masked_pos SCREAMING_SNAKE_CASE : Optional[Any] = torch.ones((1, 196) , dtype=torch.bool ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : List[Any] = model(pixel_values=_lowerCamelCase , bool_masked_pos=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = outputs.logits # verify the logits SCREAMING_SNAKE_CASE : Optional[Any] = torch.Size((1, 196, 8192) ) self.assertEqual(logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : str = torch.tensor( [[-3.2_4_3_7, 0.5_0_7_2, -1_3.9_1_7_4], [-3.2_4_5_6, 0.4_9_4_8, -1_3.9_4_0_1], [-3.2_0_3_3, 0.5_1_2_1, -1_3.8_5_5_0]] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(logits[bool_masked_pos][:3, :3] , _lowerCamelCase , atol=1e-2 ) ) @slow def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Any = BeitForImageClassification.from_pretrained('''microsoft/beit-base-patch16-224''' ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self.default_image_processor SCREAMING_SNAKE_CASE : int = prepare_img() SCREAMING_SNAKE_CASE : List[Any] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : Any = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = outputs.logits # verify the logits SCREAMING_SNAKE_CASE : int = torch.Size((1, 1000) ) self.assertEqual(logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = torch.tensor([-1.2_3_8_5, -1.0_9_8_7, -1.0_1_0_8] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(logits[0, :3] , _lowerCamelCase , atol=1e-4 ) ) SCREAMING_SNAKE_CASE : Optional[int] = 281 self.assertEqual(logits.argmax(-1 ).item() , _lowerCamelCase ) @slow def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Any = BeitForImageClassification.from_pretrained('''microsoft/beit-large-patch16-224-pt22k-ft22k''' ).to( _lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = self.default_image_processor SCREAMING_SNAKE_CASE : Dict = prepare_img() SCREAMING_SNAKE_CASE : Dict = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : int = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = outputs.logits # verify the logits SCREAMING_SNAKE_CASE : List[str] = torch.Size((1, 2_1841) ) self.assertEqual(logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.tensor([1.6_8_8_1, -0.2_7_8_7, 0.5_9_0_1] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(logits[0, :3] , _lowerCamelCase , atol=1e-4 ) ) SCREAMING_SNAKE_CASE : Optional[int] = 2396 self.assertEqual(logits.argmax(-1 ).item() , _lowerCamelCase ) @slow def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = BeitForSemanticSegmentation.from_pretrained('''microsoft/beit-base-finetuned-ade-640-640''' ) SCREAMING_SNAKE_CASE : Any = model.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = BeitImageProcessor(do_resize=_lowerCamelCase , size=640 , do_center_crop=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = load_dataset('''hf-internal-testing/fixtures_ade20k''' , split='''test''' ) SCREAMING_SNAKE_CASE : Union[str, Any] = Image.open(ds[0]['''file'''] ) SCREAMING_SNAKE_CASE : Any = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : Union[str, Any] = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = outputs.logits # verify the logits SCREAMING_SNAKE_CASE : Tuple = torch.Size((1, 150, 160, 160) ) self.assertEqual(logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = version.parse(PIL.__version__ ) < version.parse('''9.0.0''' ) if is_pillow_less_than_a: SCREAMING_SNAKE_CASE : List[str] = torch.tensor( [ [[-4.9_2_2_5, -2.3_9_5_4, -3.0_5_2_2], [-2.8_8_2_2, -1.0_0_4_6, -1.7_5_6_1], [-2.9_5_4_9, -1.3_2_2_8, -2.1_3_4_7]], [[-5.8_1_6_8, -3.4_1_2_9, -4.0_7_7_8], [-3.8_6_5_1, -2.2_2_1_4, -3.0_2_7_7], [-3.8_3_5_6, -2.4_6_4_3, -3.3_5_3_5]], [[-0.0_0_7_8, 3.9_9_5_2, 4.0_7_5_4], [2.9_8_5_6, 4.6_9_4_4, 5.0_0_3_5], [3.2_4_1_3, 4.7_8_1_3, 4.9_9_6_9]], ] , device=_lowerCamelCase , ) else: SCREAMING_SNAKE_CASE : Optional[int] = torch.tensor( [ [[-4.8_9_6_0, -2.3_6_8_8, -3.0_3_5_5], [-2.8_4_7_8, -0.9_8_3_6, -1.7_4_1_8], [-2.9_4_4_9, -1.3_3_3_2, -2.1_4_5_6]], [[-5.8_0_8_1, -3.4_1_2_4, -4.1_0_0_6], [-3.8_5_6_1, -2.2_0_8_1, -3.0_3_2_3], [-3.8_3_6_5, -2.4_6_0_1, -3.3_6_6_9]], [[-0.0_3_0_9, 3.9_8_6_8, 4.0_5_4_0], [2.9_6_4_0, 4.6_8_7_7, 4.9_9_7_6], [3.2_0_8_1, 4.7_6_9_0, 4.9_9_4_2]], ] , device=_lowerCamelCase , ) self.assertTrue(torch.allclose(logits[0, :3, :3, :3] , _lowerCamelCase , atol=1e-4 ) ) @slow def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Optional[int] = BeitForSemanticSegmentation.from_pretrained('''microsoft/beit-base-finetuned-ade-640-640''' ) SCREAMING_SNAKE_CASE : Dict = model.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = BeitImageProcessor(do_resize=_lowerCamelCase , size=640 , do_center_crop=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = load_dataset('''hf-internal-testing/fixtures_ade20k''' , split='''test''' ) SCREAMING_SNAKE_CASE : str = Image.open(ds[0]['''file'''] ) SCREAMING_SNAKE_CASE : Any = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : List[Any] = model(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = outputs.logits.detach().cpu() SCREAMING_SNAKE_CASE : List[Any] = image_processor.post_process_semantic_segmentation(outputs=_lowerCamelCase , target_sizes=[(500, 300)] ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Size((500, 300) ) self.assertEqual(segmentation[0].shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : str = image_processor.post_process_semantic_segmentation(outputs=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.Size((160, 160) ) self.assertEqual(segmentation[0].shape , _lowerCamelCase )
313
import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import XLMRobertaTokenizerFast from diffusers import DDIMScheduler, KandinskyImgaImgPipeline, KandinskyPriorPipeline, UNetaDConditionModel, VQModel from diffusers.pipelines.kandinsky.text_encoder import MCLIPConfig, MultilingualCLIP from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = KandinskyImgaImgPipeline __SCREAMING_SNAKE_CASE : str = ['prompt', 'image_embeds', 'negative_image_embeds', 'image'] __SCREAMING_SNAKE_CASE : int = [ 'prompt', 'negative_prompt', 'image_embeds', 'negative_image_embeds', 'image', ] __SCREAMING_SNAKE_CASE : int = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'negative_prompt', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->int: return 32 @property def __lowerCAmelCase ( self ) ->List[str]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Tuple: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 100 @property def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = XLMRobertaTokenizerFast.from_pretrained('''YiYiXu/tiny-random-mclip-base''' ) return tokenizer @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[Any] = MCLIPConfig( numDims=self.cross_attention_dim , transformerDimensions=self.text_embedder_hidden_size , hidden_size=self.text_embedder_hidden_size , intermediate_size=37 , num_attention_heads=4 , num_hidden_layers=5 , vocab_size=1005 , ) SCREAMING_SNAKE_CASE : Dict = MultilingualCLIP(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = text_encoder.eval() return text_encoder @property def __lowerCAmelCase ( self ) ->Union[str, Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = { '''in_channels''': 4, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''text_image''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''text_image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : Tuple = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->List[str]: return { "block_out_channels": [32, 64], "down_block_types": ["DownEncoderBlock2D", "AttnDownEncoderBlock2D"], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": [ "AttnUpDecoderBlock2D", "UpDecoderBlock2D", ], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Optional[Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.dummy_text_encoder SCREAMING_SNAKE_CASE : Any = self.dummy_tokenizer SCREAMING_SNAKE_CASE : List[Any] = self.dummy_unet SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_movq SCREAMING_SNAKE_CASE : Optional[Any] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : Optional[Any] = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = { '''text_encoder''': text_encoder, '''tokenizer''': tokenizer, '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->str: SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(seed + 1 ) ).to(_lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : str = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : str = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : str = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = { '''prompt''': '''horse''', '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : List[str] = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : List[Any] = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : List[Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Optional[int] = np.array( [0.6_1_4_7_4_9_4_3, 0.6_0_7_3_5_3_9, 0.4_3_3_0_8_5_4_4, 0.5_9_2_8_2_6_9, 0.4_7_4_9_3_5_9_5, 0.4_6_7_5_5_9_7_3, 0.4_6_1_3_8_3_8, 0.4_5_3_6_8_7_9_7, 0.5_0_1_1_9_2_3_3] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/kandinsky_img2img_frog.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : str = '''A red cartoon frog, 4k''' SCREAMING_SNAKE_CASE : Any = KandinskyPriorPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = KandinskyImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : str = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=5 , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : Dict = pipeline( _lowerCamelCase , image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=768 , width=768 , strength=0.2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Tuple = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
import torch from diffusers import EulerDiscreteScheduler from diffusers.utils import torch_device from .test_schedulers import SchedulerCommonTest class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = (EulerDiscreteScheduler,) __SCREAMING_SNAKE_CASE : Optional[int] = 10 def __lowerCAmelCase ( self , **_lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = { '''num_train_timesteps''': 1100, '''beta_start''': 0.0_0_0_1, '''beta_end''': 0.0_2, '''beta_schedule''': '''linear''', } config.update(**_lowerCamelCase ) return config def __lowerCAmelCase ( self ) ->Tuple: for timesteps in [10, 50, 100, 1000]: self.check_over_configs(num_train_timesteps=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: for beta_start, beta_end in zip([0.0_0_0_0_1, 0.0_0_0_1, 0.0_0_1] , [0.0_0_0_2, 0.0_0_2, 0.0_2] ): self.check_over_configs(beta_start=_lowerCamelCase , beta_end=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: for schedule in ["linear", "scaled_linear"]: self.check_over_configs(beta_schedule=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: for prediction_type in ["epsilon", "v_prediction"]: self.check_over_configs(prediction_type=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : str = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : int = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Any = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = self.dummy_model() SCREAMING_SNAKE_CASE : int = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : Any = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[Any] = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_scheduler_config(prediction_type='''v_prediction''' ) SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = self.dummy_model() SCREAMING_SNAKE_CASE : List[str] = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : List[str] = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : str = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = output.prev_sample SCREAMING_SNAKE_CASE : str = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 0.0_0_0_2 ) < 1e-2 assert abs(result_mean.item() - 2.2676e-06 ) < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Dict = self.get_scheduler_config() SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[Any] = self.dummy_model() SCREAMING_SNAKE_CASE : str = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : Optional[Any] = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : Dict = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Dict = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Optional[Any] = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase , use_karras_sigmas=_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_model() SCREAMING_SNAKE_CASE : Dict = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : int = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : List[Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = output.prev_sample SCREAMING_SNAKE_CASE : Optional[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_2_4.5_2_2_9_9_4_9_9_5_1_1_7_1_9 ) < 1e-2 assert abs(result_mean.item() - 0.1_6_2_1_3_9_3_2_6_3_3_3_9_9_9_6_3 ) < 1e-3
313
import argparse import json import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ViTImageProcessor, ViTMSNConfig, ViTMSNModel from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD torch.set_grad_enabled(False) def UpperCAmelCase_( a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [] for i in range(config.num_hidden_layers ): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append((F"""module.blocks.{i}.norm1.weight""", F"""vit.encoder.layer.{i}.layernorm_before.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm1.bias""", F"""vit.encoder.layer.{i}.layernorm_before.bias""") ) rename_keys.append( (F"""module.blocks.{i}.attn.proj.weight""", F"""vit.encoder.layer.{i}.attention.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.attn.proj.bias""", F"""vit.encoder.layer.{i}.attention.output.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.norm2.weight""", F"""vit.encoder.layer.{i}.layernorm_after.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm2.bias""", F"""vit.encoder.layer.{i}.layernorm_after.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.weight""", F"""vit.encoder.layer.{i}.intermediate.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.bias""", F"""vit.encoder.layer.{i}.intermediate.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.weight""", F"""vit.encoder.layer.{i}.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.bias""", F"""vit.encoder.layer.{i}.output.dense.bias""") ) # projection layer + position embeddings rename_keys.extend( [ ('''module.cls_token''', '''vit.embeddings.cls_token'''), ('''module.patch_embed.proj.weight''', '''vit.embeddings.patch_embeddings.projection.weight'''), ('''module.patch_embed.proj.bias''', '''vit.embeddings.patch_embeddings.projection.bias'''), ('''module.pos_embed''', '''vit.embeddings.position_embeddings'''), ] ) if base_model: # layernorm + pooler rename_keys.extend( [ ('''module.norm.weight''', '''layernorm.weight'''), ('''module.norm.bias''', '''layernorm.bias'''), ] ) # if just the base model, we should remove "vit" from all keys that start with "vit" SCREAMING_SNAKE_CASE : Any = [(pair[0], pair[1][4:]) if pair[1].startswith('''vit''' ) else pair for pair in rename_keys] else: # layernorm + classification head rename_keys.extend( [ ('''norm.weight''', '''vit.layernorm.weight'''), ('''norm.bias''', '''vit.layernorm.bias'''), ('''head.weight''', '''classifier.weight'''), ('''head.bias''', '''classifier.bias'''), ] ) return rename_keys def UpperCAmelCase_( a__ , a__ , a__=False ): """simple docstring""" for i in range(config.num_hidden_layers ): if base_model: SCREAMING_SNAKE_CASE : Any = '''''' else: SCREAMING_SNAKE_CASE : Optional[int] = '''vit.''' # read in weights + bias of input projection layer (in timm, this is a single matrix + bias) SCREAMING_SNAKE_CASE : Optional[Any] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.weight""" ) SCREAMING_SNAKE_CASE : List[str] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.bias""" ) # next, add query, keys and values (in that order) to the state dict SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ : config.hidden_size, : ] SCREAMING_SNAKE_CASE : Any = in_proj_bias[: config.hidden_size] SCREAMING_SNAKE_CASE : int = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] SCREAMING_SNAKE_CASE : List[Any] = in_proj_bias[ config.hidden_size : config.hidden_size * 2 ] SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ -config.hidden_size :, : ] SCREAMING_SNAKE_CASE : Optional[Any] = in_proj_bias[-config.hidden_size :] def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = ['''head.weight''', '''head.bias'''] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = [ '''module.fc.fc1.weight''', '''module.fc.fc1.bias''', '''module.fc.bn1.weight''', '''module.fc.bn1.bias''', '''module.fc.bn1.running_mean''', '''module.fc.bn1.running_var''', '''module.fc.bn1.num_batches_tracked''', '''module.fc.fc2.weight''', '''module.fc.fc2.bias''', '''module.fc.bn2.weight''', '''module.fc.bn2.bias''', '''module.fc.bn2.running_mean''', '''module.fc.bn2.running_var''', '''module.fc.bn2.num_batches_tracked''', '''module.fc.fc3.weight''', '''module.fc.fc3.bias''', ] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = dct.pop(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = val def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = ViTMSNConfig() SCREAMING_SNAKE_CASE : Optional[int] = 1_000 SCREAMING_SNAKE_CASE : str = '''datasets/huggingface/label-files''' SCREAMING_SNAKE_CASE : List[str] = '''imagenet-1k-id2label.json''' SCREAMING_SNAKE_CASE : Union[str, Any] = json.load(open(hf_hub_download(a__ , a__ ) , '''r''' ) ) SCREAMING_SNAKE_CASE : List[Any] = {int(a__ ): v for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : str = idalabel SCREAMING_SNAKE_CASE : int = {v: k for k, v in idalabel.items()} if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = 384 SCREAMING_SNAKE_CASE : Any = 1_536 SCREAMING_SNAKE_CASE : List[str] = 6 elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Optional[int] = 1_024 SCREAMING_SNAKE_CASE : Optional[int] = 4_096 SCREAMING_SNAKE_CASE : Tuple = 24 SCREAMING_SNAKE_CASE : Union[str, Any] = 16 SCREAMING_SNAKE_CASE : Dict = 0.1 elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = 4 elif "l7" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = 7 SCREAMING_SNAKE_CASE : Union[str, Any] = 1_024 SCREAMING_SNAKE_CASE : List[Any] = 4_096 SCREAMING_SNAKE_CASE : List[Any] = 24 SCREAMING_SNAKE_CASE : Tuple = 16 SCREAMING_SNAKE_CASE : Union[str, Any] = 0.1 SCREAMING_SNAKE_CASE : Union[str, Any] = ViTMSNModel(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = torch.hub.load_state_dict_from_url(a__ , map_location='''cpu''' )['''target_encoder'''] SCREAMING_SNAKE_CASE : Any = ViTImageProcessor(size=config.image_size ) remove_projection_head(a__ ) SCREAMING_SNAKE_CASE : Any = create_rename_keys(a__ , base_model=a__ ) for src, dest in rename_keys: rename_key(a__ , a__ , a__ ) read_in_q_k_v(a__ , a__ , base_model=a__ ) model.load_state_dict(a__ ) model.eval() SCREAMING_SNAKE_CASE : Any = '''http://images.cocodataset.org/val2017/000000039769.jpg''' SCREAMING_SNAKE_CASE : Dict = Image.open(requests.get(a__ , stream=a__ ).raw ) SCREAMING_SNAKE_CASE : Optional[int] = ViTImageProcessor( size=config.image_size , image_mean=a__ , image_std=a__ ) SCREAMING_SNAKE_CASE : int = image_processor(images=a__ , return_tensors='''pt''' ) # forward pass torch.manual_seed(2 ) SCREAMING_SNAKE_CASE : Tuple = model(**a__ ) SCREAMING_SNAKE_CASE : str = outputs.last_hidden_state # The following Colab Notebook was used to generate these outputs: # https://colab.research.google.com/gist/sayakpaul/3672419a04f5997827503fd84079bdd1/scratchpad.ipynb if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[-1.0_915, -1.4_876, -1.1_809]] ) elif "b16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[14.2_889, -18.9_045, 11.7_281]] ) elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.tensor([[41.5_028, -22.8_681, 45.6_475]] ) elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = torch.tensor([[-4.3_868, 5.2_932, -0.4_137]] ) else: SCREAMING_SNAKE_CASE : List[str] = torch.tensor([[-0.1_792, -0.6_465, 2.4_263]] ) # verify logits assert torch.allclose(last_hidden_state[:, 0, :3] , a__ , atol=1e-4 ) print(F"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(a__ ) print(F"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(a__ ) if __name__ == "__main__": a__ : Dict = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--checkpoint_url''', default='''https://dl.fbaipublicfiles.com/msn/vits16_800ep.pth.tar''', type=str, help='''URL of the checkpoint you\'d like to convert.''', ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.''' ) a__ : Any = parser.parse_args() convert_vit_msn_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
313
1
import unittest import numpy as np import timeout_decorator # noqa from transformers import BlenderbotConfig, is_flax_available from transformers.testing_utils import jax_device, require_flax, slow from ...generation.test_flax_utils import FlaxGenerationTesterMixin from ...test_modeling_flax_common import FlaxModelTesterMixin, ids_tensor if is_flax_available(): import os # The slow tests are often failing with OOM error on GPU # This makes JAX allocate exactly what is needed on demand, and deallocate memory that is no longer needed # but will be slower as stated here https://jax.readthedocs.io/en/latest/gpu_memory_allocation.html a__ : int = '''platform''' import jax import jax.numpy as jnp from transformers import BlenderbotTokenizer from transformers.models.blenderbot.modeling_flax_blenderbot import ( FlaxBlenderbotForConditionalGeneration, FlaxBlenderbotModel, shift_tokens_right, ) def UpperCAmelCase_( a__ , a__ , a__=None , a__=None , a__=None , a__=None , a__=None , a__=None , ): """simple docstring""" if attention_mask is None: SCREAMING_SNAKE_CASE : List[str] = np.where(input_ids != config.pad_token_id , 1 , 0 ) if decoder_attention_mask is None: SCREAMING_SNAKE_CASE : Union[str, Any] = np.where(decoder_input_ids != config.pad_token_id , 1 , 0 ) if head_mask is None: SCREAMING_SNAKE_CASE : Any = np.ones((config.encoder_layers, config.encoder_attention_heads) ) if decoder_head_mask is None: SCREAMING_SNAKE_CASE : str = np.ones((config.decoder_layers, config.decoder_attention_heads) ) if cross_attn_head_mask is None: SCREAMING_SNAKE_CASE : List[str] = np.ones((config.decoder_layers, config.decoder_attention_heads) ) return { "input_ids": input_ids, "decoder_input_ids": decoder_input_ids, "attention_mask": attention_mask, "decoder_attention_mask": attention_mask, } class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=13 , _lowerCamelCase=7 , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=99 , _lowerCamelCase=16 , _lowerCamelCase=2 , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=32 , _lowerCamelCase=2 , _lowerCamelCase=1 , _lowerCamelCase=0 , _lowerCamelCase=0.0_2 , ) ->str: SCREAMING_SNAKE_CASE : Optional[int] = parent SCREAMING_SNAKE_CASE : List[Any] = batch_size SCREAMING_SNAKE_CASE : List[Any] = seq_length SCREAMING_SNAKE_CASE : Dict = is_training SCREAMING_SNAKE_CASE : str = use_labels SCREAMING_SNAKE_CASE : Any = vocab_size SCREAMING_SNAKE_CASE : int = hidden_size SCREAMING_SNAKE_CASE : str = num_hidden_layers SCREAMING_SNAKE_CASE : Optional[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Optional[Any] = intermediate_size SCREAMING_SNAKE_CASE : int = hidden_act SCREAMING_SNAKE_CASE : List[str] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : List[str] = max_position_embeddings SCREAMING_SNAKE_CASE : Tuple = eos_token_id SCREAMING_SNAKE_CASE : Tuple = pad_token_id SCREAMING_SNAKE_CASE : List[Any] = bos_token_id SCREAMING_SNAKE_CASE : Union[str, Any] = initializer_range def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Dict = np.clip(ids_tensor([self.batch_size, self.seq_length - 1] , self.vocab_size ) , 3 , self.vocab_size ) SCREAMING_SNAKE_CASE : Any = np.concatenate((input_ids, 2 * np.ones((self.batch_size, 1) , dtype=np.intaa )) , -1 ) SCREAMING_SNAKE_CASE : Any = shift_tokens_right(_lowerCamelCase , 1 , 2 ) SCREAMING_SNAKE_CASE : List[str] = BlenderbotConfig( vocab_size=self.vocab_size , d_model=self.hidden_size , encoder_layers=self.num_hidden_layers , decoder_layers=self.num_hidden_layers , encoder_attention_heads=self.num_attention_heads , decoder_attention_heads=self.num_attention_heads , encoder_ffn_dim=self.intermediate_size , decoder_ffn_dim=self.intermediate_size , dropout=self.hidden_dropout_prob , attention_dropout=self.attention_probs_dropout_prob , max_position_embeddings=self.max_position_embeddings , eos_token_id=self.eos_token_id , bos_token_id=self.bos_token_id , pad_token_id=self.pad_token_id , initializer_range=self.initializer_range , use_cache=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Any = prepare_blenderbot_inputs_dict(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) return config, inputs_dict def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = self.prepare_config_and_inputs() return config, inputs_dict def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Optional[int]: SCREAMING_SNAKE_CASE : List[Any] = 20 SCREAMING_SNAKE_CASE : List[str] = model_class_name(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model.encode(inputs_dict['''input_ids'''] ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = ( inputs_dict['''decoder_input_ids'''], inputs_dict['''decoder_attention_mask'''], ) SCREAMING_SNAKE_CASE : Optional[int] = model.init_cache(decoder_input_ids.shape[0] , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = jnp.ones((decoder_input_ids.shape[0], max_decoder_length) , dtype='''i4''' ) SCREAMING_SNAKE_CASE : List[Any] = jnp.broadcast_to( jnp.arange(decoder_input_ids.shape[-1] - 1 )[None, :] , (decoder_input_ids.shape[0], decoder_input_ids.shape[-1] - 1) , ) SCREAMING_SNAKE_CASE : Optional[int] = model.decode( decoder_input_ids[:, :-1] , _lowerCamelCase , decoder_attention_mask=_lowerCamelCase , past_key_values=_lowerCamelCase , decoder_position_ids=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.array(decoder_input_ids.shape[0] * [[decoder_input_ids.shape[-1] - 1]] , dtype='''i4''' ) SCREAMING_SNAKE_CASE : Optional[int] = model.decode( decoder_input_ids[:, -1:] , _lowerCamelCase , decoder_attention_mask=_lowerCamelCase , past_key_values=outputs_cache.past_key_values , decoder_position_ids=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Optional[int] = model.decode(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = np.max(np.abs((outputs_cache_next[0][:, -1, :5] - outputs[0][:, -1, :5]) ) ) self.parent.assertTrue(diff < 1e-3 , msg=F"""Max diff is {diff}""" ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[Any] = 20 SCREAMING_SNAKE_CASE : List[str] = model_class_name(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = model.encode(inputs_dict['''input_ids'''] ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = ( inputs_dict['''decoder_input_ids'''], inputs_dict['''decoder_attention_mask'''], ) SCREAMING_SNAKE_CASE : str = jnp.concatenate( [ decoder_attention_mask, jnp.zeros((decoder_attention_mask.shape[0], max_decoder_length - decoder_attention_mask.shape[1]) ), ] , axis=-1 , ) SCREAMING_SNAKE_CASE : Tuple = model.init_cache(decoder_input_ids.shape[0] , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = jnp.broadcast_to( jnp.arange(decoder_input_ids.shape[-1] - 1 )[None, :] , (decoder_input_ids.shape[0], decoder_input_ids.shape[-1] - 1) , ) SCREAMING_SNAKE_CASE : Union[str, Any] = model.decode( decoder_input_ids[:, :-1] , _lowerCamelCase , decoder_attention_mask=_lowerCamelCase , past_key_values=_lowerCamelCase , decoder_position_ids=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : List[str] = jnp.array(decoder_input_ids.shape[0] * [[decoder_input_ids.shape[-1] - 1]] , dtype='''i4''' ) SCREAMING_SNAKE_CASE : Tuple = model.decode( decoder_input_ids[:, -1:] , _lowerCamelCase , past_key_values=outputs_cache.past_key_values , decoder_attention_mask=_lowerCamelCase , decoder_position_ids=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Tuple = model.decode(_lowerCamelCase , _lowerCamelCase , decoder_attention_mask=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = np.max(np.abs((outputs_cache_next[0][:, -1, :5] - outputs[0][:, -1, :5]) ) ) self.parent.assertTrue(diff < 1e-3 , msg=F"""Max diff is {diff}""" ) @require_flax class a_ ( unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 99 def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Any = np.array( [ [71, 82, 18, 33, 46, 91, 2], [68, 34, 26, 58, 30, 82, 2], [5, 97, 17, 39, 94, 40, 2], [76, 83, 94, 25, 70, 78, 2], [87, 59, 41, 35, 48, 66, 2], [55, 13, 16, 58, 5, 2, 1], # note padding [64, 27, 31, 51, 12, 75, 2], [52, 64, 86, 17, 83, 39, 2], [48, 61, 9, 24, 71, 82, 2], [26, 1, 60, 48, 22, 13, 2], [21, 5, 62, 28, 14, 76, 2], [45, 98, 37, 86, 59, 48, 2], [70, 70, 50, 9, 28, 0, 2], ] , dtype=np.intaa , ) SCREAMING_SNAKE_CASE : int = input_ids.shape[0] SCREAMING_SNAKE_CASE : Union[str, Any] = BlenderbotConfig( vocab_size=self.vocab_size , d_model=24 , encoder_layers=2 , decoder_layers=2 , encoder_attention_heads=2 , decoder_attention_heads=2 , encoder_ffn_dim=32 , decoder_ffn_dim=32 , max_position_embeddings=48 , eos_token_id=2 , pad_token_id=1 , bos_token_id=0 , ) return config, input_ids, batch_size def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self._get_config_and_data() SCREAMING_SNAKE_CASE : str = FlaxBlenderbotForConditionalGeneration(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = lm_model(input_ids=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = (batch_size, input_ids.shape[1], config.vocab_size) self.assertEqual(outputs['''logits'''].shape , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[str] = BlenderbotConfig( vocab_size=self.vocab_size , d_model=14 , encoder_layers=2 , decoder_layers=2 , encoder_attention_heads=2 , decoder_attention_heads=2 , encoder_ffn_dim=8 , decoder_ffn_dim=8 , max_position_embeddings=48 , ) SCREAMING_SNAKE_CASE : str = FlaxBlenderbotForConditionalGeneration(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = np.array([[71, 82, 18, 33, 46, 91, 2], [68, 34, 26, 58, 30, 2, 1]] , dtype=np.intaa ) SCREAMING_SNAKE_CASE : Optional[Any] = np.array([[82, 71, 82, 18, 2], [58, 68, 2, 1, 1]] , dtype=np.intaa ) SCREAMING_SNAKE_CASE : int = lm_model(input_ids=_lowerCamelCase , decoder_input_ids=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = (*summary.shape, config.vocab_size) self.assertEqual(outputs['''logits'''].shape , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Any = np.array([[71, 82, 18, 33, 2, 1, 1], [68, 34, 26, 58, 30, 82, 2]] , dtype=np.intaa ) SCREAMING_SNAKE_CASE : int = shift_tokens_right(_lowerCamelCase , 1 , 2 ) SCREAMING_SNAKE_CASE : str = np.equal(_lowerCamelCase , 1 ).astype(np.floataa ).sum() SCREAMING_SNAKE_CASE : str = np.equal(_lowerCamelCase , 1 ).astype(np.floataa ).sum() self.assertEqual(shifted.shape , input_ids.shape ) self.assertEqual(_lowerCamelCase , n_pad_before - 1 ) self.assertTrue(np.equal(shifted[:, 0] , 2 ).all() ) @require_flax class a_ ( a__ , unittest.TestCase , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Dict = True __SCREAMING_SNAKE_CASE : str = ( ( FlaxBlenderbotModel, FlaxBlenderbotForConditionalGeneration, ) if is_flax_available() else () ) __SCREAMING_SNAKE_CASE : Optional[int] = (FlaxBlenderbotForConditionalGeneration,) if is_flax_available() else () def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Optional[Any] = FlaxBlenderbotModelTester(self ) def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = self.model_tester.prepare_config_and_inputs() for model_class in self.all_model_classes: self.model_tester.check_use_cache_forward(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = self.model_tester.prepare_config_and_inputs() for model_class in self.all_model_classes: self.model_tester.check_use_cache_forward_with_attn_mask(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: with self.subTest(model_class.__name__ ): SCREAMING_SNAKE_CASE : Optional[Any] = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : int = model_class(_lowerCamelCase ) @jax.jit def encode_jitted(_lowerCamelCase , _lowerCamelCase=None , **_lowerCamelCase ): return model.encode(input_ids=_lowerCamelCase , attention_mask=_lowerCamelCase ) with self.subTest('''JIT Enabled''' ): SCREAMING_SNAKE_CASE : str = encode_jitted(**_lowerCamelCase ).to_tuple() with self.subTest('''JIT Disabled''' ): with jax.disable_jit(): SCREAMING_SNAKE_CASE : Any = encode_jitted(**_lowerCamelCase ).to_tuple() self.assertEqual(len(_lowerCamelCase ) , len(_lowerCamelCase ) ) for jitted_output, output in zip(_lowerCamelCase , _lowerCamelCase ): self.assertEqual(jitted_output.shape , output.shape ) def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: with self.subTest(model_class.__name__ ): SCREAMING_SNAKE_CASE : List[str] = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = model.encode(inputs_dict['''input_ids'''] , inputs_dict['''attention_mask'''] ) SCREAMING_SNAKE_CASE : Optional[Any] = { '''decoder_input_ids''': inputs_dict['''decoder_input_ids'''], '''decoder_attention_mask''': inputs_dict['''decoder_attention_mask'''], '''encoder_outputs''': encoder_outputs, } @jax.jit def decode_jitted(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ): return model.decode( decoder_input_ids=_lowerCamelCase , decoder_attention_mask=_lowerCamelCase , encoder_outputs=_lowerCamelCase , ) with self.subTest('''JIT Enabled''' ): SCREAMING_SNAKE_CASE : Union[str, Any] = decode_jitted(**_lowerCamelCase ).to_tuple() with self.subTest('''JIT Disabled''' ): with jax.disable_jit(): SCREAMING_SNAKE_CASE : Dict = decode_jitted(**_lowerCamelCase ).to_tuple() self.assertEqual(len(_lowerCamelCase ) , len(_lowerCamelCase ) ) for jitted_output, output in zip(_lowerCamelCase , _lowerCamelCase ): self.assertEqual(jitted_output.shape , output.shape ) @slow def __lowerCAmelCase ( self ) ->Optional[Any]: for model_class_name in self.all_model_classes: SCREAMING_SNAKE_CASE : Dict = model_class_name.from_pretrained('''facebook/blenderbot-400M-distill''' ) # FlaxBlenderbotForSequenceClassification expects eos token in input_ids SCREAMING_SNAKE_CASE : List[str] = np.ones((1, 1) ) * model.config.eos_token_id SCREAMING_SNAKE_CASE : Tuple = model(_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) @unittest.skipUnless(jax_device != '''cpu''' , '''3B test too slow on CPU.''' ) @slow def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Any = {'''num_beams''': 1, '''early_stopping''': True, '''min_length''': 15, '''max_length''': 25} SCREAMING_SNAKE_CASE : Tuple = {'''skip_special_tokens''': True, '''clean_up_tokenization_spaces''': True} SCREAMING_SNAKE_CASE : Dict = FlaxBlenderbotForConditionalGeneration.from_pretrained('''facebook/blenderbot-3B''' , from_pt=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = BlenderbotTokenizer.from_pretrained('''facebook/blenderbot-3B''' ) SCREAMING_SNAKE_CASE : List[str] = ['''Sam'''] SCREAMING_SNAKE_CASE : Union[str, Any] = tokenizer(_lowerCamelCase , return_tensors='''jax''' ) SCREAMING_SNAKE_CASE : str = model.generate(**_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = '''Sam is a great name. It means "sun" in Gaelic.''' SCREAMING_SNAKE_CASE : int = tokenizer.batch_decode(_lowerCamelCase , **_lowerCamelCase ) assert generated_txt[0].strip() == tgt_text
313
import csv import tweepy # Twitter API credentials a__ : Union[str, Any] = '''''' a__ : List[str] = '''''' a__ : Any = '''''' a__ : List[str] = '''''' def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = tweepy.OAuthHandler(a__ , a__ ) auth.set_access_token(a__ , a__ ) SCREAMING_SNAKE_CASE : List[str] = tweepy.API(a__ ) # initialize a list to hold all the tweepy Tweets SCREAMING_SNAKE_CASE : Any = [] # make initial request for most recent tweets (200 is the maximum allowed count) SCREAMING_SNAKE_CASE : List[Any] = api.user_timeline(screen_name=a__ , count=200 ) # save most recent tweets alltweets.extend(a__ ) # save the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Tuple = alltweets[-1].id - 1 # keep grabbing tweets until there are no tweets left to grab while len(a__ ) > 0: print(F"""getting tweets before {oldest}""" ) # all subsequent requests use the max_id param to prevent duplicates SCREAMING_SNAKE_CASE : Any = api.user_timeline( screen_name=a__ , count=200 , max_id=a__ ) # save most recent tweets alltweets.extend(a__ ) # update the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Dict = alltweets[-1].id - 1 print(F"""...{len(a__ )} tweets downloaded so far""" ) # transform the tweepy tweets into a 2D array that will populate the csv SCREAMING_SNAKE_CASE : Optional[Any] = [[tweet.id_str, tweet.created_at, tweet.text] for tweet in alltweets] # write the csv with open(F"""new_{screen_name}_tweets.csv""" , '''w''' ) as f: SCREAMING_SNAKE_CASE : List[Any] = csv.writer(a__ ) writer.writerow(['''id''', '''created_at''', '''text'''] ) writer.writerows(a__ ) if __name__ == "__main__": # pass in the username of the account you want to download get_all_tweets('''FirePing32''')
313
1
import unittest from transformers import ( MODEL_FOR_CAUSAL_LM_MAPPING, TF_MODEL_FOR_CAUSAL_LM_MAPPING, TextGenerationPipeline, logging, pipeline, ) from transformers.testing_utils import ( CaptureLogger, is_pipeline_test, require_accelerate, require_tf, require_torch, require_torch_gpu, require_torch_or_tf, ) from .test_pipelines_common import ANY @is_pipeline_test @require_torch_or_tf class a_ ( unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[int] = MODEL_FOR_CAUSAL_LM_MAPPING __SCREAMING_SNAKE_CASE : Dict = TF_MODEL_FOR_CAUSAL_LM_MAPPING @require_torch def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = pipeline(task='''text-generation''' , model='''sshleifer/tiny-ctrl''' , framework='''pt''' ) # Using `do_sample=False` to force deterministic output SCREAMING_SNAKE_CASE : Optional[Any] = text_generator('''This is a test''' , do_sample=_lowerCamelCase ) self.assertEqual( _lowerCamelCase , [ { '''generated_text''': ( '''This is a test ☃ ☃ segmental segmental segmental 议议eski eski flutter flutter Lacy oscope.''' ''' oscope. FiliFili@@''' ) } ] , ) SCREAMING_SNAKE_CASE : int = text_generator(['''This is a test''', '''This is a second test'''] ) self.assertEqual( _lowerCamelCase , [ [ { '''generated_text''': ( '''This is a test ☃ ☃ segmental segmental segmental 议议eski eski flutter flutter Lacy oscope.''' ''' oscope. FiliFili@@''' ) } ], [ { '''generated_text''': ( '''This is a second test ☃ segmental segmental segmental 议议eski eski flutter flutter Lacy''' ''' oscope. oscope. FiliFili@@''' ) } ], ] , ) SCREAMING_SNAKE_CASE : Tuple = text_generator('''This is a test''' , do_sample=_lowerCamelCase , num_return_sequences=2 , return_tensors=_lowerCamelCase ) self.assertEqual( _lowerCamelCase , [ {'''generated_token_ids''': ANY(_lowerCamelCase )}, {'''generated_token_ids''': ANY(_lowerCamelCase )}, ] , ) SCREAMING_SNAKE_CASE : Tuple = text_generator.model.config.eos_token_id SCREAMING_SNAKE_CASE : str = '''<pad>''' SCREAMING_SNAKE_CASE : List[Any] = text_generator( ['''This is a test''', '''This is a second test'''] , do_sample=_lowerCamelCase , num_return_sequences=2 , batch_size=2 , return_tensors=_lowerCamelCase , ) self.assertEqual( _lowerCamelCase , [ [ {'''generated_token_ids''': ANY(_lowerCamelCase )}, {'''generated_token_ids''': ANY(_lowerCamelCase )}, ], [ {'''generated_token_ids''': ANY(_lowerCamelCase )}, {'''generated_token_ids''': ANY(_lowerCamelCase )}, ], ] , ) @require_tf def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Optional[int] = pipeline(task='''text-generation''' , model='''sshleifer/tiny-ctrl''' , framework='''tf''' ) # Using `do_sample=False` to force deterministic output SCREAMING_SNAKE_CASE : Tuple = text_generator('''This is a test''' , do_sample=_lowerCamelCase ) self.assertEqual( _lowerCamelCase , [ { '''generated_text''': ( '''This is a test FeyFeyFey(Croatis.), s.), Cannes Cannes Cannes 閲閲Cannes Cannes Cannes 攵''' ''' please,''' ) } ] , ) SCREAMING_SNAKE_CASE : Tuple = text_generator(['''This is a test''', '''This is a second test'''] , do_sample=_lowerCamelCase ) self.assertEqual( _lowerCamelCase , [ [ { '''generated_text''': ( '''This is a test FeyFeyFey(Croatis.), s.), Cannes Cannes Cannes 閲閲Cannes Cannes Cannes 攵''' ''' please,''' ) } ], [ { '''generated_text''': ( '''This is a second test Chieftain Chieftain prefecture prefecture prefecture Cannes Cannes''' ''' Cannes 閲閲Cannes Cannes Cannes 攵 please,''' ) } ], ] , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->str: SCREAMING_SNAKE_CASE : str = TextGenerationPipeline(model=_lowerCamelCase , tokenizer=_lowerCamelCase ) return text_generator, ["This is a test", "Another test"] def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : int = '''Hello I believe in''' SCREAMING_SNAKE_CASE : str = pipeline('''text-generation''' , model='''hf-internal-testing/tiny-random-gpt2''' ) SCREAMING_SNAKE_CASE : Optional[int] = text_generator(_lowerCamelCase ) self.assertEqual( _lowerCamelCase , [{'''generated_text''': '''Hello I believe in fe fe fe fe fe fe fe fe fe fe fe fe'''}] , ) SCREAMING_SNAKE_CASE : Optional[int] = text_generator(_lowerCamelCase , stop_sequence=''' fe''' ) self.assertEqual(_lowerCamelCase , [{'''generated_text''': '''Hello I believe in fe'''}] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : List[Any] = text_generator.model SCREAMING_SNAKE_CASE : Tuple = text_generator.tokenizer SCREAMING_SNAKE_CASE : Tuple = text_generator('''This is a test''' ) self.assertEqual(_lowerCamelCase , [{'''generated_text''': ANY(_lowerCamelCase )}] ) self.assertTrue(outputs[0]['''generated_text'''].startswith('''This is a test''' ) ) SCREAMING_SNAKE_CASE : List[Any] = text_generator('''This is a test''' , return_full_text=_lowerCamelCase ) self.assertEqual(_lowerCamelCase , [{'''generated_text''': ANY(_lowerCamelCase )}] ) self.assertNotIn('''This is a test''' , outputs[0]['''generated_text'''] ) SCREAMING_SNAKE_CASE : Tuple = pipeline(task='''text-generation''' , model=_lowerCamelCase , tokenizer=_lowerCamelCase , return_full_text=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = text_generator('''This is a test''' ) self.assertEqual(_lowerCamelCase , [{'''generated_text''': ANY(_lowerCamelCase )}] ) self.assertNotIn('''This is a test''' , outputs[0]['''generated_text'''] ) SCREAMING_SNAKE_CASE : Union[str, Any] = text_generator('''This is a test''' , return_full_text=_lowerCamelCase ) self.assertEqual(_lowerCamelCase , [{'''generated_text''': ANY(_lowerCamelCase )}] ) self.assertTrue(outputs[0]['''generated_text'''].startswith('''This is a test''' ) ) SCREAMING_SNAKE_CASE : Optional[int] = text_generator(['''This is great !''', '''Something else'''] , num_return_sequences=2 , do_sample=_lowerCamelCase ) self.assertEqual( _lowerCamelCase , [ [{'''generated_text''': ANY(_lowerCamelCase )}, {'''generated_text''': ANY(_lowerCamelCase )}], [{'''generated_text''': ANY(_lowerCamelCase )}, {'''generated_text''': ANY(_lowerCamelCase )}], ] , ) if text_generator.tokenizer.pad_token is not None: SCREAMING_SNAKE_CASE : List[Any] = text_generator( ['''This is great !''', '''Something else'''] , num_return_sequences=2 , batch_size=2 , do_sample=_lowerCamelCase ) self.assertEqual( _lowerCamelCase , [ [{'''generated_text''': ANY(_lowerCamelCase )}, {'''generated_text''': ANY(_lowerCamelCase )}], [{'''generated_text''': ANY(_lowerCamelCase )}, {'''generated_text''': ANY(_lowerCamelCase )}], ] , ) with self.assertRaises(_lowerCamelCase ): SCREAMING_SNAKE_CASE : int = text_generator('''test''' , return_full_text=_lowerCamelCase , return_text=_lowerCamelCase ) with self.assertRaises(_lowerCamelCase ): SCREAMING_SNAKE_CASE : List[Any] = text_generator('''test''' , return_full_text=_lowerCamelCase , return_tensors=_lowerCamelCase ) with self.assertRaises(_lowerCamelCase ): SCREAMING_SNAKE_CASE : Optional[int] = text_generator('''test''' , return_text=_lowerCamelCase , return_tensors=_lowerCamelCase ) # Empty prompt is slighly special # it requires BOS token to exist. # Special case for Pegasus which will always append EOS so will # work even without BOS. if ( text_generator.tokenizer.bos_token_id is not None or "Pegasus" in tokenizer.__class__.__name__ or "Git" in model.__class__.__name__ ): SCREAMING_SNAKE_CASE : Optional[int] = text_generator('''''' ) self.assertEqual(_lowerCamelCase , [{'''generated_text''': ANY(_lowerCamelCase )}] ) else: with self.assertRaises((ValueError, AssertionError) ): SCREAMING_SNAKE_CASE : Optional[Any] = text_generator('''''' ) if text_generator.framework == "tf": # TF generation does not support max_new_tokens, and it's impossible # to control long generation with only max_length without # fancy calculation, dismissing tests for now. return # We don't care about infinite range models. # They already work. # Skip this test for XGLM, since it uses sinusoidal positional embeddings which are resized on-the-fly. SCREAMING_SNAKE_CASE : str = ['''RwkvForCausalLM''', '''XGLMForCausalLM''', '''GPTNeoXForCausalLM'''] if ( tokenizer.model_max_length < 1_0000 and text_generator.model.__class__.__name__ not in EXTRA_MODELS_CAN_HANDLE_LONG_INPUTS ): # Handling of large generations with self.assertRaises((RuntimeError, IndexError, ValueError, AssertionError) ): text_generator('''This is a test''' * 500 , max_new_tokens=20 ) SCREAMING_SNAKE_CASE : str = text_generator('''This is a test''' * 500 , handle_long_generation='''hole''' , max_new_tokens=20 ) # Hole strategy cannot work with self.assertRaises(_lowerCamelCase ): text_generator( '''This is a test''' * 500 , handle_long_generation='''hole''' , max_new_tokens=tokenizer.model_max_length + 10 , ) @require_torch @require_accelerate @require_torch_gpu def __lowerCAmelCase ( self ) ->Optional[Any]: import torch # Classic `model_kwargs` SCREAMING_SNAKE_CASE : Dict = pipeline( model='''hf-internal-testing/tiny-random-bloom''' , model_kwargs={'''device_map''': '''auto''', '''torch_dtype''': torch.bfloataa} , ) self.assertEqual(pipe.model.device , torch.device(0 ) ) self.assertEqual(pipe.model.lm_head.weight.dtype , torch.bfloataa ) SCREAMING_SNAKE_CASE : Any = pipe('''This is a test''' ) self.assertEqual( _lowerCamelCase , [ { '''generated_text''': ( '''This is a test test test test test test test test test test test test test test test test''' ''' test''' ) } ] , ) # Upgraded those two to real pipeline arguments (they just get sent for the model as they're unlikely to mean anything else.) SCREAMING_SNAKE_CASE : Optional[Any] = pipeline(model='''hf-internal-testing/tiny-random-bloom''' , device_map='''auto''' , torch_dtype=torch.bfloataa ) self.assertEqual(pipe.model.device , torch.device(0 ) ) self.assertEqual(pipe.model.lm_head.weight.dtype , torch.bfloataa ) SCREAMING_SNAKE_CASE : Any = pipe('''This is a test''' ) self.assertEqual( _lowerCamelCase , [ { '''generated_text''': ( '''This is a test test test test test test test test test test test test test test test test''' ''' test''' ) } ] , ) # torch_dtype will be automatically set to float32 if not provided - check: https://github.com/huggingface/transformers/pull/20602 SCREAMING_SNAKE_CASE : str = pipeline(model='''hf-internal-testing/tiny-random-bloom''' , device_map='''auto''' ) self.assertEqual(pipe.model.device , torch.device(0 ) ) self.assertEqual(pipe.model.lm_head.weight.dtype , torch.floataa ) SCREAMING_SNAKE_CASE : Optional[int] = pipe('''This is a test''' ) self.assertEqual( _lowerCamelCase , [ { '''generated_text''': ( '''This is a test test test test test test test test test test test test test test test test''' ''' test''' ) } ] , ) @require_torch @require_torch_gpu def __lowerCAmelCase ( self ) ->Union[str, Any]: import torch SCREAMING_SNAKE_CASE : List[str] = pipeline(model='''hf-internal-testing/tiny-random-bloom''' , device=0 , torch_dtype=torch.floataa ) pipe('''This is a test''' ) @require_torch @require_accelerate @require_torch_gpu def __lowerCAmelCase ( self ) ->Tuple: import torch SCREAMING_SNAKE_CASE : Union[str, Any] = pipeline(model='''hf-internal-testing/tiny-random-bloom''' , device_map='''auto''' , torch_dtype=torch.floataa ) pipe('''This is a test''' , do_sample=_lowerCamelCase , top_p=0.5 ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Dict = '''Hello world''' SCREAMING_SNAKE_CASE : Dict = pipeline('''text-generation''' , model='''hf-internal-testing/tiny-random-gpt2''' ) if text_generator.model.framework == "tf": SCREAMING_SNAKE_CASE : List[Any] = logging.get_logger('''transformers.generation.tf_utils''' ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = logging.get_logger('''transformers.generation.utils''' ) SCREAMING_SNAKE_CASE : Optional[Any] = '''Both `max_new_tokens`''' # The beggining of the message to be checked in this test # Both are set by the user -> log warning with CaptureLogger(_lowerCamelCase ) as cl: SCREAMING_SNAKE_CASE : str = text_generator(_lowerCamelCase , max_length=10 , max_new_tokens=1 ) self.assertIn(_lowerCamelCase , cl.out ) # The user only sets one -> no warning with CaptureLogger(_lowerCamelCase ) as cl: SCREAMING_SNAKE_CASE : str = text_generator(_lowerCamelCase , max_new_tokens=1 ) self.assertNotIn(_lowerCamelCase , cl.out ) with CaptureLogger(_lowerCamelCase ) as cl: SCREAMING_SNAKE_CASE : Optional[Any] = text_generator(_lowerCamelCase , max_length=10 ) self.assertNotIn(_lowerCamelCase , cl.out )
313
from collections import OrderedDict from typing import Mapping from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a__ : Optional[Any] = logging.get_logger(__name__) a__ : List[str] = { '''kssteven/ibert-roberta-base''': '''https://huggingface.co/kssteven/ibert-roberta-base/resolve/main/config.json''', '''kssteven/ibert-roberta-large''': '''https://huggingface.co/kssteven/ibert-roberta-large/resolve/main/config.json''', '''kssteven/ibert-roberta-large-mnli''': ( '''https://huggingface.co/kssteven/ibert-roberta-large-mnli/resolve/main/config.json''' ), } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'ibert' def __init__( self , _lowerCamelCase=3_0522 , _lowerCamelCase=768 , _lowerCamelCase=12 , _lowerCamelCase=12 , _lowerCamelCase=3072 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1e-12 , _lowerCamelCase=1 , _lowerCamelCase=0 , _lowerCamelCase=2 , _lowerCamelCase="absolute" , _lowerCamelCase=False , _lowerCamelCase="none" , **_lowerCamelCase , ) ->Any: super().__init__(pad_token_id=_lowerCamelCase , bos_token_id=_lowerCamelCase , eos_token_id=_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = vocab_size SCREAMING_SNAKE_CASE : str = hidden_size SCREAMING_SNAKE_CASE : Tuple = num_hidden_layers SCREAMING_SNAKE_CASE : List[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Dict = hidden_act SCREAMING_SNAKE_CASE : Optional[int] = intermediate_size SCREAMING_SNAKE_CASE : Optional[Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Tuple = type_vocab_size SCREAMING_SNAKE_CASE : Optional[int] = initializer_range SCREAMING_SNAKE_CASE : Union[str, Any] = layer_norm_eps SCREAMING_SNAKE_CASE : str = position_embedding_type SCREAMING_SNAKE_CASE : Optional[int] = quant_mode SCREAMING_SNAKE_CASE : Dict = force_dequant class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: if self.task == "multiple-choice": SCREAMING_SNAKE_CASE : Dict = {0: '''batch''', 1: '''choice''', 2: '''sequence'''} else: SCREAMING_SNAKE_CASE : List[Any] = {0: '''batch''', 1: '''sequence'''} return OrderedDict( [ ('''input_ids''', dynamic_axis), ('''attention_mask''', dynamic_axis), ] )
313
1
import inspect import unittest import warnings from math import ceil, floor from transformers import LevitConfig from transformers.file_utils import cached_property, is_torch_available, is_vision_available from transformers.models.auto import get_values from transformers.testing_utils import require_torch, require_vision, slow, torch_device 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 ( MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING, MODEL_MAPPING, LevitForImageClassification, LevitForImageClassificationWithTeacher, LevitModel, ) from transformers.models.levit.modeling_levit import LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST if is_vision_available(): from PIL import Image from transformers import LevitImageProcessor class a_ ( a__ ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Optional[Any] = self.config_class(**self.inputs_dict ) self.parent.assertTrue(hasattr(_lowerCamelCase , '''hidden_sizes''' ) ) self.parent.assertTrue(hasattr(_lowerCamelCase , '''num_attention_heads''' ) ) class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=13 , _lowerCamelCase=64 , _lowerCamelCase=3 , _lowerCamelCase=3 , _lowerCamelCase=2 , _lowerCamelCase=1 , _lowerCamelCase=16 , _lowerCamelCase=[128, 256, 384] , _lowerCamelCase=[4, 6, 8] , _lowerCamelCase=[2, 3, 4] , _lowerCamelCase=[16, 16, 16] , _lowerCamelCase=0 , _lowerCamelCase=[2, 2, 2] , _lowerCamelCase=[2, 2, 2] , _lowerCamelCase=0.0_2 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=2 , ) ->List[Any]: SCREAMING_SNAKE_CASE : int = parent SCREAMING_SNAKE_CASE : Optional[int] = batch_size SCREAMING_SNAKE_CASE : Tuple = image_size SCREAMING_SNAKE_CASE : List[Any] = num_channels SCREAMING_SNAKE_CASE : Optional[int] = kernel_size SCREAMING_SNAKE_CASE : Union[str, Any] = stride SCREAMING_SNAKE_CASE : int = padding SCREAMING_SNAKE_CASE : Optional[int] = hidden_sizes SCREAMING_SNAKE_CASE : Union[str, Any] = num_attention_heads SCREAMING_SNAKE_CASE : str = depths SCREAMING_SNAKE_CASE : int = key_dim SCREAMING_SNAKE_CASE : Any = drop_path_rate SCREAMING_SNAKE_CASE : Union[str, Any] = patch_size SCREAMING_SNAKE_CASE : Dict = attention_ratio SCREAMING_SNAKE_CASE : int = mlp_ratio SCREAMING_SNAKE_CASE : List[Any] = initializer_range SCREAMING_SNAKE_CASE : Optional[int] = [ ['''Subsample''', key_dim[0], hidden_sizes[0] // key_dim[0], 4, 2, 2], ['''Subsample''', key_dim[0], hidden_sizes[1] // key_dim[0], 4, 2, 2], ] SCREAMING_SNAKE_CASE : Union[str, Any] = is_training SCREAMING_SNAKE_CASE : Any = use_labels SCREAMING_SNAKE_CASE : Optional[Any] = num_labels SCREAMING_SNAKE_CASE : Union[str, Any] = initializer_range def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Optional[int] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : Optional[int] = None if self.use_labels: SCREAMING_SNAKE_CASE : Union[str, Any] = ids_tensor([self.batch_size] , self.num_labels ) SCREAMING_SNAKE_CASE : str = self.get_config() return config, pixel_values, labels def __lowerCAmelCase ( self ) ->Tuple: return LevitConfig( image_size=self.image_size , num_channels=self.num_channels , kernel_size=self.kernel_size , stride=self.stride , padding=self.padding , patch_size=self.patch_size , hidden_sizes=self.hidden_sizes , num_attention_heads=self.num_attention_heads , depths=self.depths , key_dim=self.key_dim , drop_path_rate=self.drop_path_rate , mlp_ratio=self.mlp_ratio , attention_ratio=self.attention_ratio , initializer_range=self.initializer_range , down_ops=self.down_ops , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[Any] = LevitModel(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = (self.image_size, self.image_size) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = image_size[0], image_size[1] for _ in range(4 ): SCREAMING_SNAKE_CASE : Optional[int] = floor(((height + 2 * self.padding - self.kernel_size) / self.stride) + 1 ) SCREAMING_SNAKE_CASE : str = floor(((width + 2 * self.padding - self.kernel_size) / self.stride) + 1 ) self.parent.assertEqual( result.last_hidden_state.shape , (self.batch_size, ceil(height / 4 ) * ceil(width / 4 ), self.hidden_sizes[-1]) , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Any: SCREAMING_SNAKE_CASE : Tuple = self.num_labels SCREAMING_SNAKE_CASE : Tuple = LevitForImageClassification(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : str = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Dict = self.prepare_config_and_inputs() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = config_and_inputs SCREAMING_SNAKE_CASE : Optional[int] = {'''pixel_values''': pixel_values} return config, inputs_dict @require_torch class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = ( (LevitModel, LevitForImageClassification, LevitForImageClassificationWithTeacher) if is_torch_available() else () ) __SCREAMING_SNAKE_CASE : Union[str, Any] = ( { 'feature-extraction': LevitModel, 'image-classification': (LevitForImageClassification, LevitForImageClassificationWithTeacher), } if is_torch_available() else {} ) __SCREAMING_SNAKE_CASE : List[str] = False __SCREAMING_SNAKE_CASE : List[Any] = False __SCREAMING_SNAKE_CASE : Optional[int] = False __SCREAMING_SNAKE_CASE : int = False __SCREAMING_SNAKE_CASE : int = False def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : int = LevitModelTester(self ) SCREAMING_SNAKE_CASE : Any = ConfigTester(self , config_class=_lowerCamelCase , has_text_modality=_lowerCamelCase , hidden_size=37 ) def __lowerCAmelCase ( self ) ->List[Any]: 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 __lowerCAmelCase ( self ) ->Any: return @unittest.skip(reason='''Levit does not use inputs_embeds''' ) def __lowerCAmelCase ( self ) ->str: pass @unittest.skip(reason='''Levit does not support input and output embeddings''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @unittest.skip(reason='''Levit does not output attentions''' ) def __lowerCAmelCase ( self ) ->Tuple: pass def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Union[str, Any] = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = inspect.signature(model.forward ) # signature.parameters is an OrderedDict => so arg_names order is deterministic SCREAMING_SNAKE_CASE : Dict = [*signature.parameters.keys()] SCREAMING_SNAKE_CASE : List[Any] = ['''pixel_values'''] self.assertListEqual(arg_names[:1] , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Union[str, Any]: def check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : Optional[int] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : int = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = outputs.hidden_states SCREAMING_SNAKE_CASE : Optional[int] = len(self.model_tester.depths ) + 1 self.assertEqual(len(_lowerCamelCase ) , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = (self.model_tester.image_size, self.model_tester.image_size) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = image_size[0], image_size[1] for _ in range(4 ): SCREAMING_SNAKE_CASE : Tuple = floor( ( (height + 2 * self.model_tester.padding - self.model_tester.kernel_size) / self.model_tester.stride ) + 1 ) SCREAMING_SNAKE_CASE : Union[str, Any] = floor( ( (width + 2 * self.model_tester.padding - self.model_tester.kernel_size) / self.model_tester.stride ) + 1 ) # verify the first hidden states (first block) self.assertListEqual( list(hidden_states[0].shape[-2:] ) , [ height * width, self.model_tester.hidden_sizes[0], ] , ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Union[str, Any] = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # check that output_hidden_states also work using config del inputs_dict["output_hidden_states"] SCREAMING_SNAKE_CASE : Union[str, Any] = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) @unittest.skip('''Will be fixed soon by reducing the size of the model used for common tests.''' ) def __lowerCAmelCase ( self ) ->Tuple: pass def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=False ) ->List[Any]: SCREAMING_SNAKE_CASE : List[Any] = super()._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) if return_labels: if model_class.__name__ == "LevitForImageClassificationWithTeacher": del inputs_dict["labels"] return inputs_dict def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Any = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : List[Any] = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_image_classification(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: if not self.model_tester.is_training: return SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : int = True for model_class in self.all_model_classes: # LevitForImageClassificationWithTeacher supports inference-only if ( model_class in get_values(_lowerCamelCase ) or model_class.__name__ == "LevitForImageClassificationWithTeacher" ): continue SCREAMING_SNAKE_CASE : int = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.train() SCREAMING_SNAKE_CASE : List[Any] = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = model(**_lowerCamelCase ).loss loss.backward() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs_for_common() if not self.model_tester.is_training: return SCREAMING_SNAKE_CASE : Dict = False SCREAMING_SNAKE_CASE : Dict = True for model_class in self.all_model_classes: if model_class in get_values(_lowerCamelCase ) or not model_class.supports_gradient_checkpointing: continue # LevitForImageClassificationWithTeacher supports inference-only if model_class.__name__ == "LevitForImageClassificationWithTeacher": continue SCREAMING_SNAKE_CASE : Tuple = model_class(_lowerCamelCase ) model.gradient_checkpointing_enable() model.to(_lowerCamelCase ) model.train() SCREAMING_SNAKE_CASE : Union[str, Any] = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(**_lowerCamelCase ).loss loss.backward() def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : int = [ {'''title''': '''multi_label_classification''', '''num_labels''': 2, '''dtype''': torch.float}, {'''title''': '''single_label_classification''', '''num_labels''': 1, '''dtype''': torch.long}, {'''title''': '''regression''', '''num_labels''': 1, '''dtype''': torch.float}, ] for model_class in self.all_model_classes: if ( model_class not in [ *get_values(_lowerCamelCase ), ] or model_class.__name__ == "LevitForImageClassificationWithTeacher" ): continue for problem_type in problem_types: with self.subTest(msg=F"""Testing {model_class} with {problem_type["title"]}""" ): SCREAMING_SNAKE_CASE : Union[str, Any] = problem_type['''title'''] SCREAMING_SNAKE_CASE : Optional[Any] = problem_type['''num_labels'''] SCREAMING_SNAKE_CASE : List[str] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.train() SCREAMING_SNAKE_CASE : Any = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) if problem_type["num_labels"] > 1: SCREAMING_SNAKE_CASE : List[Any] = inputs['''labels'''].unsqueeze(1 ).repeat(1 , problem_type['''num_labels'''] ) SCREAMING_SNAKE_CASE : Optional[int] = inputs['''labels'''].to(problem_type['''dtype'''] ) # This tests that we do not trigger the warning form PyTorch "Using a target size that is different # to the input size. This will likely lead to incorrect results due to broadcasting. Please ensure # they have the same size." which is a symptom something in wrong for the regression problem. # See https://github.com/huggingface/transformers/issues/11780 with warnings.catch_warnings(record=_lowerCamelCase ) as warning_list: SCREAMING_SNAKE_CASE : Any = model(**_lowerCamelCase ).loss for w in warning_list: if "Using a target size that is different to the input size" in str(w.message ): raise ValueError( F"""Something is going wrong in the regression problem: intercepted {w.message}""" ) loss.backward() @slow def __lowerCAmelCase ( self ) ->Dict: for model_name in LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: SCREAMING_SNAKE_CASE : Optional[int] = LevitModel.from_pretrained(_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) return image @require_torch @require_vision class a_ ( unittest.TestCase ): """simple docstring""" @cached_property def __lowerCAmelCase ( self ) ->List[str]: return LevitImageProcessor.from_pretrained(LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST[0] ) @slow def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = LevitForImageClassificationWithTeacher.from_pretrained(LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST[0] ).to( _lowerCamelCase ) SCREAMING_SNAKE_CASE : int = self.default_image_processor SCREAMING_SNAKE_CASE : str = prepare_img() SCREAMING_SNAKE_CASE : Union[str, Any] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ).to(_lowerCamelCase ) # forward pass with torch.no_grad(): SCREAMING_SNAKE_CASE : Any = model(**_lowerCamelCase ) # verify the logits SCREAMING_SNAKE_CASE : Any = torch.Size((1, 1000) ) self.assertEqual(outputs.logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : int = torch.tensor([1.0_4_4_8, -0.3_7_4_5, -1.8_3_1_7] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(outputs.logits[0, :3] , _lowerCamelCase , atol=1e-4 ) )
313
from collections import OrderedDict from typing import TYPE_CHECKING, Any, Mapping, Optional from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging if TYPE_CHECKING: from ... import FeatureExtractionMixin, TensorType a__ : Any = logging.get_logger(__name__) a__ : Dict = { '''openai/imagegpt-small''': '''''', '''openai/imagegpt-medium''': '''''', '''openai/imagegpt-large''': '''''', } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 'imagegpt' __SCREAMING_SNAKE_CASE : Optional[Any] = ['past_key_values'] __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'n_embd', 'max_position_embeddings': 'n_positions', 'num_attention_heads': 'n_head', 'num_hidden_layers': 'n_layer', } def __init__( self , _lowerCamelCase=512 + 1 , _lowerCamelCase=32 * 32 , _lowerCamelCase=512 , _lowerCamelCase=24 , _lowerCamelCase=8 , _lowerCamelCase=None , _lowerCamelCase="quick_gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=1e-5 , _lowerCamelCase=0.0_2 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=False , _lowerCamelCase=False , **_lowerCamelCase , ) ->str: SCREAMING_SNAKE_CASE : Any = vocab_size SCREAMING_SNAKE_CASE : Optional[int] = n_positions SCREAMING_SNAKE_CASE : Optional[int] = n_embd SCREAMING_SNAKE_CASE : List[Any] = n_layer SCREAMING_SNAKE_CASE : List[Any] = n_head SCREAMING_SNAKE_CASE : int = n_inner SCREAMING_SNAKE_CASE : Dict = activation_function SCREAMING_SNAKE_CASE : Union[str, Any] = resid_pdrop SCREAMING_SNAKE_CASE : Dict = embd_pdrop SCREAMING_SNAKE_CASE : List[str] = attn_pdrop SCREAMING_SNAKE_CASE : Optional[int] = layer_norm_epsilon SCREAMING_SNAKE_CASE : Tuple = initializer_range SCREAMING_SNAKE_CASE : int = scale_attn_weights SCREAMING_SNAKE_CASE : Optional[int] = use_cache SCREAMING_SNAKE_CASE : Optional[Any] = scale_attn_by_inverse_layer_idx SCREAMING_SNAKE_CASE : str = reorder_and_upcast_attn SCREAMING_SNAKE_CASE : List[str] = tie_word_embeddings super().__init__(tie_word_embeddings=_lowerCamelCase , **_lowerCamelCase ) class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: return OrderedDict( [ ('''input_ids''', {0: '''batch''', 1: '''sequence'''}), ] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 1 , _lowerCamelCase = -1 , _lowerCamelCase = False , _lowerCamelCase = None , _lowerCamelCase = 3 , _lowerCamelCase = 32 , _lowerCamelCase = 32 , ) ->Mapping[str, Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = self._generate_dummy_images(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = dict(preprocessor(images=_lowerCamelCase , return_tensors=_lowerCamelCase ) ) return inputs
313
1
def UpperCAmelCase_( a__ ): """simple docstring""" return " ".join(input_str.split()[::-1] ) if __name__ == "__main__": import doctest doctest.testmod()
313
from maths.prime_check import is_prime def UpperCAmelCase_( a__ ): """simple docstring""" if not isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = F"""Input value of [number={number}] must be an integer""" raise TypeError(a__ ) if is_prime(a__ ) and is_prime(number + 2 ): return number + 2 else: return -1 if __name__ == "__main__": import doctest doctest.testmod()
313
1
import argparse import os from transformers.utils import direct_transformers_import # All paths are set with the intent you should run this script from the root of the repo with the command # python utils/check_task_guides.py a__ : Optional[Any] = '''src/transformers''' a__ : Dict = '''docs/source/en/tasks''' def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" with open(a__ , '''r''' , encoding='''utf-8''' , newline='''\n''' ) as f: SCREAMING_SNAKE_CASE : Tuple = f.readlines() # Find the start prompt. SCREAMING_SNAKE_CASE : Optional[Any] = 0 while not lines[start_index].startswith(a__ ): start_index += 1 start_index += 1 SCREAMING_SNAKE_CASE : Union[str, Any] = start_index while not lines[end_index].startswith(a__ ): end_index += 1 end_index -= 1 while len(lines[start_index] ) <= 1: start_index += 1 while len(lines[end_index] ) <= 1: end_index -= 1 end_index += 1 return "".join(lines[start_index:end_index] ), start_index, end_index, lines # This is to make sure the transformers module imported is the one in the repo. a__ : str = direct_transformers_import(TRANSFORMERS_PATH) a__ : Any = { '''asr.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_CTC_MAPPING_NAMES, '''audio_classification.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMES, '''language_modeling.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_CAUSAL_LM_MAPPING_NAMES, '''image_classification.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES, '''masked_language_modeling.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_MASKED_LM_MAPPING_NAMES, '''multiple_choice.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_MULTIPLE_CHOICE_MAPPING_NAMES, '''object_detection.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES, '''question_answering.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES, '''semantic_segmentation.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING_NAMES, '''sequence_classification.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES, '''summarization.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES, '''token_classification.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES, '''translation.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES, '''video_classification.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING_NAMES, '''document_question_answering.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING_NAMES, '''monocular_depth_estimation.md''': transformers_module.models.auto.modeling_auto.MODEL_FOR_DEPTH_ESTIMATION_MAPPING_NAMES, } # This list contains model types used in some task guides that are not in `CONFIG_MAPPING_NAMES` (therefore not in any # `MODEL_MAPPING_NAMES` or any `MODEL_FOR_XXX_MAPPING_NAMES`). a__ : Tuple = { '''summarization.md''': ('''nllb''',), '''translation.md''': ('''nllb''',), } def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = TASK_GUIDE_TO_MODELS[task_guide] SCREAMING_SNAKE_CASE : List[Any] = SPECIAL_TASK_GUIDE_TO_MODEL_TYPES.get(a__ , set() ) SCREAMING_SNAKE_CASE : Dict = { code: name for code, name in transformers_module.MODEL_NAMES_MAPPING.items() if (code in model_maping_names or code in special_model_types) } return ", ".join([F"""[{name}](../model_doc/{code})""" for code, name in model_names.items()] ) + "\n" def UpperCAmelCase_( a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = _find_text_in_file( filename=os.path.join(a__ , a__ ) , start_prompt='''<!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->''' , end_prompt='''<!--End of the generated tip-->''' , ) SCREAMING_SNAKE_CASE : Optional[int] = get_model_list_for_task(a__ ) if current_list != new_list: if overwrite: with open(os.path.join(a__ , a__ ) , '''w''' , encoding='''utf-8''' , newline='''\n''' ) as f: f.writelines(lines[:start_index] + [new_list] + lines[end_index:] ) else: raise ValueError( F"""The list of models that can be used in the {task_guide} guide needs an update. Run `make fix-copies`""" ''' to fix this.''' ) if __name__ == "__main__": a__ : int = argparse.ArgumentParser() parser.add_argument('''--fix_and_overwrite''', action='''store_true''', help='''Whether to fix inconsistencies.''') a__ : List[Any] = parser.parse_args() for task_guide in TASK_GUIDE_TO_MODELS.keys(): check_model_list_for_task(task_guide, args.fix_and_overwrite)
313
import gc import random import unittest import numpy as np import torch from PIL import Image from diffusers import ( DDIMScheduler, KandinskyVaaControlnetImgaImgPipeline, KandinskyVaaPriorEmbaEmbPipeline, UNetaDConditionModel, VQModel, ) from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[Any] = KandinskyVaaControlnetImgaImgPipeline __SCREAMING_SNAKE_CASE : Optional[int] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[Any] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[str] = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->Optional[Any]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 32 @property def __lowerCAmelCase ( self ) ->str: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Dict: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Tuple: return 100 @property def __lowerCAmelCase ( self ) ->int: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Union[str, Any] = { '''in_channels''': 8, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''image_hint''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : List[str] = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->Any: return { "block_out_channels": [32, 32, 64, 64], "down_block_types": [ "DownEncoderBlock2D", "DownEncoderBlock2D", "DownEncoderBlock2D", "AttnDownEncoderBlock2D", ], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": ["AttnUpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D"], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : str = self.dummy_unet SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_movq SCREAMING_SNAKE_CASE : List[str] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : str = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = { '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->int: SCREAMING_SNAKE_CASE : str = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(seed + 1 ) ).to( _lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : Dict = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) # create hint SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : List[str] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = { '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''hint''': hint, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[int] = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : Any = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : Union[str, Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array( [0.5_4_9_8_5_0_3_4, 0.5_5_5_0_9_3_6_5, 0.5_2_5_6_1_5_0_4, 0.5_5_7_0_4_9_4, 0.5_5_9_3_8_1_8, 0.5_2_6_3_9_7_9, 0.5_0_2_8_5_6_4_3, 0.5_0_6_9_8_4_6, 0.5_1_1_9_6_7_3_6] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Union[str, Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/kandinskyv22_controlnet_img2img_robotcat_fp16.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = init_image.resize((512, 512) ) SCREAMING_SNAKE_CASE : Dict = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/hint_image_cat.png''' ) SCREAMING_SNAKE_CASE : List[Any] = torch.from_numpy(np.array(_lowerCamelCase ) ).float() / 2_5_5.0 SCREAMING_SNAKE_CASE : int = hint.permute(2 , 0 , 1 ).unsqueeze(0 ) SCREAMING_SNAKE_CASE : List[Any] = '''A robot, 4k photo''' SCREAMING_SNAKE_CASE : List[str] = KandinskyVaaPriorEmbaEmbPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = KandinskyVaaControlnetImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-controlnet-depth''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Any = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , image=_lowerCamelCase , strength=0.8_5 , generator=_lowerCamelCase , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : List[str] = pipeline( image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , hint=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=512 , width=512 , strength=0.5 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = output.images[0] assert image.shape == (512, 512, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
# limitations under the License. # NOTE: This file is deprecated and will be removed in a future version. # It only exists so that temporarely `from diffusers.pipelines import DiffusionPipeline` works from .pipelines import DiffusionPipeline, ImagePipelineOutput # noqa: F401 from .utils import deprecate deprecate( '''pipelines_utils''', '''0.22.0''', '''Importing `DiffusionPipeline` or `ImagePipelineOutput` from diffusers.pipeline_utils is deprecated. Please import from diffusers.pipelines.pipeline_utils instead.''', standard_warn=False, stacklevel=3, )
313
from typing import Any, Callable, Dict, List, Optional, Union import torch from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer from diffusers import ( AutoencoderKL, DDIMScheduler, DiffusionPipeline, LMSDiscreteScheduler, PNDMScheduler, StableDiffusionPipeline, UNetaDConditionModel, ) from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker a__ : List[str] = '''CompVis/stable-diffusion-v1-1''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-2''' a__ : Any = '''CompVis/stable-diffusion-v1-3''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-4''' class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , ) ->str: super()._init_() SCREAMING_SNAKE_CASE : Tuple = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = StableDiffusionPipeline( vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , unet=_lowerCamelCase , scheduler=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=_lowerCamelCase , requires_safety_checker=_lowerCamelCase , ) self.register_modules(pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea ) @property def __lowerCAmelCase ( self ) ->Dict[str, Any]: return {k: getattr(self , _lowerCamelCase ) for k in self.config.keys() if not k.startswith('''_''' )} def __lowerCAmelCase ( self , _lowerCamelCase = "auto" ) ->Optional[int]: if slice_size == "auto": # half the attention head size is usually a good trade-off between # speed and memory SCREAMING_SNAKE_CASE : Optional[int] = self.unet.config.attention_head_dim // 2 self.unet.set_attention_slice(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: self.enable_attention_slicing(_lowerCamelCase ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->str: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Tuple: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Optional[Any]: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: SCREAMING_SNAKE_CASE : Optional[Any] = '''cuda''' if torch.cuda.is_available() else '''cpu''' self.to(_lowerCamelCase ) # Checks if the height and width are divisible by 8 or not if height % 8 != 0 or width % 8 != 0: raise ValueError(F"""`height` and `width` must be divisible by 8 but are {height} and {width}.""" ) # Get first result from Stable Diffusion Checkpoint v1.1 SCREAMING_SNAKE_CASE : Optional[Any] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.2 SCREAMING_SNAKE_CASE : Any = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.3 SCREAMING_SNAKE_CASE : Optional[int] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.4 SCREAMING_SNAKE_CASE : str = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get all result images into a single list and pass it via StableDiffusionPipelineOutput for final result return StableDiffusionPipelineOutput([resa[0], resa[0], resa[0], resa[0]] )
313
1
import inspect import logging import os import random import shutil import tempfile import unittest import pytest import torch from torch import nn from torch.utils.data import DataLoader, TensorDataset from accelerate import Accelerator from accelerate.test_utils import execute_subprocess_async, require_cuda from accelerate.utils import ProjectConfiguration, set_seed a__ : Any = logging.getLogger(__name__) def UpperCAmelCase_( a__=2 , a__=3 , a__=16 , a__ = 10 , a__ = 2 ): """simple docstring""" def get_dataset(a__ ): SCREAMING_SNAKE_CASE : Tuple = torch.randn(batch_size * n_batches , 1 ) return TensorDataset(a__ , a * x + b + 0.1 * torch.randn(batch_size * n_batches , 1 ) ) SCREAMING_SNAKE_CASE : Optional[int] = get_dataset(a__ ) SCREAMING_SNAKE_CASE : Dict = get_dataset(a__ ) SCREAMING_SNAKE_CASE : List[Any] = DataLoader(a__ , shuffle=a__ , batch_size=a__ , num_workers=4 ) SCREAMING_SNAKE_CASE : Tuple = DataLoader(a__ , shuffle=a__ , batch_size=a__ , num_workers=4 ) return (train_dataloader, valid_dataloader) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ , a__=None ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [] for epoch in range(a__ ): # Train quickly model.train() for batch in dataloader: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = batch SCREAMING_SNAKE_CASE : Optional[int] = model(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch.nn.functional.mse_loss(a__ , a__ ) accelerator.backward(a__ ) optimizer.step() optimizer.zero_grad() rands.append(random.random() ) # Introduce some randomness if scheduler is not None: scheduler.step() return rands class a_ ( nn.Module ): """simple docstring""" def __init__( self ) ->Optional[Any]: super().__init__() SCREAMING_SNAKE_CASE : Any = nn.Parameter(torch.randn(1 ) ) SCREAMING_SNAKE_CASE : List[str] = nn.Parameter(torch.randn(1 ) ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->Tuple: return x * self.a + self.b class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Optional[int]: with tempfile.TemporaryDirectory() as tmpdir: set_seed(42 ) SCREAMING_SNAKE_CASE : Union[str, Any] = DummyModel() SCREAMING_SNAKE_CASE : Optional[int] = torch.optim.Adam(params=model.parameters() , lr=1e-3 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = dummy_dataloaders() SCREAMING_SNAKE_CASE : List[str] = ProjectConfiguration(total_limit=1 , project_dir=_lowerCamelCase , automatic_checkpoint_naming=_lowerCamelCase ) # Train baseline SCREAMING_SNAKE_CASE : Dict = Accelerator(project_config=_lowerCamelCase ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = accelerator.prepare( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # Save initial accelerator.save_state() # Save second state accelerator.save_state() self.assertEqual(len(os.listdir(accelerator.project_dir ) ) , 1 ) def __lowerCAmelCase ( self ) ->Any: with tempfile.TemporaryDirectory() as tmpdir: set_seed(42 ) SCREAMING_SNAKE_CASE : List[Any] = DummyModel() SCREAMING_SNAKE_CASE : Union[str, Any] = torch.optim.Adam(params=model.parameters() , lr=1e-3 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = dummy_dataloaders() # Train baseline SCREAMING_SNAKE_CASE : int = Accelerator() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = accelerator.prepare( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # Save initial SCREAMING_SNAKE_CASE : int = os.path.join(_lowerCamelCase , '''initial''' ) accelerator.save_state(_lowerCamelCase ) ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : Union[str, Any] = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : str = optimizer.state_dict() SCREAMING_SNAKE_CASE : Optional[Any] = train(3 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : Dict = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : List[str] = optimizer.state_dict() # Train partially set_seed(42 ) SCREAMING_SNAKE_CASE : List[str] = DummyModel() SCREAMING_SNAKE_CASE : Optional[int] = torch.optim.Adam(params=model.parameters() , lr=1e-3 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = dummy_dataloaders() SCREAMING_SNAKE_CASE : Any = Accelerator() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = accelerator.prepare( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) accelerator.load_state(_lowerCamelCase ) ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : Tuple = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : int = optimizer.state_dict() self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : str = train(2 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # Save everything SCREAMING_SNAKE_CASE : Dict = os.path.join(_lowerCamelCase , '''checkpoint''' ) accelerator.save_state(_lowerCamelCase ) # Load everything back in and make sure all states work accelerator.load_state(_lowerCamelCase ) test_rands += train(1 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : Any = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : Optional[int] = optimizer.state_dict() self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[int]: with tempfile.TemporaryDirectory() as tmpdir: set_seed(42 ) SCREAMING_SNAKE_CASE : Any = DummyModel() SCREAMING_SNAKE_CASE : Optional[Any] = torch.optim.Adam(params=model.parameters() , lr=1e-3 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = dummy_dataloaders() SCREAMING_SNAKE_CASE : str = ProjectConfiguration(automatic_checkpoint_naming=_lowerCamelCase ) # Train baseline SCREAMING_SNAKE_CASE : List[Any] = Accelerator(project_dir=_lowerCamelCase , project_config=_lowerCamelCase ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = accelerator.prepare( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # Save initial accelerator.save_state() ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : List[str] = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : Union[str, Any] = optimizer.state_dict() SCREAMING_SNAKE_CASE : Optional[Any] = train(3 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : str = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : str = optimizer.state_dict() # Train partially set_seed(42 ) SCREAMING_SNAKE_CASE : Dict = DummyModel() SCREAMING_SNAKE_CASE : List[str] = torch.optim.Adam(params=model.parameters() , lr=1e-3 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = dummy_dataloaders() SCREAMING_SNAKE_CASE : List[Any] = ProjectConfiguration(iteration=1 , automatic_checkpoint_naming=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = Accelerator(project_dir=_lowerCamelCase , project_config=_lowerCamelCase ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = accelerator.prepare( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) accelerator.load_state(os.path.join(_lowerCamelCase , '''checkpoints''' , '''checkpoint_0''' ) ) ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : Tuple = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : Optional[Any] = optimizer.state_dict() self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = train(2 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # Save everything accelerator.save_state() # Load everything back in and make sure all states work accelerator.load_state(os.path.join(_lowerCamelCase , '''checkpoints''' , '''checkpoint_1''' ) ) test_rands += train(1 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ((SCREAMING_SNAKE_CASE) , (SCREAMING_SNAKE_CASE)) : Any = model.a.item(), model.b.item() SCREAMING_SNAKE_CASE : Optional[int] = optimizer.state_dict() self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) self.assertEqual(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Optional[int] = torch.tensor([1, 2, 3] ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch.tensor([2, 3, 4] ) SCREAMING_SNAKE_CASE : Optional[int] = DummyModel() SCREAMING_SNAKE_CASE : str = torch.optim.Adam(net.parameters() ) SCREAMING_SNAKE_CASE : Optional[int] = Accelerator() with self.assertRaises(_lowerCamelCase ) as ve: accelerator.register_for_checkpointing(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : int = str(ve.exception ) self.assertTrue('''Item at index 0''' in message ) self.assertTrue('''Item at index 1''' in message ) self.assertFalse('''Item at index 2''' in message ) self.assertFalse('''Item at index 3''' in message ) def __lowerCAmelCase ( self ) ->str: with tempfile.TemporaryDirectory() as tmpdir: set_seed(42 ) SCREAMING_SNAKE_CASE : Union[str, Any] = DummyModel() SCREAMING_SNAKE_CASE : Dict = torch.optim.Adam(params=model.parameters() , lr=1e-3 ) SCREAMING_SNAKE_CASE : Any = torch.optim.lr_scheduler.StepLR(_lowerCamelCase , step_size=1 , gamma=0.9_9 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = dummy_dataloaders() SCREAMING_SNAKE_CASE : Tuple = ProjectConfiguration(automatic_checkpoint_naming=_lowerCamelCase ) # Train baseline SCREAMING_SNAKE_CASE : Any = Accelerator(project_dir=_lowerCamelCase , project_config=_lowerCamelCase ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = accelerator.prepare( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # Save initial accelerator.save_state() SCREAMING_SNAKE_CASE : Any = scheduler.state_dict() train(3 , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) self.assertNotEqual(_lowerCamelCase , scheduler.state_dict() ) # Load everything back in and make sure all states work accelerator.load_state(os.path.join(_lowerCamelCase , '''checkpoints''' , '''checkpoint_0''' ) ) self.assertEqual(_lowerCamelCase , scheduler.state_dict() ) def __lowerCAmelCase ( self ) ->str: with tempfile.TemporaryDirectory() as tmpdir: set_seed(42 ) SCREAMING_SNAKE_CASE : str = DummyModel() SCREAMING_SNAKE_CASE : Any = ProjectConfiguration(automatic_checkpoint_naming=_lowerCamelCase , total_limit=2 ) # Train baseline SCREAMING_SNAKE_CASE : Dict = Accelerator(project_dir=_lowerCamelCase , project_config=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = accelerator.prepare(_lowerCamelCase ) # Save 3 states: for _ in range(11 ): accelerator.save_state() self.assertTrue(not os.path.exists(os.path.join(_lowerCamelCase , '''checkpoints''' , '''checkpoint_0''' ) ) ) self.assertTrue(os.path.exists(os.path.join(_lowerCamelCase , '''checkpoints''' , '''checkpoint_9''' ) ) ) self.assertTrue(os.path.exists(os.path.join(_lowerCamelCase , '''checkpoints''' , '''checkpoint_10''' ) ) ) @require_cuda def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : int = ['''torchrun''', F"""--nproc_per_node={torch.cuda.device_count()}""", inspect.getfile(self.__class__ )] execute_subprocess_async(_lowerCamelCase , env=os.environ.copy() ) if __name__ == "__main__": a__ : Any = '''/tmp/accelerate/state_checkpointing''' a__ : Dict = DummyModel() a__ : Optional[Any] = torch.optim.Adam(params=model.parameters(), lr=1E-3) a__ : Optional[Any] = torch.optim.lr_scheduler.StepLR(optimizer, step_size=1, gamma=0.99) a__ , a__ : Optional[Any] = dummy_dataloaders() a__ : List[str] = ProjectConfiguration(automatic_checkpoint_naming=True) # Train baseline a__ : Tuple = Accelerator(project_dir=savedir, project_config=project_config, mixed_precision='''no''') if accelerator.process_index == 0: if os.path.exists(savedir): shutil.rmtree(savedir) os.makedirs(savedir) a__ , a__ , a__ , a__ , a__ : Union[str, Any] = accelerator.prepare( model, optimizer, train_dataloader, valid_dataloader, scheduler ) a__ , a__ : List[str] = accelerator.prepare(model, optimizer) train(3, model, train_dataloader, optimizer, accelerator, scheduler) # Check that the intial optimizer is loaded on the GPU for group in optimizer.param_groups: a__ : Dict = group['''params'''][0].device break assert param_device.type == accelerator.device.type a__ : Optional[int] = model.cpu() accelerator.wait_for_everyone() accelerator.save_state() accelerator.wait_for_everyone() # Check CPU state accelerator.load_state(os.path.join(savedir, '''checkpoints''', '''checkpoint_0'''), map_location='''cpu''') for group in optimizer.param_groups: a__ : Any = group['''params'''][0].device break assert ( param_device.type == torch.device('''cpu''').type ), F"Loaded optimizer states did not match, expected to be loaded on the CPU but got {param_device}" # Check device state model.to(accelerator.device) accelerator.load_state(os.path.join(savedir, '''checkpoints''', '''checkpoint_0'''), map_location='''on_device''') for group in optimizer.param_groups: a__ : List[str] = group['''params'''][0].device break assert ( param_device.type == accelerator.device.type ), F"Loaded optimizer states did not match, expected to be loaded on {accelerator.device} but got {param_device}" # Check error with pytest.raises(TypeError, match='''Unsupported optimizer map location passed'''): accelerator.load_state(os.path.join(savedir, '''checkpoints''', '''checkpoint_0'''), map_location='''invalid''') accelerator.wait_for_everyone() if accelerator.process_index == 0: shutil.rmtree(savedir) accelerator.wait_for_everyone()
313
from typing import Optional, Tuple, Union import flax import flax.linen as nn import jax import jax.numpy as jnp from flax.core.frozen_dict import FrozenDict from ..configuration_utils import ConfigMixin, flax_register_to_config from ..utils import BaseOutput from .embeddings_flax import FlaxTimestepEmbedding, FlaxTimesteps from .modeling_flax_utils import FlaxModelMixin from .unet_ad_blocks_flax import ( FlaxCrossAttnDownBlockaD, FlaxCrossAttnUpBlockaD, FlaxDownBlockaD, FlaxUNetMidBlockaDCrossAttn, FlaxUpBlockaD, ) @flax.struct.dataclass class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : jnp.ndarray @flax_register_to_config class a_ ( nn.Module , a__ , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = 32 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : Tuple[str] = ( "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D", ) __SCREAMING_SNAKE_CASE : Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D") __SCREAMING_SNAKE_CASE : Union[bool, Tuple[bool]] = False __SCREAMING_SNAKE_CASE : Tuple[int] = (320, 640, 1280, 1280) __SCREAMING_SNAKE_CASE : int = 2 __SCREAMING_SNAKE_CASE : Union[int, Tuple[int]] = 8 __SCREAMING_SNAKE_CASE : Optional[Union[int, Tuple[int]]] = None __SCREAMING_SNAKE_CASE : int = 1280 __SCREAMING_SNAKE_CASE : float = 0.0 __SCREAMING_SNAKE_CASE : bool = False __SCREAMING_SNAKE_CASE : jnp.dtype = jnp.floataa __SCREAMING_SNAKE_CASE : bool = True __SCREAMING_SNAKE_CASE : int = 0 __SCREAMING_SNAKE_CASE : bool = False def __lowerCAmelCase ( self , _lowerCamelCase ) ->FrozenDict: # init input tensors SCREAMING_SNAKE_CASE : List[Any] = (1, self.in_channels, self.sample_size, self.sample_size) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros(_lowerCamelCase , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Tuple = jnp.ones((1,) , dtype=jnp.intaa ) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros((1, 1, self.cross_attention_dim) , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = jax.random.split(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = {'''params''': params_rng, '''dropout''': dropout_rng} return self.init(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase )["params"] def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : List[str] = self.block_out_channels SCREAMING_SNAKE_CASE : Optional[int] = block_out_channels[0] * 4 if self.num_attention_heads is not None: raise ValueError( '''At the moment it is not possible to define the number of attention heads via `num_attention_heads` because of a naming issue as described in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131. Passing `num_attention_heads` will only be supported in diffusers v0.19.''' ) # If `num_attention_heads` is not defined (which is the case for most models) # it will default to `attention_head_dim`. This looks weird upon first reading it and it is. # The reason for this behavior is to correct for incorrectly named variables that were introduced # when this library was created. The incorrect naming was only discovered much later in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131 # Changing `attention_head_dim` to `num_attention_heads` for 40,000+ configurations is too backwards breaking # which is why we correct for the naming here. SCREAMING_SNAKE_CASE : List[str] = self.num_attention_heads or self.attention_head_dim # input SCREAMING_SNAKE_CASE : Optional[int] = nn.Conv( block_out_channels[0] , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) # time SCREAMING_SNAKE_CASE : Tuple = FlaxTimesteps( block_out_channels[0] , flip_sin_to_cos=self.flip_sin_to_cos , freq_shift=self.config.freq_shift ) SCREAMING_SNAKE_CASE : Dict = FlaxTimestepEmbedding(_lowerCamelCase , dtype=self.dtype ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.only_cross_attention if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (only_cross_attention,) * len(self.down_block_types ) if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (num_attention_heads,) * len(self.down_block_types ) # down SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Optional[Any] = block_out_channels[0] for i, down_block_type in enumerate(self.down_block_types ): SCREAMING_SNAKE_CASE : str = output_channel SCREAMING_SNAKE_CASE : int = block_out_channels[i] SCREAMING_SNAKE_CASE : List[Any] = i == len(_lowerCamelCase ) - 1 if down_block_type == "CrossAttnDownBlock2D": SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxCrossAttnDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , num_attention_heads=num_attention_heads[i] , add_downsample=not is_final_block , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , add_downsample=not is_final_block , dtype=self.dtype , ) down_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = down_blocks # mid SCREAMING_SNAKE_CASE : int = FlaxUNetMidBlockaDCrossAttn( in_channels=block_out_channels[-1] , dropout=self.dropout , num_attention_heads=num_attention_heads[-1] , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) # up SCREAMING_SNAKE_CASE : str = [] SCREAMING_SNAKE_CASE : str = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = reversed_block_out_channels[0] for i, up_block_type in enumerate(self.up_block_types ): SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Union[str, Any] = reversed_block_out_channels[i] SCREAMING_SNAKE_CASE : Tuple = reversed_block_out_channels[min(i + 1 , len(_lowerCamelCase ) - 1 )] SCREAMING_SNAKE_CASE : Dict = i == len(_lowerCamelCase ) - 1 if up_block_type == "CrossAttnUpBlock2D": SCREAMING_SNAKE_CASE : str = FlaxCrossAttnUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , num_attention_heads=reversed_num_attention_heads[i] , add_upsample=not is_final_block , dropout=self.dropout , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Optional[int] = FlaxUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , add_upsample=not is_final_block , dropout=self.dropout , dtype=self.dtype , ) up_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Tuple = up_blocks # out SCREAMING_SNAKE_CASE : Any = nn.GroupNorm(num_groups=32 , epsilon=1e-5 ) SCREAMING_SNAKE_CASE : Any = nn.Conv( self.out_channels , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) def __call__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase = True , _lowerCamelCase = False , ) ->Union[FlaxUNetaDConditionOutput, Tuple]: # 1. time if not isinstance(_lowerCamelCase , jnp.ndarray ): SCREAMING_SNAKE_CASE : int = jnp.array([timesteps] , dtype=jnp.intaa ) elif isinstance(_lowerCamelCase , jnp.ndarray ) and len(timesteps.shape ) == 0: SCREAMING_SNAKE_CASE : List[str] = timesteps.astype(dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.expand_dims(_lowerCamelCase , 0 ) SCREAMING_SNAKE_CASE : List[str] = self.time_proj(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = self.time_embedding(_lowerCamelCase ) # 2. pre-process SCREAMING_SNAKE_CASE : int = jnp.transpose(_lowerCamelCase , (0, 2, 3, 1) ) SCREAMING_SNAKE_CASE : List[Any] = self.conv_in(_lowerCamelCase ) # 3. down SCREAMING_SNAKE_CASE : Optional[int] = (sample,) for down_block in self.down_blocks: if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , deterministic=not train ) down_block_res_samples += res_samples if down_block_additional_residuals is not None: SCREAMING_SNAKE_CASE : int = () for down_block_res_sample, down_block_additional_residual in zip( _lowerCamelCase , _lowerCamelCase ): down_block_res_sample += down_block_additional_residual new_down_block_res_samples += (down_block_res_sample,) SCREAMING_SNAKE_CASE : Dict = new_down_block_res_samples # 4. mid SCREAMING_SNAKE_CASE : Optional[Any] = self.mid_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) if mid_block_additional_residual is not None: sample += mid_block_additional_residual # 5. up for up_block in self.up_blocks: SCREAMING_SNAKE_CASE : Optional[Any] = down_block_res_samples[-(self.layers_per_block + 1) :] SCREAMING_SNAKE_CASE : Optional[int] = down_block_res_samples[: -(self.layers_per_block + 1)] if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = up_block( _lowerCamelCase , temb=_lowerCamelCase , encoder_hidden_states=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train , ) else: SCREAMING_SNAKE_CASE : Optional[int] = up_block(_lowerCamelCase , temb=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train ) # 6. post-process SCREAMING_SNAKE_CASE : Optional[int] = self.conv_norm_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.silu(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self.conv_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.transpose(_lowerCamelCase , (0, 3, 1, 2) ) if not return_dict: return (sample,) return FlaxUNetaDConditionOutput(sample=_lowerCamelCase )
313
1
from ...configuration_utils import PretrainedConfig from ...utils import logging a__ : Dict = logging.get_logger(__name__) a__ : Union[str, Any] = { '''google/switch-base-8''': '''https://huggingface.co/google/switch-base-8/blob/main/config.json''', } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = 'switch_transformers' __SCREAMING_SNAKE_CASE : List[Any] = ['past_key_values'] __SCREAMING_SNAKE_CASE : str = {'hidden_size': 'd_model', 'num_attention_heads': 'num_heads', 'num_hidden_layers': 'num_layers'} def __init__( self , _lowerCamelCase=3_2128 , _lowerCamelCase=768 , _lowerCamelCase=64 , _lowerCamelCase=2048 , _lowerCamelCase=64 , _lowerCamelCase=12 , _lowerCamelCase=3 , _lowerCamelCase=12 , _lowerCamelCase=3 , _lowerCamelCase=12 , _lowerCamelCase=8 , _lowerCamelCase=False , _lowerCamelCase=0.0_1 , _lowerCamelCase="float32" , _lowerCamelCase=False , _lowerCamelCase=32 , _lowerCamelCase=128 , _lowerCamelCase=0.1 , _lowerCamelCase=1e-6 , _lowerCamelCase=0.0_0_1 , _lowerCamelCase=0.0_0_1 , _lowerCamelCase=1.0 , _lowerCamelCase="relu" , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=True , _lowerCamelCase=0 , _lowerCamelCase=1 , **_lowerCamelCase , ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : int = vocab_size SCREAMING_SNAKE_CASE : int = d_model SCREAMING_SNAKE_CASE : Union[str, Any] = d_kv SCREAMING_SNAKE_CASE : str = d_ff SCREAMING_SNAKE_CASE : Dict = num_sparse_encoder_layers SCREAMING_SNAKE_CASE : Optional[Any] = num_layers SCREAMING_SNAKE_CASE : Union[str, Any] = ( num_decoder_layers if num_decoder_layers is not None else self.num_layers ) # default = symmetry SCREAMING_SNAKE_CASE : Optional[Any] = num_sparse_decoder_layers # This tells us, each how many encoder layer we'll have to set a sparse layer. if self.num_sparse_encoder_layers > 0: SCREAMING_SNAKE_CASE : Any = self.num_layers // self.num_sparse_encoder_layers else: SCREAMING_SNAKE_CASE : Union[str, Any] = self.num_layers # HACK: this will create 0 sparse layers # This tells us, each how many encoder layer we'll have to set a sparse layer. if self.num_sparse_decoder_layers > 0: SCREAMING_SNAKE_CASE : int = self.num_decoder_layers // self.num_sparse_decoder_layers else: SCREAMING_SNAKE_CASE : Any = self.num_decoder_layers # HACK: this will create 0 sparse layers SCREAMING_SNAKE_CASE : Tuple = num_heads SCREAMING_SNAKE_CASE : Any = num_experts SCREAMING_SNAKE_CASE : str = expert_capacity SCREAMING_SNAKE_CASE : Dict = router_bias SCREAMING_SNAKE_CASE : Union[str, Any] = router_jitter_noise if router_dtype not in ["float32", "float16", "bfloat16"]: raise ValueError(F"""`router_dtype` must be one of 'float32', 'float16' or 'bfloat16', got {router_dtype}""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = router_dtype SCREAMING_SNAKE_CASE : int = router_ignore_padding_tokens SCREAMING_SNAKE_CASE : Optional[Any] = relative_attention_num_buckets SCREAMING_SNAKE_CASE : str = relative_attention_max_distance SCREAMING_SNAKE_CASE : Dict = dropout_rate SCREAMING_SNAKE_CASE : List[str] = layer_norm_epsilon SCREAMING_SNAKE_CASE : List[Any] = initializer_factor SCREAMING_SNAKE_CASE : Optional[Any] = feed_forward_proj SCREAMING_SNAKE_CASE : Optional[Any] = use_cache SCREAMING_SNAKE_CASE : Dict = add_router_probs SCREAMING_SNAKE_CASE : Tuple = router_z_loss_coef SCREAMING_SNAKE_CASE : Any = router_aux_loss_coef SCREAMING_SNAKE_CASE : Dict = self.feed_forward_proj.split('''-''' ) SCREAMING_SNAKE_CASE : Optional[Any] = act_info[-1] SCREAMING_SNAKE_CASE : Union[str, Any] = act_info[0] == '''gated''' if len(_lowerCamelCase ) > 1 and act_info[0] != "gated" or len(_lowerCamelCase ) > 2: raise ValueError( F"""`feed_forward_proj`: {feed_forward_proj} is not a valid activation function of the dense layer.""" '''Please make sure `feed_forward_proj` is of the format `gated-{ACT_FN}` or `{ACT_FN}`, e.g. ''' '''\'gated-gelu\' or \'relu\'''' ) # for backwards compatibility if feed_forward_proj == "gated-gelu": SCREAMING_SNAKE_CASE : str = '''gelu_new''' super().__init__( pad_token_id=_lowerCamelCase , eos_token_id=_lowerCamelCase , is_encoder_decoder=_lowerCamelCase , **_lowerCamelCase , )
313
import gc import random import unittest import numpy as np import torch from transformers import ( CLIPImageProcessor, CLIPTextConfig, CLIPTextModel, CLIPTokenizer, CLIPVisionConfig, CLIPVisionModelWithProjection, ) from diffusers import AutoencoderKL, DDIMScheduler, DDPMScheduler, StableUnCLIPImgaImgPipeline, UNetaDConditionModel from diffusers.pipelines.pipeline_utils import DiffusionPipeline from diffusers.pipelines.stable_diffusion.stable_unclip_image_normalizer import StableUnCLIPImageNormalizer from diffusers.utils.import_utils import is_xformers_available from diffusers.utils.testing_utils import ( enable_full_determinism, floats_tensor, load_image, load_numpy, require_torch_gpu, skip_mps, slow, torch_device, ) from ..pipeline_params import TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS from ..test_pipelines_common import ( PipelineKarrasSchedulerTesterMixin, PipelineLatentTesterMixin, PipelineTesterMixin, assert_mean_pixel_difference, ) enable_full_determinism() class a_ ( a__ , a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = StableUnCLIPImgaImgPipeline __SCREAMING_SNAKE_CASE : List[str] = TEXT_GUIDED_IMAGE_VARIATION_PARAMS __SCREAMING_SNAKE_CASE : List[Any] = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS __SCREAMING_SNAKE_CASE : Any = frozenset( [] ) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess __SCREAMING_SNAKE_CASE : Tuple = frozenset([] ) def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[Any] = 32 SCREAMING_SNAKE_CASE : Tuple = embedder_hidden_size # image encoding components SCREAMING_SNAKE_CASE : int = CLIPImageProcessor(crop_size=32 , size=32 ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = CLIPVisionModelWithProjection( CLIPVisionConfig( hidden_size=_lowerCamelCase , projection_dim=_lowerCamelCase , num_hidden_layers=5 , num_attention_heads=4 , image_size=32 , intermediate_size=37 , patch_size=1 , ) ) # regular denoising components torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = StableUnCLIPImageNormalizer(embedding_dim=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = DDPMScheduler(beta_schedule='''squaredcos_cap_v2''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = CLIPTextModel( CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=_lowerCamelCase , projection_dim=32 , intermediate_size=37 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , ) ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = UNetaDConditionModel( sample_size=32 , in_channels=4 , out_channels=4 , down_block_types=('''CrossAttnDownBlock2D''', '''DownBlock2D''') , up_block_types=('''UpBlock2D''', '''CrossAttnUpBlock2D''') , block_out_channels=(32, 64) , attention_head_dim=(2, 4) , class_embed_type='''projection''' , projection_class_embeddings_input_dim=embedder_projection_dim * 2 , cross_attention_dim=_lowerCamelCase , layers_per_block=1 , upcast_attention=_lowerCamelCase , use_linear_projection=_lowerCamelCase , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = DDIMScheduler( beta_schedule='''scaled_linear''' , beta_start=0.0_0_0_8_5 , beta_end=0.0_1_2 , prediction_type='''v_prediction''' , set_alpha_to_one=_lowerCamelCase , steps_offset=1 , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Dict = AutoencoderKL() SCREAMING_SNAKE_CASE : Optional[Any] = { # image encoding components '''feature_extractor''': feature_extractor, '''image_encoder''': image_encoder.eval(), # image noising components '''image_normalizer''': image_normalizer.eval(), '''image_noising_scheduler''': image_noising_scheduler, # regular denoising components '''tokenizer''': tokenizer, '''text_encoder''': text_encoder.eval(), '''unet''': unet.eval(), '''scheduler''': scheduler, '''vae''': vae.eval(), } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 , _lowerCamelCase=True ) ->Optional[int]: if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, 3, 32, 32) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if pil_image: SCREAMING_SNAKE_CASE : Any = input_image * 0.5 + 0.5 SCREAMING_SNAKE_CASE : int = input_image.clamp(0 , 1 ) SCREAMING_SNAKE_CASE : Union[str, Any] = input_image.cpu().permute(0 , 2 , 3 , 1 ).float().numpy() SCREAMING_SNAKE_CASE : List[str] = DiffusionPipeline.numpy_to_pil(_lowerCamelCase )[0] return { "prompt": "An anime racoon running a marathon", "image": input_image, "generator": generator, "num_inference_steps": 2, "output_type": "np", } @skip_mps def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Any = '''cpu''' # ensure determinism for the device-dependent torch.Generator SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = StableUnCLIPImgaImgPipeline(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = sd_pipe.to(_lowerCamelCase ) sd_pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = self.get_dummy_inputs(_lowerCamelCase ) inputs.update({'''image_embeds''': None} ) SCREAMING_SNAKE_CASE : Optional[int] = sd_pipe(**_lowerCamelCase ).images SCREAMING_SNAKE_CASE : Dict = image[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) SCREAMING_SNAKE_CASE : Tuple = np.array([0.3_8_7_2, 0.7_2_2_4, 0.5_6_0_1, 0.4_7_4_1, 0.6_8_7_2, 0.5_8_1_4, 0.4_6_3_6, 0.3_8_6_7, 0.5_0_7_8] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : str = torch_device in ['''cpu''', '''mps'''] self._test_attention_slicing_forward_pass(test_max_difference=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Tuple = torch_device in ['''cpu''', '''mps'''] self._test_inference_batch_single_identical(test_max_difference=_lowerCamelCase ) @unittest.skipIf( torch_device != '''cuda''' or not is_xformers_available() , reason='''XFormers attention is only available with CUDA and `xformers` installed''' , ) def __lowerCAmelCase ( self ) ->Optional[int]: self._test_xformers_attention_forwardGenerator_pass(test_max_difference=_lowerCamelCase ) @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->int: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : List[Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_l_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[Any] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-l-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : List[str] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[int] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_h_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[str] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) torch.cuda.empty_cache() torch.cuda.reset_max_memory_allocated() torch.cuda.reset_peak_memory_stats() SCREAMING_SNAKE_CASE : str = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Dict = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Dict = pipe( _lowerCamelCase , '''anime turtle''' , num_inference_steps=2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = torch.cuda.max_memory_allocated() # make sure that less than 7 GB is allocated assert mem_bytes < 7 * 10**9
313
1
import contextlib import importlib import io import unittest import transformers # Try to import everything from transformers to ensure every object can be loaded. from transformers import * # noqa F406 from transformers.testing_utils import DUMMY_UNKNOWN_IDENTIFIER, require_flax, require_tf, require_torch from transformers.utils import ContextManagers, find_labels, is_flax_available, is_tf_available, is_torch_available if is_torch_available(): from transformers import BertForPreTraining, BertForQuestionAnswering, BertForSequenceClassification if is_tf_available(): from transformers import TFBertForPreTraining, TFBertForQuestionAnswering, TFBertForSequenceClassification if is_flax_available(): from transformers import FlaxBertForPreTraining, FlaxBertForQuestionAnswering, FlaxBertForSequenceClassification a__ : Dict = DUMMY_UNKNOWN_IDENTIFIER # An actual model hosted on huggingface.co a__ : List[Any] = '''main''' # Default branch name a__ : Dict = '''f2c752cfc5c0ab6f4bdec59acea69eefbee381c2''' # One particular commit (not the top of `main`) a__ : List[Any] = '''aaaaaaa''' # This commit does not exist, so we should 404. a__ : List[Any] = '''d9e9f15bc825e4b2c9249e9578f884bbcb5e3684''' # Sha-1 of config.json on the top of `main`, for checking purposes a__ : Dict = '''4b243c475af8d0a7754e87d7d096c92e5199ec2fe168a2ee7998e3b8e9bcb1d3''' @contextlib.contextmanager def UpperCAmelCase_( ): """simple docstring""" print('''Welcome!''' ) yield print('''Bye!''' ) @contextlib.contextmanager def UpperCAmelCase_( ): """simple docstring""" print('''Bonjour!''' ) yield print('''Au revoir!''' ) class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Any: # If the spec is missing, importlib would not be able to import the module dynamically. assert transformers.__spec__ is not None assert importlib.util.find_spec('''transformers''' ) is not None class a_ ( unittest.TestCase ): """simple docstring""" @unittest.mock.patch('''sys.stdout''' , new_callable=io.StringIO ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: with ContextManagers([] ): print('''Transformers are awesome!''' ) # The print statement adds a new line at the end of the output self.assertEqual(mock_stdout.getvalue() , '''Transformers are awesome!\n''' ) @unittest.mock.patch('''sys.stdout''' , new_callable=io.StringIO ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: with ContextManagers([context_en()] ): print('''Transformers are awesome!''' ) # The output should be wrapped with an English welcome and goodbye self.assertEqual(mock_stdout.getvalue() , '''Welcome!\nTransformers are awesome!\nBye!\n''' ) @unittest.mock.patch('''sys.stdout''' , new_callable=io.StringIO ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->Tuple: with ContextManagers([context_fr(), context_en()] ): print('''Transformers are awesome!''' ) # The output should be wrapped with an English and French welcome and goodbye self.assertEqual(mock_stdout.getvalue() , '''Bonjour!\nWelcome!\nTransformers are awesome!\nBye!\nAu revoir!\n''' ) @require_torch def __lowerCAmelCase ( self ) ->Tuple: self.assertEqual(find_labels(_lowerCamelCase ) , ['''labels'''] ) self.assertEqual(find_labels(_lowerCamelCase ) , ['''labels''', '''next_sentence_label'''] ) self.assertEqual(find_labels(_lowerCamelCase ) , ['''start_positions''', '''end_positions'''] ) class a_ ( a__ ): """simple docstring""" pass self.assertEqual(find_labels(_lowerCamelCase ) , ['''labels'''] ) @require_tf def __lowerCAmelCase ( self ) ->Any: self.assertEqual(find_labels(_lowerCamelCase ) , ['''labels'''] ) self.assertEqual(find_labels(_lowerCamelCase ) , ['''labels''', '''next_sentence_label'''] ) self.assertEqual(find_labels(_lowerCamelCase ) , ['''start_positions''', '''end_positions'''] ) class a_ ( a__ ): """simple docstring""" pass self.assertEqual(find_labels(_lowerCamelCase ) , ['''labels'''] ) @require_flax def __lowerCAmelCase ( self ) ->str: # Flax models don't have labels self.assertEqual(find_labels(_lowerCamelCase ) , [] ) self.assertEqual(find_labels(_lowerCamelCase ) , [] ) self.assertEqual(find_labels(_lowerCamelCase ) , [] ) class a_ ( a__ ): """simple docstring""" pass self.assertEqual(find_labels(_lowerCamelCase ) , [] )
313
from abc import ABC, abstractmethod from typing import List, Optional class a_ ( a__ ): """simple docstring""" def __init__( self ) ->List[str]: # test for the above condition self.test() def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = False while not completed: if counter == 1: self.reset() SCREAMING_SNAKE_CASE : List[Any] = self.advance() if not self.does_advance(_lowerCamelCase ): raise Exception( '''Custom Constraint is not defined correctly. self.does_advance(self.advance()) must be true.''' ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.update(_lowerCamelCase ) counter += 1 if counter > 1_0000: raise Exception('''update() does not fulfill the constraint.''' ) if self.remaining() != 0: raise Exception('''Custom Constraint is not defined correctly.''' ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[int]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Union[str, Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Any: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->int: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`token_ids` has to be a non-empty list, but is {token_ids}.""" ) if any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ): raise ValueError(F"""Each list in `token_ids` has to be a list of positive integers, but is {token_ids}.""" ) SCREAMING_SNAKE_CASE : Optional[Any] = token_ids SCREAMING_SNAKE_CASE : Union[str, Any] = len(self.token_ids ) SCREAMING_SNAKE_CASE : Any = -1 # the index of the currently fulfilled step SCREAMING_SNAKE_CASE : Any = False def __lowerCAmelCase ( self ) ->List[Any]: if self.completed: return None return self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) if self.completed: return False return token_id == self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[str]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : str = False SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = False if self.does_advance(_lowerCamelCase ): self.fulfilled_idx += 1 SCREAMING_SNAKE_CASE : str = True if self.fulfilled_idx == (self.seqlen - 1): SCREAMING_SNAKE_CASE : Any = True SCREAMING_SNAKE_CASE : Union[str, Any] = completed else: # failed to make progress. SCREAMING_SNAKE_CASE : Dict = True self.reset() return stepped, completed, reset def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = 0 def __lowerCAmelCase ( self ) ->Any: return self.seqlen - (self.fulfilled_idx + 1) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Dict: SCREAMING_SNAKE_CASE : Any = PhrasalConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : Dict = self.seqlen SCREAMING_SNAKE_CASE : int = self.fulfilled_idx SCREAMING_SNAKE_CASE : Tuple = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=True ) ->Dict: SCREAMING_SNAKE_CASE : Any = max([len(_lowerCamelCase ) for one in nested_token_ids] ) SCREAMING_SNAKE_CASE : List[str] = {} for token_ids in nested_token_ids: SCREAMING_SNAKE_CASE : Optional[Any] = root for tidx, token_id in enumerate(_lowerCamelCase ): if token_id not in level: SCREAMING_SNAKE_CASE : Any = {} SCREAMING_SNAKE_CASE : Tuple = level[token_id] if no_subsets and self.has_subsets(_lowerCamelCase , _lowerCamelCase ): raise ValueError( '''Each list in `nested_token_ids` can\'t be a complete subset of another list, but is''' F""" {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = root def __lowerCAmelCase ( self , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = self.trie for current_token in current_seq: SCREAMING_SNAKE_CASE : int = start[current_token] SCREAMING_SNAKE_CASE : Optional[int] = list(start.keys() ) return next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.next_tokens(_lowerCamelCase ) return len(_lowerCamelCase ) == 0 def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = list(root.values() ) if len(_lowerCamelCase ) == 0: return 1 else: return sum([self.count_leaves(_lowerCamelCase ) for nn in next_nodes] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = self.count_leaves(_lowerCamelCase ) return len(_lowerCamelCase ) != leaf_count class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->str: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`nested_token_ids` has to be a non-empty list, but is {nested_token_ids}.""" ) if any(not isinstance(_lowerCamelCase , _lowerCamelCase ) for token_ids in nested_token_ids ): raise ValueError(F"""`nested_token_ids` has to be a list of lists, but is {nested_token_ids}.""" ) if any( any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ) for token_ids in nested_token_ids ): raise ValueError( F"""Each list in `nested_token_ids` has to be a list of positive integers, but is {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = DisjunctiveTrie(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = nested_token_ids SCREAMING_SNAKE_CASE : Optional[int] = self.trie.max_height SCREAMING_SNAKE_CASE : Union[str, Any] = [] SCREAMING_SNAKE_CASE : Optional[int] = False def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : str = self.trie.next_tokens(self.current_seq ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : List[str] = self.trie.next_tokens(self.current_seq ) return token_id in next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : int = False SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = False if self.does_advance(_lowerCamelCase ): self.current_seq.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = True else: SCREAMING_SNAKE_CASE : Dict = True self.reset() SCREAMING_SNAKE_CASE : Any = self.trie.reached_leaf(self.current_seq ) SCREAMING_SNAKE_CASE : List[Any] = completed return stepped, completed, reset def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = [] def __lowerCAmelCase ( self ) ->Optional[Any]: if self.completed: # since this can be completed without reaching max height return 0 else: return self.seqlen - len(self.current_seq ) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->List[str]: SCREAMING_SNAKE_CASE : str = DisjunctiveConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : str = self.seqlen SCREAMING_SNAKE_CASE : int = self.current_seq SCREAMING_SNAKE_CASE : Optional[int] = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[Any] = constraints # max # of steps required to fulfill a given constraint SCREAMING_SNAKE_CASE : str = max([c.seqlen for c in constraints] ) SCREAMING_SNAKE_CASE : List[str] = len(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = False self.init_state() def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Any = [] SCREAMING_SNAKE_CASE : List[Any] = None SCREAMING_SNAKE_CASE : Tuple = [constraint.copy(stateful=_lowerCamelCase ) for constraint in self.constraints] def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = 0 if self.inprogress_constraint: # extra points for having a constraint mid-fulfilled add += self.max_seqlen - self.inprogress_constraint.remaining() return (len(self.complete_constraints ) * self.max_seqlen) + add def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Tuple = [] if self.inprogress_constraint is None: for constraint in self.pending_constraints: # "pending" == "unfulfilled yet" SCREAMING_SNAKE_CASE : Optional[int] = constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = self.inprogress_constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: self.init_state() if token_ids is not None: for token in token_ids: # completes or steps **one** constraint SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.add(_lowerCamelCase ) # the entire list of constraints are fulfilled if self.completed: break def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` should be an `int`, but is `{token_id}`.""" ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = False, False if self.completed: SCREAMING_SNAKE_CASE : List[str] = True SCREAMING_SNAKE_CASE : Optional[int] = False return complete, stepped if self.inprogress_constraint is not None: # In the middle of fulfilling a constraint. If the `token_id` *does* makes an incremental progress to current # job, simply update the state SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.inprogress_constraint.update(_lowerCamelCase ) if reset: # 1. If the next token breaks the progress, then we must restart. # e.g. constraint = "I love pies" and sequence so far is "I love" but `token_id` == "books". # But that doesn't mean we self.init_state(), since we only reset the state for this particular # constraint, not the full list of constraints. self.pending_constraints.append(self.inprogress_constraint.copy(stateful=_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = None if complete: # 2. If the next token completes the constraint, move it to completed list, set # inprogress to None. If there are no pending constraints either, then this full list of constraints # is complete. self.complete_constraints.append(self.inprogress_constraint ) SCREAMING_SNAKE_CASE : str = None if len(self.pending_constraints ) == 0: # we're done! SCREAMING_SNAKE_CASE : Optional[Any] = True else: # Not in the middle of fulfilling a constraint. So does this `token_id` helps us step towards any of our list # of constraints? for cidx, pending_constraint in enumerate(self.pending_constraints ): if pending_constraint.does_advance(_lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = pending_constraint.update(_lowerCamelCase ) if not stepped: raise Exception( '''`constraint.update(token_id)` is not yielding incremental progress, ''' '''even though `constraint.does_advance(token_id)` is true.''' ) if complete: self.complete_constraints.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = None if not complete and stepped: SCREAMING_SNAKE_CASE : Optional[Any] = pending_constraint if complete or stepped: # If we made any progress at all, then it's at least not a "pending constraint". SCREAMING_SNAKE_CASE : Union[str, Any] = ( self.pending_constraints[:cidx] + self.pending_constraints[cidx + 1 :] ) if len(self.pending_constraints ) == 0 and self.inprogress_constraint is None: # If there's no longer any pending after this and no inprogress either, then we must be # complete. SCREAMING_SNAKE_CASE : str = True break # prevent accidentally stepping through multiple constraints with just one token. return complete, stepped def __lowerCAmelCase ( self , _lowerCamelCase=True ) ->str: SCREAMING_SNAKE_CASE : Dict = ConstraintListState(self.constraints ) # we actually never though self.constraints objects # throughout this process. So it's at initialization state. if stateful: SCREAMING_SNAKE_CASE : str = [ constraint.copy(stateful=_lowerCamelCase ) for constraint in self.complete_constraints ] if self.inprogress_constraint is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.inprogress_constraint.copy(stateful=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = [constraint.copy() for constraint in self.pending_constraints] return new_state
313
1
import json import os import sys import tempfile import unittest from pathlib import Path from shutil import copyfile from huggingface_hub import HfFolder, Repository, create_repo, delete_repo from requests.exceptions import HTTPError import transformers from transformers import ( CONFIG_MAPPING, FEATURE_EXTRACTOR_MAPPING, PROCESSOR_MAPPING, TOKENIZER_MAPPING, AutoConfig, AutoFeatureExtractor, AutoProcessor, AutoTokenizer, BertTokenizer, ProcessorMixin, WavaVecaConfig, WavaVecaFeatureExtractor, WavaVecaProcessor, ) from transformers.testing_utils import TOKEN, USER, get_tests_dir, is_staging_test from transformers.tokenization_utils import TOKENIZER_CONFIG_FILE from transformers.utils import FEATURE_EXTRACTOR_NAME, is_tokenizers_available sys.path.append(str(Path(__file__).parent.parent.parent.parent / '''utils''')) from test_module.custom_configuration import CustomConfig # noqa E402 from test_module.custom_feature_extraction import CustomFeatureExtractor # noqa E402 from test_module.custom_processing import CustomProcessor # noqa E402 from test_module.custom_tokenization import CustomTokenizer # noqa E402 a__ : List[Any] = get_tests_dir('''fixtures/dummy_feature_extractor_config.json''') a__ : Tuple = get_tests_dir('''fixtures/vocab.json''') a__ : List[str] = get_tests_dir('''fixtures''') class a_ ( unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = ['[UNK]', '[CLS]', '[SEP]', '[PAD]', '[MASK]', 'bla', 'blou'] def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : str = 0 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[Any] = AutoProcessor.from_pretrained('''facebook/wav2vec2-base-960h''' ) self.assertIsInstance(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: with tempfile.TemporaryDirectory() as tmpdirname: SCREAMING_SNAKE_CASE : int = WavaVecaConfig() SCREAMING_SNAKE_CASE : Tuple = AutoProcessor.from_pretrained('''facebook/wav2vec2-base-960h''' ) # save in new folder model_config.save_pretrained(_lowerCamelCase ) processor.save_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = AutoProcessor.from_pretrained(_lowerCamelCase ) self.assertIsInstance(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[int]: with tempfile.TemporaryDirectory() as tmpdirname: # copy relevant files copyfile(_lowerCamelCase , os.path.join(_lowerCamelCase , _lowerCamelCase ) ) copyfile(_lowerCamelCase , os.path.join(_lowerCamelCase , '''vocab.json''' ) ) SCREAMING_SNAKE_CASE : Dict = AutoProcessor.from_pretrained(_lowerCamelCase ) self.assertIsInstance(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: with tempfile.TemporaryDirectory() as tmpdirname: SCREAMING_SNAKE_CASE : List[str] = WavaVecaFeatureExtractor() SCREAMING_SNAKE_CASE : List[Any] = AutoTokenizer.from_pretrained('''facebook/wav2vec2-base-960h''' ) SCREAMING_SNAKE_CASE : str = WavaVecaProcessor(_lowerCamelCase , _lowerCamelCase ) # save in new folder processor.save_pretrained(_lowerCamelCase ) # drop `processor_class` in tokenizer with open(os.path.join(_lowerCamelCase , _lowerCamelCase ) , '''r''' ) as f: SCREAMING_SNAKE_CASE : int = json.load(_lowerCamelCase ) config_dict.pop('''processor_class''' ) with open(os.path.join(_lowerCamelCase , _lowerCamelCase ) , '''w''' ) as f: f.write(json.dumps(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = AutoProcessor.from_pretrained(_lowerCamelCase ) self.assertIsInstance(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: with tempfile.TemporaryDirectory() as tmpdirname: SCREAMING_SNAKE_CASE : List[str] = WavaVecaFeatureExtractor() SCREAMING_SNAKE_CASE : str = AutoTokenizer.from_pretrained('''facebook/wav2vec2-base-960h''' ) SCREAMING_SNAKE_CASE : Tuple = WavaVecaProcessor(_lowerCamelCase , _lowerCamelCase ) # save in new folder processor.save_pretrained(_lowerCamelCase ) # drop `processor_class` in feature extractor with open(os.path.join(_lowerCamelCase , _lowerCamelCase ) , '''r''' ) as f: SCREAMING_SNAKE_CASE : Any = json.load(_lowerCamelCase ) config_dict.pop('''processor_class''' ) with open(os.path.join(_lowerCamelCase , _lowerCamelCase ) , '''w''' ) as f: f.write(json.dumps(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = AutoProcessor.from_pretrained(_lowerCamelCase ) self.assertIsInstance(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Tuple: with tempfile.TemporaryDirectory() as tmpdirname: SCREAMING_SNAKE_CASE : Any = WavaVecaConfig(processor_class='''Wav2Vec2Processor''' ) model_config.save_pretrained(_lowerCamelCase ) # copy relevant files copyfile(_lowerCamelCase , os.path.join(_lowerCamelCase , '''vocab.json''' ) ) # create emtpy sample processor with open(os.path.join(_lowerCamelCase , _lowerCamelCase ) , '''w''' ) as f: f.write('''{}''' ) SCREAMING_SNAKE_CASE : List[str] = AutoProcessor.from_pretrained(_lowerCamelCase ) self.assertIsInstance(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: # If remote code is not set, we will time out when asking whether to load the model. with self.assertRaises(_lowerCamelCase ): SCREAMING_SNAKE_CASE : List[str] = AutoProcessor.from_pretrained('''hf-internal-testing/test_dynamic_processor''' ) # If remote code is disabled, we can't load this config. with self.assertRaises(_lowerCamelCase ): SCREAMING_SNAKE_CASE : int = AutoProcessor.from_pretrained( '''hf-internal-testing/test_dynamic_processor''' , trust_remote_code=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = AutoProcessor.from_pretrained('''hf-internal-testing/test_dynamic_processor''' , trust_remote_code=_lowerCamelCase ) self.assertTrue(processor.special_attribute_present ) self.assertEqual(processor.__class__.__name__ , '''NewProcessor''' ) SCREAMING_SNAKE_CASE : List[Any] = processor.feature_extractor self.assertTrue(feature_extractor.special_attribute_present ) self.assertEqual(feature_extractor.__class__.__name__ , '''NewFeatureExtractor''' ) SCREAMING_SNAKE_CASE : Optional[int] = processor.tokenizer self.assertTrue(tokenizer.special_attribute_present ) if is_tokenizers_available(): self.assertEqual(tokenizer.__class__.__name__ , '''NewTokenizerFast''' ) # Test we can also load the slow version SCREAMING_SNAKE_CASE : Any = AutoProcessor.from_pretrained( '''hf-internal-testing/test_dynamic_processor''' , trust_remote_code=_lowerCamelCase , use_fast=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = new_processor.tokenizer self.assertTrue(new_tokenizer.special_attribute_present ) self.assertEqual(new_tokenizer.__class__.__name__ , '''NewTokenizer''' ) else: self.assertEqual(tokenizer.__class__.__name__ , '''NewTokenizer''' ) def __lowerCAmelCase ( self ) ->Any: try: AutoConfig.register('''custom''' , _lowerCamelCase ) AutoFeatureExtractor.register(_lowerCamelCase , _lowerCamelCase ) AutoTokenizer.register(_lowerCamelCase , slow_tokenizer_class=_lowerCamelCase ) AutoProcessor.register(_lowerCamelCase , _lowerCamelCase ) # Trying to register something existing in the Transformers library will raise an error with self.assertRaises(_lowerCamelCase ): AutoProcessor.register(_lowerCamelCase , _lowerCamelCase ) # Now that the config is registered, it can be used as any other config with the auto-API SCREAMING_SNAKE_CASE : Union[str, Any] = CustomFeatureExtractor.from_pretrained(_lowerCamelCase ) with tempfile.TemporaryDirectory() as tmp_dir: SCREAMING_SNAKE_CASE : List[str] = os.path.join(_lowerCamelCase , '''vocab.txt''' ) with open(_lowerCamelCase , '''w''' , encoding='''utf-8''' ) as vocab_writer: vocab_writer.write(''''''.join([x + '''\n''' for x in self.vocab_tokens] ) ) SCREAMING_SNAKE_CASE : List[str] = CustomTokenizer(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = CustomProcessor(_lowerCamelCase , _lowerCamelCase ) with tempfile.TemporaryDirectory() as tmp_dir: processor.save_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = AutoProcessor.from_pretrained(_lowerCamelCase ) self.assertIsInstance(_lowerCamelCase , _lowerCamelCase ) finally: if "custom" in CONFIG_MAPPING._extra_content: del CONFIG_MAPPING._extra_content["custom"] if CustomConfig in FEATURE_EXTRACTOR_MAPPING._extra_content: del FEATURE_EXTRACTOR_MAPPING._extra_content[CustomConfig] if CustomConfig in TOKENIZER_MAPPING._extra_content: del TOKENIZER_MAPPING._extra_content[CustomConfig] if CustomConfig in PROCESSOR_MAPPING._extra_content: del PROCESSOR_MAPPING._extra_content[CustomConfig] def __lowerCAmelCase ( self ) ->List[str]: class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Union[str, Any] = False class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = False class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 'AutoFeatureExtractor' __SCREAMING_SNAKE_CASE : Optional[Any] = 'AutoTokenizer' __SCREAMING_SNAKE_CASE : int = False try: AutoConfig.register('''custom''' , _lowerCamelCase ) AutoFeatureExtractor.register(_lowerCamelCase , _lowerCamelCase ) AutoTokenizer.register(_lowerCamelCase , slow_tokenizer_class=_lowerCamelCase ) AutoProcessor.register(_lowerCamelCase , _lowerCamelCase ) # If remote code is not set, the default is to use local classes. SCREAMING_SNAKE_CASE : str = AutoProcessor.from_pretrained('''hf-internal-testing/test_dynamic_processor''' ) self.assertEqual(processor.__class__.__name__ , '''NewProcessor''' ) self.assertFalse(processor.special_attribute_present ) self.assertFalse(processor.feature_extractor.special_attribute_present ) self.assertFalse(processor.tokenizer.special_attribute_present ) # If remote code is disabled, we load the local ones. SCREAMING_SNAKE_CASE : List[str] = AutoProcessor.from_pretrained( '''hf-internal-testing/test_dynamic_processor''' , trust_remote_code=_lowerCamelCase ) self.assertEqual(processor.__class__.__name__ , '''NewProcessor''' ) self.assertFalse(processor.special_attribute_present ) self.assertFalse(processor.feature_extractor.special_attribute_present ) self.assertFalse(processor.tokenizer.special_attribute_present ) # If remote is enabled, we load from the Hub. SCREAMING_SNAKE_CASE : List[Any] = AutoProcessor.from_pretrained( '''hf-internal-testing/test_dynamic_processor''' , trust_remote_code=_lowerCamelCase ) self.assertEqual(processor.__class__.__name__ , '''NewProcessor''' ) self.assertTrue(processor.special_attribute_present ) self.assertTrue(processor.feature_extractor.special_attribute_present ) self.assertTrue(processor.tokenizer.special_attribute_present ) finally: if "custom" in CONFIG_MAPPING._extra_content: del CONFIG_MAPPING._extra_content["custom"] if CustomConfig in FEATURE_EXTRACTOR_MAPPING._extra_content: del FEATURE_EXTRACTOR_MAPPING._extra_content[CustomConfig] if CustomConfig in TOKENIZER_MAPPING._extra_content: del TOKENIZER_MAPPING._extra_content[CustomConfig] if CustomConfig in PROCESSOR_MAPPING._extra_content: del PROCESSOR_MAPPING._extra_content[CustomConfig] def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : str = AutoProcessor.from_pretrained('''hf-internal-testing/tiny-random-bert''' ) self.assertEqual(processor.__class__.__name__ , '''BertTokenizerFast''' ) def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : str = AutoProcessor.from_pretrained('''hf-internal-testing/tiny-random-convnext''' ) self.assertEqual(processor.__class__.__name__ , '''ConvNextImageProcessor''' ) @is_staging_test class a_ ( unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = ['[UNK]', '[CLS]', '[SEP]', '[PAD]', '[MASK]', 'bla', 'blou'] @classmethod def __lowerCAmelCase ( cls ) ->str: SCREAMING_SNAKE_CASE : List[Any] = TOKEN HfFolder.save_token(_lowerCamelCase ) @classmethod def __lowerCAmelCase ( cls ) ->int: try: delete_repo(token=cls._token , repo_id='''test-processor''' ) except HTTPError: pass try: delete_repo(token=cls._token , repo_id='''valid_org/test-processor-org''' ) except HTTPError: pass try: delete_repo(token=cls._token , repo_id='''test-dynamic-processor''' ) except HTTPError: pass def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = WavaVecaProcessor.from_pretrained(_lowerCamelCase ) with tempfile.TemporaryDirectory() as tmp_dir: processor.save_pretrained( os.path.join(_lowerCamelCase , '''test-processor''' ) , push_to_hub=_lowerCamelCase , use_auth_token=self._token ) SCREAMING_SNAKE_CASE : Optional[Any] = WavaVecaProcessor.from_pretrained(F"""{USER}/test-processor""" ) for k, v in processor.feature_extractor.__dict__.items(): self.assertEqual(_lowerCamelCase , getattr(new_processor.feature_extractor , _lowerCamelCase ) ) self.assertDictEqual(new_processor.tokenizer.get_vocab() , processor.tokenizer.get_vocab() ) def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Dict = WavaVecaProcessor.from_pretrained(_lowerCamelCase ) with tempfile.TemporaryDirectory() as tmp_dir: processor.save_pretrained( os.path.join(_lowerCamelCase , '''test-processor-org''' ) , push_to_hub=_lowerCamelCase , use_auth_token=self._token , organization='''valid_org''' , ) SCREAMING_SNAKE_CASE : Any = WavaVecaProcessor.from_pretrained('''valid_org/test-processor-org''' ) for k, v in processor.feature_extractor.__dict__.items(): self.assertEqual(_lowerCamelCase , getattr(new_processor.feature_extractor , _lowerCamelCase ) ) self.assertDictEqual(new_processor.tokenizer.get_vocab() , processor.tokenizer.get_vocab() ) def __lowerCAmelCase ( self ) ->str: CustomFeatureExtractor.register_for_auto_class() CustomTokenizer.register_for_auto_class() CustomProcessor.register_for_auto_class() SCREAMING_SNAKE_CASE : Tuple = CustomFeatureExtractor.from_pretrained(_lowerCamelCase ) with tempfile.TemporaryDirectory() as tmp_dir: SCREAMING_SNAKE_CASE : Dict = os.path.join(_lowerCamelCase , '''vocab.txt''' ) with open(_lowerCamelCase , '''w''' , encoding='''utf-8''' ) as vocab_writer: vocab_writer.write(''''''.join([x + '''\n''' for x in self.vocab_tokens] ) ) SCREAMING_SNAKE_CASE : Dict = CustomTokenizer(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = CustomProcessor(_lowerCamelCase , _lowerCamelCase ) with tempfile.TemporaryDirectory() as tmp_dir: create_repo(F"""{USER}/test-dynamic-processor""" , token=self._token ) SCREAMING_SNAKE_CASE : Optional[Any] = Repository(_lowerCamelCase , clone_from=F"""{USER}/test-dynamic-processor""" , token=self._token ) processor.save_pretrained(_lowerCamelCase ) # This has added the proper auto_map field to the feature extractor config self.assertDictEqual( processor.feature_extractor.auto_map , { '''AutoFeatureExtractor''': '''custom_feature_extraction.CustomFeatureExtractor''', '''AutoProcessor''': '''custom_processing.CustomProcessor''', } , ) # This has added the proper auto_map field to the tokenizer config with open(os.path.join(_lowerCamelCase , '''tokenizer_config.json''' ) ) as f: SCREAMING_SNAKE_CASE : List[Any] = json.load(_lowerCamelCase ) self.assertDictEqual( tokenizer_config['''auto_map'''] , { '''AutoTokenizer''': ['''custom_tokenization.CustomTokenizer''', None], '''AutoProcessor''': '''custom_processing.CustomProcessor''', } , ) # The code has been copied from fixtures self.assertTrue(os.path.isfile(os.path.join(_lowerCamelCase , '''custom_feature_extraction.py''' ) ) ) self.assertTrue(os.path.isfile(os.path.join(_lowerCamelCase , '''custom_tokenization.py''' ) ) ) self.assertTrue(os.path.isfile(os.path.join(_lowerCamelCase , '''custom_processing.py''' ) ) ) repo.push_to_hub() SCREAMING_SNAKE_CASE : Tuple = AutoProcessor.from_pretrained(F"""{USER}/test-dynamic-processor""" , trust_remote_code=_lowerCamelCase ) # Can't make an isinstance check because the new_processor is from the CustomProcessor class of a dynamic module self.assertEqual(new_processor.__class__.__name__ , '''CustomProcessor''' )
313
import argparse import random import joblib import numpy as np import torch from igf.igf import ( SecondaryLearner, collect_objective_set, compute_perplexity, generate_datasets, load_gpta, recopy_gpta, set_seed, train_secondary_learner, ) from torch.utils.data import DataLoader, RandomSampler from transformers import GPTaLMHeadModel def UpperCAmelCase_( a__=32 , a__=10 , a__=100 , a__=1_026 , a__=True , a__="data/tokenized_stories_train_wikitext103.jbl" , a__="igf_context_pairs.jbl" , ): """simple docstring""" set_seed(3 ) # generate train_data and objective_set SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = generate_datasets( a__ , a__ , number=a__ , min_len=1_026 , trim=a__ ) # keeps model same across runs set_seed(4 ) # model, lm_optimizer, lm_scheduler = recopy_gpt2(model, device, max_steps) # store original model weights # can we train on GPU? SCREAMING_SNAKE_CASE : str = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) # load pretrained model SCREAMING_SNAKE_CASE : Dict = load_gpta('''gpt2''' ).to(a__ ) print('''computing perplexity on objective set''' ) SCREAMING_SNAKE_CASE : int = compute_perplexity(a__ , a__ , a__ ).item() print('''perplexity on objective set:''' , a__ ) # collect igf pairs and save to file demo.jbl collect_objective_set(a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ ) # clean up, delete model and data we don't need anymore del model, train_data, objective_set torch.cuda.empty_cache() def UpperCAmelCase_( a__ , a__=15 , a__=128 , a__=100 , a__="igf_model.pt" , ): """simple docstring""" set_seed(42 ) # Load pre-trained model SCREAMING_SNAKE_CASE : List[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) # Initialize secondary learner to use embedding weights of model SCREAMING_SNAKE_CASE : str = SecondaryLearner(a__ ) # Train secondary learner SCREAMING_SNAKE_CASE : Union[str, Any] = train_secondary_learner( a__ , a__ , max_epochs=a__ , batch_size=a__ , eval_freq=100 , igf_model_path=a__ , ) del model, secondary_learner_train_data torch.cuda.empty_cache() return secondary_learner def UpperCAmelCase_( a__ , a__ , a__ , a__=32 , a__=1_000 , a__=16 , a__=1.0 , a__=recopy_gpta , a__=None , a__=10 , a__="gpt2_finetuned.pt" , ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) SCREAMING_SNAKE_CASE : Optional[int] = RandomSampler(a__ ) SCREAMING_SNAKE_CASE : Dict = DataLoader(a__ , sampler=a__ ) SCREAMING_SNAKE_CASE : Tuple = max_steps // (len(a__ )) + 1 SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = torch.zeros((1, context_len) , dtype=torch.long , device=a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = recopy_model(a__ , a__ , a__ ) model.train() if secondary_learner is not None: secondary_learner.to(a__ ) secondary_learner.eval() SCREAMING_SNAKE_CASE : List[str] = [] SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Tuple = [] # Compute the performance of the transformer model at the beginning SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) for epoch in range(int(a__ ) ): for step, example in enumerate(a__ ): torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Union[str, Any] = random.randint(0 , example.size(2 ) - context_len - 1 ) SCREAMING_SNAKE_CASE : Optional[int] = example[0, 0, start : start + context_len] lm_optimizer.zero_grad() SCREAMING_SNAKE_CASE : Optional[Any] = model(a__ , labels=a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = True if secondary_learner is not None: SCREAMING_SNAKE_CASE : List[str] = secondary_learner.forward( torch.tensor(a__ , dtype=torch.long , device=a__ ).unsqueeze(0 ) )[0].item() observed_qs.append(float(a__ ) ) # Here we implement the simple non-constant threshold for the predicted IG(X) value # We will decay the selectivity of our secondary learner filter from # 1 standard deviation above average to 1 below average after 10 batches. if global_step == 10: SCREAMING_SNAKE_CASE : Dict = -1 if predicted_q < threshold: SCREAMING_SNAKE_CASE : str = False # If we passed the filter, add the context to the batch! if do_backprop: contexts.append(np.array(context.cpu() ) ) SCREAMING_SNAKE_CASE : List[str] = outputs[0] lm_loss.backward() examples += 1 del outputs # Once the batch is filled with enough contexts, backprop on the batch. if examples == batch_size: torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Any = 0 # Do LM backprop torch.nn.utils.clip_grad_norm_(model.parameters() , 3.0 ) lm_optimizer.step() lm_scheduler.step() # Update learning rate schedule global_step += 1 # Compute the performance of the transformer model at this batch if global_step % eval_interval == 0: SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) # Break out of the loop after 60 batches if max_steps > 0 and global_step > 60: break if max_steps > 0 and global_step > 60: break # save finetuned transformer model torch.save(model.state_dict() , a__ ) torch.cuda.empty_cache() # Do some cleaning up so we can reinitialize for the next run of this function del lm_optimizer del lm_scheduler return model def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = argparse.ArgumentParser(description='''Fine-tune a transformer model with IGF on a language modeling task''' ) # Required parameters parser.add_argument( '''--data_dir''' , default=a__ , type=a__ , required=a__ , help='''The input data dir. Should contain data files for WikiText.''' , ) parser.add_argument( '''--model_name_or_path''' , default=a__ , type=a__ , required=a__ , help='''Path to pretrained model or model identifier from huggingface.co/models''' , ) parser.add_argument( '''--data_file''' , type=a__ , default=a__ , help=( '''A jbl file containing tokenized data which can be split as objective dataset, ''' '''train_dataset and test_dataset.''' ) , ) parser.add_argument( '''--igf_data_file''' , type=a__ , default=a__ , help='''A jbl file containing the context and information gain pairs to train secondary learner.''' , ) parser.add_argument( '''--output_dir''' , default=a__ , type=a__ , required=a__ , help='''The output directory where the final fine-tuned model is stored.''' , ) parser.add_argument( '''--tokenizer_name''' , default=a__ , type=a__ , help='''Pretrained tokenizer name or path if not the same as model_name''' , ) parser.add_argument('''--seed''' , type=a__ , default=a__ , help='''A seed for reproducible training.''' ) parser.add_argument( '''--context_len''' , default=32 , type=a__ , help=( '''The maximum total input sequence length after tokenization. Sequences longer ''' '''than this will be truncated, sequences shorter will be padded.''' ) , ) parser.add_argument( '''--size_objective_set''' , default=100 , type=a__ , help='''number of articles that are long enough to be used as our objective set''' , ) parser.add_argument( '''--eval_freq''' , default=100 , type=a__ , help='''secondary model evaluation is triggered at eval_freq''' ) parser.add_argument('''--max_steps''' , default=1_000 , type=a__ , help='''To calculate training epochs''' ) parser.add_argument( '''--secondary_learner_batch_size''' , default=128 , type=a__ , help='''batch size of training data for secondary learner''' , ) parser.add_argument( '''--batch_size''' , default=16 , type=a__ , help='''batch size of training data of language model(gpt2) ''' ) parser.add_argument( '''--eval_interval''' , default=10 , type=a__ , help=( '''decay the selectivity of our secondary learner filter from''' '''1 standard deviation above average to 1 below average after 10 batches''' ) , ) parser.add_argument( '''--number''' , default=100 , type=a__ , help='''The number of examples split to be used as objective_set/test_data''' ) parser.add_argument( '''--min_len''' , default=1_026 , type=a__ , help='''The minimum length of the article to be used as objective set''' ) parser.add_argument( '''--secondary_learner_max_epochs''' , default=15 , type=a__ , help='''number of epochs to train secondary learner''' ) parser.add_argument('''--trim''' , default=a__ , type=a__ , help='''truncate the example if it exceeds context length''' ) parser.add_argument( '''--threshold''' , default=1.0 , type=a__ , help=( '''The threshold value used by secondary learner to filter the train_data and allow only''' ''' informative data as input to the model''' ) , ) parser.add_argument('''--finetuned_model_name''' , default='''gpt2_finetuned.pt''' , type=a__ , help='''finetuned_model_name''' ) parser.add_argument( '''--recopy_model''' , default=a__ , type=a__ , help='''Reset the model to the original pretrained GPT-2 weights after each iteration''' , ) # function calls # Collecting *n* pairs of context and information gain(X, IG(X)) for training the secondary learner generate_n_pairs( context_len=32 , max_steps=10 , size_objective_set=100 , min_len=1_026 , trim=a__ , data_file='''data/tokenized_stories_train_wikitext103.jbl''' , igf_data_file='''igf_context_pairs.jbl''' , ) # Load train data for secondary learner SCREAMING_SNAKE_CASE : List[Any] = joblib.load('''data/IGF_values.jbl''' ) # Train secondary learner SCREAMING_SNAKE_CASE : Tuple = training_secondary_learner( a__ , secondary_learner_max_epochs=15 , secondary_learner_batch_size=128 , eval_freq=100 , igf_model_path='''igf_model.pt''' , ) # load pretrained gpt2 model SCREAMING_SNAKE_CASE : Optional[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) set_seed(42 ) # Generate train and test data to train and evaluate gpt2 model SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = generate_datasets( context_len=32 , file='''data/tokenized_stories_train_wikitext103.jbl''' , number=100 , min_len=1_026 , trim=a__ ) # fine-tuning of the gpt2 model using igf (Information Gain Filtration) finetune( a__ , a__ , a__ , context_len=32 , max_steps=1_000 , batch_size=16 , threshold=1.0 , recopy_model=a__ , secondary_learner=a__ , eval_interval=10 , finetuned_model_name='''gpt2_finetuned.pt''' , ) if __name__ == "__main__": main()
313
1
import collections import json import os import re from typing import TYPE_CHECKING, List, Optional, Tuple import numpy as np from ...tokenization_utils_fast import PreTrainedTokenizer from ...utils import logging if TYPE_CHECKING: from transformers.pipelines.conversational import Conversation a__ : Union[str, Any] = logging.get_logger(__name__) a__ : Tuple = {'''vocab_file''': '''vocab.txt''', '''emoji_file''': '''emoji.json'''} a__ : Any = { '''vocab_file''': { '''abeja/gpt-neox-japanese-2.7b''': '''https://huggingface.co/abeja/gpt-neox-japanese-2.7b/resolve/main/vocab.txt''', }, '''emoji_file''': { '''abeja/gpt-neox-japanese-2.7b''': '''https://huggingface.co/abeja/gpt-neox-japanese-2.7b/resolve/main/emoji.json''', }, } a__ : Any = { '''abeja/gpt-neox-japanese-2.7b''': 2_048, } def UpperCAmelCase_( a__ , a__ ): """simple docstring""" with open(a__ , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : List[str] = json.loads(f.read() ) SCREAMING_SNAKE_CASE : str = collections.OrderedDict() SCREAMING_SNAKE_CASE : Any = collections.OrderedDict() SCREAMING_SNAKE_CASE : List[str] = collections.OrderedDict() with open(a__ , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : int = f.readlines() SCREAMING_SNAKE_CASE : Dict = [[t.rstrip('''\n''' )] if (t == ''',''' or ''',''' not in t) else t.rstrip('''\n''' ).split(''',''' ) for t in token] for idx, b in enumerate(a__ ): SCREAMING_SNAKE_CASE : Union[str, Any] = b SCREAMING_SNAKE_CASE : List[str] = idx for wd in b: SCREAMING_SNAKE_CASE : Optional[Any] = idx return vocab, raw_vocab, ids_to_tokens, emoji class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Any = VOCAB_FILES_NAMES __SCREAMING_SNAKE_CASE : Union[str, Any] = PRETRAINED_VOCAB_FILES_MAP __SCREAMING_SNAKE_CASE : List[Any] = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __SCREAMING_SNAKE_CASE : Dict = ['input_ids', 'attention_mask'] def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase="<|endoftext|>" , _lowerCamelCase="<|endoftext|>" , _lowerCamelCase="<|startoftext|>" , _lowerCamelCase="<|endoftext|>" , _lowerCamelCase=False , **_lowerCamelCase , ) ->Any: super().__init__( unk_token=_lowerCamelCase , pad_token=_lowerCamelCase , bos_token=_lowerCamelCase , eos_token=_lowerCamelCase , do_clean_text=_lowerCamelCase , **_lowerCamelCase , ) if not os.path.isfile(_lowerCamelCase ): raise ValueError( F"""Can't find a vocabulary file at path '{vocab_file}'. To load the vocabulary from a Google pretrained""" ''' model use `tokenizer = GPTNeoXJapaneseokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`''' ) if not os.path.isfile(_lowerCamelCase ): raise ValueError( F"""Can't find a emoji file at path '{emoji_file}'. To load the emoji information from a Google""" ''' pretrained model use `tokenizer = GPTNeoXJapaneseokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`''' ) SCREAMING_SNAKE_CASE : str = do_clean_text SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = load_vocab_and_emoji(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = SubWordJapaneseTokenizer( vocab=self.vocab , ids_to_tokens=self.ids_to_tokens , emoji=self.emoji ) @property def __lowerCAmelCase ( self ) ->Optional[int]: # self.vocab contains support for character fluctuation unique to Japanese, and has a large number of vocab return len(self.raw_vocab ) def __lowerCAmelCase ( self ) ->str: return dict(self.raw_vocab , **self.added_tokens_encoder ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: return self.subword_tokenizer.tokenize(_lowerCamelCase , clean=self.do_clean_text ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->int: return self.vocab.get(_lowerCamelCase , self.vocab.get(self.unk_token ) ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: return self.subword_tokenizer.convert_id_to_token(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[int]: SCREAMING_SNAKE_CASE : Union[str, Any] = ''''''.join(_lowerCamelCase ).strip() return out_string def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[int]: SCREAMING_SNAKE_CASE : Dict = [] for is_user, text in conversation.iter_texts(): input_ids.extend(self.encode(_lowerCamelCase , add_special_tokens=_lowerCamelCase ) + [self.eos_token_id] ) if len(_lowerCamelCase ) > self.model_max_length: SCREAMING_SNAKE_CASE : Tuple = input_ids[-self.model_max_length :] return input_ids def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->Tuple[str]: SCREAMING_SNAKE_CASE : Any = 0 if os.path.isdir(_lowerCamelCase ): SCREAMING_SNAKE_CASE : int = os.path.join( _lowerCamelCase , (filename_prefix + '''-''' if filename_prefix else '''''') + VOCAB_FILES_NAMES['''vocab_file'''] ) SCREAMING_SNAKE_CASE : List[Any] = os.path.join( _lowerCamelCase , (filename_prefix + '''-''' if filename_prefix else '''''') + VOCAB_FILES_NAMES['''emoji_file'''] ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = ( (filename_prefix + '''-''' if filename_prefix else '''''') + save_directory + VOCAB_FILES_NAMES['''vocab_file'''] ) SCREAMING_SNAKE_CASE : Tuple = ( (filename_prefix + '''-''' if filename_prefix else '''''') + save_directory + VOCAB_FILES_NAMES['''emoji_file'''] ) with open(_lowerCamelCase , '''w''' , encoding='''utf-8''' ) as writer: for token_index, token in self.ids_to_tokens.items(): if index != token_index: logger.warning( F"""Saving vocabulary to {vocab_file}: vocabulary indices are not consecutive.""" ''' Please check that the vocabulary is not corrupted!''' ) SCREAMING_SNAKE_CASE : Optional[Any] = token_index writer.write(''','''.join(_lowerCamelCase ) + '''\n''' ) index += 1 with open(_lowerCamelCase , '''w''' , encoding='''utf-8''' ) as writer: json.dump(self.emoji , _lowerCamelCase ) return vocab_file, emoji_file class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = vocab # same as swe SCREAMING_SNAKE_CASE : List[Any] = ids_to_tokens # same as bpe SCREAMING_SNAKE_CASE : Tuple = emoji SCREAMING_SNAKE_CASE : str = np.max([len(_lowerCamelCase ) for w in self.vocab.keys()] ) SCREAMING_SNAKE_CASE : Optional[Any] = re.compile(R'''(https?|ftp)(:\/\/[-_\.!~*\'()a-zA-Z0-9;\/?:\@&=\+$,%#]+)''' ) SCREAMING_SNAKE_CASE : str = re.compile(R'''[A-Za-z0-9\._+]*@[\-_0-9A-Za-z]+(\.[A-Za-z]+)*''' ) SCREAMING_SNAKE_CASE : Dict = re.compile(R'''[\(]{0,1}[0-9]{2,4}[\)\-\(]{0,1}[0-9]{2,4}[\)\-]{0,1}[0-9]{3,4}''' ) SCREAMING_SNAKE_CASE : Tuple = re.compile( R'''([12]\d{3}[/\-年])*(0?[1-9]|1[0-2])[/\-月]((0?[1-9]|[12][0-9]|3[01])日?)*(\d{1,2}|:|\d{1,2}時|\d{1,2}分|\(日\)|\(月\)|\(火\)|\(水\)|\(木\)|\(金\)|\(土\)|㈰|㈪|㈫|㈬|㈭|㈮|㈯)*''' ) SCREAMING_SNAKE_CASE : int = re.compile( R'''(明治|大正|昭和|平成|令和|㍾|㍽|㍼|㍻|\u32ff)\d{1,2}年(0?[1-9]|1[0-2])月(0?[1-9]|[12][0-9]|3[01])日(\d{1,2}|:|\d{1,2}時|\d{1,2}分|\(日\)|\(月\)|\(火\)|\(水\)|\(木\)|\(金\)|\(土\)|㈰|㈪|㈫|㈬|㈭|㈮|㈯)*''' ) SCREAMING_SNAKE_CASE : Tuple = re.compile( R'''((0|[1-9]\d*|[1-9]\d{0,2}(,\d{3})+)*億)*((0|[1-9]\d*|[1-9]\d{0,2}(,\d{3})+)*万)*((0|[1-9]\d*|[1-9]\d{0,2}(,\d{3})+)*千)*(0|[1-9]\d*|[1-9]\d{0,2}(,\d{3})+)*(千円|万円|千万円|円|千ドル|万ドル|千万ドル|ドル|千ユーロ|万ユーロ|千万ユーロ|ユーロ)+(\(税込\)|\(税抜\)|\+tax)*''' ) SCREAMING_SNAKE_CASE : List[Any] = '''─━│┃┄┅┆┇┈┉┊┋┌┍┎┏┐┑┒┓└┕┖┗┘┙┚┛├┝┞┟┠┡┢┣┤┥┦┧┨┩┪┫┬┭┮┯┰┱┲┳┴┵┶┷┸┹┺┻┼┽┾┿╀╁╂╃╄╅╆╇╈╉╊╋╌╍╎╏═║╒╓╔╕╖╗╘╙╚╛╜╝╞╟╠╡╢╣╤╥╦╧╨╩╪╫╬╭╮╯╰╱╲╳╴╵╶╷╸╹╺╻╼╽╾╿''' SCREAMING_SNAKE_CASE : int = '''▀▁▂▃▄▅▆▇█▉▊▋▌▍▎▏▐░▒▓▔▕▖▗▘▙▚▛▜▝▞▟''' SCREAMING_SNAKE_CASE : List[Any] = str.maketrans({k: '''<BLOCK>''' for k in keisen + blocks} ) def __len__( self ) ->Union[str, Any]: return len(self.ids_to_tokens ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : str = self.content_repattera.sub('''<URL>''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : str = self.content_repattera.sub('''<EMAIL>''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = self.content_repattera.sub('''<TEL>''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = self.content_repattera.sub('''<DATE>''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = self.content_repattera.sub('''<DATE>''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = self.content_repattera.sub('''<PRICE>''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = content.translate(self.content_transa ) while "<BLOCK><BLOCK>" in content: SCREAMING_SNAKE_CASE : int = content.replace('''<BLOCK><BLOCK>''' , '''<BLOCK>''' ) return content def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=False ) ->Optional[int]: SCREAMING_SNAKE_CASE : str = text.replace(''' ''' , '''<SP>''' ) SCREAMING_SNAKE_CASE : Optional[int] = text.replace(''' ''' , '''<SP>''' ) SCREAMING_SNAKE_CASE : Optional[int] = text.replace('''\r\n''' , '''<BR>''' ) SCREAMING_SNAKE_CASE : Optional[int] = text.replace('''\n''' , '''<BR>''' ) SCREAMING_SNAKE_CASE : int = text.replace('''\r''' , '''<BR>''' ) SCREAMING_SNAKE_CASE : Optional[Any] = text.replace('''\t''' , '''<TAB>''' ) SCREAMING_SNAKE_CASE : Tuple = text.replace('''—''' , '''ー''' ) SCREAMING_SNAKE_CASE : Union[str, Any] = text.replace('''−''' , '''ー''' ) for k, v in self.emoji["emoji"].items(): if k in text: SCREAMING_SNAKE_CASE : str = text.replace(_lowerCamelCase , _lowerCamelCase ) if clean: SCREAMING_SNAKE_CASE : List[Any] = self.clean_text(_lowerCamelCase ) def check_simbol(_lowerCamelCase ): SCREAMING_SNAKE_CASE : Any = x.encode() if len(_lowerCamelCase ) == 1 and len(_lowerCamelCase ) == 2: SCREAMING_SNAKE_CASE : Union[str, Any] = (int(e[0] ) << 8) + int(e[1] ) if ( (c >= 0xc_2a1 and c <= 0xc_2bf) or (c >= 0xc_780 and c <= 0xc_783) or (c >= 0xc_ab9 and c <= 0xc_bbf) or (c >= 0xc_c80 and c <= 0xc_da2) ): return True return False def checkuae(_lowerCamelCase ): SCREAMING_SNAKE_CASE : str = x.encode() if len(_lowerCamelCase ) == 1 and len(_lowerCamelCase ) == 3: SCREAMING_SNAKE_CASE : List[Any] = (int(e[0] ) << 16) + (int(e[1] ) << 8) + int(e[2] ) if c >= 0xe28_080 and c <= 0xe2b_07f: return True return False SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : List[str] = [] while pos < len(_lowerCamelCase ): SCREAMING_SNAKE_CASE : str = min(len(_lowerCamelCase ) , pos + self.maxlen + 1 ) if text[pos] == '''<''' else pos + 3 SCREAMING_SNAKE_CASE : Optional[int] = [] # (token_id, token, pos) for e in range(_lowerCamelCase , _lowerCamelCase , -1 ): SCREAMING_SNAKE_CASE : Optional[Any] = text[pos:e] if wd in self.vocab: if wd[0] == "<" and len(_lowerCamelCase ) > 2: SCREAMING_SNAKE_CASE : Any = [(self.vocab[wd], wd, e)] break else: candidates.append((self.vocab[wd], wd, e) ) if len(_lowerCamelCase ) > 0: # the smallest token_id is adopted SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = sorted(_lowerCamelCase , key=lambda _lowerCamelCase : x[0] )[0] result.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = e else: SCREAMING_SNAKE_CASE : List[str] = pos + 1 SCREAMING_SNAKE_CASE : Dict = text[pos:end] if check_simbol(_lowerCamelCase ): result.append('''<KIGOU>''' ) elif checkuae(_lowerCamelCase ): result.append('''<U2000U2BFF>''' ) else: for i in wd.encode('''utf-8''' ): result.append('''<|byte%d|>''' % i ) SCREAMING_SNAKE_CASE : Optional[Any] = end return result def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase="\n" ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[Any] = [] SCREAMING_SNAKE_CASE : Dict = [] SCREAMING_SNAKE_CASE : Tuple = self.ids_to_tokens[index][0] if word[:6] == "<|byte" and word[-2:] == "|>": byte_tokens.append(int(word[6:-2] ) ) else: if len(_lowerCamelCase ) > 0: words.append(bytearray(_lowerCamelCase ).decode('''utf-8''' , errors='''replace''' ) ) SCREAMING_SNAKE_CASE : Dict = [] if word[:7] == "<|emoji" and word[-2:] == "|>": words.append(self.emoji['''emoji_inv'''][word] ) elif word == "<SP>": words.append(''' ''' ) elif word == "<BR>": words.append(_lowerCamelCase ) elif word == "<TAB>": words.append('''\t''' ) elif word == "<BLOCK>": words.append('''▀''' ) elif word == "<KIGOU>": words.append('''ǀ''' ) elif word == "<U2000U2BFF>": words.append('''‖''' ) else: words.append(_lowerCamelCase ) if len(_lowerCamelCase ) > 0: words.append(bytearray(_lowerCamelCase ).decode('''utf-8''' , errors='''replace''' ) ) SCREAMING_SNAKE_CASE : List[str] = ''''''.join(_lowerCamelCase ) return text
313
import logging from pathlib import Path import numpy as np import pytorch_lightning as pl import torch from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint from pytorch_lightning.utilities import rank_zero_only from utils_rag import save_json def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = filter(lambda a__ : p.requires_grad , model.parameters() ) SCREAMING_SNAKE_CASE : List[Any] = sum([np.prod(p.size() ) for p in model_parameters] ) return params a__ : Any = logging.getLogger(__name__) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if metric == "rouge2": SCREAMING_SNAKE_CASE : str = '''{val_avg_rouge2:.4f}-{step_count}''' elif metric == "bleu": SCREAMING_SNAKE_CASE : List[Any] = '''{val_avg_bleu:.4f}-{step_count}''' elif metric == "em": SCREAMING_SNAKE_CASE : int = '''{val_avg_em:.4f}-{step_count}''' elif metric == "loss": SCREAMING_SNAKE_CASE : int = '''{val_avg_loss:.4f}-{step_count}''' else: raise NotImplementedError( F"""seq2seq callbacks only support rouge2 and bleu, got {metric}, You can make your own by adding to this""" ''' function.''' ) SCREAMING_SNAKE_CASE : Dict = ModelCheckpoint( dirpath=a__ , filename=a__ , monitor=F"""val_{metric}""" , mode='''max''' , save_top_k=1 , every_n_epochs=1 , ) return checkpoint_callback def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return EarlyStopping( monitor=F"""val_{metric}""" , mode='''min''' if '''loss''' in metric else '''max''' , patience=a__ , verbose=a__ , ) class a_ ( pl.Callback ): """simple docstring""" def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = {F"""lr_group_{i}""": param['''lr'''] for i, param in enumerate(pl_module.trainer.optimizers[0].param_groups )} pl_module.logger.log_metrics(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=True ) ->None: logger.info(F"""***** {type_path} results at step {trainer.global_step:05d} *****""" ) SCREAMING_SNAKE_CASE : Optional[int] = trainer.callback_metrics trainer.logger.log_metrics({k: v for k, v in metrics.items() if k not in ['''log''', '''progress_bar''', '''preds''']} ) # Log results SCREAMING_SNAKE_CASE : List[str] = Path(pl_module.hparams.output_dir ) if type_path == "test": SCREAMING_SNAKE_CASE : Any = od / '''test_results.txt''' SCREAMING_SNAKE_CASE : Optional[int] = od / '''test_generations.txt''' else: # this never gets hit. I prefer not to save intermediate generations, and results are in metrics.json # If people want this it will be easy enough to add back. SCREAMING_SNAKE_CASE : str = od / F"""{type_path}_results/{trainer.global_step:05d}.txt""" SCREAMING_SNAKE_CASE : Tuple = od / F"""{type_path}_generations/{trainer.global_step:05d}.txt""" results_file.parent.mkdir(exist_ok=_lowerCamelCase ) generations_file.parent.mkdir(exist_ok=_lowerCamelCase ) with open(_lowerCamelCase , '''a+''' ) as writer: for key in sorted(_lowerCamelCase ): if key in ["log", "progress_bar", "preds"]: continue SCREAMING_SNAKE_CASE : Tuple = metrics[key] if isinstance(_lowerCamelCase , torch.Tensor ): SCREAMING_SNAKE_CASE : List[Any] = val.item() SCREAMING_SNAKE_CASE : Tuple = F"""{key}: {val:.6f}\n""" writer.write(_lowerCamelCase ) if not save_generations: return if "preds" in metrics: SCREAMING_SNAKE_CASE : Optional[Any] = '''\n'''.join(metrics['''preds'''] ) generations_file.open('''w+''' ).write(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: try: SCREAMING_SNAKE_CASE : Any = pl_module.model.model.num_parameters() except AttributeError: SCREAMING_SNAKE_CASE : Optional[int] = pl_module.model.num_parameters() SCREAMING_SNAKE_CASE : int = count_trainable_parameters(_lowerCamelCase ) # mp stands for million parameters trainer.logger.log_metrics({'''n_params''': npars, '''mp''': npars / 1e6, '''grad_mp''': n_trainable_pars / 1e6} ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: save_json(pl_module.metrics , pl_module.metrics_save_path ) return self._write_logs(_lowerCamelCase , _lowerCamelCase , '''test''' ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Tuple: save_json(pl_module.metrics , pl_module.metrics_save_path ) # Uncommenting this will save val generations # return self._write_logs(trainer, pl_module, "valid")
313
1
import numpy as np def UpperCAmelCase_( a__ ): """simple docstring""" return 1 / (1 + np.exp(-vector )) def UpperCAmelCase_( a__ ): """simple docstring""" return vector * sigmoid(1.702 * vector ) if __name__ == "__main__": import doctest doctest.testmod()
313
import argparse import json from typing import List from ltp import LTP from transformers import BertTokenizer def UpperCAmelCase_( a__ ): """simple docstring""" if ( (cp >= 0x4_E00 and cp <= 0x9_FFF) or (cp >= 0x3_400 and cp <= 0x4_DBF) # or (cp >= 0x20_000 and cp <= 0x2A_6DF) # or (cp >= 0x2A_700 and cp <= 0x2B_73F) # or (cp >= 0x2B_740 and cp <= 0x2B_81F) # or (cp >= 0x2B_820 and cp <= 0x2C_EAF) # or (cp >= 0xF_900 and cp <= 0xF_AFF) or (cp >= 0x2F_800 and cp <= 0x2F_A1F) # ): # return True return False def UpperCAmelCase_( a__ ): """simple docstring""" for char in word: SCREAMING_SNAKE_CASE : str = ord(a__ ) if not _is_chinese_char(a__ ): return 0 return 1 def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = set() for token in tokens: SCREAMING_SNAKE_CASE : str = len(a__ ) > 1 and is_chinese(a__ ) if chinese_word: word_set.add(a__ ) SCREAMING_SNAKE_CASE : str = list(a__ ) return word_list def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if not chinese_word_set: return bert_tokens SCREAMING_SNAKE_CASE : List[str] = max([len(a__ ) for w in chinese_word_set] ) SCREAMING_SNAKE_CASE : Tuple = bert_tokens SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = 0, len(a__ ) while start < end: SCREAMING_SNAKE_CASE : Dict = True if is_chinese(bert_word[start] ): SCREAMING_SNAKE_CASE : Optional[int] = min(end - start , a__ ) for i in range(a__ , 1 , -1 ): SCREAMING_SNAKE_CASE : Optional[int] = ''''''.join(bert_word[start : start + i] ) if whole_word in chinese_word_set: for j in range(start + 1 , start + i ): SCREAMING_SNAKE_CASE : Optional[int] = '''##''' + bert_word[j] SCREAMING_SNAKE_CASE : List[str] = start + i SCREAMING_SNAKE_CASE : Optional[Any] = False break if single_word: start += 1 return bert_word def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : Optional[Any] = ltp_tokenizer.seg(lines[i : i + 100] )[0] SCREAMING_SNAKE_CASE : Union[str, Any] = [get_chinese_word(a__ ) for r in res] ltp_res.extend(a__ ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : Any = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : int = bert_tokenizer(lines[i : i + 100] , add_special_tokens=a__ , truncation=a__ , max_length=512 ) bert_res.extend(res['''input_ids'''] ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : int = [] for input_ids, chinese_word in zip(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = [] for id in input_ids: SCREAMING_SNAKE_CASE : List[Any] = bert_tokenizer._convert_id_to_token(a__ ) input_tokens.append(a__ ) SCREAMING_SNAKE_CASE : List[str] = add_sub_symbol(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = [] # We only save pos of chinese subwords start with ##, which mean is part of a whole word. for i, token in enumerate(a__ ): if token[:2] == "##": SCREAMING_SNAKE_CASE : Optional[int] = token[2:] # save chinese tokens' pos if len(a__ ) == 1 and _is_chinese_char(ord(a__ ) ): ref_id.append(a__ ) ref_ids.append(a__ ) assert len(a__ ) == len(a__ ) return ref_ids def UpperCAmelCase_( a__ ): """simple docstring""" with open(args.file_name , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : List[str] = f.readlines() SCREAMING_SNAKE_CASE : Union[str, Any] = [line.strip() for line in data if len(a__ ) > 0 and not line.isspace()] # avoid delimiter like '\u2029' SCREAMING_SNAKE_CASE : List[str] = LTP(args.ltp ) # faster in GPU device SCREAMING_SNAKE_CASE : int = BertTokenizer.from_pretrained(args.bert ) SCREAMING_SNAKE_CASE : int = prepare_ref(a__ , a__ , a__ ) with open(args.save_path , '''w''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : Tuple = [json.dumps(a__ ) + '''\n''' for ref in ref_ids] f.writelines(a__ ) if __name__ == "__main__": a__ : int = argparse.ArgumentParser(description='''prepare_chinese_ref''') parser.add_argument( '''--file_name''', type=str, default='''./resources/chinese-demo.txt''', help='''file need process, same as training data in lm''', ) parser.add_argument( '''--ltp''', type=str, default='''./resources/ltp''', help='''resources for LTP tokenizer, usually a path''' ) parser.add_argument('''--bert''', type=str, default='''./resources/robert''', help='''resources for Bert tokenizer''') parser.add_argument('''--save_path''', type=str, default='''./resources/ref.txt''', help='''path to save res''') a__ : int = parser.parse_args() main(args)
313
1
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 a_ ( unittest.TestCase , a__ ): """simple docstring""" def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = load_tool('''text-to-speech''' ) self.tool.setup() def __lowerCAmelCase ( self ) ->Any: # SpeechT5 isn't deterministic torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : str = self.tool('''hey''' ) SCREAMING_SNAKE_CASE : List[Any] = result.to_raw() self.assertTrue( torch.allclose( resulting_tensor[:3] , torch.tensor([-0.0_0_0_5_9_6_6_6_6_8_8_3_2_1_1_5_8_2_9, -0.0_0_0_3_6_5_7_6_4_0_1_9_0_7_9_5_0_6_4, -0.0_0_0_1_3_4_3_9_5_0_2_7_9_9_8_8_3_4_8_5] ) , ) ) def __lowerCAmelCase ( self ) ->Optional[Any]: # SpeechT5 isn't deterministic torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = self.tool('''hey''' ) SCREAMING_SNAKE_CASE : str = result.to_raw() self.assertTrue( torch.allclose( resulting_tensor[:3] , torch.tensor([-0.0_0_0_5_9_6_6_6_6_8_8_3_2_1_1_5_8_2_9, -0.0_0_0_3_6_5_7_6_4_0_1_9_0_7_9_5_0_6_4, -0.0_0_0_1_3_4_3_9_5_0_2_7_9_9_8_8_3_4_8_5] ) , ) )
313
import datetime import platform import subprocess from typing import Optional, Tuple, Union import numpy as np def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Tuple = '''1''' SCREAMING_SNAKE_CASE : Union[str, Any] = '''f32le''' SCREAMING_SNAKE_CASE : List[Any] = [ '''ffmpeg''', '''-i''', '''pipe:0''', '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] try: with subprocess.Popen(a__ , stdin=subprocess.PIPE , stdout=subprocess.PIPE ) as ffmpeg_process: SCREAMING_SNAKE_CASE : Tuple = ffmpeg_process.communicate(a__ ) except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to load audio files from filename''' ) from error SCREAMING_SNAKE_CASE : Optional[Any] = output_stream[0] SCREAMING_SNAKE_CASE : Any = np.frombuffer(a__ , np.floataa ) if audio.shape[0] == 0: raise ValueError('''Malformed soundfile''' ) return audio def UpperCAmelCase_( a__ , a__ , a__ = "f32le" , ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Dict = '''1''' if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Dict = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = platform.system() if system == "Linux": SCREAMING_SNAKE_CASE : Dict = '''alsa''' SCREAMING_SNAKE_CASE : Any = '''default''' elif system == "Darwin": SCREAMING_SNAKE_CASE : Union[str, Any] = '''avfoundation''' SCREAMING_SNAKE_CASE : Optional[int] = ''':0''' elif system == "Windows": SCREAMING_SNAKE_CASE : int = '''dshow''' SCREAMING_SNAKE_CASE : Any = '''default''' SCREAMING_SNAKE_CASE : Union[str, Any] = [ '''ffmpeg''', '''-f''', format_, '''-i''', input_, '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-fflags''', '''nobuffer''', '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] SCREAMING_SNAKE_CASE : List[str] = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample SCREAMING_SNAKE_CASE : List[Any] = _ffmpeg_stream(a__ , a__ ) for item in iterator: yield item def UpperCAmelCase_( a__ , a__ , a__ = None , a__ = None , a__ = "f32le" , ): """simple docstring""" if stream_chunk_s is not None: SCREAMING_SNAKE_CASE : Tuple = stream_chunk_s else: SCREAMING_SNAKE_CASE : List[str] = chunk_length_s SCREAMING_SNAKE_CASE : Union[str, Any] = ffmpeg_microphone(a__ , a__ , format_for_conversion=a__ ) if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : Optional[int] = np.intaa SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Any = np.floataa SCREAMING_SNAKE_CASE : Union[str, Any] = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) if stride_length_s is None: SCREAMING_SNAKE_CASE : Optional[Any] = chunk_length_s / 6 SCREAMING_SNAKE_CASE : Dict = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample if isinstance(a__ , (int, float) ): SCREAMING_SNAKE_CASE : List[Any] = [stride_length_s, stride_length_s] SCREAMING_SNAKE_CASE : Any = int(round(sampling_rate * stride_length_s[0] ) ) * size_of_sample SCREAMING_SNAKE_CASE : int = int(round(sampling_rate * stride_length_s[1] ) ) * size_of_sample SCREAMING_SNAKE_CASE : Union[str, Any] = datetime.datetime.now() SCREAMING_SNAKE_CASE : Dict = datetime.timedelta(seconds=a__ ) for item in chunk_bytes_iter(a__ , a__ , stride=(stride_left, stride_right) , stream=a__ ): # Put everything back in numpy scale SCREAMING_SNAKE_CASE : Dict = np.frombuffer(item['''raw'''] , dtype=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = ( item['''stride'''][0] // size_of_sample, item['''stride'''][1] // size_of_sample, ) SCREAMING_SNAKE_CASE : Any = sampling_rate audio_time += delta if datetime.datetime.now() > audio_time + 10 * delta: # We're late !! SKIP continue yield item def UpperCAmelCase_( a__ , a__ , a__ , a__ = False ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = b'''''' SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = stride if stride_left + stride_right >= chunk_len: raise ValueError( F"""Stride needs to be strictly smaller than chunk_len: ({stride_left}, {stride_right}) vs {chunk_len}""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = 0 for raw in iterator: acc += raw if stream and len(a__ ) < chunk_len: SCREAMING_SNAKE_CASE : List[str] = (_stride_left, 0) yield {"raw": acc[:chunk_len], "stride": stride, "partial": True} else: while len(a__ ) >= chunk_len: # We are flushing the accumulator SCREAMING_SNAKE_CASE : str = (_stride_left, stride_right) SCREAMING_SNAKE_CASE : List[str] = {'''raw''': acc[:chunk_len], '''stride''': stride} if stream: SCREAMING_SNAKE_CASE : List[str] = False yield item SCREAMING_SNAKE_CASE : Dict = stride_left SCREAMING_SNAKE_CASE : int = acc[chunk_len - stride_left - stride_right :] # Last chunk if len(a__ ) > stride_left: SCREAMING_SNAKE_CASE : Optional[Any] = {'''raw''': acc, '''stride''': (_stride_left, 0)} if stream: SCREAMING_SNAKE_CASE : Union[str, Any] = False yield item def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : str = 2**24 # 16Mo try: with subprocess.Popen(a__ , stdout=subprocess.PIPE , bufsize=a__ ) as ffmpeg_process: while True: SCREAMING_SNAKE_CASE : str = ffmpeg_process.stdout.read(a__ ) if raw == b"": break yield raw except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to stream audio files from filename''' ) from error
313
1
def UpperCAmelCase_( a__ = 100 ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = 0 SCREAMING_SNAKE_CASE : List[str] = 0 for i in range(1 , n + 1 ): sum_of_squares += i**2 sum_of_ints += i return sum_of_ints**2 - sum_of_squares if __name__ == "__main__": print(F"{solution() = }")
313
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a__ : Tuple = {'''configuration_wavlm''': ['''WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''WavLMConfig''']} try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Dict = [ '''WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST''', '''WavLMForAudioFrameClassification''', '''WavLMForCTC''', '''WavLMForSequenceClassification''', '''WavLMForXVector''', '''WavLMModel''', '''WavLMPreTrainedModel''', ] if TYPE_CHECKING: from .configuration_wavlm import WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP, WavLMConfig try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_wavlm import ( WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST, WavLMForAudioFrameClassification, WavLMForCTC, WavLMForSequenceClassification, WavLMForXVector, WavLMModel, WavLMPreTrainedModel, ) else: import sys a__ : Dict = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
1
def UpperCAmelCase_( a__ ): """simple docstring""" if not numbers: return 0 if not isinstance(a__ , (list, tuple) ) or not all( isinstance(a__ , a__ ) for number in numbers ): raise ValueError('''numbers must be an iterable of integers''' ) SCREAMING_SNAKE_CASE : Tuple = numbers[0] for i in range(1 , len(a__ ) ): # update the maximum and minimum subarray products SCREAMING_SNAKE_CASE : int = numbers[i] if number < 0: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = min_till_now, max_till_now SCREAMING_SNAKE_CASE : str = max(a__ , max_till_now * number ) SCREAMING_SNAKE_CASE : Any = min(a__ , min_till_now * number ) # update the maximum product found till now SCREAMING_SNAKE_CASE : Any = max(a__ , a__ ) return max_prod
313
import copy from ...configuration_utils import PretrainedConfig from ...utils import logging from ..auto import CONFIG_MAPPING a__ : int = logging.get_logger(__name__) a__ : Optional[Any] = { '''SenseTime/deformable-detr''': '''https://huggingface.co/sensetime/deformable-detr/resolve/main/config.json''', # See all Deformable DETR models at https://huggingface.co/models?filter=deformable-detr } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'deformable_detr' __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'd_model', 'num_attention_heads': 'encoder_attention_heads', } def __init__( self , _lowerCamelCase=True , _lowerCamelCase=None , _lowerCamelCase=3 , _lowerCamelCase=300 , _lowerCamelCase=1024 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=0.0 , _lowerCamelCase=True , _lowerCamelCase="relu" , _lowerCamelCase=256 , _lowerCamelCase=0.1 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1.0 , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase="sine" , _lowerCamelCase="resnet50" , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=False , _lowerCamelCase=300 , _lowerCamelCase=False , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=1 , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=0.1 , _lowerCamelCase=0.2_5 , _lowerCamelCase=False , **_lowerCamelCase , ) ->Optional[Any]: if backbone_config is not None and use_timm_backbone: raise ValueError('''You can\'t specify both `backbone_config` and `use_timm_backbone`.''' ) if not use_timm_backbone: if backbone_config is None: logger.info('''`backbone_config` is `None`. Initializing the config with the default `ResNet` backbone.''' ) SCREAMING_SNAKE_CASE : Dict = CONFIG_MAPPING['''resnet'''](out_features=['''stage4'''] ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : List[Any] = backbone_config.get('''model_type''' ) SCREAMING_SNAKE_CASE : Optional[Any] = CONFIG_MAPPING[backbone_model_type] SCREAMING_SNAKE_CASE : int = config_class.from_dict(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = use_timm_backbone SCREAMING_SNAKE_CASE : Optional[int] = backbone_config SCREAMING_SNAKE_CASE : Union[str, Any] = num_channels SCREAMING_SNAKE_CASE : Optional[Any] = num_queries SCREAMING_SNAKE_CASE : Optional[int] = max_position_embeddings SCREAMING_SNAKE_CASE : Optional[int] = d_model SCREAMING_SNAKE_CASE : str = encoder_ffn_dim SCREAMING_SNAKE_CASE : str = encoder_layers SCREAMING_SNAKE_CASE : str = encoder_attention_heads SCREAMING_SNAKE_CASE : Optional[int] = decoder_ffn_dim SCREAMING_SNAKE_CASE : int = decoder_layers SCREAMING_SNAKE_CASE : Union[str, Any] = decoder_attention_heads SCREAMING_SNAKE_CASE : List[str] = dropout SCREAMING_SNAKE_CASE : Optional[int] = attention_dropout SCREAMING_SNAKE_CASE : str = activation_dropout SCREAMING_SNAKE_CASE : Optional[int] = activation_function SCREAMING_SNAKE_CASE : Optional[int] = init_std SCREAMING_SNAKE_CASE : List[str] = init_xavier_std SCREAMING_SNAKE_CASE : Optional[Any] = encoder_layerdrop SCREAMING_SNAKE_CASE : Union[str, Any] = auxiliary_loss SCREAMING_SNAKE_CASE : List[Any] = position_embedding_type SCREAMING_SNAKE_CASE : str = backbone SCREAMING_SNAKE_CASE : Dict = use_pretrained_backbone SCREAMING_SNAKE_CASE : Dict = dilation # deformable attributes SCREAMING_SNAKE_CASE : str = num_feature_levels SCREAMING_SNAKE_CASE : Optional[Any] = encoder_n_points SCREAMING_SNAKE_CASE : Any = decoder_n_points SCREAMING_SNAKE_CASE : str = two_stage SCREAMING_SNAKE_CASE : List[str] = two_stage_num_proposals SCREAMING_SNAKE_CASE : Dict = with_box_refine if two_stage is True and with_box_refine is False: raise ValueError('''If two_stage is True, with_box_refine must be True.''' ) # Hungarian matcher SCREAMING_SNAKE_CASE : int = class_cost SCREAMING_SNAKE_CASE : Union[str, Any] = bbox_cost SCREAMING_SNAKE_CASE : Optional[int] = giou_cost # Loss coefficients SCREAMING_SNAKE_CASE : Dict = mask_loss_coefficient SCREAMING_SNAKE_CASE : Union[str, Any] = dice_loss_coefficient SCREAMING_SNAKE_CASE : str = bbox_loss_coefficient SCREAMING_SNAKE_CASE : Tuple = giou_loss_coefficient SCREAMING_SNAKE_CASE : Optional[int] = eos_coefficient SCREAMING_SNAKE_CASE : Tuple = focal_alpha SCREAMING_SNAKE_CASE : Optional[int] = disable_custom_kernels super().__init__(is_encoder_decoder=_lowerCamelCase , **_lowerCamelCase ) @property def __lowerCAmelCase ( self ) ->int: return self.encoder_attention_heads @property def __lowerCAmelCase ( self ) ->int: return self.d_model def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : str = copy.deepcopy(self.__dict__ ) if self.backbone_config is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.backbone_config.to_dict() SCREAMING_SNAKE_CASE : Any = self.__class__.model_type return output
313
1
import os import zipfile import pytest from datasets.utils.extract import ( BzipaExtractor, Extractor, GzipExtractor, LzaExtractor, SevenZipExtractor, TarExtractor, XzExtractor, ZipExtractor, ZstdExtractor, ) from .utils import require_lza, require_pyazr, require_zstandard @pytest.mark.parametrize( '''compression_format, is_archive''' , [ ('''7z''', True), ('''bz2''', False), ('''gzip''', False), ('''lz4''', False), ('''tar''', True), ('''xz''', False), ('''zip''', True), ('''zstd''', False), ] , ) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = { '''7z''': (seven_zip_file, SevenZipExtractor), '''bz2''': (bza_file, BzipaExtractor), '''gzip''': (gz_file, GzipExtractor), '''lz4''': (lza_file, LzaExtractor), '''tar''': (tar_file, TarExtractor), '''xz''': (xz_file, XzExtractor), '''zip''': (zip_file, ZipExtractor), '''zstd''': (zstd_file, ZstdExtractor), } SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = input_paths_and_base_extractors[compression_format] if input_path is None: SCREAMING_SNAKE_CASE : Dict = F"""for '{compression_format}' compression_format, """ if compression_format == "7z": reason += require_pyazr.kwargs["reason"] elif compression_format == "lz4": reason += require_lza.kwargs["reason"] elif compression_format == "zstd": reason += require_zstandard.kwargs["reason"] pytest.skip(a__ ) assert base_extractor.is_extractable(a__ ) SCREAMING_SNAKE_CASE : Tuple = tmp_path / ('''extracted''' if is_archive else '''extracted.txt''') base_extractor.extract(a__ , a__ ) if is_archive: assert output_path.is_dir() for file_path in output_path.iterdir(): assert file_path.name == text_file.name SCREAMING_SNAKE_CASE : Union[str, Any] = file_path.read_text(encoding='''utf-8''' ) else: SCREAMING_SNAKE_CASE : List[str] = output_path.read_text(encoding='''utf-8''' ) SCREAMING_SNAKE_CASE : List[str] = text_file.read_text(encoding='''utf-8''' ) assert extracted_file_content == expected_file_content @pytest.mark.parametrize( '''compression_format, is_archive''' , [ ('''7z''', True), ('''bz2''', False), ('''gzip''', False), ('''lz4''', False), ('''tar''', True), ('''xz''', False), ('''zip''', True), ('''zstd''', False), ] , ) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ , ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = { '''7z''': seven_zip_file, '''bz2''': bza_file, '''gzip''': gz_file, '''lz4''': lza_file, '''tar''': tar_file, '''xz''': xz_file, '''zip''': zip_file, '''zstd''': zstd_file, } SCREAMING_SNAKE_CASE : Tuple = input_paths[compression_format] if input_path is None: SCREAMING_SNAKE_CASE : str = F"""for '{compression_format}' compression_format, """ if compression_format == "7z": reason += require_pyazr.kwargs["reason"] elif compression_format == "lz4": reason += require_lza.kwargs["reason"] elif compression_format == "zstd": reason += require_zstandard.kwargs["reason"] pytest.skip(a__ ) SCREAMING_SNAKE_CASE : Dict = Extractor.infer_extractor_format(a__ ) assert extractor_format is not None SCREAMING_SNAKE_CASE : Dict = tmp_path / ('''extracted''' if is_archive else '''extracted.txt''') Extractor.extract(a__ , a__ , a__ ) if is_archive: assert output_path.is_dir() for file_path in output_path.iterdir(): assert file_path.name == text_file.name SCREAMING_SNAKE_CASE : Union[str, Any] = file_path.read_text(encoding='''utf-8''' ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = output_path.read_text(encoding='''utf-8''' ) SCREAMING_SNAKE_CASE : List[str] = text_file.read_text(encoding='''utf-8''' ) assert extracted_file_content == expected_file_content @pytest.fixture def UpperCAmelCase_( a__ , a__ ): """simple docstring""" import tarfile SCREAMING_SNAKE_CASE : List[Any] = tmp_path / '''data_dot_dot''' directory.mkdir() SCREAMING_SNAKE_CASE : int = directory / '''tar_file_with_dot_dot.tar''' with tarfile.TarFile(a__ , '''w''' ) as f: f.add(a__ , arcname=os.path.join('''..''' , text_file.name ) ) return path @pytest.fixture def UpperCAmelCase_( a__ ): """simple docstring""" import tarfile SCREAMING_SNAKE_CASE : List[str] = tmp_path / '''data_sym_link''' directory.mkdir() SCREAMING_SNAKE_CASE : int = directory / '''tar_file_with_sym_link.tar''' os.symlink('''..''' , directory / '''subdir''' , target_is_directory=a__ ) with tarfile.TarFile(a__ , '''w''' ) as f: f.add(str(directory / '''subdir''' ) , arcname='''subdir''' ) # str required by os.readlink on Windows and Python < 3.8 return path @pytest.mark.parametrize( '''insecure_tar_file, error_log''' , [('''tar_file_with_dot_dot''', '''illegal path'''), ('''tar_file_with_sym_link''', '''Symlink''')] , ) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = { '''tar_file_with_dot_dot''': tar_file_with_dot_dot, '''tar_file_with_sym_link''': tar_file_with_sym_link, } SCREAMING_SNAKE_CASE : Any = insecure_tar_files[insecure_tar_file] SCREAMING_SNAKE_CASE : Dict = tmp_path / '''extracted''' TarExtractor.extract(a__ , a__ ) assert caplog.text for record in caplog.records: assert record.levelname == "ERROR" assert error_log in record.msg def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : str = tmpdir / '''not_a_zip_file''' # From: https://github.com/python/cpython/pull/5053 SCREAMING_SNAKE_CASE : Union[str, Any] = ( b'''\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00\x00\x00\x01\x00\x00''' b'''\x00\x02\x08\x06\x00\x00\x00\x99\x81\xb6\'\x00\x00\x00\x15I''' b'''DATx\x01\x01\n\x00\xf5\xff\x00PK\x05\x06\x00PK\x06\x06\x07''' b'''\xac\x01N\xc6|a\r\x00\x00\x00\x00IEND\xaeB`\x82''' ) with not_a_zip_file.open('''wb''' ) as f: f.write(a__ ) assert zipfile.is_zipfile(str(a__ ) ) # is a false positive for `zipfile` assert not ZipExtractor.is_extractable(a__ ) # but we're right
313
import torch from diffusers import EulerDiscreteScheduler from diffusers.utils import torch_device from .test_schedulers import SchedulerCommonTest class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = (EulerDiscreteScheduler,) __SCREAMING_SNAKE_CASE : Optional[int] = 10 def __lowerCAmelCase ( self , **_lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = { '''num_train_timesteps''': 1100, '''beta_start''': 0.0_0_0_1, '''beta_end''': 0.0_2, '''beta_schedule''': '''linear''', } config.update(**_lowerCamelCase ) return config def __lowerCAmelCase ( self ) ->Tuple: for timesteps in [10, 50, 100, 1000]: self.check_over_configs(num_train_timesteps=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: for beta_start, beta_end in zip([0.0_0_0_0_1, 0.0_0_0_1, 0.0_0_1] , [0.0_0_0_2, 0.0_0_2, 0.0_2] ): self.check_over_configs(beta_start=_lowerCamelCase , beta_end=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: for schedule in ["linear", "scaled_linear"]: self.check_over_configs(beta_schedule=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: for prediction_type in ["epsilon", "v_prediction"]: self.check_over_configs(prediction_type=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : str = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : int = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Any = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = self.dummy_model() SCREAMING_SNAKE_CASE : int = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : Any = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[Any] = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_scheduler_config(prediction_type='''v_prediction''' ) SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = self.dummy_model() SCREAMING_SNAKE_CASE : List[str] = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : List[str] = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : str = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = output.prev_sample SCREAMING_SNAKE_CASE : str = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 0.0_0_0_2 ) < 1e-2 assert abs(result_mean.item() - 2.2676e-06 ) < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Dict = self.get_scheduler_config() SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[Any] = self.dummy_model() SCREAMING_SNAKE_CASE : str = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : Optional[Any] = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : Dict = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Dict = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Optional[Any] = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase , use_karras_sigmas=_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_model() SCREAMING_SNAKE_CASE : Dict = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : int = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : List[Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = output.prev_sample SCREAMING_SNAKE_CASE : Optional[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_2_4.5_2_2_9_9_4_9_9_5_1_1_7_1_9 ) < 1e-2 assert abs(result_mean.item() - 0.1_6_2_1_3_9_3_2_6_3_3_3_9_9_9_6_3 ) < 1e-3
313
1
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 a__ : List[str] = logging.get_logger(__name__) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->str: super().__init__() SCREAMING_SNAKE_CASE : Dict = nn.ModuleList(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = False , _lowerCamelCase = True , ) ->Union[ControlNetOutput, Tuple]: for i, (image, scale, controlnet) in enumerate(zip(_lowerCamelCase , _lowerCamelCase , self.nets ) ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = controlnet( _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , ) # merge samples if i == 0: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = down_samples, mid_sample else: SCREAMING_SNAKE_CASE : Tuple = [ samples_prev + samples_curr for samples_prev, samples_curr in zip(_lowerCamelCase , _lowerCamelCase ) ] mid_block_res_sample += mid_sample return down_block_res_samples, mid_block_res_sample def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = False , _lowerCamelCase = None , ) ->List[str]: SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : Optional[Any] = save_directory for controlnet in self.nets: controlnet.save_pretrained( _lowerCamelCase , is_main_process=_lowerCamelCase , save_function=_lowerCamelCase , safe_serialization=_lowerCamelCase , variant=_lowerCamelCase , ) idx += 1 SCREAMING_SNAKE_CASE : Optional[int] = model_path_to_save + F"""_{idx}""" @classmethod def __lowerCAmelCase ( cls , _lowerCamelCase , **_lowerCamelCase ) ->str: SCREAMING_SNAKE_CASE : List[Any] = 0 SCREAMING_SNAKE_CASE : List[str] = [] # 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`, ... SCREAMING_SNAKE_CASE : Optional[Any] = pretrained_model_path while os.path.isdir(_lowerCamelCase ): SCREAMING_SNAKE_CASE : str = ControlNetModel.from_pretrained(_lowerCamelCase , **_lowerCamelCase ) controlnets.append(_lowerCamelCase ) idx += 1 SCREAMING_SNAKE_CASE : Union[str, Any] = pretrained_model_path + F"""_{idx}""" logger.info(F"""{len(_lowerCamelCase )} controlnets loaded from {pretrained_model_path}.""" ) if len(_lowerCamelCase ) == 0: raise ValueError( F"""No ControlNets found under {os.path.dirname(_lowerCamelCase )}. Expected at least {pretrained_model_path + "_0"}.""" ) return cls(_lowerCamelCase )
313
import json from typing import Dict, List, Optional, Tuple, Union from tokenizers import pre_tokenizers, processors from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput from ...tokenization_utils_fast import PreTrainedTokenizerFast from ...utils import PaddingStrategy, logging from .tokenization_led import LEDTokenizer a__ : Dict = logging.get_logger(__name__) a__ : Dict = {'''vocab_file''': '''vocab.json''', '''merges_file''': '''merges.txt''', '''tokenizer_file''': '''tokenizer.json'''} a__ : str = { '''vocab_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json''', }, '''merges_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt''', }, '''tokenizer_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json''', }, } a__ : Optional[int] = { '''allenai/led-base-16384''': 16_384, } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = VOCAB_FILES_NAMES __SCREAMING_SNAKE_CASE : Tuple = PRETRAINED_VOCAB_FILES_MAP __SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __SCREAMING_SNAKE_CASE : Union[str, Any] = LEDTokenizer __SCREAMING_SNAKE_CASE : Optional[int] = ['input_ids', 'attention_mask'] def __init__( self , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase="replace" , _lowerCamelCase="<s>" , _lowerCamelCase="</s>" , _lowerCamelCase="</s>" , _lowerCamelCase="<s>" , _lowerCamelCase="<unk>" , _lowerCamelCase="<pad>" , _lowerCamelCase="<mask>" , _lowerCamelCase=False , _lowerCamelCase=True , **_lowerCamelCase , ) ->Union[str, Any]: super().__init__( _lowerCamelCase , _lowerCamelCase , tokenizer_file=_lowerCamelCase , errors=_lowerCamelCase , bos_token=_lowerCamelCase , eos_token=_lowerCamelCase , sep_token=_lowerCamelCase , cls_token=_lowerCamelCase , unk_token=_lowerCamelCase , pad_token=_lowerCamelCase , mask_token=_lowerCamelCase , add_prefix_space=_lowerCamelCase , trim_offsets=_lowerCamelCase , **_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Dict = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() ) if pre_tok_state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : str = getattr(_lowerCamelCase , pre_tok_state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : Optional[int] = add_prefix_space SCREAMING_SNAKE_CASE : str = pre_tok_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = add_prefix_space # the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__` SCREAMING_SNAKE_CASE : List[Any] = '''post_processor''' SCREAMING_SNAKE_CASE : int = getattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) if tokenizer_component_instance: SCREAMING_SNAKE_CASE : Any = json.loads(tokenizer_component_instance.__getstate__() ) # The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class` if "sep" in state: SCREAMING_SNAKE_CASE : Optional[int] = tuple(state['''sep'''] ) if "cls" in state: SCREAMING_SNAKE_CASE : Optional[Any] = tuple(state['''cls'''] ) SCREAMING_SNAKE_CASE : Any = False if state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : Union[str, Any] = add_prefix_space SCREAMING_SNAKE_CASE : Union[str, Any] = True if state.get('''trim_offsets''' , _lowerCamelCase ) != trim_offsets: SCREAMING_SNAKE_CASE : List[Any] = trim_offsets SCREAMING_SNAKE_CASE : Union[str, Any] = True if changes_to_apply: SCREAMING_SNAKE_CASE : List[str] = getattr(_lowerCamelCase , state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : List[Any] = component_class(**_lowerCamelCase ) setattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) @property # Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED def __lowerCAmelCase ( self ) ->str: if self._mask_token is None: if self.verbose: logger.error('''Using mask_token, but it is not set yet.''' ) return None return str(self._mask_token ) @mask_token.setter def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : str = AddedToken(_lowerCamelCase , lstrip=_lowerCamelCase , rstrip=_lowerCamelCase ) if isinstance(_lowerCamelCase , _lowerCamelCase ) else value SCREAMING_SNAKE_CASE : List[Any] = value def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : Tuple = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._batch_encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : List[Any] = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->Tuple[str]: SCREAMING_SNAKE_CASE : Any = self._tokenizer.model.save(_lowerCamelCase , name=_lowerCamelCase ) return tuple(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=None ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id] if token_ids_a is None: return output return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id] def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->List[int]: SCREAMING_SNAKE_CASE : Any = [self.sep_token_id] SCREAMING_SNAKE_CASE : List[str] = [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 __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None , _lowerCamelCase = PaddingStrategy.DO_NOT_PAD , _lowerCamelCase = None , _lowerCamelCase = None , ) ->dict: SCREAMING_SNAKE_CASE : Tuple = super()._pad( encoded_inputs=_lowerCamelCase , max_length=_lowerCamelCase , padding_strategy=_lowerCamelCase , pad_to_multiple_of=_lowerCamelCase , return_attention_mask=_lowerCamelCase , ) # Load from model defaults if return_attention_mask is None: SCREAMING_SNAKE_CASE : Optional[Any] = '''attention_mask''' in self.model_input_names if return_attention_mask and "global_attention_mask" in encoded_inputs: SCREAMING_SNAKE_CASE : int = encoded_inputs[self.model_input_names[0]] # `global_attention_mask` need to have the same length as other (sequential) inputs. SCREAMING_SNAKE_CASE : Tuple = len(encoded_inputs['''global_attention_mask'''] ) != len(_lowerCamelCase ) if needs_to_be_padded: SCREAMING_SNAKE_CASE : int = len(_lowerCamelCase ) - len(encoded_inputs['''global_attention_mask'''] ) if self.padding_side == "right": # Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend` SCREAMING_SNAKE_CASE : str = ( encoded_inputs['''global_attention_mask'''] + [-1] * difference ) elif self.padding_side == "left": SCREAMING_SNAKE_CASE : Optional[Any] = [-1] * difference + encoded_inputs[ '''global_attention_mask''' ] else: raise ValueError('''Invalid padding strategy:''' + str(self.padding_side ) ) return encoded_inputs
313
1
import logging import os from typing import List, Tuple import numpy as np import psutil import torch import torch.distributed as dist from transformers import RagRetriever a__ : Tuple = logging.getLogger(__name__) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=None ) ->Optional[int]: super().__init__( _lowerCamelCase , question_encoder_tokenizer=_lowerCamelCase , generator_tokenizer=_lowerCamelCase , index=_lowerCamelCase , init_retrieval=_lowerCamelCase , ) SCREAMING_SNAKE_CASE : List[str] = None def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: logger.info('''initializing retrieval''' ) # initializing a separate process group for retrieval as the default # nccl backend doesn't support gather/scatter operations while gloo # is too slow to replace nccl for the core gpu communication if dist.is_initialized(): logger.info('''dist initialized''' ) # needs to be set manually SCREAMING_SNAKE_CASE : Optional[int] = self._infer_socket_ifname() # avoid clash with the NCCL port SCREAMING_SNAKE_CASE : Tuple = str(distributed_port + 1 ) SCREAMING_SNAKE_CASE : str = dist.new_group(ranks=_lowerCamelCase , backend='''gloo''' ) # initialize retriever only on the main worker if not dist.is_initialized() or self._is_main(): logger.info('''dist not initialized / main''' ) self.index.init_index() # all processes wait untill the retriever is initialized by the main process if dist.is_initialized(): torch.distributed.barrier(group=self.process_group ) def __lowerCAmelCase ( self ) ->List[Any]: return dist.get_rank(group=self.process_group ) == 0 def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=torch.floataa ) ->List[Any]: SCREAMING_SNAKE_CASE : Tuple = torch.empty(_lowerCamelCase , dtype=_lowerCamelCase ) dist.scatter(_lowerCamelCase , src=0 , scatter_list=_lowerCamelCase , group=self.process_group ) return target_tensor def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Optional[int] = psutil.net_if_addrs() # a hacky way to deal with varying network interface names SCREAMING_SNAKE_CASE : str = next((addr for addr in addrs if addr.startswith('''e''' )) , _lowerCamelCase ) return ifname def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Tuple[np.ndarray, List[dict]]: # single GPU training if not dist.is_initialized(): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = self._main_retrieve(_lowerCamelCase , _lowerCamelCase ) return retrieved_doc_embeds, doc_ids, self.index.get_doc_dicts(_lowerCamelCase ) # distributed training SCREAMING_SNAKE_CASE : str = dist.get_world_size(group=self.process_group ) # gather logic SCREAMING_SNAKE_CASE : Dict = None if self._is_main(): SCREAMING_SNAKE_CASE : List[str] = [torch.empty(question_hidden_states.shape , dtype=torch.floataa ) for _ in range(_lowerCamelCase )] dist.gather(torch.tensor(_lowerCamelCase ) , dst=0 , gather_list=_lowerCamelCase , group=self.process_group ) # scatter logic SCREAMING_SNAKE_CASE : int = question_hidden_states.shape[0] SCREAMING_SNAKE_CASE : int = [] SCREAMING_SNAKE_CASE : Optional[int] = [] if self._is_main(): assert len(_lowerCamelCase ) == world_size SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = self._main_retrieve(torch.cat(_lowerCamelCase ).numpy() , _lowerCamelCase ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = torch.tensor(_lowerCamelCase ), torch.tensor(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = self._chunk_tensor(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = self._chunk_tensor(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self._scattered(_lowerCamelCase , [n_queries, n_docs] , target_type=torch.intaa ) SCREAMING_SNAKE_CASE : Optional[Any] = self._scattered(_lowerCamelCase , [n_queries, n_docs, question_hidden_states.shape[1]] ) return retrieved_doc_embeds.numpy(), doc_ids.numpy(), self.index.get_doc_dicts(_lowerCamelCase )
313
from __future__ import annotations import math def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if len(a__ ) != 2 or len(a[0] ) != 2 or len(a__ ) != 2 or len(b[0] ) != 2: raise Exception('''Matrices are not 2x2''' ) SCREAMING_SNAKE_CASE : Dict = [ [a[0][0] * b[0][0] + a[0][1] * b[1][0], a[0][0] * b[0][1] + a[0][1] * b[1][1]], [a[1][0] * b[0][0] + a[1][1] * b[1][0], a[1][0] * b[0][1] + a[1][1] * b[1][1]], ] return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] + matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] - matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) % 2 != 0 or len(a[0] ) % 2 != 0: raise Exception('''Odd matrices are not supported!''' ) SCREAMING_SNAKE_CASE : str = len(a__ ) SCREAMING_SNAKE_CASE : Any = matrix_length // 2 SCREAMING_SNAKE_CASE : Tuple = [[a[i][j] for j in range(a__ , a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : Optional[int] = [ [a[i][j] for j in range(a__ , a__ )] for i in range(a__ , a__ ) ] SCREAMING_SNAKE_CASE : Optional[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : List[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ , a__ )] return top_left, top_right, bot_left, bot_right def UpperCAmelCase_( a__ ): """simple docstring""" return len(a__ ), len(matrix[0] ) def UpperCAmelCase_( a__ ): """simple docstring""" print('''\n'''.join(str(a__ ) for line in matrix ) ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ ) == (2, 2): return default_matrix_multiplication(a__ , a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE : Dict = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : int = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Any = actual_strassen(matrix_addition(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = matrix_addition(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = matrix_subtraction(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) # construct the new matrix from our 4 quadrants SCREAMING_SNAKE_CASE : Optional[Any] = [] for i in range(len(a__ ) ): new_matrix.append(top_left[i] + top_right[i] ) for i in range(len(a__ ) ): new_matrix.append(bot_left[i] + bot_right[i] ) return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ )[1] != matrix_dimensions(a__ )[0]: SCREAMING_SNAKE_CASE : Any = ( '''Unable to multiply these matrices, please check the dimensions.\n''' F"""Matrix A: {matrixa}\n""" F"""Matrix B: {matrixa}""" ) raise Exception(a__ ) SCREAMING_SNAKE_CASE : str = matrix_dimensions(a__ ) SCREAMING_SNAKE_CASE : Tuple = matrix_dimensions(a__ ) if dimensiona[0] == dimensiona[1] and dimensiona[0] == dimensiona[1]: return [matrixa, matrixa] SCREAMING_SNAKE_CASE : str = max(*a__ , *a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = int(math.pow(2 , math.ceil(math.loga(a__ ) ) ) ) SCREAMING_SNAKE_CASE : Optional[int] = matrixa SCREAMING_SNAKE_CASE : Tuple = matrixa # Adding zeros to the matrices so that the arrays dimensions are the same and also # power of 2 for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) SCREAMING_SNAKE_CASE : Optional[Any] = actual_strassen(a__ , a__ ) # Removing the additional zeros for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): final_matrix[i].pop() else: final_matrix.pop() return final_matrix if __name__ == "__main__": a__ : Dict = [ [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 2, 3, 1], ] a__ : Union[str, Any] = [[0, 2, 1, 1], [16, 2, 3, 3], [2, 2, 7, 7], [13, 11, 22, 4]] print(strassen(matrixa, matrixa))
313
1
import argparse import random import joblib import numpy as np import torch from igf.igf import ( SecondaryLearner, collect_objective_set, compute_perplexity, generate_datasets, load_gpta, recopy_gpta, set_seed, train_secondary_learner, ) from torch.utils.data import DataLoader, RandomSampler from transformers import GPTaLMHeadModel def UpperCAmelCase_( a__=32 , a__=10 , a__=100 , a__=1_026 , a__=True , a__="data/tokenized_stories_train_wikitext103.jbl" , a__="igf_context_pairs.jbl" , ): """simple docstring""" set_seed(3 ) # generate train_data and objective_set SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = generate_datasets( a__ , a__ , number=a__ , min_len=1_026 , trim=a__ ) # keeps model same across runs set_seed(4 ) # model, lm_optimizer, lm_scheduler = recopy_gpt2(model, device, max_steps) # store original model weights # can we train on GPU? SCREAMING_SNAKE_CASE : str = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) # load pretrained model SCREAMING_SNAKE_CASE : Dict = load_gpta('''gpt2''' ).to(a__ ) print('''computing perplexity on objective set''' ) SCREAMING_SNAKE_CASE : int = compute_perplexity(a__ , a__ , a__ ).item() print('''perplexity on objective set:''' , a__ ) # collect igf pairs and save to file demo.jbl collect_objective_set(a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ ) # clean up, delete model and data we don't need anymore del model, train_data, objective_set torch.cuda.empty_cache() def UpperCAmelCase_( a__ , a__=15 , a__=128 , a__=100 , a__="igf_model.pt" , ): """simple docstring""" set_seed(42 ) # Load pre-trained model SCREAMING_SNAKE_CASE : List[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) # Initialize secondary learner to use embedding weights of model SCREAMING_SNAKE_CASE : str = SecondaryLearner(a__ ) # Train secondary learner SCREAMING_SNAKE_CASE : Union[str, Any] = train_secondary_learner( a__ , a__ , max_epochs=a__ , batch_size=a__ , eval_freq=100 , igf_model_path=a__ , ) del model, secondary_learner_train_data torch.cuda.empty_cache() return secondary_learner def UpperCAmelCase_( a__ , a__ , a__ , a__=32 , a__=1_000 , a__=16 , a__=1.0 , a__=recopy_gpta , a__=None , a__=10 , a__="gpt2_finetuned.pt" , ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) SCREAMING_SNAKE_CASE : Optional[int] = RandomSampler(a__ ) SCREAMING_SNAKE_CASE : Dict = DataLoader(a__ , sampler=a__ ) SCREAMING_SNAKE_CASE : Tuple = max_steps // (len(a__ )) + 1 SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = torch.zeros((1, context_len) , dtype=torch.long , device=a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = recopy_model(a__ , a__ , a__ ) model.train() if secondary_learner is not None: secondary_learner.to(a__ ) secondary_learner.eval() SCREAMING_SNAKE_CASE : List[str] = [] SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Tuple = [] # Compute the performance of the transformer model at the beginning SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) for epoch in range(int(a__ ) ): for step, example in enumerate(a__ ): torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Union[str, Any] = random.randint(0 , example.size(2 ) - context_len - 1 ) SCREAMING_SNAKE_CASE : Optional[int] = example[0, 0, start : start + context_len] lm_optimizer.zero_grad() SCREAMING_SNAKE_CASE : Optional[Any] = model(a__ , labels=a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = True if secondary_learner is not None: SCREAMING_SNAKE_CASE : List[str] = secondary_learner.forward( torch.tensor(a__ , dtype=torch.long , device=a__ ).unsqueeze(0 ) )[0].item() observed_qs.append(float(a__ ) ) # Here we implement the simple non-constant threshold for the predicted IG(X) value # We will decay the selectivity of our secondary learner filter from # 1 standard deviation above average to 1 below average after 10 batches. if global_step == 10: SCREAMING_SNAKE_CASE : Dict = -1 if predicted_q < threshold: SCREAMING_SNAKE_CASE : str = False # If we passed the filter, add the context to the batch! if do_backprop: contexts.append(np.array(context.cpu() ) ) SCREAMING_SNAKE_CASE : List[str] = outputs[0] lm_loss.backward() examples += 1 del outputs # Once the batch is filled with enough contexts, backprop on the batch. if examples == batch_size: torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Any = 0 # Do LM backprop torch.nn.utils.clip_grad_norm_(model.parameters() , 3.0 ) lm_optimizer.step() lm_scheduler.step() # Update learning rate schedule global_step += 1 # Compute the performance of the transformer model at this batch if global_step % eval_interval == 0: SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) # Break out of the loop after 60 batches if max_steps > 0 and global_step > 60: break if max_steps > 0 and global_step > 60: break # save finetuned transformer model torch.save(model.state_dict() , a__ ) torch.cuda.empty_cache() # Do some cleaning up so we can reinitialize for the next run of this function del lm_optimizer del lm_scheduler return model def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = argparse.ArgumentParser(description='''Fine-tune a transformer model with IGF on a language modeling task''' ) # Required parameters parser.add_argument( '''--data_dir''' , default=a__ , type=a__ , required=a__ , help='''The input data dir. Should contain data files for WikiText.''' , ) parser.add_argument( '''--model_name_or_path''' , default=a__ , type=a__ , required=a__ , help='''Path to pretrained model or model identifier from huggingface.co/models''' , ) parser.add_argument( '''--data_file''' , type=a__ , default=a__ , help=( '''A jbl file containing tokenized data which can be split as objective dataset, ''' '''train_dataset and test_dataset.''' ) , ) parser.add_argument( '''--igf_data_file''' , type=a__ , default=a__ , help='''A jbl file containing the context and information gain pairs to train secondary learner.''' , ) parser.add_argument( '''--output_dir''' , default=a__ , type=a__ , required=a__ , help='''The output directory where the final fine-tuned model is stored.''' , ) parser.add_argument( '''--tokenizer_name''' , default=a__ , type=a__ , help='''Pretrained tokenizer name or path if not the same as model_name''' , ) parser.add_argument('''--seed''' , type=a__ , default=a__ , help='''A seed for reproducible training.''' ) parser.add_argument( '''--context_len''' , default=32 , type=a__ , help=( '''The maximum total input sequence length after tokenization. Sequences longer ''' '''than this will be truncated, sequences shorter will be padded.''' ) , ) parser.add_argument( '''--size_objective_set''' , default=100 , type=a__ , help='''number of articles that are long enough to be used as our objective set''' , ) parser.add_argument( '''--eval_freq''' , default=100 , type=a__ , help='''secondary model evaluation is triggered at eval_freq''' ) parser.add_argument('''--max_steps''' , default=1_000 , type=a__ , help='''To calculate training epochs''' ) parser.add_argument( '''--secondary_learner_batch_size''' , default=128 , type=a__ , help='''batch size of training data for secondary learner''' , ) parser.add_argument( '''--batch_size''' , default=16 , type=a__ , help='''batch size of training data of language model(gpt2) ''' ) parser.add_argument( '''--eval_interval''' , default=10 , type=a__ , help=( '''decay the selectivity of our secondary learner filter from''' '''1 standard deviation above average to 1 below average after 10 batches''' ) , ) parser.add_argument( '''--number''' , default=100 , type=a__ , help='''The number of examples split to be used as objective_set/test_data''' ) parser.add_argument( '''--min_len''' , default=1_026 , type=a__ , help='''The minimum length of the article to be used as objective set''' ) parser.add_argument( '''--secondary_learner_max_epochs''' , default=15 , type=a__ , help='''number of epochs to train secondary learner''' ) parser.add_argument('''--trim''' , default=a__ , type=a__ , help='''truncate the example if it exceeds context length''' ) parser.add_argument( '''--threshold''' , default=1.0 , type=a__ , help=( '''The threshold value used by secondary learner to filter the train_data and allow only''' ''' informative data as input to the model''' ) , ) parser.add_argument('''--finetuned_model_name''' , default='''gpt2_finetuned.pt''' , type=a__ , help='''finetuned_model_name''' ) parser.add_argument( '''--recopy_model''' , default=a__ , type=a__ , help='''Reset the model to the original pretrained GPT-2 weights after each iteration''' , ) # function calls # Collecting *n* pairs of context and information gain(X, IG(X)) for training the secondary learner generate_n_pairs( context_len=32 , max_steps=10 , size_objective_set=100 , min_len=1_026 , trim=a__ , data_file='''data/tokenized_stories_train_wikitext103.jbl''' , igf_data_file='''igf_context_pairs.jbl''' , ) # Load train data for secondary learner SCREAMING_SNAKE_CASE : List[Any] = joblib.load('''data/IGF_values.jbl''' ) # Train secondary learner SCREAMING_SNAKE_CASE : Tuple = training_secondary_learner( a__ , secondary_learner_max_epochs=15 , secondary_learner_batch_size=128 , eval_freq=100 , igf_model_path='''igf_model.pt''' , ) # load pretrained gpt2 model SCREAMING_SNAKE_CASE : Optional[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) set_seed(42 ) # Generate train and test data to train and evaluate gpt2 model SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = generate_datasets( context_len=32 , file='''data/tokenized_stories_train_wikitext103.jbl''' , number=100 , min_len=1_026 , trim=a__ ) # fine-tuning of the gpt2 model using igf (Information Gain Filtration) finetune( a__ , a__ , a__ , context_len=32 , max_steps=1_000 , batch_size=16 , threshold=1.0 , recopy_model=a__ , secondary_learner=a__ , eval_interval=10 , finetuned_model_name='''gpt2_finetuned.pt''' , ) if __name__ == "__main__": main()
313
# We ignore warnings about stepping the scheduler since we step it ourselves during gradient accumulation import warnings from .state import AcceleratorState, GradientState warnings.filterwarnings('''ignore''', category=UserWarning, module='''torch.optim.lr_scheduler''') class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , _lowerCamelCase = False ) ->Any: SCREAMING_SNAKE_CASE : str = scheduler SCREAMING_SNAKE_CASE : List[str] = optimizers if isinstance(_lowerCamelCase , (list, tuple) ) else [optimizers] SCREAMING_SNAKE_CASE : Union[str, Any] = split_batches SCREAMING_SNAKE_CASE : List[Any] = step_with_optimizer SCREAMING_SNAKE_CASE : List[str] = GradientState() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->Optional[Any]: if not self.step_with_optimizer: # No link between scheduler and optimizer -> just step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) return # Otherwise, first make sure the optimizer was stepped. if not self.gradient_state.sync_gradients: if self.gradient_state.adjust_scheduler: self.scheduler._step_count += 1 return for opt in self.optimizers: if opt.step_was_skipped: return if self.split_batches: # Split batches -> the training dataloader batch size is not changed so one step per training step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: # Otherwise the training dataloader batch size was multiplied by `num_processes`, so we need to do # num_processes steps per training step SCREAMING_SNAKE_CASE : List[str] = AcceleratorState().num_processes for _ in range(_lowerCamelCase ): # Special case when using OneCycle and `drop_last` was not used if hasattr(self.scheduler , '''total_steps''' ): if self.scheduler._step_count <= self.scheduler.total_steps: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Union[str, Any]: return self.scheduler.get_last_lr() def __lowerCAmelCase ( self ) ->List[str]: return self.scheduler.state_dict() def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: self.scheduler.load_state_dict(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: return self.scheduler.get_lr() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->List[str]: return self.scheduler.print_lr(*_lowerCamelCase , **_lowerCamelCase )
313
1
from ...configuration_utils import PretrainedConfig from ...utils import logging from ...utils.backbone_utils import BackboneConfigMixin, get_aligned_output_features_output_indices a__ : str = logging.get_logger(__name__) a__ : Union[str, Any] = { '''shi-labs/nat-mini-in1k-224''': '''https://huggingface.co/shi-labs/nat-mini-in1k-224/resolve/main/config.json''', # See all Nat models at https://huggingface.co/models?filter=nat } class a_ ( a__ , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 'nat' __SCREAMING_SNAKE_CASE : Optional[int] = { 'num_attention_heads': 'num_heads', 'num_hidden_layers': 'num_layers', } def __init__( self , _lowerCamelCase=4 , _lowerCamelCase=3 , _lowerCamelCase=64 , _lowerCamelCase=[3, 4, 6, 5] , _lowerCamelCase=[2, 4, 8, 16] , _lowerCamelCase=7 , _lowerCamelCase=3.0 , _lowerCamelCase=True , _lowerCamelCase=0.0 , _lowerCamelCase=0.0 , _lowerCamelCase=0.1 , _lowerCamelCase="gelu" , _lowerCamelCase=0.0_2 , _lowerCamelCase=1e-5 , _lowerCamelCase=0.0 , _lowerCamelCase=None , _lowerCamelCase=None , **_lowerCamelCase , ) ->str: super().__init__(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = patch_size SCREAMING_SNAKE_CASE : Dict = num_channels SCREAMING_SNAKE_CASE : Dict = embed_dim SCREAMING_SNAKE_CASE : Tuple = depths SCREAMING_SNAKE_CASE : Any = len(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = num_heads SCREAMING_SNAKE_CASE : Any = kernel_size SCREAMING_SNAKE_CASE : List[Any] = mlp_ratio SCREAMING_SNAKE_CASE : int = qkv_bias SCREAMING_SNAKE_CASE : Dict = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = drop_path_rate SCREAMING_SNAKE_CASE : int = hidden_act SCREAMING_SNAKE_CASE : int = layer_norm_eps SCREAMING_SNAKE_CASE : List[Any] = initializer_range # we set the hidden_size attribute in order to make Nat work with VisionEncoderDecoderModel # this indicates the channel dimension after the last stage of the model SCREAMING_SNAKE_CASE : List[Any] = int(embed_dim * 2 ** (len(_lowerCamelCase ) - 1) ) SCREAMING_SNAKE_CASE : List[Any] = layer_scale_init_value SCREAMING_SNAKE_CASE : Dict = ['''stem'''] + [F"""stage{idx}""" for idx in range(1 , len(_lowerCamelCase ) + 1 )] SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = get_aligned_output_features_output_indices( out_features=_lowerCamelCase , out_indices=_lowerCamelCase , stage_names=self.stage_names )
313
import argparse import os from pathlib import Path from typing import Dict import tensorflow as tf import torch from tqdm import tqdm from transformers import PegasusConfig, PegasusForConditionalGeneration, PegasusTokenizer from transformers.models.pegasus.configuration_pegasus import DEFAULTS, task_specific_params a__ : Optional[Any] = [ # replace left string with right string to get the relevant state_dict key (identical state dict to bart) ['''memory_attention''', '''encoder_attn'''], ['''attention''', '''attn'''], ['''/''', '''.'''], ['''.LayerNorm.gamma''', '''_layer_norm.weight'''], ['''.LayerNorm.beta''', '''_layer_norm.bias'''], ['''r.layer_''', '''r.layers.'''], ['''output_proj''', '''out_proj'''], ['''ffn.dense_1.''', '''fc2.'''], ['''ffn.dense.''', '''fc1.'''], ['''ffn_layer_norm''', '''final_layer_norm'''], ['''kernel''', '''weight'''], ['''encoder_layer_norm.''', '''encoder.layer_norm.'''], ['''decoder_layer_norm.''', '''decoder.layer_norm.'''], ['''embeddings.weights''', '''shared.weight'''], ] def UpperCAmelCase_( a__ ): """simple docstring""" for pegasus_name, hf_name in PATTERNS: SCREAMING_SNAKE_CASE : Union[str, Any] = k.replace(a__ , a__ ) return k def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = DEFAULTS.copy() cfg_kwargs.update(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = PegasusConfig(**a__ ) SCREAMING_SNAKE_CASE : Optional[int] = PegasusForConditionalGeneration(a__ ) SCREAMING_SNAKE_CASE : Dict = torch_model.model.state_dict() SCREAMING_SNAKE_CASE : List[str] = {} for k, v in tf_weights.items(): SCREAMING_SNAKE_CASE : int = rename_state_dict_key(a__ ) if new_k not in sd: raise ValueError(F"""could not find new key {new_k} in state dict. (converted from {k})""" ) if "dense" in k or "proj" in new_k: SCREAMING_SNAKE_CASE : Dict = v.T SCREAMING_SNAKE_CASE : Tuple = torch.tensor(a__ , dtype=sd[new_k].dtype ) assert v.shape == sd[new_k].shape, F"""{new_k}, {k}, {v.shape}, {sd[new_k].shape}""" # make sure embedding.padding_idx is respected SCREAMING_SNAKE_CASE : Tuple = torch.zeros_like(mapping['''shared.weight'''][cfg.pad_token_id + 1] ) SCREAMING_SNAKE_CASE : int = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Union[str, Any] = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Optional[Any] = {k: torch.zeros_like(a__ ) for k, v in sd.items() if k.endswith('''bias''' ) and k not in mapping} mapping.update(**a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = torch_model.model.load_state_dict(a__ , strict=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = [ k for k in missing if k not in ['''encoder.embed_positions.weight''', '''decoder.embed_positions.weight'''] ] assert unexpected_missing == [], F"""no matches found for the following torch keys {unexpected_missing}""" assert extra == [], F"""no matches found for the following tf keys {extra}""" return torch_model def UpperCAmelCase_( a__="./ckpt/aeslc/model.ckpt-32000" ): """simple docstring""" SCREAMING_SNAKE_CASE : str = tf.train.list_variables(a__ ) SCREAMING_SNAKE_CASE : str = {} SCREAMING_SNAKE_CASE : List[Any] = ['''Adafactor''', '''global_step'''] for name, shape in tqdm(a__ , desc='''converting tf checkpoint to dict''' ): SCREAMING_SNAKE_CASE : Union[str, Any] = any(pat in name for pat in ignore_name ) if skip_key: continue SCREAMING_SNAKE_CASE : Dict = tf.train.load_variable(a__ , a__ ) SCREAMING_SNAKE_CASE : Any = array return tf_weights def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = Path(a__ ).parent.name SCREAMING_SNAKE_CASE : Union[str, Any] = task_specific_params[F"""summarization_{dataset}"""]['''max_position_embeddings'''] SCREAMING_SNAKE_CASE : Dict = PegasusTokenizer.from_pretrained('''sshleifer/pegasus''' , model_max_length=a__ ) assert tok.model_max_length == desired_max_model_length tok.save_pretrained(a__ ) # convert model SCREAMING_SNAKE_CASE : Any = get_tf_weights_as_numpy(a__ ) SCREAMING_SNAKE_CASE : List[str] = task_specific_params[F"""summarization_{dataset}"""] if dataset == "large": SCREAMING_SNAKE_CASE : int = task_specific_params SCREAMING_SNAKE_CASE : List[str] = convert_pegasus(a__ , a__ ) torch_model.save_pretrained(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch_model.state_dict() sd.pop('''model.decoder.embed_positions.weight''' ) sd.pop('''model.encoder.embed_positions.weight''' ) torch.save(a__ , Path(a__ ) / '''pytorch_model.bin''' ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument('''tf_ckpt_path''', type=str, help='''passed to tf.train.list_variables''') parser.add_argument('''save_dir''', default=None, type=str, help='''Path to the output PyTorch model.''') a__ : List[str] = parser.parse_args() if args.save_dir is None: a__ : Any = Path(args.tf_ckpt_path).parent.name a__ : int = os.path.join('''pegasus''', dataset) convert_pegasus_ckpt_to_pytorch(args.tf_ckpt_path, args.save_dir)
313
1
# Note: if you intend to run this script make sure you look under scripts/fsmt/ # to locate the appropriate script to do the work correctly. There is a set of scripts to: # - download and prepare data and run the conversion script # - perform eval to get the best hparam into the config # - generate model_cards - useful if you have multiple models from the same paper import argparse import json import os import re from collections import OrderedDict from os.path import basename, dirname import fairseq import torch from fairseq import hub_utils from fairseq.data.dictionary import Dictionary from transformers import FSMTConfig, FSMTForConditionalGeneration from transformers.models.fsmt.tokenization_fsmt import VOCAB_FILES_NAMES from transformers.tokenization_utils_base import TOKENIZER_CONFIG_FILE from transformers.utils import WEIGHTS_NAME, logging logging.set_verbosity_warning() a__ : str = 2 # based on the results of a search on a range of `num_beams`, `length_penalty` and `early_stopping` # values against wmt19 test data to obtain the best BLEU scores, we will use the following defaults: # # * `num_beams`: 5 (higher scores better, but requires more memory/is slower, can be adjusted by users) # * `early_stopping`: `False` consistently scored better # * `length_penalty` varied, so will assign the best one depending on the model a__ : List[Any] = { # fairseq: '''wmt19-ru-en''': {'''length_penalty''': 1.1}, '''wmt19-en-ru''': {'''length_penalty''': 1.15}, '''wmt19-en-de''': {'''length_penalty''': 1.0}, '''wmt19-de-en''': {'''length_penalty''': 1.1}, # allenai: '''wmt16-en-de-dist-12-1''': {'''length_penalty''': 0.6}, '''wmt16-en-de-dist-6-1''': {'''length_penalty''': 0.6}, '''wmt16-en-de-12-1''': {'''length_penalty''': 0.8}, '''wmt19-de-en-6-6-base''': {'''length_penalty''': 0.6}, '''wmt19-de-en-6-6-big''': {'''length_penalty''': 0.6}, } # this remaps the different models to their organization names a__ : Dict = {} for m in ["wmt19-ru-en", "wmt19-en-ru", "wmt19-en-de", "wmt19-de-en"]: a__ : str = '''facebook''' for m in [ "wmt16-en-de-dist-12-1", "wmt16-en-de-dist-6-1", "wmt16-en-de-12-1", "wmt19-de-en-6-6-base", "wmt19-de-en-6-6-big", ]: a__ : Tuple = '''allenai''' def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = dict((re.sub(r'''@@$''' , '''''' , a__ ), v) if k.endswith('''@@''' ) else (re.sub(r'''$''' , '''</w>''' , a__ ), v) for k, v in d.items() ) SCREAMING_SNAKE_CASE : str = '''<s> <pad> </s> <unk>'''.split() # restore the special tokens for k in keep_keys: del da[F"""{k}</w>"""] SCREAMING_SNAKE_CASE : Any = d[k] # restore return da def UpperCAmelCase_( a__ , a__ ): """simple docstring""" assert os.path.exists(a__ ) os.makedirs(a__ , exist_ok=a__ ) print(F"""Writing results to {pytorch_dump_folder_path}""" ) # handle various types of models SCREAMING_SNAKE_CASE : List[str] = basename(a__ ) SCREAMING_SNAKE_CASE : List[Any] = dirname(a__ ) SCREAMING_SNAKE_CASE : List[Any] = fairseq.model_parallel.models.transformer.ModelParallelTransformerModel SCREAMING_SNAKE_CASE : str = cls.hub_models() SCREAMING_SNAKE_CASE : Dict = {'''bpe''': '''fastbpe''', '''tokenizer''': '''moses'''} SCREAMING_SNAKE_CASE : Union[str, Any] = '''.''' # note: since the model dump is old, fairseq has upgraded its model some # time later, and it does a whole lot of rewrites and splits on the saved # weights, therefore we can't use torch.load() directly on the model file. # see: upgrade_state_dict(state_dict) in fairseq_model.py print(F"""using checkpoint {checkpoint_file}""" ) SCREAMING_SNAKE_CASE : int = hub_utils.from_pretrained( a__ , a__ , a__ , archive_map=a__ , **a__ ) SCREAMING_SNAKE_CASE : Optional[int] = vars(chkpt['''args''']['''model'''] ) SCREAMING_SNAKE_CASE : Dict = args['''source_lang'''] SCREAMING_SNAKE_CASE : List[Any] = args['''target_lang'''] SCREAMING_SNAKE_CASE : Optional[int] = dirname(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = basename(a__ ) # dicts SCREAMING_SNAKE_CASE : Any = os.path.join(a__ , F"""dict.{src_lang}.txt""" ) SCREAMING_SNAKE_CASE : Any = os.path.join(a__ , F"""dict.{tgt_lang}.txt""" ) SCREAMING_SNAKE_CASE : Dict = Dictionary.load(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = rewrite_dict_keys(src_dict.indices ) SCREAMING_SNAKE_CASE : Dict = len(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = os.path.join(a__ , '''vocab-src.json''' ) print(F"""Generating {src_vocab_file} of {src_vocab_size} of {src_lang} records""" ) with open(a__ , '''w''' , encoding='''utf-8''' ) as f: f.write(json.dumps(a__ , ensure_ascii=a__ , indent=a__ ) ) # detect whether this is a do_lower_case situation, which can be derived by checking whether we # have at least one uppercase letter in the source vocab SCREAMING_SNAKE_CASE : List[str] = True for k in src_vocab.keys(): if not k.islower(): SCREAMING_SNAKE_CASE : List[Any] = False break SCREAMING_SNAKE_CASE : Optional[int] = Dictionary.load(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = rewrite_dict_keys(tgt_dict.indices ) SCREAMING_SNAKE_CASE : str = len(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = os.path.join(a__ , '''vocab-tgt.json''' ) print(F"""Generating {tgt_vocab_file} of {tgt_vocab_size} of {tgt_lang} records""" ) with open(a__ , '''w''' , encoding='''utf-8''' ) as f: f.write(json.dumps(a__ , ensure_ascii=a__ , indent=a__ ) ) # merges_file (bpecodes) SCREAMING_SNAKE_CASE : Tuple = os.path.join(a__ , VOCAB_FILES_NAMES['''merges_file'''] ) for fn in ["bpecodes", "code"]: # older fairseq called the merges file "code" SCREAMING_SNAKE_CASE : str = os.path.join(a__ , a__ ) if os.path.exists(a__ ): break with open(a__ , encoding='''utf-8''' ) as fin: SCREAMING_SNAKE_CASE : Union[str, Any] = fin.read() SCREAMING_SNAKE_CASE : Optional[int] = re.sub(r''' \d+$''' , '''''' , a__ , 0 , re.M ) # remove frequency number print(F"""Generating {merges_file}""" ) with open(a__ , '''w''' , encoding='''utf-8''' ) as fout: fout.write(a__ ) # model config SCREAMING_SNAKE_CASE : int = os.path.join(a__ , '''config.json''' ) # validate bpe/tokenizer config, as currently it's hardcoded to moses+fastbpe - # may have to modify the tokenizer if a different type is used by a future model assert args["bpe"] == "fastbpe", F"""need to extend tokenizer to support bpe={args["bpe"]}""" assert args["tokenizer"] == "moses", F"""need to extend tokenizer to support bpe={args["tokenizer"]}""" SCREAMING_SNAKE_CASE : Dict = { '''architectures''': ['''FSMTForConditionalGeneration'''], '''model_type''': '''fsmt''', '''activation_dropout''': args['''activation_dropout'''], '''activation_function''': '''relu''', '''attention_dropout''': args['''attention_dropout'''], '''d_model''': args['''decoder_embed_dim'''], '''dropout''': args['''dropout'''], '''init_std''': 0.02, '''max_position_embeddings''': args['''max_source_positions'''], '''num_hidden_layers''': args['''encoder_layers'''], '''src_vocab_size''': src_vocab_size, '''tgt_vocab_size''': tgt_vocab_size, '''langs''': [src_lang, tgt_lang], '''encoder_attention_heads''': args['''encoder_attention_heads'''], '''encoder_ffn_dim''': args['''encoder_ffn_embed_dim'''], '''encoder_layerdrop''': args['''encoder_layerdrop'''], '''encoder_layers''': args['''encoder_layers'''], '''decoder_attention_heads''': args['''decoder_attention_heads'''], '''decoder_ffn_dim''': args['''decoder_ffn_embed_dim'''], '''decoder_layerdrop''': args['''decoder_layerdrop'''], '''decoder_layers''': args['''decoder_layers'''], '''bos_token_id''': 0, '''pad_token_id''': 1, '''eos_token_id''': 2, '''is_encoder_decoder''': True, '''scale_embedding''': not args['''no_scale_embedding'''], '''tie_word_embeddings''': args['''share_all_embeddings'''], } # good hparam defaults to start with SCREAMING_SNAKE_CASE : Tuple = 5 SCREAMING_SNAKE_CASE : int = False if model_dir in best_score_hparams and "length_penalty" in best_score_hparams[model_dir]: SCREAMING_SNAKE_CASE : str = best_score_hparams[model_dir]['''length_penalty'''] else: SCREAMING_SNAKE_CASE : int = 1.0 print(F"""Generating {fsmt_model_config_file}""" ) with open(a__ , '''w''' , encoding='''utf-8''' ) as f: f.write(json.dumps(a__ , ensure_ascii=a__ , indent=a__ ) ) # tokenizer config SCREAMING_SNAKE_CASE : Tuple = os.path.join(a__ , a__ ) SCREAMING_SNAKE_CASE : Any = { '''langs''': [src_lang, tgt_lang], '''model_max_length''': 1_024, '''do_lower_case''': do_lower_case, } print(F"""Generating {fsmt_tokenizer_config_file}""" ) with open(a__ , '''w''' , encoding='''utf-8''' ) as f: f.write(json.dumps(a__ , ensure_ascii=a__ , indent=a__ ) ) # model SCREAMING_SNAKE_CASE : int = chkpt['''models'''][0] SCREAMING_SNAKE_CASE : str = model.state_dict() # rename keys to start with 'model.' SCREAMING_SNAKE_CASE : Any = OrderedDict(('''model.''' + k, v) for k, v in model_state_dict.items() ) # remove unneeded keys SCREAMING_SNAKE_CASE : List[str] = [ '''model.model''', '''model.encoder.version''', '''model.decoder.version''', '''model.encoder_embed_tokens.weight''', '''model.decoder_embed_tokens.weight''', '''model.encoder.embed_positions._float_tensor''', '''model.decoder.embed_positions._float_tensor''', ] for k in ignore_keys: model_state_dict.pop(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[int] = FSMTConfig.from_pretrained(a__ ) SCREAMING_SNAKE_CASE : List[Any] = FSMTForConditionalGeneration(a__ ) # check that it loads ok model_new.load_state_dict(a__ , strict=a__ ) # save SCREAMING_SNAKE_CASE : List[str] = os.path.join(a__ , a__ ) print(F"""Generating {pytorch_weights_dump_path}""" ) torch.save(a__ , a__ ) print('''Conversion is done!''' ) print('''\nLast step is to upload the files to s3''' ) print(F"""cd {data_root}""" ) print(F"""transformers-cli upload {model_dir}""" ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--fsmt_checkpoint_path''', default=None, type=str, required=True, help=( '''Path to the official PyTorch checkpoint file which is expected to reside in the dump dir with dicts,''' ''' bpecodes, etc.''' ), ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, required=True, help='''Path to the output PyTorch model.''' ) a__ : str = parser.parse_args() convert_fsmt_checkpoint_to_pytorch(args.fsmt_checkpoint_path, args.pytorch_dump_folder_path)
313
import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import XLMRobertaTokenizerFast from diffusers import DDIMScheduler, KandinskyImgaImgPipeline, KandinskyPriorPipeline, UNetaDConditionModel, VQModel from diffusers.pipelines.kandinsky.text_encoder import MCLIPConfig, MultilingualCLIP from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = KandinskyImgaImgPipeline __SCREAMING_SNAKE_CASE : str = ['prompt', 'image_embeds', 'negative_image_embeds', 'image'] __SCREAMING_SNAKE_CASE : int = [ 'prompt', 'negative_prompt', 'image_embeds', 'negative_image_embeds', 'image', ] __SCREAMING_SNAKE_CASE : int = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'negative_prompt', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->int: return 32 @property def __lowerCAmelCase ( self ) ->List[str]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Tuple: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 100 @property def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = XLMRobertaTokenizerFast.from_pretrained('''YiYiXu/tiny-random-mclip-base''' ) return tokenizer @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[Any] = MCLIPConfig( numDims=self.cross_attention_dim , transformerDimensions=self.text_embedder_hidden_size , hidden_size=self.text_embedder_hidden_size , intermediate_size=37 , num_attention_heads=4 , num_hidden_layers=5 , vocab_size=1005 , ) SCREAMING_SNAKE_CASE : Dict = MultilingualCLIP(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = text_encoder.eval() return text_encoder @property def __lowerCAmelCase ( self ) ->Union[str, Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = { '''in_channels''': 4, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''text_image''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''text_image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : Tuple = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->List[str]: return { "block_out_channels": [32, 64], "down_block_types": ["DownEncoderBlock2D", "AttnDownEncoderBlock2D"], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": [ "AttnUpDecoderBlock2D", "UpDecoderBlock2D", ], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Optional[Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.dummy_text_encoder SCREAMING_SNAKE_CASE : Any = self.dummy_tokenizer SCREAMING_SNAKE_CASE : List[Any] = self.dummy_unet SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_movq SCREAMING_SNAKE_CASE : Optional[Any] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : Optional[Any] = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = { '''text_encoder''': text_encoder, '''tokenizer''': tokenizer, '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->str: SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(seed + 1 ) ).to(_lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : str = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : str = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : str = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = { '''prompt''': '''horse''', '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : List[str] = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : List[Any] = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : List[Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Optional[int] = np.array( [0.6_1_4_7_4_9_4_3, 0.6_0_7_3_5_3_9, 0.4_3_3_0_8_5_4_4, 0.5_9_2_8_2_6_9, 0.4_7_4_9_3_5_9_5, 0.4_6_7_5_5_9_7_3, 0.4_6_1_3_8_3_8, 0.4_5_3_6_8_7_9_7, 0.5_0_1_1_9_2_3_3] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/kandinsky_img2img_frog.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : str = '''A red cartoon frog, 4k''' SCREAMING_SNAKE_CASE : Any = KandinskyPriorPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = KandinskyImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : str = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=5 , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : Dict = pipeline( _lowerCamelCase , image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=768 , width=768 , strength=0.2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Tuple = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
from __future__ import annotations from collections.abc import Callable from typing import Generic, TypeVar a__ : Dict = TypeVar('''T''') a__ : Optional[int] = TypeVar('''U''') class a_ ( Generic[T, U] ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : Dict = key SCREAMING_SNAKE_CASE : Any = val SCREAMING_SNAKE_CASE : DoubleLinkedListNode[T, U] | None = None SCREAMING_SNAKE_CASE : DoubleLinkedListNode[T, U] | None = None def __repr__( self ) ->str: return ( F"""Node: key: {self.key}, val: {self.val}, """ F"""has next: {bool(self.next )}, has prev: {bool(self.prev )}""" ) class a_ ( Generic[T, U] ): """simple docstring""" def __init__( self ) ->None: SCREAMING_SNAKE_CASE : DoubleLinkedListNode[T, U] = DoubleLinkedListNode(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : DoubleLinkedListNode[T, U] = DoubleLinkedListNode(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = self.rear, self.head def __repr__( self ) ->str: SCREAMING_SNAKE_CASE : List[Any] = ['''DoubleLinkedList'''] SCREAMING_SNAKE_CASE : Tuple = self.head while node.next is not None: rep.append(str(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[Any] = node.next rep.append(str(self.rear ) ) return ",\n ".join(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase ) ->None: SCREAMING_SNAKE_CASE : str = self.rear.prev # All nodes other than self.head are guaranteed to have non-None previous assert previous is not None SCREAMING_SNAKE_CASE : Union[str, Any] = node SCREAMING_SNAKE_CASE : Dict = previous SCREAMING_SNAKE_CASE : List[Any] = node SCREAMING_SNAKE_CASE : Tuple = self.rear def __lowerCAmelCase ( self , _lowerCamelCase ) ->DoubleLinkedListNode[T, U] | None: if node.prev is None or node.next is None: return None SCREAMING_SNAKE_CASE : Union[str, Any] = node.next SCREAMING_SNAKE_CASE : int = node.prev SCREAMING_SNAKE_CASE : Union[str, Any] = None SCREAMING_SNAKE_CASE : Tuple = None return node class a_ ( Generic[T, U] ): """simple docstring""" __SCREAMING_SNAKE_CASE : dict[Callable[[T], U], LRUCache[T, U]] = {} def __init__( self , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : DoubleLinkedList[T, U] = DoubleLinkedList() SCREAMING_SNAKE_CASE : Union[str, Any] = capacity SCREAMING_SNAKE_CASE : List[str] = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = 0 SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : dict[T, DoubleLinkedListNode[T, U]] = {} def __repr__( self ) ->str: return ( F"""CacheInfo(hits={self.hits}, misses={self.miss}, """ F"""capacity={self.capacity}, current size={self.num_keys})""" ) def __contains__( self , _lowerCamelCase ) ->bool: return key in self.cache def __lowerCAmelCase ( self , _lowerCamelCase ) ->U | None: # Note: pythonic interface would throw KeyError rather than return None if key in self.cache: self.hits += 1 SCREAMING_SNAKE_CASE : DoubleLinkedListNode[T, U] = self.cache[key] SCREAMING_SNAKE_CASE : Dict = self.list.remove(self.cache[key] ) assert node == value_node # node is guaranteed not None because it is in self.cache assert node is not None self.list.add(_lowerCamelCase ) return node.val self.miss += 1 return None def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->None: if key not in self.cache: if self.num_keys >= self.capacity: # delete first node (oldest) when over capacity SCREAMING_SNAKE_CASE : Optional[Any] = self.list.head.next # guaranteed to have a non-None first node when num_keys > 0 # explain to type checker via assertions assert first_node is not None assert first_node.key is not None assert ( self.list.remove(_lowerCamelCase ) is not None ) # node guaranteed to be in list assert node.key is not None del self.cache[first_node.key] self.num_keys -= 1 SCREAMING_SNAKE_CASE : Any = DoubleLinkedListNode(_lowerCamelCase , _lowerCamelCase ) self.list.add(self.cache[key] ) self.num_keys += 1 else: # bump node to the end of the list, update value SCREAMING_SNAKE_CASE : Any = self.list.remove(self.cache[key] ) assert node is not None # node guaranteed to be in list SCREAMING_SNAKE_CASE : Any = value self.list.add(_lowerCamelCase ) @classmethod def __lowerCAmelCase ( cls , _lowerCamelCase = 128 ) ->Callable[[Callable[[T], U]], Callable[..., U]]: def cache_decorator_inner(_lowerCamelCase ) -> Callable[..., U]: def cache_decorator_wrapper(*_lowerCamelCase ) -> U: if func not in cls.decorator_function_to_instance_map: SCREAMING_SNAKE_CASE : Any = LRUCache(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = cls.decorator_function_to_instance_map[func].get(args[0] ) if result is None: SCREAMING_SNAKE_CASE : Union[str, Any] = func(*_lowerCamelCase ) cls.decorator_function_to_instance_map[func].put(args[0] , _lowerCamelCase ) return result def cache_info() -> LRUCache[T, U]: return cls.decorator_function_to_instance_map[func] setattr(_lowerCamelCase , '''cache_info''' , _lowerCamelCase ) # noqa: B010 return cache_decorator_wrapper return cache_decorator_inner if __name__ == "__main__": import doctest doctest.testmod()
313
import argparse import json import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ViTImageProcessor, ViTMSNConfig, ViTMSNModel from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD torch.set_grad_enabled(False) def UpperCAmelCase_( a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [] for i in range(config.num_hidden_layers ): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append((F"""module.blocks.{i}.norm1.weight""", F"""vit.encoder.layer.{i}.layernorm_before.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm1.bias""", F"""vit.encoder.layer.{i}.layernorm_before.bias""") ) rename_keys.append( (F"""module.blocks.{i}.attn.proj.weight""", F"""vit.encoder.layer.{i}.attention.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.attn.proj.bias""", F"""vit.encoder.layer.{i}.attention.output.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.norm2.weight""", F"""vit.encoder.layer.{i}.layernorm_after.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm2.bias""", F"""vit.encoder.layer.{i}.layernorm_after.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.weight""", F"""vit.encoder.layer.{i}.intermediate.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.bias""", F"""vit.encoder.layer.{i}.intermediate.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.weight""", F"""vit.encoder.layer.{i}.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.bias""", F"""vit.encoder.layer.{i}.output.dense.bias""") ) # projection layer + position embeddings rename_keys.extend( [ ('''module.cls_token''', '''vit.embeddings.cls_token'''), ('''module.patch_embed.proj.weight''', '''vit.embeddings.patch_embeddings.projection.weight'''), ('''module.patch_embed.proj.bias''', '''vit.embeddings.patch_embeddings.projection.bias'''), ('''module.pos_embed''', '''vit.embeddings.position_embeddings'''), ] ) if base_model: # layernorm + pooler rename_keys.extend( [ ('''module.norm.weight''', '''layernorm.weight'''), ('''module.norm.bias''', '''layernorm.bias'''), ] ) # if just the base model, we should remove "vit" from all keys that start with "vit" SCREAMING_SNAKE_CASE : Any = [(pair[0], pair[1][4:]) if pair[1].startswith('''vit''' ) else pair for pair in rename_keys] else: # layernorm + classification head rename_keys.extend( [ ('''norm.weight''', '''vit.layernorm.weight'''), ('''norm.bias''', '''vit.layernorm.bias'''), ('''head.weight''', '''classifier.weight'''), ('''head.bias''', '''classifier.bias'''), ] ) return rename_keys def UpperCAmelCase_( a__ , a__ , a__=False ): """simple docstring""" for i in range(config.num_hidden_layers ): if base_model: SCREAMING_SNAKE_CASE : Any = '''''' else: SCREAMING_SNAKE_CASE : Optional[int] = '''vit.''' # read in weights + bias of input projection layer (in timm, this is a single matrix + bias) SCREAMING_SNAKE_CASE : Optional[Any] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.weight""" ) SCREAMING_SNAKE_CASE : List[str] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.bias""" ) # next, add query, keys and values (in that order) to the state dict SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ : config.hidden_size, : ] SCREAMING_SNAKE_CASE : Any = in_proj_bias[: config.hidden_size] SCREAMING_SNAKE_CASE : int = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] SCREAMING_SNAKE_CASE : List[Any] = in_proj_bias[ config.hidden_size : config.hidden_size * 2 ] SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ -config.hidden_size :, : ] SCREAMING_SNAKE_CASE : Optional[Any] = in_proj_bias[-config.hidden_size :] def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = ['''head.weight''', '''head.bias'''] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = [ '''module.fc.fc1.weight''', '''module.fc.fc1.bias''', '''module.fc.bn1.weight''', '''module.fc.bn1.bias''', '''module.fc.bn1.running_mean''', '''module.fc.bn1.running_var''', '''module.fc.bn1.num_batches_tracked''', '''module.fc.fc2.weight''', '''module.fc.fc2.bias''', '''module.fc.bn2.weight''', '''module.fc.bn2.bias''', '''module.fc.bn2.running_mean''', '''module.fc.bn2.running_var''', '''module.fc.bn2.num_batches_tracked''', '''module.fc.fc3.weight''', '''module.fc.fc3.bias''', ] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = dct.pop(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = val def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = ViTMSNConfig() SCREAMING_SNAKE_CASE : Optional[int] = 1_000 SCREAMING_SNAKE_CASE : str = '''datasets/huggingface/label-files''' SCREAMING_SNAKE_CASE : List[str] = '''imagenet-1k-id2label.json''' SCREAMING_SNAKE_CASE : Union[str, Any] = json.load(open(hf_hub_download(a__ , a__ ) , '''r''' ) ) SCREAMING_SNAKE_CASE : List[Any] = {int(a__ ): v for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : str = idalabel SCREAMING_SNAKE_CASE : int = {v: k for k, v in idalabel.items()} if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = 384 SCREAMING_SNAKE_CASE : Any = 1_536 SCREAMING_SNAKE_CASE : List[str] = 6 elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Optional[int] = 1_024 SCREAMING_SNAKE_CASE : Optional[int] = 4_096 SCREAMING_SNAKE_CASE : Tuple = 24 SCREAMING_SNAKE_CASE : Union[str, Any] = 16 SCREAMING_SNAKE_CASE : Dict = 0.1 elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = 4 elif "l7" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = 7 SCREAMING_SNAKE_CASE : Union[str, Any] = 1_024 SCREAMING_SNAKE_CASE : List[Any] = 4_096 SCREAMING_SNAKE_CASE : List[Any] = 24 SCREAMING_SNAKE_CASE : Tuple = 16 SCREAMING_SNAKE_CASE : Union[str, Any] = 0.1 SCREAMING_SNAKE_CASE : Union[str, Any] = ViTMSNModel(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = torch.hub.load_state_dict_from_url(a__ , map_location='''cpu''' )['''target_encoder'''] SCREAMING_SNAKE_CASE : Any = ViTImageProcessor(size=config.image_size ) remove_projection_head(a__ ) SCREAMING_SNAKE_CASE : Any = create_rename_keys(a__ , base_model=a__ ) for src, dest in rename_keys: rename_key(a__ , a__ , a__ ) read_in_q_k_v(a__ , a__ , base_model=a__ ) model.load_state_dict(a__ ) model.eval() SCREAMING_SNAKE_CASE : Any = '''http://images.cocodataset.org/val2017/000000039769.jpg''' SCREAMING_SNAKE_CASE : Dict = Image.open(requests.get(a__ , stream=a__ ).raw ) SCREAMING_SNAKE_CASE : Optional[int] = ViTImageProcessor( size=config.image_size , image_mean=a__ , image_std=a__ ) SCREAMING_SNAKE_CASE : int = image_processor(images=a__ , return_tensors='''pt''' ) # forward pass torch.manual_seed(2 ) SCREAMING_SNAKE_CASE : Tuple = model(**a__ ) SCREAMING_SNAKE_CASE : str = outputs.last_hidden_state # The following Colab Notebook was used to generate these outputs: # https://colab.research.google.com/gist/sayakpaul/3672419a04f5997827503fd84079bdd1/scratchpad.ipynb if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[-1.0_915, -1.4_876, -1.1_809]] ) elif "b16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[14.2_889, -18.9_045, 11.7_281]] ) elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.tensor([[41.5_028, -22.8_681, 45.6_475]] ) elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = torch.tensor([[-4.3_868, 5.2_932, -0.4_137]] ) else: SCREAMING_SNAKE_CASE : List[str] = torch.tensor([[-0.1_792, -0.6_465, 2.4_263]] ) # verify logits assert torch.allclose(last_hidden_state[:, 0, :3] , a__ , atol=1e-4 ) print(F"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(a__ ) print(F"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(a__ ) if __name__ == "__main__": a__ : Dict = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--checkpoint_url''', default='''https://dl.fbaipublicfiles.com/msn/vits16_800ep.pth.tar''', type=str, help='''URL of the checkpoint you\'d like to convert.''', ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.''' ) a__ : Any = parser.parse_args() convert_vit_msn_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
313
1
import unittest from .lib import ( Matrix, Vector, axpy, square_zero_matrix, unit_basis_vector, zero_vector, ) class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Dict = Vector([1, 2, 3] ) self.assertEqual(x.component(0 ) , 1 ) self.assertEqual(x.component(2 ) , 3 ) SCREAMING_SNAKE_CASE : List[Any] = Vector() def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Optional[int] = Vector([0, 0, 0, 0, 0, 1] ) self.assertEqual(str(_lowerCamelCase ) , '''(0,0,0,0,0,1)''' ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Tuple = Vector([1, 2, 3, 4] ) self.assertEqual(len(_lowerCamelCase ) , 4 ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Dict = Vector([1, 2] ) SCREAMING_SNAKE_CASE : str = Vector([1, 2, 3, 4, 5] ) SCREAMING_SNAKE_CASE : List[Any] = Vector([0, 0, 0, 0, 0, 0, 0, 0, 0, 0] ) SCREAMING_SNAKE_CASE : Tuple = Vector([1, -1, 1, -1, 2, -3, 4, -5] ) self.assertAlmostEqual(x.euclidean_length() , 2.2_3_6 , 3 ) self.assertAlmostEqual(y.euclidean_length() , 7.4_1_6 , 3 ) self.assertEqual(z.euclidean_length() , 0 ) self.assertAlmostEqual(w.euclidean_length() , 7.6_1_6 , 3 ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Dict = Vector([1, 2, 3] ) SCREAMING_SNAKE_CASE : List[str] = Vector([1, 1, 1] ) self.assertEqual((x + y).component(0 ) , 2 ) self.assertEqual((x + y).component(1 ) , 3 ) self.assertEqual((x + y).component(2 ) , 4 ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : List[str] = Vector([1, 2, 3] ) SCREAMING_SNAKE_CASE : List[Any] = Vector([1, 1, 1] ) self.assertEqual((x - y).component(0 ) , 0 ) self.assertEqual((x - y).component(1 ) , 1 ) self.assertEqual((x - y).component(2 ) , 2 ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : int = Vector([1, 2, 3] ) SCREAMING_SNAKE_CASE : Optional[int] = Vector([2, -1, 4] ) # for test of dot product SCREAMING_SNAKE_CASE : str = Vector([1, -2, -1] ) self.assertEqual(str(x * 3.0 ) , '''(3.0,6.0,9.0)''' ) self.assertEqual((a * b) , 0 ) def __lowerCAmelCase ( self ) ->None: self.assertEqual(str(zero_vector(10 ) ).count('''0''' ) , 10 ) def __lowerCAmelCase ( self ) ->None: self.assertEqual(str(unit_basis_vector(3 , 1 ) ) , '''(0,1,0)''' ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : int = Vector([1, 2, 3] ) SCREAMING_SNAKE_CASE : Union[str, Any] = Vector([1, 0, 1] ) self.assertEqual(str(axpy(2 , _lowerCamelCase , _lowerCamelCase ) ) , '''(3,4,7)''' ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Tuple = Vector([1, 0, 0, 0, 0, 0] ) SCREAMING_SNAKE_CASE : str = x.copy() self.assertEqual(str(_lowerCamelCase ) , str(_lowerCamelCase ) ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Dict = Vector([1, 0, 0] ) x.change_component(0 , 0 ) x.change_component(1 , 1 ) self.assertEqual(str(_lowerCamelCase ) , '''(0,1,0)''' ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : int = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) self.assertEqual('''|1,2,3|\n|2,4,5|\n|6,7,8|\n''' , str(_lowerCamelCase ) ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Any = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) SCREAMING_SNAKE_CASE : Optional[int] = [[-3, -14, -10], [-5, -10, -5], [-2, -1, 0]] for x in range(a.height() ): for y in range(a.width() ): self.assertEqual(minors[x][y] , a.minor(_lowerCamelCase , _lowerCamelCase ) ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : List[str] = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) SCREAMING_SNAKE_CASE : str = [[-3, 14, -10], [5, -10, 5], [-2, 1, 0]] for x in range(a.height() ): for y in range(a.width() ): self.assertEqual(cofactors[x][y] , a.cofactor(_lowerCamelCase , _lowerCamelCase ) ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Tuple = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) self.assertEqual(-5 , a.determinant() ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : int = Matrix([[1, 2, 3], [4, 5, 6], [7, 8, 9]] , 3 , 3 ) SCREAMING_SNAKE_CASE : Optional[Any] = Vector([1, 2, 3] ) self.assertEqual('''(14,32,50)''' , str(a * x ) ) self.assertEqual('''|2,4,6|\n|8,10,12|\n|14,16,18|\n''' , str(a * 2 ) ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Union[str, Any] = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) a.change_component(0 , 2 , 5 ) self.assertEqual('''|1,2,5|\n|2,4,5|\n|6,7,8|\n''' , str(_lowerCamelCase ) ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : List[str] = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) self.assertEqual(7 , a.component(2 , 1 ) , 0.0_1 ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : str = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) SCREAMING_SNAKE_CASE : List[Any] = Matrix([[1, 2, 7], [2, 4, 5], [6, 7, 10]] , 3 , 3 ) self.assertEqual('''|2,4,10|\n|4,8,10|\n|12,14,18|\n''' , str(a + b ) ) def __lowerCAmelCase ( self ) ->None: SCREAMING_SNAKE_CASE : Union[str, Any] = Matrix([[1, 2, 3], [2, 4, 5], [6, 7, 8]] , 3 , 3 ) SCREAMING_SNAKE_CASE : int = Matrix([[1, 2, 7], [2, 4, 5], [6, 7, 10]] , 3 , 3 ) self.assertEqual('''|0,0,-4|\n|0,0,0|\n|0,0,-2|\n''' , str(a - b ) ) def __lowerCAmelCase ( self ) ->None: self.assertEqual( '''|0,0,0,0,0|\n|0,0,0,0,0|\n|0,0,0,0,0|\n|0,0,0,0,0|\n|0,0,0,0,0|\n''' , str(square_zero_matrix(5 ) ) , ) if __name__ == "__main__": unittest.main()
313
import csv import tweepy # Twitter API credentials a__ : Union[str, Any] = '''''' a__ : List[str] = '''''' a__ : Any = '''''' a__ : List[str] = '''''' def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = tweepy.OAuthHandler(a__ , a__ ) auth.set_access_token(a__ , a__ ) SCREAMING_SNAKE_CASE : List[str] = tweepy.API(a__ ) # initialize a list to hold all the tweepy Tweets SCREAMING_SNAKE_CASE : Any = [] # make initial request for most recent tweets (200 is the maximum allowed count) SCREAMING_SNAKE_CASE : List[Any] = api.user_timeline(screen_name=a__ , count=200 ) # save most recent tweets alltweets.extend(a__ ) # save the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Tuple = alltweets[-1].id - 1 # keep grabbing tweets until there are no tweets left to grab while len(a__ ) > 0: print(F"""getting tweets before {oldest}""" ) # all subsequent requests use the max_id param to prevent duplicates SCREAMING_SNAKE_CASE : Any = api.user_timeline( screen_name=a__ , count=200 , max_id=a__ ) # save most recent tweets alltweets.extend(a__ ) # update the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Dict = alltweets[-1].id - 1 print(F"""...{len(a__ )} tweets downloaded so far""" ) # transform the tweepy tweets into a 2D array that will populate the csv SCREAMING_SNAKE_CASE : Optional[Any] = [[tweet.id_str, tweet.created_at, tweet.text] for tweet in alltweets] # write the csv with open(F"""new_{screen_name}_tweets.csv""" , '''w''' ) as f: SCREAMING_SNAKE_CASE : List[Any] = csv.writer(a__ ) writer.writerow(['''id''', '''created_at''', '''text'''] ) writer.writerows(a__ ) if __name__ == "__main__": # pass in the username of the account you want to download get_all_tweets('''FirePing32''')
313
1
import json import multiprocessing as mp import re from collections import defaultdict from functools import partial from typing import Dict, List, Optional, Set, Tuple, Type from datasets import Dataset from datasketch import MinHash, MinHashLSH from dpu_utils.utils.iterators import ThreadedIterator from tqdm import tqdm a__ : Tuple = re.compile('''[^A-Za-z_0-9]''') # parameters used in DuplicationIndex a__ : str = 10 a__ : Optional[int] = 256 def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) < MIN_NUM_TOKENS: return None SCREAMING_SNAKE_CASE : Dict = MinHash(num_perm=a__ ) for token in set(a__ ): min_hash.update(token.encode() ) return min_hash def UpperCAmelCase_( a__ ): """simple docstring""" return {t for t in NON_ALPHA.split(a__ ) if len(t.strip() ) > 0} class a_ : """simple docstring""" def __init__( self , *, _lowerCamelCase = 0.8_5 , ) ->Tuple: SCREAMING_SNAKE_CASE : str = duplication_jaccard_threshold SCREAMING_SNAKE_CASE : int = NUM_PERM SCREAMING_SNAKE_CASE : str = MinHashLSH(threshold=self._duplication_jaccard_threshold , num_perm=self._num_perm ) SCREAMING_SNAKE_CASE : List[str] = defaultdict(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->None: SCREAMING_SNAKE_CASE : Tuple = self._index.query(_lowerCamelCase ) if code_key in self._index.keys: print(F"""Duplicate key {code_key}""" ) return self._index.insert(_lowerCamelCase , _lowerCamelCase ) if len(_lowerCamelCase ) > 0: for base_duplicate in close_duplicates: if base_duplicate in self._duplicate_clusters: self._duplicate_clusters[base_duplicate].add(_lowerCamelCase ) break else: self._duplicate_clusters[close_duplicates[0]].add(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[List[Dict]]: SCREAMING_SNAKE_CASE : Optional[int] = [] for base, duplicates in self._duplicate_clusters.items(): SCREAMING_SNAKE_CASE : str = [base] + list(_lowerCamelCase ) # reformat the cluster to be a list of dict SCREAMING_SNAKE_CASE : Dict = [{'''base_index''': el[0], '''repo_name''': el[1], '''path''': el[2]} for el in cluster] duplicate_clusters.append(_lowerCamelCase ) return duplicate_clusters def __lowerCAmelCase ( self , _lowerCamelCase ) ->None: SCREAMING_SNAKE_CASE : Tuple = self.get_duplicate_clusters() with open(_lowerCamelCase , '''w''' ) as f: json.dump(_lowerCamelCase , _lowerCamelCase ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = element SCREAMING_SNAKE_CASE : str = get_min_hash([t for t in NON_ALPHA.split(data['''content'''] ) if len(t.strip() ) > 0] ) if min_hash is not None: return (index, data["repo_name"], data["path"]), min_hash def UpperCAmelCase_( a__ ): """simple docstring""" with mp.Pool() as pool: for data in pool.imap_unordered( _compute_min_hash , ThreadedIterator(a__ , max_queue_size=10_000 ) , chunksize=100 , ): if data is not None: yield data def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = DuplicationIndex(duplication_jaccard_threshold=a__ ) for filename, min_hash in tqdm(ThreadedIterator(minhash_iter(enumerate(a__ ) ) , max_queue_size=100 ) ): di.add(a__ , a__ ) # Returns a List[Cluster] where Cluster is List[str] with the filenames. return di.get_duplicate_clusters() def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = get_tokens(a__ ) SCREAMING_SNAKE_CASE : Dict = get_tokens(a__ ) return len(tokensa & tokensa ) / len(tokensa | tokensa ) a__ : Optional[Any] = None def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = [] for elementa in cluster: SCREAMING_SNAKE_CASE : Any = _shared_dataset[elementa['''base_index''']]['''content'''] for elementa in extremes: SCREAMING_SNAKE_CASE : Dict = _shared_dataset[elementa['''base_index''']]['''content'''] if jaccard_similarity(a__ , a__ ) >= jaccard_threshold: elementa["copies"] += 1 break else: SCREAMING_SNAKE_CASE : Dict = 1 extremes.append(a__ ) return extremes def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" global _shared_dataset SCREAMING_SNAKE_CASE : Dict = dataset SCREAMING_SNAKE_CASE : Optional[Any] = [] SCREAMING_SNAKE_CASE : Tuple = partial(_find_cluster_extremes_shared , jaccard_threshold=a__ ) with mp.Pool() as pool: for extremes in tqdm( pool.imap_unordered( a__ , a__ , ) , total=len(a__ ) , ): extremes_list.append(a__ ) return extremes_list def UpperCAmelCase_( a__ , a__ = 0.85 ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = make_duplicate_clusters(a__ , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = {x['''base_index'''] for cluster in duplicate_clusters for x in cluster} SCREAMING_SNAKE_CASE : int = {} SCREAMING_SNAKE_CASE : List[Any] = find_extremes(a__ , a__ , a__ ) for extremes in extremes_clusters: for element in extremes: SCREAMING_SNAKE_CASE : int = element SCREAMING_SNAKE_CASE : Optional[int] = duplicate_indices - set(extreme_dict.keys() ) SCREAMING_SNAKE_CASE : List[Any] = dataset.filter(lambda a__ , a__ : idx not in remove_indices , with_indices=a__ ) # update duplicate_clusters for cluster in duplicate_clusters: for element in cluster: SCREAMING_SNAKE_CASE : Tuple = element['''base_index'''] in extreme_dict if element["is_extreme"]: SCREAMING_SNAKE_CASE : Union[str, Any] = extreme_dict[element['''base_index''']]['''copies'''] print(F"""Original dataset size: {len(a__ )}""" ) print(F"""Number of duplicate clusters: {len(a__ )}""" ) print(F"""Files in duplicate cluster: {len(a__ )}""" ) print(F"""Unique files in duplicate cluster: {len(a__ )}""" ) print(F"""Filtered dataset size: {len(a__ )}""" ) return ds_filter, duplicate_clusters
313
from collections import OrderedDict from typing import Mapping from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a__ : Optional[Any] = logging.get_logger(__name__) a__ : List[str] = { '''kssteven/ibert-roberta-base''': '''https://huggingface.co/kssteven/ibert-roberta-base/resolve/main/config.json''', '''kssteven/ibert-roberta-large''': '''https://huggingface.co/kssteven/ibert-roberta-large/resolve/main/config.json''', '''kssteven/ibert-roberta-large-mnli''': ( '''https://huggingface.co/kssteven/ibert-roberta-large-mnli/resolve/main/config.json''' ), } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'ibert' def __init__( self , _lowerCamelCase=3_0522 , _lowerCamelCase=768 , _lowerCamelCase=12 , _lowerCamelCase=12 , _lowerCamelCase=3072 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1e-12 , _lowerCamelCase=1 , _lowerCamelCase=0 , _lowerCamelCase=2 , _lowerCamelCase="absolute" , _lowerCamelCase=False , _lowerCamelCase="none" , **_lowerCamelCase , ) ->Any: super().__init__(pad_token_id=_lowerCamelCase , bos_token_id=_lowerCamelCase , eos_token_id=_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = vocab_size SCREAMING_SNAKE_CASE : str = hidden_size SCREAMING_SNAKE_CASE : Tuple = num_hidden_layers SCREAMING_SNAKE_CASE : List[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Dict = hidden_act SCREAMING_SNAKE_CASE : Optional[int] = intermediate_size SCREAMING_SNAKE_CASE : Optional[Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Tuple = type_vocab_size SCREAMING_SNAKE_CASE : Optional[int] = initializer_range SCREAMING_SNAKE_CASE : Union[str, Any] = layer_norm_eps SCREAMING_SNAKE_CASE : str = position_embedding_type SCREAMING_SNAKE_CASE : Optional[int] = quant_mode SCREAMING_SNAKE_CASE : Dict = force_dequant class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: if self.task == "multiple-choice": SCREAMING_SNAKE_CASE : Dict = {0: '''batch''', 1: '''choice''', 2: '''sequence'''} else: SCREAMING_SNAKE_CASE : List[Any] = {0: '''batch''', 1: '''sequence'''} return OrderedDict( [ ('''input_ids''', dynamic_axis), ('''attention_mask''', dynamic_axis), ] )
313
1
import inspect import unittest from transformers import SegformerConfig, is_torch_available, is_vision_available from transformers.models.auto import get_values from transformers.testing_utils import require_torch, slow, torch_device 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 ( MODEL_MAPPING, SegformerForImageClassification, SegformerForSemanticSegmentation, SegformerModel, ) from transformers.models.segformer.modeling_segformer import SEGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST if is_vision_available(): from PIL import Image from transformers import SegformerImageProcessor class a_ ( a__ ): """simple docstring""" def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Optional[int] = self.config_class(**self.inputs_dict ) self.parent.assertTrue(hasattr(_lowerCamelCase , '''hidden_sizes''' ) ) self.parent.assertTrue(hasattr(_lowerCamelCase , '''num_attention_heads''' ) ) self.parent.assertTrue(hasattr(_lowerCamelCase , '''num_encoder_blocks''' ) ) class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=13 , _lowerCamelCase=64 , _lowerCamelCase=3 , _lowerCamelCase=4 , _lowerCamelCase=[2, 2, 2, 2] , _lowerCamelCase=[8, 4, 2, 1] , _lowerCamelCase=[16, 32, 64, 128] , _lowerCamelCase=[1, 4, 8, 16] , _lowerCamelCase=[1, 2, 4, 8] , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=0.0_2 , _lowerCamelCase=3 , _lowerCamelCase=None , ) ->Optional[Any]: SCREAMING_SNAKE_CASE : List[str] = parent SCREAMING_SNAKE_CASE : Optional[int] = batch_size SCREAMING_SNAKE_CASE : Any = image_size SCREAMING_SNAKE_CASE : str = num_channels SCREAMING_SNAKE_CASE : Tuple = num_encoder_blocks SCREAMING_SNAKE_CASE : Tuple = sr_ratios SCREAMING_SNAKE_CASE : Optional[int] = depths SCREAMING_SNAKE_CASE : List[Any] = hidden_sizes SCREAMING_SNAKE_CASE : Dict = downsampling_rates SCREAMING_SNAKE_CASE : Union[str, Any] = num_attention_heads SCREAMING_SNAKE_CASE : int = is_training SCREAMING_SNAKE_CASE : Optional[Any] = use_labels SCREAMING_SNAKE_CASE : Any = hidden_act SCREAMING_SNAKE_CASE : Any = hidden_dropout_prob SCREAMING_SNAKE_CASE : Union[str, Any] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Dict = initializer_range SCREAMING_SNAKE_CASE : Union[str, Any] = num_labels SCREAMING_SNAKE_CASE : Optional[Any] = scope def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : List[str] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) SCREAMING_SNAKE_CASE : List[str] = None if self.use_labels: SCREAMING_SNAKE_CASE : List[str] = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels ) SCREAMING_SNAKE_CASE : Optional[Any] = self.get_config() return config, pixel_values, labels def __lowerCAmelCase ( self ) ->Optional[int]: return SegformerConfig( image_size=self.image_size , num_channels=self.num_channels , num_encoder_blocks=self.num_encoder_blocks , depths=self.depths , hidden_sizes=self.hidden_sizes , num_attention_heads=self.num_attention_heads , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , initializer_range=self.initializer_range , ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = SegformerModel(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = self.image_size // (self.downsampling_rates[-1] * 2) self.parent.assertEqual( result.last_hidden_state.shape , (self.batch_size, self.hidden_sizes[-1], expected_height, expected_width) ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[Any] = self.num_labels SCREAMING_SNAKE_CASE : List[Any] = SegformerForSemanticSegmentation(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase ) self.parent.assertEqual( result.logits.shape , (self.batch_size, self.num_labels, self.image_size // 4, self.image_size // 4) ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertEqual( result.logits.shape , (self.batch_size, self.num_labels, self.image_size // 4, self.image_size // 4) ) self.parent.assertGreater(result.loss , 0.0 ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) ->str: SCREAMING_SNAKE_CASE : List[Any] = 1 SCREAMING_SNAKE_CASE : List[str] = SegformerForSemanticSegmentation(config=_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() SCREAMING_SNAKE_CASE : Optional[Any] = torch.randint(0 , 1 , (self.batch_size, self.image_size, self.image_size) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , labels=_lowerCamelCase ) self.parent.assertGreater(result.loss , 0.0 ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : List[Any] = self.prepare_config_and_inputs() SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = config_and_inputs SCREAMING_SNAKE_CASE : List[Any] = {'''pixel_values''': pixel_values} return config, inputs_dict @require_torch class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = ( ( SegformerModel, SegformerForSemanticSegmentation, SegformerForImageClassification, ) if is_torch_available() else () ) __SCREAMING_SNAKE_CASE : List[Any] = ( { 'feature-extraction': SegformerModel, 'image-classification': SegformerForImageClassification, 'image-segmentation': SegformerForSemanticSegmentation, } if is_torch_available() else {} ) __SCREAMING_SNAKE_CASE : int = True __SCREAMING_SNAKE_CASE : Optional[int] = False __SCREAMING_SNAKE_CASE : Union[str, Any] = False __SCREAMING_SNAKE_CASE : Any = False def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Optional[int] = SegformerModelTester(self ) SCREAMING_SNAKE_CASE : Tuple = SegformerConfigTester(self , config_class=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: self.config_tester.run_common_tests() def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : int = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_binary_image_segmentation(*_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Dict = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_image_segmentation(*_lowerCamelCase ) @unittest.skip('''SegFormer does not use inputs_embeds''' ) def __lowerCAmelCase ( self ) ->Tuple: pass @unittest.skip('''SegFormer does not have get_input_embeddings method and get_output_embeddings methods''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Optional[Any] = model_class(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = inspect.signature(model.forward ) # signature.parameters is an OrderedDict => so arg_names order is deterministic SCREAMING_SNAKE_CASE : Dict = [*signature.parameters.keys()] SCREAMING_SNAKE_CASE : Any = ['''pixel_values'''] self.assertListEqual(arg_names[:1] , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : Tuple = True for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : Dict = True SCREAMING_SNAKE_CASE : Optional[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = True SCREAMING_SNAKE_CASE : Any = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : List[Any] = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = outputs.attentions SCREAMING_SNAKE_CASE : List[str] = sum(self.model_tester.depths ) self.assertEqual(len(_lowerCamelCase ) , _lowerCamelCase ) # check that output_attentions also work using config del inputs_dict["output_attentions"] SCREAMING_SNAKE_CASE : List[Any] = True SCREAMING_SNAKE_CASE : Union[str, Any] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : Optional[Any] = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = outputs.attentions self.assertEqual(len(_lowerCamelCase ) , _lowerCamelCase ) # verify the first attentions (first block, first layer) SCREAMING_SNAKE_CASE : int = (self.model_tester.image_size // 4) ** 2 SCREAMING_SNAKE_CASE : Optional[int] = (self.model_tester.image_size // (4 * self.model_tester.sr_ratios[0])) ** 2 self.assertListEqual( list(attentions[0].shape[-3:] ) , [self.model_tester.num_attention_heads[0], expected_seq_len, expected_reduced_seq_len] , ) # verify the last attentions (last block, last layer) SCREAMING_SNAKE_CASE : Union[str, Any] = (self.model_tester.image_size // 32) ** 2 SCREAMING_SNAKE_CASE : List[str] = (self.model_tester.image_size // (32 * self.model_tester.sr_ratios[-1])) ** 2 self.assertListEqual( list(attentions[-1].shape[-3:] ) , [self.model_tester.num_attention_heads[-1], expected_seq_len, expected_reduced_seq_len] , ) SCREAMING_SNAKE_CASE : int = len(_lowerCamelCase ) # Check attention is always last and order is fine SCREAMING_SNAKE_CASE : Dict = True SCREAMING_SNAKE_CASE : int = True SCREAMING_SNAKE_CASE : List[str] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : List[Any] = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) self.assertEqual(out_len + 1 , len(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = outputs.attentions self.assertEqual(len(_lowerCamelCase ) , _lowerCamelCase ) # verify the first attentions (first block, first layer) SCREAMING_SNAKE_CASE : List[str] = (self.model_tester.image_size // 4) ** 2 SCREAMING_SNAKE_CASE : Dict = (self.model_tester.image_size // (4 * self.model_tester.sr_ratios[0])) ** 2 self.assertListEqual( list(self_attentions[0].shape[-3:] ) , [self.model_tester.num_attention_heads[0], expected_seq_len, expected_reduced_seq_len] , ) def __lowerCAmelCase ( self ) ->Tuple: def check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : Optional[int] = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.eval() with torch.no_grad(): SCREAMING_SNAKE_CASE : List[Any] = model(**self._prepare_for_class(_lowerCamelCase , _lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[Any] = outputs.hidden_states SCREAMING_SNAKE_CASE : List[str] = self.model_tester.num_encoder_blocks self.assertEqual(len(_lowerCamelCase ) , _lowerCamelCase ) # verify the first hidden states (first block) self.assertListEqual( list(hidden_states[0].shape[-3:] ) , [ self.model_tester.hidden_sizes[0], self.model_tester.image_size // 4, self.model_tester.image_size // 4, ] , ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: SCREAMING_SNAKE_CASE : int = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) # check that output_hidden_states also work using config del inputs_dict["output_hidden_states"] SCREAMING_SNAKE_CASE : Any = True check_hidden_states_output(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Dict: if not self.model_tester.is_training: return SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common() SCREAMING_SNAKE_CASE : Optional[int] = True for model_class in self.all_model_classes: if model_class in get_values(_lowerCamelCase ): continue SCREAMING_SNAKE_CASE : Tuple = model_class(_lowerCamelCase ) model.to(_lowerCamelCase ) model.train() SCREAMING_SNAKE_CASE : Optional[Any] = self._prepare_for_class(_lowerCamelCase , _lowerCamelCase , return_labels=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = model(**_lowerCamelCase ).loss loss.backward() @unittest.skip('''Will be fixed soon by reducing the size of the model used for common tests.''' ) def __lowerCAmelCase ( self ) ->List[Any]: pass @slow def __lowerCAmelCase ( self ) ->List[Any]: for model_name in SEGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: SCREAMING_SNAKE_CASE : Optional[Any] = SegformerModel.from_pretrained(_lowerCamelCase ) self.assertIsNotNone(_lowerCamelCase ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) return image @require_torch class a_ ( unittest.TestCase ): """simple docstring""" @slow def __lowerCAmelCase ( self ) ->str: # only resize + normalize SCREAMING_SNAKE_CASE : Dict = SegformerImageProcessor( image_scale=(512, 512) , keep_ratio=_lowerCamelCase , align=_lowerCamelCase , do_random_crop=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = SegformerForSemanticSegmentation.from_pretrained('''nvidia/segformer-b0-finetuned-ade-512-512''' ).to( _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = prepare_img() SCREAMING_SNAKE_CASE : Optional[int] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ) SCREAMING_SNAKE_CASE : int = encoded_inputs.pixel_values.to(_lowerCamelCase ) with torch.no_grad(): SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = torch.Size((1, model.config.num_labels, 128, 128) ) self.assertEqual(outputs.logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = torch.tensor( [ [[-4.6_3_1_0, -5.5_2_3_2, -6.2_3_5_6], [-5.1_9_2_1, -6.1_4_4_4, -6.5_9_9_6], [-5.4_4_2_4, -6.2_7_9_0, -6.7_5_7_4]], [[-1_2.1_3_9_1, -1_3.3_1_2_2, -1_3.9_5_5_4], [-1_2.8_7_3_2, -1_3.9_3_5_2, -1_4.3_5_6_3], [-1_2.9_4_3_8, -1_3.8_2_2_6, -1_4.2_5_1_3]], [[-1_2.5_1_3_4, -1_3.4_6_8_6, -1_4.4_9_1_5], [-1_2.8_6_6_9, -1_4.4_3_4_3, -1_4.7_7_5_8], [-1_3.2_5_2_3, -1_4.5_8_1_9, -1_5.0_6_9_4]], ] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(outputs.logits[0, :3, :3, :3] , _lowerCamelCase , atol=1e-4 ) ) @slow def __lowerCAmelCase ( self ) ->str: # only resize + normalize SCREAMING_SNAKE_CASE : Union[str, Any] = SegformerImageProcessor( image_scale=(512, 512) , keep_ratio=_lowerCamelCase , align=_lowerCamelCase , do_random_crop=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = SegformerForSemanticSegmentation.from_pretrained( '''nvidia/segformer-b1-finetuned-cityscapes-1024-1024''' ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = prepare_img() SCREAMING_SNAKE_CASE : Optional[Any] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ) SCREAMING_SNAKE_CASE : Tuple = encoded_inputs.pixel_values.to(_lowerCamelCase ) with torch.no_grad(): SCREAMING_SNAKE_CASE : int = model(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = torch.Size((1, model.config.num_labels, 128, 128) ) self.assertEqual(outputs.logits.shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.tensor( [ [[-1_3.5_7_4_8, -1_3.9_1_1_1, -1_2.6_5_0_0], [-1_4.3_5_0_0, -1_5.3_6_8_3, -1_4.2_3_2_8], [-1_4.7_5_3_2, -1_6.0_4_2_4, -1_5.6_0_8_7]], [[-1_7.1_6_5_1, -1_5.8_7_2_5, -1_2.9_6_5_3], [-1_7.2_5_8_0, -1_7.3_7_1_8, -1_4.8_2_2_3], [-1_6.6_0_5_8, -1_6.8_7_8_3, -1_6.7_4_5_2]], [[-3.6_4_5_6, -3.0_2_0_9, -1.4_2_0_3], [-3.0_7_9_7, -3.1_9_5_9, -2.0_0_0_0], [-1.8_7_5_7, -1.9_2_1_7, -1.6_9_9_7]], ] ).to(_lowerCamelCase ) self.assertTrue(torch.allclose(outputs.logits[0, :3, :3, :3] , _lowerCamelCase , atol=1e-1 ) ) @slow def __lowerCAmelCase ( self ) ->str: # only resize + normalize SCREAMING_SNAKE_CASE : str = SegformerImageProcessor( image_scale=(512, 512) , keep_ratio=_lowerCamelCase , align=_lowerCamelCase , do_random_crop=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = SegformerForSemanticSegmentation.from_pretrained('''nvidia/segformer-b0-finetuned-ade-512-512''' ).to( _lowerCamelCase ) SCREAMING_SNAKE_CASE : int = prepare_img() SCREAMING_SNAKE_CASE : List[Any] = image_processor(images=_lowerCamelCase , return_tensors='''pt''' ) SCREAMING_SNAKE_CASE : Any = encoded_inputs.pixel_values.to(_lowerCamelCase ) with torch.no_grad(): SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = outputs.logits.detach().cpu() SCREAMING_SNAKE_CASE : Dict = image_processor.post_process_semantic_segmentation(outputs=_lowerCamelCase , target_sizes=[(500, 300)] ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Size((500, 300) ) self.assertEqual(segmentation[0].shape , _lowerCamelCase ) SCREAMING_SNAKE_CASE : str = image_processor.post_process_semantic_segmentation(outputs=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = torch.Size((128, 128) ) self.assertEqual(segmentation[0].shape , _lowerCamelCase )
313
from collections import OrderedDict from typing import TYPE_CHECKING, Any, Mapping, Optional from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging if TYPE_CHECKING: from ... import FeatureExtractionMixin, TensorType a__ : Any = logging.get_logger(__name__) a__ : Dict = { '''openai/imagegpt-small''': '''''', '''openai/imagegpt-medium''': '''''', '''openai/imagegpt-large''': '''''', } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 'imagegpt' __SCREAMING_SNAKE_CASE : Optional[Any] = ['past_key_values'] __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'n_embd', 'max_position_embeddings': 'n_positions', 'num_attention_heads': 'n_head', 'num_hidden_layers': 'n_layer', } def __init__( self , _lowerCamelCase=512 + 1 , _lowerCamelCase=32 * 32 , _lowerCamelCase=512 , _lowerCamelCase=24 , _lowerCamelCase=8 , _lowerCamelCase=None , _lowerCamelCase="quick_gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=1e-5 , _lowerCamelCase=0.0_2 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=False , _lowerCamelCase=False , **_lowerCamelCase , ) ->str: SCREAMING_SNAKE_CASE : Any = vocab_size SCREAMING_SNAKE_CASE : Optional[int] = n_positions SCREAMING_SNAKE_CASE : Optional[int] = n_embd SCREAMING_SNAKE_CASE : List[Any] = n_layer SCREAMING_SNAKE_CASE : List[Any] = n_head SCREAMING_SNAKE_CASE : int = n_inner SCREAMING_SNAKE_CASE : Dict = activation_function SCREAMING_SNAKE_CASE : Union[str, Any] = resid_pdrop SCREAMING_SNAKE_CASE : Dict = embd_pdrop SCREAMING_SNAKE_CASE : List[str] = attn_pdrop SCREAMING_SNAKE_CASE : Optional[int] = layer_norm_epsilon SCREAMING_SNAKE_CASE : Tuple = initializer_range SCREAMING_SNAKE_CASE : int = scale_attn_weights SCREAMING_SNAKE_CASE : Optional[int] = use_cache SCREAMING_SNAKE_CASE : Optional[Any] = scale_attn_by_inverse_layer_idx SCREAMING_SNAKE_CASE : str = reorder_and_upcast_attn SCREAMING_SNAKE_CASE : List[str] = tie_word_embeddings super().__init__(tie_word_embeddings=_lowerCamelCase , **_lowerCamelCase ) class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: return OrderedDict( [ ('''input_ids''', {0: '''batch''', 1: '''sequence'''}), ] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 1 , _lowerCamelCase = -1 , _lowerCamelCase = False , _lowerCamelCase = None , _lowerCamelCase = 3 , _lowerCamelCase = 32 , _lowerCamelCase = 32 , ) ->Mapping[str, Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = self._generate_dummy_images(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = dict(preprocessor(images=_lowerCamelCase , return_tensors=_lowerCamelCase ) ) return inputs
313
1
import unittest from transformers.testing_utils import require_bsa from transformers.utils import is_bsa_available from ...test_feature_extraction_common import FeatureExtractionSavingTestMixin if is_bsa_available(): from transformers import MarkupLMFeatureExtractor class a_ ( unittest.TestCase ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->Optional[int]: SCREAMING_SNAKE_CASE : Any = parent def __lowerCAmelCase ( self ) ->Union[str, Any]: return {} def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : int = '''<HTML> <HEAD> <TITLE>sample document</TITLE> </HEAD> <BODY BGCOLOR="FFFFFF"> <HR> <a href="http://google.com">Goog</a> <H1>This is one header</H1> <H2>This is a another Header</H2> <P>Travel from <P> <B>SFO to JFK</B> <BR> <B><I>on May 2, 2015 at 2:00 pm. For details go to confirm.com </I></B> <HR> <div style="color:#0000FF"> <h3>Traveler <b> name </b> is <p> John Doe </p> </div>''' SCREAMING_SNAKE_CASE : Any = ''' <!DOCTYPE html> <html> <body> <h1>My First Heading</h1> <p>My first paragraph.</p> </body> </html> ''' return [html_string_a, html_string_a] @require_bsa class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = MarkupLMFeatureExtractor if is_bsa_available() else None def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = MarkupLMFeatureExtractionTester(self ) @property def __lowerCAmelCase ( self ) ->Tuple: return self.feature_extract_tester.prepare_feat_extract_dict() def __lowerCAmelCase ( self ) ->Any: # Initialize feature_extractor SCREAMING_SNAKE_CASE : int = self.feature_extraction_class() # Test not batched input SCREAMING_SNAKE_CASE : Tuple = get_html_strings()[0] SCREAMING_SNAKE_CASE : str = feature_extractor(_lowerCamelCase ) # fmt: off SCREAMING_SNAKE_CASE : int = [['''sample document''', '''Goog''', '''This is one header''', '''This is a another Header''', '''Travel from''', '''SFO to JFK''', '''on May 2, 2015 at 2:00 pm. For details go to confirm.com''', '''Traveler''', '''name''', '''is''', '''John Doe''']] SCREAMING_SNAKE_CASE : Any = [['''/html/head/title''', '''/html/body/a''', '''/html/body/h1''', '''/html/body/h2''', '''/html/body/p''', '''/html/body/p/p/b[1]''', '''/html/body/p/p/b[2]/i''', '''/html/body/p/p/div/h3''', '''/html/body/p/p/div/h3/b''', '''/html/body/p/p/div/h3''', '''/html/body/p/p/div/h3/p''']] # fmt: on self.assertEqual(encoding.nodes , _lowerCamelCase ) self.assertEqual(encoding.xpaths , _lowerCamelCase ) # Test batched SCREAMING_SNAKE_CASE : Optional[int] = get_html_strings() SCREAMING_SNAKE_CASE : Union[str, Any] = feature_extractor(_lowerCamelCase ) # fmt: off SCREAMING_SNAKE_CASE : List[str] = expected_nodes + [['''My First Heading''', '''My first paragraph.''']] SCREAMING_SNAKE_CASE : int = expected_xpaths + [['''/html/body/h1''', '''/html/body/p''']] self.assertEqual(len(encoding.nodes ) , 2 ) self.assertEqual(len(encoding.xpaths ) , 2 ) self.assertEqual(encoding.nodes , _lowerCamelCase ) self.assertEqual(encoding.xpaths , _lowerCamelCase )
313
from maths.prime_check import is_prime def UpperCAmelCase_( a__ ): """simple docstring""" if not isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = F"""Input value of [number={number}] must be an integer""" raise TypeError(a__ ) if is_prime(a__ ) and is_prime(number + 2 ): return number + 2 else: return -1 if __name__ == "__main__": import doctest doctest.testmod()
313
1
import warnings from ...utils import logging from .image_processing_donut import DonutImageProcessor a__ : str = logging.get_logger(__name__) class a_ ( a__ ): """simple docstring""" def __init__( self , *_lowerCamelCase , **_lowerCamelCase ) ->None: warnings.warn( '''The class DonutFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please''' ''' use DonutImageProcessor instead.''' , _lowerCamelCase , ) super().__init__(*_lowerCamelCase , **_lowerCamelCase )
313
import gc import random import unittest import numpy as np import torch from PIL import Image from diffusers import ( DDIMScheduler, KandinskyVaaControlnetImgaImgPipeline, KandinskyVaaPriorEmbaEmbPipeline, UNetaDConditionModel, VQModel, ) from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[Any] = KandinskyVaaControlnetImgaImgPipeline __SCREAMING_SNAKE_CASE : Optional[int] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[Any] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[str] = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->Optional[Any]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 32 @property def __lowerCAmelCase ( self ) ->str: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Dict: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Tuple: return 100 @property def __lowerCAmelCase ( self ) ->int: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Union[str, Any] = { '''in_channels''': 8, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''image_hint''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : List[str] = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->Any: return { "block_out_channels": [32, 32, 64, 64], "down_block_types": [ "DownEncoderBlock2D", "DownEncoderBlock2D", "DownEncoderBlock2D", "AttnDownEncoderBlock2D", ], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": ["AttnUpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D"], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : str = self.dummy_unet SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_movq SCREAMING_SNAKE_CASE : List[str] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : str = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = { '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->int: SCREAMING_SNAKE_CASE : str = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(seed + 1 ) ).to( _lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : Dict = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) # create hint SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : List[str] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = { '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''hint''': hint, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[int] = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : Any = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : Union[str, Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array( [0.5_4_9_8_5_0_3_4, 0.5_5_5_0_9_3_6_5, 0.5_2_5_6_1_5_0_4, 0.5_5_7_0_4_9_4, 0.5_5_9_3_8_1_8, 0.5_2_6_3_9_7_9, 0.5_0_2_8_5_6_4_3, 0.5_0_6_9_8_4_6, 0.5_1_1_9_6_7_3_6] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Union[str, Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/kandinskyv22_controlnet_img2img_robotcat_fp16.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = init_image.resize((512, 512) ) SCREAMING_SNAKE_CASE : Dict = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/hint_image_cat.png''' ) SCREAMING_SNAKE_CASE : List[Any] = torch.from_numpy(np.array(_lowerCamelCase ) ).float() / 2_5_5.0 SCREAMING_SNAKE_CASE : int = hint.permute(2 , 0 , 1 ).unsqueeze(0 ) SCREAMING_SNAKE_CASE : List[Any] = '''A robot, 4k photo''' SCREAMING_SNAKE_CASE : List[str] = KandinskyVaaPriorEmbaEmbPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = KandinskyVaaControlnetImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-controlnet-depth''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Any = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , image=_lowerCamelCase , strength=0.8_5 , generator=_lowerCamelCase , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : List[str] = pipeline( image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , hint=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=512 , width=512 , strength=0.5 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = output.images[0] assert image.shape == (512, 512, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
import json import logging import math import os import sys from dataclasses import dataclass, field from typing import Optional from datasets import Dataset, load_dataset import transformers from transformers import ( CONFIG_MAPPING, MODEL_FOR_MASKED_LM_MAPPING, AutoConfig, AutoModelForMaskedLM, AutoTokenizer, DataCollatorForWholeWordMask, HfArgumentParser, Trainer, TrainingArguments, set_seed, ) from transformers.trainer_utils import get_last_checkpoint, is_main_process a__ : Dict = logging.getLogger(__name__) a__ : Union[str, Any] = list(MODEL_FOR_MASKED_LM_MAPPING.keys()) a__ : Union[str, Any] = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES) @dataclass class a_ : """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={ 'help': ( 'The model checkpoint for weights initialization.Don\'t set if you want to train a model from scratch.' ) } , ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'If training from scratch, pass a model type from the list: ' + ', '.join(a__ )} , ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , 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 : Optional[str] = field( default=a__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , ) __SCREAMING_SNAKE_CASE : bool = field( default=a__ , metadata={'help': 'Whether to use one of the fast tokenizer (backed by the tokenizers library) or not.'} , ) __SCREAMING_SNAKE_CASE : str = field( default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , ) __SCREAMING_SNAKE_CASE : bool = field( default=a__ , metadata={ 'help': ( 'Will use the token generated when running `huggingface-cli login` (necessary to use this script ' 'with private models).' ) } , ) def __lowerCAmelCase ( self ) ->Any: if self.config_overrides is not None and (self.config_name is not None or self.model_name_or_path is not None): raise ValueError( '''--config_overrides can\'t be used in combination with --config_name or --model_name_or_path''' ) @dataclass class a_ : """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'The name of the dataset to use (via the datasets library).'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field(default=a__ , metadata={'help': 'The input training data file (a text file).'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'An optional input evaluation data file to evaluate the perplexity on (a text file).'} , ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'An optional input train ref data file for whole word masking in Chinese.'} , ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'An optional input validation ref data file for whole word masking in Chinese.'} , ) __SCREAMING_SNAKE_CASE : bool = field( default=a__ , metadata={'help': 'Overwrite the cached training and evaluation sets'} ) __SCREAMING_SNAKE_CASE : Optional[int] = field( default=5 , metadata={ 'help': 'The percentage of the train set used as validation set in case there\'s no validation split' } , ) __SCREAMING_SNAKE_CASE : Optional[int] = field( default=a__ , metadata={ 'help': ( 'The maximum total input sequence length after tokenization. Sequences longer ' 'than this will be truncated. Default to the max input length of the model.' ) } , ) __SCREAMING_SNAKE_CASE : Optional[int] = field( default=a__ , metadata={'help': 'The number of processes to use for the preprocessing.'} , ) __SCREAMING_SNAKE_CASE : float = field( default=0.1_5 , metadata={'help': 'Ratio of tokens to mask for masked language modeling loss'} ) __SCREAMING_SNAKE_CASE : bool = field( default=a__ , 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.' ) } , ) def __lowerCAmelCase ( self ) ->Dict: if self.train_file is not None: SCREAMING_SNAKE_CASE : Any = self.train_file.split('''.''' )[-1] assert extension in ["csv", "json", "txt"], "`train_file` should be a csv, a json or a txt file." if self.validation_file is not None: SCREAMING_SNAKE_CASE : Optional[Any] = self.validation_file.split('''.''' )[-1] assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, a json or a txt file." def UpperCAmelCase_( a__ , a__ ): """simple docstring""" with open(a__ , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : Any = [json.loads(a__ ) for line in f.read().splitlines() if (len(a__ ) > 0 and not line.isspace())] assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : int = {c: dataset[c] for c in dataset.column_names} SCREAMING_SNAKE_CASE : Tuple = refs return Dataset.from_dict(a__ ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = 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. SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) ) else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = parser.parse_args_into_dataclasses() # Detecting last checkpoint. SCREAMING_SNAKE_CASE : Dict = None if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir: SCREAMING_SNAKE_CASE : List[Any] = 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: 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.''' ) # 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 )] , ) logger.setLevel(logging.INFO if is_main_process(training_args.local_rank ) else logging.WARN ) # 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}""" ) # Set the verbosity to info of the Transformers logger (on main process only): if is_main_process(training_args.local_rank ): transformers.utils.logging.set_verbosity_info() transformers.utils.logging.enable_default_handler() transformers.utils.logging.enable_explicit_format() logger.info('''Training/evaluation parameters %s''' , a__ ) # Set seed before initializing model. set_seed(training_args.seed ) # Get the datasets: you can either provide your own CSV/JSON/TXT training and evaluation files (see below) # or just provide the name of one of the public datasets available on the hub at https://huggingface.co/datasets/ # (the dataset will be downloaded automatically from the datasets Hub). # # For CSV/JSON files, this script will use the column called 'text' or the first column if no column called # 'text' is found. 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. SCREAMING_SNAKE_CASE : List[Any] = load_dataset(data_args.dataset_name , data_args.dataset_config_name ) if "validation" not in datasets.keys(): SCREAMING_SNAKE_CASE : str = load_dataset( data_args.dataset_name , data_args.dataset_config_name , split=F"""train[:{data_args.validation_split_percentage}%]""" , ) SCREAMING_SNAKE_CASE : int = load_dataset( data_args.dataset_name , data_args.dataset_config_name , split=F"""train[{data_args.validation_split_percentage}%:]""" , ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = {} if data_args.train_file is not None: SCREAMING_SNAKE_CASE : str = data_args.train_file if data_args.validation_file is not None: SCREAMING_SNAKE_CASE : str = data_args.validation_file SCREAMING_SNAKE_CASE : List[str] = data_args.train_file.split('''.''' )[-1] if extension == "txt": SCREAMING_SNAKE_CASE : str = '''text''' SCREAMING_SNAKE_CASE : Dict = load_dataset(a__ , data_files=a__ ) # See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at # https://huggingface.co/docs/datasets/loading_datasets.html. # Load pretrained model and tokenizer # # Distributed training: # The .from_pretrained methods guarantee that only one local process can concurrently # download model & vocab. SCREAMING_SNAKE_CASE : Any = { '''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: SCREAMING_SNAKE_CASE : int = AutoConfig.from_pretrained(model_args.config_name , **a__ ) elif model_args.model_name_or_path: SCREAMING_SNAKE_CASE : Optional[Any] = AutoConfig.from_pretrained(model_args.model_name_or_path , **a__ ) else: SCREAMING_SNAKE_CASE : List[Any] = CONFIG_MAPPING[model_args.model_type]() 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}""" ) SCREAMING_SNAKE_CASE : str = { '''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, } if model_args.tokenizer_name: SCREAMING_SNAKE_CASE : Optional[int] = AutoTokenizer.from_pretrained(model_args.tokenizer_name , **a__ ) elif model_args.model_name_or_path: SCREAMING_SNAKE_CASE : str = AutoTokenizer.from_pretrained(model_args.model_name_or_path , **a__ ) else: raise ValueError( '''You are instantiating a new tokenizer from scratch. This is not supported by this script.''' '''You can do it from another script, save it, and load it from here, using --tokenizer_name.''' ) if model_args.model_name_or_path: SCREAMING_SNAKE_CASE : str = AutoModelForMaskedLM.from_pretrained( model_args.model_name_or_path , from_tf=bool('''.ckpt''' in model_args.model_name_or_path ) , config=a__ , 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''' ) SCREAMING_SNAKE_CASE : Dict = AutoModelForMaskedLM.from_config(a__ ) model.resize_token_embeddings(len(a__ ) ) # Preprocessing the datasets. # First we tokenize all the texts. if training_args.do_train: SCREAMING_SNAKE_CASE : str = datasets['''train'''].column_names else: SCREAMING_SNAKE_CASE : int = datasets['''validation'''].column_names SCREAMING_SNAKE_CASE : List[str] = '''text''' if '''text''' in column_names else column_names[0] SCREAMING_SNAKE_CASE : int = '''max_length''' if data_args.pad_to_max_length else False def tokenize_function(a__ ): # Remove empty lines SCREAMING_SNAKE_CASE : Optional[int] = [line for line in examples['''text'''] if len(a__ ) > 0 and not line.isspace()] return tokenizer(examples['''text'''] , padding=a__ , truncation=a__ , max_length=data_args.max_seq_length ) SCREAMING_SNAKE_CASE : Any = datasets.map( a__ , batched=a__ , num_proc=data_args.preprocessing_num_workers , remove_columns=[text_column_name] , load_from_cache_file=not data_args.overwrite_cache , ) # Add the chinese references if provided if data_args.train_ref_file is not None: SCREAMING_SNAKE_CASE : str = add_chinese_references(tokenized_datasets['''train'''] , data_args.train_ref_file ) if data_args.validation_ref_file is not None: SCREAMING_SNAKE_CASE : str = add_chinese_references( tokenized_datasets['''validation'''] , data_args.validation_ref_file ) # If we have ref files, need to avoid it removed by trainer SCREAMING_SNAKE_CASE : str = data_args.train_ref_file or data_args.validation_ref_file if has_ref: SCREAMING_SNAKE_CASE : Any = False # Data collator # This one will take care of randomly masking the tokens. SCREAMING_SNAKE_CASE : Optional[Any] = DataCollatorForWholeWordMask(tokenizer=a__ , mlm_probability=data_args.mlm_probability ) # Initialize our Trainer SCREAMING_SNAKE_CASE : Optional[int] = Trainer( model=a__ , args=a__ , train_dataset=tokenized_datasets['''train'''] if training_args.do_train else None , eval_dataset=tokenized_datasets['''validation'''] if training_args.do_eval else None , tokenizer=a__ , data_collator=a__ , ) # Training if training_args.do_train: if last_checkpoint is not None: SCREAMING_SNAKE_CASE : str = last_checkpoint elif model_args.model_name_or_path is not None and os.path.isdir(model_args.model_name_or_path ): SCREAMING_SNAKE_CASE : str = model_args.model_name_or_path else: SCREAMING_SNAKE_CASE : Dict = None SCREAMING_SNAKE_CASE : Tuple = trainer.train(resume_from_checkpoint=a__ ) trainer.save_model() # Saves the tokenizer too for easy upload SCREAMING_SNAKE_CASE : Optional[int] = os.path.join(training_args.output_dir , '''train_results.txt''' ) if trainer.is_world_process_zero(): with open(a__ , '''w''' ) as writer: logger.info('''***** Train results *****''' ) for key, value in sorted(train_result.metrics.items() ): logger.info(F""" {key} = {value}""" ) writer.write(F"""{key} = {value}\n""" ) # Need to save the state, since Trainer.save_model saves only the tokenizer with the model trainer.state.save_to_json(os.path.join(training_args.output_dir , '''trainer_state.json''' ) ) # Evaluation SCREAMING_SNAKE_CASE : Any = {} if training_args.do_eval: logger.info('''*** Evaluate ***''' ) SCREAMING_SNAKE_CASE : Dict = trainer.evaluate() SCREAMING_SNAKE_CASE : Any = math.exp(eval_output['''eval_loss'''] ) SCREAMING_SNAKE_CASE : Optional[int] = perplexity SCREAMING_SNAKE_CASE : Tuple = os.path.join(training_args.output_dir , '''eval_results_mlm_wwm.txt''' ) if trainer.is_world_process_zero(): with open(a__ , '''w''' ) as writer: logger.info('''***** Eval results *****''' ) for key, value in sorted(results.items() ): logger.info(F""" {key} = {value}""" ) writer.write(F"""{key} = {value}\n""" ) return results def UpperCAmelCase_( a__ ): """simple docstring""" main() if __name__ == "__main__": main()
313
from typing import Any, Callable, Dict, List, Optional, Union import torch from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer from diffusers import ( AutoencoderKL, DDIMScheduler, DiffusionPipeline, LMSDiscreteScheduler, PNDMScheduler, StableDiffusionPipeline, UNetaDConditionModel, ) from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker a__ : List[str] = '''CompVis/stable-diffusion-v1-1''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-2''' a__ : Any = '''CompVis/stable-diffusion-v1-3''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-4''' class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , ) ->str: super()._init_() SCREAMING_SNAKE_CASE : Tuple = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = StableDiffusionPipeline( vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , unet=_lowerCamelCase , scheduler=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=_lowerCamelCase , requires_safety_checker=_lowerCamelCase , ) self.register_modules(pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea ) @property def __lowerCAmelCase ( self ) ->Dict[str, Any]: return {k: getattr(self , _lowerCamelCase ) for k in self.config.keys() if not k.startswith('''_''' )} def __lowerCAmelCase ( self , _lowerCamelCase = "auto" ) ->Optional[int]: if slice_size == "auto": # half the attention head size is usually a good trade-off between # speed and memory SCREAMING_SNAKE_CASE : Optional[int] = self.unet.config.attention_head_dim // 2 self.unet.set_attention_slice(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: self.enable_attention_slicing(_lowerCamelCase ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->str: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Tuple: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Optional[Any]: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: SCREAMING_SNAKE_CASE : Optional[Any] = '''cuda''' if torch.cuda.is_available() else '''cpu''' self.to(_lowerCamelCase ) # Checks if the height and width are divisible by 8 or not if height % 8 != 0 or width % 8 != 0: raise ValueError(F"""`height` and `width` must be divisible by 8 but are {height} and {width}.""" ) # Get first result from Stable Diffusion Checkpoint v1.1 SCREAMING_SNAKE_CASE : Optional[Any] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.2 SCREAMING_SNAKE_CASE : Any = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.3 SCREAMING_SNAKE_CASE : Optional[int] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.4 SCREAMING_SNAKE_CASE : str = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get all result images into a single list and pass it via StableDiffusionPipelineOutput for final result return StableDiffusionPipelineOutput([resa[0], resa[0], resa[0], resa[0]] )
313
1
from typing import List from ...configuration_utils import PretrainedConfig from ...utils import logging a__ : str = logging.get_logger(__name__) a__ : str = { '''snap-research/efficientformer-l1-300''': ( '''https://huggingface.co/snap-research/efficientformer-l1-300/resolve/main/config.json''' ), } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'efficientformer' def __init__( self , _lowerCamelCase = [3, 2, 6, 4] , _lowerCamelCase = [48, 96, 224, 448] , _lowerCamelCase = [True, True, True, True] , _lowerCamelCase = 448 , _lowerCamelCase = 32 , _lowerCamelCase = 4 , _lowerCamelCase = 7 , _lowerCamelCase = 5 , _lowerCamelCase = 8 , _lowerCamelCase = 4 , _lowerCamelCase = 0.0 , _lowerCamelCase = 16 , _lowerCamelCase = 3 , _lowerCamelCase = 3 , _lowerCamelCase = 3 , _lowerCamelCase = 2 , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = 1 , _lowerCamelCase = True , _lowerCamelCase = True , _lowerCamelCase = 1e-5 , _lowerCamelCase = "gelu" , _lowerCamelCase = 0.0_2 , _lowerCamelCase = 1e-12 , _lowerCamelCase = 224 , _lowerCamelCase = 1e-05 , **_lowerCamelCase , ) ->None: super().__init__(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = hidden_act SCREAMING_SNAKE_CASE : Union[str, Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Any = hidden_sizes SCREAMING_SNAKE_CASE : Any = num_hidden_layers SCREAMING_SNAKE_CASE : Optional[int] = num_attention_heads SCREAMING_SNAKE_CASE : Union[str, Any] = initializer_range SCREAMING_SNAKE_CASE : Any = layer_norm_eps SCREAMING_SNAKE_CASE : Optional[int] = patch_size SCREAMING_SNAKE_CASE : Optional[int] = num_channels SCREAMING_SNAKE_CASE : str = depths SCREAMING_SNAKE_CASE : List[Any] = mlp_expansion_ratio SCREAMING_SNAKE_CASE : Any = downsamples SCREAMING_SNAKE_CASE : Optional[int] = dim SCREAMING_SNAKE_CASE : Tuple = key_dim SCREAMING_SNAKE_CASE : Optional[int] = attention_ratio SCREAMING_SNAKE_CASE : Optional[Any] = resolution SCREAMING_SNAKE_CASE : int = pool_size SCREAMING_SNAKE_CASE : int = downsample_patch_size SCREAMING_SNAKE_CASE : Optional[Any] = downsample_stride SCREAMING_SNAKE_CASE : str = downsample_pad SCREAMING_SNAKE_CASE : Optional[int] = drop_path_rate SCREAMING_SNAKE_CASE : Any = num_metaad_blocks SCREAMING_SNAKE_CASE : Any = distillation SCREAMING_SNAKE_CASE : Optional[Any] = use_layer_scale SCREAMING_SNAKE_CASE : int = layer_scale_init_value SCREAMING_SNAKE_CASE : Union[str, Any] = image_size SCREAMING_SNAKE_CASE : Dict = batch_norm_eps
313
from typing import Optional, Tuple, Union import flax import flax.linen as nn import jax import jax.numpy as jnp from flax.core.frozen_dict import FrozenDict from ..configuration_utils import ConfigMixin, flax_register_to_config from ..utils import BaseOutput from .embeddings_flax import FlaxTimestepEmbedding, FlaxTimesteps from .modeling_flax_utils import FlaxModelMixin from .unet_ad_blocks_flax import ( FlaxCrossAttnDownBlockaD, FlaxCrossAttnUpBlockaD, FlaxDownBlockaD, FlaxUNetMidBlockaDCrossAttn, FlaxUpBlockaD, ) @flax.struct.dataclass class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : jnp.ndarray @flax_register_to_config class a_ ( nn.Module , a__ , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = 32 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : Tuple[str] = ( "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D", ) __SCREAMING_SNAKE_CASE : Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D") __SCREAMING_SNAKE_CASE : Union[bool, Tuple[bool]] = False __SCREAMING_SNAKE_CASE : Tuple[int] = (320, 640, 1280, 1280) __SCREAMING_SNAKE_CASE : int = 2 __SCREAMING_SNAKE_CASE : Union[int, Tuple[int]] = 8 __SCREAMING_SNAKE_CASE : Optional[Union[int, Tuple[int]]] = None __SCREAMING_SNAKE_CASE : int = 1280 __SCREAMING_SNAKE_CASE : float = 0.0 __SCREAMING_SNAKE_CASE : bool = False __SCREAMING_SNAKE_CASE : jnp.dtype = jnp.floataa __SCREAMING_SNAKE_CASE : bool = True __SCREAMING_SNAKE_CASE : int = 0 __SCREAMING_SNAKE_CASE : bool = False def __lowerCAmelCase ( self , _lowerCamelCase ) ->FrozenDict: # init input tensors SCREAMING_SNAKE_CASE : List[Any] = (1, self.in_channels, self.sample_size, self.sample_size) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros(_lowerCamelCase , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Tuple = jnp.ones((1,) , dtype=jnp.intaa ) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros((1, 1, self.cross_attention_dim) , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = jax.random.split(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = {'''params''': params_rng, '''dropout''': dropout_rng} return self.init(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase )["params"] def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : List[str] = self.block_out_channels SCREAMING_SNAKE_CASE : Optional[int] = block_out_channels[0] * 4 if self.num_attention_heads is not None: raise ValueError( '''At the moment it is not possible to define the number of attention heads via `num_attention_heads` because of a naming issue as described in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131. Passing `num_attention_heads` will only be supported in diffusers v0.19.''' ) # If `num_attention_heads` is not defined (which is the case for most models) # it will default to `attention_head_dim`. This looks weird upon first reading it and it is. # The reason for this behavior is to correct for incorrectly named variables that were introduced # when this library was created. The incorrect naming was only discovered much later in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131 # Changing `attention_head_dim` to `num_attention_heads` for 40,000+ configurations is too backwards breaking # which is why we correct for the naming here. SCREAMING_SNAKE_CASE : List[str] = self.num_attention_heads or self.attention_head_dim # input SCREAMING_SNAKE_CASE : Optional[int] = nn.Conv( block_out_channels[0] , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) # time SCREAMING_SNAKE_CASE : Tuple = FlaxTimesteps( block_out_channels[0] , flip_sin_to_cos=self.flip_sin_to_cos , freq_shift=self.config.freq_shift ) SCREAMING_SNAKE_CASE : Dict = FlaxTimestepEmbedding(_lowerCamelCase , dtype=self.dtype ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.only_cross_attention if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (only_cross_attention,) * len(self.down_block_types ) if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (num_attention_heads,) * len(self.down_block_types ) # down SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Optional[Any] = block_out_channels[0] for i, down_block_type in enumerate(self.down_block_types ): SCREAMING_SNAKE_CASE : str = output_channel SCREAMING_SNAKE_CASE : int = block_out_channels[i] SCREAMING_SNAKE_CASE : List[Any] = i == len(_lowerCamelCase ) - 1 if down_block_type == "CrossAttnDownBlock2D": SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxCrossAttnDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , num_attention_heads=num_attention_heads[i] , add_downsample=not is_final_block , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , add_downsample=not is_final_block , dtype=self.dtype , ) down_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = down_blocks # mid SCREAMING_SNAKE_CASE : int = FlaxUNetMidBlockaDCrossAttn( in_channels=block_out_channels[-1] , dropout=self.dropout , num_attention_heads=num_attention_heads[-1] , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) # up SCREAMING_SNAKE_CASE : str = [] SCREAMING_SNAKE_CASE : str = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = reversed_block_out_channels[0] for i, up_block_type in enumerate(self.up_block_types ): SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Union[str, Any] = reversed_block_out_channels[i] SCREAMING_SNAKE_CASE : Tuple = reversed_block_out_channels[min(i + 1 , len(_lowerCamelCase ) - 1 )] SCREAMING_SNAKE_CASE : Dict = i == len(_lowerCamelCase ) - 1 if up_block_type == "CrossAttnUpBlock2D": SCREAMING_SNAKE_CASE : str = FlaxCrossAttnUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , num_attention_heads=reversed_num_attention_heads[i] , add_upsample=not is_final_block , dropout=self.dropout , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Optional[int] = FlaxUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , add_upsample=not is_final_block , dropout=self.dropout , dtype=self.dtype , ) up_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Tuple = up_blocks # out SCREAMING_SNAKE_CASE : Any = nn.GroupNorm(num_groups=32 , epsilon=1e-5 ) SCREAMING_SNAKE_CASE : Any = nn.Conv( self.out_channels , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) def __call__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase = True , _lowerCamelCase = False , ) ->Union[FlaxUNetaDConditionOutput, Tuple]: # 1. time if not isinstance(_lowerCamelCase , jnp.ndarray ): SCREAMING_SNAKE_CASE : int = jnp.array([timesteps] , dtype=jnp.intaa ) elif isinstance(_lowerCamelCase , jnp.ndarray ) and len(timesteps.shape ) == 0: SCREAMING_SNAKE_CASE : List[str] = timesteps.astype(dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.expand_dims(_lowerCamelCase , 0 ) SCREAMING_SNAKE_CASE : List[str] = self.time_proj(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = self.time_embedding(_lowerCamelCase ) # 2. pre-process SCREAMING_SNAKE_CASE : int = jnp.transpose(_lowerCamelCase , (0, 2, 3, 1) ) SCREAMING_SNAKE_CASE : List[Any] = self.conv_in(_lowerCamelCase ) # 3. down SCREAMING_SNAKE_CASE : Optional[int] = (sample,) for down_block in self.down_blocks: if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , deterministic=not train ) down_block_res_samples += res_samples if down_block_additional_residuals is not None: SCREAMING_SNAKE_CASE : int = () for down_block_res_sample, down_block_additional_residual in zip( _lowerCamelCase , _lowerCamelCase ): down_block_res_sample += down_block_additional_residual new_down_block_res_samples += (down_block_res_sample,) SCREAMING_SNAKE_CASE : Dict = new_down_block_res_samples # 4. mid SCREAMING_SNAKE_CASE : Optional[Any] = self.mid_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) if mid_block_additional_residual is not None: sample += mid_block_additional_residual # 5. up for up_block in self.up_blocks: SCREAMING_SNAKE_CASE : Optional[Any] = down_block_res_samples[-(self.layers_per_block + 1) :] SCREAMING_SNAKE_CASE : Optional[int] = down_block_res_samples[: -(self.layers_per_block + 1)] if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = up_block( _lowerCamelCase , temb=_lowerCamelCase , encoder_hidden_states=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train , ) else: SCREAMING_SNAKE_CASE : Optional[int] = up_block(_lowerCamelCase , temb=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train ) # 6. post-process SCREAMING_SNAKE_CASE : Optional[int] = self.conv_norm_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.silu(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self.conv_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.transpose(_lowerCamelCase , (0, 3, 1, 2) ) if not return_dict: return (sample,) return FlaxUNetaDConditionOutput(sample=_lowerCamelCase )
313
1
from ...configuration_utils import PretrainedConfig from ...utils import logging a__ : Optional[int] = logging.get_logger(__name__) a__ : List[Any] = { '''funnel-transformer/small''': '''https://huggingface.co/funnel-transformer/small/resolve/main/config.json''', '''funnel-transformer/small-base''': '''https://huggingface.co/funnel-transformer/small-base/resolve/main/config.json''', '''funnel-transformer/medium''': '''https://huggingface.co/funnel-transformer/medium/resolve/main/config.json''', '''funnel-transformer/medium-base''': '''https://huggingface.co/funnel-transformer/medium-base/resolve/main/config.json''', '''funnel-transformer/intermediate''': ( '''https://huggingface.co/funnel-transformer/intermediate/resolve/main/config.json''' ), '''funnel-transformer/intermediate-base''': ( '''https://huggingface.co/funnel-transformer/intermediate-base/resolve/main/config.json''' ), '''funnel-transformer/large''': '''https://huggingface.co/funnel-transformer/large/resolve/main/config.json''', '''funnel-transformer/large-base''': '''https://huggingface.co/funnel-transformer/large-base/resolve/main/config.json''', '''funnel-transformer/xlarge''': '''https://huggingface.co/funnel-transformer/xlarge/resolve/main/config.json''', '''funnel-transformer/xlarge-base''': '''https://huggingface.co/funnel-transformer/xlarge-base/resolve/main/config.json''', } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Union[str, Any] = 'funnel' __SCREAMING_SNAKE_CASE : List[str] = { 'hidden_size': 'd_model', 'num_attention_heads': 'n_head', } def __init__( self , _lowerCamelCase=3_0522 , _lowerCamelCase=[4, 4, 4] , _lowerCamelCase=None , _lowerCamelCase=2 , _lowerCamelCase=768 , _lowerCamelCase=12 , _lowerCamelCase=64 , _lowerCamelCase=3072 , _lowerCamelCase="gelu_new" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=0.0 , _lowerCamelCase=0.1 , _lowerCamelCase=None , _lowerCamelCase=1e-9 , _lowerCamelCase="mean" , _lowerCamelCase="relative_shift" , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=True , **_lowerCamelCase , ) ->Tuple: SCREAMING_SNAKE_CASE : Tuple = vocab_size SCREAMING_SNAKE_CASE : Any = block_sizes SCREAMING_SNAKE_CASE : Union[str, Any] = [1] * len(_lowerCamelCase ) if block_repeats is None else block_repeats assert len(_lowerCamelCase ) == len( self.block_repeats ), "`block_sizes` and `block_repeats` should have the same length." SCREAMING_SNAKE_CASE : Tuple = num_decoder_layers SCREAMING_SNAKE_CASE : Dict = d_model SCREAMING_SNAKE_CASE : int = n_head SCREAMING_SNAKE_CASE : Optional[int] = d_head SCREAMING_SNAKE_CASE : List[str] = d_inner SCREAMING_SNAKE_CASE : Optional[int] = hidden_act SCREAMING_SNAKE_CASE : Dict = hidden_dropout SCREAMING_SNAKE_CASE : Dict = attention_dropout SCREAMING_SNAKE_CASE : int = activation_dropout SCREAMING_SNAKE_CASE : Optional[Any] = initializer_range SCREAMING_SNAKE_CASE : Tuple = initializer_std SCREAMING_SNAKE_CASE : List[str] = layer_norm_eps assert pooling_type in [ "mean", "max", ], F"""Got {pooling_type} for `pooling_type` but only 'mean' and 'max' are supported.""" SCREAMING_SNAKE_CASE : List[Any] = pooling_type assert attention_type in [ "relative_shift", "factorized", ], F"""Got {attention_type} for `attention_type` but only 'relative_shift' and 'factorized' are supported.""" SCREAMING_SNAKE_CASE : Union[str, Any] = attention_type SCREAMING_SNAKE_CASE : int = separate_cls SCREAMING_SNAKE_CASE : List[Any] = truncate_seq SCREAMING_SNAKE_CASE : int = pool_q_only super().__init__(**_lowerCamelCase ) @property def __lowerCAmelCase ( self ) ->List[Any]: return sum(self.block_sizes ) @num_hidden_layers.setter def __lowerCAmelCase ( self , _lowerCamelCase ) ->str: raise NotImplementedError( '''This model does not support the setting of `num_hidden_layers`. Please set `block_sizes`.''' ) @property def __lowerCAmelCase ( self ) ->Tuple: return len(self.block_sizes ) @num_blocks.setter def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: raise NotImplementedError('''This model does not support the setting of `num_blocks`. Please set `block_sizes`.''' )
313
import gc import random import unittest import numpy as np import torch from transformers import ( CLIPImageProcessor, CLIPTextConfig, CLIPTextModel, CLIPTokenizer, CLIPVisionConfig, CLIPVisionModelWithProjection, ) from diffusers import AutoencoderKL, DDIMScheduler, DDPMScheduler, StableUnCLIPImgaImgPipeline, UNetaDConditionModel from diffusers.pipelines.pipeline_utils import DiffusionPipeline from diffusers.pipelines.stable_diffusion.stable_unclip_image_normalizer import StableUnCLIPImageNormalizer from diffusers.utils.import_utils import is_xformers_available from diffusers.utils.testing_utils import ( enable_full_determinism, floats_tensor, load_image, load_numpy, require_torch_gpu, skip_mps, slow, torch_device, ) from ..pipeline_params import TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS from ..test_pipelines_common import ( PipelineKarrasSchedulerTesterMixin, PipelineLatentTesterMixin, PipelineTesterMixin, assert_mean_pixel_difference, ) enable_full_determinism() class a_ ( a__ , a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = StableUnCLIPImgaImgPipeline __SCREAMING_SNAKE_CASE : List[str] = TEXT_GUIDED_IMAGE_VARIATION_PARAMS __SCREAMING_SNAKE_CASE : List[Any] = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS __SCREAMING_SNAKE_CASE : Any = frozenset( [] ) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess __SCREAMING_SNAKE_CASE : Tuple = frozenset([] ) def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[Any] = 32 SCREAMING_SNAKE_CASE : Tuple = embedder_hidden_size # image encoding components SCREAMING_SNAKE_CASE : int = CLIPImageProcessor(crop_size=32 , size=32 ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = CLIPVisionModelWithProjection( CLIPVisionConfig( hidden_size=_lowerCamelCase , projection_dim=_lowerCamelCase , num_hidden_layers=5 , num_attention_heads=4 , image_size=32 , intermediate_size=37 , patch_size=1 , ) ) # regular denoising components torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = StableUnCLIPImageNormalizer(embedding_dim=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = DDPMScheduler(beta_schedule='''squaredcos_cap_v2''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = CLIPTextModel( CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=_lowerCamelCase , projection_dim=32 , intermediate_size=37 , layer_norm_eps=1e-05 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , ) ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = UNetaDConditionModel( sample_size=32 , in_channels=4 , out_channels=4 , down_block_types=('''CrossAttnDownBlock2D''', '''DownBlock2D''') , up_block_types=('''UpBlock2D''', '''CrossAttnUpBlock2D''') , block_out_channels=(32, 64) , attention_head_dim=(2, 4) , class_embed_type='''projection''' , projection_class_embeddings_input_dim=embedder_projection_dim * 2 , cross_attention_dim=_lowerCamelCase , layers_per_block=1 , upcast_attention=_lowerCamelCase , use_linear_projection=_lowerCamelCase , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = DDIMScheduler( beta_schedule='''scaled_linear''' , beta_start=0.0_0_0_8_5 , beta_end=0.0_1_2 , prediction_type='''v_prediction''' , set_alpha_to_one=_lowerCamelCase , steps_offset=1 , ) torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Dict = AutoencoderKL() SCREAMING_SNAKE_CASE : Optional[Any] = { # image encoding components '''feature_extractor''': feature_extractor, '''image_encoder''': image_encoder.eval(), # image noising components '''image_normalizer''': image_normalizer.eval(), '''image_noising_scheduler''': image_noising_scheduler, # regular denoising components '''tokenizer''': tokenizer, '''text_encoder''': text_encoder.eval(), '''unet''': unet.eval(), '''scheduler''': scheduler, '''vae''': vae.eval(), } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 , _lowerCamelCase=True ) ->Optional[int]: if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, 3, 32, 32) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if pil_image: SCREAMING_SNAKE_CASE : Any = input_image * 0.5 + 0.5 SCREAMING_SNAKE_CASE : int = input_image.clamp(0 , 1 ) SCREAMING_SNAKE_CASE : Union[str, Any] = input_image.cpu().permute(0 , 2 , 3 , 1 ).float().numpy() SCREAMING_SNAKE_CASE : List[str] = DiffusionPipeline.numpy_to_pil(_lowerCamelCase )[0] return { "prompt": "An anime racoon running a marathon", "image": input_image, "generator": generator, "num_inference_steps": 2, "output_type": "np", } @skip_mps def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Any = '''cpu''' # ensure determinism for the device-dependent torch.Generator SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = StableUnCLIPImgaImgPipeline(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = sd_pipe.to(_lowerCamelCase ) sd_pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = self.get_dummy_inputs(_lowerCamelCase ) inputs.update({'''image_embeds''': None} ) SCREAMING_SNAKE_CASE : Optional[int] = sd_pipe(**_lowerCamelCase ).images SCREAMING_SNAKE_CASE : Dict = image[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) SCREAMING_SNAKE_CASE : Tuple = np.array([0.3_8_7_2, 0.7_2_2_4, 0.5_6_0_1, 0.4_7_4_1, 0.6_8_7_2, 0.5_8_1_4, 0.4_6_3_6, 0.3_8_6_7, 0.5_0_7_8] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : str = torch_device in ['''cpu''', '''mps'''] self._test_attention_slicing_forward_pass(test_max_difference=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : Tuple = torch_device in ['''cpu''', '''mps'''] self._test_inference_batch_single_identical(test_max_difference=_lowerCamelCase ) @unittest.skipIf( torch_device != '''cuda''' or not is_xformers_available() , reason='''XFormers attention is only available with CUDA and `xformers` installed''' , ) def __lowerCAmelCase ( self ) ->Optional[int]: self._test_xformers_attention_forwardGenerator_pass(test_max_difference=_lowerCamelCase ) @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->int: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : List[Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_l_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[Any] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-l-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Union[str, Any] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : List[str] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) SCREAMING_SNAKE_CASE : Optional[int] = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/stable_unclip_2_1_h_img2img_anime_turtle_fp16.npy''' ) SCREAMING_SNAKE_CASE : List[str] = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) # stable unclip will oom when integration tests are run on a V100, # so turn on memory savings pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = pipe(_lowerCamelCase , '''anime turle''' , generator=_lowerCamelCase , output_type='''np''' ) SCREAMING_SNAKE_CASE : List[str] = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) torch.cuda.empty_cache() torch.cuda.reset_max_memory_allocated() torch.cuda.reset_peak_memory_stats() SCREAMING_SNAKE_CASE : str = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Dict = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() SCREAMING_SNAKE_CASE : Dict = pipe( _lowerCamelCase , '''anime turtle''' , num_inference_steps=2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = torch.cuda.max_memory_allocated() # make sure that less than 7 GB is allocated assert mem_bytes < 7 * 10**9
313
1
from __future__ import annotations def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = 2 SCREAMING_SNAKE_CASE : Optional[Any] = [] while i * i <= n: if n % i: i += 1 else: n //= i factors.append(a__ ) if n > 1: factors.append(a__ ) return factors if __name__ == "__main__": import doctest doctest.testmod()
313
from abc import ABC, abstractmethod from typing import List, Optional class a_ ( a__ ): """simple docstring""" def __init__( self ) ->List[str]: # test for the above condition self.test() def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = False while not completed: if counter == 1: self.reset() SCREAMING_SNAKE_CASE : List[Any] = self.advance() if not self.does_advance(_lowerCamelCase ): raise Exception( '''Custom Constraint is not defined correctly. self.does_advance(self.advance()) must be true.''' ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.update(_lowerCamelCase ) counter += 1 if counter > 1_0000: raise Exception('''update() does not fulfill the constraint.''' ) if self.remaining() != 0: raise Exception('''Custom Constraint is not defined correctly.''' ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[int]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Union[str, Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Any: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->int: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`token_ids` has to be a non-empty list, but is {token_ids}.""" ) if any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ): raise ValueError(F"""Each list in `token_ids` has to be a list of positive integers, but is {token_ids}.""" ) SCREAMING_SNAKE_CASE : Optional[Any] = token_ids SCREAMING_SNAKE_CASE : Union[str, Any] = len(self.token_ids ) SCREAMING_SNAKE_CASE : Any = -1 # the index of the currently fulfilled step SCREAMING_SNAKE_CASE : Any = False def __lowerCAmelCase ( self ) ->List[Any]: if self.completed: return None return self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) if self.completed: return False return token_id == self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[str]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : str = False SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = False if self.does_advance(_lowerCamelCase ): self.fulfilled_idx += 1 SCREAMING_SNAKE_CASE : str = True if self.fulfilled_idx == (self.seqlen - 1): SCREAMING_SNAKE_CASE : Any = True SCREAMING_SNAKE_CASE : Union[str, Any] = completed else: # failed to make progress. SCREAMING_SNAKE_CASE : Dict = True self.reset() return stepped, completed, reset def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = 0 def __lowerCAmelCase ( self ) ->Any: return self.seqlen - (self.fulfilled_idx + 1) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Dict: SCREAMING_SNAKE_CASE : Any = PhrasalConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : Dict = self.seqlen SCREAMING_SNAKE_CASE : int = self.fulfilled_idx SCREAMING_SNAKE_CASE : Tuple = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=True ) ->Dict: SCREAMING_SNAKE_CASE : Any = max([len(_lowerCamelCase ) for one in nested_token_ids] ) SCREAMING_SNAKE_CASE : List[str] = {} for token_ids in nested_token_ids: SCREAMING_SNAKE_CASE : Optional[Any] = root for tidx, token_id in enumerate(_lowerCamelCase ): if token_id not in level: SCREAMING_SNAKE_CASE : Any = {} SCREAMING_SNAKE_CASE : Tuple = level[token_id] if no_subsets and self.has_subsets(_lowerCamelCase , _lowerCamelCase ): raise ValueError( '''Each list in `nested_token_ids` can\'t be a complete subset of another list, but is''' F""" {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = root def __lowerCAmelCase ( self , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = self.trie for current_token in current_seq: SCREAMING_SNAKE_CASE : int = start[current_token] SCREAMING_SNAKE_CASE : Optional[int] = list(start.keys() ) return next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.next_tokens(_lowerCamelCase ) return len(_lowerCamelCase ) == 0 def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = list(root.values() ) if len(_lowerCamelCase ) == 0: return 1 else: return sum([self.count_leaves(_lowerCamelCase ) for nn in next_nodes] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = self.count_leaves(_lowerCamelCase ) return len(_lowerCamelCase ) != leaf_count class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->str: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`nested_token_ids` has to be a non-empty list, but is {nested_token_ids}.""" ) if any(not isinstance(_lowerCamelCase , _lowerCamelCase ) for token_ids in nested_token_ids ): raise ValueError(F"""`nested_token_ids` has to be a list of lists, but is {nested_token_ids}.""" ) if any( any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ) for token_ids in nested_token_ids ): raise ValueError( F"""Each list in `nested_token_ids` has to be a list of positive integers, but is {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = DisjunctiveTrie(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = nested_token_ids SCREAMING_SNAKE_CASE : Optional[int] = self.trie.max_height SCREAMING_SNAKE_CASE : Union[str, Any] = [] SCREAMING_SNAKE_CASE : Optional[int] = False def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : str = self.trie.next_tokens(self.current_seq ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : List[str] = self.trie.next_tokens(self.current_seq ) return token_id in next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : int = False SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = False if self.does_advance(_lowerCamelCase ): self.current_seq.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = True else: SCREAMING_SNAKE_CASE : Dict = True self.reset() SCREAMING_SNAKE_CASE : Any = self.trie.reached_leaf(self.current_seq ) SCREAMING_SNAKE_CASE : List[Any] = completed return stepped, completed, reset def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = [] def __lowerCAmelCase ( self ) ->Optional[Any]: if self.completed: # since this can be completed without reaching max height return 0 else: return self.seqlen - len(self.current_seq ) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->List[str]: SCREAMING_SNAKE_CASE : str = DisjunctiveConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : str = self.seqlen SCREAMING_SNAKE_CASE : int = self.current_seq SCREAMING_SNAKE_CASE : Optional[int] = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[Any] = constraints # max # of steps required to fulfill a given constraint SCREAMING_SNAKE_CASE : str = max([c.seqlen for c in constraints] ) SCREAMING_SNAKE_CASE : List[str] = len(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = False self.init_state() def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Any = [] SCREAMING_SNAKE_CASE : List[Any] = None SCREAMING_SNAKE_CASE : Tuple = [constraint.copy(stateful=_lowerCamelCase ) for constraint in self.constraints] def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = 0 if self.inprogress_constraint: # extra points for having a constraint mid-fulfilled add += self.max_seqlen - self.inprogress_constraint.remaining() return (len(self.complete_constraints ) * self.max_seqlen) + add def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Tuple = [] if self.inprogress_constraint is None: for constraint in self.pending_constraints: # "pending" == "unfulfilled yet" SCREAMING_SNAKE_CASE : Optional[int] = constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = self.inprogress_constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: self.init_state() if token_ids is not None: for token in token_ids: # completes or steps **one** constraint SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.add(_lowerCamelCase ) # the entire list of constraints are fulfilled if self.completed: break def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` should be an `int`, but is `{token_id}`.""" ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = False, False if self.completed: SCREAMING_SNAKE_CASE : List[str] = True SCREAMING_SNAKE_CASE : Optional[int] = False return complete, stepped if self.inprogress_constraint is not None: # In the middle of fulfilling a constraint. If the `token_id` *does* makes an incremental progress to current # job, simply update the state SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.inprogress_constraint.update(_lowerCamelCase ) if reset: # 1. If the next token breaks the progress, then we must restart. # e.g. constraint = "I love pies" and sequence so far is "I love" but `token_id` == "books". # But that doesn't mean we self.init_state(), since we only reset the state for this particular # constraint, not the full list of constraints. self.pending_constraints.append(self.inprogress_constraint.copy(stateful=_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = None if complete: # 2. If the next token completes the constraint, move it to completed list, set # inprogress to None. If there are no pending constraints either, then this full list of constraints # is complete. self.complete_constraints.append(self.inprogress_constraint ) SCREAMING_SNAKE_CASE : str = None if len(self.pending_constraints ) == 0: # we're done! SCREAMING_SNAKE_CASE : Optional[Any] = True else: # Not in the middle of fulfilling a constraint. So does this `token_id` helps us step towards any of our list # of constraints? for cidx, pending_constraint in enumerate(self.pending_constraints ): if pending_constraint.does_advance(_lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = pending_constraint.update(_lowerCamelCase ) if not stepped: raise Exception( '''`constraint.update(token_id)` is not yielding incremental progress, ''' '''even though `constraint.does_advance(token_id)` is true.''' ) if complete: self.complete_constraints.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = None if not complete and stepped: SCREAMING_SNAKE_CASE : Optional[Any] = pending_constraint if complete or stepped: # If we made any progress at all, then it's at least not a "pending constraint". SCREAMING_SNAKE_CASE : Union[str, Any] = ( self.pending_constraints[:cidx] + self.pending_constraints[cidx + 1 :] ) if len(self.pending_constraints ) == 0 and self.inprogress_constraint is None: # If there's no longer any pending after this and no inprogress either, then we must be # complete. SCREAMING_SNAKE_CASE : str = True break # prevent accidentally stepping through multiple constraints with just one token. return complete, stepped def __lowerCAmelCase ( self , _lowerCamelCase=True ) ->str: SCREAMING_SNAKE_CASE : Dict = ConstraintListState(self.constraints ) # we actually never though self.constraints objects # throughout this process. So it's at initialization state. if stateful: SCREAMING_SNAKE_CASE : str = [ constraint.copy(stateful=_lowerCamelCase ) for constraint in self.complete_constraints ] if self.inprogress_constraint is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.inprogress_constraint.copy(stateful=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = [constraint.copy() for constraint in self.pending_constraints] return new_state
313
1
import unittest import numpy as np import torch from transformers import CLIPTextConfig, CLIPTextModel from diffusers import DDIMScheduler, LDMPipeline, UNetaDModel, VQModel from diffusers.utils.testing_utils import enable_full_determinism, require_torch, slow, torch_device enable_full_determinism() class a_ ( unittest.TestCase ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Optional[int]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = UNetaDModel( block_out_channels=(32, 64) , layers_per_block=2 , sample_size=32 , in_channels=3 , out_channels=3 , down_block_types=('''DownBlock2D''', '''AttnDownBlock2D''') , up_block_types=('''AttnUpBlock2D''', '''UpBlock2D''') , ) return model @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = VQModel( block_out_channels=[32, 64] , in_channels=3 , out_channels=3 , down_block_types=['''DownEncoderBlock2D''', '''DownEncoderBlock2D'''] , up_block_types=['''UpDecoderBlock2D''', '''UpDecoderBlock2D'''] , latent_channels=3 , ) return model @property def __lowerCAmelCase ( self ) ->Dict: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = 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 , ) return CLIPTextModel(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : List[str] = self.dummy_uncond_unet SCREAMING_SNAKE_CASE : Union[str, Any] = DDIMScheduler() SCREAMING_SNAKE_CASE : str = self.dummy_vq_model SCREAMING_SNAKE_CASE : Optional[int] = LDMPipeline(unet=_lowerCamelCase , vqvae=_lowerCamelCase , scheduler=_lowerCamelCase ) ldm.to(_lowerCamelCase ) ldm.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = ldm(generator=_lowerCamelCase , num_inference_steps=2 , output_type='''numpy''' ).images SCREAMING_SNAKE_CASE : Union[str, Any] = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Dict = ldm(generator=_lowerCamelCase , num_inference_steps=2 , output_type='''numpy''' , return_dict=_lowerCamelCase )[0] SCREAMING_SNAKE_CASE : Union[str, Any] = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : Any = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array([0.8_5_1_2, 0.8_1_8, 0.6_4_1_1, 0.6_8_0_8, 0.4_4_6_5, 0.5_6_1_8, 0.4_6, 0.6_2_3_1, 0.5_1_7_2] ) SCREAMING_SNAKE_CASE : Optional[int] = 1e-2 if torch_device != '''mps''' else 3e-2 assert np.abs(image_slice.flatten() - expected_slice ).max() < tolerance assert np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < tolerance @slow @require_torch class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Union[str, Any] = LDMPipeline.from_pretrained('''CompVis/ldm-celebahq-256''' ) ldm.to(_lowerCamelCase ) ldm.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Tuple = ldm(generator=_lowerCamelCase , num_inference_steps=5 , output_type='''numpy''' ).images SCREAMING_SNAKE_CASE : Optional[Any] = image[0, -3:, -3:, -1] assert image.shape == (1, 256, 256, 3) SCREAMING_SNAKE_CASE : List[str] = np.array([0.4_3_9_9, 0.4_4_9_7_5, 0.4_6_8_2_5, 0.4_7_4, 0.4_3_5_9, 0.4_5_8_1, 0.4_5_0_9_5, 0.4_3_4_1, 0.4_4_4_7] ) SCREAMING_SNAKE_CASE : Optional[int] = 1e-2 if torch_device != '''mps''' else 3e-2 assert np.abs(image_slice.flatten() - expected_slice ).max() < tolerance
313
import argparse import random import joblib import numpy as np import torch from igf.igf import ( SecondaryLearner, collect_objective_set, compute_perplexity, generate_datasets, load_gpta, recopy_gpta, set_seed, train_secondary_learner, ) from torch.utils.data import DataLoader, RandomSampler from transformers import GPTaLMHeadModel def UpperCAmelCase_( a__=32 , a__=10 , a__=100 , a__=1_026 , a__=True , a__="data/tokenized_stories_train_wikitext103.jbl" , a__="igf_context_pairs.jbl" , ): """simple docstring""" set_seed(3 ) # generate train_data and objective_set SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = generate_datasets( a__ , a__ , number=a__ , min_len=1_026 , trim=a__ ) # keeps model same across runs set_seed(4 ) # model, lm_optimizer, lm_scheduler = recopy_gpt2(model, device, max_steps) # store original model weights # can we train on GPU? SCREAMING_SNAKE_CASE : str = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) # load pretrained model SCREAMING_SNAKE_CASE : Dict = load_gpta('''gpt2''' ).to(a__ ) print('''computing perplexity on objective set''' ) SCREAMING_SNAKE_CASE : int = compute_perplexity(a__ , a__ , a__ ).item() print('''perplexity on objective set:''' , a__ ) # collect igf pairs and save to file demo.jbl collect_objective_set(a__ , a__ , a__ , a__ , a__ , a__ , a__ , a__ ) # clean up, delete model and data we don't need anymore del model, train_data, objective_set torch.cuda.empty_cache() def UpperCAmelCase_( a__ , a__=15 , a__=128 , a__=100 , a__="igf_model.pt" , ): """simple docstring""" set_seed(42 ) # Load pre-trained model SCREAMING_SNAKE_CASE : List[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) # Initialize secondary learner to use embedding weights of model SCREAMING_SNAKE_CASE : str = SecondaryLearner(a__ ) # Train secondary learner SCREAMING_SNAKE_CASE : Union[str, Any] = train_secondary_learner( a__ , a__ , max_epochs=a__ , batch_size=a__ , eval_freq=100 , igf_model_path=a__ , ) del model, secondary_learner_train_data torch.cuda.empty_cache() return secondary_learner def UpperCAmelCase_( a__ , a__ , a__ , a__=32 , a__=1_000 , a__=16 , a__=1.0 , a__=recopy_gpta , a__=None , a__=10 , a__="gpt2_finetuned.pt" , ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = torch.device('''cuda:0''' if torch.cuda.is_available() else '''cpu''' ) SCREAMING_SNAKE_CASE : Optional[int] = RandomSampler(a__ ) SCREAMING_SNAKE_CASE : Dict = DataLoader(a__ , sampler=a__ ) SCREAMING_SNAKE_CASE : Tuple = max_steps // (len(a__ )) + 1 SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = torch.zeros((1, context_len) , dtype=torch.long , device=a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : str = recopy_model(a__ , a__ , a__ ) model.train() if secondary_learner is not None: secondary_learner.to(a__ ) secondary_learner.eval() SCREAMING_SNAKE_CASE : List[str] = [] SCREAMING_SNAKE_CASE : Optional[int] = 0 SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Tuple = [] # Compute the performance of the transformer model at the beginning SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) for epoch in range(int(a__ ) ): for step, example in enumerate(a__ ): torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Union[str, Any] = random.randint(0 , example.size(2 ) - context_len - 1 ) SCREAMING_SNAKE_CASE : Optional[int] = example[0, 0, start : start + context_len] lm_optimizer.zero_grad() SCREAMING_SNAKE_CASE : Optional[Any] = model(a__ , labels=a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = True if secondary_learner is not None: SCREAMING_SNAKE_CASE : List[str] = secondary_learner.forward( torch.tensor(a__ , dtype=torch.long , device=a__ ).unsqueeze(0 ) )[0].item() observed_qs.append(float(a__ ) ) # Here we implement the simple non-constant threshold for the predicted IG(X) value # We will decay the selectivity of our secondary learner filter from # 1 standard deviation above average to 1 below average after 10 batches. if global_step == 10: SCREAMING_SNAKE_CASE : Dict = -1 if predicted_q < threshold: SCREAMING_SNAKE_CASE : str = False # If we passed the filter, add the context to the batch! if do_backprop: contexts.append(np.array(context.cpu() ) ) SCREAMING_SNAKE_CASE : List[str] = outputs[0] lm_loss.backward() examples += 1 del outputs # Once the batch is filled with enough contexts, backprop on the batch. if examples == batch_size: torch.cuda.empty_cache() SCREAMING_SNAKE_CASE : Any = 0 # Do LM backprop torch.nn.utils.clip_grad_norm_(model.parameters() , 3.0 ) lm_optimizer.step() lm_scheduler.step() # Update learning rate schedule global_step += 1 # Compute the performance of the transformer model at this batch if global_step % eval_interval == 0: SCREAMING_SNAKE_CASE : str = compute_perplexity(a__ , a__ , a__ ) test_perps.append(a__ ) print('''Test perplexity, step''' , a__ , ''':''' , a__ ) # Break out of the loop after 60 batches if max_steps > 0 and global_step > 60: break if max_steps > 0 and global_step > 60: break # save finetuned transformer model torch.save(model.state_dict() , a__ ) torch.cuda.empty_cache() # Do some cleaning up so we can reinitialize for the next run of this function del lm_optimizer del lm_scheduler return model def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = argparse.ArgumentParser(description='''Fine-tune a transformer model with IGF on a language modeling task''' ) # Required parameters parser.add_argument( '''--data_dir''' , default=a__ , type=a__ , required=a__ , help='''The input data dir. Should contain data files for WikiText.''' , ) parser.add_argument( '''--model_name_or_path''' , default=a__ , type=a__ , required=a__ , help='''Path to pretrained model or model identifier from huggingface.co/models''' , ) parser.add_argument( '''--data_file''' , type=a__ , default=a__ , help=( '''A jbl file containing tokenized data which can be split as objective dataset, ''' '''train_dataset and test_dataset.''' ) , ) parser.add_argument( '''--igf_data_file''' , type=a__ , default=a__ , help='''A jbl file containing the context and information gain pairs to train secondary learner.''' , ) parser.add_argument( '''--output_dir''' , default=a__ , type=a__ , required=a__ , help='''The output directory where the final fine-tuned model is stored.''' , ) parser.add_argument( '''--tokenizer_name''' , default=a__ , type=a__ , help='''Pretrained tokenizer name or path if not the same as model_name''' , ) parser.add_argument('''--seed''' , type=a__ , default=a__ , help='''A seed for reproducible training.''' ) parser.add_argument( '''--context_len''' , default=32 , type=a__ , help=( '''The maximum total input sequence length after tokenization. Sequences longer ''' '''than this will be truncated, sequences shorter will be padded.''' ) , ) parser.add_argument( '''--size_objective_set''' , default=100 , type=a__ , help='''number of articles that are long enough to be used as our objective set''' , ) parser.add_argument( '''--eval_freq''' , default=100 , type=a__ , help='''secondary model evaluation is triggered at eval_freq''' ) parser.add_argument('''--max_steps''' , default=1_000 , type=a__ , help='''To calculate training epochs''' ) parser.add_argument( '''--secondary_learner_batch_size''' , default=128 , type=a__ , help='''batch size of training data for secondary learner''' , ) parser.add_argument( '''--batch_size''' , default=16 , type=a__ , help='''batch size of training data of language model(gpt2) ''' ) parser.add_argument( '''--eval_interval''' , default=10 , type=a__ , help=( '''decay the selectivity of our secondary learner filter from''' '''1 standard deviation above average to 1 below average after 10 batches''' ) , ) parser.add_argument( '''--number''' , default=100 , type=a__ , help='''The number of examples split to be used as objective_set/test_data''' ) parser.add_argument( '''--min_len''' , default=1_026 , type=a__ , help='''The minimum length of the article to be used as objective set''' ) parser.add_argument( '''--secondary_learner_max_epochs''' , default=15 , type=a__ , help='''number of epochs to train secondary learner''' ) parser.add_argument('''--trim''' , default=a__ , type=a__ , help='''truncate the example if it exceeds context length''' ) parser.add_argument( '''--threshold''' , default=1.0 , type=a__ , help=( '''The threshold value used by secondary learner to filter the train_data and allow only''' ''' informative data as input to the model''' ) , ) parser.add_argument('''--finetuned_model_name''' , default='''gpt2_finetuned.pt''' , type=a__ , help='''finetuned_model_name''' ) parser.add_argument( '''--recopy_model''' , default=a__ , type=a__ , help='''Reset the model to the original pretrained GPT-2 weights after each iteration''' , ) # function calls # Collecting *n* pairs of context and information gain(X, IG(X)) for training the secondary learner generate_n_pairs( context_len=32 , max_steps=10 , size_objective_set=100 , min_len=1_026 , trim=a__ , data_file='''data/tokenized_stories_train_wikitext103.jbl''' , igf_data_file='''igf_context_pairs.jbl''' , ) # Load train data for secondary learner SCREAMING_SNAKE_CASE : List[Any] = joblib.load('''data/IGF_values.jbl''' ) # Train secondary learner SCREAMING_SNAKE_CASE : Tuple = training_secondary_learner( a__ , secondary_learner_max_epochs=15 , secondary_learner_batch_size=128 , eval_freq=100 , igf_model_path='''igf_model.pt''' , ) # load pretrained gpt2 model SCREAMING_SNAKE_CASE : Optional[Any] = GPTaLMHeadModel.from_pretrained('''gpt2''' ) set_seed(42 ) # Generate train and test data to train and evaluate gpt2 model SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = generate_datasets( context_len=32 , file='''data/tokenized_stories_train_wikitext103.jbl''' , number=100 , min_len=1_026 , trim=a__ ) # fine-tuning of the gpt2 model using igf (Information Gain Filtration) finetune( a__ , a__ , a__ , context_len=32 , max_steps=1_000 , batch_size=16 , threshold=1.0 , recopy_model=a__ , secondary_learner=a__ , eval_interval=10 , finetuned_model_name='''gpt2_finetuned.pt''' , ) if __name__ == "__main__": main()
313
1
import random from typing import Any def UpperCAmelCase_( a__ ): """simple docstring""" for _ in range(len(a__ ) ): SCREAMING_SNAKE_CASE : str = random.randint(0 , len(a__ ) - 1 ) SCREAMING_SNAKE_CASE : int = random.randint(0 , len(a__ ) - 1 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = data[b], data[a] return data if __name__ == "__main__": a__ : List[Any] = [0, 1, 2, 3, 4, 5, 6, 7] a__ : Tuple = ['''python''', '''says''', '''hello''', '''!'''] print('''Fisher-Yates Shuffle:''') print('''List''', integers, strings) print('''FY Shuffle''', fisher_yates_shuffle(integers), fisher_yates_shuffle(strings))
313
import logging from pathlib import Path import numpy as np import pytorch_lightning as pl import torch from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint from pytorch_lightning.utilities import rank_zero_only from utils_rag import save_json def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = filter(lambda a__ : p.requires_grad , model.parameters() ) SCREAMING_SNAKE_CASE : List[Any] = sum([np.prod(p.size() ) for p in model_parameters] ) return params a__ : Any = logging.getLogger(__name__) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if metric == "rouge2": SCREAMING_SNAKE_CASE : str = '''{val_avg_rouge2:.4f}-{step_count}''' elif metric == "bleu": SCREAMING_SNAKE_CASE : List[Any] = '''{val_avg_bleu:.4f}-{step_count}''' elif metric == "em": SCREAMING_SNAKE_CASE : int = '''{val_avg_em:.4f}-{step_count}''' elif metric == "loss": SCREAMING_SNAKE_CASE : int = '''{val_avg_loss:.4f}-{step_count}''' else: raise NotImplementedError( F"""seq2seq callbacks only support rouge2 and bleu, got {metric}, You can make your own by adding to this""" ''' function.''' ) SCREAMING_SNAKE_CASE : Dict = ModelCheckpoint( dirpath=a__ , filename=a__ , monitor=F"""val_{metric}""" , mode='''max''' , save_top_k=1 , every_n_epochs=1 , ) return checkpoint_callback def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return EarlyStopping( monitor=F"""val_{metric}""" , mode='''min''' if '''loss''' in metric else '''max''' , patience=a__ , verbose=a__ , ) class a_ ( pl.Callback ): """simple docstring""" def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = {F"""lr_group_{i}""": param['''lr'''] for i, param in enumerate(pl_module.trainer.optimizers[0].param_groups )} pl_module.logger.log_metrics(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=True ) ->None: logger.info(F"""***** {type_path} results at step {trainer.global_step:05d} *****""" ) SCREAMING_SNAKE_CASE : Optional[int] = trainer.callback_metrics trainer.logger.log_metrics({k: v for k, v in metrics.items() if k not in ['''log''', '''progress_bar''', '''preds''']} ) # Log results SCREAMING_SNAKE_CASE : List[str] = Path(pl_module.hparams.output_dir ) if type_path == "test": SCREAMING_SNAKE_CASE : Any = od / '''test_results.txt''' SCREAMING_SNAKE_CASE : Optional[int] = od / '''test_generations.txt''' else: # this never gets hit. I prefer not to save intermediate generations, and results are in metrics.json # If people want this it will be easy enough to add back. SCREAMING_SNAKE_CASE : str = od / F"""{type_path}_results/{trainer.global_step:05d}.txt""" SCREAMING_SNAKE_CASE : Tuple = od / F"""{type_path}_generations/{trainer.global_step:05d}.txt""" results_file.parent.mkdir(exist_ok=_lowerCamelCase ) generations_file.parent.mkdir(exist_ok=_lowerCamelCase ) with open(_lowerCamelCase , '''a+''' ) as writer: for key in sorted(_lowerCamelCase ): if key in ["log", "progress_bar", "preds"]: continue SCREAMING_SNAKE_CASE : Tuple = metrics[key] if isinstance(_lowerCamelCase , torch.Tensor ): SCREAMING_SNAKE_CASE : List[Any] = val.item() SCREAMING_SNAKE_CASE : Tuple = F"""{key}: {val:.6f}\n""" writer.write(_lowerCamelCase ) if not save_generations: return if "preds" in metrics: SCREAMING_SNAKE_CASE : Optional[Any] = '''\n'''.join(metrics['''preds'''] ) generations_file.open('''w+''' ).write(_lowerCamelCase ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: try: SCREAMING_SNAKE_CASE : Any = pl_module.model.model.num_parameters() except AttributeError: SCREAMING_SNAKE_CASE : Optional[int] = pl_module.model.num_parameters() SCREAMING_SNAKE_CASE : int = count_trainable_parameters(_lowerCamelCase ) # mp stands for million parameters trainer.logger.log_metrics({'''n_params''': npars, '''mp''': npars / 1e6, '''grad_mp''': n_trainable_pars / 1e6} ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->List[Any]: save_json(pl_module.metrics , pl_module.metrics_save_path ) return self._write_logs(_lowerCamelCase , _lowerCamelCase , '''test''' ) @rank_zero_only def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Tuple: save_json(pl_module.metrics , pl_module.metrics_save_path ) # Uncommenting this will save val generations # return self._write_logs(trainer, pl_module, "valid")
313
1
import os import torch from ..logging import get_logger from .constants import FSDP_PYTORCH_VERSION, MODEL_NAME, OPTIMIZER_NAME from .versions import is_torch_version if is_torch_version('''>=''', FSDP_PYTORCH_VERSION): import torch.distributed.checkpoint as dist_cp from torch.distributed.checkpoint.default_planner import DefaultLoadPlanner, DefaultSavePlanner from torch.distributed.checkpoint.optimizer import load_sharded_optimizer_state_dict from torch.distributed.fsdp.fully_sharded_data_parallel import FullyShardedDataParallel as FSDP from torch.distributed.fsdp.fully_sharded_data_parallel import StateDictType a__ : Union[str, Any] = get_logger(__name__) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__=0 ): """simple docstring""" os.makedirs(a__ , exist_ok=a__ ) with FSDP.state_dict_type( a__ , fsdp_plugin.state_dict_type , fsdp_plugin.state_dict_config , fsdp_plugin.optim_state_dict_config ): SCREAMING_SNAKE_CASE : Union[str, Any] = model.state_dict() if fsdp_plugin.state_dict_type == StateDictType.FULL_STATE_DICT: SCREAMING_SNAKE_CASE : Dict = F"""{MODEL_NAME}.bin""" if model_index == 0 else F"""{MODEL_NAME}_{model_index}.bin""" SCREAMING_SNAKE_CASE : Optional[int] = os.path.join(a__ , a__ ) if accelerator.process_index == 0: logger.info(F"""Saving model to {output_model_file}""" ) torch.save(a__ , a__ ) logger.info(F"""Model saved to {output_model_file}""" ) elif fsdp_plugin.state_dict_type == StateDictType.LOCAL_STATE_DICT: SCREAMING_SNAKE_CASE : Union[str, Any] = ( F"""{MODEL_NAME}_rank{accelerator.process_index}.bin""" if model_index == 0 else F"""{MODEL_NAME}_{model_index}_rank{accelerator.process_index}.bin""" ) SCREAMING_SNAKE_CASE : Tuple = os.path.join(a__ , a__ ) logger.info(F"""Saving model to {output_model_file}""" ) torch.save(a__ , a__ ) logger.info(F"""Model saved to {output_model_file}""" ) elif fsdp_plugin.state_dict_type == StateDictType.SHARDED_STATE_DICT: SCREAMING_SNAKE_CASE : Any = os.path.join(a__ , F"""{MODEL_NAME}_{model_index}""" ) os.makedirs(a__ , exist_ok=a__ ) logger.info(F"""Saving model to {ckpt_dir}""" ) SCREAMING_SNAKE_CASE : List[Any] = {'''model''': state_dict} dist_cp.save_state_dict( state_dict=a__ , storage_writer=dist_cp.FileSystemWriter(a__ ) , planner=DefaultSavePlanner() , ) logger.info(F"""Model saved to {ckpt_dir}""" ) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__=0 ): """simple docstring""" accelerator.wait_for_everyone() with FSDP.state_dict_type( a__ , fsdp_plugin.state_dict_type , fsdp_plugin.state_dict_config , fsdp_plugin.optim_state_dict_config ): if fsdp_plugin.state_dict_type == StateDictType.FULL_STATE_DICT: if type(a__ ) != FSDP and accelerator.process_index != 0: if not fsdp_plugin.sync_module_states: raise ValueError( '''Set the `sync_module_states` flag to `True` so that model states are synced across processes when ''' '''initializing FSDP object''' ) return SCREAMING_SNAKE_CASE : Optional[Any] = F"""{MODEL_NAME}.bin""" if model_index == 0 else F"""{MODEL_NAME}_{model_index}.bin""" SCREAMING_SNAKE_CASE : int = os.path.join(a__ , a__ ) logger.info(F"""Loading model from {input_model_file}""" ) SCREAMING_SNAKE_CASE : List[str] = torch.load(a__ ) logger.info(F"""Model loaded from {input_model_file}""" ) elif fsdp_plugin.state_dict_type == StateDictType.LOCAL_STATE_DICT: SCREAMING_SNAKE_CASE : Optional[Any] = ( F"""{MODEL_NAME}_rank{accelerator.process_index}.bin""" if model_index == 0 else F"""{MODEL_NAME}_{model_index}_rank{accelerator.process_index}.bin""" ) SCREAMING_SNAKE_CASE : List[str] = os.path.join(a__ , a__ ) logger.info(F"""Loading model from {input_model_file}""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch.load(a__ ) logger.info(F"""Model loaded from {input_model_file}""" ) elif fsdp_plugin.state_dict_type == StateDictType.SHARDED_STATE_DICT: SCREAMING_SNAKE_CASE : Union[str, Any] = ( os.path.join(a__ , F"""{MODEL_NAME}_{model_index}""" ) if F"""{MODEL_NAME}""" not in input_dir else input_dir ) logger.info(F"""Loading model from {ckpt_dir}""" ) SCREAMING_SNAKE_CASE : str = {'''model''': model.state_dict()} dist_cp.load_state_dict( state_dict=a__ , storage_reader=dist_cp.FileSystemReader(a__ ) , planner=DefaultLoadPlanner() , ) SCREAMING_SNAKE_CASE : Tuple = state_dict['''model'''] logger.info(F"""Model loaded from {ckpt_dir}""" ) model.load_state_dict(a__ ) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ , a__=0 ): """simple docstring""" os.makedirs(a__ , exist_ok=a__ ) with FSDP.state_dict_type( a__ , fsdp_plugin.state_dict_type , fsdp_plugin.state_dict_config , fsdp_plugin.optim_state_dict_config ): SCREAMING_SNAKE_CASE : Union[str, Any] = FSDP.optim_state_dict(a__ , a__ ) if fsdp_plugin.state_dict_type == StateDictType.FULL_STATE_DICT: if accelerator.process_index == 0: SCREAMING_SNAKE_CASE : Union[str, Any] = ( F"""{OPTIMIZER_NAME}.bin""" if optimizer_index == 0 else F"""{OPTIMIZER_NAME}_{optimizer_index}.bin""" ) SCREAMING_SNAKE_CASE : List[Any] = os.path.join(a__ , a__ ) logger.info(F"""Saving Optimizer state to {output_optimizer_file}""" ) torch.save(a__ , a__ ) logger.info(F"""Optimizer state saved in {output_optimizer_file}""" ) else: SCREAMING_SNAKE_CASE : List[Any] = os.path.join(a__ , F"""{OPTIMIZER_NAME}_{optimizer_index}""" ) os.makedirs(a__ , exist_ok=a__ ) logger.info(F"""Saving Optimizer state to {ckpt_dir}""" ) dist_cp.save_state_dict( state_dict={'''optimizer''': optim_state} , storage_writer=dist_cp.FileSystemWriter(a__ ) , planner=DefaultSavePlanner() , ) logger.info(F"""Optimizer state saved in {ckpt_dir}""" ) def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ , a__=0 ): """simple docstring""" accelerator.wait_for_everyone() with FSDP.state_dict_type( a__ , fsdp_plugin.state_dict_type , fsdp_plugin.state_dict_config , fsdp_plugin.optim_state_dict_config ): if fsdp_plugin.state_dict_type == StateDictType.FULL_STATE_DICT: SCREAMING_SNAKE_CASE : Union[str, Any] = None # below check should work but currently it isn't working (mostly opytorch issue), # in the meantime disabling it at the cost of excess memory usage # if accelerator.process_index == 0 or not fsdp_plugin.optim_state_dict_config.rank0_only: SCREAMING_SNAKE_CASE : List[Any] = ( F"""{OPTIMIZER_NAME}.bin""" if optimizer_index == 0 else F"""{OPTIMIZER_NAME}_{optimizer_index}.bin""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = os.path.join(a__ , a__ ) logger.info(F"""Loading Optimizer state from {input_optimizer_file}""" ) SCREAMING_SNAKE_CASE : Tuple = torch.load(a__ ) logger.info(F"""Optimizer state loaded from {input_optimizer_file}""" ) else: SCREAMING_SNAKE_CASE : Optional[int] = ( os.path.join(a__ , F"""{OPTIMIZER_NAME}_{optimizer_index}""" ) if F"""{OPTIMIZER_NAME}""" not in input_dir else input_dir ) logger.info(F"""Loading Optimizer from {ckpt_dir}""" ) SCREAMING_SNAKE_CASE : Optional[Any] = load_sharded_optimizer_state_dict( model_state_dict=model.state_dict() , optimizer_key='''optimizer''' , storage_reader=dist_cp.FileSystemReader(a__ ) , ) SCREAMING_SNAKE_CASE : Any = optim_state['''optimizer'''] logger.info(F"""Optimizer loaded from {ckpt_dir}""" ) SCREAMING_SNAKE_CASE : List[str] = FSDP.optim_state_dict_to_load(a__ , a__ , a__ ) optimizer.load_state_dict(a__ )
313
import argparse import json from typing import List from ltp import LTP from transformers import BertTokenizer def UpperCAmelCase_( a__ ): """simple docstring""" if ( (cp >= 0x4_E00 and cp <= 0x9_FFF) or (cp >= 0x3_400 and cp <= 0x4_DBF) # or (cp >= 0x20_000 and cp <= 0x2A_6DF) # or (cp >= 0x2A_700 and cp <= 0x2B_73F) # or (cp >= 0x2B_740 and cp <= 0x2B_81F) # or (cp >= 0x2B_820 and cp <= 0x2C_EAF) # or (cp >= 0xF_900 and cp <= 0xF_AFF) or (cp >= 0x2F_800 and cp <= 0x2F_A1F) # ): # return True return False def UpperCAmelCase_( a__ ): """simple docstring""" for char in word: SCREAMING_SNAKE_CASE : str = ord(a__ ) if not _is_chinese_char(a__ ): return 0 return 1 def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = set() for token in tokens: SCREAMING_SNAKE_CASE : str = len(a__ ) > 1 and is_chinese(a__ ) if chinese_word: word_set.add(a__ ) SCREAMING_SNAKE_CASE : str = list(a__ ) return word_list def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if not chinese_word_set: return bert_tokens SCREAMING_SNAKE_CASE : List[str] = max([len(a__ ) for w in chinese_word_set] ) SCREAMING_SNAKE_CASE : Tuple = bert_tokens SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Any = 0, len(a__ ) while start < end: SCREAMING_SNAKE_CASE : Dict = True if is_chinese(bert_word[start] ): SCREAMING_SNAKE_CASE : Optional[int] = min(end - start , a__ ) for i in range(a__ , 1 , -1 ): SCREAMING_SNAKE_CASE : Optional[int] = ''''''.join(bert_word[start : start + i] ) if whole_word in chinese_word_set: for j in range(start + 1 , start + i ): SCREAMING_SNAKE_CASE : Optional[int] = '''##''' + bert_word[j] SCREAMING_SNAKE_CASE : List[str] = start + i SCREAMING_SNAKE_CASE : Optional[Any] = False break if single_word: start += 1 return bert_word def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : Optional[Any] = ltp_tokenizer.seg(lines[i : i + 100] )[0] SCREAMING_SNAKE_CASE : Union[str, Any] = [get_chinese_word(a__ ) for r in res] ltp_res.extend(a__ ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : Any = [] for i in range(0 , len(a__ ) , 100 ): SCREAMING_SNAKE_CASE : int = bert_tokenizer(lines[i : i + 100] , add_special_tokens=a__ , truncation=a__ , max_length=512 ) bert_res.extend(res['''input_ids'''] ) assert len(a__ ) == len(a__ ) SCREAMING_SNAKE_CASE : int = [] for input_ids, chinese_word in zip(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = [] for id in input_ids: SCREAMING_SNAKE_CASE : List[Any] = bert_tokenizer._convert_id_to_token(a__ ) input_tokens.append(a__ ) SCREAMING_SNAKE_CASE : List[str] = add_sub_symbol(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = [] # We only save pos of chinese subwords start with ##, which mean is part of a whole word. for i, token in enumerate(a__ ): if token[:2] == "##": SCREAMING_SNAKE_CASE : Optional[int] = token[2:] # save chinese tokens' pos if len(a__ ) == 1 and _is_chinese_char(ord(a__ ) ): ref_id.append(a__ ) ref_ids.append(a__ ) assert len(a__ ) == len(a__ ) return ref_ids def UpperCAmelCase_( a__ ): """simple docstring""" with open(args.file_name , '''r''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : List[str] = f.readlines() SCREAMING_SNAKE_CASE : Union[str, Any] = [line.strip() for line in data if len(a__ ) > 0 and not line.isspace()] # avoid delimiter like '\u2029' SCREAMING_SNAKE_CASE : List[str] = LTP(args.ltp ) # faster in GPU device SCREAMING_SNAKE_CASE : int = BertTokenizer.from_pretrained(args.bert ) SCREAMING_SNAKE_CASE : int = prepare_ref(a__ , a__ , a__ ) with open(args.save_path , '''w''' , encoding='''utf-8''' ) as f: SCREAMING_SNAKE_CASE : Tuple = [json.dumps(a__ ) + '''\n''' for ref in ref_ids] f.writelines(a__ ) if __name__ == "__main__": a__ : int = argparse.ArgumentParser(description='''prepare_chinese_ref''') parser.add_argument( '''--file_name''', type=str, default='''./resources/chinese-demo.txt''', help='''file need process, same as training data in lm''', ) parser.add_argument( '''--ltp''', type=str, default='''./resources/ltp''', help='''resources for LTP tokenizer, usually a path''' ) parser.add_argument('''--bert''', type=str, default='''./resources/robert''', help='''resources for Bert tokenizer''') parser.add_argument('''--save_path''', type=str, default='''./resources/ref.txt''', help='''path to save res''') a__ : int = parser.parse_args() main(args)
313
1
def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = abs(a__ ) SCREAMING_SNAKE_CASE : int = 0 while n > 0: res += n % 10 n //= 10 return res def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = abs(a__ ) return n if n < 10 else n % 10 + sum_of_digits(n // 10 ) def UpperCAmelCase_( a__ ): """simple docstring""" return sum(int(a__ ) for c in str(abs(a__ ) ) ) def UpperCAmelCase_( ): """simple docstring""" from collections.abc import Callable from timeit import timeit def benchmark_a_function(a__ , a__ ) -> None: SCREAMING_SNAKE_CASE : Dict = F"""{func.__name__}({value})""" SCREAMING_SNAKE_CASE : str = timeit(F"""__main__.{call}""" , setup='''import __main__''' ) print(F"""{call:56} = {func(a__ )} -- {timing:.4f} seconds""" ) for value in (262_144, 1_125_899_906_842_624, 1_267_650_600_228_229_401_496_703_205_376): for func in (sum_of_digits, sum_of_digits_recursion, sum_of_digits_compact): benchmark_a_function(a__ , a__ ) print() if __name__ == "__main__": import doctest doctest.testmod() benchmark()
313
import datetime import platform import subprocess from typing import Optional, Tuple, Union import numpy as np def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Tuple = '''1''' SCREAMING_SNAKE_CASE : Union[str, Any] = '''f32le''' SCREAMING_SNAKE_CASE : List[Any] = [ '''ffmpeg''', '''-i''', '''pipe:0''', '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] try: with subprocess.Popen(a__ , stdin=subprocess.PIPE , stdout=subprocess.PIPE ) as ffmpeg_process: SCREAMING_SNAKE_CASE : Tuple = ffmpeg_process.communicate(a__ ) except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to load audio files from filename''' ) from error SCREAMING_SNAKE_CASE : Optional[Any] = output_stream[0] SCREAMING_SNAKE_CASE : Any = np.frombuffer(a__ , np.floataa ) if audio.shape[0] == 0: raise ValueError('''Malformed soundfile''' ) return audio def UpperCAmelCase_( a__ , a__ , a__ = "f32le" , ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = F"""{sampling_rate}""" SCREAMING_SNAKE_CASE : Dict = '''1''' if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Dict = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = platform.system() if system == "Linux": SCREAMING_SNAKE_CASE : Dict = '''alsa''' SCREAMING_SNAKE_CASE : Any = '''default''' elif system == "Darwin": SCREAMING_SNAKE_CASE : Union[str, Any] = '''avfoundation''' SCREAMING_SNAKE_CASE : Optional[int] = ''':0''' elif system == "Windows": SCREAMING_SNAKE_CASE : int = '''dshow''' SCREAMING_SNAKE_CASE : Any = '''default''' SCREAMING_SNAKE_CASE : Union[str, Any] = [ '''ffmpeg''', '''-f''', format_, '''-i''', input_, '''-ac''', ac, '''-ar''', ar, '''-f''', format_for_conversion, '''-fflags''', '''nobuffer''', '''-hide_banner''', '''-loglevel''', '''quiet''', '''pipe:1''', ] SCREAMING_SNAKE_CASE : List[str] = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample SCREAMING_SNAKE_CASE : List[Any] = _ffmpeg_stream(a__ , a__ ) for item in iterator: yield item def UpperCAmelCase_( a__ , a__ , a__ = None , a__ = None , a__ = "f32le" , ): """simple docstring""" if stream_chunk_s is not None: SCREAMING_SNAKE_CASE : Tuple = stream_chunk_s else: SCREAMING_SNAKE_CASE : List[str] = chunk_length_s SCREAMING_SNAKE_CASE : Union[str, Any] = ffmpeg_microphone(a__ , a__ , format_for_conversion=a__ ) if format_for_conversion == "s16le": SCREAMING_SNAKE_CASE : Optional[int] = np.intaa SCREAMING_SNAKE_CASE : List[Any] = 2 elif format_for_conversion == "f32le": SCREAMING_SNAKE_CASE : Any = np.floataa SCREAMING_SNAKE_CASE : Union[str, Any] = 4 else: raise ValueError(F"""Unhandled format `{format_for_conversion}`. Please use `s16le` or `f32le`""" ) if stride_length_s is None: SCREAMING_SNAKE_CASE : Optional[Any] = chunk_length_s / 6 SCREAMING_SNAKE_CASE : Dict = int(round(sampling_rate * chunk_length_s ) ) * size_of_sample if isinstance(a__ , (int, float) ): SCREAMING_SNAKE_CASE : List[Any] = [stride_length_s, stride_length_s] SCREAMING_SNAKE_CASE : Any = int(round(sampling_rate * stride_length_s[0] ) ) * size_of_sample SCREAMING_SNAKE_CASE : int = int(round(sampling_rate * stride_length_s[1] ) ) * size_of_sample SCREAMING_SNAKE_CASE : Union[str, Any] = datetime.datetime.now() SCREAMING_SNAKE_CASE : Dict = datetime.timedelta(seconds=a__ ) for item in chunk_bytes_iter(a__ , a__ , stride=(stride_left, stride_right) , stream=a__ ): # Put everything back in numpy scale SCREAMING_SNAKE_CASE : Dict = np.frombuffer(item['''raw'''] , dtype=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = ( item['''stride'''][0] // size_of_sample, item['''stride'''][1] // size_of_sample, ) SCREAMING_SNAKE_CASE : Any = sampling_rate audio_time += delta if datetime.datetime.now() > audio_time + 10 * delta: # We're late !! SKIP continue yield item def UpperCAmelCase_( a__ , a__ , a__ , a__ = False ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = b'''''' SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = stride if stride_left + stride_right >= chunk_len: raise ValueError( F"""Stride needs to be strictly smaller than chunk_len: ({stride_left}, {stride_right}) vs {chunk_len}""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = 0 for raw in iterator: acc += raw if stream and len(a__ ) < chunk_len: SCREAMING_SNAKE_CASE : List[str] = (_stride_left, 0) yield {"raw": acc[:chunk_len], "stride": stride, "partial": True} else: while len(a__ ) >= chunk_len: # We are flushing the accumulator SCREAMING_SNAKE_CASE : str = (_stride_left, stride_right) SCREAMING_SNAKE_CASE : List[str] = {'''raw''': acc[:chunk_len], '''stride''': stride} if stream: SCREAMING_SNAKE_CASE : List[str] = False yield item SCREAMING_SNAKE_CASE : Dict = stride_left SCREAMING_SNAKE_CASE : int = acc[chunk_len - stride_left - stride_right :] # Last chunk if len(a__ ) > stride_left: SCREAMING_SNAKE_CASE : Optional[Any] = {'''raw''': acc, '''stride''': (_stride_left, 0)} if stream: SCREAMING_SNAKE_CASE : Union[str, Any] = False yield item def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : str = 2**24 # 16Mo try: with subprocess.Popen(a__ , stdout=subprocess.PIPE , bufsize=a__ ) as ffmpeg_process: while True: SCREAMING_SNAKE_CASE : str = ffmpeg_process.stdout.read(a__ ) if raw == b"": break yield raw except FileNotFoundError as error: raise ValueError('''ffmpeg was not found but is required to stream audio files from filename''' ) from error
313
1
import unittest from transformers import is_vision_available from transformers.pipelines import pipeline from transformers.testing_utils import ( is_pipeline_test, nested_simplify, require_tf, require_torch, require_vision, slow, ) from .test_pipelines_common import ANY if is_vision_available(): from PIL import Image else: class a_ : """simple docstring""" @staticmethod def __lowerCAmelCase ( *_lowerCamelCase , **_lowerCamelCase ) ->str: pass @is_pipeline_test @require_vision class a_ ( unittest.TestCase ): """simple docstring""" @require_torch def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Tuple = pipeline( model='''hf-internal-testing/tiny-random-clip-zero-shot-image-classification''' , ) SCREAMING_SNAKE_CASE : int = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) SCREAMING_SNAKE_CASE : Any = image_classifier(_lowerCamelCase , candidate_labels=['''a''', '''b''', '''c'''] ) # The floating scores are so close, we enter floating error approximation and the order is not guaranteed across # python and torch versions. self.assertIn( nested_simplify(_lowerCamelCase ) , [ [{'''score''': 0.3_3_3, '''label''': '''a'''}, {'''score''': 0.3_3_3, '''label''': '''b'''}, {'''score''': 0.3_3_3, '''label''': '''c'''}], [{'''score''': 0.3_3_3, '''label''': '''a'''}, {'''score''': 0.3_3_3, '''label''': '''c'''}, {'''score''': 0.3_3_3, '''label''': '''b'''}], ] , ) SCREAMING_SNAKE_CASE : Optional[int] = image_classifier([image] * 5 , candidate_labels=['''A''', '''B''', '''C'''] , batch_size=2 ) self.assertEqual( nested_simplify(_lowerCamelCase ) , [ [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], ] , ) @require_tf def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[str] = pipeline( model='''hf-internal-testing/tiny-random-clip-zero-shot-image-classification''' , framework='''tf''' ) SCREAMING_SNAKE_CASE : Tuple = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) SCREAMING_SNAKE_CASE : int = image_classifier(_lowerCamelCase , candidate_labels=['''a''', '''b''', '''c'''] ) self.assertEqual( nested_simplify(_lowerCamelCase ) , [{'''score''': 0.3_3_3, '''label''': '''a'''}, {'''score''': 0.3_3_3, '''label''': '''b'''}, {'''score''': 0.3_3_3, '''label''': '''c'''}] , ) SCREAMING_SNAKE_CASE : List[Any] = image_classifier([image] * 5 , candidate_labels=['''A''', '''B''', '''C'''] , batch_size=2 ) self.assertEqual( nested_simplify(_lowerCamelCase ) , [ [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], [ {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, {'''score''': 0.3_3_3, '''label''': ANY(_lowerCamelCase )}, ], ] , ) @slow @require_torch def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Any = pipeline( task='''zero-shot-image-classification''' , model='''openai/clip-vit-base-patch32''' , ) # This is an image of 2 cats with remotes and no planes SCREAMING_SNAKE_CASE : Dict = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = image_classifier(_lowerCamelCase , candidate_labels=['''cat''', '''plane''', '''remote'''] ) self.assertEqual( nested_simplify(_lowerCamelCase ) , [ {'''score''': 0.5_1_1, '''label''': '''remote'''}, {'''score''': 0.4_8_5, '''label''': '''cat'''}, {'''score''': 0.0_0_4, '''label''': '''plane'''}, ] , ) SCREAMING_SNAKE_CASE : int = image_classifier([image] * 5 , candidate_labels=['''cat''', '''plane''', '''remote'''] , batch_size=2 ) self.assertEqual( nested_simplify(_lowerCamelCase ) , [ [ {'''score''': 0.5_1_1, '''label''': '''remote'''}, {'''score''': 0.4_8_5, '''label''': '''cat'''}, {'''score''': 0.0_0_4, '''label''': '''plane'''}, ], ] * 5 , ) @slow @require_tf def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Optional[int] = pipeline( task='''zero-shot-image-classification''' , model='''openai/clip-vit-base-patch32''' , framework='''tf''' ) # This is an image of 2 cats with remotes and no planes SCREAMING_SNAKE_CASE : Union[str, Any] = Image.open('''./tests/fixtures/tests_samples/COCO/000000039769.png''' ) SCREAMING_SNAKE_CASE : List[Any] = image_classifier(_lowerCamelCase , candidate_labels=['''cat''', '''plane''', '''remote'''] ) self.assertEqual( nested_simplify(_lowerCamelCase ) , [ {'''score''': 0.5_1_1, '''label''': '''remote'''}, {'''score''': 0.4_8_5, '''label''': '''cat'''}, {'''score''': 0.0_0_4, '''label''': '''plane'''}, ] , ) SCREAMING_SNAKE_CASE : List[Any] = image_classifier([image] * 5 , candidate_labels=['''cat''', '''plane''', '''remote'''] , batch_size=2 ) self.assertEqual( nested_simplify(_lowerCamelCase ) , [ [ {'''score''': 0.5_1_1, '''label''': '''remote'''}, {'''score''': 0.4_8_5, '''label''': '''cat'''}, {'''score''': 0.0_0_4, '''label''': '''plane'''}, ], ] * 5 , )
313
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a__ : Tuple = {'''configuration_wavlm''': ['''WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP''', '''WavLMConfig''']} try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Dict = [ '''WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST''', '''WavLMForAudioFrameClassification''', '''WavLMForCTC''', '''WavLMForSequenceClassification''', '''WavLMForXVector''', '''WavLMModel''', '''WavLMPreTrainedModel''', ] if TYPE_CHECKING: from .configuration_wavlm import WAVLM_PRETRAINED_CONFIG_ARCHIVE_MAP, WavLMConfig try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_wavlm import ( WAVLM_PRETRAINED_MODEL_ARCHIVE_LIST, WavLMForAudioFrameClassification, WavLMForCTC, WavLMForSequenceClassification, WavLMForXVector, WavLMModel, WavLMPreTrainedModel, ) else: import sys a__ : Dict = _LazyModule(__name__, globals()['''__file__'''], _import_structure, module_spec=__spec__)
313
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 warnings from typing import List from unittest.mock import Mock import torch from torch.utils.data import DataLoader, IterableDataset, TensorDataset from accelerate.accelerator import Accelerator from accelerate.utils.dataclasses import DistributedType class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->Any: SCREAMING_SNAKE_CASE : int = data def __iter__( self ) ->Dict: for element in self.data: yield element def UpperCAmelCase_( a__=True ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = Accelerator(even_batches=a__ ) assert accelerator.num_processes == 2, "this script expects that two GPUs are available" return accelerator def UpperCAmelCase_( a__ , a__ , a__ , a__ = False ): """simple docstring""" if iterable: SCREAMING_SNAKE_CASE : int = DummyIterableDataset(torch.as_tensor(range(a__ ) ) ) else: SCREAMING_SNAKE_CASE : int = TensorDataset(torch.as_tensor(range(a__ ) ) ) SCREAMING_SNAKE_CASE : List[Any] = DataLoader(a__ , batch_size=a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = accelerator.prepare(a__ ) return dl def UpperCAmelCase_( a__ , a__ , a__ , a__ , a__ , ): """simple docstring""" SCREAMING_SNAKE_CASE : str = create_dataloader(accelerator=a__ , dataset_size=a__ , batch_size=a__ ) SCREAMING_SNAKE_CASE : Optional[int] = [len(batch[0] ) for batch in dl] if accelerator.process_index == 0: assert batch_sizes == process_0_expected_batch_sizes elif accelerator.process_index == 1: assert batch_sizes == process_1_expected_batch_sizes def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[Any] = create_accelerator() # without padding, we would expect a different number of batches verify_dataloader_batch_sizes( a__ , dataset_size=3 , batch_size=1 , process_0_expected_batch_sizes=[1, 1] , process_1_expected_batch_sizes=[1, 1] , ) # without padding, we would expect the same number of batches, but different sizes verify_dataloader_batch_sizes( a__ , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 2] , ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = create_accelerator(even_batches=a__ ) verify_dataloader_batch_sizes( a__ , dataset_size=3 , batch_size=1 , process_0_expected_batch_sizes=[1, 1] , process_1_expected_batch_sizes=[1] , ) verify_dataloader_batch_sizes( a__ , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 1] , ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = create_accelerator(even_batches=a__ ) SCREAMING_SNAKE_CASE : str = torch.nn.Linear(1 , 1 ) SCREAMING_SNAKE_CASE : Dict = accelerator.prepare(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = create_dataloader(a__ , dataset_size=3 , batch_size=1 ) SCREAMING_SNAKE_CASE : List[Any] = [] with accelerator.join_uneven_inputs([ddp_model] ): for batch_idx, batch in enumerate(a__ ): SCREAMING_SNAKE_CASE : Union[str, Any] = ddp_model(batch[0].float() ) SCREAMING_SNAKE_CASE : Dict = output.sum() loss.backward() batch_idxs.append(a__ ) accelerator.wait_for_everyone() if accelerator.process_index == 0: assert batch_idxs == [0, 1] elif accelerator.process_index == 1: assert batch_idxs == [0] def UpperCAmelCase_( a__ ): """simple docstring""" with warnings.catch_warnings(record=a__ ) as w: with accelerator.join_uneven_inputs([Mock()] ): pass assert issubclass(w[-1].category , a__ ) assert "only supported for multi-GPU" in str(w[-1].message ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = True SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = create_accelerator(even_batches=a__ ) SCREAMING_SNAKE_CASE : List[Any] = torch.nn.Linear(1 , 1 ) SCREAMING_SNAKE_CASE : str = accelerator.prepare(a__ ) SCREAMING_SNAKE_CASE : int = create_dataloader(a__ , dataset_size=3 , batch_size=1 ) SCREAMING_SNAKE_CASE : Dict = create_dataloader(a__ , dataset_size=3 , batch_size=1 ) with accelerator.join_uneven_inputs([ddp_model] , even_batches=a__ ): SCREAMING_SNAKE_CASE : Tuple = train_dl.batch_sampler.even_batches SCREAMING_SNAKE_CASE : List[Any] = valid_dl.batch_sampler.even_batches assert train_dl_overridden_value == overridden_even_batches assert valid_dl_overridden_value == overridden_even_batches assert train_dl.batch_sampler.even_batches == default_even_batches assert valid_dl.batch_sampler.even_batches == default_even_batches def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = True SCREAMING_SNAKE_CASE : int = False SCREAMING_SNAKE_CASE : Tuple = create_accelerator(even_batches=a__ ) SCREAMING_SNAKE_CASE : int = torch.nn.Linear(1 , 1 ) SCREAMING_SNAKE_CASE : Dict = accelerator.prepare(a__ ) create_dataloader(a__ , dataset_size=3 , batch_size=1 , iterable=a__ ) SCREAMING_SNAKE_CASE : Any = create_dataloader(a__ , dataset_size=3 , batch_size=1 ) with warnings.catch_warnings(): warnings.filterwarnings('''ignore''' ) try: with accelerator.join_uneven_inputs([ddp_model] , even_batches=a__ ): SCREAMING_SNAKE_CASE : Optional[int] = batch_dl.batch_sampler.even_batches except AttributeError: # ensure attribute error is not raised when processing iterable dl raise AssertionError assert batch_dl_overridden_value == overridden_even_batches assert batch_dl.batch_sampler.even_batches == default_even_batches def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = create_accelerator() SCREAMING_SNAKE_CASE : Optional[Any] = torch.nn.Linear(1 , 1 ) SCREAMING_SNAKE_CASE : int = accelerator.prepare(a__ ) create_dataloader(a__ , dataset_size=3 , batch_size=1 , iterable=a__ ) with warnings.catch_warnings(record=a__ ) as w: with accelerator.join_uneven_inputs([ddp_model] , even_batches=a__ ): pass assert issubclass(w[-1].category , a__ ) assert "only supported for map-style datasets" in str(w[-1].message ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = create_accelerator() accelerator.print('''Test that even_batches variable ensures uniform batches across processes''' ) test_default_ensures_even_batch_sizes() accelerator.print('''Run tests with even_batches disabled''' ) test_can_disable_even_batches() accelerator.print('''Test joining uneven inputs''' ) test_can_join_uneven_inputs() accelerator.print('''Test overriding even_batches when joining uneven inputs''' ) test_join_can_override_even_batches() accelerator.print('''Test overriding even_batches for mixed dataloader types''' ) test_join_can_override_for_mixed_type_dataloaders() accelerator.print('''Test overriding even_batches raises a warning for iterable dataloaders''' ) test_join_raises_warning_for_iterable_when_overriding_even_batches() accelerator.print('''Test join with non DDP distributed raises warning''' ) SCREAMING_SNAKE_CASE : Tuple = accelerator.state.distributed_type SCREAMING_SNAKE_CASE : Optional[int] = DistributedType.FSDP test_join_raises_warning_for_non_ddp_distributed(a__ ) SCREAMING_SNAKE_CASE : Tuple = original_state if __name__ == "__main__": main()
313
import copy from ...configuration_utils import PretrainedConfig from ...utils import logging from ..auto import CONFIG_MAPPING a__ : int = logging.get_logger(__name__) a__ : Optional[Any] = { '''SenseTime/deformable-detr''': '''https://huggingface.co/sensetime/deformable-detr/resolve/main/config.json''', # See all Deformable DETR models at https://huggingface.co/models?filter=deformable-detr } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'deformable_detr' __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'd_model', 'num_attention_heads': 'encoder_attention_heads', } def __init__( self , _lowerCamelCase=True , _lowerCamelCase=None , _lowerCamelCase=3 , _lowerCamelCase=300 , _lowerCamelCase=1024 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=6 , _lowerCamelCase=1024 , _lowerCamelCase=8 , _lowerCamelCase=0.0 , _lowerCamelCase=True , _lowerCamelCase="relu" , _lowerCamelCase=256 , _lowerCamelCase=0.1 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1.0 , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase="sine" , _lowerCamelCase="resnet50" , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=4 , _lowerCamelCase=False , _lowerCamelCase=300 , _lowerCamelCase=False , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=1 , _lowerCamelCase=1 , _lowerCamelCase=5 , _lowerCamelCase=2 , _lowerCamelCase=0.1 , _lowerCamelCase=0.2_5 , _lowerCamelCase=False , **_lowerCamelCase , ) ->Optional[Any]: if backbone_config is not None and use_timm_backbone: raise ValueError('''You can\'t specify both `backbone_config` and `use_timm_backbone`.''' ) if not use_timm_backbone: if backbone_config is None: logger.info('''`backbone_config` is `None`. Initializing the config with the default `ResNet` backbone.''' ) SCREAMING_SNAKE_CASE : Dict = CONFIG_MAPPING['''resnet'''](out_features=['''stage4'''] ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : List[Any] = backbone_config.get('''model_type''' ) SCREAMING_SNAKE_CASE : Optional[Any] = CONFIG_MAPPING[backbone_model_type] SCREAMING_SNAKE_CASE : int = config_class.from_dict(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = use_timm_backbone SCREAMING_SNAKE_CASE : Optional[int] = backbone_config SCREAMING_SNAKE_CASE : Union[str, Any] = num_channels SCREAMING_SNAKE_CASE : Optional[Any] = num_queries SCREAMING_SNAKE_CASE : Optional[int] = max_position_embeddings SCREAMING_SNAKE_CASE : Optional[int] = d_model SCREAMING_SNAKE_CASE : str = encoder_ffn_dim SCREAMING_SNAKE_CASE : str = encoder_layers SCREAMING_SNAKE_CASE : str = encoder_attention_heads SCREAMING_SNAKE_CASE : Optional[int] = decoder_ffn_dim SCREAMING_SNAKE_CASE : int = decoder_layers SCREAMING_SNAKE_CASE : Union[str, Any] = decoder_attention_heads SCREAMING_SNAKE_CASE : List[str] = dropout SCREAMING_SNAKE_CASE : Optional[int] = attention_dropout SCREAMING_SNAKE_CASE : str = activation_dropout SCREAMING_SNAKE_CASE : Optional[int] = activation_function SCREAMING_SNAKE_CASE : Optional[int] = init_std SCREAMING_SNAKE_CASE : List[str] = init_xavier_std SCREAMING_SNAKE_CASE : Optional[Any] = encoder_layerdrop SCREAMING_SNAKE_CASE : Union[str, Any] = auxiliary_loss SCREAMING_SNAKE_CASE : List[Any] = position_embedding_type SCREAMING_SNAKE_CASE : str = backbone SCREAMING_SNAKE_CASE : Dict = use_pretrained_backbone SCREAMING_SNAKE_CASE : Dict = dilation # deformable attributes SCREAMING_SNAKE_CASE : str = num_feature_levels SCREAMING_SNAKE_CASE : Optional[Any] = encoder_n_points SCREAMING_SNAKE_CASE : Any = decoder_n_points SCREAMING_SNAKE_CASE : str = two_stage SCREAMING_SNAKE_CASE : List[str] = two_stage_num_proposals SCREAMING_SNAKE_CASE : Dict = with_box_refine if two_stage is True and with_box_refine is False: raise ValueError('''If two_stage is True, with_box_refine must be True.''' ) # Hungarian matcher SCREAMING_SNAKE_CASE : int = class_cost SCREAMING_SNAKE_CASE : Union[str, Any] = bbox_cost SCREAMING_SNAKE_CASE : Optional[int] = giou_cost # Loss coefficients SCREAMING_SNAKE_CASE : Dict = mask_loss_coefficient SCREAMING_SNAKE_CASE : Union[str, Any] = dice_loss_coefficient SCREAMING_SNAKE_CASE : str = bbox_loss_coefficient SCREAMING_SNAKE_CASE : Tuple = giou_loss_coefficient SCREAMING_SNAKE_CASE : Optional[int] = eos_coefficient SCREAMING_SNAKE_CASE : Tuple = focal_alpha SCREAMING_SNAKE_CASE : Optional[int] = disable_custom_kernels super().__init__(is_encoder_decoder=_lowerCamelCase , **_lowerCamelCase ) @property def __lowerCAmelCase ( self ) ->int: return self.encoder_attention_heads @property def __lowerCAmelCase ( self ) ->int: return self.d_model def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : str = copy.deepcopy(self.__dict__ ) if self.backbone_config is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.backbone_config.to_dict() SCREAMING_SNAKE_CASE : Any = self.__class__.model_type return output
313
1
import argparse import torch from transformers import ( WavaVecaConfig, WavaVecaFeatureExtractor, WavaVecaForAudioFrameClassification, WavaVecaForSequenceClassification, WavaVecaForXVector, logging, ) logging.set_verbosity_info() a__ : str = logging.get_logger(__name__) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = WavaVecaForSequenceClassification.from_pretrained(a__ , config=a__ ) SCREAMING_SNAKE_CASE : int = downstream_dict['''projector.weight'''] SCREAMING_SNAKE_CASE : Any = downstream_dict['''projector.bias'''] SCREAMING_SNAKE_CASE : List[str] = downstream_dict['''model.post_net.linear.weight'''] SCREAMING_SNAKE_CASE : str = downstream_dict['''model.post_net.linear.bias'''] return model def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = WavaVecaForAudioFrameClassification.from_pretrained(a__ , config=a__ ) SCREAMING_SNAKE_CASE : Dict = downstream_dict['''model.linear.weight'''] SCREAMING_SNAKE_CASE : str = downstream_dict['''model.linear.bias'''] return model def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Union[str, Any] = WavaVecaForXVector.from_pretrained(a__ , config=a__ ) SCREAMING_SNAKE_CASE : int = downstream_dict['''connector.weight'''] SCREAMING_SNAKE_CASE : Tuple = downstream_dict['''connector.bias'''] for i, kernel_size in enumerate(hf_config.tdnn_kernel ): SCREAMING_SNAKE_CASE : Dict = downstream_dict[ F"""model.framelevel_feature_extractor.module.{i}.kernel.weight""" ] SCREAMING_SNAKE_CASE : Union[str, Any] = downstream_dict[F"""model.framelevel_feature_extractor.module.{i}.kernel.bias"""] SCREAMING_SNAKE_CASE : Tuple = downstream_dict['''model.utterancelevel_feature_extractor.linear1.weight'''] SCREAMING_SNAKE_CASE : int = downstream_dict['''model.utterancelevel_feature_extractor.linear1.bias'''] SCREAMING_SNAKE_CASE : Union[str, Any] = downstream_dict['''model.utterancelevel_feature_extractor.linear2.weight'''] SCREAMING_SNAKE_CASE : Optional[int] = downstream_dict['''model.utterancelevel_feature_extractor.linear2.bias'''] SCREAMING_SNAKE_CASE : Union[str, Any] = downstream_dict['''objective.W'''] return model @torch.no_grad() def UpperCAmelCase_( a__ , a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = torch.load(a__ , map_location='''cpu''' ) SCREAMING_SNAKE_CASE : Tuple = checkpoint['''Downstream'''] SCREAMING_SNAKE_CASE : str = WavaVecaConfig.from_pretrained(a__ ) SCREAMING_SNAKE_CASE : List[str] = WavaVecaFeatureExtractor.from_pretrained( a__ , return_attention_mask=a__ , do_normalize=a__ ) SCREAMING_SNAKE_CASE : Tuple = hf_config.architectures[0] if arch.endswith('''ForSequenceClassification''' ): SCREAMING_SNAKE_CASE : int = convert_classification(a__ , a__ , a__ ) elif arch.endswith('''ForAudioFrameClassification''' ): SCREAMING_SNAKE_CASE : List[Any] = convert_diarization(a__ , a__ , a__ ) elif arch.endswith('''ForXVector''' ): SCREAMING_SNAKE_CASE : Optional[Any] = convert_xvector(a__ , a__ , a__ ) else: raise NotImplementedError(F"""S3PRL weights conversion is not supported for {arch}""" ) if hf_config.use_weighted_layer_sum: SCREAMING_SNAKE_CASE : Any = checkpoint['''Featurizer''']['''weights'''] hf_feature_extractor.save_pretrained(a__ ) hf_model.save_pretrained(a__ ) if __name__ == "__main__": a__ : Optional[Any] = argparse.ArgumentParser() parser.add_argument( '''--base_model_name''', default=None, type=str, help='''Name of the huggingface pretrained base model.''' ) parser.add_argument('''--config_path''', default=None, type=str, help='''Path to the huggingface classifier config.''') parser.add_argument('''--checkpoint_path''', default=None, type=str, help='''Path to the s3prl checkpoint.''') parser.add_argument('''--model_dump_path''', default=None, type=str, help='''Path to the final converted model.''') a__ : Union[str, Any] = parser.parse_args() convert_saprl_checkpoint(args.base_model_name, args.config_path, args.checkpoint_path, args.model_dump_path)
313
import torch from diffusers import EulerDiscreteScheduler from diffusers.utils import torch_device from .test_schedulers import SchedulerCommonTest class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[Any] = (EulerDiscreteScheduler,) __SCREAMING_SNAKE_CASE : Optional[int] = 10 def __lowerCAmelCase ( self , **_lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = { '''num_train_timesteps''': 1100, '''beta_start''': 0.0_0_0_1, '''beta_end''': 0.0_2, '''beta_schedule''': '''linear''', } config.update(**_lowerCamelCase ) return config def __lowerCAmelCase ( self ) ->Tuple: for timesteps in [10, 50, 100, 1000]: self.check_over_configs(num_train_timesteps=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: for beta_start, beta_end in zip([0.0_0_0_0_1, 0.0_0_0_1, 0.0_0_1] , [0.0_0_0_2, 0.0_0_2, 0.0_2] ): self.check_over_configs(beta_start=_lowerCamelCase , beta_end=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: for schedule in ["linear", "scaled_linear"]: self.check_over_configs(beta_schedule=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Optional[Any]: for prediction_type in ["epsilon", "v_prediction"]: self.check_over_configs(prediction_type=_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : str = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : int = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Any = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = self.dummy_model() SCREAMING_SNAKE_CASE : int = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : Any = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[Any] = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_scheduler_config(prediction_type='''v_prediction''' ) SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps ) SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : int = self.dummy_model() SCREAMING_SNAKE_CASE : List[str] = self.dummy_sample_deter * scheduler.init_noise_sigma SCREAMING_SNAKE_CASE : List[str] = sample.to(_lowerCamelCase ) for i, t in enumerate(scheduler.timesteps ): SCREAMING_SNAKE_CASE : str = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = output.prev_sample SCREAMING_SNAKE_CASE : str = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 0.0_0_0_2 ) < 1e-2 assert abs(result_mean.item() - 2.2676e-06 ) < 1e-3 def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Optional[int] = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Dict = self.get_scheduler_config() SCREAMING_SNAKE_CASE : Tuple = scheduler_class(**_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[Any] = self.dummy_model() SCREAMING_SNAKE_CASE : str = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : Optional[Any] = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : Dict = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = output.prev_sample SCREAMING_SNAKE_CASE : List[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_0.0_8_0_7 ) < 1e-2 assert abs(result_mean.item() - 0.0_1_3_1 ) < 1e-3 def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Dict = self.scheduler_classes[0] SCREAMING_SNAKE_CASE : Optional[Any] = self.get_scheduler_config() SCREAMING_SNAKE_CASE : List[Any] = scheduler_class(**_lowerCamelCase , use_karras_sigmas=_lowerCamelCase ) scheduler.set_timesteps(self.num_inference_steps , device=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_model() SCREAMING_SNAKE_CASE : Dict = self.dummy_sample_deter * scheduler.init_noise_sigma.cpu() SCREAMING_SNAKE_CASE : int = sample.to(_lowerCamelCase ) for t in scheduler.timesteps: SCREAMING_SNAKE_CASE : List[Any] = scheduler.scale_model_input(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = model(_lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = scheduler.step(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , generator=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = output.prev_sample SCREAMING_SNAKE_CASE : Optional[Any] = torch.sum(torch.abs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Any = torch.mean(torch.abs(_lowerCamelCase ) ) assert abs(result_sum.item() - 1_2_4.5_2_2_9_9_4_9_9_5_1_1_7_1_9 ) < 1e-2 assert abs(result_mean.item() - 0.1_6_2_1_3_9_3_2_6_3_3_3_9_9_9_6_3 ) < 1e-3
313
1
from ...processing_utils import ProcessorMixin class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Dict = 'SpeechT5FeatureExtractor' __SCREAMING_SNAKE_CASE : int = 'SpeechT5Tokenizer' def __init__( self , _lowerCamelCase , _lowerCamelCase ) ->Any: super().__init__(_lowerCamelCase , _lowerCamelCase ) def __call__( self , *_lowerCamelCase , **_lowerCamelCase ) ->Tuple: SCREAMING_SNAKE_CASE : Dict = kwargs.pop('''audio''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = kwargs.pop('''text''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = kwargs.pop('''text_target''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = kwargs.pop('''audio_target''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = kwargs.pop('''sampling_rate''' , _lowerCamelCase ) if audio is not None and text is not None: raise ValueError( '''Cannot process both `audio` and `text` inputs. Did you mean `audio_target` or `text_target`?''' ) if audio_target is not None and text_target is not None: raise ValueError( '''Cannot process both `audio_target` and `text_target` inputs. Did you mean `audio` or `text`?''' ) if audio is None and audio_target is None and text is None and text_target is None: raise ValueError( '''You need to specify either an `audio`, `audio_target`, `text`, or `text_target` input to process.''' ) if audio is not None: SCREAMING_SNAKE_CASE : List[str] = self.feature_extractor(_lowerCamelCase , *_lowerCamelCase , sampling_rate=_lowerCamelCase , **_lowerCamelCase ) elif text is not None: SCREAMING_SNAKE_CASE : List[str] = self.tokenizer(_lowerCamelCase , **_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = None if audio_target is not None: SCREAMING_SNAKE_CASE : Any = self.feature_extractor(audio_target=_lowerCamelCase , *_lowerCamelCase , sampling_rate=_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = targets['''input_values'''] elif text_target is not None: SCREAMING_SNAKE_CASE : List[Any] = self.tokenizer(_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = targets['''input_ids'''] else: SCREAMING_SNAKE_CASE : int = None if inputs is None: return targets if targets is not None: SCREAMING_SNAKE_CASE : List[Any] = labels SCREAMING_SNAKE_CASE : str = targets.get('''attention_mask''' ) if decoder_attention_mask is not None: SCREAMING_SNAKE_CASE : Dict = decoder_attention_mask return inputs def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->str: SCREAMING_SNAKE_CASE : str = kwargs.pop('''input_values''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = kwargs.pop('''input_ids''' , _lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = kwargs.pop('''labels''' , _lowerCamelCase ) if input_values is not None and input_ids is not None: raise ValueError('''Cannot process both `input_values` and `input_ids` inputs.''' ) if input_values is None and input_ids is None and labels is None: raise ValueError( '''You need to specify either an `input_values`, `input_ids`, or `labels` input to be padded.''' ) if input_values is not None: SCREAMING_SNAKE_CASE : Union[str, Any] = self.feature_extractor.pad(_lowerCamelCase , *_lowerCamelCase , **_lowerCamelCase ) elif input_ids is not None: SCREAMING_SNAKE_CASE : Any = self.tokenizer.pad(_lowerCamelCase , **_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : int = None if labels is not None: if "input_ids" in labels or (isinstance(_lowerCamelCase , _lowerCamelCase ) and "input_ids" in labels[0]): SCREAMING_SNAKE_CASE : List[Any] = self.tokenizer.pad(_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = targets['''input_ids'''] else: SCREAMING_SNAKE_CASE : Dict = self.feature_extractor.feature_size SCREAMING_SNAKE_CASE : Any = self.feature_extractor.num_mel_bins SCREAMING_SNAKE_CASE : Dict = self.feature_extractor.pad(_lowerCamelCase , *_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = feature_size_hack SCREAMING_SNAKE_CASE : List[str] = targets['''input_values'''] else: SCREAMING_SNAKE_CASE : List[str] = None if inputs is None: return targets if targets is not None: SCREAMING_SNAKE_CASE : Any = labels SCREAMING_SNAKE_CASE : Optional[int] = targets.get('''attention_mask''' ) if decoder_attention_mask is not None: SCREAMING_SNAKE_CASE : List[str] = decoder_attention_mask return inputs def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->int: return self.tokenizer.batch_decode(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->List[str]: return self.tokenizer.decode(*_lowerCamelCase , **_lowerCamelCase )
313
import json from typing import Dict, List, Optional, Tuple, Union from tokenizers import pre_tokenizers, processors from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput from ...tokenization_utils_fast import PreTrainedTokenizerFast from ...utils import PaddingStrategy, logging from .tokenization_led import LEDTokenizer a__ : Dict = logging.get_logger(__name__) a__ : Dict = {'''vocab_file''': '''vocab.json''', '''merges_file''': '''merges.txt''', '''tokenizer_file''': '''tokenizer.json'''} a__ : str = { '''vocab_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json''', }, '''merges_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt''', }, '''tokenizer_file''': { '''allenai/led-base-16384''': '''https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json''', }, } a__ : Optional[int] = { '''allenai/led-base-16384''': 16_384, } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = VOCAB_FILES_NAMES __SCREAMING_SNAKE_CASE : Tuple = PRETRAINED_VOCAB_FILES_MAP __SCREAMING_SNAKE_CASE : List[str] = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __SCREAMING_SNAKE_CASE : Union[str, Any] = LEDTokenizer __SCREAMING_SNAKE_CASE : Optional[int] = ['input_ids', 'attention_mask'] def __init__( self , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase="replace" , _lowerCamelCase="<s>" , _lowerCamelCase="</s>" , _lowerCamelCase="</s>" , _lowerCamelCase="<s>" , _lowerCamelCase="<unk>" , _lowerCamelCase="<pad>" , _lowerCamelCase="<mask>" , _lowerCamelCase=False , _lowerCamelCase=True , **_lowerCamelCase , ) ->Union[str, Any]: super().__init__( _lowerCamelCase , _lowerCamelCase , tokenizer_file=_lowerCamelCase , errors=_lowerCamelCase , bos_token=_lowerCamelCase , eos_token=_lowerCamelCase , sep_token=_lowerCamelCase , cls_token=_lowerCamelCase , unk_token=_lowerCamelCase , pad_token=_lowerCamelCase , mask_token=_lowerCamelCase , add_prefix_space=_lowerCamelCase , trim_offsets=_lowerCamelCase , **_lowerCamelCase , ) SCREAMING_SNAKE_CASE : Dict = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() ) if pre_tok_state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : str = getattr(_lowerCamelCase , pre_tok_state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : Optional[int] = add_prefix_space SCREAMING_SNAKE_CASE : str = pre_tok_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = add_prefix_space # the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__` SCREAMING_SNAKE_CASE : List[Any] = '''post_processor''' SCREAMING_SNAKE_CASE : int = getattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) if tokenizer_component_instance: SCREAMING_SNAKE_CASE : Any = json.loads(tokenizer_component_instance.__getstate__() ) # The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class` if "sep" in state: SCREAMING_SNAKE_CASE : Optional[int] = tuple(state['''sep'''] ) if "cls" in state: SCREAMING_SNAKE_CASE : Optional[Any] = tuple(state['''cls'''] ) SCREAMING_SNAKE_CASE : Any = False if state.get('''add_prefix_space''' , _lowerCamelCase ) != add_prefix_space: SCREAMING_SNAKE_CASE : Union[str, Any] = add_prefix_space SCREAMING_SNAKE_CASE : Union[str, Any] = True if state.get('''trim_offsets''' , _lowerCamelCase ) != trim_offsets: SCREAMING_SNAKE_CASE : List[Any] = trim_offsets SCREAMING_SNAKE_CASE : Union[str, Any] = True if changes_to_apply: SCREAMING_SNAKE_CASE : List[str] = getattr(_lowerCamelCase , state.pop('''type''' ) ) SCREAMING_SNAKE_CASE : List[Any] = component_class(**_lowerCamelCase ) setattr(self.backend_tokenizer , _lowerCamelCase , _lowerCamelCase ) @property # Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED def __lowerCAmelCase ( self ) ->str: if self._mask_token is None: if self.verbose: logger.error('''Using mask_token, but it is not set yet.''' ) return None return str(self._mask_token ) @mask_token.setter def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: SCREAMING_SNAKE_CASE : str = AddedToken(_lowerCamelCase , lstrip=_lowerCamelCase , rstrip=_lowerCamelCase ) if isinstance(_lowerCamelCase , _lowerCamelCase ) else value SCREAMING_SNAKE_CASE : List[Any] = value def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : Tuple = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._batch_encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->BatchEncoding: SCREAMING_SNAKE_CASE : List[Any] = kwargs.get('''is_split_into_words''' , _lowerCamelCase ) if is_split_into_words and not self.add_prefix_space: raise ValueError( F"""You need to instantiate {self.__class__.__name__} with add_prefix_space=True """ '''to use it with pretokenized inputs.''' ) return super()._encode_plus(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->Tuple[str]: SCREAMING_SNAKE_CASE : Any = self._tokenizer.model.save(_lowerCamelCase , name=_lowerCamelCase ) return tuple(_lowerCamelCase ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=None ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id] if token_ids_a is None: return output return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id] def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None ) ->List[int]: SCREAMING_SNAKE_CASE : Any = [self.sep_token_id] SCREAMING_SNAKE_CASE : List[str] = [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 __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = None , _lowerCamelCase = PaddingStrategy.DO_NOT_PAD , _lowerCamelCase = None , _lowerCamelCase = None , ) ->dict: SCREAMING_SNAKE_CASE : Tuple = super()._pad( encoded_inputs=_lowerCamelCase , max_length=_lowerCamelCase , padding_strategy=_lowerCamelCase , pad_to_multiple_of=_lowerCamelCase , return_attention_mask=_lowerCamelCase , ) # Load from model defaults if return_attention_mask is None: SCREAMING_SNAKE_CASE : Optional[Any] = '''attention_mask''' in self.model_input_names if return_attention_mask and "global_attention_mask" in encoded_inputs: SCREAMING_SNAKE_CASE : int = encoded_inputs[self.model_input_names[0]] # `global_attention_mask` need to have the same length as other (sequential) inputs. SCREAMING_SNAKE_CASE : Tuple = len(encoded_inputs['''global_attention_mask'''] ) != len(_lowerCamelCase ) if needs_to_be_padded: SCREAMING_SNAKE_CASE : int = len(_lowerCamelCase ) - len(encoded_inputs['''global_attention_mask'''] ) if self.padding_side == "right": # Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend` SCREAMING_SNAKE_CASE : str = ( encoded_inputs['''global_attention_mask'''] + [-1] * difference ) elif self.padding_side == "left": SCREAMING_SNAKE_CASE : Optional[Any] = [-1] * difference + encoded_inputs[ '''global_attention_mask''' ] else: raise ValueError('''Invalid padding strategy:''' + str(self.padding_side ) ) return encoded_inputs
313
1
from __future__ import annotations import math def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if len(a__ ) != 2 or len(a[0] ) != 2 or len(a__ ) != 2 or len(b[0] ) != 2: raise Exception('''Matrices are not 2x2''' ) SCREAMING_SNAKE_CASE : Dict = [ [a[0][0] * b[0][0] + a[0][1] * b[1][0], a[0][0] * b[0][1] + a[0][1] * b[1][1]], [a[1][0] * b[0][0] + a[1][1] * b[1][0], a[1][0] * b[0][1] + a[1][1] * b[1][1]], ] return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] + matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] - matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) % 2 != 0 or len(a[0] ) % 2 != 0: raise Exception('''Odd matrices are not supported!''' ) SCREAMING_SNAKE_CASE : str = len(a__ ) SCREAMING_SNAKE_CASE : Any = matrix_length // 2 SCREAMING_SNAKE_CASE : Tuple = [[a[i][j] for j in range(a__ , a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : Optional[int] = [ [a[i][j] for j in range(a__ , a__ )] for i in range(a__ , a__ ) ] SCREAMING_SNAKE_CASE : Optional[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : List[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ , a__ )] return top_left, top_right, bot_left, bot_right def UpperCAmelCase_( a__ ): """simple docstring""" return len(a__ ), len(matrix[0] ) def UpperCAmelCase_( a__ ): """simple docstring""" print('''\n'''.join(str(a__ ) for line in matrix ) ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ ) == (2, 2): return default_matrix_multiplication(a__ , a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE : Dict = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : int = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Any = actual_strassen(matrix_addition(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = matrix_addition(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = matrix_subtraction(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) # construct the new matrix from our 4 quadrants SCREAMING_SNAKE_CASE : Optional[Any] = [] for i in range(len(a__ ) ): new_matrix.append(top_left[i] + top_right[i] ) for i in range(len(a__ ) ): new_matrix.append(bot_left[i] + bot_right[i] ) return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ )[1] != matrix_dimensions(a__ )[0]: SCREAMING_SNAKE_CASE : Any = ( '''Unable to multiply these matrices, please check the dimensions.\n''' F"""Matrix A: {matrixa}\n""" F"""Matrix B: {matrixa}""" ) raise Exception(a__ ) SCREAMING_SNAKE_CASE : str = matrix_dimensions(a__ ) SCREAMING_SNAKE_CASE : Tuple = matrix_dimensions(a__ ) if dimensiona[0] == dimensiona[1] and dimensiona[0] == dimensiona[1]: return [matrixa, matrixa] SCREAMING_SNAKE_CASE : str = max(*a__ , *a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = int(math.pow(2 , math.ceil(math.loga(a__ ) ) ) ) SCREAMING_SNAKE_CASE : Optional[int] = matrixa SCREAMING_SNAKE_CASE : Tuple = matrixa # Adding zeros to the matrices so that the arrays dimensions are the same and also # power of 2 for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) SCREAMING_SNAKE_CASE : Optional[Any] = actual_strassen(a__ , a__ ) # Removing the additional zeros for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): final_matrix[i].pop() else: final_matrix.pop() return final_matrix if __name__ == "__main__": a__ : Dict = [ [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 2, 3, 1], ] a__ : Union[str, Any] = [[0, 2, 1, 1], [16, 2, 3, 3], [2, 2, 7, 7], [13, 11, 22, 4]] print(strassen(matrixa, matrixa))
313
from __future__ import annotations import math def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if len(a__ ) != 2 or len(a[0] ) != 2 or len(a__ ) != 2 or len(b[0] ) != 2: raise Exception('''Matrices are not 2x2''' ) SCREAMING_SNAKE_CASE : Dict = [ [a[0][0] * b[0][0] + a[0][1] * b[1][0], a[0][0] * b[0][1] + a[0][1] * b[1][1]], [a[1][0] * b[0][0] + a[1][1] * b[1][0], a[1][0] * b[0][1] + a[1][1] * b[1][1]], ] return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] + matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ , a__ ): """simple docstring""" return [ [matrix_a[row][col] - matrix_b[row][col] for col in range(len(matrix_a[row] ) )] for row in range(len(a__ ) ) ] def UpperCAmelCase_( a__ ): """simple docstring""" if len(a__ ) % 2 != 0 or len(a[0] ) % 2 != 0: raise Exception('''Odd matrices are not supported!''' ) SCREAMING_SNAKE_CASE : str = len(a__ ) SCREAMING_SNAKE_CASE : Any = matrix_length // 2 SCREAMING_SNAKE_CASE : Tuple = [[a[i][j] for j in range(a__ , a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : Optional[int] = [ [a[i][j] for j in range(a__ , a__ )] for i in range(a__ , a__ ) ] SCREAMING_SNAKE_CASE : Optional[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ )] SCREAMING_SNAKE_CASE : List[Any] = [[a[i][j] for j in range(a__ )] for i in range(a__ , a__ )] return top_left, top_right, bot_left, bot_right def UpperCAmelCase_( a__ ): """simple docstring""" return len(a__ ), len(matrix[0] ) def UpperCAmelCase_( a__ ): """simple docstring""" print('''\n'''.join(str(a__ ) for line in matrix ) ) def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ ) == (2, 2): return default_matrix_multiplication(a__ , a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = split_matrix(a__ ) SCREAMING_SNAKE_CASE : Dict = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = actual_strassen(matrix_addition(a__ , a__ ) , a__ ) SCREAMING_SNAKE_CASE : int = actual_strassen(a__ , matrix_subtraction(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Any = actual_strassen(matrix_addition(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : Tuple = actual_strassen(matrix_subtraction(a__ , a__ ) , matrix_addition(a__ , a__ ) ) SCREAMING_SNAKE_CASE : List[Any] = matrix_addition(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = matrix_addition(a__ , a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = matrix_subtraction(matrix_subtraction(matrix_addition(a__ , a__ ) , a__ ) , a__ ) # construct the new matrix from our 4 quadrants SCREAMING_SNAKE_CASE : Optional[Any] = [] for i in range(len(a__ ) ): new_matrix.append(top_left[i] + top_right[i] ) for i in range(len(a__ ) ): new_matrix.append(bot_left[i] + bot_right[i] ) return new_matrix def UpperCAmelCase_( a__ , a__ ): """simple docstring""" if matrix_dimensions(a__ )[1] != matrix_dimensions(a__ )[0]: SCREAMING_SNAKE_CASE : Any = ( '''Unable to multiply these matrices, please check the dimensions.\n''' F"""Matrix A: {matrixa}\n""" F"""Matrix B: {matrixa}""" ) raise Exception(a__ ) SCREAMING_SNAKE_CASE : str = matrix_dimensions(a__ ) SCREAMING_SNAKE_CASE : Tuple = matrix_dimensions(a__ ) if dimensiona[0] == dimensiona[1] and dimensiona[0] == dimensiona[1]: return [matrixa, matrixa] SCREAMING_SNAKE_CASE : str = max(*a__ , *a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = int(math.pow(2 , math.ceil(math.loga(a__ ) ) ) ) SCREAMING_SNAKE_CASE : Optional[int] = matrixa SCREAMING_SNAKE_CASE : Tuple = matrixa # Adding zeros to the matrices so that the arrays dimensions are the same and also # power of 2 for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): new_matrixa[i].append(0 ) else: new_matrixa.append([0] * maxim ) SCREAMING_SNAKE_CASE : Optional[Any] = actual_strassen(a__ , a__ ) # Removing the additional zeros for i in range(0 , a__ ): if i < dimensiona[0]: for _ in range(dimensiona[1] , a__ ): final_matrix[i].pop() else: final_matrix.pop() return final_matrix if __name__ == "__main__": a__ : Dict = [ [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 4, 3, 1], [2, 3, 6, 7], [3, 1, 2, 4], [2, 3, 4, 5], [6, 2, 3, 1], ] a__ : Union[str, Any] = [[0, 2, 1, 1], [16, 2, 3, 3], [2, 2, 7, 7], [13, 11, 22, 4]] print(strassen(matrixa, matrixa))
313
1
def UpperCAmelCase_( a__ ): """simple docstring""" if any(not isinstance(a__ , a__ ) or x < 0 for x in sequence ): raise TypeError('''Sequence must be list of non-negative integers''' ) for _ in range(len(a__ ) ): for i, (rod_upper, rod_lower) in enumerate(zip(a__ , sequence[1:] ) ): if rod_upper > rod_lower: sequence[i] -= rod_upper - rod_lower sequence[i + 1] += rod_upper - rod_lower return sequence if __name__ == "__main__": assert bead_sort([5, 4, 3, 2, 1]) == [1, 2, 3, 4, 5] assert bead_sort([7, 9, 4, 3, 5]) == [3, 4, 5, 7, 9]
313
# We ignore warnings about stepping the scheduler since we step it ourselves during gradient accumulation import warnings from .state import AcceleratorState, GradientState warnings.filterwarnings('''ignore''', category=UserWarning, module='''torch.optim.lr_scheduler''') class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , _lowerCamelCase = False ) ->Any: SCREAMING_SNAKE_CASE : str = scheduler SCREAMING_SNAKE_CASE : List[str] = optimizers if isinstance(_lowerCamelCase , (list, tuple) ) else [optimizers] SCREAMING_SNAKE_CASE : Union[str, Any] = split_batches SCREAMING_SNAKE_CASE : List[Any] = step_with_optimizer SCREAMING_SNAKE_CASE : List[str] = GradientState() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->Optional[Any]: if not self.step_with_optimizer: # No link between scheduler and optimizer -> just step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) return # Otherwise, first make sure the optimizer was stepped. if not self.gradient_state.sync_gradients: if self.gradient_state.adjust_scheduler: self.scheduler._step_count += 1 return for opt in self.optimizers: if opt.step_was_skipped: return if self.split_batches: # Split batches -> the training dataloader batch size is not changed so one step per training step self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: # Otherwise the training dataloader batch size was multiplied by `num_processes`, so we need to do # num_processes steps per training step SCREAMING_SNAKE_CASE : List[str] = AcceleratorState().num_processes for _ in range(_lowerCamelCase ): # Special case when using OneCycle and `drop_last` was not used if hasattr(self.scheduler , '''total_steps''' ): if self.scheduler._step_count <= self.scheduler.total_steps: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) else: self.scheduler.step(*_lowerCamelCase , **_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Union[str, Any]: return self.scheduler.get_last_lr() def __lowerCAmelCase ( self ) ->List[str]: return self.scheduler.state_dict() def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: self.scheduler.load_state_dict(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->Any: return self.scheduler.get_lr() def __lowerCAmelCase ( self , *_lowerCamelCase , **_lowerCamelCase ) ->List[str]: return self.scheduler.print_lr(*_lowerCamelCase , **_lowerCamelCase )
313
1
from abc import ABC, abstractmethod from typing import List, Optional class a_ ( a__ ): """simple docstring""" def __init__( self ) ->List[str]: # test for the above condition self.test() def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : int = 0 SCREAMING_SNAKE_CASE : Union[str, Any] = False while not completed: if counter == 1: self.reset() SCREAMING_SNAKE_CASE : List[Any] = self.advance() if not self.does_advance(_lowerCamelCase ): raise Exception( '''Custom Constraint is not defined correctly. self.does_advance(self.advance()) must be true.''' ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = self.update(_lowerCamelCase ) counter += 1 if counter > 1_0000: raise Exception('''update() does not fulfill the constraint.''' ) if self.remaining() != 0: raise Exception('''Custom Constraint is not defined correctly.''' ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[int]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Optional[Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self ) ->Union[str, Any]: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) @abstractmethod def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Any: raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->int: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`token_ids` has to be a non-empty list, but is {token_ids}.""" ) if any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ): raise ValueError(F"""Each list in `token_ids` has to be a list of positive integers, but is {token_ids}.""" ) SCREAMING_SNAKE_CASE : Optional[Any] = token_ids SCREAMING_SNAKE_CASE : Union[str, Any] = len(self.token_ids ) SCREAMING_SNAKE_CASE : Any = -1 # the index of the currently fulfilled step SCREAMING_SNAKE_CASE : Any = False def __lowerCAmelCase ( self ) ->List[Any]: if self.completed: return None return self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) if self.completed: return False return token_id == self.token_ids[self.fulfilled_idx + 1] def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[str]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` has to be an `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : str = False SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = False if self.does_advance(_lowerCamelCase ): self.fulfilled_idx += 1 SCREAMING_SNAKE_CASE : str = True if self.fulfilled_idx == (self.seqlen - 1): SCREAMING_SNAKE_CASE : Any = True SCREAMING_SNAKE_CASE : Union[str, Any] = completed else: # failed to make progress. SCREAMING_SNAKE_CASE : Dict = True self.reset() return stepped, completed, reset def __lowerCAmelCase ( self ) ->List[Any]: SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = 0 def __lowerCAmelCase ( self ) ->Any: return self.seqlen - (self.fulfilled_idx + 1) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->Dict: SCREAMING_SNAKE_CASE : Any = PhrasalConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : Dict = self.seqlen SCREAMING_SNAKE_CASE : int = self.fulfilled_idx SCREAMING_SNAKE_CASE : Tuple = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase=True ) ->Dict: SCREAMING_SNAKE_CASE : Any = max([len(_lowerCamelCase ) for one in nested_token_ids] ) SCREAMING_SNAKE_CASE : List[str] = {} for token_ids in nested_token_ids: SCREAMING_SNAKE_CASE : Optional[Any] = root for tidx, token_id in enumerate(_lowerCamelCase ): if token_id not in level: SCREAMING_SNAKE_CASE : Any = {} SCREAMING_SNAKE_CASE : Tuple = level[token_id] if no_subsets and self.has_subsets(_lowerCamelCase , _lowerCamelCase ): raise ValueError( '''Each list in `nested_token_ids` can\'t be a complete subset of another list, but is''' F""" {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = root def __lowerCAmelCase ( self , _lowerCamelCase ) ->int: SCREAMING_SNAKE_CASE : List[Any] = self.trie for current_token in current_seq: SCREAMING_SNAKE_CASE : int = start[current_token] SCREAMING_SNAKE_CASE : Optional[int] = list(start.keys() ) return next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.next_tokens(_lowerCamelCase ) return len(_lowerCamelCase ) == 0 def __lowerCAmelCase ( self , _lowerCamelCase ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = list(root.values() ) if len(_lowerCamelCase ) == 0: return 1 else: return sum([self.count_leaves(_lowerCamelCase ) for nn in next_nodes] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase ) ->Dict: SCREAMING_SNAKE_CASE : List[str] = self.count_leaves(_lowerCamelCase ) return len(_lowerCamelCase ) != leaf_count class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->str: super(_lowerCamelCase , self ).__init__() if not isinstance(_lowerCamelCase , _lowerCamelCase ) or len(_lowerCamelCase ) == 0: raise ValueError(F"""`nested_token_ids` has to be a non-empty list, but is {nested_token_ids}.""" ) if any(not isinstance(_lowerCamelCase , _lowerCamelCase ) for token_ids in nested_token_ids ): raise ValueError(F"""`nested_token_ids` has to be a list of lists, but is {nested_token_ids}.""" ) if any( any((not isinstance(_lowerCamelCase , _lowerCamelCase ) or token_id < 0) for token_id in token_ids ) for token_ids in nested_token_ids ): raise ValueError( F"""Each list in `nested_token_ids` has to be a list of positive integers, but is {nested_token_ids}.""" ) SCREAMING_SNAKE_CASE : List[Any] = DisjunctiveTrie(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = nested_token_ids SCREAMING_SNAKE_CASE : Optional[int] = self.trie.max_height SCREAMING_SNAKE_CASE : Union[str, Any] = [] SCREAMING_SNAKE_CASE : Optional[int] = False def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : str = self.trie.next_tokens(self.current_seq ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Dict: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : List[str] = self.trie.next_tokens(self.current_seq ) return token_id in next_tokens def __lowerCAmelCase ( self , _lowerCamelCase ) ->Any: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` is supposed to be type `int`, but is {token_id} of type {type(_lowerCamelCase )}""" ) SCREAMING_SNAKE_CASE : int = False SCREAMING_SNAKE_CASE : List[Any] = False SCREAMING_SNAKE_CASE : Union[str, Any] = False if self.does_advance(_lowerCamelCase ): self.current_seq.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = True else: SCREAMING_SNAKE_CASE : Dict = True self.reset() SCREAMING_SNAKE_CASE : Any = self.trie.reached_leaf(self.current_seq ) SCREAMING_SNAKE_CASE : List[Any] = completed return stepped, completed, reset def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Any = False SCREAMING_SNAKE_CASE : List[Any] = [] def __lowerCAmelCase ( self ) ->Optional[Any]: if self.completed: # since this can be completed without reaching max height return 0 else: return self.seqlen - len(self.current_seq ) def __lowerCAmelCase ( self , _lowerCamelCase=False ) ->List[str]: SCREAMING_SNAKE_CASE : str = DisjunctiveConstraint(self.token_ids ) if stateful: SCREAMING_SNAKE_CASE : str = self.seqlen SCREAMING_SNAKE_CASE : int = self.current_seq SCREAMING_SNAKE_CASE : Optional[int] = self.completed return new_constraint class a_ : """simple docstring""" def __init__( self , _lowerCamelCase ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : List[Any] = constraints # max # of steps required to fulfill a given constraint SCREAMING_SNAKE_CASE : str = max([c.seqlen for c in constraints] ) SCREAMING_SNAKE_CASE : List[str] = len(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = False self.init_state() def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : Any = [] SCREAMING_SNAKE_CASE : List[Any] = None SCREAMING_SNAKE_CASE : Tuple = [constraint.copy(stateful=_lowerCamelCase ) for constraint in self.constraints] def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : str = 0 if self.inprogress_constraint: # extra points for having a constraint mid-fulfilled add += self.max_seqlen - self.inprogress_constraint.remaining() return (len(self.complete_constraints ) * self.max_seqlen) + add def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Tuple = [] if self.inprogress_constraint is None: for constraint in self.pending_constraints: # "pending" == "unfulfilled yet" SCREAMING_SNAKE_CASE : Optional[int] = constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = self.inprogress_constraint.advance() if isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.append(_lowerCamelCase ) elif isinstance(_lowerCamelCase , _lowerCamelCase ): token_list.extend(_lowerCamelCase ) if len(_lowerCamelCase ) == 0: return None else: return token_list def __lowerCAmelCase ( self , _lowerCamelCase ) ->Union[str, Any]: self.init_state() if token_ids is not None: for token in token_ids: # completes or steps **one** constraint SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[str] = self.add(_lowerCamelCase ) # the entire list of constraints are fulfilled if self.completed: break def __lowerCAmelCase ( self , _lowerCamelCase ) ->List[Any]: if not isinstance(_lowerCamelCase , _lowerCamelCase ): raise ValueError(F"""`token_id` should be an `int`, but is `{token_id}`.""" ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = False, False if self.completed: SCREAMING_SNAKE_CASE : List[str] = True SCREAMING_SNAKE_CASE : Optional[int] = False return complete, stepped if self.inprogress_constraint is not None: # In the middle of fulfilling a constraint. If the `token_id` *does* makes an incremental progress to current # job, simply update the state SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = self.inprogress_constraint.update(_lowerCamelCase ) if reset: # 1. If the next token breaks the progress, then we must restart. # e.g. constraint = "I love pies" and sequence so far is "I love" but `token_id` == "books". # But that doesn't mean we self.init_state(), since we only reset the state for this particular # constraint, not the full list of constraints. self.pending_constraints.append(self.inprogress_constraint.copy(stateful=_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = None if complete: # 2. If the next token completes the constraint, move it to completed list, set # inprogress to None. If there are no pending constraints either, then this full list of constraints # is complete. self.complete_constraints.append(self.inprogress_constraint ) SCREAMING_SNAKE_CASE : str = None if len(self.pending_constraints ) == 0: # we're done! SCREAMING_SNAKE_CASE : Optional[Any] = True else: # Not in the middle of fulfilling a constraint. So does this `token_id` helps us step towards any of our list # of constraints? for cidx, pending_constraint in enumerate(self.pending_constraints ): if pending_constraint.does_advance(_lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = pending_constraint.update(_lowerCamelCase ) if not stepped: raise Exception( '''`constraint.update(token_id)` is not yielding incremental progress, ''' '''even though `constraint.does_advance(token_id)` is true.''' ) if complete: self.complete_constraints.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = None if not complete and stepped: SCREAMING_SNAKE_CASE : Optional[Any] = pending_constraint if complete or stepped: # If we made any progress at all, then it's at least not a "pending constraint". SCREAMING_SNAKE_CASE : Union[str, Any] = ( self.pending_constraints[:cidx] + self.pending_constraints[cidx + 1 :] ) if len(self.pending_constraints ) == 0 and self.inprogress_constraint is None: # If there's no longer any pending after this and no inprogress either, then we must be # complete. SCREAMING_SNAKE_CASE : str = True break # prevent accidentally stepping through multiple constraints with just one token. return complete, stepped def __lowerCAmelCase ( self , _lowerCamelCase=True ) ->str: SCREAMING_SNAKE_CASE : Dict = ConstraintListState(self.constraints ) # we actually never though self.constraints objects # throughout this process. So it's at initialization state. if stateful: SCREAMING_SNAKE_CASE : str = [ constraint.copy(stateful=_lowerCamelCase ) for constraint in self.complete_constraints ] if self.inprogress_constraint is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.inprogress_constraint.copy(stateful=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = [constraint.copy() for constraint in self.pending_constraints] return new_state
313
import argparse import os from pathlib import Path from typing import Dict import tensorflow as tf import torch from tqdm import tqdm from transformers import PegasusConfig, PegasusForConditionalGeneration, PegasusTokenizer from transformers.models.pegasus.configuration_pegasus import DEFAULTS, task_specific_params a__ : Optional[Any] = [ # replace left string with right string to get the relevant state_dict key (identical state dict to bart) ['''memory_attention''', '''encoder_attn'''], ['''attention''', '''attn'''], ['''/''', '''.'''], ['''.LayerNorm.gamma''', '''_layer_norm.weight'''], ['''.LayerNorm.beta''', '''_layer_norm.bias'''], ['''r.layer_''', '''r.layers.'''], ['''output_proj''', '''out_proj'''], ['''ffn.dense_1.''', '''fc2.'''], ['''ffn.dense.''', '''fc1.'''], ['''ffn_layer_norm''', '''final_layer_norm'''], ['''kernel''', '''weight'''], ['''encoder_layer_norm.''', '''encoder.layer_norm.'''], ['''decoder_layer_norm.''', '''decoder.layer_norm.'''], ['''embeddings.weights''', '''shared.weight'''], ] def UpperCAmelCase_( a__ ): """simple docstring""" for pegasus_name, hf_name in PATTERNS: SCREAMING_SNAKE_CASE : Union[str, Any] = k.replace(a__ , a__ ) return k def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = DEFAULTS.copy() cfg_kwargs.update(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = PegasusConfig(**a__ ) SCREAMING_SNAKE_CASE : Optional[int] = PegasusForConditionalGeneration(a__ ) SCREAMING_SNAKE_CASE : Dict = torch_model.model.state_dict() SCREAMING_SNAKE_CASE : List[str] = {} for k, v in tf_weights.items(): SCREAMING_SNAKE_CASE : int = rename_state_dict_key(a__ ) if new_k not in sd: raise ValueError(F"""could not find new key {new_k} in state dict. (converted from {k})""" ) if "dense" in k or "proj" in new_k: SCREAMING_SNAKE_CASE : Dict = v.T SCREAMING_SNAKE_CASE : Tuple = torch.tensor(a__ , dtype=sd[new_k].dtype ) assert v.shape == sd[new_k].shape, F"""{new_k}, {k}, {v.shape}, {sd[new_k].shape}""" # make sure embedding.padding_idx is respected SCREAMING_SNAKE_CASE : Tuple = torch.zeros_like(mapping['''shared.weight'''][cfg.pad_token_id + 1] ) SCREAMING_SNAKE_CASE : int = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Union[str, Any] = mapping['''shared.weight'''] SCREAMING_SNAKE_CASE : Optional[Any] = {k: torch.zeros_like(a__ ) for k, v in sd.items() if k.endswith('''bias''' ) and k not in mapping} mapping.update(**a__ ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[Any] = torch_model.model.load_state_dict(a__ , strict=a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = [ k for k in missing if k not in ['''encoder.embed_positions.weight''', '''decoder.embed_positions.weight'''] ] assert unexpected_missing == [], F"""no matches found for the following torch keys {unexpected_missing}""" assert extra == [], F"""no matches found for the following tf keys {extra}""" return torch_model def UpperCAmelCase_( a__="./ckpt/aeslc/model.ckpt-32000" ): """simple docstring""" SCREAMING_SNAKE_CASE : str = tf.train.list_variables(a__ ) SCREAMING_SNAKE_CASE : str = {} SCREAMING_SNAKE_CASE : List[Any] = ['''Adafactor''', '''global_step'''] for name, shape in tqdm(a__ , desc='''converting tf checkpoint to dict''' ): SCREAMING_SNAKE_CASE : Union[str, Any] = any(pat in name for pat in ignore_name ) if skip_key: continue SCREAMING_SNAKE_CASE : Dict = tf.train.load_variable(a__ , a__ ) SCREAMING_SNAKE_CASE : Any = array return tf_weights def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = Path(a__ ).parent.name SCREAMING_SNAKE_CASE : Union[str, Any] = task_specific_params[F"""summarization_{dataset}"""]['''max_position_embeddings'''] SCREAMING_SNAKE_CASE : Dict = PegasusTokenizer.from_pretrained('''sshleifer/pegasus''' , model_max_length=a__ ) assert tok.model_max_length == desired_max_model_length tok.save_pretrained(a__ ) # convert model SCREAMING_SNAKE_CASE : Any = get_tf_weights_as_numpy(a__ ) SCREAMING_SNAKE_CASE : List[str] = task_specific_params[F"""summarization_{dataset}"""] if dataset == "large": SCREAMING_SNAKE_CASE : int = task_specific_params SCREAMING_SNAKE_CASE : List[str] = convert_pegasus(a__ , a__ ) torch_model.save_pretrained(a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = torch_model.state_dict() sd.pop('''model.decoder.embed_positions.weight''' ) sd.pop('''model.encoder.embed_positions.weight''' ) torch.save(a__ , Path(a__ ) / '''pytorch_model.bin''' ) if __name__ == "__main__": a__ : Union[str, Any] = argparse.ArgumentParser() # Required parameters parser.add_argument('''tf_ckpt_path''', type=str, help='''passed to tf.train.list_variables''') parser.add_argument('''save_dir''', default=None, type=str, help='''Path to the output PyTorch model.''') a__ : List[str] = parser.parse_args() if args.save_dir is None: a__ : Any = Path(args.tf_ckpt_path).parent.name a__ : int = os.path.join('''pegasus''', dataset) convert_pegasus_ckpt_to_pytorch(args.tf_ckpt_path, args.save_dir)
313
1
import operator as op a__ : Dict = '''scaler.pt''' a__ : Dict = '''pytorch_model''' a__ : Any = '''random_states''' a__ : Optional[int] = '''optimizer''' a__ : Tuple = '''scheduler''' a__ : int = '''pytorch_model.bin''' a__ : str = '''pytorch_model.bin.index.json''' a__ : Optional[Any] = '''model.safetensors''' a__ : str = '''model.safetensors.index.json''' a__ : Optional[int] = '''1.10.2''' a__ : int = '''py38''' a__ : str = '''4.17.0''' a__ : Optional[int] = ['''ml.p3.16xlarge''', '''ml.p3dn.24xlarge''', '''ml.p4dn.24xlarge'''] a__ : List[Any] = ['''FULL_SHARD''', '''SHARD_GRAD_OP''', '''NO_SHARD''', '''HYBRID_SHARD''', '''HYBRID_SHARD_ZERO2'''] a__ : List[str] = ['''TRANSFORMER_BASED_WRAP''', '''SIZE_BASED_WRAP''', '''NO_WRAP'''] a__ : int = ['''BACKWARD_PRE''', '''BACKWARD_POST''', '''NO_PREFETCH'''] a__ : List[str] = ['''FULL_STATE_DICT''', '''LOCAL_STATE_DICT''', '''SHARDED_STATE_DICT'''] a__ : int = '''2.0.1''' a__ : Union[str, Any] = ['''pdsh''', '''standard''', '''openmpi''', '''mvapich'''] a__ : Optional[Any] = ['''default''', '''reduce-overhead''', '''max-autotune'''] a__ : Union[str, Any] = {'''>''': op.gt, '''>=''': op.ge, '''==''': op.eq, '''!=''': op.ne, '''<=''': op.le, '''<''': op.lt} # These are the args for `torch.distributed.launch` for pytorch < 1.9 a__ : List[Any] = [ '''nnodes''', '''nproc_per_node''', '''rdzv_backend''', '''rdzv_endpoint''', '''rdzv_id''', '''rdzv_conf''', '''standalone''', '''max_restarts''', '''monitor_interval''', '''start_method''', '''role''', '''module''', '''m''', '''no_python''', '''run_path''', '''log_dir''', '''r''', '''redirects''', '''t''', '''tee''', '''node_rank''', '''master_addr''', '''master_port''', ] a__ : Optional[int] = ['''DEEPSPEED''', '''MULTI_GPU''', '''FSDP''', '''MEGATRON_LM'''] a__ : Any = ['''DEEPSPEED''', '''MULTI_XPU''', '''FSDP''']
313
import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import XLMRobertaTokenizerFast from diffusers import DDIMScheduler, KandinskyImgaImgPipeline, KandinskyPriorPipeline, UNetaDConditionModel, VQModel from diffusers.pipelines.kandinsky.text_encoder import MCLIPConfig, MultilingualCLIP from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : str = KandinskyImgaImgPipeline __SCREAMING_SNAKE_CASE : str = ['prompt', 'image_embeds', 'negative_image_embeds', 'image'] __SCREAMING_SNAKE_CASE : int = [ 'prompt', 'negative_prompt', 'image_embeds', 'negative_image_embeds', 'image', ] __SCREAMING_SNAKE_CASE : int = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'negative_prompt', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->int: return 32 @property def __lowerCAmelCase ( self ) ->List[str]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Tuple: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 100 @property def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = XLMRobertaTokenizerFast.from_pretrained('''YiYiXu/tiny-random-mclip-base''' ) return tokenizer @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[Any] = MCLIPConfig( numDims=self.cross_attention_dim , transformerDimensions=self.text_embedder_hidden_size , hidden_size=self.text_embedder_hidden_size , intermediate_size=37 , num_attention_heads=4 , num_hidden_layers=5 , vocab_size=1005 , ) SCREAMING_SNAKE_CASE : Dict = MultilingualCLIP(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = text_encoder.eval() return text_encoder @property def __lowerCAmelCase ( self ) ->Union[str, Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Any = { '''in_channels''': 4, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''text_image''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''text_image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : Tuple = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->List[str]: return { "block_out_channels": [32, 64], "down_block_types": ["DownEncoderBlock2D", "AttnDownEncoderBlock2D"], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": [ "AttnUpDecoderBlock2D", "UpDecoderBlock2D", ], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Optional[Any]: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : List[str] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Any = self.dummy_text_encoder SCREAMING_SNAKE_CASE : Any = self.dummy_tokenizer SCREAMING_SNAKE_CASE : List[Any] = self.dummy_unet SCREAMING_SNAKE_CASE : Optional[int] = self.dummy_movq SCREAMING_SNAKE_CASE : Optional[Any] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : Optional[Any] = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = { '''text_encoder''': text_encoder, '''tokenizer''': tokenizer, '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->str: SCREAMING_SNAKE_CASE : List[str] = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = floats_tensor((1, self.cross_attention_dim) , rng=random.Random(seed + 1 ) ).to(_lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : str = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : str = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Optional[int] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : str = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = { '''prompt''': '''horse''', '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : str = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : List[str] = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Dict = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : List[Any] = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : List[Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Optional[int] = np.array( [0.6_1_4_7_4_9_4_3, 0.6_0_7_3_5_3_9, 0.4_3_3_0_8_5_4_4, 0.5_9_2_8_2_6_9, 0.4_7_4_9_3_5_9_5, 0.4_6_7_5_5_9_7_3, 0.4_6_1_3_8_3_8, 0.4_5_3_6_8_7_9_7, 0.5_0_1_1_9_2_3_3] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->List[Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/kandinsky_img2img_frog.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : str = '''A red cartoon frog, 4k''' SCREAMING_SNAKE_CASE : Any = KandinskyPriorPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = KandinskyImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-1''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : str = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=5 , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : Dict = pipeline( _lowerCamelCase , image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=768 , width=768 , strength=0.2 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Tuple = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
import logging import os import sys from dataclasses import dataclass, field from typing import Optional import evaluate import numpy as np import torch from datasets import load_dataset from PIL import Image from torchvision.transforms import ( CenterCrop, Compose, Normalize, RandomHorizontalFlip, RandomResizedCrop, Resize, ToTensor, ) import transformers from transformers import ( MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING, AutoConfig, AutoImageProcessor, AutoModelForImageClassification, HfArgumentParser, Trainer, TrainingArguments, set_seed, ) 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 a__ : 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-classification/requirements.txt''') a__ : List[Any] = list(MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING.keys()) a__ : List[str] = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES) def UpperCAmelCase_( a__ ): """simple docstring""" with open(a__ , '''rb''' ) as f: SCREAMING_SNAKE_CASE : Dict = Image.open(a__ ) return im.convert('''RGB''' ) @dataclass class a_ : """simple docstring""" __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={ 'help': 'Name of a dataset from the hub (could be your own, possibly private dataset hosted on the hub).' } , ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field(default=a__ , metadata={'help': 'A folder containing the training data.'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field(default=a__ , metadata={'help': 'A folder containing the validation data.'} ) __SCREAMING_SNAKE_CASE : Optional[float] = field( default=0.1_5 , metadata={'help': 'Percent to split off of train for validation.'} ) __SCREAMING_SNAKE_CASE : Optional[int] = field( default=a__ , metadata={ 'help': ( 'For debugging purposes or quicker training, truncate the number of training examples to this ' 'value if set.' ) } , ) __SCREAMING_SNAKE_CASE : Optional[int] = field( default=a__ , metadata={ 'help': ( 'For debugging purposes or quicker training, truncate the number of evaluation examples to this ' 'value if set.' ) } , ) def __lowerCAmelCase ( self ) ->Tuple: if self.dataset_name is None and (self.train_dir is None and self.validation_dir is None): raise ValueError( '''You must specify either a dataset name from the hub or a train and/or validation directory.''' ) @dataclass class a_ : """simple docstring""" __SCREAMING_SNAKE_CASE : str = field( default='google/vit-base-patch16-224-in21k' , metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} , ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'If training from scratch, pass a model type from the list: ' + ', '.join(a__ )} , ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} ) __SCREAMING_SNAKE_CASE : Optional[str] = field( default=a__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from s3'} ) __SCREAMING_SNAKE_CASE : str = field( default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , ) __SCREAMING_SNAKE_CASE : str = field(default=a__ , metadata={'help': 'Name or path of preprocessor config.'} ) __SCREAMING_SNAKE_CASE : bool = field( default=a__ , metadata={ 'help': ( 'Will use the token generated when running `huggingface-cli login` (necessary to use this script ' 'with private models).' ) } , ) __SCREAMING_SNAKE_CASE : bool = field( default=a__ , metadata={'help': 'Will enable to load a pretrained model whose head dimensions are different.'} , ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : List[Any] = torch.stack([example['''pixel_values'''] for example in examples] ) SCREAMING_SNAKE_CASE : Dict = torch.tensor([example['''labels'''] for example in examples] ) return {"pixel_values": pixel_values, "labels": labels} def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = 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. SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Optional[int] = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) ) else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = 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_image_classification''' , a__ , a__ ) # 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() SCREAMING_SNAKE_CASE : Any = training_args.get_process_log_level() logger.setLevel(a__ ) transformers.utils.logging.set_verbosity(a__ ) 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. SCREAMING_SNAKE_CASE : List[Any] = None if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir: SCREAMING_SNAKE_CASE : int = 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 ) # Initialize our dataset and prepare it for the 'image-classification' task. if data_args.dataset_name is not None: SCREAMING_SNAKE_CASE : int = load_dataset( data_args.dataset_name , data_args.dataset_config_name , cache_dir=model_args.cache_dir , task='''image-classification''' , use_auth_token=True if model_args.use_auth_token else None , ) else: SCREAMING_SNAKE_CASE : Any = {} if data_args.train_dir is not None: SCREAMING_SNAKE_CASE : int = os.path.join(data_args.train_dir , '''**''' ) if data_args.validation_dir is not None: SCREAMING_SNAKE_CASE : List[str] = os.path.join(data_args.validation_dir , '''**''' ) SCREAMING_SNAKE_CASE : Tuple = load_dataset( '''imagefolder''' , data_files=a__ , cache_dir=model_args.cache_dir , task='''image-classification''' , ) # If we don't have a validation split, split off a percentage of train as validation. SCREAMING_SNAKE_CASE : Dict = None if '''validation''' in dataset.keys() else data_args.train_val_split if isinstance(data_args.train_val_split , a__ ) and data_args.train_val_split > 0.0: SCREAMING_SNAKE_CASE : Any = dataset['''train'''].train_test_split(data_args.train_val_split ) SCREAMING_SNAKE_CASE : Any = split['''train'''] SCREAMING_SNAKE_CASE : Union[str, Any] = split['''test'''] # Prepare label mappings. # We'll include these in the model's config to get human readable labels in the Inference API. SCREAMING_SNAKE_CASE : Dict = dataset['''train'''].features['''labels'''].names SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = {}, {} for i, label in enumerate(a__ ): SCREAMING_SNAKE_CASE : List[Any] = str(a__ ) SCREAMING_SNAKE_CASE : Tuple = label # Load the accuracy metric from the datasets package SCREAMING_SNAKE_CASE : Union[str, Any] = evaluate.load('''accuracy''' ) # Define our 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(a__ ): return metric.compute(predictions=np.argmax(p.predictions , axis=1 ) , references=p.label_ids ) SCREAMING_SNAKE_CASE : int = AutoConfig.from_pretrained( model_args.config_name or model_args.model_name_or_path , num_labels=len(a__ ) , labelaid=a__ , idalabel=a__ , finetuning_task='''image-classification''' , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , ) SCREAMING_SNAKE_CASE : Any = AutoModelForImageClassification.from_pretrained( model_args.model_name_or_path , from_tf=bool('''.ckpt''' in model_args.model_name_or_path ) , config=a__ , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , ignore_mismatched_sizes=model_args.ignore_mismatched_sizes , ) SCREAMING_SNAKE_CASE : Optional[int] = AutoImageProcessor.from_pretrained( model_args.image_processor_name or model_args.model_name_or_path , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , ) # Define torchvision transforms to be applied to each image. if "shortest_edge" in image_processor.size: SCREAMING_SNAKE_CASE : List[str] = image_processor.size['''shortest_edge'''] else: SCREAMING_SNAKE_CASE : Union[str, Any] = (image_processor.size['''height'''], image_processor.size['''width''']) SCREAMING_SNAKE_CASE : Optional[Any] = Normalize(mean=image_processor.image_mean , std=image_processor.image_std ) SCREAMING_SNAKE_CASE : Tuple = Compose( [ RandomResizedCrop(a__ ), RandomHorizontalFlip(), ToTensor(), normalize, ] ) SCREAMING_SNAKE_CASE : int = Compose( [ Resize(a__ ), CenterCrop(a__ ), ToTensor(), normalize, ] ) def train_transforms(a__ ): SCREAMING_SNAKE_CASE : Dict = [ _train_transforms(pil_img.convert('''RGB''' ) ) for pil_img in example_batch['''image'''] ] return example_batch def val_transforms(a__ ): SCREAMING_SNAKE_CASE : List[str] = [_val_transforms(pil_img.convert('''RGB''' ) ) for pil_img in example_batch['''image''']] return example_batch if training_args.do_train: if "train" not in dataset: raise ValueError('''--do_train requires a train dataset''' ) if data_args.max_train_samples is not None: SCREAMING_SNAKE_CASE : Optional[int] = ( dataset['''train'''].shuffle(seed=training_args.seed ).select(range(data_args.max_train_samples ) ) ) # Set the training transforms dataset["train"].set_transform(a__ ) if training_args.do_eval: if "validation" not in dataset: raise ValueError('''--do_eval requires a validation dataset''' ) if data_args.max_eval_samples is not None: SCREAMING_SNAKE_CASE : List[str] = ( dataset['''validation'''].shuffle(seed=training_args.seed ).select(range(data_args.max_eval_samples ) ) ) # Set the validation transforms dataset["validation"].set_transform(a__ ) # Initalize our trainer SCREAMING_SNAKE_CASE : Dict = Trainer( model=a__ , args=a__ , train_dataset=dataset['''train'''] if training_args.do_train else None , eval_dataset=dataset['''validation'''] if training_args.do_eval else None , compute_metrics=a__ , tokenizer=a__ , data_collator=a__ , ) # Training if training_args.do_train: SCREAMING_SNAKE_CASE : List[Any] = None if training_args.resume_from_checkpoint is not None: SCREAMING_SNAKE_CASE : Dict = training_args.resume_from_checkpoint elif last_checkpoint is not None: SCREAMING_SNAKE_CASE : str = last_checkpoint SCREAMING_SNAKE_CASE : Optional[Any] = trainer.train(resume_from_checkpoint=a__ ) 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: SCREAMING_SNAKE_CASE : Any = trainer.evaluate() trainer.log_metrics('''eval''' , a__ ) trainer.save_metrics('''eval''' , a__ ) # Write model card and (optionally) push to hub SCREAMING_SNAKE_CASE : Optional[int] = { '''finetuned_from''': model_args.model_name_or_path, '''tasks''': '''image-classification''', '''dataset''': data_args.dataset_name, '''tags''': ['''image-classification''', '''vision'''], } if training_args.push_to_hub: trainer.push_to_hub(**a__ ) else: trainer.create_model_card(**a__ ) if __name__ == "__main__": main()
313
import argparse import json import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ViTImageProcessor, ViTMSNConfig, ViTMSNModel from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD torch.set_grad_enabled(False) def UpperCAmelCase_( a__ , a__=False ): """simple docstring""" SCREAMING_SNAKE_CASE : int = [] for i in range(config.num_hidden_layers ): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append((F"""module.blocks.{i}.norm1.weight""", F"""vit.encoder.layer.{i}.layernorm_before.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm1.bias""", F"""vit.encoder.layer.{i}.layernorm_before.bias""") ) rename_keys.append( (F"""module.blocks.{i}.attn.proj.weight""", F"""vit.encoder.layer.{i}.attention.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.attn.proj.bias""", F"""vit.encoder.layer.{i}.attention.output.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.norm2.weight""", F"""vit.encoder.layer.{i}.layernorm_after.weight""") ) rename_keys.append((F"""module.blocks.{i}.norm2.bias""", F"""vit.encoder.layer.{i}.layernorm_after.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.weight""", F"""vit.encoder.layer.{i}.intermediate.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc1.bias""", F"""vit.encoder.layer.{i}.intermediate.dense.bias""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.weight""", F"""vit.encoder.layer.{i}.output.dense.weight""") ) rename_keys.append((F"""module.blocks.{i}.mlp.fc2.bias""", F"""vit.encoder.layer.{i}.output.dense.bias""") ) # projection layer + position embeddings rename_keys.extend( [ ('''module.cls_token''', '''vit.embeddings.cls_token'''), ('''module.patch_embed.proj.weight''', '''vit.embeddings.patch_embeddings.projection.weight'''), ('''module.patch_embed.proj.bias''', '''vit.embeddings.patch_embeddings.projection.bias'''), ('''module.pos_embed''', '''vit.embeddings.position_embeddings'''), ] ) if base_model: # layernorm + pooler rename_keys.extend( [ ('''module.norm.weight''', '''layernorm.weight'''), ('''module.norm.bias''', '''layernorm.bias'''), ] ) # if just the base model, we should remove "vit" from all keys that start with "vit" SCREAMING_SNAKE_CASE : Any = [(pair[0], pair[1][4:]) if pair[1].startswith('''vit''' ) else pair for pair in rename_keys] else: # layernorm + classification head rename_keys.extend( [ ('''norm.weight''', '''vit.layernorm.weight'''), ('''norm.bias''', '''vit.layernorm.bias'''), ('''head.weight''', '''classifier.weight'''), ('''head.bias''', '''classifier.bias'''), ] ) return rename_keys def UpperCAmelCase_( a__ , a__ , a__=False ): """simple docstring""" for i in range(config.num_hidden_layers ): if base_model: SCREAMING_SNAKE_CASE : Any = '''''' else: SCREAMING_SNAKE_CASE : Optional[int] = '''vit.''' # read in weights + bias of input projection layer (in timm, this is a single matrix + bias) SCREAMING_SNAKE_CASE : Optional[Any] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.weight""" ) SCREAMING_SNAKE_CASE : List[str] = state_dict.pop(F"""module.blocks.{i}.attn.qkv.bias""" ) # next, add query, keys and values (in that order) to the state dict SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ : config.hidden_size, : ] SCREAMING_SNAKE_CASE : Any = in_proj_bias[: config.hidden_size] SCREAMING_SNAKE_CASE : int = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] SCREAMING_SNAKE_CASE : List[Any] = in_proj_bias[ config.hidden_size : config.hidden_size * 2 ] SCREAMING_SNAKE_CASE : List[str] = in_proj_weight[ -config.hidden_size :, : ] SCREAMING_SNAKE_CASE : Optional[Any] = in_proj_bias[-config.hidden_size :] def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = ['''head.weight''', '''head.bias'''] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = [ '''module.fc.fc1.weight''', '''module.fc.fc1.bias''', '''module.fc.bn1.weight''', '''module.fc.bn1.bias''', '''module.fc.bn1.running_mean''', '''module.fc.bn1.running_var''', '''module.fc.bn1.num_batches_tracked''', '''module.fc.fc2.weight''', '''module.fc.fc2.bias''', '''module.fc.bn2.weight''', '''module.fc.bn2.bias''', '''module.fc.bn2.running_mean''', '''module.fc.bn2.running_var''', '''module.fc.bn2.num_batches_tracked''', '''module.fc.fc3.weight''', '''module.fc.fc3.bias''', ] for k in ignore_keys: state_dict.pop(a__ , a__ ) def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : int = dct.pop(a__ ) SCREAMING_SNAKE_CASE : Optional[Any] = val def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Any = ViTMSNConfig() SCREAMING_SNAKE_CASE : Optional[int] = 1_000 SCREAMING_SNAKE_CASE : str = '''datasets/huggingface/label-files''' SCREAMING_SNAKE_CASE : List[str] = '''imagenet-1k-id2label.json''' SCREAMING_SNAKE_CASE : Union[str, Any] = json.load(open(hf_hub_download(a__ , a__ ) , '''r''' ) ) SCREAMING_SNAKE_CASE : List[Any] = {int(a__ ): v for k, v in idalabel.items()} SCREAMING_SNAKE_CASE : str = idalabel SCREAMING_SNAKE_CASE : int = {v: k for k, v in idalabel.items()} if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = 384 SCREAMING_SNAKE_CASE : Any = 1_536 SCREAMING_SNAKE_CASE : List[str] = 6 elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Optional[int] = 1_024 SCREAMING_SNAKE_CASE : Optional[int] = 4_096 SCREAMING_SNAKE_CASE : Tuple = 24 SCREAMING_SNAKE_CASE : Union[str, Any] = 16 SCREAMING_SNAKE_CASE : Dict = 0.1 elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = 4 elif "l7" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = 7 SCREAMING_SNAKE_CASE : Union[str, Any] = 1_024 SCREAMING_SNAKE_CASE : List[Any] = 4_096 SCREAMING_SNAKE_CASE : List[Any] = 24 SCREAMING_SNAKE_CASE : Tuple = 16 SCREAMING_SNAKE_CASE : Union[str, Any] = 0.1 SCREAMING_SNAKE_CASE : Union[str, Any] = ViTMSNModel(a__ ) SCREAMING_SNAKE_CASE : Optional[int] = torch.hub.load_state_dict_from_url(a__ , map_location='''cpu''' )['''target_encoder'''] SCREAMING_SNAKE_CASE : Any = ViTImageProcessor(size=config.image_size ) remove_projection_head(a__ ) SCREAMING_SNAKE_CASE : Any = create_rename_keys(a__ , base_model=a__ ) for src, dest in rename_keys: rename_key(a__ , a__ , a__ ) read_in_q_k_v(a__ , a__ , base_model=a__ ) model.load_state_dict(a__ ) model.eval() SCREAMING_SNAKE_CASE : Any = '''http://images.cocodataset.org/val2017/000000039769.jpg''' SCREAMING_SNAKE_CASE : Dict = Image.open(requests.get(a__ , stream=a__ ).raw ) SCREAMING_SNAKE_CASE : Optional[int] = ViTImageProcessor( size=config.image_size , image_mean=a__ , image_std=a__ ) SCREAMING_SNAKE_CASE : int = image_processor(images=a__ , return_tensors='''pt''' ) # forward pass torch.manual_seed(2 ) SCREAMING_SNAKE_CASE : Tuple = model(**a__ ) SCREAMING_SNAKE_CASE : str = outputs.last_hidden_state # The following Colab Notebook was used to generate these outputs: # https://colab.research.google.com/gist/sayakpaul/3672419a04f5997827503fd84079bdd1/scratchpad.ipynb if "s16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[-1.0_915, -1.4_876, -1.1_809]] ) elif "b16" in checkpoint_url: SCREAMING_SNAKE_CASE : Tuple = torch.tensor([[14.2_889, -18.9_045, 11.7_281]] ) elif "l16" in checkpoint_url: SCREAMING_SNAKE_CASE : Union[str, Any] = torch.tensor([[41.5_028, -22.8_681, 45.6_475]] ) elif "b4" in checkpoint_url: SCREAMING_SNAKE_CASE : str = torch.tensor([[-4.3_868, 5.2_932, -0.4_137]] ) else: SCREAMING_SNAKE_CASE : List[str] = torch.tensor([[-0.1_792, -0.6_465, 2.4_263]] ) # verify logits assert torch.allclose(last_hidden_state[:, 0, :3] , a__ , atol=1e-4 ) print(F"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(a__ ) print(F"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(a__ ) if __name__ == "__main__": a__ : Dict = argparse.ArgumentParser() # Required parameters parser.add_argument( '''--checkpoint_url''', default='''https://dl.fbaipublicfiles.com/msn/vits16_800ep.pth.tar''', type=str, help='''URL of the checkpoint you\'d like to convert.''', ) parser.add_argument( '''--pytorch_dump_folder_path''', default=None, type=str, help='''Path to the output PyTorch model directory.''' ) a__ : Any = parser.parse_args() convert_vit_msn_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
313
1
import random import unittest import torch from diffusers import IFInpaintingSuperResolutionPipeline from diffusers.utils import floats_tensor from diffusers.utils.import_utils import is_xformers_available from diffusers.utils.testing_utils import skip_mps, torch_device from ..pipeline_params import ( TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS, TEXT_GUIDED_IMAGE_INPAINTING_PARAMS, ) from ..test_pipelines_common import PipelineTesterMixin from . import IFPipelineTesterMixin @skip_mps class a_ ( a__ , a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : Union[str, Any] = IFInpaintingSuperResolutionPipeline __SCREAMING_SNAKE_CASE : Any = TEXT_GUIDED_IMAGE_INPAINTING_PARAMS - {'width', 'height'} __SCREAMING_SNAKE_CASE : str = TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS.union({'original_image'} ) __SCREAMING_SNAKE_CASE : Any = PipelineTesterMixin.required_optional_params - {'latents'} def __lowerCAmelCase ( self ) ->Optional[Any]: return self._get_superresolution_dummy_components() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->List[Any]: if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : Tuple = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : str = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = floats_tensor((1, 3, 16, 16) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = floats_tensor((1, 3, 32, 32) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = floats_tensor((1, 3, 32, 32) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = { '''prompt''': '''A painting of a squirrel eating a burger''', '''image''': image, '''original_image''': original_image, '''mask_image''': mask_image, '''generator''': generator, '''num_inference_steps''': 2, '''output_type''': '''numpy''', } return inputs @unittest.skipIf( torch_device != '''cuda''' or not is_xformers_available() , reason='''XFormers attention is only available with CUDA and `xformers` installed''' , ) def __lowerCAmelCase ( self ) ->Optional[int]: self._test_xformers_attention_forwardGenerator_pass(expected_max_diff=1e-3 ) def __lowerCAmelCase ( self ) ->str: self._test_save_load_optional_components() @unittest.skipIf(torch_device != '''cuda''' , reason='''float16 requires CUDA''' ) def __lowerCAmelCase ( self ) ->Union[str, Any]: # 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 __lowerCAmelCase ( self ) ->Dict: self._test_attention_slicing_forward_pass(expected_max_diff=1e-2 ) def __lowerCAmelCase ( self ) ->Dict: self._test_save_load_local() def __lowerCAmelCase ( self ) ->Dict: self._test_inference_batch_single_identical( expected_max_diff=1e-2 , )
313
import csv import tweepy # Twitter API credentials a__ : Union[str, Any] = '''''' a__ : List[str] = '''''' a__ : Any = '''''' a__ : List[str] = '''''' def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = tweepy.OAuthHandler(a__ , a__ ) auth.set_access_token(a__ , a__ ) SCREAMING_SNAKE_CASE : List[str] = tweepy.API(a__ ) # initialize a list to hold all the tweepy Tweets SCREAMING_SNAKE_CASE : Any = [] # make initial request for most recent tweets (200 is the maximum allowed count) SCREAMING_SNAKE_CASE : List[Any] = api.user_timeline(screen_name=a__ , count=200 ) # save most recent tweets alltweets.extend(a__ ) # save the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Tuple = alltweets[-1].id - 1 # keep grabbing tweets until there are no tweets left to grab while len(a__ ) > 0: print(F"""getting tweets before {oldest}""" ) # all subsequent requests use the max_id param to prevent duplicates SCREAMING_SNAKE_CASE : Any = api.user_timeline( screen_name=a__ , count=200 , max_id=a__ ) # save most recent tweets alltweets.extend(a__ ) # update the id of the oldest tweet less one SCREAMING_SNAKE_CASE : Dict = alltweets[-1].id - 1 print(F"""...{len(a__ )} tweets downloaded so far""" ) # transform the tweepy tweets into a 2D array that will populate the csv SCREAMING_SNAKE_CASE : Optional[Any] = [[tweet.id_str, tweet.created_at, tweet.text] for tweet in alltweets] # write the csv with open(F"""new_{screen_name}_tweets.csv""" , '''w''' ) as f: SCREAMING_SNAKE_CASE : List[Any] = csv.writer(a__ ) writer.writerow(['''id''', '''created_at''', '''text'''] ) writer.writerows(a__ ) if __name__ == "__main__": # pass in the username of the account you want to download get_all_tweets('''FirePing32''')
313
1
from typing import Optional, Union import torch from torch import nn from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss from ...activations import ACTaFN from ...modeling_outputs import BaseModelOutputWithPoolingAndNoAttention, 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_mobilenet_va import MobileNetVaConfig a__ : Optional[int] = logging.get_logger(__name__) # General docstring a__ : str = '''MobileNetV1Config''' # Base docstring a__ : Union[str, Any] = '''google/mobilenet_v1_1.0_224''' a__ : Tuple = [1, 1_024, 7, 7] # Image classification docstring a__ : Union[str, Any] = '''google/mobilenet_v1_1.0_224''' a__ : Any = '''tabby, tabby cat''' a__ : int = [ '''google/mobilenet_v1_1.0_224''', '''google/mobilenet_v1_0.75_192''', # See all MobileNetV1 models at https://huggingface.co/models?filter=mobilenet_v1 ] def UpperCAmelCase_( a__ , a__ , a__=None ): """simple docstring""" SCREAMING_SNAKE_CASE : Optional[int] = {} if isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : int = model.mobilenet_va else: SCREAMING_SNAKE_CASE : Optional[Any] = model SCREAMING_SNAKE_CASE : int = '''MobilenetV1/Conv2d_0/''' SCREAMING_SNAKE_CASE : int = backbone.conv_stem.convolution.weight SCREAMING_SNAKE_CASE : Union[str, Any] = backbone.conv_stem.normalization.bias SCREAMING_SNAKE_CASE : Union[str, Any] = backbone.conv_stem.normalization.weight SCREAMING_SNAKE_CASE : Optional[Any] = backbone.conv_stem.normalization.running_mean SCREAMING_SNAKE_CASE : int = backbone.conv_stem.normalization.running_var for i in range(13 ): SCREAMING_SNAKE_CASE : Union[str, Any] = i + 1 SCREAMING_SNAKE_CASE : int = i * 2 SCREAMING_SNAKE_CASE : int = backbone.layer[pt_index] SCREAMING_SNAKE_CASE : Union[str, Any] = F"""MobilenetV1/Conv2d_{tf_index}_depthwise/""" SCREAMING_SNAKE_CASE : int = pointer.convolution.weight SCREAMING_SNAKE_CASE : Union[str, Any] = pointer.normalization.bias SCREAMING_SNAKE_CASE : str = pointer.normalization.weight SCREAMING_SNAKE_CASE : int = pointer.normalization.running_mean SCREAMING_SNAKE_CASE : Any = pointer.normalization.running_var SCREAMING_SNAKE_CASE : int = backbone.layer[pt_index + 1] SCREAMING_SNAKE_CASE : Any = F"""MobilenetV1/Conv2d_{tf_index}_pointwise/""" SCREAMING_SNAKE_CASE : List[Any] = pointer.convolution.weight SCREAMING_SNAKE_CASE : Any = pointer.normalization.bias SCREAMING_SNAKE_CASE : Tuple = pointer.normalization.weight SCREAMING_SNAKE_CASE : List[Any] = pointer.normalization.running_mean SCREAMING_SNAKE_CASE : str = pointer.normalization.running_var if isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : Dict = '''MobilenetV1/Logits/Conv2d_1c_1x1/''' SCREAMING_SNAKE_CASE : Dict = model.classifier.weight SCREAMING_SNAKE_CASE : Union[str, Any] = model.classifier.bias return tf_to_pt_map def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" try: import numpy as np import tensorflow as tf except ImportError: logger.error( '''Loading a TensorFlow models in PyTorch, requires TensorFlow to be installed. Please see ''' '''https://www.tensorflow.org/install/ for installation instructions.''' ) raise # Load weights from TF model SCREAMING_SNAKE_CASE : Optional[int] = tf.train.list_variables(a__ ) SCREAMING_SNAKE_CASE : str = {} for name, shape in init_vars: logger.info(F"""Loading TF weight {name} with shape {shape}""" ) SCREAMING_SNAKE_CASE : Union[str, Any] = tf.train.load_variable(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = array # Build TF to PyTorch weights loading map SCREAMING_SNAKE_CASE : Optional[Any] = _build_tf_to_pytorch_map(a__ , a__ , a__ ) for name, pointer in tf_to_pt_map.items(): logger.info(F"""Importing {name}""" ) if name not in tf_weights: logger.info(F"""{name} not in tf pre-trained weights, skipping""" ) continue SCREAMING_SNAKE_CASE : Any = tf_weights[name] if "depthwise_weights" in name: logger.info('''Transposing depthwise''' ) SCREAMING_SNAKE_CASE : str = np.transpose(a__ , (2, 3, 0, 1) ) elif "weights" in name: logger.info('''Transposing''' ) if len(pointer.shape ) == 2: # copying into linear layer SCREAMING_SNAKE_CASE : List[Any] = array.squeeze().transpose() else: SCREAMING_SNAKE_CASE : Tuple = np.transpose(a__ , (3, 2, 0, 1) ) if pointer.shape != array.shape: raise ValueError(F"""Pointer shape {pointer.shape} and array shape {array.shape} mismatched""" ) logger.info(F"""Initialize PyTorch weight {name} {array.shape}""" ) SCREAMING_SNAKE_CASE : List[str] = torch.from_numpy(a__ ) tf_weights.pop(a__ , a__ ) tf_weights.pop(name + '''/RMSProp''' , a__ ) tf_weights.pop(name + '''/RMSProp_1''' , a__ ) tf_weights.pop(name + '''/ExponentialMovingAverage''' , a__ ) logger.info(F"""Weights not copied to PyTorch model: {", ".join(tf_weights.keys() )}""" ) return model def UpperCAmelCase_( a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : int = features.shape[-2:] SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = conv_layer.stride SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = conv_layer.kernel_size if in_height % stride_height == 0: SCREAMING_SNAKE_CASE : Union[str, Any] = max(kernel_height - stride_height , 0 ) else: SCREAMING_SNAKE_CASE : List[str] = max(kernel_height - (in_height % stride_height) , 0 ) if in_width % stride_width == 0: SCREAMING_SNAKE_CASE : Optional[Any] = max(kernel_width - stride_width , 0 ) else: SCREAMING_SNAKE_CASE : List[str] = max(kernel_width - (in_width % stride_width) , 0 ) SCREAMING_SNAKE_CASE : int = pad_along_width // 2 SCREAMING_SNAKE_CASE : Any = pad_along_width - pad_left SCREAMING_SNAKE_CASE : Dict = pad_along_height // 2 SCREAMING_SNAKE_CASE : Dict = pad_along_height - pad_top SCREAMING_SNAKE_CASE : Any = (pad_left, pad_right, pad_top, pad_bottom) return nn.functional.pad(a__ , a__ , '''constant''' , 0.0 ) class a_ ( nn.Module ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = 1 , _lowerCamelCase = 1 , _lowerCamelCase = False , _lowerCamelCase = True , _lowerCamelCase = True , ) ->None: super().__init__() SCREAMING_SNAKE_CASE : Union[str, Any] = config if in_channels % groups != 0: raise ValueError(F"""Input channels ({in_channels}) are not divisible by {groups} groups.""" ) if out_channels % groups != 0: raise ValueError(F"""Output channels ({out_channels}) are not divisible by {groups} groups.""" ) SCREAMING_SNAKE_CASE : List[Any] = 0 if config.tf_padding else int((kernel_size - 1) / 2 ) SCREAMING_SNAKE_CASE : Tuple = nn.Convad( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , kernel_size=_lowerCamelCase , stride=_lowerCamelCase , padding=_lowerCamelCase , groups=_lowerCamelCase , bias=_lowerCamelCase , padding_mode='''zeros''' , ) if use_normalization: SCREAMING_SNAKE_CASE : List[Any] = nn.BatchNormad( num_features=_lowerCamelCase , eps=config.layer_norm_eps , momentum=0.9_9_9_7 , affine=_lowerCamelCase , track_running_stats=_lowerCamelCase , ) else: SCREAMING_SNAKE_CASE : List[Any] = None if use_activation: if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : Optional[int] = ACTaFN[use_activation] elif isinstance(config.hidden_act , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = ACTaFN[config.hidden_act] else: SCREAMING_SNAKE_CASE : int = config.hidden_act else: SCREAMING_SNAKE_CASE : Union[str, Any] = None def __lowerCAmelCase ( self , _lowerCamelCase ) ->torch.Tensor: if self.config.tf_padding: SCREAMING_SNAKE_CASE : Any = apply_tf_padding(_lowerCamelCase , self.convolution ) SCREAMING_SNAKE_CASE : Optional[int] = self.convolution(_lowerCamelCase ) if self.normalization is not None: SCREAMING_SNAKE_CASE : Optional[int] = self.normalization(_lowerCamelCase ) if self.activation is not None: SCREAMING_SNAKE_CASE : Any = self.activation(_lowerCamelCase ) return features class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Union[str, Any] = MobileNetVaConfig __SCREAMING_SNAKE_CASE : Dict = load_tf_weights_in_mobilenet_va __SCREAMING_SNAKE_CASE : List[str] = 'mobilenet_v1' __SCREAMING_SNAKE_CASE : Dict = 'pixel_values' __SCREAMING_SNAKE_CASE : Union[str, Any] = False def __lowerCAmelCase ( self , _lowerCamelCase ) ->None: if isinstance(_lowerCamelCase , (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(_lowerCamelCase , nn.BatchNormad ): module.bias.data.zero_() module.weight.data.fill_(1.0 ) a__ : Any = r''' This model is a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior. Parameters: config ([`MobileNetV1Config`]): 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. ''' a__ : Optional[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 [`MobileNetV1ImageProcessor.__call__`] for details. output_hidden_states (`bool`, *optional*): Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for more detail. return_dict (`bool`, *optional*): Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. ''' @add_start_docstrings( 'The bare MobileNetV1 model outputting raw hidden-states without any specific head on top.' , a__ , ) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase = True ) ->List[Any]: super().__init__(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = config SCREAMING_SNAKE_CASE : int = 32 SCREAMING_SNAKE_CASE : Optional[Any] = max(int(depth * config.depth_multiplier ) , config.min_depth ) SCREAMING_SNAKE_CASE : int = MobileNetVaConvLayer( _lowerCamelCase , in_channels=config.num_channels , out_channels=_lowerCamelCase , kernel_size=3 , stride=2 , ) SCREAMING_SNAKE_CASE : Optional[Any] = [1, 2, 1, 2, 1, 2, 1, 1, 1, 1, 1, 2, 1] SCREAMING_SNAKE_CASE : Union[str, Any] = nn.ModuleList() for i in range(13 ): SCREAMING_SNAKE_CASE : Tuple = out_channels if strides[i] == 2 or i == 0: depth *= 2 SCREAMING_SNAKE_CASE : Union[str, Any] = max(int(depth * config.depth_multiplier ) , config.min_depth ) self.layer.append( MobileNetVaConvLayer( _lowerCamelCase , in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , kernel_size=3 , stride=strides[i] , groups=_lowerCamelCase , ) ) self.layer.append( MobileNetVaConvLayer( _lowerCamelCase , in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , kernel_size=1 , ) ) SCREAMING_SNAKE_CASE : int = nn.AdaptiveAvgPoolad((1, 1) ) if add_pooling_layer else None # Initialize weights and apply final processing self.post_init() def __lowerCAmelCase ( self , _lowerCamelCase ) ->Tuple: raise NotImplementedError @add_start_docstrings_to_model_forward(_lowerCamelCase ) @add_code_sample_docstrings( checkpoint=_CHECKPOINT_FOR_DOC , output_type=_lowerCamelCase , config_class=_CONFIG_FOR_DOC , modality='''vision''' , expected_output=_EXPECTED_OUTPUT_SHAPE , ) def __lowerCAmelCase ( self , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , ) ->Union[tuple, BaseModelOutputWithPoolingAndNoAttention]: SCREAMING_SNAKE_CASE : Union[str, Any] = ( output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states ) SCREAMING_SNAKE_CASE : List[str] = 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''' ) SCREAMING_SNAKE_CASE : Dict = self.conv_stem(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = () if output_hidden_states else None for i, layer_module in enumerate(self.layer ): SCREAMING_SNAKE_CASE : Dict = layer_module(_lowerCamelCase ) if output_hidden_states: SCREAMING_SNAKE_CASE : Dict = all_hidden_states + (hidden_states,) SCREAMING_SNAKE_CASE : Dict = hidden_states if self.pooler is not None: SCREAMING_SNAKE_CASE : Tuple = torch.flatten(self.pooler(_lowerCamelCase ) , start_dim=1 ) else: SCREAMING_SNAKE_CASE : Optional[Any] = None if not return_dict: return tuple(v for v in [last_hidden_state, pooled_output, all_hidden_states] if v is not None ) return BaseModelOutputWithPoolingAndNoAttention( last_hidden_state=_lowerCamelCase , pooler_output=_lowerCamelCase , hidden_states=_lowerCamelCase , ) @add_start_docstrings( '\n MobileNetV1 model with an image classification head on top (a linear layer on top of the pooled features), e.g. for\n ImageNet.\n ' , a__ , ) class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase ) ->None: super().__init__(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = config.num_labels SCREAMING_SNAKE_CASE : List[Any] = MobileNetVaModel(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = self.mobilenet_va.layer[-1].convolution.out_channels # Classifier head SCREAMING_SNAKE_CASE : Any = nn.Dropout(config.classifier_dropout_prob , inplace=_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[str] = nn.Linear(_lowerCamelCase , 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(_lowerCamelCase ) @add_code_sample_docstrings( checkpoint=_IMAGE_CLASS_CHECKPOINT , output_type=_lowerCamelCase , config_class=_CONFIG_FOR_DOC , expected_output=_IMAGE_CLASS_EXPECTED_OUTPUT , ) def __lowerCAmelCase ( self , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = None , ) ->Union[tuple, ImageClassifierOutputWithNoAttention]: SCREAMING_SNAKE_CASE : Union[str, Any] = return_dict if return_dict is not None else self.config.use_return_dict SCREAMING_SNAKE_CASE : str = self.mobilenet_va(_lowerCamelCase , output_hidden_states=_lowerCamelCase , return_dict=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = outputs.pooler_output if return_dict else outputs[1] SCREAMING_SNAKE_CASE : List[Any] = self.classifier(self.dropout(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = None if labels is not None: if self.config.problem_type is None: if self.num_labels == 1: SCREAMING_SNAKE_CASE : List[Any] = '''regression''' elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int): SCREAMING_SNAKE_CASE : Dict = '''single_label_classification''' else: SCREAMING_SNAKE_CASE : List[Any] = '''multi_label_classification''' if self.config.problem_type == "regression": SCREAMING_SNAKE_CASE : Optional[Any] = MSELoss() if self.num_labels == 1: SCREAMING_SNAKE_CASE : Tuple = loss_fct(logits.squeeze() , labels.squeeze() ) else: SCREAMING_SNAKE_CASE : str = loss_fct(_lowerCamelCase , _lowerCamelCase ) elif self.config.problem_type == "single_label_classification": SCREAMING_SNAKE_CASE : Optional[int] = CrossEntropyLoss() SCREAMING_SNAKE_CASE : Optional[Any] = loss_fct(logits.view(-1 , self.num_labels ) , labels.view(-1 ) ) elif self.config.problem_type == "multi_label_classification": SCREAMING_SNAKE_CASE : Optional[Any] = BCEWithLogitsLoss() SCREAMING_SNAKE_CASE : Optional[Any] = loss_fct(_lowerCamelCase , _lowerCamelCase ) if not return_dict: SCREAMING_SNAKE_CASE : List[Any] = (logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return ImageClassifierOutputWithNoAttention( loss=_lowerCamelCase , logits=_lowerCamelCase , hidden_states=outputs.hidden_states , )
313
from collections import OrderedDict from typing import Mapping from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a__ : Optional[Any] = logging.get_logger(__name__) a__ : List[str] = { '''kssteven/ibert-roberta-base''': '''https://huggingface.co/kssteven/ibert-roberta-base/resolve/main/config.json''', '''kssteven/ibert-roberta-large''': '''https://huggingface.co/kssteven/ibert-roberta-large/resolve/main/config.json''', '''kssteven/ibert-roberta-large-mnli''': ( '''https://huggingface.co/kssteven/ibert-roberta-large-mnli/resolve/main/config.json''' ), } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : Tuple = 'ibert' def __init__( self , _lowerCamelCase=3_0522 , _lowerCamelCase=768 , _lowerCamelCase=12 , _lowerCamelCase=12 , _lowerCamelCase=3072 , _lowerCamelCase="gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=512 , _lowerCamelCase=2 , _lowerCamelCase=0.0_2 , _lowerCamelCase=1e-12 , _lowerCamelCase=1 , _lowerCamelCase=0 , _lowerCamelCase=2 , _lowerCamelCase="absolute" , _lowerCamelCase=False , _lowerCamelCase="none" , **_lowerCamelCase , ) ->Any: super().__init__(pad_token_id=_lowerCamelCase , bos_token_id=_lowerCamelCase , eos_token_id=_lowerCamelCase , **_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = vocab_size SCREAMING_SNAKE_CASE : str = hidden_size SCREAMING_SNAKE_CASE : Tuple = num_hidden_layers SCREAMING_SNAKE_CASE : List[Any] = num_attention_heads SCREAMING_SNAKE_CASE : Dict = hidden_act SCREAMING_SNAKE_CASE : Optional[int] = intermediate_size SCREAMING_SNAKE_CASE : Optional[Any] = hidden_dropout_prob SCREAMING_SNAKE_CASE : Optional[int] = attention_probs_dropout_prob SCREAMING_SNAKE_CASE : Optional[Any] = max_position_embeddings SCREAMING_SNAKE_CASE : Tuple = type_vocab_size SCREAMING_SNAKE_CASE : Optional[int] = initializer_range SCREAMING_SNAKE_CASE : Union[str, Any] = layer_norm_eps SCREAMING_SNAKE_CASE : str = position_embedding_type SCREAMING_SNAKE_CASE : Optional[int] = quant_mode SCREAMING_SNAKE_CASE : Dict = force_dequant class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: if self.task == "multiple-choice": SCREAMING_SNAKE_CASE : Dict = {0: '''batch''', 1: '''choice''', 2: '''sequence'''} else: SCREAMING_SNAKE_CASE : List[Any] = {0: '''batch''', 1: '''sequence'''} return OrderedDict( [ ('''input_ids''', dynamic_axis), ('''attention_mask''', dynamic_axis), ] )
313
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.test_utils import execute_subprocess_async def UpperCAmelCase_( a__=None ): """simple docstring""" if subparsers is not None: SCREAMING_SNAKE_CASE : List[str] = subparsers.add_parser('''test''' ) else: SCREAMING_SNAKE_CASE : List[str] = argparse.ArgumentParser('''Accelerate test command''' ) parser.add_argument( '''--config_file''' , default=a__ , 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=a__ ) return parser def UpperCAmelCase_( a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Dict = os.path.sep.join(__file__.split(os.path.sep )[:-2] + ['''test_utils''', '''scripts''', '''test_script.py'''] ) if args.config_file is None: SCREAMING_SNAKE_CASE : Union[str, Any] = script_name else: SCREAMING_SNAKE_CASE : List[str] = F"""--config_file={args.config_file} {script_name}""" SCREAMING_SNAKE_CASE : str = ['''accelerate-launch'''] + test_args.split() SCREAMING_SNAKE_CASE : Optional[int] = execute_subprocess_async(a__ , env=os.environ.copy() ) if result.returncode == 0: print('''Test is a success! You are ready for your distributed training!''' ) def UpperCAmelCase_( ): """simple docstring""" SCREAMING_SNAKE_CASE : List[str] = test_command_parser() SCREAMING_SNAKE_CASE : Tuple = parser.parse_args() test_command(a__ ) if __name__ == "__main__": main()
313
from collections import OrderedDict from typing import TYPE_CHECKING, Any, Mapping, Optional from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging if TYPE_CHECKING: from ... import FeatureExtractionMixin, TensorType a__ : Any = logging.get_logger(__name__) a__ : Dict = { '''openai/imagegpt-small''': '''''', '''openai/imagegpt-medium''': '''''', '''openai/imagegpt-large''': '''''', } class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[str] = 'imagegpt' __SCREAMING_SNAKE_CASE : Optional[Any] = ['past_key_values'] __SCREAMING_SNAKE_CASE : Union[str, Any] = { 'hidden_size': 'n_embd', 'max_position_embeddings': 'n_positions', 'num_attention_heads': 'n_head', 'num_hidden_layers': 'n_layer', } def __init__( self , _lowerCamelCase=512 + 1 , _lowerCamelCase=32 * 32 , _lowerCamelCase=512 , _lowerCamelCase=24 , _lowerCamelCase=8 , _lowerCamelCase=None , _lowerCamelCase="quick_gelu" , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=0.1 , _lowerCamelCase=1e-5 , _lowerCamelCase=0.0_2 , _lowerCamelCase=True , _lowerCamelCase=True , _lowerCamelCase=False , _lowerCamelCase=False , _lowerCamelCase=False , **_lowerCamelCase , ) ->str: SCREAMING_SNAKE_CASE : Any = vocab_size SCREAMING_SNAKE_CASE : Optional[int] = n_positions SCREAMING_SNAKE_CASE : Optional[int] = n_embd SCREAMING_SNAKE_CASE : List[Any] = n_layer SCREAMING_SNAKE_CASE : List[Any] = n_head SCREAMING_SNAKE_CASE : int = n_inner SCREAMING_SNAKE_CASE : Dict = activation_function SCREAMING_SNAKE_CASE : Union[str, Any] = resid_pdrop SCREAMING_SNAKE_CASE : Dict = embd_pdrop SCREAMING_SNAKE_CASE : List[str] = attn_pdrop SCREAMING_SNAKE_CASE : Optional[int] = layer_norm_epsilon SCREAMING_SNAKE_CASE : Tuple = initializer_range SCREAMING_SNAKE_CASE : int = scale_attn_weights SCREAMING_SNAKE_CASE : Optional[int] = use_cache SCREAMING_SNAKE_CASE : Optional[Any] = scale_attn_by_inverse_layer_idx SCREAMING_SNAKE_CASE : str = reorder_and_upcast_attn SCREAMING_SNAKE_CASE : List[str] = tie_word_embeddings super().__init__(tie_word_embeddings=_lowerCamelCase , **_lowerCamelCase ) class a_ ( a__ ): """simple docstring""" @property def __lowerCAmelCase ( self ) ->Mapping[str, Mapping[int, str]]: return OrderedDict( [ ('''input_ids''', {0: '''batch''', 1: '''sequence'''}), ] ) def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 1 , _lowerCamelCase = -1 , _lowerCamelCase = False , _lowerCamelCase = None , _lowerCamelCase = 3 , _lowerCamelCase = 32 , _lowerCamelCase = 32 , ) ->Mapping[str, Any]: SCREAMING_SNAKE_CASE : Union[str, Any] = self._generate_dummy_images(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = dict(preprocessor(images=_lowerCamelCase , return_tensors=_lowerCamelCase ) ) return inputs
313
1
from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_tf_available, is_torch_available, ) a__ : int = { '''configuration_vision_text_dual_encoder''': ['''VisionTextDualEncoderConfig'''], '''processing_vision_text_dual_encoder''': ['''VisionTextDualEncoderProcessor'''], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : int = ['''VisionTextDualEncoderModel'''] try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : Dict = ['''FlaxVisionTextDualEncoderModel'''] try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a__ : str = ['''TFVisionTextDualEncoderModel'''] if TYPE_CHECKING: from .configuration_vision_text_dual_encoder import VisionTextDualEncoderConfig from .processing_vision_text_dual_encoder import VisionTextDualEncoderProcessor try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_vision_text_dual_encoder import VisionTextDualEncoderModel try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_flax_vision_text_dual_encoder import FlaxVisionTextDualEncoderModel try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_tf_vision_text_dual_encoder import TFVisionTextDualEncoderModel else: import sys a__ : Any = _LazyModule(__name__, globals()['''__file__'''], _import_structure)
313
from maths.prime_check import is_prime def UpperCAmelCase_( a__ ): """simple docstring""" if not isinstance(a__ , a__ ): SCREAMING_SNAKE_CASE : List[Any] = F"""Input value of [number={number}] must be an integer""" raise TypeError(a__ ) if is_prime(a__ ) and is_prime(number + 2 ): return number + 2 else: return -1 if __name__ == "__main__": import doctest doctest.testmod()
313
1
import json import os import shutil import tempfile import unittest import numpy as np import pytest from transformers import CLIPTokenizer, CLIPTokenizerFast from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES from transformers.testing_utils import require_vision from transformers.utils import IMAGE_PROCESSOR_NAME, is_vision_available if is_vision_available(): from PIL import Image from transformers import OwlViTImageProcessor, OwlViTProcessor @require_vision class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : List[Any] = tempfile.mkdtemp() # fmt: off SCREAMING_SNAKE_CASE : List[Any] = ['''''', '''l''', '''o''', '''w''', '''e''', '''r''', '''s''', '''t''', '''i''', '''d''', '''n''', '''lo''', '''l</w>''', '''w</w>''', '''r</w>''', '''t</w>''', '''low</w>''', '''er</w>''', '''lowest</w>''', '''newer</w>''', '''wider''', '''<unk>''', '''<|startoftext|>''', '''<|endoftext|>'''] # fmt: on SCREAMING_SNAKE_CASE : Dict = dict(zip(_lowerCamelCase , range(len(_lowerCamelCase ) ) ) ) SCREAMING_SNAKE_CASE : Dict = ['''#version: 0.2''', '''l o''', '''lo w</w>''', '''e r</w>''', ''''''] SCREAMING_SNAKE_CASE : Dict = {'''unk_token''': '''<unk>'''} SCREAMING_SNAKE_CASE : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['''vocab_file'''] ) SCREAMING_SNAKE_CASE : Optional[int] = 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(_lowerCamelCase ) + '''\n''' ) with open(self.merges_file , '''w''' , encoding='''utf-8''' ) as fp: fp.write('''\n'''.join(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : Optional[int] = { '''do_resize''': True, '''size''': 20, '''do_center_crop''': True, '''crop_size''': 18, '''do_normalize''': True, '''image_mean''': [0.4_8_1_4_5_4_6_6, 0.4_5_7_8_2_7_5, 0.4_0_8_2_1_0_7_3], '''image_std''': [0.2_6_8_6_2_9_5_4, 0.2_6_1_3_0_2_5_8, 0.2_7_5_7_7_7_1_1], } SCREAMING_SNAKE_CASE : Union[str, Any] = os.path.join(self.tmpdirname , _lowerCamelCase ) with open(self.image_processor_file , '''w''' , encoding='''utf-8''' ) as fp: json.dump(_lowerCamelCase , _lowerCamelCase ) def __lowerCAmelCase ( self , **_lowerCamelCase ) ->Optional[int]: return CLIPTokenizer.from_pretrained(self.tmpdirname , pad_token='''!''' , **_lowerCamelCase ) def __lowerCAmelCase ( self , **_lowerCamelCase ) ->List[str]: return CLIPTokenizerFast.from_pretrained(self.tmpdirname , pad_token='''!''' , **_lowerCamelCase ) def __lowerCAmelCase ( self , **_lowerCamelCase ) ->Optional[Any]: return OwlViTImageProcessor.from_pretrained(self.tmpdirname , **_lowerCamelCase ) def __lowerCAmelCase ( self ) ->int: shutil.rmtree(self.tmpdirname ) def __lowerCAmelCase ( self ) ->Tuple: SCREAMING_SNAKE_CASE : Dict = [np.random.randint(255 , size=(3, 30, 400) , dtype=np.uinta )] SCREAMING_SNAKE_CASE : List[Any] = [Image.fromarray(np.moveaxis(_lowerCamelCase , 0 , -1 ) ) for x in image_inputs] return image_inputs def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_tokenizer() SCREAMING_SNAKE_CASE : List[str] = self.get_rust_tokenizer() SCREAMING_SNAKE_CASE : Tuple = self.get_image_processor() SCREAMING_SNAKE_CASE : Optional[int] = OwlViTProcessor(tokenizer=_lowerCamelCase , image_processor=_lowerCamelCase ) processor_slow.save_pretrained(self.tmpdirname ) SCREAMING_SNAKE_CASE : List[str] = OwlViTProcessor.from_pretrained(self.tmpdirname , use_fast=_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = OwlViTProcessor(tokenizer=_lowerCamelCase , image_processor=_lowerCamelCase ) processor_fast.save_pretrained(self.tmpdirname ) SCREAMING_SNAKE_CASE : Tuple = OwlViTProcessor.from_pretrained(self.tmpdirname ) self.assertEqual(processor_slow.tokenizer.get_vocab() , tokenizer_slow.get_vocab() ) self.assertEqual(processor_fast.tokenizer.get_vocab() , tokenizer_fast.get_vocab() ) self.assertEqual(tokenizer_slow.get_vocab() , tokenizer_fast.get_vocab() ) self.assertIsInstance(processor_slow.tokenizer , _lowerCamelCase ) self.assertIsInstance(processor_fast.tokenizer , _lowerCamelCase ) self.assertEqual(processor_slow.image_processor.to_json_string() , image_processor.to_json_string() ) self.assertEqual(processor_fast.image_processor.to_json_string() , image_processor.to_json_string() ) self.assertIsInstance(processor_slow.image_processor , _lowerCamelCase ) self.assertIsInstance(processor_fast.image_processor , _lowerCamelCase ) def __lowerCAmelCase ( self ) ->str: SCREAMING_SNAKE_CASE : Dict = OwlViTProcessor(tokenizer=self.get_tokenizer() , image_processor=self.get_image_processor() ) processor.save_pretrained(self.tmpdirname ) SCREAMING_SNAKE_CASE : Tuple = self.get_tokenizer(bos_token='''(BOS)''' , eos_token='''(EOS)''' ) SCREAMING_SNAKE_CASE : List[str] = self.get_image_processor(do_normalize=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = OwlViTProcessor.from_pretrained( self.tmpdirname , bos_token='''(BOS)''' , eos_token='''(EOS)''' , do_normalize=_lowerCamelCase ) 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 __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : str = self.get_image_processor() SCREAMING_SNAKE_CASE : Any = self.get_tokenizer() SCREAMING_SNAKE_CASE : Dict = OwlViTProcessor(tokenizer=_lowerCamelCase , image_processor=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.prepare_image_inputs() SCREAMING_SNAKE_CASE : List[Any] = image_processor(_lowerCamelCase , return_tensors='''np''' ) SCREAMING_SNAKE_CASE : List[str] = processor(images=_lowerCamelCase , return_tensors='''np''' ) for key in input_image_proc.keys(): self.assertAlmostEqual(input_image_proc[key].sum() , input_processor[key].sum() , delta=1e-2 ) def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : Optional[Any] = self.get_image_processor() SCREAMING_SNAKE_CASE : int = self.get_tokenizer() SCREAMING_SNAKE_CASE : int = OwlViTProcessor(tokenizer=_lowerCamelCase , image_processor=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = '''lower newer''' SCREAMING_SNAKE_CASE : str = processor(text=_lowerCamelCase , return_tensors='''np''' ) SCREAMING_SNAKE_CASE : int = tokenizer(_lowerCamelCase , return_tensors='''np''' ) for key in encoded_tok.keys(): self.assertListEqual(encoded_tok[key][0].tolist() , encoded_processor[key][0].tolist() ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : str = self.get_image_processor() SCREAMING_SNAKE_CASE : Union[str, Any] = self.get_tokenizer() SCREAMING_SNAKE_CASE : int = OwlViTProcessor(tokenizer=_lowerCamelCase , image_processor=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = '''lower newer''' SCREAMING_SNAKE_CASE : int = self.prepare_image_inputs() SCREAMING_SNAKE_CASE : Dict = processor(text=_lowerCamelCase , images=_lowerCamelCase ) self.assertListEqual(list(inputs.keys() ) , ['''input_ids''', '''attention_mask''', '''pixel_values'''] ) # test if it raises when no input is passed with pytest.raises(_lowerCamelCase ): processor() def __lowerCAmelCase ( self ) ->Union[str, Any]: SCREAMING_SNAKE_CASE : int = '''google/owlvit-base-patch32''' SCREAMING_SNAKE_CASE : int = OwlViTProcessor.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = ['''cat''', '''nasa badge'''] SCREAMING_SNAKE_CASE : List[str] = processor(text=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = 16 self.assertListEqual(list(inputs.keys() ) , ['''input_ids''', '''attention_mask'''] ) self.assertEqual(inputs['''input_ids'''].shape , (2, seq_length) ) # test if it raises when no input is passed with pytest.raises(_lowerCamelCase ): processor() def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : Tuple = '''google/owlvit-base-patch32''' SCREAMING_SNAKE_CASE : Any = OwlViTProcessor.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = [['''cat''', '''nasa badge'''], ['''person''']] SCREAMING_SNAKE_CASE : int = processor(text=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = 16 SCREAMING_SNAKE_CASE : List[str] = len(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = max([len(_lowerCamelCase ) for texts in input_texts] ) self.assertListEqual(list(inputs.keys() ) , ['''input_ids''', '''attention_mask'''] ) self.assertEqual(inputs['''input_ids'''].shape , (batch_size * num_max_text_queries, seq_length) ) # test if it raises when no input is passed with pytest.raises(_lowerCamelCase ): processor() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : Optional[int] = '''google/owlvit-base-patch32''' SCREAMING_SNAKE_CASE : List[Any] = OwlViTProcessor.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = ['''cat''', '''nasa badge'''] SCREAMING_SNAKE_CASE : int = processor(text=_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = 16 SCREAMING_SNAKE_CASE : str = inputs['''input_ids'''] SCREAMING_SNAKE_CASE : Tuple = [ [4_9406, 2368, 4_9407, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [4_9406, 6841, 1_1301, 4_9407, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], ] self.assertListEqual(list(inputs.keys() ) , ['''input_ids''', '''attention_mask'''] ) self.assertEqual(inputs['''input_ids'''].shape , (2, seq_length) ) self.assertListEqual(list(input_ids[0] ) , predicted_ids[0] ) self.assertListEqual(list(input_ids[1] ) , predicted_ids[1] ) def __lowerCAmelCase ( self ) ->Optional[Any]: SCREAMING_SNAKE_CASE : Tuple = self.get_image_processor() SCREAMING_SNAKE_CASE : str = self.get_tokenizer() SCREAMING_SNAKE_CASE : int = OwlViTProcessor(tokenizer=_lowerCamelCase , image_processor=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self.prepare_image_inputs() SCREAMING_SNAKE_CASE : str = self.prepare_image_inputs() SCREAMING_SNAKE_CASE : str = processor(images=_lowerCamelCase , query_images=_lowerCamelCase ) self.assertListEqual(list(inputs.keys() ) , ['''query_pixel_values''', '''pixel_values'''] ) # test if it raises when no input is passed with pytest.raises(_lowerCamelCase ): processor() def __lowerCAmelCase ( self ) ->Any: SCREAMING_SNAKE_CASE : List[Any] = self.get_image_processor() SCREAMING_SNAKE_CASE : int = self.get_tokenizer() SCREAMING_SNAKE_CASE : Dict = OwlViTProcessor(tokenizer=_lowerCamelCase , image_processor=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = [[1, 4, 5, 8, 1, 0, 8], [3, 4, 3, 1, 1, 8, 9]] SCREAMING_SNAKE_CASE : List[str] = processor.batch_decode(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = tokenizer.batch_decode(_lowerCamelCase ) self.assertListEqual(_lowerCamelCase , _lowerCamelCase )
313
import gc import random import unittest import numpy as np import torch from PIL import Image from diffusers import ( DDIMScheduler, KandinskyVaaControlnetImgaImgPipeline, KandinskyVaaPriorEmbaEmbPipeline, UNetaDConditionModel, VQModel, ) from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference enable_full_determinism() class a_ ( a__ , unittest.TestCase ): """simple docstring""" __SCREAMING_SNAKE_CASE : List[Any] = KandinskyVaaControlnetImgaImgPipeline __SCREAMING_SNAKE_CASE : Optional[int] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[Any] = ['image_embeds', 'negative_image_embeds', 'image', 'hint'] __SCREAMING_SNAKE_CASE : List[str] = [ 'generator', 'height', 'width', 'strength', 'guidance_scale', 'num_inference_steps', 'return_dict', 'guidance_scale', 'num_images_per_prompt', 'output_type', 'return_dict', ] __SCREAMING_SNAKE_CASE : List[Any] = False @property def __lowerCAmelCase ( self ) ->Optional[Any]: return 32 @property def __lowerCAmelCase ( self ) ->Optional[int]: return 32 @property def __lowerCAmelCase ( self ) ->str: return self.time_input_dim @property def __lowerCAmelCase ( self ) ->Dict: return self.time_input_dim * 4 @property def __lowerCAmelCase ( self ) ->Tuple: return 100 @property def __lowerCAmelCase ( self ) ->int: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Union[str, Any] = { '''in_channels''': 8, # Out channels is double in channels because predicts mean and variance '''out_channels''': 8, '''addition_embed_type''': '''image_hint''', '''down_block_types''': ('''ResnetDownsampleBlock2D''', '''SimpleCrossAttnDownBlock2D'''), '''up_block_types''': ('''SimpleCrossAttnUpBlock2D''', '''ResnetUpsampleBlock2D'''), '''mid_block_type''': '''UNetMidBlock2DSimpleCrossAttn''', '''block_out_channels''': (self.block_out_channels_a, self.block_out_channels_a * 2), '''layers_per_block''': 1, '''encoder_hid_dim''': self.text_embedder_hidden_size, '''encoder_hid_dim_type''': '''image_proj''', '''cross_attention_dim''': self.cross_attention_dim, '''attention_head_dim''': 4, '''resnet_time_scale_shift''': '''scale_shift''', '''class_embed_type''': None, } SCREAMING_SNAKE_CASE : List[str] = UNetaDConditionModel(**_lowerCamelCase ) return model @property def __lowerCAmelCase ( self ) ->Any: return { "block_out_channels": [32, 32, 64, 64], "down_block_types": [ "DownEncoderBlock2D", "DownEncoderBlock2D", "DownEncoderBlock2D", "AttnDownEncoderBlock2D", ], "in_channels": 3, "latent_channels": 4, "layers_per_block": 1, "norm_num_groups": 8, "norm_type": "spatial", "num_vq_embeddings": 12, "out_channels": 3, "up_block_types": ["AttnUpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D", "UpDecoderBlock2D"], "vq_embed_dim": 4, } @property def __lowerCAmelCase ( self ) ->Tuple: torch.manual_seed(0 ) SCREAMING_SNAKE_CASE : Optional[int] = VQModel(**self.dummy_movq_kwargs ) return model def __lowerCAmelCase ( self ) ->List[str]: SCREAMING_SNAKE_CASE : str = self.dummy_unet SCREAMING_SNAKE_CASE : Union[str, Any] = self.dummy_movq SCREAMING_SNAKE_CASE : List[str] = { '''num_train_timesteps''': 1000, '''beta_schedule''': '''linear''', '''beta_start''': 0.0_0_0_8_5, '''beta_end''': 0.0_1_2, '''clip_sample''': False, '''set_alpha_to_one''': False, '''steps_offset''': 0, '''prediction_type''': '''epsilon''', '''thresholding''': False, } SCREAMING_SNAKE_CASE : str = DDIMScheduler(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = { '''unet''': unet, '''scheduler''': scheduler, '''movq''': movq, } return components def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase=0 ) ->int: SCREAMING_SNAKE_CASE : str = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Union[str, Any] = floats_tensor((1, self.text_embedder_hidden_size) , rng=random.Random(seed + 1 ) ).to( _lowerCamelCase ) # create init_image SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = image.cpu().permute(0 , 2 , 3 , 1 )[0] SCREAMING_SNAKE_CASE : Dict = Image.fromarray(np.uinta(_lowerCamelCase ) ).convert('''RGB''' ).resize((256, 256) ) # create hint SCREAMING_SNAKE_CASE : Any = floats_tensor((1, 3, 64, 64) , rng=random.Random(_lowerCamelCase ) ).to(_lowerCamelCase ) if str(_lowerCamelCase ).startswith('''mps''' ): SCREAMING_SNAKE_CASE : List[str] = torch.manual_seed(_lowerCamelCase ) else: SCREAMING_SNAKE_CASE : List[str] = torch.Generator(device=_lowerCamelCase ).manual_seed(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = { '''image''': init_image, '''image_embeds''': image_embeds, '''negative_image_embeds''': negative_image_embeds, '''hint''': hint, '''generator''': generator, '''height''': 64, '''width''': 64, '''num_inference_steps''': 10, '''guidance_scale''': 7.0, '''strength''': 0.2, '''output_type''': '''np''', } return inputs def __lowerCAmelCase ( self ) ->Dict: SCREAMING_SNAKE_CASE : Optional[int] = '''cpu''' SCREAMING_SNAKE_CASE : Optional[Any] = self.get_dummy_components() SCREAMING_SNAKE_CASE : Tuple = self.pipeline_class(**_lowerCamelCase ) SCREAMING_SNAKE_CASE : int = pipe.to(_lowerCamelCase ) pipe.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = pipe(**self.get_dummy_inputs(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : str = output.images SCREAMING_SNAKE_CASE : Any = pipe( **self.get_dummy_inputs(_lowerCamelCase ) , return_dict=_lowerCamelCase , )[0] SCREAMING_SNAKE_CASE : Any = image[0, -3:, -3:, -1] SCREAMING_SNAKE_CASE : Union[str, Any] = image_from_tuple[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) SCREAMING_SNAKE_CASE : Union[str, Any] = np.array( [0.5_4_9_8_5_0_3_4, 0.5_5_5_0_9_3_6_5, 0.5_2_5_6_1_5_0_4, 0.5_5_7_0_4_9_4, 0.5_5_9_3_8_1_8, 0.5_2_6_3_9_7_9, 0.5_0_2_8_5_6_4_3, 0.5_0_6_9_8_4_6, 0.5_1_1_9_6_7_3_6] ) assert ( np.abs(image_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_slice.flatten()}""" assert ( np.abs(image_from_tuple_slice.flatten() - expected_slice ).max() < 1e-2 ), F""" expected_slice {expected_slice}, but got {image_from_tuple_slice.flatten()}""" @slow @require_torch_gpu class a_ ( unittest.TestCase ): """simple docstring""" def __lowerCAmelCase ( self ) ->Union[str, Any]: # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def __lowerCAmelCase ( self ) ->Optional[int]: SCREAMING_SNAKE_CASE : int = load_numpy( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/kandinskyv22_controlnet_img2img_robotcat_fp16.npy''' ) SCREAMING_SNAKE_CASE : int = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinsky/cat.png''' ) SCREAMING_SNAKE_CASE : Optional[Any] = init_image.resize((512, 512) ) SCREAMING_SNAKE_CASE : Dict = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main''' '''/kandinskyv22/hint_image_cat.png''' ) SCREAMING_SNAKE_CASE : List[Any] = torch.from_numpy(np.array(_lowerCamelCase ) ).float() / 2_5_5.0 SCREAMING_SNAKE_CASE : int = hint.permute(2 , 0 , 1 ).unsqueeze(0 ) SCREAMING_SNAKE_CASE : List[Any] = '''A robot, 4k photo''' SCREAMING_SNAKE_CASE : List[str] = KandinskyVaaPriorEmbaEmbPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-prior''' , torch_dtype=torch.floataa ) pipe_prior.to(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = KandinskyVaaControlnetImgaImgPipeline.from_pretrained( '''kandinsky-community/kandinsky-2-2-controlnet-depth''' , torch_dtype=torch.floataa ) SCREAMING_SNAKE_CASE : Any = pipeline.to(_lowerCamelCase ) pipeline.set_progress_bar_config(disable=_lowerCamelCase ) SCREAMING_SNAKE_CASE : Any = torch.Generator(device='''cpu''' ).manual_seed(0 ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Union[str, Any] = pipe_prior( _lowerCamelCase , image=_lowerCamelCase , strength=0.8_5 , generator=_lowerCamelCase , negative_prompt='''''' , ).to_tuple() SCREAMING_SNAKE_CASE : List[str] = pipeline( image=_lowerCamelCase , image_embeds=_lowerCamelCase , negative_image_embeds=_lowerCamelCase , hint=_lowerCamelCase , generator=_lowerCamelCase , num_inference_steps=100 , height=512 , width=512 , strength=0.5 , output_type='''np''' , ) SCREAMING_SNAKE_CASE : Any = output.images[0] assert image.shape == (512, 512, 3) assert_mean_pixel_difference(_lowerCamelCase , _lowerCamelCase )
313
1
from __future__ import annotations import numpy as np def UpperCAmelCase_( a__ ): """simple docstring""" return np.maximum(0 , a__ ) if __name__ == "__main__": print(np.array(relu([-1, 0, 5]))) # --> [0, 0, 5]
313
from typing import Any, Callable, Dict, List, Optional, Union import torch from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer from diffusers import ( AutoencoderKL, DDIMScheduler, DiffusionPipeline, LMSDiscreteScheduler, PNDMScheduler, StableDiffusionPipeline, UNetaDConditionModel, ) from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker a__ : List[str] = '''CompVis/stable-diffusion-v1-1''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-2''' a__ : Any = '''CompVis/stable-diffusion-v1-3''' a__ : Optional[Any] = '''CompVis/stable-diffusion-v1-4''' class a_ ( a__ ): """simple docstring""" def __init__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase = True , ) ->str: super()._init_() SCREAMING_SNAKE_CASE : Tuple = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : List[Any] = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = StableDiffusionPipeline.from_pretrained(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = StableDiffusionPipeline( vae=_lowerCamelCase , text_encoder=_lowerCamelCase , tokenizer=_lowerCamelCase , unet=_lowerCamelCase , scheduler=_lowerCamelCase , safety_checker=_lowerCamelCase , feature_extractor=_lowerCamelCase , requires_safety_checker=_lowerCamelCase , ) self.register_modules(pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea , pipelinea=self.pipea ) @property def __lowerCAmelCase ( self ) ->Dict[str, Any]: return {k: getattr(self , _lowerCamelCase ) for k in self.config.keys() if not k.startswith('''_''' )} def __lowerCAmelCase ( self , _lowerCamelCase = "auto" ) ->Optional[int]: if slice_size == "auto": # half the attention head size is usually a good trade-off between # speed and memory SCREAMING_SNAKE_CASE : Optional[int] = self.unet.config.attention_head_dim // 2 self.unet.set_attention_slice(_lowerCamelCase ) def __lowerCAmelCase ( self ) ->List[str]: self.enable_attention_slicing(_lowerCamelCase ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->str: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Tuple: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Optional[Any]: return self.pipea( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) @torch.no_grad() def __lowerCAmelCase ( self , _lowerCamelCase , _lowerCamelCase = 512 , _lowerCamelCase = 512 , _lowerCamelCase = 50 , _lowerCamelCase = 7.5 , _lowerCamelCase = None , _lowerCamelCase = 1 , _lowerCamelCase = 0.0 , _lowerCamelCase = None , _lowerCamelCase = None , _lowerCamelCase = "pil" , _lowerCamelCase = True , _lowerCamelCase = None , _lowerCamelCase = 1 , **_lowerCamelCase , ) ->Dict: SCREAMING_SNAKE_CASE : Optional[Any] = '''cuda''' if torch.cuda.is_available() else '''cpu''' self.to(_lowerCamelCase ) # Checks if the height and width are divisible by 8 or not if height % 8 != 0 or width % 8 != 0: raise ValueError(F"""`height` and `width` must be divisible by 8 but are {height} and {width}.""" ) # Get first result from Stable Diffusion Checkpoint v1.1 SCREAMING_SNAKE_CASE : Optional[Any] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.2 SCREAMING_SNAKE_CASE : Any = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.3 SCREAMING_SNAKE_CASE : Optional[int] = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get first result from Stable Diffusion Checkpoint v1.4 SCREAMING_SNAKE_CASE : str = self.textaimg_sda_a( prompt=_lowerCamelCase , height=_lowerCamelCase , width=_lowerCamelCase , num_inference_steps=_lowerCamelCase , guidance_scale=_lowerCamelCase , negative_prompt=_lowerCamelCase , num_images_per_prompt=_lowerCamelCase , eta=_lowerCamelCase , generator=_lowerCamelCase , latents=_lowerCamelCase , output_type=_lowerCamelCase , return_dict=_lowerCamelCase , callback=_lowerCamelCase , callback_steps=_lowerCamelCase , **_lowerCamelCase , ) # Get all result images into a single list and pass it via StableDiffusionPipelineOutput for final result return StableDiffusionPipelineOutput([resa[0], resa[0], resa[0], resa[0]] )
313
1
from random import randint from tempfile import TemporaryFile import numpy as np def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : Tuple = 0 if start < end: SCREAMING_SNAKE_CASE : List[str] = randint(a__ , a__ ) SCREAMING_SNAKE_CASE : Union[str, Any] = a[end] SCREAMING_SNAKE_CASE : Dict = a[pivot] SCREAMING_SNAKE_CASE : Optional[Any] = temp SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Dict = _in_place_partition(a__ , a__ , a__ ) count += _in_place_quick_sort(a__ , a__ , p - 1 ) count += _in_place_quick_sort(a__ , p + 1 , a__ ) return count def UpperCAmelCase_( a__ , a__ , a__ ): """simple docstring""" SCREAMING_SNAKE_CASE : str = 0 SCREAMING_SNAKE_CASE : Optional[int] = randint(a__ , a__ ) SCREAMING_SNAKE_CASE : Dict = a[end] SCREAMING_SNAKE_CASE : Dict = a[pivot] SCREAMING_SNAKE_CASE : Optional[int] = temp SCREAMING_SNAKE_CASE : int = start - 1 for index in range(a__ , a__ ): count += 1 if a[index] < a[end]: # check if current val is less than pivot value SCREAMING_SNAKE_CASE : Optional[Any] = new_pivot_index + 1 SCREAMING_SNAKE_CASE : Optional[Any] = a[new_pivot_index] SCREAMING_SNAKE_CASE : Union[str, Any] = a[index] SCREAMING_SNAKE_CASE : str = temp SCREAMING_SNAKE_CASE : Optional[int] = a[new_pivot_index + 1] SCREAMING_SNAKE_CASE : Tuple = a[end] SCREAMING_SNAKE_CASE : List[str] = temp return new_pivot_index + 1, count a__ : List[Any] = TemporaryFile() a__ : Tuple = 100 # 1000 elements are to be sorted a__ , a__ : Optional[int] = 0, 1 # mean and standard deviation a__ : int = np.random.normal(mu, sigma, p) np.save(outfile, X) print('''The array is''') print(X) outfile.seek(0) # using the same array a__ : str = np.load(outfile) a__ : Optional[int] = len(M) - 1 a__ : List[Any] = _in_place_quick_sort(M, 0, r) print( '''No of Comparisons for 100 elements selected from a standard normal distribution''' '''is :''' ) print(z)
313
from typing import Optional, Tuple, Union import flax import flax.linen as nn import jax import jax.numpy as jnp from flax.core.frozen_dict import FrozenDict from ..configuration_utils import ConfigMixin, flax_register_to_config from ..utils import BaseOutput from .embeddings_flax import FlaxTimestepEmbedding, FlaxTimesteps from .modeling_flax_utils import FlaxModelMixin from .unet_ad_blocks_flax import ( FlaxCrossAttnDownBlockaD, FlaxCrossAttnUpBlockaD, FlaxDownBlockaD, FlaxUNetMidBlockaDCrossAttn, FlaxUpBlockaD, ) @flax.struct.dataclass class a_ ( a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : jnp.ndarray @flax_register_to_config class a_ ( nn.Module , a__ , a__ ): """simple docstring""" __SCREAMING_SNAKE_CASE : int = 32 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : int = 4 __SCREAMING_SNAKE_CASE : Tuple[str] = ( "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D", ) __SCREAMING_SNAKE_CASE : Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D") __SCREAMING_SNAKE_CASE : Union[bool, Tuple[bool]] = False __SCREAMING_SNAKE_CASE : Tuple[int] = (320, 640, 1280, 1280) __SCREAMING_SNAKE_CASE : int = 2 __SCREAMING_SNAKE_CASE : Union[int, Tuple[int]] = 8 __SCREAMING_SNAKE_CASE : Optional[Union[int, Tuple[int]]] = None __SCREAMING_SNAKE_CASE : int = 1280 __SCREAMING_SNAKE_CASE : float = 0.0 __SCREAMING_SNAKE_CASE : bool = False __SCREAMING_SNAKE_CASE : jnp.dtype = jnp.floataa __SCREAMING_SNAKE_CASE : bool = True __SCREAMING_SNAKE_CASE : int = 0 __SCREAMING_SNAKE_CASE : bool = False def __lowerCAmelCase ( self , _lowerCamelCase ) ->FrozenDict: # init input tensors SCREAMING_SNAKE_CASE : List[Any] = (1, self.in_channels, self.sample_size, self.sample_size) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros(_lowerCamelCase , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Tuple = jnp.ones((1,) , dtype=jnp.intaa ) SCREAMING_SNAKE_CASE : List[Any] = jnp.zeros((1, 1, self.cross_attention_dim) , dtype=jnp.floataa ) SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : List[Any] = jax.random.split(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Tuple = {'''params''': params_rng, '''dropout''': dropout_rng} return self.init(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase )["params"] def __lowerCAmelCase ( self ) ->int: SCREAMING_SNAKE_CASE : List[str] = self.block_out_channels SCREAMING_SNAKE_CASE : Optional[int] = block_out_channels[0] * 4 if self.num_attention_heads is not None: raise ValueError( '''At the moment it is not possible to define the number of attention heads via `num_attention_heads` because of a naming issue as described in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131. Passing `num_attention_heads` will only be supported in diffusers v0.19.''' ) # If `num_attention_heads` is not defined (which is the case for most models) # it will default to `attention_head_dim`. This looks weird upon first reading it and it is. # The reason for this behavior is to correct for incorrectly named variables that were introduced # when this library was created. The incorrect naming was only discovered much later in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131 # Changing `attention_head_dim` to `num_attention_heads` for 40,000+ configurations is too backwards breaking # which is why we correct for the naming here. SCREAMING_SNAKE_CASE : List[str] = self.num_attention_heads or self.attention_head_dim # input SCREAMING_SNAKE_CASE : Optional[int] = nn.Conv( block_out_channels[0] , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) # time SCREAMING_SNAKE_CASE : Tuple = FlaxTimesteps( block_out_channels[0] , flip_sin_to_cos=self.flip_sin_to_cos , freq_shift=self.config.freq_shift ) SCREAMING_SNAKE_CASE : Dict = FlaxTimestepEmbedding(_lowerCamelCase , dtype=self.dtype ) SCREAMING_SNAKE_CASE : Union[str, Any] = self.only_cross_attention if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (only_cross_attention,) * len(self.down_block_types ) if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = (num_attention_heads,) * len(self.down_block_types ) # down SCREAMING_SNAKE_CASE : Optional[int] = [] SCREAMING_SNAKE_CASE : Optional[Any] = block_out_channels[0] for i, down_block_type in enumerate(self.down_block_types ): SCREAMING_SNAKE_CASE : str = output_channel SCREAMING_SNAKE_CASE : int = block_out_channels[i] SCREAMING_SNAKE_CASE : List[Any] = i == len(_lowerCamelCase ) - 1 if down_block_type == "CrossAttnDownBlock2D": SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxCrossAttnDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , num_attention_heads=num_attention_heads[i] , add_downsample=not is_final_block , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Union[str, Any] = FlaxDownBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , dropout=self.dropout , num_layers=self.layers_per_block , add_downsample=not is_final_block , dtype=self.dtype , ) down_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : str = down_blocks # mid SCREAMING_SNAKE_CASE : int = FlaxUNetMidBlockaDCrossAttn( in_channels=block_out_channels[-1] , dropout=self.dropout , num_attention_heads=num_attention_heads[-1] , use_linear_projection=self.use_linear_projection , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) # up SCREAMING_SNAKE_CASE : str = [] SCREAMING_SNAKE_CASE : str = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : int = list(reversed(_lowerCamelCase ) ) SCREAMING_SNAKE_CASE : List[str] = reversed_block_out_channels[0] for i, up_block_type in enumerate(self.up_block_types ): SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Union[str, Any] = reversed_block_out_channels[i] SCREAMING_SNAKE_CASE : Tuple = reversed_block_out_channels[min(i + 1 , len(_lowerCamelCase ) - 1 )] SCREAMING_SNAKE_CASE : Dict = i == len(_lowerCamelCase ) - 1 if up_block_type == "CrossAttnUpBlock2D": SCREAMING_SNAKE_CASE : str = FlaxCrossAttnUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , num_attention_heads=reversed_num_attention_heads[i] , add_upsample=not is_final_block , dropout=self.dropout , use_linear_projection=self.use_linear_projection , only_cross_attention=only_cross_attention[i] , use_memory_efficient_attention=self.use_memory_efficient_attention , dtype=self.dtype , ) else: SCREAMING_SNAKE_CASE : Optional[int] = FlaxUpBlockaD( in_channels=_lowerCamelCase , out_channels=_lowerCamelCase , prev_output_channel=_lowerCamelCase , num_layers=self.layers_per_block + 1 , add_upsample=not is_final_block , dropout=self.dropout , dtype=self.dtype , ) up_blocks.append(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = output_channel SCREAMING_SNAKE_CASE : Tuple = up_blocks # out SCREAMING_SNAKE_CASE : Any = nn.GroupNorm(num_groups=32 , epsilon=1e-5 ) SCREAMING_SNAKE_CASE : Any = nn.Conv( self.out_channels , kernel_size=(3, 3) , strides=(1, 1) , padding=((1, 1), (1, 1)) , dtype=self.dtype , ) def __call__( self , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase , _lowerCamelCase=None , _lowerCamelCase=None , _lowerCamelCase = True , _lowerCamelCase = False , ) ->Union[FlaxUNetaDConditionOutput, Tuple]: # 1. time if not isinstance(_lowerCamelCase , jnp.ndarray ): SCREAMING_SNAKE_CASE : int = jnp.array([timesteps] , dtype=jnp.intaa ) elif isinstance(_lowerCamelCase , jnp.ndarray ) and len(timesteps.shape ) == 0: SCREAMING_SNAKE_CASE : List[str] = timesteps.astype(dtype=jnp.floataa ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.expand_dims(_lowerCamelCase , 0 ) SCREAMING_SNAKE_CASE : List[str] = self.time_proj(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = self.time_embedding(_lowerCamelCase ) # 2. pre-process SCREAMING_SNAKE_CASE : int = jnp.transpose(_lowerCamelCase , (0, 2, 3, 1) ) SCREAMING_SNAKE_CASE : List[Any] = self.conv_in(_lowerCamelCase ) # 3. down SCREAMING_SNAKE_CASE : Optional[int] = (sample,) for down_block in self.down_blocks: if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) else: SCREAMING_SNAKE_CASE , SCREAMING_SNAKE_CASE : Tuple = down_block(_lowerCamelCase , _lowerCamelCase , deterministic=not train ) down_block_res_samples += res_samples if down_block_additional_residuals is not None: SCREAMING_SNAKE_CASE : int = () for down_block_res_sample, down_block_additional_residual in zip( _lowerCamelCase , _lowerCamelCase ): down_block_res_sample += down_block_additional_residual new_down_block_res_samples += (down_block_res_sample,) SCREAMING_SNAKE_CASE : Dict = new_down_block_res_samples # 4. mid SCREAMING_SNAKE_CASE : Optional[Any] = self.mid_block(_lowerCamelCase , _lowerCamelCase , _lowerCamelCase , deterministic=not train ) if mid_block_additional_residual is not None: sample += mid_block_additional_residual # 5. up for up_block in self.up_blocks: SCREAMING_SNAKE_CASE : Optional[Any] = down_block_res_samples[-(self.layers_per_block + 1) :] SCREAMING_SNAKE_CASE : Optional[int] = down_block_res_samples[: -(self.layers_per_block + 1)] if isinstance(_lowerCamelCase , _lowerCamelCase ): SCREAMING_SNAKE_CASE : str = up_block( _lowerCamelCase , temb=_lowerCamelCase , encoder_hidden_states=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train , ) else: SCREAMING_SNAKE_CASE : Optional[int] = up_block(_lowerCamelCase , temb=_lowerCamelCase , res_hidden_states_tuple=_lowerCamelCase , deterministic=not train ) # 6. post-process SCREAMING_SNAKE_CASE : Optional[int] = self.conv_norm_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[int] = nn.silu(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Dict = self.conv_out(_lowerCamelCase ) SCREAMING_SNAKE_CASE : Optional[Any] = jnp.transpose(_lowerCamelCase , (0, 3, 1, 2) ) if not return_dict: return (sample,) return FlaxUNetaDConditionOutput(sample=_lowerCamelCase )
313
1