code
stringlengths
81
54k
code_codestyle
int64
0
721
style_context
stringlengths
91
41.9k
style_context_codestyle
int64
0
699
label
int64
0
1
"""simple docstring""" from pathlib import Path import json import tempfile from transformers import FSMTTokenizer, FSMTConfig, FSMTForConditionalGeneration from transformers.models.fsmt.tokenization_fsmt import VOCAB_FILES_NAMES a_ = "tiny-wmt19-en-ru" # Build # borrowed from a test a_ = [ "l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "w</w>", "r</w>", "t</w>", "lo", "low", "er</w>", "low</w>", "lowest</w>", "newer</w>", "wider</w>", "<unk>", ] a_ = dict(zip(vocab, range(len(vocab)))) a_ = ["l o 123", "lo w 1456", "e r</w> 1789", ""] with tempfile.TemporaryDirectory() as tmpdirname: a_ = Path(tmpdirname) a_ = build_dir / VOCAB_FILES_NAMES["src_vocab_file"] a_ = build_dir / VOCAB_FILES_NAMES["tgt_vocab_file"] a_ = build_dir / VOCAB_FILES_NAMES["merges_file"] with open(src_vocab_file, "w") as fp: fp.write(json.dumps(vocab_tokens)) with open(tgt_vocab_file, "w") as fp: fp.write(json.dumps(vocab_tokens)) with open(merges_file, "w") as fp: fp.write("\n".join(merges)) a_ = FSMTTokenizer( langs=["en", "ru"], src_vocab_size=len(vocab), tgt_vocab_size=len(vocab), src_vocab_file=src_vocab_file, tgt_vocab_file=tgt_vocab_file, merges_file=merges_file, ) a_ = FSMTConfig( langs=["ru", "en"], src_vocab_size=10_00, tgt_vocab_size=10_00, d_model=4, encoder_layers=1, decoder_layers=1, encoder_ffn_dim=4, decoder_ffn_dim=4, encoder_attention_heads=1, decoder_attention_heads=1, ) a_ = FSMTForConditionalGeneration(config) print(f'''num of params {tiny_model.num_parameters()}''') # Test a_ = tokenizer(["Making tiny model"], return_tensors="pt") a_ = tiny_model(**batch) print("test output:", len(outputs.logits[0])) # Save tiny_model.half() # makes it smaller tiny_model.save_pretrained(mname_tiny) tokenizer.save_pretrained(mname_tiny) print(f'''Generated {mname_tiny}''') # Upload # transformers-cli upload tiny-wmt19-en-ru
719
"""simple docstring""" import os import re import warnings from shutil import copyfile from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import PreTrainedTokenizer if TYPE_CHECKING: from ...tokenization_utils_base import TextInput from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"vocab_file": "spiece.model"} a_ = { "vocab_file": { "t5-small": "https://huggingface.co/t5-small/resolve/main/spiece.model", "t5-base": "https://huggingface.co/t5-base/resolve/main/spiece.model", "t5-large": "https://huggingface.co/t5-large/resolve/main/spiece.model", "t5-3b": "https://huggingface.co/t5-3b/resolve/main/spiece.model", "t5-11b": "https://huggingface.co/t5-11b/resolve/main/spiece.model", } } # TODO(PVP) - this should be removed in Transformers v5 a_ = { "t5-small": 5_12, "t5-base": 5_12, "t5-large": 5_12, "t5-3b": 5_12, "t5-11b": 5_12, } a_ = "▁" class snake_case ( _UpperCamelCase): __UpperCamelCase = VOCAB_FILES_NAMES __UpperCamelCase = PRETRAINED_VOCAB_FILES_MAP __UpperCamelCase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __UpperCamelCase = ['input_ids', 'attention_mask'] def __init__( self : List[str] , a__ : Optional[int] , a__ : Union[str, Any]="</s>" , a__ : Union[str, Any]="<unk>" , a__ : str="<pad>" , a__ : Optional[int]=1_00 , a__ : List[Any]=None , a__ : Optional[Dict[str, Any]] = None , a__ : Any=True , **a__ : Optional[int] , ) -> None: '''simple docstring''' if extra_ids > 0 and additional_special_tokens is None: _A = [F"""<extra_id_{i}>""" for i in range(a__ )] elif extra_ids > 0 and additional_special_tokens is not None: # Check that we have the right number of extra_id special tokens _A = len(set(filter(lambda a__ : bool("extra_id" in str(a__ ) ) , a__ ) ) ) if extra_tokens != extra_ids: raise ValueError( F"""Both extra_ids ({extra_ids}) and additional_special_tokens ({additional_special_tokens}) are""" " provided to T5Tokenizer. In this case the additional_special_tokens must include the extra_ids" " tokens" ) if legacy: logger.warning_once( F"""You are using the legacy behaviour of the {self.__class__}. This means that tokens that come after special tokens will not be properly handled. We recommend you to""" " read the related pull request available at https://github.com/huggingface/transformers/pull/24565" ) _A = legacy _A = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( eos_token=a__ , unk_token=a__ , pad_token=a__ , extra_ids=a__ , additional_special_tokens=a__ , sp_model_kwargs=self.sp_model_kwargs , legacy=a__ , **a__ , ) _A = vocab_file _A = extra_ids _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(a__ ) @staticmethod def a_ ( a__ : List[str] , a__ : Optional[int] , a__ : Tuple ) -> Tuple: '''simple docstring''' if pretrained_model_name_or_path in TaTokenizer.max_model_input_sizes: _A = TaTokenizer.max_model_input_sizes[pretrained_model_name_or_path] if init_max_model_length is not None and init_max_model_length != max_model_length: return init_max_model_length elif init_max_model_length is None: warnings.warn( "This tokenizer was incorrectly instantiated with a model max length of" F""" {deprecated_max_model_length} which will be corrected in Transformers v5.\nFor now, this""" " behavior is kept to avoid breaking backwards compatibility when padding/encoding with" " `truncation is True`.\n- Be aware that you SHOULD NOT rely on" F""" {pretrained_model_name_or_path} automatically truncating your input to""" F""" {deprecated_max_model_length} when padding/encoding.\n- If you want to encode/pad to sequences""" F""" longer than {deprecated_max_model_length} you can either instantiate this tokenizer with""" " `model_max_length` or pass `max_length` when encoding/padding.\n- To avoid this warning, please" " instantiate this tokenizer with `model_max_length` set to your preferred value." , a__ , ) return max_model_length @property def a_ ( self : List[Any] ) -> Dict: '''simple docstring''' return self.sp_model.get_piece_size() + self._extra_ids def a_ ( self : Dict ) -> Optional[Any]: '''simple docstring''' _A = {self.convert_ids_to_tokens(a__ ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def a_ ( self : Optional[Any] , a__ : List[int] , a__ : Optional[List[int]] = None , a__ : bool = False ) -> List[int]: '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=a__ , token_ids_a=a__ , already_has_special_tokens=a__ ) # normal case: some special tokens if token_ids_a is None: return ([0] * len(a__ )) + [1] return ([0] * len(a__ )) + [1] + ([0] * len(a__ )) + [1] def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' return list( set(filter(lambda a__ : bool(re.search(r"<extra_id_\d+>" , a__ ) ) is not None , self.additional_special_tokens ) ) ) def a_ ( self : str ) -> List[Any]: '''simple docstring''' return [self._convert_token_to_id(a__ ) for token in self.get_sentinel_tokens()] def a_ ( self : List[Any] , a__ : List[int] ) -> List[int]: '''simple docstring''' if len(a__ ) > 0 and token_ids[-1] == self.eos_token_id: warnings.warn( F"""This sequence already has {self.eos_token}. In future versions this behavior may lead to duplicated""" " eos tokens being added." ) return token_ids else: return token_ids + [self.eos_token_id] def a_ ( self : int , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = [self.eos_token_id] if token_ids_a is None: return len(token_ids_a + eos ) * [0] return len(token_ids_a + eos + token_ids_a + eos ) * [0] def a_ ( self : Union[str, Any] , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = self._add_eos_if_not_present(a__ ) if token_ids_a is None: return token_ids_a else: _A = self._add_eos_if_not_present(a__ ) return token_ids_a + token_ids_a def __getstate__( self : Dict ) -> Union[str, Any]: '''simple docstring''' _A = self.__dict__.copy() _A = None return state def __setstate__( self : int , a__ : Optional[int] ) -> Union[str, Any]: '''simple docstring''' _A = d # for backward compatibility if not hasattr(self , "sp_model_kwargs" ): _A = {} _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def a_ ( self : int , a__ : "TextInput" , **a__ : List[str] ) -> List[str]: '''simple docstring''' if not self.legacy: _A = SPIECE_UNDERLINE + text.replace(a__ , " " ) return super().tokenize(a__ , **a__ ) def a_ ( self : str , a__ : Dict , **a__ : Optional[int] ) -> Any: '''simple docstring''' if not self.legacy: _A = text.startswith(a__ ) if is_first: _A = text[1:] _A = self.sp_model.encode(a__ , out_type=a__ ) if not self.legacy and not is_first and not text.startswith(" " ) and tokens[0].startswith(a__ ): _A = ([tokens[0][1:]] if len(tokens[0] ) > 1 else []) + tokens[1:] return tokens def a_ ( self : int , a__ : List[Any] ) -> List[str]: '''simple docstring''' if token.startswith("<extra_id_" ): _A = re.match(r"<extra_id_(\d+)>" , a__ ) _A = int(match.group(1 ) ) return self.vocab_size - num - 1 return self.sp_model.piece_to_id(a__ ) def a_ ( self : Dict , a__ : Union[str, Any] ) -> Any: '''simple docstring''' if index < self.sp_model.get_piece_size(): _A = self.sp_model.IdToPiece(a__ ) else: _A = F"""<extra_id_{self.vocab_size - 1 - index}>""" return token def a_ ( self : Optional[int] , a__ : Tuple ) -> List[str]: '''simple docstring''' _A = [] _A = "" _A = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(a__ ) + token _A = True _A = [] else: current_sub_tokens.append(a__ ) _A = False out_string += self.sp_model.decode(a__ ) return out_string.strip() def a_ ( self : Dict , a__ : str , a__ : Optional[str] = None ) -> Tuple[str]: '''simple docstring''' if not os.path.isdir(a__ ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return _A = os.path.join( a__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(a__ ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , a__ ) elif not os.path.isfile(self.vocab_file ): with open(a__ , "wb" ) as fi: _A = self.sp_model.serialized_model_proto() fi.write(a__ ) return (out_vocab_file,)
621
0
"""simple docstring""" import inspect import jax import jax.lax as lax import jax.numpy as jnp from ..utils import add_start_docstrings from ..utils.logging import get_logger a_ = get_logger(__name__) a_ = R"\n Args:\n input_ids (`jnp.ndarray` of shape `(batch_size, sequence_length)`):\n Indices of input sequence tokens in the vocabulary.\n\n Indices can be obtained using [`PreTrainedTokenizer`]. See [`PreTrainedTokenizer.encode`] and\n [`PreTrainedTokenizer.__call__`] for details.\n\n [What are input IDs?](../glossary#input-ids)\n scores (`jnp.ndarray` of shape `(batch_size, config.vocab_size)`):\n Prediction scores of a language modeling head. These can be logits for each vocabulary when not using beam\n search or log softmax for each vocabulary token when using beam search\n kwargs (`Dict[str, Any]`, *optional*):\n Additional logits processor specific kwargs.\n\n Return:\n `jnp.ndarray` of shape `(batch_size, config.vocab_size)`: The processed prediction scores.\n\n" class snake_case : @add_start_docstrings(a__ ) def __call__( self : Tuple , a__ : jnp.ndarray , a__ : jnp.ndarray ) -> jnp.ndarray: '''simple docstring''' raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) class snake_case : @add_start_docstrings(a__ ) def __call__( self : Any , a__ : jnp.ndarray , a__ : jnp.ndarray ) -> jnp.ndarray: '''simple docstring''' raise NotImplementedError( F"""{self.__class__} is an abstract class. Only classes inheriting this class can be called.""" ) class snake_case ( _UpperCamelCase): @add_start_docstrings(a__ ) def __call__( self : Optional[int] , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int , **a__ : str ) -> jnp.ndarray: '''simple docstring''' for processor in self: _A = inspect.signature(processor.__call__ ).parameters if len(a__ ) > 3: if not all(arg in kwargs for arg in list(function_args.keys() )[2:] ): raise ValueError( F"""Make sure that all the required parameters: {list(function_args.keys() )} for """ F"""{processor.__class__} are passed to the logits processor.""" ) _A = processor(a__ , a__ , a__ , **a__ ) else: _A = processor(a__ , a__ , a__ ) return scores class snake_case ( _UpperCamelCase): def __init__( self : Optional[Any] , a__ : float ) -> int: '''simple docstring''' if not isinstance(a__ , a__ ) or not (temperature > 0): raise ValueError(F"""`temperature` has to be a strictly positive float, but is {temperature}""" ) _A = temperature def __call__( self : str , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' _A = scores / self.temperature return scores class snake_case ( _UpperCamelCase): def __init__( self : Any , a__ : float , a__ : float = -float("Inf" ) , a__ : int = 1 ) -> List[Any]: '''simple docstring''' if not isinstance(a__ , a__ ) or (top_p < 0 or top_p > 1.0): raise ValueError(F"""`top_p` has to be a float > 0 and < 1, but is {top_p}""" ) if not isinstance(a__ , a__ ) or (min_tokens_to_keep < 1): raise ValueError(F"""`min_tokens_to_keep` has to be a positive integer, but is {min_tokens_to_keep}""" ) _A = top_p _A = filter_value _A = min_tokens_to_keep def __call__( self : str , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' _A , _A = lax.top_k(a__ , scores.shape[-1] ) _A = jnp.full_like(a__ , self.filter_value ) _A = jax.nn.softmax(a__ , axis=-1 ).cumsum(axis=-1 ) _A = cumulative_probs < self.top_p # include the token that is higher than top_p as well _A = jnp.roll(a__ , 1 ) score_mask |= score_mask.at[:, 0].set(a__ ) # min tokens to keep _A = score_mask.at[:, : self.min_tokens_to_keep].set(a__ ) _A = jnp.where(a__ , a__ , a__ ) _A = jax.lax.sort_key_val(a__ , a__ )[-1] return next_scores class snake_case ( _UpperCamelCase): def __init__( self : Union[str, Any] , a__ : int , a__ : float = -float("Inf" ) , a__ : int = 1 ) -> Optional[Any]: '''simple docstring''' if not isinstance(a__ , a__ ) or top_k <= 0: raise ValueError(F"""`top_k` has to be a strictly positive integer, but is {top_k}""" ) _A = max(a__ , a__ ) _A = filter_value def __call__( self : Union[str, Any] , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' _A , _A = scores.shape _A = jnp.full(batch_size * vocab_size , self.filter_value ) _A = min(self.top_k , scores.shape[-1] ) # Safety check _A , _A = lax.top_k(a__ , a__ ) _A = jnp.broadcast_to((jnp.arange(a__ ) * vocab_size)[:, None] , (batch_size, topk) ).flatten() _A = topk_scores.flatten() _A = topk_indices.flatten() + shift _A = next_scores_flat.at[topk_indices_flat].set(a__ ) _A = next_scores_flat.reshape(a__ , a__ ) return next_scores class snake_case ( _UpperCamelCase): def __init__( self : Any , a__ : int ) -> Any: '''simple docstring''' _A = bos_token_id def __call__( self : str , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' _A = jnp.full(scores.shape , -float("inf" ) ) _A = 1 - jnp.bool_(cur_len - 1 ) _A = jnp.where(a__ , new_scores.at[:, self.bos_token_id].set(0 ) , a__ ) return scores class snake_case ( _UpperCamelCase): def __init__( self : int , a__ : int , a__ : int ) -> Dict: '''simple docstring''' _A = max_length _A = eos_token_id def __call__( self : Any , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' _A = jnp.full(scores.shape , -float("inf" ) ) _A = 1 - jnp.bool_(cur_len - self.max_length + 1 ) _A = jnp.where(a__ , new_scores.at[:, self.eos_token_id].set(0 ) , a__ ) return scores class snake_case ( _UpperCamelCase): def __init__( self : Any , a__ : int , a__ : int ) -> Union[str, Any]: '''simple docstring''' if not isinstance(a__ , a__ ) or min_length < 0: raise ValueError(F"""`min_length` has to be a positive integer, but is {min_length}""" ) if not isinstance(a__ , a__ ) or eos_token_id < 0: raise ValueError(F"""`eos_token_id` has to be a positive integer, but is {eos_token_id}""" ) _A = min_length _A = eos_token_id def __call__( self : int , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' _A = 1 - jnp.clip(cur_len - self.min_length , 0 , 1 ) _A = jnp.where(a__ , scores.at[:, self.eos_token_id].set(-float("inf" ) ) , a__ ) return scores class snake_case ( _UpperCamelCase): def __init__( self : Any , a__ : int , a__ : str ) -> List[Any]: '''simple docstring''' _A = list(a__ ) _A = begin_index def __call__( self : List[Any] , a__ : Tuple , a__ : Tuple , a__ : int ) -> Dict: '''simple docstring''' _A = 1 - jnp.bool_(cur_len - self.begin_index ) _A = jnp.where(a__ , scores.at[:, self.begin_suppress_tokens].set(-float("inf" ) ) , a__ ) return scores class snake_case ( _UpperCamelCase): def __init__( self : List[str] , a__ : list ) -> Union[str, Any]: '''simple docstring''' _A = list(a__ ) def __call__( self : List[Any] , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' _A = scores.at[..., self.suppress_tokens].set(-float("inf" ) ) return scores class snake_case ( _UpperCamelCase): def __init__( self : int , a__ : Optional[int] ) -> Optional[int]: '''simple docstring''' _A = dict(a__ ) # Converts the dictionary of format {index: token} containing the tokens to be forced to an array, where the # index of the array corresponds to the index of the token to be forced, for XLA compatibility. # Indexes without forced tokens will have a negative value. _A = jnp.ones((max(force_token_map.keys() ) + 1) , dtype=jnp.intaa ) * -1 for index, token in force_token_map.items(): if token is not None: _A = force_token_array.at[index].set(a__ ) _A = jnp.intaa(a__ ) def __call__( self : Union[str, Any] , a__ : jnp.ndarray , a__ : jnp.ndarray , a__ : int ) -> jnp.ndarray: '''simple docstring''' def _force_token(a__ : Tuple ): _A = scores.shape[0] _A = self.force_token_array[generation_idx] _A = jnp.ones_like(a__ , dtype=scores.dtype ) * -float("inf" ) _A = jnp.zeros((batch_size, 1) , dtype=scores.dtype ) _A = lax.dynamic_update_slice(a__ , a__ , (0, current_token) ) return new_scores _A = lax.cond( cur_len >= self.force_token_array.shape[0] , lambda: scores , lambda: lax.cond( self.force_token_array[cur_len] >= 0 , lambda: _force_token(a__ ) , lambda: scores , ) , ) return scores class snake_case ( _UpperCamelCase): def __init__( self : int , a__ : Any , a__ : Dict , a__ : List[str] ) -> str: '''simple docstring''' _A = generate_config.eos_token_id _A = generate_config.no_timestamps_token_id _A = generate_config.no_timestamps_token_id + 1 _A = decoder_input_length + 1 if generate_config.is_multilingual: # room for language token and task token self.begin_index += 2 if hasattr(a__ , "max_initial_timestamp_index" ): _A = generate_config.max_initial_timestamp_index else: _A = model_config.vocab_size if self.max_initial_timestamp_index is None: _A = model_config.vocab_size def __call__( self : Dict , a__ : Any , a__ : int , a__ : List[str] ) -> Dict: '''simple docstring''' _A = scores.at[:, self.no_timestamps_token_id].set(-float("inf" ) ) def handle_pairs(a__ : Optional[Any] , a__ : Dict ): _A = jnp.where((cur_len - self.begin_index) >= 1 , a__ , a__ ) _A = jnp.where( input_ids_k[cur_len - 1] >= self.timestamp_begin , True and last_was_timestamp , a__ , ) _A = jnp.where((cur_len - self.begin_index) < 2 , a__ , a__ ) _A = jnp.where( input_ids_k[cur_len - 2] >= self.timestamp_begin , a__ , a__ , ) return jnp.where( a__ , jnp.where( penultimate_was_timestamp > 0 , scores_k.at[self.timestamp_begin :].set(-float("inf" ) ) , scores_k.at[: self.eos_token_id].set(-float("inf" ) ) , ) , a__ , ) _A = jax.vmap(a__ )(a__ , a__ ) _A = jnp.where(cur_len == self.begin_index , a__ , a__ ) _A = jnp.where( self.max_initial_timestamp_index is not None , True and apply_max_initial_timestamp , a__ , ) _A = self.timestamp_begin + self.max_initial_timestamp_index _A = jnp.where( a__ , scores.at[:, last_allowed + 1 :].set(-float("inf" ) ) , a__ , ) # if sum of probability over timestamps is above any other token, sample timestamp _A = jax.nn.log_softmax(a__ , axis=-1 ) def handle_cumulative_probs(a__ : int , a__ : Tuple ): _A = jax.nn.logsumexp(logprobs_k[self.timestamp_begin :] , axis=-1 ) _A = jnp.max(logprobs_k[: self.timestamp_begin] ) return jnp.where( timestamp_logprob > max_text_token_logprob , scores_k.at[: self.timestamp_begin].set(-float("inf" ) ) , a__ , ) _A = jax.vmap(a__ )(a__ , a__ ) return scores
720
"""simple docstring""" import argparse import json import os from collections import OrderedDict import numpy as np import tensorflow as tf import torch def a__ ( __lowercase ) -> List[Any]: _A = os.path.join(args.tf_model_dir , "parameters.json" ) _A = json.loads(open(__lowercase ).read() ) if not params: raise ValueError( f"""It seems that the json file at {parameter_file} is empty. Make sure you have a correct json file.""" ) if not args.output.endswith(".pt" ): _A = args.output + ".pt" _A = OrderedDict() with tf.device("/CPU:0" ): _A = tf.train.load_checkpoint(args.tf_model_dir ) _A = reader.get_variable_to_shape_map() for key_name in shapes.keys(): _A = reader.get_tensor(__lowercase ).astype(np.floataa ) if key_name.endswith("/adam_m" ) or key_name.endswith("/adam_v" ): continue if key_name.startswith("pasts/" ): if key_name.startswith("pasts/mlp" ): _A = int(key_name[9] ) elif key_name.startswith("pasts/out" ): _A = 8 _A = "model.sqout.%d.weight" % (player * 2) # enter to nn.Sequencial with Tanh, so 2 at a time _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/moe" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/switch_gating/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.router.classifier.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/softmlp/kernel" ): _A = "model.blocks.%d.feed_forward.soft_bypass_mlp.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/wo/kernel" ) or key_name.endswith("/wi/kernel" ): _A = key_name[-9:-7] for i in range(16 ): _A = "model.blocks.%d.feed_forward.mlp.experts.expert_%d.%s.weight" % (player, i, nlayer) _A = ( vnp[i].transpose([1, 0] ).copy() ) # In Mesh-Tensorflow, it is one array, so it is divided _A = torch.tensor(__lowercase ) elif key_name.startswith("model/mlp" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/p1/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wi.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p1/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wi.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wo.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wo.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/ln" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.feed_forward.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.feed_forward.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/att" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/qkv/kernel" ): _A = vnp.copy() # Compute same dimension as Mesh-tensorflow using einsum _A = state[:, 0, :, :] _A = state[:, 1, :, :] _A = state[:, 2, :, :] _A = ( state_q.reshape([state_q.shape[0], state_q.shape[1] * state_q.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_k.reshape([state_k.shape[0], state_k.shape[1] * state_k.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_v.reshape([state_v.shape[0], state_v.shape[1] * state_v.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = "model.blocks.%d.self_attn.self_attn.q_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.k_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.v_proj.weight" % player _A = torch.tensor(__lowercase ) elif key_name.endswith("/o/kernel" ): _A = "model.blocks.%d.self_attn.self_attn.out_proj.weight" % player _A = ( vnp.reshape([vnp.shape[0] * vnp.shape[1], vnp.shape[2]] ).transpose([1, 0] ).copy() ) # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/an" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.self_attn.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.self_attn.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif ( key_name.startswith("model/wte" ) or key_name.startswith("model/wpe" ) or key_name.startswith("model/ete" ) ): _A = {"wte": "embed_tokens", "wpe": "position_embeddings", "ete": "extra_position_embeddings"}[ key_name[-3:] ] _A = "model.%s.weight" % nlayer _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) if key_name.startswith("model/wte" ): _A = "lm_head.weight" _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) elif key_name.startswith("model/wob" ): _A = "final_logits_bias" _A = vnp.copy() # same in embedded _A = state.reshape((1, -1) ) _A = torch.tensor(__lowercase ) elif key_name == "model/dense/kernel": _A = "model.last_project.weight" _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name == "model/dense_1/bias": _A = "model.last_project.bias" _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) torch.save(__lowercase , args.output ) if __name__ == "__main__": a_ = argparse.ArgumentParser( description="model converter.", formatter_class=argparse.ArgumentDefaultsHelpFormatter ) parser.add_argument("--tf_model_dir", metavar="PATH", type=str, required=True, help="import model") parser.add_argument("--output", metavar="PATH", type=str, required=True, help="output model") a_ = parser.parse_args() convert_tf_gptsan_to_pt(args)
621
0
"""simple docstring""" from __future__ import annotations a_ = { "A": ["B", "C", "E"], "B": ["A", "D", "E"], "C": ["A", "F", "G"], "D": ["B"], "E": ["A", "B", "D"], "F": ["C"], "G": ["C"], } class snake_case : def __init__( self : Optional[int] , a__ : dict[str, list[str]] , a__ : str ) -> None: '''simple docstring''' _A = graph # mapping node to its parent in resulting breadth first tree _A = {} _A = source_vertex def a_ ( self : List[str] ) -> None: '''simple docstring''' _A = {self.source_vertex} _A = None _A = [self.source_vertex] # first in first out queue while queue: _A = queue.pop(0 ) for adjacent_vertex in self.graph[vertex]: if adjacent_vertex not in visited: visited.add(a__ ) _A = vertex queue.append(a__ ) def a_ ( self : List[Any] , a__ : str ) -> str: '''simple docstring''' if target_vertex == self.source_vertex: return self.source_vertex _A = self.parent.get(a__ ) if target_vertex_parent is None: _A = ( F"""No path from vertex: {self.source_vertex} to vertex: {target_vertex}""" ) raise ValueError(a__ ) return self.shortest_path(a__ ) + F"""->{target_vertex}""" if __name__ == "__main__": a_ = Graph(graph, "G") g.breath_first_search() print(g.shortest_path("D")) print(g.shortest_path("G")) print(g.shortest_path("Foo"))
721
"""simple docstring""" import argparse import torch from transformers import GPTaLMHeadModel, RobertaForMaskedLM if __name__ == "__main__": a_ = argparse.ArgumentParser( description=( "Extraction some layers of the full RobertaForMaskedLM or GPT2LMHeadModel for Transfer Learned" " Distillation" ) ) parser.add_argument("--model_type", default="roberta", choices=["roberta", "gpt2"]) parser.add_argument("--model_name", default="roberta-large", type=str) parser.add_argument("--dump_checkpoint", default="serialization_dir/tf_roberta_048131723.pth", type=str) parser.add_argument("--vocab_transform", action="store_true") a_ = parser.parse_args() if args.model_type == "roberta": a_ = RobertaForMaskedLM.from_pretrained(args.model_name) a_ = "roberta" elif args.model_type == "gpt2": a_ = GPTaLMHeadModel.from_pretrained(args.model_name) a_ = "transformer" a_ = model.state_dict() a_ = {} # Embeddings # if args.model_type == "gpt2": for param_name in ["wte.weight", "wpe.weight"]: a_ = state_dict[f'''{prefix}.{param_name}'''] else: for w in ["word_embeddings", "position_embeddings", "token_type_embeddings"]: a_ = f'''{prefix}.embeddings.{w}.weight''' a_ = state_dict[param_name] for w in ["weight", "bias"]: a_ = f'''{prefix}.embeddings.LayerNorm.{w}''' a_ = state_dict[param_name] # Transformer Blocks # a_ = 0 for teacher_idx in [0, 2, 4, 7, 9, 11]: if args.model_type == "gpt2": for layer in ["ln_1", "attn.c_attn", "attn.c_proj", "ln_2", "mlp.c_fc", "mlp.c_proj"]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.h.{teacher_idx}.{layer}.{w}''' ] a_ = state_dict[f'''{prefix}.h.{teacher_idx}.attn.bias'''] else: for layer in [ "attention.self.query", "attention.self.key", "attention.self.value", "attention.output.dense", "attention.output.LayerNorm", "intermediate.dense", "output.dense", "output.LayerNorm", ]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.encoder.layer.{teacher_idx}.{layer}.{w}''' ] std_idx += 1 # Language Modeling Head ###s if args.model_type == "roberta": for layer in ["lm_head.decoder.weight", "lm_head.bias"]: a_ = state_dict[f'''{layer}'''] if args.vocab_transform: for w in ["weight", "bias"]: a_ = state_dict[f'''lm_head.dense.{w}'''] a_ = state_dict[f'''lm_head.layer_norm.{w}'''] elif args.model_type == "gpt2": for w in ["weight", "bias"]: a_ = state_dict[f'''{prefix}.ln_f.{w}'''] a_ = state_dict["lm_head.weight"] print(f'''N layers selected for distillation: {std_idx}''') print(f'''Number of params transferred for distillation: {len(compressed_sd.keys())}''') print(f'''Save transferred checkpoint to {args.dump_checkpoint}.''') torch.save(compressed_sd, args.dump_checkpoint)
621
0
"""simple docstring""" import numpy as np from cva import destroyAllWindows, imread, imshow, waitKey class snake_case : def __init__( self : Union[str, Any] , a__ : List[str] , a__ : int , a__ : int ) -> List[str]: '''simple docstring''' if dst_width < 0 or dst_height < 0: raise ValueError("Destination width/height should be > 0" ) _A = img _A = img.shape[1] _A = img.shape[0] _A = dst_width _A = dst_height _A = self.src_w / self.dst_w _A = self.src_h / self.dst_h _A = _A = ( np.ones((self.dst_h, self.dst_w, 3) , np.uinta ) * 2_55 ) def a_ ( self : Dict ) -> Dict: '''simple docstring''' for i in range(self.dst_h ): for j in range(self.dst_w ): _A = self.img[self.get_y(a__ )][self.get_x(a__ )] def a_ ( self : Tuple , a__ : int ) -> int: '''simple docstring''' return int(self.ratio_x * x ) def a_ ( self : Any , a__ : int ) -> int: '''simple docstring''' return int(self.ratio_y * y ) if __name__ == "__main__": a_ , a_ = 8_00, 6_00 a_ = imread("image_data/lena.jpg", 1) a_ = NearestNeighbour(im, dst_w, dst_h) n.process() imshow( f'''Image resized from: {im.shape[1]}x{im.shape[0]} to {dst_w}x{dst_h}''', n.output ) waitKey(0) destroyAllWindows()
700
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a_ = { "configuration_upernet": ["UperNetConfig"], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "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_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
621
0
"""simple docstring""" class snake_case : def __init__( self : Tuple , a__ : int ) -> None: '''simple docstring''' _A = size _A = [0] * size _A = [0] * size @staticmethod def a_ ( a__ : int ) -> int: '''simple docstring''' return index | (index + 1) @staticmethod def a_ ( a__ : int ) -> int: '''simple docstring''' return (index & (index + 1)) - 1 def a_ ( self : Any , a__ : int , a__ : int ) -> None: '''simple docstring''' _A = value while index < self.size: _A = self.get_prev(a__ ) + 1 if current_left_border == index: _A = value else: _A = max(a__ , a__ , a__ ) _A = self.get_next(a__ ) def a_ ( self : int , a__ : int , a__ : int ) -> int: '''simple docstring''' right -= 1 # Because of right is exclusive _A = 0 while left <= right: _A = self.get_prev(a__ ) if left <= current_left: _A = max(a__ , self.tree[right] ) _A = current_left else: _A = max(a__ , self.arr[right] ) right -= 1 return result if __name__ == "__main__": import doctest doctest.testmod()
701
"""simple docstring""" import random import unittest from torch.utils.data import BatchSampler, DataLoader, IterableDataset from accelerate import Accelerator from accelerate.data_loader import ( BatchSamplerShard, DataLoaderDispatcher, DataLoaderShard, IterableDatasetShard, SkipBatchSampler, SkipDataLoader, skip_first_batches, ) class snake_case ( _UpperCamelCase): def __init__( self : Optional[int] , a__ : str=0.0_1 , a__ : str=10_00 ) -> int: '''simple docstring''' _A = p_stop _A = max_length def __iter__( self : Any ) -> Optional[Any]: '''simple docstring''' _A = 0 _A = False while not stop and count < self.max_length: yield count count += 1 _A = random.random() < self.p_stop class snake_case ( unittest.TestCase): def a_ ( self : List[Any] , a__ : Union[str, Any] , a__ : Union[str, Any] , a__ : List[str]=False , a__ : str=True ) -> Union[str, Any]: '''simple docstring''' _A = [ BatchSamplerShard(a__ , 2 , a__ , split_batches=a__ , even_batches=a__ ) for i in range(2 ) ] _A = [list(a__ ) for batch_sampler_shard in batch_sampler_shards] if not split_batches: self.assertListEqual([len(a__ ) for shard in batch_sampler_shards] , [len(a__ ) for e in expected] ) self.assertListEqual(a__ , a__ ) def a_ ( self : List[Any] ) -> str: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 22, 23]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is a round multiple of batch size but not total batch size. _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [0, 1, 2]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is not a round multiple of batch size but has a multiple of # num_processes batch. _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 0, 1]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is not a round multiple of batch size but and has not a multiple of # num_processes batch. _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 0]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [1, 2, 3]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[[0, 1, 0]], [[1, 0, 1]]] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ ) def a_ ( self : int ) -> int: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size. _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [0, 1]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size or num_processes. _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 0]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [1, 2]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[[0, 1]], [[0, 1]]] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is a round multiple of batch size but not total batch size. _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size but has a multiple of # num_processes batch. _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size but and has not a multiple of # num_processes batch. _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[[0, 1]], []] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) def a_ ( self : List[str] ) -> str: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size. _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size or num_processes. _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[[0, 1]], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) def a_ ( self : Union[str, Any] ) -> str: '''simple docstring''' _A = [[0, 1, 2], [3, 4], [5, 6, 7, 8], [9, 10, 11], [12, 13]] _A = [BatchSamplerShard(a__ , 2 , a__ , even_batches=a__ ) for i in range(2 )] self.assertEqual(len(batch_sampler_shards[0] ) , 3 ) self.assertEqual(len(batch_sampler_shards[1] ) , 2 ) self.assertListEqual(list(batch_sampler_shards[0] ) , [[0, 1, 2], [5, 6, 7, 8], [12, 13]] ) self.assertListEqual(list(batch_sampler_shards[1] ) , [[3, 4], [9, 10, 11]] ) def a_ ( self : Optional[int] , a__ : Optional[int] , a__ : Tuple , a__ : Optional[int] , a__ : Union[str, Any]=False , a__ : int=2 , a__ : List[Any]=False ) -> str: '''simple docstring''' random.seed(a__ ) _A = list(a__ ) _A = [ IterableDatasetShard( a__ , batch_size=a__ , drop_last=a__ , num_processes=a__ , process_index=a__ , split_batches=a__ , ) for i in range(a__ ) ] _A = [] for iterable_dataset_shard in iterable_dataset_shards: # Since our random iterable dataset will be... random... we need to use a seed to get reproducible results. random.seed(a__ ) iterable_dataset_lists.append(list(a__ ) ) _A = batch_size // num_processes if split_batches else batch_size # All iterable dataset shard should have the same length, a round multiple of shard_batch_size _A = iterable_dataset_lists[0] for l in iterable_dataset_lists[1:]: self.assertEqual(len(a__ ) , len(a__ ) ) self.assertTrue(len(a__ ) % shard_batch_size == 0 ) _A = [] for idx in range(0 , len(a__ ) , a__ ): for l in iterable_dataset_lists: observed += l[idx : idx + shard_batch_size] if not drop_last: while len(a__ ) < len(a__ ): reference += reference self.assertListEqual(a__ , reference[: len(a__ )] ) def a_ ( self : List[str] ) -> List[Any]: '''simple docstring''' _A = 42 _A = RandomIterableDataset() self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) # Edge case with a very small dataset _A = RandomIterableDataset(max_length=2 ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) def a_ ( self : List[str] ) -> Dict: '''simple docstring''' _A = BatchSampler(range(16 ) , batch_size=4 , drop_last=a__ ) _A = SkipBatchSampler(a__ , 2 ) self.assertListEqual(list(a__ ) , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : int ) -> Union[str, Any]: '''simple docstring''' _A = SkipDataLoader(list(range(16 ) ) , batch_size=4 , skip_batches=2 ) self.assertListEqual([t.tolist() for t in dataloader] , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : int ) -> Optional[int]: '''simple docstring''' _A = DataLoader(list(range(16 ) ) , batch_size=4 ) _A = skip_first_batches(a__ , num_batches=2 ) self.assertListEqual([t.tolist() for t in new_dataloader] , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : Optional[Any] ) -> Optional[int]: '''simple docstring''' _A = DataLoaderShard(list(range(16 ) ) , batch_size=4 ) for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) # Test it also works on the second iteration for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) def a_ ( self : int ) -> int: '''simple docstring''' Accelerator() _A = DataLoaderDispatcher(range(16 ) , batch_size=4 ) for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) # Test it also works on the second iteration for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 )
621
0
"""simple docstring""" from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { "naver-clova-ix/donut-base": "https://huggingface.co/naver-clova-ix/donut-base/resolve/main/config.json", # See all Donut models at https://huggingface.co/models?filter=donut-swin } class snake_case ( _UpperCamelCase): __UpperCamelCase = 'donut-swin' __UpperCamelCase = { 'num_attention_heads': 'num_heads', 'num_hidden_layers': 'num_layers', } def __init__( self : Union[str, Any] , a__ : Any=2_24 , a__ : Optional[int]=4 , a__ : Dict=3 , a__ : str=96 , a__ : Optional[Any]=[2, 2, 6, 2] , a__ : Optional[Any]=[3, 6, 12, 24] , a__ : Union[str, Any]=7 , a__ : Any=4.0 , a__ : Tuple=True , a__ : List[str]=0.0 , a__ : Dict=0.0 , a__ : Any=0.1 , a__ : Dict="gelu" , a__ : Optional[Any]=False , a__ : List[Any]=0.0_2 , a__ : str=1E-5 , **a__ : int , ) -> List[str]: '''simple docstring''' super().__init__(**a__ ) _A = image_size _A = patch_size _A = num_channels _A = embed_dim _A = depths _A = len(a__ ) _A = num_heads _A = window_size _A = mlp_ratio _A = qkv_bias _A = hidden_dropout_prob _A = attention_probs_dropout_prob _A = drop_path_rate _A = hidden_act _A = use_absolute_embeddings _A = layer_norm_eps _A = initializer_range # we set the hidden_size attribute in order to make Swin work with VisionEncoderDecoderModel # this indicates the channel dimension after the last stage of the model _A = int(embed_dim * 2 ** (len(a__ ) - 1) )
702
"""simple docstring""" import gc import tempfile import unittest import numpy as np import torch from diffusers import VersatileDiffusionPipeline from diffusers.utils.testing_utils import load_image, nightly, require_torch_gpu, torch_device a_ = False class snake_case ( unittest.TestCase): pass @nightly @require_torch_gpu class snake_case ( unittest.TestCase): def a_ ( self : Optional[int] ) -> str: '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def a_ ( self : Tuple ) -> Any: '''simple docstring''' _A = VersatileDiffusionPipeline.from_pretrained("shi-labs/versatile-diffusion" , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg" ) _A = torch.manual_seed(0 ) _A = pipe.dual_guided( prompt="first prompt" , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" , ).images with tempfile.TemporaryDirectory() as tmpdirname: pipe.save_pretrained(a__ ) _A = VersatileDiffusionPipeline.from_pretrained(a__ , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = generator.manual_seed(0 ) _A = pipe.dual_guided( prompt="first prompt" , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" , ).images assert np.abs(image - new_image ).sum() < 1E-5, "Models don't have the same forward pass" def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' _A = VersatileDiffusionPipeline.from_pretrained("shi-labs/versatile-diffusion" , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = "cyberpunk 2077" _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg" ) _A = torch.manual_seed(0 ) _A = pipe.dual_guided( prompt=a__ , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=50 , output_type="numpy" , ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.1_4_4_8, 0.1_6_1_9, 0.1_7_4_1, 0.1_0_8_6, 0.1_1_4_7, 0.1_1_2_8, 0.1_1_9_9, 0.1_1_6_5, 0.1_0_0_1] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 _A = "A painting of a squirrel eating a burger " _A = torch.manual_seed(0 ) _A = pipe.text_to_image( prompt=a__ , generator=a__ , guidance_scale=7.5 , num_inference_steps=50 , output_type="numpy" ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.3_3_6_7, 0.3_1_6_9, 0.2_6_5_6, 0.3_8_7_0, 0.4_7_9_0, 0.3_7_9_6, 0.4_0_0_9, 0.4_8_7_8, 0.4_7_7_8] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 _A = pipe.image_variation(a__ , generator=a__ , output_type="numpy" ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.3_0_7_6, 0.3_1_2_3, 0.3_2_8_4, 0.3_7_8_2, 0.3_7_7_0, 0.3_8_9_4, 0.4_2_9_7, 0.4_3_3_1, 0.4_4_5_6] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1
621
0
"""simple docstring""" from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = {} class snake_case ( _UpperCamelCase): __UpperCamelCase = 'llama' __UpperCamelCase = ['past_key_values'] def __init__( self : Tuple , a__ : List[Any]=3_20_00 , a__ : Optional[int]=40_96 , a__ : Optional[Any]=1_10_08 , a__ : Optional[int]=32 , a__ : Union[str, Any]=32 , a__ : Union[str, Any]=None , a__ : Optional[Any]="silu" , a__ : str=20_48 , a__ : Optional[int]=0.0_2 , a__ : List[Any]=1E-6 , a__ : Any=True , a__ : Optional[int]=0 , a__ : Optional[int]=1 , a__ : Any=2 , a__ : Optional[int]=1 , a__ : Optional[Any]=False , a__ : Dict=None , **a__ : str , ) -> Dict: '''simple docstring''' _A = vocab_size _A = max_position_embeddings _A = hidden_size _A = intermediate_size _A = num_hidden_layers _A = num_attention_heads # for backward compatibility if num_key_value_heads is None: _A = num_attention_heads _A = num_key_value_heads _A = hidden_act _A = initializer_range _A = rms_norm_eps _A = pretraining_tp _A = use_cache _A = rope_scaling self._rope_scaling_validation() super().__init__( pad_token_id=a__ , bos_token_id=a__ , eos_token_id=a__ , tie_word_embeddings=a__ , **a__ , ) def a_ ( self : Dict ) -> Tuple: '''simple docstring''' if self.rope_scaling is None: return if not isinstance(self.rope_scaling , a__ ) or len(self.rope_scaling ) != 2: raise ValueError( "`rope_scaling` must be a dictionary with with two fields, `name` and `factor`, " F"""got {self.rope_scaling}""" ) _A = self.rope_scaling.get("type" , a__ ) _A = self.rope_scaling.get("factor" , a__ ) if rope_scaling_type is None or rope_scaling_type not in ["linear", "dynamic"]: raise ValueError( F"""`rope_scaling`'s name field must be one of ['linear', 'dynamic'], got {rope_scaling_type}""" ) if rope_scaling_factor is None or not isinstance(a__ , a__ ) or rope_scaling_factor <= 1.0: raise ValueError(F"""`rope_scaling`'s factor field must be an float > 1, got {rope_scaling_factor}""" )
703
"""simple docstring""" import os import time import warnings from dataclasses import dataclass, field from enum import Enum from typing import List, Optional, Union import torch from filelock import FileLock from torch.utils.data import Dataset from ...tokenization_utils_base import PreTrainedTokenizerBase from ...utils import logging from ..processors.glue import glue_convert_examples_to_features, glue_output_modes, glue_processors from ..processors.utils import InputFeatures a_ = logging.get_logger(__name__) @dataclass class snake_case : __UpperCamelCase = field(metadata={'help': 'The name of the task to train on: ' + ', '.join(glue_processors.keys())}) __UpperCamelCase = field( metadata={'help': 'The input data dir. Should contain the .tsv files (or other data files) for the task.'}) __UpperCamelCase = field( default=128 , metadata={ 'help': ( 'The maximum total input sequence length after tokenization. Sequences longer ' 'than this will be truncated, sequences shorter will be padded.' ) } , ) __UpperCamelCase = field( default=_UpperCamelCase , metadata={'help': 'Overwrite the cached training and evaluation sets'}) def a_ ( self : Optional[Any] ) -> Tuple: '''simple docstring''' _A = self.task_name.lower() class snake_case ( _UpperCamelCase): __UpperCamelCase = 'train' __UpperCamelCase = 'dev' __UpperCamelCase = 'test' class snake_case ( _UpperCamelCase): __UpperCamelCase = 42 __UpperCamelCase = 42 __UpperCamelCase = 42 def __init__( self : Optional[int] , a__ : GlueDataTrainingArguments , a__ : PreTrainedTokenizerBase , a__ : Optional[int] = None , a__ : Union[str, Split] = Split.train , a__ : Optional[str] = None , ) -> Tuple: '''simple docstring''' warnings.warn( "This dataset will be removed from the library soon, preprocessing should be handled with the 🤗 Datasets " "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" , a__ , ) _A = args _A = glue_processors[args.task_name]() _A = glue_output_modes[args.task_name] if isinstance(a__ , a__ ): try: _A = Split[mode] except KeyError: raise KeyError("mode is not a valid split name" ) # Load data features from cache or dataset file _A = os.path.join( cache_dir if cache_dir is not None else args.data_dir , F"""cached_{mode.value}_{tokenizer.__class__.__name__}_{args.max_seq_length}_{args.task_name}""" , ) _A = self.processor.get_labels() if args.task_name in ["mnli", "mnli-mm"] and tokenizer.__class__.__name__ in ( "RobertaTokenizer", "RobertaTokenizerFast", "XLMRobertaTokenizer", "BartTokenizer", "BartTokenizerFast", ): # HACK(label indices are swapped in RoBERTa pretrained model) _A , _A = label_list[2], label_list[1] _A = label_list # Make sure only the first process in distributed training processes the dataset, # and the others will use the cache. _A = cached_features_file + ".lock" with FileLock(a__ ): if os.path.exists(a__ ) and not args.overwrite_cache: _A = time.time() _A = torch.load(a__ ) logger.info( F"""Loading features from cached file {cached_features_file} [took %.3f s]""" , time.time() - start ) else: logger.info(F"""Creating features from dataset file at {args.data_dir}""" ) if mode == Split.dev: _A = self.processor.get_dev_examples(args.data_dir ) elif mode == Split.test: _A = self.processor.get_test_examples(args.data_dir ) else: _A = self.processor.get_train_examples(args.data_dir ) if limit_length is not None: _A = examples[:limit_length] _A = glue_convert_examples_to_features( a__ , a__ , max_length=args.max_seq_length , label_list=a__ , output_mode=self.output_mode , ) _A = time.time() torch.save(self.features , a__ ) # ^ This seems to take a lot of time so I want to investigate why and how we can improve. logger.info( F"""Saving features into cached file {cached_features_file} [took {time.time() - start:.3f} s]""" ) def __len__( self : List[Any] ) -> Any: '''simple docstring''' return len(self.features ) def __getitem__( self : Tuple , a__ : Union[str, Any] ) -> InputFeatures: '''simple docstring''' return self.features[i] def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' return self.label_list
621
0
"""simple docstring""" from ...configuration_utils import PretrainedConfig from ...utils import logging __snake_case = logging.get_logger(__name__) __snake_case = { "microsoft/cvt-13": "https://huggingface.co/microsoft/cvt-13/resolve/main/config.json", # See all Cvt models at https://huggingface.co/models?filter=cvt } class snake_case ( _UpperCamelCase): __UpperCamelCase = 'cvt' def __init__( self : List[Any] , a__ : List[str]=3 , a__ : List[str]=[7, 3, 3] , a__ : List[Any]=[4, 2, 2] , a__ : str=[2, 1, 1] , a__ : str=[64, 1_92, 3_84] , a__ : Optional[int]=[1, 3, 6] , a__ : str=[1, 2, 10] , a__ : str=[4.0, 4.0, 4.0] , a__ : Optional[int]=[0.0, 0.0, 0.0] , a__ : Any=[0.0, 0.0, 0.0] , a__ : Tuple=[0.0, 0.0, 0.1] , a__ : Optional[int]=[True, True, True] , a__ : Optional[int]=[False, False, True] , a__ : Any=["dw_bn", "dw_bn", "dw_bn"] , a__ : int=[3, 3, 3] , a__ : Any=[1, 1, 1] , a__ : Any=[2, 2, 2] , a__ : List[Any]=[1, 1, 1] , a__ : Dict=[1, 1, 1] , a__ : Tuple=0.0_2 , a__ : str=1E-1_2 , **a__ : str , ) -> Optional[int]: '''simple docstring''' super().__init__(**a__ ) _A = num_channels _A = patch_sizes _A = patch_stride _A = patch_padding _A = embed_dim _A = num_heads _A = depth _A = mlp_ratio _A = attention_drop_rate _A = drop_rate _A = drop_path_rate _A = qkv_bias _A = cls_token _A = qkv_projection_method _A = kernel_qkv _A = padding_kv _A = stride_kv _A = padding_q _A = stride_q _A = initializer_range _A = layer_norm_eps
704
"""simple docstring""" def a__ ( __lowercase , __lowercase , __lowercase , __lowercase ) -> str: # Return True if there is node that has not iterated. _A = [False] * len(__lowercase ) _A = [] queue.append(__lowercase ) _A = True while queue: _A = queue.pop(0 ) for ind in range(len(graph[u] ) ): if visited[ind] is False and graph[u][ind] > 0: queue.append(__lowercase ) _A = True _A = u return visited[t] def a__ ( __lowercase , __lowercase , __lowercase ) -> int: # This array is filled by BFS and to store path _A = [-1] * (len(__lowercase )) _A = 0 while bfs(__lowercase , __lowercase , __lowercase , __lowercase ): _A = float("Inf" ) _A = sink while s != source: # Find the minimum value in select path _A = min(__lowercase , graph[parent[s]][s] ) _A = parent[s] max_flow += path_flow _A = sink while v != source: _A = parent[v] graph[u][v] -= path_flow graph[v][u] += path_flow _A = parent[v] return max_flow a_ = [ [0, 16, 13, 0, 0, 0], [0, 0, 10, 12, 0, 0], [0, 4, 0, 0, 14, 0], [0, 0, 9, 0, 0, 20], [0, 0, 0, 7, 0, 4], [0, 0, 0, 0, 0, 0], ] a_ , a_ = 0, 5 print(ford_fulkerson(graph, source, sink))
621
0
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> int: if len(__lowercase ) != len(__lowercase ): raise ValueError("String lengths must match!" ) _A = 0 for chara, chara in zip(__lowercase , __lowercase ): if chara != chara: count += 1 return count if __name__ == "__main__": import doctest doctest.testmod()
705
"""simple docstring""" import argparse import json from collections import OrderedDict from pathlib import Path import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ( ConditionalDetrConfig, ConditionalDetrForObjectDetection, ConditionalDetrForSegmentation, ConditionalDetrImageProcessor, ) from transformers.utils import logging logging.set_verbosity_info() a_ = logging.get_logger(__name__) # here we list all keys to be renamed (original name on the left, our name on the right) a_ = [] for i in range(6): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.weight''', f'''encoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.bias''', f'''encoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.weight''', f'''encoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.bias''', f'''encoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.weight''', f'''encoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.bias''', f'''encoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.encoder.layers.{i}.norm1.weight''', f'''encoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.norm1.bias''', f'''encoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.weight''', f'''encoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.bias''', f'''encoder.layers.{i}.final_layer_norm.bias''')) # decoder layers: 2 times output projection, 2 feedforward neural networks and 3 layernorms rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.weight''', f'''decoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.bias''', f'''decoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.weight''', f'''decoder.layers.{i}.encoder_attn.out_proj.weight''', ) ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.bias''', f'''decoder.layers.{i}.encoder_attn.out_proj.bias''', ) ) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.weight''', f'''decoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.bias''', f'''decoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.weight''', f'''decoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.bias''', f'''decoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm1.weight''', f'''decoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm1.bias''', f'''decoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.weight''', f'''decoder.layers.{i}.encoder_attn_layer_norm.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.bias''', f'''decoder.layers.{i}.encoder_attn_layer_norm.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.weight''', f'''decoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.bias''', f'''decoder.layers.{i}.final_layer_norm.bias''')) # q, k, v projections in self/cross-attention in decoder for conditional DETR rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.weight''', f'''decoder.layers.{i}.sa_qcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.weight''', f'''decoder.layers.{i}.sa_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qpos_proj.weight''', f'''decoder.layers.{i}.sa_qpos_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kpos_proj.weight''', f'''decoder.layers.{i}.sa_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.weight''', f'''decoder.layers.{i}.sa_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.weight''', f'''decoder.layers.{i}.ca_qcontent_proj.weight''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.weight", f"decoder.layers.{i}.ca_qpos_proj.weight")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.weight''', f'''decoder.layers.{i}.ca_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kpos_proj.weight''', f'''decoder.layers.{i}.ca_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.weight''', f'''decoder.layers.{i}.ca_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.weight''', f'''decoder.layers.{i}.ca_qpos_sine_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.bias''', f'''decoder.layers.{i}.sa_qcontent_proj.bias''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.bias''', f'''decoder.layers.{i}.sa_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_qpos_proj.bias''', f'''decoder.layers.{i}.sa_qpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_kpos_proj.bias''', f'''decoder.layers.{i}.sa_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.bias''', f'''decoder.layers.{i}.sa_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.bias''', f'''decoder.layers.{i}.ca_qcontent_proj.bias''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.bias", f"decoder.layers.{i}.ca_qpos_proj.bias")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.bias''', f'''decoder.layers.{i}.ca_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_kpos_proj.bias''', f'''decoder.layers.{i}.ca_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.bias''', f'''decoder.layers.{i}.ca_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.bias''', f'''decoder.layers.{i}.ca_qpos_sine_proj.bias''') ) # convolutional projection + query embeddings + layernorm of decoder + class and bounding box heads # for conditional DETR, also convert reference point head and query scale MLP rename_keys.extend( [ ("input_proj.weight", "input_projection.weight"), ("input_proj.bias", "input_projection.bias"), ("query_embed.weight", "query_position_embeddings.weight"), ("transformer.decoder.norm.weight", "decoder.layernorm.weight"), ("transformer.decoder.norm.bias", "decoder.layernorm.bias"), ("class_embed.weight", "class_labels_classifier.weight"), ("class_embed.bias", "class_labels_classifier.bias"), ("bbox_embed.layers.0.weight", "bbox_predictor.layers.0.weight"), ("bbox_embed.layers.0.bias", "bbox_predictor.layers.0.bias"), ("bbox_embed.layers.1.weight", "bbox_predictor.layers.1.weight"), ("bbox_embed.layers.1.bias", "bbox_predictor.layers.1.bias"), ("bbox_embed.layers.2.weight", "bbox_predictor.layers.2.weight"), ("bbox_embed.layers.2.bias", "bbox_predictor.layers.2.bias"), ("transformer.decoder.ref_point_head.layers.0.weight", "decoder.ref_point_head.layers.0.weight"), ("transformer.decoder.ref_point_head.layers.0.bias", "decoder.ref_point_head.layers.0.bias"), ("transformer.decoder.ref_point_head.layers.1.weight", "decoder.ref_point_head.layers.1.weight"), ("transformer.decoder.ref_point_head.layers.1.bias", "decoder.ref_point_head.layers.1.bias"), ("transformer.decoder.query_scale.layers.0.weight", "decoder.query_scale.layers.0.weight"), ("transformer.decoder.query_scale.layers.0.bias", "decoder.query_scale.layers.0.bias"), ("transformer.decoder.query_scale.layers.1.weight", "decoder.query_scale.layers.1.weight"), ("transformer.decoder.query_scale.layers.1.bias", "decoder.query_scale.layers.1.bias"), ("transformer.decoder.layers.0.ca_qpos_proj.weight", "decoder.layers.0.ca_qpos_proj.weight"), ("transformer.decoder.layers.0.ca_qpos_proj.bias", "decoder.layers.0.ca_qpos_proj.bias"), ] ) def a__ ( __lowercase , __lowercase , __lowercase ) -> List[str]: _A = state_dict.pop(__lowercase ) _A = val def a__ ( __lowercase ) -> List[str]: _A = OrderedDict() for key, value in state_dict.items(): if "backbone.0.body" in key: _A = key.replace("backbone.0.body" , "backbone.conv_encoder.model" ) _A = value else: _A = value return new_state_dict def a__ ( __lowercase , __lowercase=False ) -> Any: _A = "" if is_panoptic: _A = "conditional_detr." # first: transformer encoder for i in range(6 ): # read in weights + bias of input projection layer (in PyTorch's MultiHeadAttention, this is a single matrix + bias) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_weight""" ) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_bias""" ) # next, add query, keys and values (in that order) to the state dict _A = in_proj_weight[:256, :] _A = in_proj_bias[:256] _A = in_proj_weight[256:512, :] _A = in_proj_bias[256:512] _A = in_proj_weight[-256:, :] _A = in_proj_bias[-256:] def a__ ( ) -> int: _A = "http://images.cocodataset.org/val2017/000000039769.jpg" _A = Image.open(requests.get(__lowercase , stream=__lowercase ).raw ) return im @torch.no_grad() def a__ ( __lowercase , __lowercase ) -> Any: _A = ConditionalDetrConfig() # set backbone and dilation attributes if "resnet101" in model_name: _A = "resnet101" if "dc5" in model_name: _A = True _A = "panoptic" in model_name if is_panoptic: _A = 250 else: _A = 91 _A = "huggingface/label-files" _A = "coco-detection-id2label.json" _A = json.load(open(hf_hub_download(__lowercase , __lowercase , repo_type="dataset" ) , "r" ) ) _A = {int(__lowercase ): v for k, v in idalabel.items()} _A = idalabel _A = {v: k for k, v in idalabel.items()} # load image processor _A = "coco_panoptic" if is_panoptic else "coco_detection" _A = ConditionalDetrImageProcessor(format=__lowercase ) # prepare image _A = prepare_img() _A = image_processor(images=__lowercase , return_tensors="pt" ) _A = encoding["pixel_values"] logger.info(f"""Converting model {model_name}...""" ) # load original model from torch hub _A = torch.hub.load("DeppMeng/ConditionalDETR" , __lowercase , pretrained=__lowercase ).eval() _A = conditional_detr.state_dict() # rename keys for src, dest in rename_keys: if is_panoptic: _A = "conditional_detr." + src rename_key(__lowercase , __lowercase , __lowercase ) _A = rename_backbone_keys(__lowercase ) # query, key and value matrices need special treatment read_in_q_k_v(__lowercase , is_panoptic=__lowercase ) # important: we need to prepend a prefix to each of the base model keys as the head models use different attributes for them _A = "conditional_detr.model." if is_panoptic else "model." for key in state_dict.copy().keys(): if is_panoptic: if ( key.startswith("conditional_detr" ) and not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ) ): _A = state_dict.pop(__lowercase ) _A = val elif "class_labels_classifier" in key or "bbox_predictor" in key: _A = state_dict.pop(__lowercase ) _A = val elif key.startswith("bbox_attention" ) or key.startswith("mask_head" ): continue else: _A = state_dict.pop(__lowercase ) _A = val else: if not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ): _A = state_dict.pop(__lowercase ) _A = val # finally, create HuggingFace model and load state dict _A = ConditionalDetrForSegmentation(__lowercase ) if is_panoptic else ConditionalDetrForObjectDetection(__lowercase ) model.load_state_dict(__lowercase ) model.eval() model.push_to_hub(repo_id=__lowercase , organization="DepuMeng" , commit_message="Add model" ) # verify our conversion _A = conditional_detr(__lowercase ) _A = model(__lowercase ) assert torch.allclose(outputs.logits , original_outputs["pred_logits"] , atol=1E-4 ) assert torch.allclose(outputs.pred_boxes , original_outputs["pred_boxes"] , atol=1E-4 ) if is_panoptic: assert torch.allclose(outputs.pred_masks , original_outputs["pred_masks"] , atol=1E-4 ) # Save model and image processor logger.info(f"""Saving PyTorch model and image processor to {pytorch_dump_folder_path}...""" ) Path(__lowercase ).mkdir(exist_ok=__lowercase ) model.save_pretrained(__lowercase ) image_processor.save_pretrained(__lowercase ) if __name__ == "__main__": a_ = argparse.ArgumentParser() parser.add_argument( "--model_name", default="conditional_detr_resnet50", type=str, help="Name of the CONDITIONAL_DETR model you'd like to convert.", ) parser.add_argument( "--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model." ) a_ = parser.parse_args() convert_conditional_detr_checkpoint(args.model_name, args.pytorch_dump_folder_path)
621
0
"""simple docstring""" from collections import OrderedDict from typing import Mapping from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { "distilbert-base-uncased": "https://huggingface.co/distilbert-base-uncased/resolve/main/config.json", "distilbert-base-uncased-distilled-squad": ( "https://huggingface.co/distilbert-base-uncased-distilled-squad/resolve/main/config.json" ), "distilbert-base-cased": "https://huggingface.co/distilbert-base-cased/resolve/main/config.json", "distilbert-base-cased-distilled-squad": ( "https://huggingface.co/distilbert-base-cased-distilled-squad/resolve/main/config.json" ), "distilbert-base-german-cased": "https://huggingface.co/distilbert-base-german-cased/resolve/main/config.json", "distilbert-base-multilingual-cased": ( "https://huggingface.co/distilbert-base-multilingual-cased/resolve/main/config.json" ), "distilbert-base-uncased-finetuned-sst-2-english": ( "https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english/resolve/main/config.json" ), } class snake_case ( _UpperCamelCase): __UpperCamelCase = 'distilbert' __UpperCamelCase = { 'hidden_size': 'dim', 'num_attention_heads': 'n_heads', 'num_hidden_layers': 'n_layers', } def __init__( self : Any , a__ : Union[str, Any]=3_05_22 , a__ : Optional[Any]=5_12 , a__ : Union[str, Any]=False , a__ : Optional[Any]=6 , a__ : Any=12 , a__ : Union[str, Any]=7_68 , a__ : List[str]=4 * 7_68 , a__ : Any=0.1 , a__ : Dict=0.1 , a__ : List[Any]="gelu" , a__ : Dict=0.0_2 , a__ : int=0.1 , a__ : Optional[Any]=0.2 , a__ : Union[str, Any]=0 , **a__ : int , ) -> List[Any]: '''simple docstring''' _A = vocab_size _A = max_position_embeddings _A = sinusoidal_pos_embds _A = n_layers _A = n_heads _A = dim _A = hidden_dim _A = dropout _A = attention_dropout _A = activation _A = initializer_range _A = qa_dropout _A = seq_classif_dropout super().__init__(**a__ , pad_token_id=a__ ) class snake_case ( _UpperCamelCase): @property def a_ ( self : Optional[Any] ) -> Mapping[str, Mapping[int, str]]: '''simple docstring''' if self.task == "multiple-choice": _A = {0: "batch", 1: "choice", 2: "sequence"} else: _A = {0: "batch", 1: "sequence"} return OrderedDict( [ ("input_ids", dynamic_axis), ("attention_mask", dynamic_axis), ] )
706
"""simple docstring""" import random def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = a[left_index] _A = left_index + 1 for j in range(left_index + 1 , __lowercase ): if a[j] < pivot: _A , _A = a[i], a[j] i += 1 _A , _A = a[i - 1], a[left_index] return i - 1 def a__ ( __lowercase , __lowercase , __lowercase ) -> int: if left < right: _A = random.randint(__lowercase , right - 1 ) _A , _A = ( a[left], a[pivot], ) # switches the pivot with the left most bound _A = partition(__lowercase , __lowercase , __lowercase ) quick_sort_random( __lowercase , __lowercase , __lowercase ) # recursive quicksort to the left of the pivot point quick_sort_random( __lowercase , pivot_index + 1 , __lowercase ) # recursive quicksort to the right of the pivot point def a__ ( ) -> Dict: _A = input("Enter numbers separated by a comma:\n" ).strip() _A = [int(__lowercase ) for item in user_input.split("," )] quick_sort_random(__lowercase , 0 , len(__lowercase ) ) print(__lowercase ) if __name__ == "__main__": main()
621
0
"""simple docstring""" import argparse import json from collections import OrderedDict import torch from huggingface_hub import cached_download, hf_hub_url from transformers import AutoImageProcessor, CvtConfig, CvtForImageClassification def a__ ( __lowercase ) -> int: _A = [] embed.append( ( f"""cvt.encoder.stages.{idx}.embedding.convolution_embeddings.projection.weight""", f"""stage{idx}.patch_embed.proj.weight""", ) ) embed.append( ( f"""cvt.encoder.stages.{idx}.embedding.convolution_embeddings.projection.bias""", f"""stage{idx}.patch_embed.proj.bias""", ) ) embed.append( ( f"""cvt.encoder.stages.{idx}.embedding.convolution_embeddings.normalization.weight""", f"""stage{idx}.patch_embed.norm.weight""", ) ) embed.append( ( f"""cvt.encoder.stages.{idx}.embedding.convolution_embeddings.normalization.bias""", f"""stage{idx}.patch_embed.norm.bias""", ) ) return embed def a__ ( __lowercase , __lowercase ) -> str: _A = [] attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_query.convolution_projection.convolution.weight""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_q.conv.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_query.convolution_projection.normalization.weight""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_q.bn.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_query.convolution_projection.normalization.bias""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_q.bn.bias""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_query.convolution_projection.normalization.running_mean""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_q.bn.running_mean""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_query.convolution_projection.normalization.running_var""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_q.bn.running_var""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_query.convolution_projection.normalization.num_batches_tracked""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_q.bn.num_batches_tracked""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_key.convolution_projection.convolution.weight""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_k.conv.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_key.convolution_projection.normalization.weight""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_k.bn.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_key.convolution_projection.normalization.bias""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_k.bn.bias""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_key.convolution_projection.normalization.running_mean""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_k.bn.running_mean""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_key.convolution_projection.normalization.running_var""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_k.bn.running_var""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_key.convolution_projection.normalization.num_batches_tracked""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_k.bn.num_batches_tracked""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_value.convolution_projection.convolution.weight""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_v.conv.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_value.convolution_projection.normalization.weight""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_v.bn.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_value.convolution_projection.normalization.bias""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_v.bn.bias""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_value.convolution_projection.normalization.running_mean""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_v.bn.running_mean""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_value.convolution_projection.normalization.running_var""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_v.bn.running_var""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.convolution_projection_value.convolution_projection.normalization.num_batches_tracked""", f"""stage{idx}.blocks.{cnt}.attn.conv_proj_v.bn.num_batches_tracked""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.projection_query.weight""", f"""stage{idx}.blocks.{cnt}.attn.proj_q.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.projection_query.bias""", f"""stage{idx}.blocks.{cnt}.attn.proj_q.bias""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.projection_key.weight""", f"""stage{idx}.blocks.{cnt}.attn.proj_k.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.projection_key.bias""", f"""stage{idx}.blocks.{cnt}.attn.proj_k.bias""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.projection_value.weight""", f"""stage{idx}.blocks.{cnt}.attn.proj_v.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.attention.projection_value.bias""", f"""stage{idx}.blocks.{cnt}.attn.proj_v.bias""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.output.dense.weight""", f"""stage{idx}.blocks.{cnt}.attn.proj.weight""", ) ) attention_weights.append( ( f"""cvt.encoder.stages.{idx}.layers.{cnt}.attention.output.dense.bias""", f"""stage{idx}.blocks.{cnt}.attn.proj.bias""", ) ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.intermediate.dense.weight""", f"""stage{idx}.blocks.{cnt}.mlp.fc1.weight""") ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.intermediate.dense.bias""", f"""stage{idx}.blocks.{cnt}.mlp.fc1.bias""") ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.output.dense.weight""", f"""stage{idx}.blocks.{cnt}.mlp.fc2.weight""") ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.output.dense.bias""", f"""stage{idx}.blocks.{cnt}.mlp.fc2.bias""") ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.layernorm_before.weight""", f"""stage{idx}.blocks.{cnt}.norm1.weight""") ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.layernorm_before.bias""", f"""stage{idx}.blocks.{cnt}.norm1.bias""") ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.layernorm_after.weight""", f"""stage{idx}.blocks.{cnt}.norm2.weight""") ) attention_weights.append( (f"""cvt.encoder.stages.{idx}.layers.{cnt}.layernorm_after.bias""", f"""stage{idx}.blocks.{cnt}.norm2.bias""") ) return attention_weights def a__ ( __lowercase ) -> Dict: _A = [] token.append((f"""cvt.encoder.stages.{idx}.cls_token""", "stage2.cls_token") ) return token def a__ ( ) -> Dict: _A = [] head.append(("layernorm.weight", "norm.weight") ) head.append(("layernorm.bias", "norm.bias") ) head.append(("classifier.weight", "head.weight") ) head.append(("classifier.bias", "head.bias") ) return head def a__ ( __lowercase , __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = "imagenet-1k-id2label.json" _A = 1000 _A = "huggingface/label-files" _A = num_labels _A = json.load(open(cached_download(hf_hub_url(__lowercase , __lowercase , repo_type="dataset" ) ) , "r" ) ) _A = {int(__lowercase ): v for k, v in idalabel.items()} _A = idalabel _A = {v: k for k, v in idalabel.items()} _A = _A = CvtConfig(num_labels=__lowercase , idalabel=__lowercase , labelaid=__lowercase ) # For depth size 13 (13 = 1+2+10) if cvt_model.rsplit("/" , 1 )[-1][4:6] == "13": _A = [1, 2, 10] # For depth size 21 (21 = 1+4+16) elif cvt_model.rsplit("/" , 1 )[-1][4:6] == "21": _A = [1, 4, 16] # For wide cvt (similar to wide-resnet) depth size 24 (w24 = 2 + 2 20) else: _A = [2, 2, 20] _A = [3, 12, 16] _A = [192, 768, 1024] _A = CvtForImageClassification(__lowercase ) _A = AutoImageProcessor.from_pretrained("facebook/convnext-base-224-22k-1k" ) _A = image_size _A = torch.load(__lowercase , map_location=torch.device("cpu" ) ) _A = OrderedDict() _A = [] for idx in range(len(config.depth ) ): if config.cls_token[idx]: _A = list_of_state_dict + cls_token(__lowercase ) _A = list_of_state_dict + embeddings(__lowercase ) for cnt in range(config.depth[idx] ): _A = list_of_state_dict + attention(__lowercase , __lowercase ) _A = list_of_state_dict + final() for gg in list_of_state_dict: print(__lowercase ) for i in range(len(__lowercase ) ): _A = original_weights[list_of_state_dict[i][1]] model.load_state_dict(__lowercase ) model.save_pretrained(__lowercase ) image_processor.save_pretrained(__lowercase ) # Download the weights from zoo: https://1drv.ms/u/s!AhIXJn_J-blW9RzF3rMW7SsLHa8h?e=blQ0Al if __name__ == "__main__": a_ = argparse.ArgumentParser() parser.add_argument( "--cvt_model", default="cvt-w24", type=str, help="Name of the cvt model you'd like to convert.", ) parser.add_argument( "--image_size", default=3_84, type=int, help="Input Image Size", ) parser.add_argument( "--cvt_file_name", default=r"cvtmodels\CvT-w24-384x384-IN-22k.pth", type=str, help="Input Image Size", ) parser.add_argument( "--pytorch_dump_folder_path", default=None, type=str, help="Path to the output PyTorch model directory." ) a_ = parser.parse_args() convert_cvt_checkpoint(args.cvt_model, args.image_size, args.cvt_file_name, args.pytorch_dump_folder_path)
707
"""simple docstring""" import copy from typing import Any, Dict, List, Optional, Union import numpy as np 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_ = logging.get_logger(__name__) class snake_case ( _UpperCamelCase): __UpperCamelCase = ['input_features'] def __init__( self : int , a__ : Optional[Any]=80 , a__ : Optional[int]=1_60_00 , a__ : int=1_60 , a__ : Union[str, Any]=30 , a__ : Tuple=4_00 , a__ : List[Any]=0.0 , a__ : Optional[Any]=False , **a__ : List[Any] , ) -> str: '''simple docstring''' super().__init__( feature_size=a__ , sampling_rate=a__ , padding_value=a__ , return_attention_mask=a__ , **a__ , ) _A = n_fft _A = hop_length _A = chunk_length _A = chunk_length * sampling_rate _A = self.n_samples // hop_length _A = sampling_rate _A = mel_filter_bank( num_frequency_bins=1 + n_fft // 2 , num_mel_filters=a__ , min_frequency=0.0 , max_frequency=8_0_0_0.0 , sampling_rate=a__ , norm="slaney" , mel_scale="slaney" , ) def a_ ( self : int , a__ : np.array ) -> np.ndarray: '''simple docstring''' _A = spectrogram( a__ , window_function(self.n_fft , "hann" ) , frame_length=self.n_fft , hop_length=self.hop_length , power=2.0 , mel_filters=self.mel_filters , log_mel="log10" , ) _A = log_spec[:, :-1] _A = np.maximum(a__ , log_spec.max() - 8.0 ) _A = (log_spec + 4.0) / 4.0 return log_spec @staticmethod # Copied from transformers.models.wav2vec2.feature_extraction_wav2vec2.Wav2Vec2FeatureExtractor.zero_mean_unit_var_norm def a_ ( a__ : List[np.ndarray] , a__ : List[np.ndarray] , a__ : float = 0.0 ) -> List[np.ndarray]: '''simple docstring''' if attention_mask is not None: _A = np.array(a__ , np.intaa ) _A = [] for vector, length in zip(a__ , attention_mask.sum(-1 ) ): _A = (vector - vector[:length].mean()) / np.sqrt(vector[:length].var() + 1E-7 ) if length < normed_slice.shape[0]: _A = padding_value normed_input_values.append(a__ ) else: _A = [(x - x.mean()) / np.sqrt(x.var() + 1E-7 ) for x in input_values] return normed_input_values def __call__( self : Optional[int] , a__ : Union[np.ndarray, List[float], List[np.ndarray], List[List[float]]] , a__ : bool = True , a__ : Optional[int] = None , a__ : Optional[Union[str, TensorType]] = None , a__ : Optional[bool] = None , a__ : Optional[str] = "max_length" , a__ : Optional[int] = None , a__ : Optional[int] = None , a__ : Optional[bool] = None , **a__ : Dict , ) -> BatchFeature: '''simple docstring''' 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." ) _A = isinstance(a__ , 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}""" ) _A = is_batched_numpy or ( isinstance(a__ , (list, tuple) ) and (isinstance(raw_speech[0] , (np.ndarray, tuple, list) )) ) if is_batched: _A = [np.asarray([speech] , dtype=np.floataa ).T for speech in raw_speech] elif not is_batched and not isinstance(a__ , np.ndarray ): _A = np.asarray(a__ , dtype=np.floataa ) elif isinstance(a__ , np.ndarray ) and raw_speech.dtype is np.dtype(np.floataa ): _A = raw_speech.astype(np.floataa ) # always return batch if not is_batched: _A = [np.asarray([raw_speech] ).T] _A = BatchFeature({"input_features": raw_speech} ) # convert into correct format for padding _A = self.pad( a__ , padding=a__ , max_length=max_length if max_length else self.n_samples , truncation=a__ , pad_to_multiple_of=a__ , return_attention_mask=return_attention_mask or do_normalize , ) # zero-mean and unit-variance normalization if do_normalize: _A = self.zero_mean_unit_var_norm( padded_inputs["input_features"] , attention_mask=padded_inputs["attention_mask"] , padding_value=self.padding_value , ) _A = np.stack(padded_inputs["input_features"] , axis=0 ) # make sure list is in array format _A = padded_inputs.get("input_features" ).transpose(2 , 0 , 1 ) _A = [self._np_extract_fbank_features(a__ ) for waveform in input_features[0]] if isinstance(input_features[0] , a__ ): _A = [np.asarray(a__ , dtype=np.floataa ) for feature in input_features] else: _A = input_features if return_attention_mask: # rescale from sample (48000) to feature (3000) _A = padded_inputs["attention_mask"][:, :: self.hop_length] if return_tensors is not None: _A = padded_inputs.convert_to_tensors(a__ ) return padded_inputs def a_ ( self : Dict ) -> Dict[str, Any]: '''simple docstring''' _A = copy.deepcopy(self.__dict__ ) _A = self.__class__.__name__ if "mel_filters" in output: del output["mel_filters"] return output
621
0
"""simple docstring""" import warnings from diffusers import StableDiffusionImgaImgPipeline # noqa F401 warnings.warn( "The `image_to_image.py` script is outdated. Please use directly `from diffusers import" " StableDiffusionImg2ImgPipeline` instead." )
708
"""simple docstring""" from __future__ import annotations def a__ ( __lowercase , __lowercase ) -> float: _A = sorted(numsa + numsa ) _A , _A = divmod(len(__lowercase ) , 2 ) if mod == 1: return all_numbers[div] else: return (all_numbers[div] + all_numbers[div - 1]) / 2 if __name__ == "__main__": import doctest doctest.testmod() a_ = [float(x) for x in input("Enter the elements of first array: ").split()] a_ = [float(x) for x in input("Enter the elements of second array: ").split()] print(f'''The median of two arrays is: {median_of_two_arrays(array_a, array_a)}''')
621
0
"""simple docstring""" import argparse import shutil from pathlib import Path from tqdm import tqdm from transformers import AutoTokenizer def a__ ( __lowercase , __lowercase , __lowercase , __lowercase=1024 ) -> Any: _A , _A = [], [] _A = list(zip(__lowercase , __lowercase ) ) _A , _A = sorted_examples[0] def is_too_big(__lowercase ): return tok(__lowercase , return_tensors="pt" ).input_ids.shape[1] > max_tokens for src, tgt in tqdm(sorted_examples[1:] ): _A = new_src + " " + src _A = new_tgt + " " + tgt if is_too_big(__lowercase ) or is_too_big(__lowercase ): # cant fit, finalize example finished_src.append(__lowercase ) finished_tgt.append(__lowercase ) _A , _A = src, tgt else: # can fit, keep adding _A , _A = cand_src, cand_tgt # cleanup if new_src: assert new_tgt finished_src.append(__lowercase ) finished_tgt.append(__lowercase ) return finished_src, finished_tgt def a__ ( __lowercase , __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = Path(__lowercase ) save_path.mkdir(exist_ok=__lowercase ) for split in ["train"]: _A , _A = data_dir / f"""{split}.source""", data_dir / f"""{split}.target""" _A = [x.rstrip() for x in Path(__lowercase ).open().readlines()] _A = [x.rstrip() for x in Path(__lowercase ).open().readlines()] _A , _A = pack_examples(__lowercase , __lowercase , __lowercase , __lowercase ) print(f"""packed {split} split from {len(__lowercase )} examples -> {len(__lowercase )}.""" ) Path(save_path / f"""{split}.source""" ).open("w" ).write("\n".join(__lowercase ) ) Path(save_path / f"""{split}.target""" ).open("w" ).write("\n".join(__lowercase ) ) for split in ["val", "test"]: _A , _A = data_dir / f"""{split}.source""", data_dir / f"""{split}.target""" shutil.copyfile(__lowercase , save_path / f"""{split}.source""" ) shutil.copyfile(__lowercase , save_path / f"""{split}.target""" ) def a__ ( ) -> Tuple: _A = argparse.ArgumentParser() parser.add_argument("--tok_name" , type=__lowercase , help="like facebook/bart-large-cnn,t5-base, etc." ) parser.add_argument("--max_seq_len" , type=__lowercase , default=128 ) parser.add_argument("--data_dir" , type=__lowercase ) parser.add_argument("--save_path" , type=__lowercase ) _A = parser.parse_args() _A = AutoTokenizer.from_pretrained(args.tok_name ) return pack_data_dir(__lowercase , Path(args.data_dir ) , args.max_seq_len , args.save_path ) if __name__ == "__main__": packer_cli()
709
"""simple docstring""" import copy import os from typing import Union from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { "Salesforce/blip-vqa-base": "https://huggingface.co/Salesforce/blip-vqa-base/resolve/main/config.json", "Salesforce/blip-vqa-capfit-large": ( "https://huggingface.co/Salesforce/blip-vqa-base-capfit/resolve/main/config.json" ), "Salesforce/blip-image-captioning-base": ( "https://huggingface.co/Salesforce/blip-image-captioning-base/resolve/main/config.json" ), "Salesforce/blip-image-captioning-large": ( "https://huggingface.co/Salesforce/blip-image-captioning-large/resolve/main/config.json" ), "Salesforce/blip-itm-base-coco": "https://huggingface.co/Salesforce/blip-itm-base-coco/resolve/main/config.json", "Salesforce/blip-itm-large-coco": "https://huggingface.co/Salesforce/blip-itm-large-coco/resolve/main/config.json", "Salesforce/blip-itm-base-flikr": "https://huggingface.co/Salesforce/blip-itm-base-flikr/resolve/main/config.json", "Salesforce/blip-itm-large-flikr": ( "https://huggingface.co/Salesforce/blip-itm-large-flikr/resolve/main/config.json" ), } class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip_text_model' def __init__( self : int , a__ : List[str]=3_05_24 , a__ : List[str]=7_68 , a__ : List[Any]=7_68 , a__ : int=30_72 , a__ : List[str]=7_68 , a__ : Dict=12 , a__ : Optional[int]=8 , a__ : Optional[Any]=5_12 , a__ : List[Any]="gelu" , a__ : Optional[Any]=1E-1_2 , a__ : Any=0.0 , a__ : int=0.0 , a__ : Dict=0.0_2 , a__ : Optional[Any]=3_05_22 , a__ : Any=2 , a__ : int=0 , a__ : Union[str, Any]=1_02 , a__ : Tuple=True , a__ : Optional[int]=True , **a__ : Any , ) -> List[Any]: '''simple docstring''' super().__init__( pad_token_id=a__ , bos_token_id=a__ , eos_token_id=a__ , sep_token_id=a__ , **a__ , ) _A = vocab_size _A = hidden_size _A = encoder_hidden_size _A = intermediate_size _A = projection_dim _A = hidden_dropout_prob _A = num_hidden_layers _A = num_attention_heads _A = max_position_embeddings _A = layer_norm_eps _A = hidden_act _A = initializer_range _A = attention_probs_dropout_prob _A = is_decoder _A = use_cache @classmethod def a_ ( cls : Optional[Any] , a__ : Union[str, os.PathLike] , **a__ : Optional[Any] ) -> "PretrainedConfig": '''simple docstring''' cls._set_token_in_kwargs(a__ ) _A , _A = cls.get_config_dict(a__ , **a__ ) # get the text config dict if we are loading from BlipConfig if config_dict.get("model_type" ) == "blip": _A = config_dict["text_config"] if "model_type" in config_dict and hasattr(cls , "model_type" ) and config_dict["model_type"] != cls.model_type: logger.warning( F"""You are using a model of type {config_dict['model_type']} to instantiate a model of type """ F"""{cls.model_type}. This is not supported for all configurations of models and can yield errors.""" ) return cls.from_dict(a__ , **a__ ) class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip_vision_model' def __init__( self : Optional[Any] , a__ : Any=7_68 , a__ : List[str]=30_72 , a__ : str=5_12 , a__ : Any=12 , a__ : int=12 , a__ : int=3_84 , a__ : Tuple=16 , a__ : str="gelu" , a__ : Tuple=1E-5 , a__ : List[str]=0.0 , a__ : List[Any]=1E-1_0 , **a__ : int , ) -> List[str]: '''simple docstring''' super().__init__(**a__ ) _A = hidden_size _A = intermediate_size _A = projection_dim _A = num_hidden_layers _A = num_attention_heads _A = patch_size _A = image_size _A = initializer_range _A = attention_dropout _A = layer_norm_eps _A = hidden_act @classmethod def a_ ( cls : Any , a__ : Union[str, os.PathLike] , **a__ : int ) -> "PretrainedConfig": '''simple docstring''' cls._set_token_in_kwargs(a__ ) _A , _A = cls.get_config_dict(a__ , **a__ ) # get the vision config dict if we are loading from BlipConfig if config_dict.get("model_type" ) == "blip": _A = config_dict["vision_config"] if "model_type" in config_dict and hasattr(cls , "model_type" ) and config_dict["model_type"] != cls.model_type: logger.warning( F"""You are using a model of type {config_dict['model_type']} to instantiate a model of type """ F"""{cls.model_type}. This is not supported for all configurations of models and can yield errors.""" ) return cls.from_dict(a__ , **a__ ) class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip' __UpperCamelCase = True def __init__( self : List[Any] , a__ : Optional[int]=None , a__ : str=None , a__ : List[str]=5_12 , a__ : Any=2.6_5_9_2 , a__ : str=2_56 , **a__ : Optional[int] , ) -> Dict: '''simple docstring''' super().__init__(**a__ ) if text_config is None: _A = {} logger.info("`text_config` is `None`. Initializing the `BlipTextConfig` with default values." ) if vision_config is None: _A = {} logger.info("`vision_config` is `None`. Initializing the `BlipVisionConfig` with default values." ) _A = BlipTextConfig(**a__ ) _A = BlipVisionConfig(**a__ ) _A = self.vision_config.hidden_size _A = projection_dim _A = logit_scale_init_value _A = 1.0 _A = 0.0_2 _A = image_text_hidden_size @classmethod def a_ ( cls : Tuple , a__ : BlipTextConfig , a__ : BlipVisionConfig , **a__ : Optional[int] ) -> str: '''simple docstring''' return cls(text_config=text_config.to_dict() , vision_config=vision_config.to_dict() , **a__ ) def a_ ( self : Union[str, Any] ) -> List[Any]: '''simple docstring''' _A = copy.deepcopy(self.__dict__ ) _A = self.text_config.to_dict() _A = self.vision_config.to_dict() _A = self.__class__.model_type return output
621
0
"""simple docstring""" import json import os import unittest from transformers import BatchEncoding, MvpTokenizer, MvpTokenizerFast from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES from transformers.testing_utils import require_tokenizers, require_torch from transformers.utils import cached_property from ...test_tokenization_common import TokenizerTesterMixin, filter_roberta_detectors @require_tokenizers class snake_case ( _UpperCamelCase , unittest.TestCase): __UpperCamelCase = MvpTokenizer __UpperCamelCase = MvpTokenizerFast __UpperCamelCase = True __UpperCamelCase = filter_roberta_detectors def a_ ( self : List[Any] ) -> List[Any]: '''simple docstring''' super().setUp() _A = [ "l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "\u0120", "\u0120l", "\u0120n", "\u0120lo", "\u0120low", "er", "\u0120lowest", "\u0120newer", "\u0120wider", "<unk>", ] _A = dict(zip(a__ , range(len(a__ ) ) ) ) _A = ["#version: 0.2", "\u0120 l", "\u0120l o", "\u0120lo w", "e r", ""] _A = {"unk_token": "<unk>"} _A = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["vocab_file"] ) _A = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["merges_file"] ) with open(self.vocab_file , "w" , encoding="utf-8" ) as fp: fp.write(json.dumps(a__ ) + "\n" ) with open(self.merges_file , "w" , encoding="utf-8" ) as fp: fp.write("\n".join(a__ ) ) def a_ ( self : Optional[Any] , **a__ : List[Any] ) -> Tuple: '''simple docstring''' kwargs.update(self.special_tokens_map ) return self.tokenizer_class.from_pretrained(self.tmpdirname , **a__ ) def a_ ( self : int , **a__ : int ) -> Union[str, Any]: '''simple docstring''' kwargs.update(self.special_tokens_map ) return self.rust_tokenizer_class.from_pretrained(self.tmpdirname , **a__ ) def a_ ( self : str , a__ : int ) -> Union[str, Any]: '''simple docstring''' return "lower newer", "lower newer" @cached_property def a_ ( self : int ) -> Optional[int]: '''simple docstring''' return MvpTokenizer.from_pretrained("RUCAIBox/mvp" ) @cached_property def a_ ( self : Tuple ) -> Any: '''simple docstring''' return MvpTokenizerFast.from_pretrained("RUCAIBox/mvp" ) @require_torch def a_ ( self : Optional[int] ) -> Tuple: '''simple docstring''' _A = ["A long paragraph for summarization.", "Another paragraph for summarization."] _A = [0, 2_50, 2_51, 1_78_18, 13, 3_91_86, 19_38, 4, 2] for tokenizer in [self.default_tokenizer, self.default_tokenizer_fast]: _A = tokenizer(a__ , max_length=len(a__ ) , padding=a__ , return_tensors="pt" ) self.assertIsInstance(a__ , a__ ) self.assertEqual((2, 9) , batch.input_ids.shape ) self.assertEqual((2, 9) , batch.attention_mask.shape ) _A = batch.input_ids.tolist()[0] self.assertListEqual(a__ , a__ ) # Test that special tokens are reset @require_torch def a_ ( self : str ) -> Dict: '''simple docstring''' _A = ["A long paragraph for summarization.", "Another paragraph for summarization."] for tokenizer in [self.default_tokenizer, self.default_tokenizer_fast]: _A = tokenizer(a__ , padding=a__ , return_tensors="pt" ) # check if input_ids are returned and no labels self.assertIn("input_ids" , a__ ) self.assertIn("attention_mask" , a__ ) self.assertNotIn("labels" , a__ ) self.assertNotIn("decoder_attention_mask" , a__ ) @require_torch def a_ ( self : Optional[Any] ) -> int: '''simple docstring''' _A = [ "Summary of the text.", "Another summary.", ] for tokenizer in [self.default_tokenizer, self.default_tokenizer_fast]: _A = tokenizer(text_target=a__ , max_length=32 , padding="max_length" , return_tensors="pt" ) self.assertEqual(32 , targets["input_ids"].shape[1] ) @require_torch def a_ ( self : List[str] ) -> Optional[int]: '''simple docstring''' for tokenizer in [self.default_tokenizer, self.default_tokenizer_fast]: _A = tokenizer( ["I am a small frog" * 10_24, "I am a small frog"] , padding=a__ , truncation=a__ , return_tensors="pt" ) self.assertIsInstance(a__ , a__ ) self.assertEqual(batch.input_ids.shape , (2, 10_24) ) @require_torch def a_ ( self : List[str] ) -> Tuple: '''simple docstring''' _A = ["A long paragraph for summarization."] _A = [ "Summary of the text.", ] for tokenizer in [self.default_tokenizer, self.default_tokenizer_fast]: _A = tokenizer(a__ , text_target=a__ , return_tensors="pt" ) _A = inputs["input_ids"] _A = inputs["labels"] self.assertTrue((input_ids[:, 0] == tokenizer.bos_token_id).all().item() ) self.assertTrue((labels[:, 0] == tokenizer.bos_token_id).all().item() ) self.assertTrue((input_ids[:, -1] == tokenizer.eos_token_id).all().item() ) self.assertTrue((labels[:, -1] == tokenizer.eos_token_id).all().item() ) def a_ ( self : str ) -> Union[str, Any]: '''simple docstring''' pass def a_ ( self : int ) -> Union[str, Any]: '''simple docstring''' for tokenizer, pretrained_name, kwargs in self.tokenizers_list: with self.subTest(F"""{tokenizer.__class__.__name__} ({pretrained_name})""" ): _A = self.rust_tokenizer_class.from_pretrained(a__ , **a__ ) _A = self.tokenizer_class.from_pretrained(a__ , **a__ ) _A = "A, <mask> AllenNLP sentence." _A = tokenizer_r.encode_plus(a__ , add_special_tokens=a__ , return_token_type_ids=a__ ) _A = tokenizer_p.encode_plus(a__ , add_special_tokens=a__ , return_token_type_ids=a__ ) # token_type_ids should put 0 everywhere self.assertEqual(sum(tokens_r["token_type_ids"] ) , sum(tokens_p["token_type_ids"] ) ) # attention_mask should put 1 everywhere, so sum over length should be 1 self.assertEqual( sum(tokens_r["attention_mask"] ) / len(tokens_r["attention_mask"] ) , sum(tokens_p["attention_mask"] ) / len(tokens_p["attention_mask"] ) , ) _A = tokenizer_r.convert_ids_to_tokens(tokens_r["input_ids"] ) _A = tokenizer_p.convert_ids_to_tokens(tokens_p["input_ids"] ) # Rust correctly handles the space before the mask while python doesnt self.assertSequenceEqual(tokens_p["input_ids"] , [0, 2_50, 6, 5_02_64, 38_23, 4_87, 2_19_92, 36_45, 4, 2] ) self.assertSequenceEqual(tokens_r["input_ids"] , [0, 2_50, 6, 5_02_64, 38_23, 4_87, 2_19_92, 36_45, 4, 2] ) self.assertSequenceEqual( a__ , ["<s>", "A", ",", "<mask>", "ĠAllen", "N", "LP", "Ġsentence", ".", "</s>"] ) self.assertSequenceEqual( a__ , ["<s>", "A", ",", "<mask>", "ĠAllen", "N", "LP", "Ġsentence", ".", "</s>"] )
710
"""simple docstring""" import unittest from transformers import load_tool from .test_tools_common import ToolTesterMixin class snake_case ( unittest.TestCase , _UpperCamelCase): def a_ ( self : Optional[Any] ) -> List[str]: '''simple docstring''' _A = load_tool("text-classification" ) self.tool.setup() _A = load_tool("text-classification" , remote=a__ ) def a_ ( self : Optional[int] ) -> Dict: '''simple docstring''' _A = self.tool("That's quite cool" , ["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Optional[Any] ) -> Dict: '''simple docstring''' _A = self.remote_tool("That's quite cool" , ["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Dict ) -> Optional[int]: '''simple docstring''' _A = self.tool(text="That's quite cool" , labels=["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Dict ) -> Any: '''simple docstring''' _A = self.remote_tool(text="That's quite cool" , labels=["positive", "negative"] ) self.assertEqual(a__ , "positive" )
621
0
"""simple docstring""" import os import re import warnings from shutil import copyfile from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import PreTrainedTokenizer if TYPE_CHECKING: from ...tokenization_utils_base import TextInput from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"vocab_file": "spiece.model"} a_ = { "vocab_file": { "t5-small": "https://huggingface.co/t5-small/resolve/main/spiece.model", "t5-base": "https://huggingface.co/t5-base/resolve/main/spiece.model", "t5-large": "https://huggingface.co/t5-large/resolve/main/spiece.model", "t5-3b": "https://huggingface.co/t5-3b/resolve/main/spiece.model", "t5-11b": "https://huggingface.co/t5-11b/resolve/main/spiece.model", } } # TODO(PVP) - this should be removed in Transformers v5 a_ = { "t5-small": 5_12, "t5-base": 5_12, "t5-large": 5_12, "t5-3b": 5_12, "t5-11b": 5_12, } a_ = "▁" class snake_case ( _UpperCamelCase): __UpperCamelCase = VOCAB_FILES_NAMES __UpperCamelCase = PRETRAINED_VOCAB_FILES_MAP __UpperCamelCase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __UpperCamelCase = ['input_ids', 'attention_mask'] def __init__( self : List[str] , a__ : Optional[int] , a__ : Union[str, Any]="</s>" , a__ : Union[str, Any]="<unk>" , a__ : str="<pad>" , a__ : Optional[int]=1_00 , a__ : List[Any]=None , a__ : Optional[Dict[str, Any]] = None , a__ : Any=True , **a__ : Optional[int] , ) -> None: '''simple docstring''' if extra_ids > 0 and additional_special_tokens is None: _A = [F"""<extra_id_{i}>""" for i in range(a__ )] elif extra_ids > 0 and additional_special_tokens is not None: # Check that we have the right number of extra_id special tokens _A = len(set(filter(lambda a__ : bool("extra_id" in str(a__ ) ) , a__ ) ) ) if extra_tokens != extra_ids: raise ValueError( F"""Both extra_ids ({extra_ids}) and additional_special_tokens ({additional_special_tokens}) are""" " provided to T5Tokenizer. In this case the additional_special_tokens must include the extra_ids" " tokens" ) if legacy: logger.warning_once( F"""You are using the legacy behaviour of the {self.__class__}. This means that tokens that come after special tokens will not be properly handled. We recommend you to""" " read the related pull request available at https://github.com/huggingface/transformers/pull/24565" ) _A = legacy _A = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( eos_token=a__ , unk_token=a__ , pad_token=a__ , extra_ids=a__ , additional_special_tokens=a__ , sp_model_kwargs=self.sp_model_kwargs , legacy=a__ , **a__ , ) _A = vocab_file _A = extra_ids _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(a__ ) @staticmethod def a_ ( a__ : List[str] , a__ : Optional[int] , a__ : Tuple ) -> Tuple: '''simple docstring''' if pretrained_model_name_or_path in TaTokenizer.max_model_input_sizes: _A = TaTokenizer.max_model_input_sizes[pretrained_model_name_or_path] if init_max_model_length is not None and init_max_model_length != max_model_length: return init_max_model_length elif init_max_model_length is None: warnings.warn( "This tokenizer was incorrectly instantiated with a model max length of" F""" {deprecated_max_model_length} which will be corrected in Transformers v5.\nFor now, this""" " behavior is kept to avoid breaking backwards compatibility when padding/encoding with" " `truncation is True`.\n- Be aware that you SHOULD NOT rely on" F""" {pretrained_model_name_or_path} automatically truncating your input to""" F""" {deprecated_max_model_length} when padding/encoding.\n- If you want to encode/pad to sequences""" F""" longer than {deprecated_max_model_length} you can either instantiate this tokenizer with""" " `model_max_length` or pass `max_length` when encoding/padding.\n- To avoid this warning, please" " instantiate this tokenizer with `model_max_length` set to your preferred value." , a__ , ) return max_model_length @property def a_ ( self : List[Any] ) -> Dict: '''simple docstring''' return self.sp_model.get_piece_size() + self._extra_ids def a_ ( self : Dict ) -> Optional[Any]: '''simple docstring''' _A = {self.convert_ids_to_tokens(a__ ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def a_ ( self : Optional[Any] , a__ : List[int] , a__ : Optional[List[int]] = None , a__ : bool = False ) -> List[int]: '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=a__ , token_ids_a=a__ , already_has_special_tokens=a__ ) # normal case: some special tokens if token_ids_a is None: return ([0] * len(a__ )) + [1] return ([0] * len(a__ )) + [1] + ([0] * len(a__ )) + [1] def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' return list( set(filter(lambda a__ : bool(re.search(r"<extra_id_\d+>" , a__ ) ) is not None , self.additional_special_tokens ) ) ) def a_ ( self : str ) -> List[Any]: '''simple docstring''' return [self._convert_token_to_id(a__ ) for token in self.get_sentinel_tokens()] def a_ ( self : List[Any] , a__ : List[int] ) -> List[int]: '''simple docstring''' if len(a__ ) > 0 and token_ids[-1] == self.eos_token_id: warnings.warn( F"""This sequence already has {self.eos_token}. In future versions this behavior may lead to duplicated""" " eos tokens being added." ) return token_ids else: return token_ids + [self.eos_token_id] def a_ ( self : int , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = [self.eos_token_id] if token_ids_a is None: return len(token_ids_a + eos ) * [0] return len(token_ids_a + eos + token_ids_a + eos ) * [0] def a_ ( self : Union[str, Any] , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = self._add_eos_if_not_present(a__ ) if token_ids_a is None: return token_ids_a else: _A = self._add_eos_if_not_present(a__ ) return token_ids_a + token_ids_a def __getstate__( self : Dict ) -> Union[str, Any]: '''simple docstring''' _A = self.__dict__.copy() _A = None return state def __setstate__( self : int , a__ : Optional[int] ) -> Union[str, Any]: '''simple docstring''' _A = d # for backward compatibility if not hasattr(self , "sp_model_kwargs" ): _A = {} _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def a_ ( self : int , a__ : "TextInput" , **a__ : List[str] ) -> List[str]: '''simple docstring''' if not self.legacy: _A = SPIECE_UNDERLINE + text.replace(a__ , " " ) return super().tokenize(a__ , **a__ ) def a_ ( self : str , a__ : Dict , **a__ : Optional[int] ) -> Any: '''simple docstring''' if not self.legacy: _A = text.startswith(a__ ) if is_first: _A = text[1:] _A = self.sp_model.encode(a__ , out_type=a__ ) if not self.legacy and not is_first and not text.startswith(" " ) and tokens[0].startswith(a__ ): _A = ([tokens[0][1:]] if len(tokens[0] ) > 1 else []) + tokens[1:] return tokens def a_ ( self : int , a__ : List[Any] ) -> List[str]: '''simple docstring''' if token.startswith("<extra_id_" ): _A = re.match(r"<extra_id_(\d+)>" , a__ ) _A = int(match.group(1 ) ) return self.vocab_size - num - 1 return self.sp_model.piece_to_id(a__ ) def a_ ( self : Dict , a__ : Union[str, Any] ) -> Any: '''simple docstring''' if index < self.sp_model.get_piece_size(): _A = self.sp_model.IdToPiece(a__ ) else: _A = F"""<extra_id_{self.vocab_size - 1 - index}>""" return token def a_ ( self : Optional[int] , a__ : Tuple ) -> List[str]: '''simple docstring''' _A = [] _A = "" _A = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(a__ ) + token _A = True _A = [] else: current_sub_tokens.append(a__ ) _A = False out_string += self.sp_model.decode(a__ ) return out_string.strip() def a_ ( self : Dict , a__ : str , a__ : Optional[str] = None ) -> Tuple[str]: '''simple docstring''' if not os.path.isdir(a__ ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return _A = os.path.join( a__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(a__ ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , a__ ) elif not os.path.isfile(self.vocab_file ): with open(a__ , "wb" ) as fi: _A = self.sp_model.serialized_model_proto() fi.write(a__ ) return (out_vocab_file,)
711
"""simple docstring""" import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer from diffusers import AutoencoderKL, PNDMScheduler, StableDiffusionInpaintPipeline, UNetaDConditionModel from diffusers.utils import floats_tensor, load_image, load_numpy, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu, slow from ..pipeline_params import TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS, TEXT_GUIDED_IMAGE_INPAINTING_PARAMS from ..test_pipelines_common import PipelineKarrasSchedulerTesterMixin, PipelineLatentTesterMixin, PipelineTesterMixin enable_full_determinism() class snake_case ( _UpperCamelCase , _UpperCamelCase , _UpperCamelCase , unittest.TestCase): __UpperCamelCase = StableDiffusionInpaintPipeline __UpperCamelCase = TEXT_GUIDED_IMAGE_INPAINTING_PARAMS __UpperCamelCase = TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS __UpperCamelCase = frozenset( []) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess __UpperCamelCase = frozenset([]) def a_ ( self : Dict ) -> Optional[int]: '''simple docstring''' torch.manual_seed(0 ) _A = UNetaDConditionModel( block_out_channels=(32, 64) , layers_per_block=2 , sample_size=32 , in_channels=9 , out_channels=4 , down_block_types=("DownBlock2D", "CrossAttnDownBlock2D") , up_block_types=("CrossAttnUpBlock2D", "UpBlock2D") , cross_attention_dim=32 , attention_head_dim=(2, 4) , use_linear_projection=a__ , ) _A = PNDMScheduler(skip_prk_steps=a__ ) torch.manual_seed(0 ) _A = AutoencoderKL( block_out_channels=[32, 64] , in_channels=3 , out_channels=3 , down_block_types=["DownEncoderBlock2D", "DownEncoderBlock2D"] , up_block_types=["UpDecoderBlock2D", "UpDecoderBlock2D"] , latent_channels=4 , sample_size=1_28 , ) torch.manual_seed(0 ) _A = CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=32 , intermediate_size=37 , layer_norm_eps=1E-0_5 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=10_00 , hidden_act="gelu" , projection_dim=5_12 , ) _A = CLIPTextModel(a__ ) _A = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip" ) _A = { "unet": unet, "scheduler": scheduler, "vae": vae, "text_encoder": text_encoder, "tokenizer": tokenizer, "safety_checker": None, "feature_extractor": None, } return components def a_ ( self : Optional[Any] , a__ : List[str] , a__ : Tuple=0 ) -> int: '''simple docstring''' _A = floats_tensor((1, 3, 32, 32) , rng=random.Random(a__ ) ).to(a__ ) _A = image.cpu().permute(0 , 2 , 3 , 1 )[0] _A = Image.fromarray(np.uinta(a__ ) ).convert("RGB" ).resize((64, 64) ) _A = Image.fromarray(np.uinta(image + 4 ) ).convert("RGB" ).resize((64, 64) ) if str(a__ ).startswith("mps" ): _A = torch.manual_seed(a__ ) else: _A = torch.Generator(device=a__ ).manual_seed(a__ ) _A = { "prompt": "A painting of a squirrel eating a burger", "image": init_image, "mask_image": mask_image, "generator": generator, "num_inference_steps": 2, "guidance_scale": 6.0, "output_type": "numpy", } return inputs def a_ ( self : Optional[Any] ) -> Optional[Any]: '''simple docstring''' _A = "cpu" # ensure determinism for the device-dependent torch.Generator _A = self.get_dummy_components() _A = StableDiffusionInpaintPipeline(**a__ ) _A = sd_pipe.to(a__ ) sd_pipe.set_progress_bar_config(disable=a__ ) _A = self.get_dummy_inputs(a__ ) _A = sd_pipe(**a__ ).images _A = image[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) _A = np.array([0.4_7_2_7, 0.5_7_3_5, 0.3_9_4_1, 0.5_4_4_6, 0.5_9_2_6, 0.4_3_9_4, 0.5_0_6_2, 0.4_6_5_4, 0.4_4_7_6] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-2 def a_ ( self : str ) -> Union[str, Any]: '''simple docstring''' super().test_inference_batch_single_identical(expected_max_diff=3E-3 ) @slow @require_torch_gpu class snake_case ( unittest.TestCase): def a_ ( self : List[Any] ) -> Any: '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def a_ ( self : Union[str, Any] ) -> Tuple: '''simple docstring''' _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = load_numpy( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint" "/yellow_cat_sitting_on_a_park_bench.npy" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = StableDiffusionInpaintPipeline.from_pretrained(a__ , safety_checker=a__ ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , output_type="np" , ) _A = output.images[0] assert image.shape == (5_12, 5_12, 3) assert np.abs(expected_image - image ).max() < 9E-3 def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = load_numpy( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint" "/yellow_cat_sitting_on_a_park_bench_fp16.npy" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = StableDiffusionInpaintPipeline.from_pretrained( a__ , torch_dtype=torch.floataa , safety_checker=a__ , ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , output_type="np" , ) _A = output.images[0] assert image.shape == (5_12, 5_12, 3) assert np.abs(expected_image - image ).max() < 5E-1 def a_ ( self : List[Any] ) -> Union[str, Any]: '''simple docstring''' torch.cuda.empty_cache() torch.cuda.reset_max_memory_allocated() torch.cuda.reset_peak_memory_stats() _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = PNDMScheduler.from_pretrained(a__ , subfolder="scheduler" ) _A = StableDiffusionInpaintPipeline.from_pretrained( a__ , safety_checker=a__ , scheduler=a__ , torch_dtype=torch.floataa , ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing(1 ) pipe.enable_sequential_cpu_offload() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , num_inference_steps=2 , output_type="np" , ) _A = torch.cuda.max_memory_allocated() # make sure that less than 2.65 GB is allocated assert mem_bytes < 2.6_5 * 10**9
621
0
"""simple docstring""" import inspect from typing import Callable, List, Optional, Union import torch from transformers import ( CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, WhisperForConditionalGeneration, WhisperProcessor, ) from diffusers import ( AutoencoderKL, DDIMScheduler, DiffusionPipeline, LMSDiscreteScheduler, PNDMScheduler, UNetaDConditionModel, ) from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import StableDiffusionPipelineOutput from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker from diffusers.utils import logging a_ = logging.get_logger(__name__) # pylint: disable=invalid-name class snake_case ( _UpperCamelCase): def __init__( self : int , a__ : WhisperForConditionalGeneration , a__ : WhisperProcessor , a__ : AutoencoderKL , a__ : CLIPTextModel , a__ : CLIPTokenizer , a__ : UNetaDConditionModel , a__ : Union[DDIMScheduler, PNDMScheduler, LMSDiscreteScheduler] , a__ : StableDiffusionSafetyChecker , a__ : CLIPImageProcessor , ) -> List[str]: '''simple docstring''' super().__init__() 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( speech_model=a__ , speech_processor=a__ , vae=a__ , text_encoder=a__ , tokenizer=a__ , unet=a__ , scheduler=a__ , feature_extractor=a__ , ) def a_ ( self : List[str] , a__ : Optional[Union[str, int]] = "auto" ) -> List[str]: '''simple docstring''' if slice_size == "auto": _A = self.unet.config.attention_head_dim // 2 self.unet.set_attention_slice(a__ ) def a_ ( self : Union[str, Any] ) -> Optional[Any]: '''simple docstring''' self.enable_attention_slicing(a__ ) @torch.no_grad() def __call__( self : Dict , a__ : Optional[int] , a__ : List[str]=1_60_00 , a__ : int = 5_12 , a__ : int = 5_12 , a__ : int = 50 , a__ : float = 7.5 , a__ : Optional[Union[str, List[str]]] = None , a__ : Optional[int] = 1 , a__ : float = 0.0 , a__ : Optional[torch.Generator] = None , a__ : Optional[torch.FloatTensor] = None , a__ : Optional[str] = "pil" , a__ : bool = True , a__ : Optional[Callable[[int, int, torch.FloatTensor], None]] = None , a__ : int = 1 , **a__ : Any , ) -> Union[str, Any]: '''simple docstring''' _A = self.speech_processor.feature_extractor( a__ , return_tensors="pt" , sampling_rate=a__ ).input_features.to(self.device ) _A = self.speech_model.generate(a__ , max_length=48_00_00 ) _A = self.speech_processor.tokenizer.batch_decode(a__ , skip_special_tokens=a__ , normalize=a__ )[ 0 ] if isinstance(a__ , a__ ): _A = 1 elif isinstance(a__ , a__ ): _A = len(a__ ) else: raise ValueError(F"""`prompt` has to be of type `str` or `list` but is {type(a__ )}""" ) if height % 8 != 0 or width % 8 != 0: raise ValueError(F"""`height` and `width` have to be divisible by 8 but are {height} and {width}.""" ) if (callback_steps is None) or ( callback_steps is not None and (not isinstance(a__ , a__ ) or callback_steps <= 0) ): raise ValueError( F"""`callback_steps` has to be a positive integer but is {callback_steps} of type""" F""" {type(a__ )}.""" ) # get prompt text embeddings _A = self.tokenizer( a__ , padding="max_length" , max_length=self.tokenizer.model_max_length , return_tensors="pt" , ) _A = text_inputs.input_ids if text_input_ids.shape[-1] > self.tokenizer.model_max_length: _A = self.tokenizer.batch_decode(text_input_ids[:, self.tokenizer.model_max_length :] ) logger.warning( "The following part of your input was truncated because CLIP can only handle sequences up to" F""" {self.tokenizer.model_max_length} tokens: {removed_text}""" ) _A = text_input_ids[:, : self.tokenizer.model_max_length] _A = self.text_encoder(text_input_ids.to(self.device ) )[0] # duplicate text embeddings for each generation per prompt, using mps friendly method _A , _A , _A = text_embeddings.shape _A = text_embeddings.repeat(1 , a__ , 1 ) _A = text_embeddings.view(bs_embed * num_images_per_prompt , a__ , -1 ) # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2) # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1` # corresponds to doing no classifier free guidance. _A = guidance_scale > 1.0 # get unconditional embeddings for classifier free guidance if do_classifier_free_guidance: _A = 42 if negative_prompt is None: _A = [""] * batch_size elif type(a__ ) is not type(a__ ): raise TypeError( F"""`negative_prompt` should be the same type to `prompt`, but got {type(a__ )} !=""" F""" {type(a__ )}.""" ) elif isinstance(a__ , a__ ): _A = [negative_prompt] elif batch_size != len(a__ ): raise ValueError( F"""`negative_prompt`: {negative_prompt} has batch size {len(a__ )}, but `prompt`:""" F""" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches""" " the batch size of `prompt`." ) else: _A = negative_prompt _A = text_input_ids.shape[-1] _A = self.tokenizer( a__ , padding="max_length" , max_length=a__ , truncation=a__ , return_tensors="pt" , ) _A = self.text_encoder(uncond_input.input_ids.to(self.device ) )[0] # duplicate unconditional embeddings for each generation per prompt, using mps friendly method _A = uncond_embeddings.shape[1] _A = uncond_embeddings.repeat(1 , a__ , 1 ) _A = uncond_embeddings.view(batch_size * num_images_per_prompt , a__ , -1 ) # For classifier free guidance, we need to do two forward passes. # Here we concatenate the unconditional and text embeddings into a single batch # to avoid doing two forward passes _A = torch.cat([uncond_embeddings, text_embeddings] ) # get the initial random noise unless the user supplied it # Unlike in other pipelines, latents need to be generated in the target device # for 1-to-1 results reproducibility with the CompVis implementation. # However this currently doesn't work in `mps`. _A = (batch_size * num_images_per_prompt, self.unet.config.in_channels, height // 8, width // 8) _A = text_embeddings.dtype if latents is None: if self.device.type == "mps": # randn does not exist on mps _A = torch.randn(a__ , generator=a__ , device="cpu" , dtype=a__ ).to( self.device ) else: _A = torch.randn(a__ , generator=a__ , device=self.device , dtype=a__ ) else: if latents.shape != latents_shape: raise ValueError(F"""Unexpected latents shape, got {latents.shape}, expected {latents_shape}""" ) _A = latents.to(self.device ) # set timesteps self.scheduler.set_timesteps(a__ ) # Some schedulers like PNDM have timesteps as arrays # It's more optimized to move all timesteps to correct device beforehand _A = self.scheduler.timesteps.to(self.device ) # scale the initial noise by the standard deviation required by the scheduler _A = latents * self.scheduler.init_noise_sigma # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers. # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502 # and should be between [0, 1] _A = "eta" in set(inspect.signature(self.scheduler.step ).parameters.keys() ) _A = {} if accepts_eta: _A = eta for i, t in enumerate(self.progress_bar(a__ ) ): # expand the latents if we are doing classifier free guidance _A = torch.cat([latents] * 2 ) if do_classifier_free_guidance else latents _A = self.scheduler.scale_model_input(a__ , a__ ) # predict the noise residual _A = self.unet(a__ , a__ , encoder_hidden_states=a__ ).sample # perform guidance if do_classifier_free_guidance: _A , _A = noise_pred.chunk(2 ) _A = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond) # compute the previous noisy sample x_t -> x_t-1 _A = self.scheduler.step(a__ , a__ , a__ , **a__ ).prev_sample # call the callback, if provided if callback is not None and i % callback_steps == 0: callback(a__ , a__ , a__ ) _A = 1 / 0.1_8_2_1_5 * latents _A = self.vae.decode(a__ ).sample _A = (image / 2 + 0.5).clamp(0 , 1 ) # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16 _A = image.cpu().permute(0 , 2 , 3 , 1 ).float().numpy() if output_type == "pil": _A = self.numpy_to_pil(a__ ) if not return_dict: return image return StableDiffusionPipelineOutput(images=a__ , nsfw_content_detected=a__ )
712
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> int: while a != 0: _A , _A = b % a, a return b def a__ ( __lowercase , __lowercase ) -> int: if gcd(__lowercase , __lowercase ) != 1: _A = f"""mod inverse of {a!r} and {m!r} does not exist""" raise ValueError(__lowercase ) _A , _A , _A = 1, 0, a _A , _A , _A = 0, 1, m while va != 0: _A = ua // va _A , _A , _A , _A , _A , _A = (ua - q * va), (ua - q * va), (ua - q * va), va, va, va return ua % m
621
0
"""simple docstring""" a_ : List[str] = "\n# Transformers installation\n! pip install transformers datasets\n# To install from source instead of the last release, comment the command above and uncomment the following one.\n# ! pip install git+https://github.com/huggingface/transformers.git\n" a_ : int = [{"type": "code", "content": INSTALL_CONTENT}] a_ : List[Any] = { "{processor_class}": "FakeProcessorClass", "{model_class}": "FakeModelClass", "{object_class}": "FakeObjectClass", }
713
"""simple docstring""" # 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 snake_case ( _UpperCamelCase): def __init__( self : List[Any] , a__ : Any ) -> Any: '''simple docstring''' _A = data def __iter__( self : List[str] ) -> str: '''simple docstring''' for element in self.data: yield element def a__ ( __lowercase=True ) -> Tuple: _A = Accelerator(even_batches=__lowercase ) assert accelerator.num_processes == 2, "this script expects that two GPUs are available" return accelerator def a__ ( __lowercase , __lowercase , __lowercase , __lowercase = False ) -> Union[str, Any]: if iterable: _A = DummyIterableDataset(torch.as_tensor(range(__lowercase ) ) ) else: _A = TensorDataset(torch.as_tensor(range(__lowercase ) ) ) _A = DataLoader(__lowercase , batch_size=__lowercase ) _A = accelerator.prepare(__lowercase ) return dl def a__ ( __lowercase , __lowercase , __lowercase , __lowercase , __lowercase , ) -> Dict: _A = create_dataloader(accelerator=__lowercase , dataset_size=__lowercase , batch_size=__lowercase ) _A = [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 a__ ( ) -> List[str]: _A = create_accelerator() # without padding, we would expect a different number of batches verify_dataloader_batch_sizes( __lowercase , 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( __lowercase , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 2] , ) def a__ ( ) -> List[Any]: _A = create_accelerator(even_batches=__lowercase ) verify_dataloader_batch_sizes( __lowercase , dataset_size=3 , batch_size=1 , process_0_expected_batch_sizes=[1, 1] , process_1_expected_batch_sizes=[1] , ) verify_dataloader_batch_sizes( __lowercase , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 1] , ) def a__ ( ) -> int: _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) _A = [] with accelerator.join_uneven_inputs([ddp_model] ): for batch_idx, batch in enumerate(__lowercase ): _A = ddp_model(batch[0].float() ) _A = output.sum() loss.backward() batch_idxs.append(__lowercase ) accelerator.wait_for_everyone() if accelerator.process_index == 0: assert batch_idxs == [0, 1] elif accelerator.process_index == 1: assert batch_idxs == [0] def a__ ( __lowercase ) -> List[str]: with warnings.catch_warnings(record=__lowercase ) as w: with accelerator.join_uneven_inputs([Mock()] ): pass assert issubclass(w[-1].category , __lowercase ) assert "only supported for multi-GPU" in str(w[-1].message ) def a__ ( ) -> Tuple: _A = True _A = False _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): _A = train_dl.batch_sampler.even_batches _A = 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 a__ ( ) -> int: _A = True _A = False _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) create_dataloader(__lowercase , dataset_size=3 , batch_size=1 , iterable=__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) with warnings.catch_warnings(): warnings.filterwarnings("ignore" ) try: with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): _A = 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 a__ ( ) -> Optional[Any]: _A = create_accelerator() _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) create_dataloader(__lowercase , dataset_size=3 , batch_size=1 , iterable=__lowercase ) with warnings.catch_warnings(record=__lowercase ) as w: with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): pass assert issubclass(w[-1].category , __lowercase ) assert "only supported for map-style datasets" in str(w[-1].message ) def a__ ( ) -> Optional[Any]: _A = 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" ) _A = accelerator.state.distributed_type _A = DistributedType.FSDP test_join_raises_warning_for_non_ddp_distributed(__lowercase ) _A = original_state if __name__ == "__main__": main()
621
0
"""simple docstring""" import inspect import unittest import torch import torch.nn as nn from accelerate.hooks import ( AlignDevicesHook, ModelHook, SequentialHook, add_hook_to_module, attach_align_device_hook, remove_hook_from_module, remove_hook_from_submodules, ) from accelerate.test_utils import require_multi_gpu class snake_case ( nn.Module): def __init__( self : Optional[Any] ) -> Union[str, Any]: '''simple docstring''' super().__init__() _A = nn.Linear(3 , 4 ) _A = nn.BatchNormad(4 ) _A = nn.Linear(4 , 5 ) def a_ ( self : str , a__ : str ) -> Dict: '''simple docstring''' return self.lineara(self.batchnorm(self.lineara(a__ ) ) ) class snake_case ( _UpperCamelCase): def a_ ( self : List[str] , a__ : int , *a__ : Tuple , **a__ : Optional[Any] ) -> Union[str, Any]: '''simple docstring''' return (args[0] + 1,) + args[1:], kwargs class snake_case ( _UpperCamelCase): def a_ ( self : Optional[int] , a__ : int , a__ : Tuple ) -> int: '''simple docstring''' return output + 1 class snake_case ( unittest.TestCase): def a_ ( self : int ) -> List[str]: '''simple docstring''' _A = ModelForTest() _A = ModelHook() add_hook_to_module(a__ , a__ ) self.assertEqual(test_model._hf_hook , a__ ) self.assertTrue(hasattr(a__ , "_old_forward" ) ) # Check adding the hook did not change the name or the signature self.assertEqual(test_model.forward.__name__ , "forward" ) self.assertListEqual(list(inspect.signature(test_model.forward ).parameters ) , ["x"] ) remove_hook_from_module(a__ ) self.assertFalse(hasattr(a__ , "_hf_hook" ) ) self.assertFalse(hasattr(a__ , "_old_forward" ) ) def a_ ( self : List[str] ) -> Optional[int]: '''simple docstring''' _A = ModelForTest() _A = ModelHook() add_hook_to_module(a__ , a__ ) add_hook_to_module(a__ , a__ , append=a__ ) self.assertEqual(isinstance(test_model._hf_hook , a__ ) , a__ ) self.assertEqual(len(test_model._hf_hook.hooks ) , 2 ) self.assertTrue(hasattr(a__ , "_old_forward" ) ) # Check adding the hook did not change the name or the signature self.assertEqual(test_model.forward.__name__ , "forward" ) self.assertListEqual(list(inspect.signature(test_model.forward ).parameters ) , ["x"] ) remove_hook_from_module(a__ ) self.assertFalse(hasattr(a__ , "_hf_hook" ) ) self.assertFalse(hasattr(a__ , "_old_forward" ) ) def a_ ( self : Optional[Any] ) -> Any: '''simple docstring''' _A = ModelForTest() _A = torch.randn(2 , 3 ) _A = test_model(x + 1 ) _A = test_model(x + 2 ) _A = PreForwardHook() add_hook_to_module(a__ , a__ ) _A = test_model(a__ ) self.assertTrue(torch.allclose(a__ , a__ , atol=1E-5 ) ) # Attaching a hook to a model when it already has one replaces, does not chain _A = PreForwardHook() add_hook_to_module(a__ , a__ ) _A = test_model(a__ ) self.assertTrue(torch.allclose(a__ , a__ , atol=1E-5 ) ) # You need to use the sequential hook to chain two or more hooks _A = SequentialHook(PreForwardHook() , PreForwardHook() ) add_hook_to_module(a__ , a__ ) _A = test_model(a__ ) assert torch.allclose(a__ , a__ , atol=1E-5 ) def a_ ( self : Union[str, Any] ) -> int: '''simple docstring''' _A = ModelForTest() _A = torch.randn(2 , 3 ) _A = test_model(a__ ) _A = PostForwardHook() add_hook_to_module(a__ , a__ ) _A = test_model(a__ ) self.assertTrue(torch.allclose(a__ , output + 1 , atol=1E-5 ) ) # Attaching a hook to a model when it already has one replaces, does not chain _A = PostForwardHook() add_hook_to_module(a__ , a__ ) _A = test_model(a__ ) self.assertTrue(torch.allclose(a__ , output + 1 , atol=1E-5 ) ) # You need to use the sequential hook to chain two or more hooks _A = SequentialHook(PostForwardHook() , PostForwardHook() ) add_hook_to_module(a__ , a__ ) _A = test_model(a__ ) assert torch.allclose(a__ , output + 2 , atol=1E-5 ) def a_ ( self : List[Any] ) -> Optional[int]: '''simple docstring''' _A = ModelForTest() _A = torch.randn(2 , 3 ) _A = test_model(a__ ) _A = PostForwardHook() add_hook_to_module(a__ , a__ ) _A = test_model(a__ ) self.assertTrue(torch.allclose(a__ , output + 1 ) ) self.assertTrue(outputa.requires_grad ) _A = True _A = test_model(a__ ) self.assertFalse(outputa.requires_grad ) @require_multi_gpu def a_ ( self : Optional[Any] ) -> Dict: '''simple docstring''' _A = ModelForTest() # Everything is on CPU self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) # This will move each submodule on different devices add_hook_to_module(model.lineara , AlignDevicesHook(execution_device=0 ) ) add_hook_to_module(model.batchnorm , AlignDevicesHook(execution_device=0 ) ) add_hook_to_module(model.lineara , AlignDevicesHook(execution_device=1 ) ) self.assertEqual(model.lineara.weight.device , torch.device(0 ) ) self.assertEqual(model.batchnorm.weight.device , torch.device(0 ) ) self.assertEqual(model.batchnorm.running_mean.device , torch.device(0 ) ) self.assertEqual(model.lineara.weight.device , torch.device(1 ) ) # We can still make a forward pass. The input does not need to be on any particular device _A = torch.randn(2 , 3 ) _A = model(a__ ) self.assertEqual(output.device , torch.device(1 ) ) # We can add a general hook to put back output on same device as input. add_hook_to_module(a__ , AlignDevicesHook(io_same_device=a__ ) ) _A = torch.randn(2 , 3 ).to(0 ) _A = model(a__ ) self.assertEqual(output.device , torch.device(0 ) ) def a_ ( self : Optional[Any] ) -> Optional[int]: '''simple docstring''' _A = ModelForTest() # Everything is on CPU self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) # This will move each submodule on different devices _A = {"execution_device": 0 if torch.cuda.is_available() else "cpu", "offload": True} add_hook_to_module(model.lineara , AlignDevicesHook(**a__ ) ) add_hook_to_module(model.batchnorm , AlignDevicesHook(**a__ ) ) add_hook_to_module(model.lineara , AlignDevicesHook(**a__ ) ) # Parameters have been offloaded, so on the meta device self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("meta" ) ) self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) # Buffers are not included in the offload by default, so are on the execution device _A = torch.device(hook_kwargs["execution_device"] ) self.assertEqual(model.batchnorm.running_mean.device , a__ ) _A = torch.randn(2 , 3 ) _A = model(a__ ) self.assertEqual(output.device , a__ ) # Removing hooks loads back the weights in the model. remove_hook_from_module(model.lineara ) remove_hook_from_module(model.batchnorm ) remove_hook_from_module(model.lineara ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) # Now test with buffers included in the offload _A = { "execution_device": 0 if torch.cuda.is_available() else "cpu", "offload": True, "offload_buffers": True, } add_hook_to_module(model.lineara , AlignDevicesHook(**a__ ) ) add_hook_to_module(model.batchnorm , AlignDevicesHook(**a__ ) ) add_hook_to_module(model.lineara , AlignDevicesHook(**a__ ) ) # Parameters have been offloaded, so on the meta device, buffers included self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("meta" ) ) self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.running_mean.device , torch.device("meta" ) ) _A = torch.randn(2 , 3 ) _A = model(a__ ) self.assertEqual(output.device , a__ ) # Removing hooks loads back the weights in the model. remove_hook_from_module(model.lineara ) remove_hook_from_module(model.batchnorm ) remove_hook_from_module(model.lineara ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) def a_ ( self : List[Any] ) -> Union[str, Any]: '''simple docstring''' _A = ModelForTest() # Everything is on CPU self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) # This will move each submodule on different devices _A = 0 if torch.cuda.is_available() else "cpu" attach_align_device_hook(a__ , execution_device=a__ , offload=a__ ) # Parameters have been offloaded, so on the meta device self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("meta" ) ) self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) # Buffers are not included in the offload by default, so are on the execution device _A = torch.device(a__ ) self.assertEqual(model.batchnorm.running_mean.device , a__ ) _A = torch.randn(2 , 3 ) _A = model(a__ ) self.assertEqual(output.device , a__ ) # Removing hooks loads back the weights in the model. remove_hook_from_submodules(a__ ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) # Now test with buffers included in the offload attach_align_device_hook(a__ , execution_device=a__ , offload=a__ , offload_buffers=a__ ) # Parameters have been offloaded, so on the meta device, buffers included self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("meta" ) ) self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.running_mean.device , torch.device("meta" ) ) _A = torch.randn(2 , 3 ) _A = model(a__ ) self.assertEqual(output.device , a__ ) # Removing hooks loads back the weights in the model. remove_hook_from_submodules(a__ ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) def a_ ( self : str ) -> Optional[Any]: '''simple docstring''' _A = ModelForTest() # Everything is on CPU self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) # This will move each submodule on different devices _A = 0 if torch.cuda.is_available() else "cpu" attach_align_device_hook( a__ , execution_device=a__ , offload=a__ , weights_map=model.state_dict() ) # Parameters have been offloaded, so on the meta device self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("meta" ) ) self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) # Buffers are not included in the offload by default, so are on the execution device _A = torch.device(a__ ) self.assertEqual(model.batchnorm.running_mean.device , a__ ) _A = torch.randn(2 , 3 ) _A = model(a__ ) self.assertEqual(output.device , a__ ) # Removing hooks loads back the weights in the model. remove_hook_from_submodules(a__ ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) # Now test with buffers included in the offload attach_align_device_hook( a__ , execution_device=a__ , offload=a__ , weights_map=model.state_dict() , offload_buffers=a__ , ) # Parameters have been offloaded, so on the meta device, buffers included self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("meta" ) ) self.assertEqual(model.lineara.weight.device , torch.device("meta" ) ) self.assertEqual(model.batchnorm.running_mean.device , torch.device("meta" ) ) _A = torch.randn(2 , 3 ) _A = model(a__ ) self.assertEqual(output.device , a__ ) # Removing hooks loads back the weights in the model. remove_hook_from_submodules(a__ ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) ) self.assertEqual(model.batchnorm.weight.device , torch.device("cpu" ) ) self.assertEqual(model.lineara.weight.device , torch.device("cpu" ) )
714
"""simple docstring""" class snake_case : def __init__( self : Optional[int] , a__ : List[Any] , a__ : List[str] , a__ : Tuple ) -> Optional[Any]: '''simple docstring''' _A = None _A = None _A = graph self._normalize_graph(a__ , a__ ) _A = len(a__ ) _A = None def a_ ( self : str , a__ : List[str] , a__ : List[Any] ) -> Dict: '''simple docstring''' if sources is int: _A = [sources] if sinks is int: _A = [sinks] if len(a__ ) == 0 or len(a__ ) == 0: return _A = sources[0] _A = sinks[0] # make fake vertex if there are more # than one source or sink if len(a__ ) > 1 or len(a__ ) > 1: _A = 0 for i in sources: max_input_flow += sum(self.graph[i] ) _A = len(self.graph ) + 1 for room in self.graph: room.insert(0 , 0 ) self.graph.insert(0 , [0] * size ) for i in sources: _A = max_input_flow _A = 0 _A = len(self.graph ) + 1 for room in self.graph: room.append(0 ) self.graph.append([0] * size ) for i in sinks: _A = max_input_flow _A = size - 1 def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' if self.maximum_flow_algorithm is None: raise Exception("You need to set maximum flow algorithm before." ) if self.source_index is None or self.sink_index is None: return 0 self.maximum_flow_algorithm.execute() return self.maximum_flow_algorithm.getMaximumFlow() def a_ ( self : List[Any] , a__ : Optional[Any] ) -> str: '''simple docstring''' _A = algorithm(self ) class snake_case : def __init__( self : List[str] , a__ : List[str] ) -> Union[str, Any]: '''simple docstring''' _A = flow_network _A = flow_network.verticesCount _A = flow_network.sourceIndex _A = flow_network.sinkIndex # it's just a reference, so you shouldn't change # it in your algorithms, use deep copy before doing that _A = flow_network.graph _A = False def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' if not self.executed: self._algorithm() _A = True def a_ ( self : Any ) -> int: '''simple docstring''' pass class snake_case ( _UpperCamelCase): def __init__( self : Optional[Any] , a__ : Dict ) -> List[str]: '''simple docstring''' super().__init__(a__ ) # use this to save your result _A = -1 def a_ ( self : Any ) -> List[str]: '''simple docstring''' if not self.executed: raise Exception("You should execute algorithm before using its result!" ) return self.maximum_flow class snake_case ( _UpperCamelCase): def __init__( self : Union[str, Any] , a__ : Union[str, Any] ) -> Dict: '''simple docstring''' super().__init__(a__ ) _A = [[0] * self.verticies_count for i in range(self.verticies_count )] _A = [0] * self.verticies_count _A = [0] * self.verticies_count def a_ ( self : Any ) -> Dict: '''simple docstring''' _A = self.verticies_count # push some substance to graph for nextvertex_index, bandwidth in enumerate(self.graph[self.source_index] ): self.preflow[self.source_index][nextvertex_index] += bandwidth self.preflow[nextvertex_index][self.source_index] -= bandwidth self.excesses[nextvertex_index] += bandwidth # Relabel-to-front selection rule _A = [ i for i in range(self.verticies_count ) if i != self.source_index and i != self.sink_index ] # move through list _A = 0 while i < len(a__ ): _A = vertices_list[i] _A = self.heights[vertex_index] self.process_vertex(a__ ) if self.heights[vertex_index] > previous_height: # if it was relabeled, swap elements # and start from 0 index vertices_list.insert(0 , vertices_list.pop(a__ ) ) _A = 0 else: i += 1 _A = sum(self.preflow[self.source_index] ) def a_ ( self : Dict , a__ : Any ) -> Optional[int]: '''simple docstring''' while self.excesses[vertex_index] > 0: for neighbour_index in range(self.verticies_count ): # if it's neighbour and current vertex is higher if ( self.graph[vertex_index][neighbour_index] - self.preflow[vertex_index][neighbour_index] > 0 and self.heights[vertex_index] > self.heights[neighbour_index] ): self.push(a__ , a__ ) self.relabel(a__ ) def a_ ( self : str , a__ : Optional[int] , a__ : List[Any] ) -> Optional[int]: '''simple docstring''' _A = min( self.excesses[from_index] , self.graph[from_index][to_index] - self.preflow[from_index][to_index] , ) self.preflow[from_index][to_index] += preflow_delta self.preflow[to_index][from_index] -= preflow_delta self.excesses[from_index] -= preflow_delta self.excesses[to_index] += preflow_delta def a_ ( self : Any , a__ : Dict ) -> Any: '''simple docstring''' _A = None for to_index in range(self.verticies_count ): if ( self.graph[vertex_index][to_index] - self.preflow[vertex_index][to_index] > 0 ) and (min_height is None or self.heights[to_index] < min_height): _A = self.heights[to_index] if min_height is not None: _A = min_height + 1 if __name__ == "__main__": a_ = [0] a_ = [3] # graph = [ # [0, 0, 4, 6, 0, 0], # [0, 0, 5, 2, 0, 0], # [0, 0, 0, 0, 4, 4], # [0, 0, 0, 0, 6, 6], # [0, 0, 0, 0, 0, 0], # [0, 0, 0, 0, 0, 0], # ] a_ = [[0, 7, 0, 0], [0, 0, 6, 0], [0, 0, 0, 8], [9, 0, 0, 0]] # prepare our network a_ = FlowNetwork(graph, entrances, exits) # set algorithm flow_network.set_maximum_flow_algorithm(PushRelabelExecutor) # and calculate a_ = flow_network.find_maximum_flow() print(f'''maximum flow is {maximum_flow}''')
621
0
"""simple docstring""" def a__ ( __lowercase ) -> tuple[int, int]: try: _A = float(__lowercase ) except ValueError: raise ValueError("Please enter a valid number" ) _A = decimal - int(__lowercase ) if fractional_part == 0: return int(__lowercase ), 1 else: _A = len(str(__lowercase ).split("." )[1] ) _A = int(decimal * (10**number_of_frac_digits) ) _A = 10**number_of_frac_digits _A , _A = denominator, numerator while True: _A = dividend % divisor if remainder == 0: break _A , _A = divisor, remainder _A , _A = numerator / divisor, denominator / divisor return int(__lowercase ), int(__lowercase ) if __name__ == "__main__": print(f'''{decimal_to_fraction(2) = }''') print(f'''{decimal_to_fraction(89.0) = }''') print(f'''{decimal_to_fraction("67") = }''') print(f'''{decimal_to_fraction("45.0") = }''') print(f'''{decimal_to_fraction(1.5) = }''') print(f'''{decimal_to_fraction("6.25") = }''') print(f'''{decimal_to_fraction("78td") = }''')
715
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_tf_available, is_tokenizers_available, is_torch_available, ) a_ = { "configuration_roformer": ["ROFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP", "RoFormerConfig", "RoFormerOnnxConfig"], "tokenization_roformer": ["RoFormerTokenizer"], } try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = ["RoFormerTokenizerFast"] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "RoFormerForCausalLM", "RoFormerForMaskedLM", "RoFormerForMultipleChoice", "RoFormerForQuestionAnswering", "RoFormerForSequenceClassification", "RoFormerForTokenClassification", "RoFormerLayer", "RoFormerModel", "RoFormerPreTrainedModel", "load_tf_weights_in_roformer", ] try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "TF_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "TFRoFormerForCausalLM", "TFRoFormerForMaskedLM", "TFRoFormerForMultipleChoice", "TFRoFormerForQuestionAnswering", "TFRoFormerForSequenceClassification", "TFRoFormerForTokenClassification", "TFRoFormerLayer", "TFRoFormerModel", "TFRoFormerPreTrainedModel", ] try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "FLAX_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "FlaxRoFormerForMaskedLM", "FlaxRoFormerForMultipleChoice", "FlaxRoFormerForQuestionAnswering", "FlaxRoFormerForSequenceClassification", "FlaxRoFormerForTokenClassification", "FlaxRoFormerModel", "FlaxRoFormerPreTrainedModel", ] if TYPE_CHECKING: from .configuration_roformer import ROFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, RoFormerConfig, RoFormerOnnxConfig from .tokenization_roformer import RoFormerTokenizer try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_roformer_fast import RoFormerTokenizerFast try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_roformer import ( ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, RoFormerForCausalLM, RoFormerForMaskedLM, RoFormerForMultipleChoice, RoFormerForQuestionAnswering, RoFormerForSequenceClassification, RoFormerForTokenClassification, RoFormerLayer, RoFormerModel, RoFormerPreTrainedModel, load_tf_weights_in_roformer, ) try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_tf_roformer import ( TF_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, TFRoFormerForCausalLM, TFRoFormerForMaskedLM, TFRoFormerForMultipleChoice, TFRoFormerForQuestionAnswering, TFRoFormerForSequenceClassification, TFRoFormerForTokenClassification, TFRoFormerLayer, TFRoFormerModel, TFRoFormerPreTrainedModel, ) try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_flax_roformer import ( FLAX_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, FlaxRoFormerForMaskedLM, FlaxRoFormerForMultipleChoice, FlaxRoFormerForQuestionAnswering, FlaxRoFormerForSequenceClassification, FlaxRoFormerForTokenClassification, FlaxRoFormerModel, FlaxRoFormerPreTrainedModel, ) else: import sys a_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
621
0
"""simple docstring""" import inspect import unittest from transformers import RegNetConfig, is_flax_available from transformers.testing_utils import require_flax, slow from transformers.utils import cached_property, is_vision_available from ...test_configuration_common import ConfigTester from ...test_modeling_flax_common import FlaxModelTesterMixin, floats_tensor if is_flax_available(): import jax import jax.numpy as jnp from transformers.models.regnet.modeling_flax_regnet import FlaxRegNetForImageClassification, FlaxRegNetModel if is_vision_available(): from PIL import Image from transformers import AutoImageProcessor class snake_case ( unittest.TestCase): def __init__( self : List[Any] , a__ : int , a__ : int=3 , a__ : Tuple=32 , a__ : Optional[Any]=3 , a__ : Optional[Any]=10 , a__ : Optional[int]=[10, 20, 30, 40] , a__ : List[Any]=[1, 1, 2, 1] , a__ : Optional[Any]=True , a__ : List[Any]=True , a__ : Dict="relu" , a__ : int=3 , a__ : Optional[int]=None , ) -> str: '''simple docstring''' _A = parent _A = batch_size _A = image_size _A = num_channels _A = embeddings_size _A = hidden_sizes _A = depths _A = is_training _A = use_labels _A = hidden_act _A = num_labels _A = scope _A = len(a__ ) def a_ ( self : List[Any] ) -> List[str]: '''simple docstring''' _A = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] ) _A = self.get_config() return config, pixel_values def a_ ( self : str ) -> Union[str, Any]: '''simple docstring''' return RegNetConfig( 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 a_ ( self : List[Any] , a__ : Union[str, Any] , a__ : Optional[Any] ) -> Tuple: '''simple docstring''' _A = FlaxRegNetModel(config=a__ ) _A = model(a__ ) # Output shape (b, c, h, w) self.parent.assertEqual( result.last_hidden_state.shape , (self.batch_size, self.hidden_sizes[-1], self.image_size // 32, self.image_size // 32) , ) def a_ ( self : List[Any] , a__ : List[str] , a__ : Optional[Any] ) -> Dict: '''simple docstring''' _A = self.num_labels _A = FlaxRegNetForImageClassification(config=a__ ) _A = model(a__ ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) ) def a_ ( self : Any ) -> int: '''simple docstring''' _A = self.prepare_config_and_inputs() _A , _A = config_and_inputs _A = {"pixel_values": pixel_values} return config, inputs_dict @require_flax class snake_case ( _UpperCamelCase , unittest.TestCase): __UpperCamelCase = (FlaxRegNetModel, FlaxRegNetForImageClassification) if is_flax_available() else () __UpperCamelCase = False __UpperCamelCase = False __UpperCamelCase = False def a_ ( self : Dict ) -> None: '''simple docstring''' _A = FlaxRegNetModelTester(self ) _A = ConfigTester(self , config_class=a__ , has_text_modality=a__ ) def a_ ( self : str ) -> Any: '''simple docstring''' self.create_and_test_config_common_properties() self.config_tester.create_and_test_config_to_json_string() self.config_tester.create_and_test_config_to_json_file() self.config_tester.create_and_test_config_from_and_save_pretrained() self.config_tester.create_and_test_config_with_num_labels() self.config_tester.check_config_can_be_init_without_params() self.config_tester.check_config_arguments_init() def a_ ( self : int ) -> Dict: '''simple docstring''' return def a_ ( self : Optional[int] ) -> Union[str, Any]: '''simple docstring''' _A = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*a__ ) def a_ ( self : Optional[Any] ) -> List[str]: '''simple docstring''' _A = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_image_classification(*a__ ) @unittest.skip(reason="RegNet does not use inputs_embeds" ) def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' pass @unittest.skip(reason="RegNet does not support input and output embeddings" ) def a_ ( self : List[str] ) -> Tuple: '''simple docstring''' pass def a_ ( self : int ) -> int: '''simple docstring''' _A , _A = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: _A = model_class(a__ ) _A = inspect.signature(model.__call__ ) # signature.parameters is an OrderedDict => so arg_names order is deterministic _A = [*signature.parameters.keys()] _A = ["pixel_values"] self.assertListEqual(arg_names[:1] , a__ ) def a_ ( self : str ) -> Optional[Any]: '''simple docstring''' def check_hidden_states_output(a__ : List[Any] , a__ : str , a__ : Union[str, Any] ): _A = model_class(a__ ) _A = model(**self._prepare_for_class(a__ , a__ ) ) _A = outputs.encoder_hidden_states if config.is_encoder_decoder else outputs.hidden_states _A = self.model_tester.num_stages self.assertEqual(len(a__ ) , expected_num_stages + 1 ) _A , _A = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: _A = True check_hidden_states_output(a__ , a__ , a__ ) # check that output_hidden_states also work using config del inputs_dict["output_hidden_states"] _A = True check_hidden_states_output(a__ , a__ , a__ ) def a_ ( self : Union[str, Any] ) -> List[str]: '''simple docstring''' _A , _A = self.model_tester.prepare_config_and_inputs_for_common() for model_class in self.all_model_classes: with self.subTest(model_class.__name__ ): _A = self._prepare_for_class(a__ , a__ ) _A = model_class(a__ ) @jax.jit def model_jitted(a__ : Dict , **a__ : List[Any] ): return model(pixel_values=a__ , **a__ ) with self.subTest("JIT Enabled" ): _A = model_jitted(**a__ ).to_tuple() with self.subTest("JIT Disabled" ): with jax.disable_jit(): _A = model_jitted(**a__ ).to_tuple() self.assertEqual(len(a__ ) , len(a__ ) ) for jitted_output, output in zip(a__ , a__ ): self.assertEqual(jitted_output.shape , output.shape ) def a__ ( ) -> Optional[int]: _A = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" ) return image @require_flax class snake_case ( unittest.TestCase): @cached_property def a_ ( self : Optional[int] ) -> Optional[int]: '''simple docstring''' return AutoImageProcessor.from_pretrained("facebook/regnet-y-040" ) if is_vision_available() else None @slow def a_ ( self : str ) -> List[Any]: '''simple docstring''' _A = FlaxRegNetForImageClassification.from_pretrained("facebook/regnet-y-040" ) _A = self.default_image_processor _A = prepare_img() _A = image_processor(images=a__ , return_tensors="np" ) _A = model(**a__ ) # verify the logits _A = (1, 10_00) self.assertEqual(outputs.logits.shape , a__ ) _A = jnp.array([-0.4_1_8_0, -1.5_0_5_1, -3.4_8_3_6] ) self.assertTrue(jnp.allclose(outputs.logits[0, :3] , a__ , atol=1E-4 ) )
716
"""simple docstring""" import warnings from ...utils import logging from .image_processing_glpn import GLPNImageProcessor a_ = logging.get_logger(__name__) class snake_case ( _UpperCamelCase): def __init__( self : str , *a__ : Dict , **a__ : Optional[int] ) -> None: '''simple docstring''' warnings.warn( "The class GLPNFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please" " use GLPNImageProcessor instead." , a__ , ) super().__init__(*a__ , **a__ )
621
0
"""simple docstring""" import json import os from pathlib import Path import pytest from datasets.download.download_config import DownloadConfig from datasets.download.download_manager import DownloadManager from datasets.utils.file_utils import hash_url_to_filename a_ = "http://www.mocksite.com/file1.txt" a_ = "\"text\": [\"foo\", \"foo\"]" a_ = "6d8ce9aa78a471c7477201efbeabd3bb01ac2e7d100a6dc024ba1608361f90a8" class snake_case : __UpperCamelCase = 200 __UpperCamelCase = {'Content-Length': '100'} __UpperCamelCase = {} def a_ ( self : Dict , **a__ : Tuple ) -> List[str]: '''simple docstring''' return [bytes(a__ , "utf-8" )] def a__ ( *__lowercase , **__lowercase ) -> Any: return MockResponse() @pytest.mark.parametrize("urls_type" , [str, list, dict] ) def a__ ( __lowercase , __lowercase , __lowercase ) -> Tuple: import requests monkeypatch.setattr(__lowercase , "request" , __lowercase ) _A = URL if issubclass(__lowercase , __lowercase ): _A = url elif issubclass(__lowercase , __lowercase ): _A = [url] elif issubclass(__lowercase , __lowercase ): _A = {"train": url} _A = "dummy" _A = "downloads" _A = tmp_path _A = DownloadConfig( cache_dir=os.path.join(__lowercase , __lowercase ) , use_etag=__lowercase , ) _A = DownloadManager(dataset_name=__lowercase , download_config=__lowercase ) _A = dl_manager.download(__lowercase ) _A = urls for downloaded_paths in [downloaded_paths]: if isinstance(__lowercase , __lowercase ): _A = [downloaded_paths] _A = [urls] elif isinstance(__lowercase , __lowercase ): assert "train" in downloaded_paths.keys() _A = downloaded_paths.values() _A = urls.values() assert downloaded_paths for downloaded_path, input_url in zip(__lowercase , __lowercase ): assert downloaded_path == dl_manager.downloaded_paths[input_url] _A = Path(__lowercase ) _A = downloaded_path.parts assert parts[-1] == HASH assert parts[-2] == cache_subdir assert downloaded_path.exists() _A = downloaded_path.read_text() assert content == CONTENT _A = downloaded_path.with_suffix(".json" ) assert metadata_downloaded_path.exists() _A = json.loads(metadata_downloaded_path.read_text() ) assert metadata_content == {"url": URL, "etag": None} @pytest.mark.parametrize("paths_type" , [str, list, dict] ) def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[int]: _A = str(__lowercase ) if issubclass(__lowercase , __lowercase ): _A = filename elif issubclass(__lowercase , __lowercase ): _A = [filename] elif issubclass(__lowercase , __lowercase ): _A = {"train": filename} _A = "dummy" _A = xz_file.parent _A = "extracted" _A = DownloadConfig( cache_dir=__lowercase , use_etag=__lowercase , ) _A = DownloadManager(dataset_name=__lowercase , download_config=__lowercase ) _A = dl_manager.extract(__lowercase ) _A = paths for extracted_paths in [extracted_paths]: if isinstance(__lowercase , __lowercase ): _A = [extracted_paths] _A = [paths] elif isinstance(__lowercase , __lowercase ): assert "train" in extracted_paths.keys() _A = extracted_paths.values() _A = paths.values() assert extracted_paths for extracted_path, input_path in zip(__lowercase , __lowercase ): assert extracted_path == dl_manager.extracted_paths[input_path] _A = Path(__lowercase ) _A = extracted_path.parts assert parts[-1] == hash_url_to_filename(__lowercase , etag=__lowercase ) assert parts[-2] == extracted_subdir assert extracted_path.exists() _A = extracted_path.read_text() _A = text_file.read_text() assert extracted_file_content == expected_file_content def a__ ( __lowercase , __lowercase ) -> Optional[Any]: assert path.endswith(".jsonl" ) for num_items, line in enumerate(__lowercase , start=1 ): _A = json.loads(line.decode("utf-8" ) ) assert item.keys() == {"col_1", "col_2", "col_3"} assert num_items == 4 @pytest.mark.parametrize("archive_jsonl" , ["tar_jsonl_path", "zip_jsonl_path"] ) def a__ ( __lowercase , __lowercase ) -> List[Any]: _A = request.getfixturevalue(__lowercase ) _A = DownloadManager() for num_jsonl, (path, file) in enumerate(dl_manager.iter_archive(__lowercase ) , start=1 ): _test_jsonl(__lowercase , __lowercase ) assert num_jsonl == 2 @pytest.mark.parametrize("archive_nested_jsonl" , ["tar_nested_jsonl_path", "zip_nested_jsonl_path"] ) def a__ ( __lowercase , __lowercase ) -> List[Any]: _A = request.getfixturevalue(__lowercase ) _A = DownloadManager() for num_tar, (path, file) in enumerate(dl_manager.iter_archive(__lowercase ) , start=1 ): for num_jsonl, (subpath, subfile) in enumerate(dl_manager.iter_archive(__lowercase ) , start=1 ): _test_jsonl(__lowercase , __lowercase ) assert num_tar == 1 assert num_jsonl == 2 def a__ ( __lowercase ) -> Dict: _A = DownloadManager() for num_file, file in enumerate(dl_manager.iter_files(__lowercase ) , start=1 ): assert os.path.basename(__lowercase ) == ("test.txt" if num_file == 1 else "train.txt") assert num_file == 2
717
"""simple docstring""" import argparse import torch from torch import nn from transformers import MBartConfig, MBartForConditionalGeneration def a__ ( __lowercase ) -> Optional[int]: _A = [ "encoder.version", "decoder.version", "model.encoder.version", "model.decoder.version", "_float_tensor", "decoder.output_projection.weight", ] for k in ignore_keys: state_dict.pop(__lowercase , __lowercase ) def a__ ( __lowercase ) -> List[Any]: _A , _A = emb.weight.shape _A = nn.Linear(__lowercase , __lowercase , bias=__lowercase ) _A = emb.weight.data return lin_layer def a__ ( __lowercase , __lowercase="facebook/mbart-large-en-ro" , __lowercase=False , __lowercase=False ) -> List[str]: _A = torch.load(__lowercase , map_location="cpu" )["model"] remove_ignore_keys_(__lowercase ) _A = state_dict["encoder.embed_tokens.weight"].shape[0] _A = MBartConfig.from_pretrained(__lowercase , vocab_size=__lowercase ) if mbart_aa and finetuned: _A = "relu" _A = state_dict["decoder.embed_tokens.weight"] _A = MBartForConditionalGeneration(__lowercase ) model.model.load_state_dict(__lowercase ) if finetuned: _A = make_linear_from_emb(model.model.shared ) return model if __name__ == "__main__": a_ = argparse.ArgumentParser() # Required parameters parser.add_argument( "fairseq_path", type=str, help="bart.large, bart.large.cnn or a path to a model.pt on local filesystem." ) parser.add_argument("pytorch_dump_folder_path", default=None, type=str, help="Path to the output PyTorch model.") parser.add_argument( "--hf_config", default="facebook/mbart-large-cc25", type=str, help="Which huggingface architecture to use: mbart-large", ) parser.add_argument("--mbart_50", action="store_true", help="whether the model is mMART-50 checkpoint") parser.add_argument("--finetuned", action="store_true", help="whether the model is a fine-tuned checkpoint") a_ = parser.parse_args() a_ = convert_fairseq_mbart_checkpoint_from_disk( args.fairseq_path, hf_config_path=args.hf_config, finetuned=args.finetuned, mbart_aa=args.mbart_aa ) model.save_pretrained(args.pytorch_dump_folder_path)
621
0
"""simple docstring""" import importlib import shutil import threading import warnings from typing import List import fsspec import fsspec.asyn from . import compression from .hffilesystem import HfFileSystem a_ = importlib.util.find_spec("s3fs") is not None if _has_safs: from .safilesystem import SaFileSystem # noqa: F401 a_ = [ compression.BzaFileSystem, compression.GzipFileSystem, compression.LzaFileSystem, compression.XzFileSystem, compression.ZstdFileSystem, ] # Register custom filesystems for fs_class in COMPRESSION_FILESYSTEMS + [HfFileSystem]: if fs_class.protocol in fsspec.registry and fsspec.registry[fs_class.protocol] is not fs_class: warnings.warn(f'''A filesystem protocol was already set for {fs_class.protocol} and will be overwritten.''') fsspec.register_implementation(fs_class.protocol, fs_class, clobber=True) def a__ ( __lowercase ) -> str: if "://" in dataset_path: _A = dataset_path.split("://" )[1] return dataset_path def a__ ( __lowercase ) -> bool: if fs is not None and fs.protocol != "file": return True else: return False def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = not is_remote_filesystem(__lowercase ) if is_local: # LocalFileSystem.mv does copy + rm, it is more efficient to simply move a local directory shutil.move(fs._strip_protocol(__lowercase ) , fs._strip_protocol(__lowercase ) ) else: fs.mv(__lowercase , __lowercase , recursive=__lowercase ) def a__ ( ) -> None: if hasattr(fsspec.asyn , "reset_lock" ): # for future fsspec>2022.05.0 fsspec.asyn.reset_lock() else: _A = None _A = None _A = threading.Lock()
718
"""simple docstring""" import numpy as np def a__ ( __lowercase , __lowercase ) -> np.ndarray: return np.where(vector > 0 , __lowercase , (alpha * (np.exp(__lowercase ) - 1)) ) if __name__ == "__main__": import doctest doctest.testmod()
621
0
"""simple docstring""" from unittest import TestCase from datasets import Dataset from minhash_deduplication import deduplicate_dataset, make_duplicate_clusters def a__ ( ) -> Tuple: _A = { "repo_name": ["test_repo1", "test_repo2", "test_repo3"], "path": ["test_1.py", "test_2.py", "unit_test.py"], "content": ["a " * 20, "a " * 30, "b " * 7], } _A = Dataset.from_dict(__lowercase ) return dataset class snake_case ( _UpperCamelCase): '''simple docstring''' def a_ ( self : Any ) -> Any: '''simple docstring''' _A = get_dataset() _A = make_duplicate_clusters(a__ , 0.8_5 ) self.assertEqual(len(duplicate_clusters[0] ) , 2 ) def a_ ( self : List[Any] ) -> Tuple: '''simple docstring''' _A = get_dataset() _A , _A = deduplicate_dataset(a__ ) self.assertEqual(len(a__ ) , 2 ) print(a__ ) self.assertEqual(duplicate_clusters[0][0]["copies"] , 2 ) self.assertEqual(duplicate_clusters[0][0]["is_extreme"] , a__ )
719
"""simple docstring""" import os import re import warnings from shutil import copyfile from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import PreTrainedTokenizer if TYPE_CHECKING: from ...tokenization_utils_base import TextInput from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"vocab_file": "spiece.model"} a_ = { "vocab_file": { "t5-small": "https://huggingface.co/t5-small/resolve/main/spiece.model", "t5-base": "https://huggingface.co/t5-base/resolve/main/spiece.model", "t5-large": "https://huggingface.co/t5-large/resolve/main/spiece.model", "t5-3b": "https://huggingface.co/t5-3b/resolve/main/spiece.model", "t5-11b": "https://huggingface.co/t5-11b/resolve/main/spiece.model", } } # TODO(PVP) - this should be removed in Transformers v5 a_ = { "t5-small": 5_12, "t5-base": 5_12, "t5-large": 5_12, "t5-3b": 5_12, "t5-11b": 5_12, } a_ = "▁" class snake_case ( _UpperCamelCase): __UpperCamelCase = VOCAB_FILES_NAMES __UpperCamelCase = PRETRAINED_VOCAB_FILES_MAP __UpperCamelCase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __UpperCamelCase = ['input_ids', 'attention_mask'] def __init__( self : List[str] , a__ : Optional[int] , a__ : Union[str, Any]="</s>" , a__ : Union[str, Any]="<unk>" , a__ : str="<pad>" , a__ : Optional[int]=1_00 , a__ : List[Any]=None , a__ : Optional[Dict[str, Any]] = None , a__ : Any=True , **a__ : Optional[int] , ) -> None: '''simple docstring''' if extra_ids > 0 and additional_special_tokens is None: _A = [F"""<extra_id_{i}>""" for i in range(a__ )] elif extra_ids > 0 and additional_special_tokens is not None: # Check that we have the right number of extra_id special tokens _A = len(set(filter(lambda a__ : bool("extra_id" in str(a__ ) ) , a__ ) ) ) if extra_tokens != extra_ids: raise ValueError( F"""Both extra_ids ({extra_ids}) and additional_special_tokens ({additional_special_tokens}) are""" " provided to T5Tokenizer. In this case the additional_special_tokens must include the extra_ids" " tokens" ) if legacy: logger.warning_once( F"""You are using the legacy behaviour of the {self.__class__}. This means that tokens that come after special tokens will not be properly handled. We recommend you to""" " read the related pull request available at https://github.com/huggingface/transformers/pull/24565" ) _A = legacy _A = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( eos_token=a__ , unk_token=a__ , pad_token=a__ , extra_ids=a__ , additional_special_tokens=a__ , sp_model_kwargs=self.sp_model_kwargs , legacy=a__ , **a__ , ) _A = vocab_file _A = extra_ids _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(a__ ) @staticmethod def a_ ( a__ : List[str] , a__ : Optional[int] , a__ : Tuple ) -> Tuple: '''simple docstring''' if pretrained_model_name_or_path in TaTokenizer.max_model_input_sizes: _A = TaTokenizer.max_model_input_sizes[pretrained_model_name_or_path] if init_max_model_length is not None and init_max_model_length != max_model_length: return init_max_model_length elif init_max_model_length is None: warnings.warn( "This tokenizer was incorrectly instantiated with a model max length of" F""" {deprecated_max_model_length} which will be corrected in Transformers v5.\nFor now, this""" " behavior is kept to avoid breaking backwards compatibility when padding/encoding with" " `truncation is True`.\n- Be aware that you SHOULD NOT rely on" F""" {pretrained_model_name_or_path} automatically truncating your input to""" F""" {deprecated_max_model_length} when padding/encoding.\n- If you want to encode/pad to sequences""" F""" longer than {deprecated_max_model_length} you can either instantiate this tokenizer with""" " `model_max_length` or pass `max_length` when encoding/padding.\n- To avoid this warning, please" " instantiate this tokenizer with `model_max_length` set to your preferred value." , a__ , ) return max_model_length @property def a_ ( self : List[Any] ) -> Dict: '''simple docstring''' return self.sp_model.get_piece_size() + self._extra_ids def a_ ( self : Dict ) -> Optional[Any]: '''simple docstring''' _A = {self.convert_ids_to_tokens(a__ ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def a_ ( self : Optional[Any] , a__ : List[int] , a__ : Optional[List[int]] = None , a__ : bool = False ) -> List[int]: '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=a__ , token_ids_a=a__ , already_has_special_tokens=a__ ) # normal case: some special tokens if token_ids_a is None: return ([0] * len(a__ )) + [1] return ([0] * len(a__ )) + [1] + ([0] * len(a__ )) + [1] def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' return list( set(filter(lambda a__ : bool(re.search(r"<extra_id_\d+>" , a__ ) ) is not None , self.additional_special_tokens ) ) ) def a_ ( self : str ) -> List[Any]: '''simple docstring''' return [self._convert_token_to_id(a__ ) for token in self.get_sentinel_tokens()] def a_ ( self : List[Any] , a__ : List[int] ) -> List[int]: '''simple docstring''' if len(a__ ) > 0 and token_ids[-1] == self.eos_token_id: warnings.warn( F"""This sequence already has {self.eos_token}. In future versions this behavior may lead to duplicated""" " eos tokens being added." ) return token_ids else: return token_ids + [self.eos_token_id] def a_ ( self : int , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = [self.eos_token_id] if token_ids_a is None: return len(token_ids_a + eos ) * [0] return len(token_ids_a + eos + token_ids_a + eos ) * [0] def a_ ( self : Union[str, Any] , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = self._add_eos_if_not_present(a__ ) if token_ids_a is None: return token_ids_a else: _A = self._add_eos_if_not_present(a__ ) return token_ids_a + token_ids_a def __getstate__( self : Dict ) -> Union[str, Any]: '''simple docstring''' _A = self.__dict__.copy() _A = None return state def __setstate__( self : int , a__ : Optional[int] ) -> Union[str, Any]: '''simple docstring''' _A = d # for backward compatibility if not hasattr(self , "sp_model_kwargs" ): _A = {} _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def a_ ( self : int , a__ : "TextInput" , **a__ : List[str] ) -> List[str]: '''simple docstring''' if not self.legacy: _A = SPIECE_UNDERLINE + text.replace(a__ , " " ) return super().tokenize(a__ , **a__ ) def a_ ( self : str , a__ : Dict , **a__ : Optional[int] ) -> Any: '''simple docstring''' if not self.legacy: _A = text.startswith(a__ ) if is_first: _A = text[1:] _A = self.sp_model.encode(a__ , out_type=a__ ) if not self.legacy and not is_first and not text.startswith(" " ) and tokens[0].startswith(a__ ): _A = ([tokens[0][1:]] if len(tokens[0] ) > 1 else []) + tokens[1:] return tokens def a_ ( self : int , a__ : List[Any] ) -> List[str]: '''simple docstring''' if token.startswith("<extra_id_" ): _A = re.match(r"<extra_id_(\d+)>" , a__ ) _A = int(match.group(1 ) ) return self.vocab_size - num - 1 return self.sp_model.piece_to_id(a__ ) def a_ ( self : Dict , a__ : Union[str, Any] ) -> Any: '''simple docstring''' if index < self.sp_model.get_piece_size(): _A = self.sp_model.IdToPiece(a__ ) else: _A = F"""<extra_id_{self.vocab_size - 1 - index}>""" return token def a_ ( self : Optional[int] , a__ : Tuple ) -> List[str]: '''simple docstring''' _A = [] _A = "" _A = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(a__ ) + token _A = True _A = [] else: current_sub_tokens.append(a__ ) _A = False out_string += self.sp_model.decode(a__ ) return out_string.strip() def a_ ( self : Dict , a__ : str , a__ : Optional[str] = None ) -> Tuple[str]: '''simple docstring''' if not os.path.isdir(a__ ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return _A = os.path.join( a__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(a__ ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , a__ ) elif not os.path.isfile(self.vocab_file ): with open(a__ , "wb" ) as fi: _A = self.sp_model.serialized_model_proto() fi.write(a__ ) return (out_vocab_file,)
621
0
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_sentencepiece_available, is_tokenizers_available, is_torch_available, ) a_ = {"configuration_fnet": ["FNET_PRETRAINED_CONFIG_ARCHIVE_MAP", "FNetConfig"]} try: if not is_sentencepiece_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = ["FNetTokenizer"] try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = ["FNetTokenizerFast"] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "FNET_PRETRAINED_MODEL_ARCHIVE_LIST", "FNetForMaskedLM", "FNetForMultipleChoice", "FNetForNextSentencePrediction", "FNetForPreTraining", "FNetForQuestionAnswering", "FNetForSequenceClassification", "FNetForTokenClassification", "FNetLayer", "FNetModel", "FNetPreTrainedModel", ] if TYPE_CHECKING: from .configuration_fnet import FNET_PRETRAINED_CONFIG_ARCHIVE_MAP, FNetConfig try: if not is_sentencepiece_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_fnet import FNetTokenizer try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_fnet_fast import FNetTokenizerFast try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_fnet import ( FNET_PRETRAINED_MODEL_ARCHIVE_LIST, FNetForMaskedLM, FNetForMultipleChoice, FNetForNextSentencePrediction, FNetForPreTraining, FNetForQuestionAnswering, FNetForSequenceClassification, FNetForTokenClassification, FNetLayer, FNetModel, FNetPreTrainedModel, ) else: import sys a_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
720
"""simple docstring""" import argparse import json import os from collections import OrderedDict import numpy as np import tensorflow as tf import torch def a__ ( __lowercase ) -> List[Any]: _A = os.path.join(args.tf_model_dir , "parameters.json" ) _A = json.loads(open(__lowercase ).read() ) if not params: raise ValueError( f"""It seems that the json file at {parameter_file} is empty. Make sure you have a correct json file.""" ) if not args.output.endswith(".pt" ): _A = args.output + ".pt" _A = OrderedDict() with tf.device("/CPU:0" ): _A = tf.train.load_checkpoint(args.tf_model_dir ) _A = reader.get_variable_to_shape_map() for key_name in shapes.keys(): _A = reader.get_tensor(__lowercase ).astype(np.floataa ) if key_name.endswith("/adam_m" ) or key_name.endswith("/adam_v" ): continue if key_name.startswith("pasts/" ): if key_name.startswith("pasts/mlp" ): _A = int(key_name[9] ) elif key_name.startswith("pasts/out" ): _A = 8 _A = "model.sqout.%d.weight" % (player * 2) # enter to nn.Sequencial with Tanh, so 2 at a time _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/moe" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/switch_gating/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.router.classifier.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/softmlp/kernel" ): _A = "model.blocks.%d.feed_forward.soft_bypass_mlp.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/wo/kernel" ) or key_name.endswith("/wi/kernel" ): _A = key_name[-9:-7] for i in range(16 ): _A = "model.blocks.%d.feed_forward.mlp.experts.expert_%d.%s.weight" % (player, i, nlayer) _A = ( vnp[i].transpose([1, 0] ).copy() ) # In Mesh-Tensorflow, it is one array, so it is divided _A = torch.tensor(__lowercase ) elif key_name.startswith("model/mlp" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/p1/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wi.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p1/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wi.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wo.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wo.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/ln" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.feed_forward.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.feed_forward.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/att" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/qkv/kernel" ): _A = vnp.copy() # Compute same dimension as Mesh-tensorflow using einsum _A = state[:, 0, :, :] _A = state[:, 1, :, :] _A = state[:, 2, :, :] _A = ( state_q.reshape([state_q.shape[0], state_q.shape[1] * state_q.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_k.reshape([state_k.shape[0], state_k.shape[1] * state_k.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_v.reshape([state_v.shape[0], state_v.shape[1] * state_v.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = "model.blocks.%d.self_attn.self_attn.q_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.k_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.v_proj.weight" % player _A = torch.tensor(__lowercase ) elif key_name.endswith("/o/kernel" ): _A = "model.blocks.%d.self_attn.self_attn.out_proj.weight" % player _A = ( vnp.reshape([vnp.shape[0] * vnp.shape[1], vnp.shape[2]] ).transpose([1, 0] ).copy() ) # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/an" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.self_attn.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.self_attn.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif ( key_name.startswith("model/wte" ) or key_name.startswith("model/wpe" ) or key_name.startswith("model/ete" ) ): _A = {"wte": "embed_tokens", "wpe": "position_embeddings", "ete": "extra_position_embeddings"}[ key_name[-3:] ] _A = "model.%s.weight" % nlayer _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) if key_name.startswith("model/wte" ): _A = "lm_head.weight" _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) elif key_name.startswith("model/wob" ): _A = "final_logits_bias" _A = vnp.copy() # same in embedded _A = state.reshape((1, -1) ) _A = torch.tensor(__lowercase ) elif key_name == "model/dense/kernel": _A = "model.last_project.weight" _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name == "model/dense_1/bias": _A = "model.last_project.bias" _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) torch.save(__lowercase , args.output ) if __name__ == "__main__": a_ = argparse.ArgumentParser( description="model converter.", formatter_class=argparse.ArgumentDefaultsHelpFormatter ) parser.add_argument("--tf_model_dir", metavar="PATH", type=str, required=True, help="import model") parser.add_argument("--output", metavar="PATH", type=str, required=True, help="output model") a_ = parser.parse_args() convert_tf_gptsan_to_pt(args)
621
0
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available a_ = {"configuration_yolos": ["YOLOS_PRETRAINED_CONFIG_ARCHIVE_MAP", "YolosConfig", "YolosOnnxConfig"]} try: if not is_vision_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = ["YolosFeatureExtractor"] a_ = ["YolosImageProcessor"] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "YOLOS_PRETRAINED_MODEL_ARCHIVE_LIST", "YolosForObjectDetection", "YolosModel", "YolosPreTrainedModel", ] if TYPE_CHECKING: from .configuration_yolos import YOLOS_PRETRAINED_CONFIG_ARCHIVE_MAP, YolosConfig, YolosOnnxConfig try: if not is_vision_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .feature_extraction_yolos import YolosFeatureExtractor from .image_processing_yolos import YolosImageProcessor try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_yolos import ( YOLOS_PRETRAINED_MODEL_ARCHIVE_LIST, YolosForObjectDetection, YolosModel, YolosPreTrainedModel, ) else: import sys a_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
721
"""simple docstring""" import argparse import torch from transformers import GPTaLMHeadModel, RobertaForMaskedLM if __name__ == "__main__": a_ = argparse.ArgumentParser( description=( "Extraction some layers of the full RobertaForMaskedLM or GPT2LMHeadModel for Transfer Learned" " Distillation" ) ) parser.add_argument("--model_type", default="roberta", choices=["roberta", "gpt2"]) parser.add_argument("--model_name", default="roberta-large", type=str) parser.add_argument("--dump_checkpoint", default="serialization_dir/tf_roberta_048131723.pth", type=str) parser.add_argument("--vocab_transform", action="store_true") a_ = parser.parse_args() if args.model_type == "roberta": a_ = RobertaForMaskedLM.from_pretrained(args.model_name) a_ = "roberta" elif args.model_type == "gpt2": a_ = GPTaLMHeadModel.from_pretrained(args.model_name) a_ = "transformer" a_ = model.state_dict() a_ = {} # Embeddings # if args.model_type == "gpt2": for param_name in ["wte.weight", "wpe.weight"]: a_ = state_dict[f'''{prefix}.{param_name}'''] else: for w in ["word_embeddings", "position_embeddings", "token_type_embeddings"]: a_ = f'''{prefix}.embeddings.{w}.weight''' a_ = state_dict[param_name] for w in ["weight", "bias"]: a_ = f'''{prefix}.embeddings.LayerNorm.{w}''' a_ = state_dict[param_name] # Transformer Blocks # a_ = 0 for teacher_idx in [0, 2, 4, 7, 9, 11]: if args.model_type == "gpt2": for layer in ["ln_1", "attn.c_attn", "attn.c_proj", "ln_2", "mlp.c_fc", "mlp.c_proj"]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.h.{teacher_idx}.{layer}.{w}''' ] a_ = state_dict[f'''{prefix}.h.{teacher_idx}.attn.bias'''] else: for layer in [ "attention.self.query", "attention.self.key", "attention.self.value", "attention.output.dense", "attention.output.LayerNorm", "intermediate.dense", "output.dense", "output.LayerNorm", ]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.encoder.layer.{teacher_idx}.{layer}.{w}''' ] std_idx += 1 # Language Modeling Head ###s if args.model_type == "roberta": for layer in ["lm_head.decoder.weight", "lm_head.bias"]: a_ = state_dict[f'''{layer}'''] if args.vocab_transform: for w in ["weight", "bias"]: a_ = state_dict[f'''lm_head.dense.{w}'''] a_ = state_dict[f'''lm_head.layer_norm.{w}'''] elif args.model_type == "gpt2": for w in ["weight", "bias"]: a_ = state_dict[f'''{prefix}.ln_f.{w}'''] a_ = state_dict["lm_head.weight"] print(f'''N layers selected for distillation: {std_idx}''') print(f'''Number of params transferred for distillation: {len(compressed_sd.keys())}''') print(f'''Save transferred checkpoint to {args.dump_checkpoint}.''') torch.save(compressed_sd, args.dump_checkpoint)
621
0
"""simple docstring""" import torch from diffusers import DPMSolverSDEScheduler from diffusers.utils import torch_device from diffusers.utils.testing_utils import require_torchsde from .test_schedulers import SchedulerCommonTest @require_torchsde class snake_case ( _UpperCamelCase): __UpperCamelCase = (DPMSolverSDEScheduler,) __UpperCamelCase = 10 def a_ ( self : int , **a__ : Optional[int] ) -> Dict: '''simple docstring''' _A = { "num_train_timesteps": 11_00, "beta_start": 0.0_0_0_1, "beta_end": 0.0_2, "beta_schedule": "linear", "noise_sampler_seed": 0, } config.update(**a__ ) return config def a_ ( self : Dict ) -> Dict: '''simple docstring''' for timesteps in [10, 50, 1_00, 10_00]: self.check_over_configs(num_train_timesteps=a__ ) def a_ ( self : Dict ) -> Tuple: '''simple docstring''' 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=a__ , beta_end=a__ ) def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' for schedule in ["linear", "scaled_linear"]: self.check_over_configs(beta_schedule=a__ ) def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' for prediction_type in ["epsilon", "v_prediction"]: self.check_over_configs(prediction_type=a__ ) def a_ ( self : Union[str, Any] ) -> List[str]: '''simple docstring''' _A = self.scheduler_classes[0] _A = self.get_scheduler_config() _A = scheduler_class(**a__ ) scheduler.set_timesteps(self.num_inference_steps ) _A = self.dummy_model() _A = self.dummy_sample_deter * scheduler.init_noise_sigma _A = sample.to(a__ ) for i, t in enumerate(scheduler.timesteps ): _A = scheduler.scale_model_input(a__ , a__ ) _A = model(a__ , a__ ) _A = scheduler.step(a__ , a__ , a__ ) _A = output.prev_sample _A = torch.sum(torch.abs(a__ ) ) _A = torch.mean(torch.abs(a__ ) ) if torch_device in ["mps"]: assert abs(result_sum.item() - 1_67.47_82_10_44_92_18_75 ) < 1E-2 assert abs(result_mean.item() - 0.2_1_7_8_7_0_5_9_6_4_5_6_5_2_7_7 ) < 1E-3 elif torch_device in ["cuda"]: assert abs(result_sum.item() - 1_71.59_35_21_11_81_64_06 ) < 1E-2 assert abs(result_mean.item() - 0.2_2_3_4_2_9_0_6_8_9_2_2_9_9_6_5_2 ) < 1E-3 else: assert abs(result_sum.item() - 1_62.52_38_34_22_85_15_62 ) < 1E-2 assert abs(result_mean.item() - 0.2_1_1_6_1_9_5_7_0_8_5_1_3_2_6 ) < 1E-3 def a_ ( self : Optional[int] ) -> str: '''simple docstring''' _A = self.scheduler_classes[0] _A = self.get_scheduler_config(prediction_type="v_prediction" ) _A = scheduler_class(**a__ ) scheduler.set_timesteps(self.num_inference_steps ) _A = self.dummy_model() _A = self.dummy_sample_deter * scheduler.init_noise_sigma _A = sample.to(a__ ) for i, t in enumerate(scheduler.timesteps ): _A = scheduler.scale_model_input(a__ , a__ ) _A = model(a__ , a__ ) _A = scheduler.step(a__ , a__ , a__ ) _A = output.prev_sample _A = torch.sum(torch.abs(a__ ) ) _A = torch.mean(torch.abs(a__ ) ) if torch_device in ["mps"]: assert abs(result_sum.item() - 1_24.77_14_92_00_43_94_53 ) < 1E-2 assert abs(result_mean.item() - 0.1_6_2_2_6_2_8_9_0_1_4_8_1_6_2_8_4 ) < 1E-3 elif torch_device in ["cuda"]: assert abs(result_sum.item() - 1_28.1_66_33_60_59_57_03 ) < 1E-2 assert abs(result_mean.item() - 0.1_6_6_8_8_3_2_6_0_0_1_1_6_7_2_9_7 ) < 1E-3 else: assert abs(result_sum.item() - 1_19.8_48_75_48_82_81_25 ) < 1E-2 assert abs(result_mean.item() - 0.1_5_6_0_5_3_0_6_6_2_5_3_6_6_2_1 ) < 1E-3 def a_ ( self : List[str] ) -> Any: '''simple docstring''' _A = self.scheduler_classes[0] _A = self.get_scheduler_config() _A = scheduler_class(**a__ ) scheduler.set_timesteps(self.num_inference_steps , device=a__ ) _A = self.dummy_model() _A = self.dummy_sample_deter.to(a__ ) * scheduler.init_noise_sigma for t in scheduler.timesteps: _A = scheduler.scale_model_input(a__ , a__ ) _A = model(a__ , a__ ) _A = scheduler.step(a__ , a__ , a__ ) _A = output.prev_sample _A = torch.sum(torch.abs(a__ ) ) _A = torch.mean(torch.abs(a__ ) ) if torch_device in ["mps"]: assert abs(result_sum.item() - 1_67.46_95_73_97_46_09_38 ) < 1E-2 assert abs(result_mean.item() - 0.2_1_8_0_5_9_3_4_6_0_7_9_8_2_6_3_5 ) < 1E-3 elif torch_device in ["cuda"]: assert abs(result_sum.item() - 1_71.59_35_36_37_69_53_12 ) < 1E-2 assert abs(result_mean.item() - 0.2_2_3_4_2_9_0_8_3_8_2_4_1_5_7_7_1 ) < 1E-3 else: assert abs(result_sum.item() - 1_62.52_38_34_22_85_15_62 ) < 1E-2 assert abs(result_mean.item() - 0.2_1_1_6_1_9_5_7_0_8_5_1_3_2_6 ) < 1E-3 def a_ ( self : int ) -> Optional[int]: '''simple docstring''' _A = self.scheduler_classes[0] _A = self.get_scheduler_config() _A = scheduler_class(**a__ , use_karras_sigmas=a__ ) scheduler.set_timesteps(self.num_inference_steps , device=a__ ) _A = self.dummy_model() _A = self.dummy_sample_deter.to(a__ ) * scheduler.init_noise_sigma _A = sample.to(a__ ) for t in scheduler.timesteps: _A = scheduler.scale_model_input(a__ , a__ ) _A = model(a__ , a__ ) _A = scheduler.step(a__ , a__ , a__ ) _A = output.prev_sample _A = torch.sum(torch.abs(a__ ) ) _A = torch.mean(torch.abs(a__ ) ) if torch_device in ["mps"]: assert abs(result_sum.item() - 1_76.66_97_41_35_74_21_88 ) < 1E-2 assert abs(result_mean.item() - 0.2_3_0_0_3_8_7_2_7_3_0_9_8_1_8_1_1 ) < 1E-2 elif torch_device in ["cuda"]: assert abs(result_sum.item() - 1_77.63_65_35_64_45_31_25 ) < 1E-2 assert abs(result_mean.item() - 0.2_3_0_0_3_8_7_2_7_3_0_9_8_1_8_1_1 ) < 1E-2 else: assert abs(result_sum.item() - 1_70.3_13_52_23_38_86_72 ) < 1E-2 assert abs(result_mean.item() - 0.2_3_0_0_3_8_7_2_7_3_0_9_8_1_8_1_1 ) < 1E-2
700
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a_ = { "configuration_upernet": ["UperNetConfig"], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "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_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
621
0
"""simple docstring""" def a__ ( __lowercase = 1000 ) -> int: return sum(e for e in range(3 , __lowercase ) if e % 3 == 0 or e % 5 == 0 ) if __name__ == "__main__": print(f'''{solution() = }''')
701
"""simple docstring""" import random import unittest from torch.utils.data import BatchSampler, DataLoader, IterableDataset from accelerate import Accelerator from accelerate.data_loader import ( BatchSamplerShard, DataLoaderDispatcher, DataLoaderShard, IterableDatasetShard, SkipBatchSampler, SkipDataLoader, skip_first_batches, ) class snake_case ( _UpperCamelCase): def __init__( self : Optional[int] , a__ : str=0.0_1 , a__ : str=10_00 ) -> int: '''simple docstring''' _A = p_stop _A = max_length def __iter__( self : Any ) -> Optional[Any]: '''simple docstring''' _A = 0 _A = False while not stop and count < self.max_length: yield count count += 1 _A = random.random() < self.p_stop class snake_case ( unittest.TestCase): def a_ ( self : List[Any] , a__ : Union[str, Any] , a__ : Union[str, Any] , a__ : List[str]=False , a__ : str=True ) -> Union[str, Any]: '''simple docstring''' _A = [ BatchSamplerShard(a__ , 2 , a__ , split_batches=a__ , even_batches=a__ ) for i in range(2 ) ] _A = [list(a__ ) for batch_sampler_shard in batch_sampler_shards] if not split_batches: self.assertListEqual([len(a__ ) for shard in batch_sampler_shards] , [len(a__ ) for e in expected] ) self.assertListEqual(a__ , a__ ) def a_ ( self : List[Any] ) -> str: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 22, 23]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is a round multiple of batch size but not total batch size. _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [0, 1, 2]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is not a round multiple of batch size but has a multiple of # num_processes batch. _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 0, 1]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is not a round multiple of batch size but and has not a multiple of # num_processes batch. _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 0]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [1, 2, 3]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[[0, 1, 0]], [[1, 0, 1]]] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ ) def a_ ( self : int ) -> int: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size. _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [0, 1]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size or num_processes. _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 0]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [1, 2]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[[0, 1]], [[0, 1]]] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is a round multiple of batch size but not total batch size. _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size but has a multiple of # num_processes batch. _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size but and has not a multiple of # num_processes batch. _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[[0, 1]], []] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) def a_ ( self : List[str] ) -> str: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size. _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size or num_processes. _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[[0, 1]], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) def a_ ( self : Union[str, Any] ) -> str: '''simple docstring''' _A = [[0, 1, 2], [3, 4], [5, 6, 7, 8], [9, 10, 11], [12, 13]] _A = [BatchSamplerShard(a__ , 2 , a__ , even_batches=a__ ) for i in range(2 )] self.assertEqual(len(batch_sampler_shards[0] ) , 3 ) self.assertEqual(len(batch_sampler_shards[1] ) , 2 ) self.assertListEqual(list(batch_sampler_shards[0] ) , [[0, 1, 2], [5, 6, 7, 8], [12, 13]] ) self.assertListEqual(list(batch_sampler_shards[1] ) , [[3, 4], [9, 10, 11]] ) def a_ ( self : Optional[int] , a__ : Optional[int] , a__ : Tuple , a__ : Optional[int] , a__ : Union[str, Any]=False , a__ : int=2 , a__ : List[Any]=False ) -> str: '''simple docstring''' random.seed(a__ ) _A = list(a__ ) _A = [ IterableDatasetShard( a__ , batch_size=a__ , drop_last=a__ , num_processes=a__ , process_index=a__ , split_batches=a__ , ) for i in range(a__ ) ] _A = [] for iterable_dataset_shard in iterable_dataset_shards: # Since our random iterable dataset will be... random... we need to use a seed to get reproducible results. random.seed(a__ ) iterable_dataset_lists.append(list(a__ ) ) _A = batch_size // num_processes if split_batches else batch_size # All iterable dataset shard should have the same length, a round multiple of shard_batch_size _A = iterable_dataset_lists[0] for l in iterable_dataset_lists[1:]: self.assertEqual(len(a__ ) , len(a__ ) ) self.assertTrue(len(a__ ) % shard_batch_size == 0 ) _A = [] for idx in range(0 , len(a__ ) , a__ ): for l in iterable_dataset_lists: observed += l[idx : idx + shard_batch_size] if not drop_last: while len(a__ ) < len(a__ ): reference += reference self.assertListEqual(a__ , reference[: len(a__ )] ) def a_ ( self : List[str] ) -> List[Any]: '''simple docstring''' _A = 42 _A = RandomIterableDataset() self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) # Edge case with a very small dataset _A = RandomIterableDataset(max_length=2 ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) def a_ ( self : List[str] ) -> Dict: '''simple docstring''' _A = BatchSampler(range(16 ) , batch_size=4 , drop_last=a__ ) _A = SkipBatchSampler(a__ , 2 ) self.assertListEqual(list(a__ ) , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : int ) -> Union[str, Any]: '''simple docstring''' _A = SkipDataLoader(list(range(16 ) ) , batch_size=4 , skip_batches=2 ) self.assertListEqual([t.tolist() for t in dataloader] , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : int ) -> Optional[int]: '''simple docstring''' _A = DataLoader(list(range(16 ) ) , batch_size=4 ) _A = skip_first_batches(a__ , num_batches=2 ) self.assertListEqual([t.tolist() for t in new_dataloader] , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : Optional[Any] ) -> Optional[int]: '''simple docstring''' _A = DataLoaderShard(list(range(16 ) ) , batch_size=4 ) for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) # Test it also works on the second iteration for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) def a_ ( self : int ) -> int: '''simple docstring''' Accelerator() _A = DataLoaderDispatcher(range(16 ) , batch_size=4 ) for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) # Test it also works on the second iteration for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 )
621
0
"""simple docstring""" import hashlib import unittest from transformers import MODEL_FOR_DEPTH_ESTIMATION_MAPPING, is_torch_available, is_vision_available from transformers.pipelines import DepthEstimationPipeline, pipeline from transformers.testing_utils import ( is_pipeline_test, nested_simplify, require_tf, require_timm, require_torch, require_vision, slow, ) from .test_pipelines_common import ANY if is_torch_available(): import torch if is_vision_available(): from PIL import Image else: class snake_case : @staticmethod def a_ ( *a__ : Tuple , **a__ : List[str] ) -> Tuple: '''simple docstring''' pass def a__ ( __lowercase ) -> str: _A = hashlib.mda(image.tobytes() ) return m.hexdigest() @is_pipeline_test @require_vision @require_timm @require_torch class snake_case ( unittest.TestCase): __UpperCamelCase = MODEL_FOR_DEPTH_ESTIMATION_MAPPING def a_ ( self : Dict , a__ : Tuple , a__ : Optional[int] , a__ : str ) -> Optional[Any]: '''simple docstring''' _A = DepthEstimationPipeline(model=a__ , image_processor=a__ ) return depth_estimator, [ "./tests/fixtures/tests_samples/COCO/000000039769.png", "./tests/fixtures/tests_samples/COCO/000000039769.png", ] def a_ ( self : Tuple , a__ : Optional[int] , a__ : Dict ) -> int: '''simple docstring''' _A = depth_estimator("./tests/fixtures/tests_samples/COCO/000000039769.png" ) self.assertEqual({"predicted_depth": ANY(torch.Tensor ), "depth": ANY(Image.Image )} , a__ ) import datasets _A = datasets.load_dataset("hf-internal-testing/fixtures_image_utils" , "image" , split="test" ) _A = depth_estimator( [ Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" ), "http://images.cocodataset.org/val2017/000000039769.jpg", # RGBA dataset[0]["file"], # LA dataset[1]["file"], # L dataset[2]["file"], ] ) self.assertEqual( [ {"predicted_depth": ANY(torch.Tensor ), "depth": ANY(Image.Image )}, {"predicted_depth": ANY(torch.Tensor ), "depth": ANY(Image.Image )}, {"predicted_depth": ANY(torch.Tensor ), "depth": ANY(Image.Image )}, {"predicted_depth": ANY(torch.Tensor ), "depth": ANY(Image.Image )}, {"predicted_depth": ANY(torch.Tensor ), "depth": ANY(Image.Image )}, ] , a__ , ) @require_tf @unittest.skip("Depth estimation is not implemented in TF" ) def a_ ( self : List[str] ) -> Optional[int]: '''simple docstring''' pass @slow @require_torch def a_ ( self : Optional[Any] ) -> str: '''simple docstring''' _A = "Intel/dpt-large" _A = pipeline("depth-estimation" , model=a__ ) _A = depth_estimator("http://images.cocodataset.org/val2017/000000039769.jpg" ) _A = hashimage(outputs["depth"] ) # This seems flaky. # self.assertEqual(outputs["depth"], "1a39394e282e9f3b0741a90b9f108977") self.assertEqual(nested_simplify(outputs["predicted_depth"].max().item() ) , 29.3_04 ) self.assertEqual(nested_simplify(outputs["predicted_depth"].min().item() ) , 2.6_6_2 ) @require_torch def a_ ( self : Optional[Any] ) -> Optional[int]: '''simple docstring''' self.skipTest("There is not hf-internal-testing tiny model for either GLPN nor DPT" )
702
"""simple docstring""" import gc import tempfile import unittest import numpy as np import torch from diffusers import VersatileDiffusionPipeline from diffusers.utils.testing_utils import load_image, nightly, require_torch_gpu, torch_device a_ = False class snake_case ( unittest.TestCase): pass @nightly @require_torch_gpu class snake_case ( unittest.TestCase): def a_ ( self : Optional[int] ) -> str: '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def a_ ( self : Tuple ) -> Any: '''simple docstring''' _A = VersatileDiffusionPipeline.from_pretrained("shi-labs/versatile-diffusion" , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg" ) _A = torch.manual_seed(0 ) _A = pipe.dual_guided( prompt="first prompt" , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" , ).images with tempfile.TemporaryDirectory() as tmpdirname: pipe.save_pretrained(a__ ) _A = VersatileDiffusionPipeline.from_pretrained(a__ , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = generator.manual_seed(0 ) _A = pipe.dual_guided( prompt="first prompt" , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" , ).images assert np.abs(image - new_image ).sum() < 1E-5, "Models don't have the same forward pass" def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' _A = VersatileDiffusionPipeline.from_pretrained("shi-labs/versatile-diffusion" , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = "cyberpunk 2077" _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg" ) _A = torch.manual_seed(0 ) _A = pipe.dual_guided( prompt=a__ , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=50 , output_type="numpy" , ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.1_4_4_8, 0.1_6_1_9, 0.1_7_4_1, 0.1_0_8_6, 0.1_1_4_7, 0.1_1_2_8, 0.1_1_9_9, 0.1_1_6_5, 0.1_0_0_1] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 _A = "A painting of a squirrel eating a burger " _A = torch.manual_seed(0 ) _A = pipe.text_to_image( prompt=a__ , generator=a__ , guidance_scale=7.5 , num_inference_steps=50 , output_type="numpy" ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.3_3_6_7, 0.3_1_6_9, 0.2_6_5_6, 0.3_8_7_0, 0.4_7_9_0, 0.3_7_9_6, 0.4_0_0_9, 0.4_8_7_8, 0.4_7_7_8] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 _A = pipe.image_variation(a__ , generator=a__ , output_type="numpy" ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.3_0_7_6, 0.3_1_2_3, 0.3_2_8_4, 0.3_7_8_2, 0.3_7_7_0, 0.3_8_9_4, 0.4_2_9_7, 0.4_3_3_1, 0.4_4_5_6] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1
621
0
"""simple docstring""" import math import os import re import sys import unittest from pathlib import Path from typing import Tuple from unittest.mock import patch from parameterized import parameterized from transformers.testing_utils import ( CaptureStderr, ExtendSysPath, TestCasePlus, execute_subprocess_async, get_gpu_count, get_torch_dist_unique_port, require_apex, require_bitsandbytes, require_fairscale, require_torch, require_torch_gpu, require_torch_multi_gpu, require_torch_non_multi_gpu, slow, ) from transformers.trainer_callback import TrainerState from transformers.trainer_utils import set_seed a_ = os.path.abspath(os.path.dirname(__file__)) with ExtendSysPath(f'''{bindir}/../../examples/pytorch/translation'''): from run_translation import main # noqa set_seed(42) a_ = "sshleifer/student_marian_en_ro_6_1" a_ = "sshleifer/tiny-mbart" @require_torch class snake_case ( _UpperCamelCase): def a_ ( self : str , a__ : str=False , a__ : str=None , a__ : Tuple=True , a__ : Optional[Any]=True , a__ : Union[str, Any]=True , a__ : Union[str, Any]=True , ) -> Any: '''simple docstring''' _A = self.run_trainer( eval_steps=1 , max_len=12 , model_name=a__ , num_train_epochs=1 , distributed=a__ , extra_args_str=a__ , predict_with_generate=a__ , do_train=a__ , do_eval=a__ , do_predict=a__ , ) _A = TrainerState.load_from_json(os.path.join(a__ , "trainer_state.json" ) ).log_history if not do_eval: return _A = [log for log in logs if "eval_loss" in log.keys()] _A = eval_metrics[0] if predict_with_generate: assert "eval_bleu" in first_step_stats _A = eval_metrics[-1] assert isinstance(last_step_stats["eval_bleu"] , a__ ) assert not math.isnan(float(last_step_stats["eval_loss"] ) ), "eval_loss must not be `nan`" @require_torch_non_multi_gpu def a_ ( self : Dict ) -> Union[str, Any]: '''simple docstring''' self.run_seqaseq_quick() @require_torch_multi_gpu def a_ ( self : Dict ) -> Optional[Any]: '''simple docstring''' self.run_seqaseq_quick(distributed=a__ ) @require_torch_multi_gpu def a_ ( self : str ) -> List[str]: '''simple docstring''' self.run_seqaseq_quick(distributed=a__ ) @unittest.skip("Requires an update of the env running those tests" ) @require_torch_multi_gpu @require_fairscale def a_ ( self : Any ) -> List[Any]: '''simple docstring''' self.run_seqaseq_quick(distributed=a__ , extra_args_str="--sharded_ddp simple" ) @unittest.skip("Requires an update of the env running those tests" ) @require_torch_multi_gpu @require_fairscale def a_ ( self : Dict ) -> Union[str, Any]: '''simple docstring''' self.run_seqaseq_quick(distributed=a__ , extra_args_str="--sharded_ddp simple --fp16" ) @unittest.skip("Requires an update of the env running those tests" ) @require_torch_multi_gpu @require_fairscale def a_ ( self : str ) -> Union[str, Any]: '''simple docstring''' self.run_seqaseq_quick(distributed=a__ , extra_args_str="--sharded_ddp zero_dp_2" , predict_with_generate=a__ ) @unittest.skip("Requires an update of the env running those tests" ) @require_torch_multi_gpu @require_fairscale def a_ ( self : int ) -> Optional[int]: '''simple docstring''' self.run_seqaseq_quick( distributed=a__ , extra_args_str="--sharded_ddp zero_dp_2 --fp16" , predict_with_generate=a__ ) @require_apex @require_torch_gpu def a_ ( self : str ) -> int: '''simple docstring''' self.run_seqaseq_quick(distributed=a__ , extra_args_str="--fp16 --fp16_backend=apex" ) # test 2nd time - was getting eval_loss': nan' # to reproduce the problem set distributed=False self.run_seqaseq_quick(distributed=a__ , extra_args_str="--fp16 --fp16_backend=apex" ) @parameterized.expand(["base", "low", "high", "mixed"] ) @require_torch_multi_gpu def a_ ( self : Optional[Any] , a__ : Tuple ) -> List[str]: '''simple docstring''' _A = { # test with the default log_level - should be info and thus log info once "base": {"extra_args_str": "", "n_matches": 1}, # test with low log_level and log_level_replica - should be noisy on all processes # now the info string should appear twice on 2 processes "low": {"extra_args_str": "--log_level debug --log_level_replica debug", "n_matches": 2}, # test with high log_level and low log_level_replica # now the info string should appear once only on the replica "high": {"extra_args_str": "--log_level error --log_level_replica debug", "n_matches": 1}, # test with high log_level and log_level_replica - should be quiet on all processes "mixed": {"extra_args_str": "--log_level error --log_level_replica error", "n_matches": 0}, } _A = experiments[experiment_id] _A = {"distributed": True, "predict_with_generate": False, "do_eval": False, "do_predict": False} _A = "Running training" with CaptureStderr() as cl: self.run_seqaseq_quick(**a__ , extra_args_str=data["extra_args_str"] ) _A = len(re.findall(a__ , cl.err ) ) self.assertEqual(a__ , data["n_matches"] ) @slow def a_ ( self : Tuple ) -> Union[str, Any]: '''simple docstring''' _A = self.run_trainer( eval_steps=2 , max_len=1_28 , model_name=a__ , learning_rate=3E-4 , num_train_epochs=10 , distributed=a__ , ) # Check metrics _A = TrainerState.load_from_json(os.path.join(a__ , "trainer_state.json" ) ).log_history _A = [log for log in logs if "eval_loss" in log.keys()] _A = eval_metrics[0] _A = eval_metrics[-1] assert first_step_stats["eval_loss"] > last_step_stats["eval_loss"], "model learned nothing" assert isinstance(last_step_stats["eval_bleu"] , a__ ) # test if do_predict saves generations and metrics _A = os.listdir(a__ ) _A = {os.path.basename(a__ ) for p in contents} assert "generated_predictions.txt" in contents assert "predict_results.json" in contents @slow @require_bitsandbytes def a_ ( self : List[Any] ) -> str: '''simple docstring''' from transformers.training_args import OptimizerNames def train_and_return_metrics(a__ : str ) -> Tuple[int, float]: _A = "--skip_memory_metrics 0" _A = self.run_trainer( max_len=1_28 , model_name=a__ , learning_rate=3E-4 , num_train_epochs=1 , optim=a__ , distributed=a__ , extra_args_str=a__ , do_eval=a__ , do_predict=a__ , n_gpus_to_use=1 , ) # Check metrics _A = TrainerState.load_from_json(Path(a__ , "trainer_state.json" ) ).log_history _A = int(logs[0]["train_mem_gpu_peaked_delta"] / 2**20 ) _A = int(logs[0]["train_mem_gpu_alloc_delta"] / 2**20 ) _A = logs[0]["train_loss"] return gpu_peak_mem_mb, gpu_alloc_mem_mb, loss _A , _A , _A = train_and_return_metrics(OptimizerNames.ADAMW_TORCH.value ) _A , _A , _A = train_and_return_metrics(OptimizerNames.ADAMW_BNB.value ) _A = gpu_alloc_mem_orig - gpu_alloc_mem_bnb _A = gpu_peak_mem_orig + gpu_alloc_mem_orig _A = gpu_peak_mem_bnb + gpu_alloc_mem_bnb _A = gpu_total_mem_orig - gpu_total_mem_bnb # sshleifer/student_marian_en_ro_6_1 has 54M parameter, 29M of which is `nn.Embedding` which # doesn't get quantized and remains in fp32. Therefore we only have 25M parameters quantized # in 2 bytes and the diff in optim memory usage is derived as so: # # - normal 25*8=~200MB (8 bytes per param) # - bnb 25*2= ~50MB (2 bytes per param) # # Thus we should expect ~150MB total memory saved. # # Peak memory should be the same - the total should be different by about that same margin # # After leaving a small margin to accommodate for differences between gpus let's check # that we have at least 120MB in savings _A = 1_20 # uncomment the following if this test starts failing - requires py38 for a new print feature # gpu_peak_mem_diff = gpu_peak_mem_orig - gpu_peak_mem_bnb # print(f"{gpu_alloc_mem_orig=}MB {gpu_peak_mem_orig=}MB {gpu_alloc_mem_orig+gpu_peak_mem_orig=}MB") # print(f" {gpu_alloc_mem_bnb=}MB {gpu_peak_mem_bnb=}MB {gpu_alloc_mem_bnb+gpu_peak_mem_bnb=}MB") # print(f"{gpu_alloc_mem_diff=}MB") # print(f"{gpu_peak_mem_diff=}MB") # print(f"{gpu_total_mem_orig=}MB, {gpu_total_mem_bnb=}MB") # print(f"{gpu_total_mem_diff=}MB, {gpu_total_mem_diff=}MB") self.assertGreater( a__ , a__ , "should use ~150MB less alloc gpu memory with BNB, compared to without it for this model but got" F""" a difference of {gpu_alloc_mem_diff}MB, with gpu_alloc_mem_orig={gpu_alloc_mem_orig}MB and""" F""" gpu_alloc_mem_bnb={gpu_alloc_mem_bnb}MB""" , ) self.assertGreater( a__ , a__ , "should use ~150MB less total gpu memory with BNB, compared to without it for this model but got" F""" a difference of {gpu_total_mem_diff}MB, with gpu_total_mem_orig={gpu_total_mem_orig}MB and""" F""" gpu_total_mem_bnb={gpu_total_mem_bnb}MB""" , ) self.assertEqual( a__ , a__ , F"""loss should be the same, but got loss_orig={loss_orig}, loss_bnb={loss_bnb}""" ) def a_ ( self : Any , a__ : int , a__ : str , a__ : int , a__ : float = 3E-3 , a__ : str = "adafactor" , a__ : bool = False , a__ : str = None , a__ : int = 0 , a__ : bool = True , a__ : bool = True , a__ : bool = True , a__ : bool = True , a__ : int = None , ) -> str: '''simple docstring''' _A = self.test_file_dir / "../fixtures/tests_samples/wmt_en_ro" _A = self.get_auto_remove_tmp_dir() _A = F""" --model_name_or_path {model_name} --train_file {data_dir}/train.json --validation_file {data_dir}/val.json --test_file {data_dir}/test.json --output_dir {output_dir} --overwrite_output_dir --max_train_samples 8 --max_source_length {max_len} --max_target_length {max_len} --do_train --num_train_epochs {str(a__ )} --per_device_train_batch_size 4 --learning_rate {learning_rate} --warmup_steps 8 --logging_steps 0 --logging_strategy no --save_steps {str(a__ )} --group_by_length --label_smoothing_factor 0.1 --target_lang ro_RO --source_lang en_XX """.split() _A = F""" --do_eval --per_device_eval_batch_size 4 --max_eval_samples 8 --val_max_target_length {max_len} --evaluation_strategy steps --eval_steps {str(a__ )} """.split() _A = "\n --do_predict\n ".split() _A = [] if do_train: args += args_train if do_eval: args += args_eval if do_predict: args += args_predict if predict_with_generate: args += "--predict_with_generate".split() if do_train: if optim == "adafactor": args += "--adafactor".split() else: args += F"""--optim {optim}""".split() if extra_args_str is not None: args += extra_args_str.split() if distributed: if n_gpus_to_use is None: _A = get_gpu_count() _A = get_torch_dist_unique_port() _A = F""" -m torch.distributed.run --nproc_per_node={n_gpus_to_use} --master_port={master_port} {self.examples_dir_str}/pytorch/translation/run_translation.py """.split() _A = [sys.executable] + distributed_args + args # keep for quick debug # print(" ".join([f"\nPYTHONPATH={self.src_dir_str}"] +cmd)); die execute_subprocess_async(a__ , env=self.get_env() ) else: _A = ["run_translation.py"] + args with patch.object(a__ , "argv" , a__ ): main() return output_dir
703
"""simple docstring""" import os import time import warnings from dataclasses import dataclass, field from enum import Enum from typing import List, Optional, Union import torch from filelock import FileLock from torch.utils.data import Dataset from ...tokenization_utils_base import PreTrainedTokenizerBase from ...utils import logging from ..processors.glue import glue_convert_examples_to_features, glue_output_modes, glue_processors from ..processors.utils import InputFeatures a_ = logging.get_logger(__name__) @dataclass class snake_case : __UpperCamelCase = field(metadata={'help': 'The name of the task to train on: ' + ', '.join(glue_processors.keys())}) __UpperCamelCase = field( metadata={'help': 'The input data dir. Should contain the .tsv files (or other data files) for the task.'}) __UpperCamelCase = field( default=128 , metadata={ 'help': ( 'The maximum total input sequence length after tokenization. Sequences longer ' 'than this will be truncated, sequences shorter will be padded.' ) } , ) __UpperCamelCase = field( default=_UpperCamelCase , metadata={'help': 'Overwrite the cached training and evaluation sets'}) def a_ ( self : Optional[Any] ) -> Tuple: '''simple docstring''' _A = self.task_name.lower() class snake_case ( _UpperCamelCase): __UpperCamelCase = 'train' __UpperCamelCase = 'dev' __UpperCamelCase = 'test' class snake_case ( _UpperCamelCase): __UpperCamelCase = 42 __UpperCamelCase = 42 __UpperCamelCase = 42 def __init__( self : Optional[int] , a__ : GlueDataTrainingArguments , a__ : PreTrainedTokenizerBase , a__ : Optional[int] = None , a__ : Union[str, Split] = Split.train , a__ : Optional[str] = None , ) -> Tuple: '''simple docstring''' warnings.warn( "This dataset will be removed from the library soon, preprocessing should be handled with the 🤗 Datasets " "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" , a__ , ) _A = args _A = glue_processors[args.task_name]() _A = glue_output_modes[args.task_name] if isinstance(a__ , a__ ): try: _A = Split[mode] except KeyError: raise KeyError("mode is not a valid split name" ) # Load data features from cache or dataset file _A = os.path.join( cache_dir if cache_dir is not None else args.data_dir , F"""cached_{mode.value}_{tokenizer.__class__.__name__}_{args.max_seq_length}_{args.task_name}""" , ) _A = self.processor.get_labels() if args.task_name in ["mnli", "mnli-mm"] and tokenizer.__class__.__name__ in ( "RobertaTokenizer", "RobertaTokenizerFast", "XLMRobertaTokenizer", "BartTokenizer", "BartTokenizerFast", ): # HACK(label indices are swapped in RoBERTa pretrained model) _A , _A = label_list[2], label_list[1] _A = label_list # Make sure only the first process in distributed training processes the dataset, # and the others will use the cache. _A = cached_features_file + ".lock" with FileLock(a__ ): if os.path.exists(a__ ) and not args.overwrite_cache: _A = time.time() _A = torch.load(a__ ) logger.info( F"""Loading features from cached file {cached_features_file} [took %.3f s]""" , time.time() - start ) else: logger.info(F"""Creating features from dataset file at {args.data_dir}""" ) if mode == Split.dev: _A = self.processor.get_dev_examples(args.data_dir ) elif mode == Split.test: _A = self.processor.get_test_examples(args.data_dir ) else: _A = self.processor.get_train_examples(args.data_dir ) if limit_length is not None: _A = examples[:limit_length] _A = glue_convert_examples_to_features( a__ , a__ , max_length=args.max_seq_length , label_list=a__ , output_mode=self.output_mode , ) _A = time.time() torch.save(self.features , a__ ) # ^ This seems to take a lot of time so I want to investigate why and how we can improve. logger.info( F"""Saving features into cached file {cached_features_file} [took {time.time() - start:.3f} s]""" ) def __len__( self : List[Any] ) -> Any: '''simple docstring''' return len(self.features ) def __getitem__( self : Tuple , a__ : Union[str, Any] ) -> InputFeatures: '''simple docstring''' return self.features[i] def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' return self.label_list
621
0
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_tokenizers_available, is_torch_available __snake_case = { "configuration_graphormer": ["GRAPHORMER_PRETRAINED_CONFIG_ARCHIVE_MAP", "GraphormerConfig"], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: __snake_case = [ "GRAPHORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "GraphormerForGraphClassification", "GraphormerModel", "GraphormerPreTrainedModel", ] if TYPE_CHECKING: from .configuration_graphormer import GRAPHORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, GraphormerConfig try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_graphormer import ( GRAPHORMER_PRETRAINED_MODEL_ARCHIVE_LIST, GraphormerForGraphClassification, GraphormerModel, GraphormerPreTrainedModel, ) else: import sys __snake_case = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
704
"""simple docstring""" def a__ ( __lowercase , __lowercase , __lowercase , __lowercase ) -> str: # Return True if there is node that has not iterated. _A = [False] * len(__lowercase ) _A = [] queue.append(__lowercase ) _A = True while queue: _A = queue.pop(0 ) for ind in range(len(graph[u] ) ): if visited[ind] is False and graph[u][ind] > 0: queue.append(__lowercase ) _A = True _A = u return visited[t] def a__ ( __lowercase , __lowercase , __lowercase ) -> int: # This array is filled by BFS and to store path _A = [-1] * (len(__lowercase )) _A = 0 while bfs(__lowercase , __lowercase , __lowercase , __lowercase ): _A = float("Inf" ) _A = sink while s != source: # Find the minimum value in select path _A = min(__lowercase , graph[parent[s]][s] ) _A = parent[s] max_flow += path_flow _A = sink while v != source: _A = parent[v] graph[u][v] -= path_flow graph[v][u] += path_flow _A = parent[v] return max_flow a_ = [ [0, 16, 13, 0, 0, 0], [0, 0, 10, 12, 0, 0], [0, 4, 0, 0, 14, 0], [0, 0, 9, 0, 0, 20], [0, 0, 0, 7, 0, 4], [0, 0, 0, 0, 0, 0], ] a_ , a_ = 0, 5 print(ford_fulkerson(graph, source, sink))
621
0
"""simple docstring""" def a__ ( __lowercase , __lowercase , __lowercase , __lowercase ) -> str: # Return True if there is node that has not iterated. _A = [False] * len(__lowercase ) _A = [] queue.append(__lowercase ) _A = True while queue: _A = queue.pop(0 ) for ind in range(len(graph[u] ) ): if visited[ind] is False and graph[u][ind] > 0: queue.append(__lowercase ) _A = True _A = u return visited[t] def a__ ( __lowercase , __lowercase , __lowercase ) -> int: # This array is filled by BFS and to store path _A = [-1] * (len(__lowercase )) _A = 0 while bfs(__lowercase , __lowercase , __lowercase , __lowercase ): _A = float("Inf" ) _A = sink while s != source: # Find the minimum value in select path _A = min(__lowercase , graph[parent[s]][s] ) _A = parent[s] max_flow += path_flow _A = sink while v != source: _A = parent[v] graph[u][v] -= path_flow graph[v][u] += path_flow _A = parent[v] return max_flow a_ = [ [0, 16, 13, 0, 0, 0], [0, 0, 10, 12, 0, 0], [0, 4, 0, 0, 14, 0], [0, 0, 9, 0, 0, 20], [0, 0, 0, 7, 0, 4], [0, 0, 0, 0, 0, 0], ] a_ , a_ = 0, 5 print(ford_fulkerson(graph, source, sink))
705
"""simple docstring""" import argparse import json from collections import OrderedDict from pathlib import Path import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ( ConditionalDetrConfig, ConditionalDetrForObjectDetection, ConditionalDetrForSegmentation, ConditionalDetrImageProcessor, ) from transformers.utils import logging logging.set_verbosity_info() a_ = logging.get_logger(__name__) # here we list all keys to be renamed (original name on the left, our name on the right) a_ = [] for i in range(6): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.weight''', f'''encoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.bias''', f'''encoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.weight''', f'''encoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.bias''', f'''encoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.weight''', f'''encoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.bias''', f'''encoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.encoder.layers.{i}.norm1.weight''', f'''encoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.norm1.bias''', f'''encoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.weight''', f'''encoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.bias''', f'''encoder.layers.{i}.final_layer_norm.bias''')) # decoder layers: 2 times output projection, 2 feedforward neural networks and 3 layernorms rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.weight''', f'''decoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.bias''', f'''decoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.weight''', f'''decoder.layers.{i}.encoder_attn.out_proj.weight''', ) ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.bias''', f'''decoder.layers.{i}.encoder_attn.out_proj.bias''', ) ) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.weight''', f'''decoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.bias''', f'''decoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.weight''', f'''decoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.bias''', f'''decoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm1.weight''', f'''decoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm1.bias''', f'''decoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.weight''', f'''decoder.layers.{i}.encoder_attn_layer_norm.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.bias''', f'''decoder.layers.{i}.encoder_attn_layer_norm.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.weight''', f'''decoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.bias''', f'''decoder.layers.{i}.final_layer_norm.bias''')) # q, k, v projections in self/cross-attention in decoder for conditional DETR rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.weight''', f'''decoder.layers.{i}.sa_qcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.weight''', f'''decoder.layers.{i}.sa_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qpos_proj.weight''', f'''decoder.layers.{i}.sa_qpos_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kpos_proj.weight''', f'''decoder.layers.{i}.sa_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.weight''', f'''decoder.layers.{i}.sa_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.weight''', f'''decoder.layers.{i}.ca_qcontent_proj.weight''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.weight", f"decoder.layers.{i}.ca_qpos_proj.weight")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.weight''', f'''decoder.layers.{i}.ca_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kpos_proj.weight''', f'''decoder.layers.{i}.ca_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.weight''', f'''decoder.layers.{i}.ca_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.weight''', f'''decoder.layers.{i}.ca_qpos_sine_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.bias''', f'''decoder.layers.{i}.sa_qcontent_proj.bias''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.bias''', f'''decoder.layers.{i}.sa_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_qpos_proj.bias''', f'''decoder.layers.{i}.sa_qpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_kpos_proj.bias''', f'''decoder.layers.{i}.sa_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.bias''', f'''decoder.layers.{i}.sa_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.bias''', f'''decoder.layers.{i}.ca_qcontent_proj.bias''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.bias", f"decoder.layers.{i}.ca_qpos_proj.bias")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.bias''', f'''decoder.layers.{i}.ca_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_kpos_proj.bias''', f'''decoder.layers.{i}.ca_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.bias''', f'''decoder.layers.{i}.ca_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.bias''', f'''decoder.layers.{i}.ca_qpos_sine_proj.bias''') ) # convolutional projection + query embeddings + layernorm of decoder + class and bounding box heads # for conditional DETR, also convert reference point head and query scale MLP rename_keys.extend( [ ("input_proj.weight", "input_projection.weight"), ("input_proj.bias", "input_projection.bias"), ("query_embed.weight", "query_position_embeddings.weight"), ("transformer.decoder.norm.weight", "decoder.layernorm.weight"), ("transformer.decoder.norm.bias", "decoder.layernorm.bias"), ("class_embed.weight", "class_labels_classifier.weight"), ("class_embed.bias", "class_labels_classifier.bias"), ("bbox_embed.layers.0.weight", "bbox_predictor.layers.0.weight"), ("bbox_embed.layers.0.bias", "bbox_predictor.layers.0.bias"), ("bbox_embed.layers.1.weight", "bbox_predictor.layers.1.weight"), ("bbox_embed.layers.1.bias", "bbox_predictor.layers.1.bias"), ("bbox_embed.layers.2.weight", "bbox_predictor.layers.2.weight"), ("bbox_embed.layers.2.bias", "bbox_predictor.layers.2.bias"), ("transformer.decoder.ref_point_head.layers.0.weight", "decoder.ref_point_head.layers.0.weight"), ("transformer.decoder.ref_point_head.layers.0.bias", "decoder.ref_point_head.layers.0.bias"), ("transformer.decoder.ref_point_head.layers.1.weight", "decoder.ref_point_head.layers.1.weight"), ("transformer.decoder.ref_point_head.layers.1.bias", "decoder.ref_point_head.layers.1.bias"), ("transformer.decoder.query_scale.layers.0.weight", "decoder.query_scale.layers.0.weight"), ("transformer.decoder.query_scale.layers.0.bias", "decoder.query_scale.layers.0.bias"), ("transformer.decoder.query_scale.layers.1.weight", "decoder.query_scale.layers.1.weight"), ("transformer.decoder.query_scale.layers.1.bias", "decoder.query_scale.layers.1.bias"), ("transformer.decoder.layers.0.ca_qpos_proj.weight", "decoder.layers.0.ca_qpos_proj.weight"), ("transformer.decoder.layers.0.ca_qpos_proj.bias", "decoder.layers.0.ca_qpos_proj.bias"), ] ) def a__ ( __lowercase , __lowercase , __lowercase ) -> List[str]: _A = state_dict.pop(__lowercase ) _A = val def a__ ( __lowercase ) -> List[str]: _A = OrderedDict() for key, value in state_dict.items(): if "backbone.0.body" in key: _A = key.replace("backbone.0.body" , "backbone.conv_encoder.model" ) _A = value else: _A = value return new_state_dict def a__ ( __lowercase , __lowercase=False ) -> Any: _A = "" if is_panoptic: _A = "conditional_detr." # first: transformer encoder for i in range(6 ): # read in weights + bias of input projection layer (in PyTorch's MultiHeadAttention, this is a single matrix + bias) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_weight""" ) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_bias""" ) # next, add query, keys and values (in that order) to the state dict _A = in_proj_weight[:256, :] _A = in_proj_bias[:256] _A = in_proj_weight[256:512, :] _A = in_proj_bias[256:512] _A = in_proj_weight[-256:, :] _A = in_proj_bias[-256:] def a__ ( ) -> int: _A = "http://images.cocodataset.org/val2017/000000039769.jpg" _A = Image.open(requests.get(__lowercase , stream=__lowercase ).raw ) return im @torch.no_grad() def a__ ( __lowercase , __lowercase ) -> Any: _A = ConditionalDetrConfig() # set backbone and dilation attributes if "resnet101" in model_name: _A = "resnet101" if "dc5" in model_name: _A = True _A = "panoptic" in model_name if is_panoptic: _A = 250 else: _A = 91 _A = "huggingface/label-files" _A = "coco-detection-id2label.json" _A = json.load(open(hf_hub_download(__lowercase , __lowercase , repo_type="dataset" ) , "r" ) ) _A = {int(__lowercase ): v for k, v in idalabel.items()} _A = idalabel _A = {v: k for k, v in idalabel.items()} # load image processor _A = "coco_panoptic" if is_panoptic else "coco_detection" _A = ConditionalDetrImageProcessor(format=__lowercase ) # prepare image _A = prepare_img() _A = image_processor(images=__lowercase , return_tensors="pt" ) _A = encoding["pixel_values"] logger.info(f"""Converting model {model_name}...""" ) # load original model from torch hub _A = torch.hub.load("DeppMeng/ConditionalDETR" , __lowercase , pretrained=__lowercase ).eval() _A = conditional_detr.state_dict() # rename keys for src, dest in rename_keys: if is_panoptic: _A = "conditional_detr." + src rename_key(__lowercase , __lowercase , __lowercase ) _A = rename_backbone_keys(__lowercase ) # query, key and value matrices need special treatment read_in_q_k_v(__lowercase , is_panoptic=__lowercase ) # important: we need to prepend a prefix to each of the base model keys as the head models use different attributes for them _A = "conditional_detr.model." if is_panoptic else "model." for key in state_dict.copy().keys(): if is_panoptic: if ( key.startswith("conditional_detr" ) and not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ) ): _A = state_dict.pop(__lowercase ) _A = val elif "class_labels_classifier" in key or "bbox_predictor" in key: _A = state_dict.pop(__lowercase ) _A = val elif key.startswith("bbox_attention" ) or key.startswith("mask_head" ): continue else: _A = state_dict.pop(__lowercase ) _A = val else: if not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ): _A = state_dict.pop(__lowercase ) _A = val # finally, create HuggingFace model and load state dict _A = ConditionalDetrForSegmentation(__lowercase ) if is_panoptic else ConditionalDetrForObjectDetection(__lowercase ) model.load_state_dict(__lowercase ) model.eval() model.push_to_hub(repo_id=__lowercase , organization="DepuMeng" , commit_message="Add model" ) # verify our conversion _A = conditional_detr(__lowercase ) _A = model(__lowercase ) assert torch.allclose(outputs.logits , original_outputs["pred_logits"] , atol=1E-4 ) assert torch.allclose(outputs.pred_boxes , original_outputs["pred_boxes"] , atol=1E-4 ) if is_panoptic: assert torch.allclose(outputs.pred_masks , original_outputs["pred_masks"] , atol=1E-4 ) # Save model and image processor logger.info(f"""Saving PyTorch model and image processor to {pytorch_dump_folder_path}...""" ) Path(__lowercase ).mkdir(exist_ok=__lowercase ) model.save_pretrained(__lowercase ) image_processor.save_pretrained(__lowercase ) if __name__ == "__main__": a_ = argparse.ArgumentParser() parser.add_argument( "--model_name", default="conditional_detr_resnet50", type=str, help="Name of the CONDITIONAL_DETR model you'd like to convert.", ) parser.add_argument( "--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model." ) a_ = parser.parse_args() convert_conditional_detr_checkpoint(args.model_name, args.pytorch_dump_folder_path)
621
0
"""simple docstring""" from typing import Any class snake_case : def __init__( self : Optional[int] , a__ : Any ) -> str: '''simple docstring''' _A = data _A = None class snake_case : def __init__( self : Optional[int] ) -> str: '''simple docstring''' _A = None def a_ ( self : List[Any] ) -> Any: '''simple docstring''' _A = self.head while temp is not None: print(temp.data , end=" " ) _A = temp.next print() def a_ ( self : Optional[Any] , a__ : Any ) -> Dict: '''simple docstring''' _A = Node(a__ ) _A = self.head _A = new_node def a_ ( self : List[str] , a__ : Optional[Any] , a__ : List[Any] ) -> Tuple: '''simple docstring''' if node_data_a == node_data_a: return else: _A = self.head while node_a is not None and node_a.data != node_data_a: _A = node_a.next _A = self.head while node_a is not None and node_a.data != node_data_a: _A = node_a.next if node_a is None or node_a is None: return _A , _A = node_a.data, node_a.data if __name__ == "__main__": a_ = LinkedList() for i in range(5, 0, -1): ll.push(i) ll.print_list() ll.swap_nodes(1, 4) print("After swapping") ll.print_list()
706
"""simple docstring""" import random def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = a[left_index] _A = left_index + 1 for j in range(left_index + 1 , __lowercase ): if a[j] < pivot: _A , _A = a[i], a[j] i += 1 _A , _A = a[i - 1], a[left_index] return i - 1 def a__ ( __lowercase , __lowercase , __lowercase ) -> int: if left < right: _A = random.randint(__lowercase , right - 1 ) _A , _A = ( a[left], a[pivot], ) # switches the pivot with the left most bound _A = partition(__lowercase , __lowercase , __lowercase ) quick_sort_random( __lowercase , __lowercase , __lowercase ) # recursive quicksort to the left of the pivot point quick_sort_random( __lowercase , pivot_index + 1 , __lowercase ) # recursive quicksort to the right of the pivot point def a__ ( ) -> Dict: _A = input("Enter numbers separated by a comma:\n" ).strip() _A = [int(__lowercase ) for item in user_input.split("," )] quick_sort_random(__lowercase , 0 , len(__lowercase ) ) print(__lowercase ) if __name__ == "__main__": main()
621
0
"""simple docstring""" import unittest from transformers import BertGenerationTokenizer from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_torch, slow from transformers.utils import cached_property from ...test_tokenization_common import TokenizerTesterMixin a_ = "▁" a_ = get_tests_dir("fixtures/test_sentencepiece.model") @require_sentencepiece class snake_case ( _UpperCamelCase , unittest.TestCase): __UpperCamelCase = BertGenerationTokenizer __UpperCamelCase = False __UpperCamelCase = True def a_ ( self : Any ) -> Optional[int]: '''simple docstring''' super().setUp() _A = BertGenerationTokenizer(a__ , keep_accents=a__ ) tokenizer.save_pretrained(self.tmpdirname ) def a_ ( self : Dict ) -> Dict: '''simple docstring''' _A = "<s>" _A = 1 self.assertEqual(self.get_tokenizer()._convert_token_to_id(a__ ) , a__ ) self.assertEqual(self.get_tokenizer()._convert_id_to_token(a__ ) , a__ ) def a_ ( self : int ) -> int: '''simple docstring''' _A = list(self.get_tokenizer().get_vocab().keys() ) self.assertEqual(vocab_keys[0] , "<unk>" ) self.assertEqual(vocab_keys[1] , "<s>" ) self.assertEqual(vocab_keys[-1] , "<pad>" ) self.assertEqual(len(a__ ) , 10_02 ) def a_ ( self : List[Any] ) -> List[Any]: '''simple docstring''' self.assertEqual(self.get_tokenizer().vocab_size , 10_00 ) def a_ ( self : List[str] ) -> Tuple: '''simple docstring''' _A = BertGenerationTokenizer(a__ , keep_accents=a__ ) _A = tokenizer.tokenize("This is a test" ) self.assertListEqual(a__ , ["▁This", "▁is", "▁a", "▁t", "est"] ) self.assertListEqual( tokenizer.convert_tokens_to_ids(a__ ) , [2_85, 46, 10, 1_70, 3_82] , ) _A = tokenizer.tokenize("I was born in 92000, and this is falsé." ) self.assertListEqual( a__ , [ SPIECE_UNDERLINE + "I", SPIECE_UNDERLINE + "was", SPIECE_UNDERLINE + "b", "or", "n", SPIECE_UNDERLINE + "in", SPIECE_UNDERLINE + "", "9", "2", "0", "0", "0", ",", SPIECE_UNDERLINE + "and", SPIECE_UNDERLINE + "this", SPIECE_UNDERLINE + "is", SPIECE_UNDERLINE + "f", "al", "s", "é", ".", ] , ) _A = tokenizer.convert_tokens_to_ids(a__ ) self.assertListEqual( a__ , [8, 21, 84, 55, 24, 19, 7, 0, 6_02, 3_47, 3_47, 3_47, 3, 12, 66, 46, 72, 80, 6, 0, 4] , ) _A = tokenizer.convert_ids_to_tokens(a__ ) self.assertListEqual( a__ , [ SPIECE_UNDERLINE + "I", SPIECE_UNDERLINE + "was", SPIECE_UNDERLINE + "b", "or", "n", SPIECE_UNDERLINE + "in", SPIECE_UNDERLINE + "", "<unk>", "2", "0", "0", "0", ",", SPIECE_UNDERLINE + "and", SPIECE_UNDERLINE + "this", SPIECE_UNDERLINE + "is", SPIECE_UNDERLINE + "f", "al", "s", "<unk>", ".", ] , ) @cached_property def a_ ( self : str ) -> List[Any]: '''simple docstring''' return BertGenerationTokenizer.from_pretrained("google/bert_for_seq_generation_L-24_bbc_encoder" ) @slow def a_ ( self : List[Any] ) -> Tuple: '''simple docstring''' _A = "Hello World!" _A = [1_85_36, 22_60, 1_01] self.assertListEqual(a__ , self.big_tokenizer.encode(a__ ) ) @slow def a_ ( self : Optional[Any] ) -> Optional[Any]: '''simple docstring''' _A = ( "This is a very long text with a lot of weird characters, such as: . , ~ ? ( ) \" [ ] ! : - . Also we will" " add words that should not exsist and be tokenized to <unk>, such as saoneuhaoesuth" ) _A = [ 8_71, 4_19, 3_58, 9_46, 9_91, 25_21, 4_52, 3_58, 13_57, 3_87, 77_51, 35_36, 1_12, 9_85, 4_56, 1_26, 8_65, 9_38, 54_00, 57_34, 4_58, 13_68, 4_67, 7_86, 24_62, 52_46, 11_59, 6_33, 8_65, 45_19, 4_57, 5_82, 8_52, 25_57, 4_27, 9_16, 5_08, 4_05, 3_43_24, 4_97, 3_91, 4_08, 1_13_42, 12_44, 3_85, 1_00, 9_38, 9_85, 4_56, 5_74, 3_62, 1_25_97, 32_00, 31_29, 11_72, ] self.assertListEqual(a__ , self.big_tokenizer.encode(a__ ) ) @require_torch @slow def a_ ( self : Dict ) -> List[str]: '''simple docstring''' import torch from transformers import BertGenerationConfig, BertGenerationEncoder # Build sequence _A = list(self.big_tokenizer.get_vocab().keys() )[:10] _A = " ".join(a__ ) _A = self.big_tokenizer.encode_plus(a__ , return_tensors="pt" , return_token_type_ids=a__ ) _A = self.big_tokenizer.batch_encode_plus( [sequence + " " + sequence] , return_tensors="pt" , return_token_type_ids=a__ ) _A = BertGenerationConfig() _A = BertGenerationEncoder(a__ ) assert model.get_input_embeddings().weight.shape[0] >= self.big_tokenizer.vocab_size with torch.no_grad(): model(**a__ ) model(**a__ ) @slow def a_ ( self : Tuple ) -> Optional[int]: '''simple docstring''' _A = {"input_ids": [[3_92_86, 4_58, 3_63_35, 20_01, 4_56, 1_30_73, 1_32_66, 4_55, 1_13, 77_46, 17_41, 1_11_57, 3_91, 1_30_73, 1_32_66, 4_55, 1_13, 39_67, 3_54_12, 1_13, 49_36, 1_09, 38_70, 23_77, 1_13, 3_00_84, 4_57_20, 4_58, 1_34, 1_74_96, 1_12, 5_03, 1_16_72, 1_13, 1_18, 1_12, 56_65, 1_33_47, 3_86_87, 1_12, 14_96, 3_13_89, 1_12, 32_68, 4_72_64, 1_34, 9_62, 1_12, 1_63_77, 80_35, 2_31_30, 4_30, 1_21_69, 1_55_18, 2_85_92, 4_58, 1_46, 4_16_97, 1_09, 3_91, 1_21_69, 1_55_18, 1_66_89, 4_58, 1_46, 4_13_58, 1_09, 4_52, 7_26, 40_34, 1_11, 7_63, 3_54_12, 50_82, 3_88, 19_03, 1_11, 90_51, 3_91, 28_70, 4_89_18, 19_00, 11_23, 5_50, 9_98, 1_12, 95_86, 1_59_85, 4_55, 3_91, 4_10, 2_29_55, 3_76_36, 1_14], [4_48, 1_74_96, 4_19, 36_63, 3_85, 7_63, 1_13, 2_75_33, 28_70, 32_83, 1_30_43, 16_39, 2_47_13, 5_23, 6_56, 2_40_13, 1_85_50, 25_21, 5_17, 2_70_14, 2_12_44, 4_20, 12_12, 14_65, 3_91, 9_27, 48_33, 3_88, 5_78, 1_17_86, 1_14, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [4_84, 21_69, 76_87, 2_19_32, 1_81_46, 7_26, 3_63, 1_70_32, 33_91, 1_14, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], "attention_mask": [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]} # noqa: E501 # fmt: on self.tokenizer_integration_test_util( expected_encoding=a__ , model_name="google/bert_for_seq_generation_L-24_bbc_encoder" , revision="c817d1fd1be2ffa69431227a1fe320544943d4db" , )
707
"""simple docstring""" import copy from typing import Any, Dict, List, Optional, Union import numpy as np 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_ = logging.get_logger(__name__) class snake_case ( _UpperCamelCase): __UpperCamelCase = ['input_features'] def __init__( self : int , a__ : Optional[Any]=80 , a__ : Optional[int]=1_60_00 , a__ : int=1_60 , a__ : Union[str, Any]=30 , a__ : Tuple=4_00 , a__ : List[Any]=0.0 , a__ : Optional[Any]=False , **a__ : List[Any] , ) -> str: '''simple docstring''' super().__init__( feature_size=a__ , sampling_rate=a__ , padding_value=a__ , return_attention_mask=a__ , **a__ , ) _A = n_fft _A = hop_length _A = chunk_length _A = chunk_length * sampling_rate _A = self.n_samples // hop_length _A = sampling_rate _A = mel_filter_bank( num_frequency_bins=1 + n_fft // 2 , num_mel_filters=a__ , min_frequency=0.0 , max_frequency=8_0_0_0.0 , sampling_rate=a__ , norm="slaney" , mel_scale="slaney" , ) def a_ ( self : int , a__ : np.array ) -> np.ndarray: '''simple docstring''' _A = spectrogram( a__ , window_function(self.n_fft , "hann" ) , frame_length=self.n_fft , hop_length=self.hop_length , power=2.0 , mel_filters=self.mel_filters , log_mel="log10" , ) _A = log_spec[:, :-1] _A = np.maximum(a__ , log_spec.max() - 8.0 ) _A = (log_spec + 4.0) / 4.0 return log_spec @staticmethod # Copied from transformers.models.wav2vec2.feature_extraction_wav2vec2.Wav2Vec2FeatureExtractor.zero_mean_unit_var_norm def a_ ( a__ : List[np.ndarray] , a__ : List[np.ndarray] , a__ : float = 0.0 ) -> List[np.ndarray]: '''simple docstring''' if attention_mask is not None: _A = np.array(a__ , np.intaa ) _A = [] for vector, length in zip(a__ , attention_mask.sum(-1 ) ): _A = (vector - vector[:length].mean()) / np.sqrt(vector[:length].var() + 1E-7 ) if length < normed_slice.shape[0]: _A = padding_value normed_input_values.append(a__ ) else: _A = [(x - x.mean()) / np.sqrt(x.var() + 1E-7 ) for x in input_values] return normed_input_values def __call__( self : Optional[int] , a__ : Union[np.ndarray, List[float], List[np.ndarray], List[List[float]]] , a__ : bool = True , a__ : Optional[int] = None , a__ : Optional[Union[str, TensorType]] = None , a__ : Optional[bool] = None , a__ : Optional[str] = "max_length" , a__ : Optional[int] = None , a__ : Optional[int] = None , a__ : Optional[bool] = None , **a__ : Dict , ) -> BatchFeature: '''simple docstring''' 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." ) _A = isinstance(a__ , 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}""" ) _A = is_batched_numpy or ( isinstance(a__ , (list, tuple) ) and (isinstance(raw_speech[0] , (np.ndarray, tuple, list) )) ) if is_batched: _A = [np.asarray([speech] , dtype=np.floataa ).T for speech in raw_speech] elif not is_batched and not isinstance(a__ , np.ndarray ): _A = np.asarray(a__ , dtype=np.floataa ) elif isinstance(a__ , np.ndarray ) and raw_speech.dtype is np.dtype(np.floataa ): _A = raw_speech.astype(np.floataa ) # always return batch if not is_batched: _A = [np.asarray([raw_speech] ).T] _A = BatchFeature({"input_features": raw_speech} ) # convert into correct format for padding _A = self.pad( a__ , padding=a__ , max_length=max_length if max_length else self.n_samples , truncation=a__ , pad_to_multiple_of=a__ , return_attention_mask=return_attention_mask or do_normalize , ) # zero-mean and unit-variance normalization if do_normalize: _A = self.zero_mean_unit_var_norm( padded_inputs["input_features"] , attention_mask=padded_inputs["attention_mask"] , padding_value=self.padding_value , ) _A = np.stack(padded_inputs["input_features"] , axis=0 ) # make sure list is in array format _A = padded_inputs.get("input_features" ).transpose(2 , 0 , 1 ) _A = [self._np_extract_fbank_features(a__ ) for waveform in input_features[0]] if isinstance(input_features[0] , a__ ): _A = [np.asarray(a__ , dtype=np.floataa ) for feature in input_features] else: _A = input_features if return_attention_mask: # rescale from sample (48000) to feature (3000) _A = padded_inputs["attention_mask"][:, :: self.hop_length] if return_tensors is not None: _A = padded_inputs.convert_to_tensors(a__ ) return padded_inputs def a_ ( self : Dict ) -> Dict[str, Any]: '''simple docstring''' _A = copy.deepcopy(self.__dict__ ) _A = self.__class__.__name__ if "mel_filters" in output: del output["mel_filters"] return output
621
0
"""simple docstring""" from typing import List, Optional, Union from ...image_utils import ImageInput from ...processing_utils import ProcessorMixin from ...tokenization_utils_base import BatchEncoding, PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy from ...utils import TensorType class snake_case ( _UpperCamelCase): __UpperCamelCase = ['image_processor', 'tokenizer'] __UpperCamelCase = 'BlipImageProcessor' __UpperCamelCase = ('BertTokenizer', 'BertTokenizerFast') def __init__( self : str , a__ : Any , a__ : int ) -> str: '''simple docstring''' _A = False super().__init__(a__ , a__ ) _A = self.image_processor def __call__( self : List[Any] , a__ : ImageInput = None , a__ : Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]] = None , a__ : bool = True , a__ : Union[bool, str, PaddingStrategy] = False , a__ : Union[bool, str, TruncationStrategy] = None , a__ : Optional[int] = None , a__ : int = 0 , a__ : Optional[int] = None , a__ : Optional[bool] = None , a__ : bool = False , a__ : bool = False , a__ : bool = False , a__ : bool = False , a__ : bool = False , a__ : bool = True , a__ : Optional[Union[str, TensorType]] = None , **a__ : Dict , ) -> BatchEncoding: '''simple docstring''' if images is None and text is None: raise ValueError("You have to specify either images or text." ) # Get only text if images is None: _A = self.tokenizer _A = self.tokenizer( text=a__ , add_special_tokens=a__ , padding=a__ , truncation=a__ , max_length=a__ , stride=a__ , pad_to_multiple_of=a__ , return_attention_mask=a__ , return_overflowing_tokens=a__ , return_special_tokens_mask=a__ , return_offsets_mapping=a__ , return_token_type_ids=a__ , return_length=a__ , verbose=a__ , return_tensors=a__ , **a__ , ) return text_encoding # add pixel_values _A = self.image_processor(a__ , return_tensors=a__ ) if text is not None: _A = self.tokenizer( text=a__ , add_special_tokens=a__ , padding=a__ , truncation=a__ , max_length=a__ , stride=a__ , pad_to_multiple_of=a__ , return_attention_mask=a__ , return_overflowing_tokens=a__ , return_special_tokens_mask=a__ , return_offsets_mapping=a__ , return_token_type_ids=a__ , return_length=a__ , verbose=a__ , return_tensors=a__ , **a__ , ) else: _A = None if text_encoding is not None: encoding_image_processor.update(a__ ) return encoding_image_processor def a_ ( self : Optional[int] , *a__ : int , **a__ : Tuple ) -> List[Any]: '''simple docstring''' return self.tokenizer.batch_decode(*a__ , **a__ ) def a_ ( self : List[Any] , *a__ : List[str] , **a__ : Optional[Any] ) -> Any: '''simple docstring''' return self.tokenizer.decode(*a__ , **a__ ) @property def a_ ( self : Union[str, Any] ) -> Optional[Any]: '''simple docstring''' _A = self.tokenizer.model_input_names _A = self.image_processor.model_input_names return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names ) )
708
"""simple docstring""" from __future__ import annotations def a__ ( __lowercase , __lowercase ) -> float: _A = sorted(numsa + numsa ) _A , _A = divmod(len(__lowercase ) , 2 ) if mod == 1: return all_numbers[div] else: return (all_numbers[div] + all_numbers[div - 1]) / 2 if __name__ == "__main__": import doctest doctest.testmod() a_ = [float(x) for x in input("Enter the elements of first array: ").split()] a_ = [float(x) for x in input("Enter the elements of second array: ").split()] print(f'''The median of two arrays is: {median_of_two_arrays(array_a, array_a)}''')
621
0
"""simple docstring""" from ...processing_utils import ProcessorMixin class snake_case ( _UpperCamelCase): __UpperCamelCase = 'WhisperFeatureExtractor' __UpperCamelCase = 'WhisperTokenizer' def __init__( self : List[Any] , a__ : List[Any] , a__ : List[Any] ) -> int: '''simple docstring''' super().__init__(a__ , a__ ) _A = self.feature_extractor _A = False def a_ ( self : Any , a__ : Tuple=None , a__ : Any=None , a__ : Optional[int]=True ) -> Optional[int]: '''simple docstring''' return self.tokenizer.get_decoder_prompt_ids(task=a__ , language=a__ , no_timestamps=a__ ) def __call__( self : int , *a__ : Optional[Any] , **a__ : List[Any] ) -> Any: '''simple docstring''' if self._in_target_context_manager: return self.current_processor(*a__ , **a__ ) _A = kwargs.pop("audio" , a__ ) _A = kwargs.pop("sampling_rate" , a__ ) _A = kwargs.pop("text" , a__ ) if len(a__ ) > 0: _A = args[0] _A = args[1:] if audio is None and text is None: raise ValueError("You need to specify either an `audio` or `text` input to process." ) if audio is not None: _A = self.feature_extractor(a__ , *a__ , sampling_rate=a__ , **a__ ) if text is not None: _A = self.tokenizer(a__ , **a__ ) if text is None: return inputs elif audio is None: return encodings else: _A = encodings["input_ids"] return inputs def a_ ( self : Dict , *a__ : Optional[int] , **a__ : Tuple ) -> Tuple: '''simple docstring''' return self.tokenizer.batch_decode(*a__ , **a__ ) def a_ ( self : Dict , *a__ : Optional[int] , **a__ : Optional[int] ) -> Optional[Any]: '''simple docstring''' return self.tokenizer.decode(*a__ , **a__ ) def a_ ( self : List[Any] , a__ : str , a__ : Union[str, Any]="np" ) -> Union[str, Any]: '''simple docstring''' return self.tokenizer.get_prompt_ids(a__ , return_tensors=a__ )
709
"""simple docstring""" import copy import os from typing import Union from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { "Salesforce/blip-vqa-base": "https://huggingface.co/Salesforce/blip-vqa-base/resolve/main/config.json", "Salesforce/blip-vqa-capfit-large": ( "https://huggingface.co/Salesforce/blip-vqa-base-capfit/resolve/main/config.json" ), "Salesforce/blip-image-captioning-base": ( "https://huggingface.co/Salesforce/blip-image-captioning-base/resolve/main/config.json" ), "Salesforce/blip-image-captioning-large": ( "https://huggingface.co/Salesforce/blip-image-captioning-large/resolve/main/config.json" ), "Salesforce/blip-itm-base-coco": "https://huggingface.co/Salesforce/blip-itm-base-coco/resolve/main/config.json", "Salesforce/blip-itm-large-coco": "https://huggingface.co/Salesforce/blip-itm-large-coco/resolve/main/config.json", "Salesforce/blip-itm-base-flikr": "https://huggingface.co/Salesforce/blip-itm-base-flikr/resolve/main/config.json", "Salesforce/blip-itm-large-flikr": ( "https://huggingface.co/Salesforce/blip-itm-large-flikr/resolve/main/config.json" ), } class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip_text_model' def __init__( self : int , a__ : List[str]=3_05_24 , a__ : List[str]=7_68 , a__ : List[Any]=7_68 , a__ : int=30_72 , a__ : List[str]=7_68 , a__ : Dict=12 , a__ : Optional[int]=8 , a__ : Optional[Any]=5_12 , a__ : List[Any]="gelu" , a__ : Optional[Any]=1E-1_2 , a__ : Any=0.0 , a__ : int=0.0 , a__ : Dict=0.0_2 , a__ : Optional[Any]=3_05_22 , a__ : Any=2 , a__ : int=0 , a__ : Union[str, Any]=1_02 , a__ : Tuple=True , a__ : Optional[int]=True , **a__ : Any , ) -> List[Any]: '''simple docstring''' super().__init__( pad_token_id=a__ , bos_token_id=a__ , eos_token_id=a__ , sep_token_id=a__ , **a__ , ) _A = vocab_size _A = hidden_size _A = encoder_hidden_size _A = intermediate_size _A = projection_dim _A = hidden_dropout_prob _A = num_hidden_layers _A = num_attention_heads _A = max_position_embeddings _A = layer_norm_eps _A = hidden_act _A = initializer_range _A = attention_probs_dropout_prob _A = is_decoder _A = use_cache @classmethod def a_ ( cls : Optional[Any] , a__ : Union[str, os.PathLike] , **a__ : Optional[Any] ) -> "PretrainedConfig": '''simple docstring''' cls._set_token_in_kwargs(a__ ) _A , _A = cls.get_config_dict(a__ , **a__ ) # get the text config dict if we are loading from BlipConfig if config_dict.get("model_type" ) == "blip": _A = config_dict["text_config"] if "model_type" in config_dict and hasattr(cls , "model_type" ) and config_dict["model_type"] != cls.model_type: logger.warning( F"""You are using a model of type {config_dict['model_type']} to instantiate a model of type """ F"""{cls.model_type}. This is not supported for all configurations of models and can yield errors.""" ) return cls.from_dict(a__ , **a__ ) class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip_vision_model' def __init__( self : Optional[Any] , a__ : Any=7_68 , a__ : List[str]=30_72 , a__ : str=5_12 , a__ : Any=12 , a__ : int=12 , a__ : int=3_84 , a__ : Tuple=16 , a__ : str="gelu" , a__ : Tuple=1E-5 , a__ : List[str]=0.0 , a__ : List[Any]=1E-1_0 , **a__ : int , ) -> List[str]: '''simple docstring''' super().__init__(**a__ ) _A = hidden_size _A = intermediate_size _A = projection_dim _A = num_hidden_layers _A = num_attention_heads _A = patch_size _A = image_size _A = initializer_range _A = attention_dropout _A = layer_norm_eps _A = hidden_act @classmethod def a_ ( cls : Any , a__ : Union[str, os.PathLike] , **a__ : int ) -> "PretrainedConfig": '''simple docstring''' cls._set_token_in_kwargs(a__ ) _A , _A = cls.get_config_dict(a__ , **a__ ) # get the vision config dict if we are loading from BlipConfig if config_dict.get("model_type" ) == "blip": _A = config_dict["vision_config"] if "model_type" in config_dict and hasattr(cls , "model_type" ) and config_dict["model_type"] != cls.model_type: logger.warning( F"""You are using a model of type {config_dict['model_type']} to instantiate a model of type """ F"""{cls.model_type}. This is not supported for all configurations of models and can yield errors.""" ) return cls.from_dict(a__ , **a__ ) class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip' __UpperCamelCase = True def __init__( self : List[Any] , a__ : Optional[int]=None , a__ : str=None , a__ : List[str]=5_12 , a__ : Any=2.6_5_9_2 , a__ : str=2_56 , **a__ : Optional[int] , ) -> Dict: '''simple docstring''' super().__init__(**a__ ) if text_config is None: _A = {} logger.info("`text_config` is `None`. Initializing the `BlipTextConfig` with default values." ) if vision_config is None: _A = {} logger.info("`vision_config` is `None`. Initializing the `BlipVisionConfig` with default values." ) _A = BlipTextConfig(**a__ ) _A = BlipVisionConfig(**a__ ) _A = self.vision_config.hidden_size _A = projection_dim _A = logit_scale_init_value _A = 1.0 _A = 0.0_2 _A = image_text_hidden_size @classmethod def a_ ( cls : Tuple , a__ : BlipTextConfig , a__ : BlipVisionConfig , **a__ : Optional[int] ) -> str: '''simple docstring''' return cls(text_config=text_config.to_dict() , vision_config=vision_config.to_dict() , **a__ ) def a_ ( self : Union[str, Any] ) -> List[Any]: '''simple docstring''' _A = copy.deepcopy(self.__dict__ ) _A = self.text_config.to_dict() _A = self.vision_config.to_dict() _A = self.__class__.model_type return output
621
0
"""simple docstring""" a_ = "ABCDEFGHIJKLMNOPQRSTUVWXYZ" def a__ ( ) -> None: _A = input("Enter message: " ) _A = input("Enter key [alphanumeric]: " ) _A = input("Encrypt/Decrypt [e/d]: " ) if mode.lower().startswith("e" ): _A = "encrypt" _A = encrypt_message(__lowercase , __lowercase ) elif mode.lower().startswith("d" ): _A = "decrypt" _A = decrypt_message(__lowercase , __lowercase ) print(f"""\n{mode.title()}ed message:""" ) print(__lowercase ) def a__ ( __lowercase , __lowercase ) -> str: return translate_message(__lowercase , __lowercase , "encrypt" ) def a__ ( __lowercase , __lowercase ) -> str: return translate_message(__lowercase , __lowercase , "decrypt" ) def a__ ( __lowercase , __lowercase , __lowercase ) -> str: _A = [] _A = 0 _A = key.upper() for symbol in message: _A = LETTERS.find(symbol.upper() ) if num != -1: if mode == "encrypt": num += LETTERS.find(key[key_index] ) elif mode == "decrypt": num -= LETTERS.find(key[key_index] ) num %= len(__lowercase ) if symbol.isupper(): translated.append(LETTERS[num] ) elif symbol.islower(): translated.append(LETTERS[num].lower() ) key_index += 1 if key_index == len(__lowercase ): _A = 0 else: translated.append(__lowercase ) return "".join(__lowercase ) if __name__ == "__main__": main()
710
"""simple docstring""" import unittest from transformers import load_tool from .test_tools_common import ToolTesterMixin class snake_case ( unittest.TestCase , _UpperCamelCase): def a_ ( self : Optional[Any] ) -> List[str]: '''simple docstring''' _A = load_tool("text-classification" ) self.tool.setup() _A = load_tool("text-classification" , remote=a__ ) def a_ ( self : Optional[int] ) -> Dict: '''simple docstring''' _A = self.tool("That's quite cool" , ["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Optional[Any] ) -> Dict: '''simple docstring''' _A = self.remote_tool("That's quite cool" , ["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Dict ) -> Optional[int]: '''simple docstring''' _A = self.tool(text="That's quite cool" , labels=["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Dict ) -> Any: '''simple docstring''' _A = self.remote_tool(text="That's quite cool" , labels=["positive", "negative"] ) self.assertEqual(a__ , "positive" )
621
0
"""simple docstring""" import warnings from ...utils import logging from .image_processing_glpn import GLPNImageProcessor a_ = logging.get_logger(__name__) class snake_case ( _UpperCamelCase): def __init__( self : str , *a__ : Dict , **a__ : Optional[int] ) -> None: '''simple docstring''' warnings.warn( "The class GLPNFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please" " use GLPNImageProcessor instead." , a__ , ) super().__init__(*a__ , **a__ )
711
"""simple docstring""" import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer from diffusers import AutoencoderKL, PNDMScheduler, StableDiffusionInpaintPipeline, UNetaDConditionModel from diffusers.utils import floats_tensor, load_image, load_numpy, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu, slow from ..pipeline_params import TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS, TEXT_GUIDED_IMAGE_INPAINTING_PARAMS from ..test_pipelines_common import PipelineKarrasSchedulerTesterMixin, PipelineLatentTesterMixin, PipelineTesterMixin enable_full_determinism() class snake_case ( _UpperCamelCase , _UpperCamelCase , _UpperCamelCase , unittest.TestCase): __UpperCamelCase = StableDiffusionInpaintPipeline __UpperCamelCase = TEXT_GUIDED_IMAGE_INPAINTING_PARAMS __UpperCamelCase = TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS __UpperCamelCase = frozenset( []) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess __UpperCamelCase = frozenset([]) def a_ ( self : Dict ) -> Optional[int]: '''simple docstring''' torch.manual_seed(0 ) _A = UNetaDConditionModel( block_out_channels=(32, 64) , layers_per_block=2 , sample_size=32 , in_channels=9 , out_channels=4 , down_block_types=("DownBlock2D", "CrossAttnDownBlock2D") , up_block_types=("CrossAttnUpBlock2D", "UpBlock2D") , cross_attention_dim=32 , attention_head_dim=(2, 4) , use_linear_projection=a__ , ) _A = PNDMScheduler(skip_prk_steps=a__ ) torch.manual_seed(0 ) _A = AutoencoderKL( block_out_channels=[32, 64] , in_channels=3 , out_channels=3 , down_block_types=["DownEncoderBlock2D", "DownEncoderBlock2D"] , up_block_types=["UpDecoderBlock2D", "UpDecoderBlock2D"] , latent_channels=4 , sample_size=1_28 , ) torch.manual_seed(0 ) _A = CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=32 , intermediate_size=37 , layer_norm_eps=1E-0_5 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=10_00 , hidden_act="gelu" , projection_dim=5_12 , ) _A = CLIPTextModel(a__ ) _A = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip" ) _A = { "unet": unet, "scheduler": scheduler, "vae": vae, "text_encoder": text_encoder, "tokenizer": tokenizer, "safety_checker": None, "feature_extractor": None, } return components def a_ ( self : Optional[Any] , a__ : List[str] , a__ : Tuple=0 ) -> int: '''simple docstring''' _A = floats_tensor((1, 3, 32, 32) , rng=random.Random(a__ ) ).to(a__ ) _A = image.cpu().permute(0 , 2 , 3 , 1 )[0] _A = Image.fromarray(np.uinta(a__ ) ).convert("RGB" ).resize((64, 64) ) _A = Image.fromarray(np.uinta(image + 4 ) ).convert("RGB" ).resize((64, 64) ) if str(a__ ).startswith("mps" ): _A = torch.manual_seed(a__ ) else: _A = torch.Generator(device=a__ ).manual_seed(a__ ) _A = { "prompt": "A painting of a squirrel eating a burger", "image": init_image, "mask_image": mask_image, "generator": generator, "num_inference_steps": 2, "guidance_scale": 6.0, "output_type": "numpy", } return inputs def a_ ( self : Optional[Any] ) -> Optional[Any]: '''simple docstring''' _A = "cpu" # ensure determinism for the device-dependent torch.Generator _A = self.get_dummy_components() _A = StableDiffusionInpaintPipeline(**a__ ) _A = sd_pipe.to(a__ ) sd_pipe.set_progress_bar_config(disable=a__ ) _A = self.get_dummy_inputs(a__ ) _A = sd_pipe(**a__ ).images _A = image[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) _A = np.array([0.4_7_2_7, 0.5_7_3_5, 0.3_9_4_1, 0.5_4_4_6, 0.5_9_2_6, 0.4_3_9_4, 0.5_0_6_2, 0.4_6_5_4, 0.4_4_7_6] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-2 def a_ ( self : str ) -> Union[str, Any]: '''simple docstring''' super().test_inference_batch_single_identical(expected_max_diff=3E-3 ) @slow @require_torch_gpu class snake_case ( unittest.TestCase): def a_ ( self : List[Any] ) -> Any: '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def a_ ( self : Union[str, Any] ) -> Tuple: '''simple docstring''' _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = load_numpy( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint" "/yellow_cat_sitting_on_a_park_bench.npy" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = StableDiffusionInpaintPipeline.from_pretrained(a__ , safety_checker=a__ ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , output_type="np" , ) _A = output.images[0] assert image.shape == (5_12, 5_12, 3) assert np.abs(expected_image - image ).max() < 9E-3 def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = load_numpy( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint" "/yellow_cat_sitting_on_a_park_bench_fp16.npy" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = StableDiffusionInpaintPipeline.from_pretrained( a__ , torch_dtype=torch.floataa , safety_checker=a__ , ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , output_type="np" , ) _A = output.images[0] assert image.shape == (5_12, 5_12, 3) assert np.abs(expected_image - image ).max() < 5E-1 def a_ ( self : List[Any] ) -> Union[str, Any]: '''simple docstring''' torch.cuda.empty_cache() torch.cuda.reset_max_memory_allocated() torch.cuda.reset_peak_memory_stats() _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = PNDMScheduler.from_pretrained(a__ , subfolder="scheduler" ) _A = StableDiffusionInpaintPipeline.from_pretrained( a__ , safety_checker=a__ , scheduler=a__ , torch_dtype=torch.floataa , ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing(1 ) pipe.enable_sequential_cpu_offload() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , num_inference_steps=2 , output_type="np" , ) _A = torch.cuda.max_memory_allocated() # make sure that less than 2.65 GB is allocated assert mem_bytes < 2.6_5 * 10**9
621
0
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> int: while a != 0: _A , _A = b % a, a return b def a__ ( __lowercase , __lowercase ) -> int: if gcd(__lowercase , __lowercase ) != 1: _A = f"""mod inverse of {a!r} and {m!r} does not exist""" raise ValueError(__lowercase ) _A , _A , _A = 1, 0, a _A , _A , _A = 0, 1, m while va != 0: _A = ua // va _A , _A , _A , _A , _A , _A = (ua - q * va), (ua - q * va), (ua - q * va), va, va, va return ua % m
712
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> int: while a != 0: _A , _A = b % a, a return b def a__ ( __lowercase , __lowercase ) -> int: if gcd(__lowercase , __lowercase ) != 1: _A = f"""mod inverse of {a!r} and {m!r} does not exist""" raise ValueError(__lowercase ) _A , _A , _A = 1, 0, a _A , _A , _A = 0, 1, m while va != 0: _A = ua // va _A , _A , _A , _A , _A , _A = (ua - q * va), (ua - q * va), (ua - q * va), va, va, va return ua % m
621
0
"""simple docstring""" import argparse import json from collections import OrderedDict from pathlib import Path import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ( ConditionalDetrConfig, ConditionalDetrForObjectDetection, ConditionalDetrForSegmentation, ConditionalDetrImageProcessor, ) from transformers.utils import logging logging.set_verbosity_info() a_ : Union[str, Any] = logging.get_logger(__name__) # here we list all keys to be renamed (original name on the left, our name on the right) a_ : Optional[Any] = [] for i in range(6): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.weight''', f'''encoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.bias''', f'''encoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.weight''', f'''encoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.bias''', f'''encoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.weight''', f'''encoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.bias''', f'''encoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.encoder.layers.{i}.norm1.weight''', f'''encoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.norm1.bias''', f'''encoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.weight''', f'''encoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.bias''', f'''encoder.layers.{i}.final_layer_norm.bias''')) # decoder layers: 2 times output projection, 2 feedforward neural networks and 3 layernorms rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.weight''', f'''decoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.bias''', f'''decoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.weight''', f'''decoder.layers.{i}.encoder_attn.out_proj.weight''', ) ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.bias''', f'''decoder.layers.{i}.encoder_attn.out_proj.bias''', ) ) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.weight''', f'''decoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.bias''', f'''decoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.weight''', f'''decoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.bias''', f'''decoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm1.weight''', f'''decoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm1.bias''', f'''decoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.weight''', f'''decoder.layers.{i}.encoder_attn_layer_norm.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.bias''', f'''decoder.layers.{i}.encoder_attn_layer_norm.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.weight''', f'''decoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.bias''', f'''decoder.layers.{i}.final_layer_norm.bias''')) # q, k, v projections in self/cross-attention in decoder for conditional DETR rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.weight''', f'''decoder.layers.{i}.sa_qcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.weight''', f'''decoder.layers.{i}.sa_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qpos_proj.weight''', f'''decoder.layers.{i}.sa_qpos_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kpos_proj.weight''', f'''decoder.layers.{i}.sa_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.weight''', f'''decoder.layers.{i}.sa_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.weight''', f'''decoder.layers.{i}.ca_qcontent_proj.weight''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.weight", f"decoder.layers.{i}.ca_qpos_proj.weight")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.weight''', f'''decoder.layers.{i}.ca_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kpos_proj.weight''', f'''decoder.layers.{i}.ca_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.weight''', f'''decoder.layers.{i}.ca_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.weight''', f'''decoder.layers.{i}.ca_qpos_sine_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.bias''', f'''decoder.layers.{i}.sa_qcontent_proj.bias''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.bias''', f'''decoder.layers.{i}.sa_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_qpos_proj.bias''', f'''decoder.layers.{i}.sa_qpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_kpos_proj.bias''', f'''decoder.layers.{i}.sa_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.bias''', f'''decoder.layers.{i}.sa_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.bias''', f'''decoder.layers.{i}.ca_qcontent_proj.bias''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.bias", f"decoder.layers.{i}.ca_qpos_proj.bias")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.bias''', f'''decoder.layers.{i}.ca_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_kpos_proj.bias''', f'''decoder.layers.{i}.ca_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.bias''', f'''decoder.layers.{i}.ca_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.bias''', f'''decoder.layers.{i}.ca_qpos_sine_proj.bias''') ) # convolutional projection + query embeddings + layernorm of decoder + class and bounding box heads # for conditional DETR, also convert reference point head and query scale MLP rename_keys.extend( [ ("input_proj.weight", "input_projection.weight"), ("input_proj.bias", "input_projection.bias"), ("query_embed.weight", "query_position_embeddings.weight"), ("transformer.decoder.norm.weight", "decoder.layernorm.weight"), ("transformer.decoder.norm.bias", "decoder.layernorm.bias"), ("class_embed.weight", "class_labels_classifier.weight"), ("class_embed.bias", "class_labels_classifier.bias"), ("bbox_embed.layers.0.weight", "bbox_predictor.layers.0.weight"), ("bbox_embed.layers.0.bias", "bbox_predictor.layers.0.bias"), ("bbox_embed.layers.1.weight", "bbox_predictor.layers.1.weight"), ("bbox_embed.layers.1.bias", "bbox_predictor.layers.1.bias"), ("bbox_embed.layers.2.weight", "bbox_predictor.layers.2.weight"), ("bbox_embed.layers.2.bias", "bbox_predictor.layers.2.bias"), ("transformer.decoder.ref_point_head.layers.0.weight", "decoder.ref_point_head.layers.0.weight"), ("transformer.decoder.ref_point_head.layers.0.bias", "decoder.ref_point_head.layers.0.bias"), ("transformer.decoder.ref_point_head.layers.1.weight", "decoder.ref_point_head.layers.1.weight"), ("transformer.decoder.ref_point_head.layers.1.bias", "decoder.ref_point_head.layers.1.bias"), ("transformer.decoder.query_scale.layers.0.weight", "decoder.query_scale.layers.0.weight"), ("transformer.decoder.query_scale.layers.0.bias", "decoder.query_scale.layers.0.bias"), ("transformer.decoder.query_scale.layers.1.weight", "decoder.query_scale.layers.1.weight"), ("transformer.decoder.query_scale.layers.1.bias", "decoder.query_scale.layers.1.bias"), ("transformer.decoder.layers.0.ca_qpos_proj.weight", "decoder.layers.0.ca_qpos_proj.weight"), ("transformer.decoder.layers.0.ca_qpos_proj.bias", "decoder.layers.0.ca_qpos_proj.bias"), ] ) def a__ ( __lowercase , __lowercase , __lowercase ) -> List[str]: _A = state_dict.pop(__lowercase ) _A = val def a__ ( __lowercase ) -> List[str]: _A = OrderedDict() for key, value in state_dict.items(): if "backbone.0.body" in key: _A = key.replace("backbone.0.body" , "backbone.conv_encoder.model" ) _A = value else: _A = value return new_state_dict def a__ ( __lowercase , __lowercase=False ) -> Any: _A = "" if is_panoptic: _A = "conditional_detr." # first: transformer encoder for i in range(6 ): # read in weights + bias of input projection layer (in PyTorch's MultiHeadAttention, this is a single matrix + bias) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_weight""" ) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_bias""" ) # next, add query, keys and values (in that order) to the state dict _A = in_proj_weight[:256, :] _A = in_proj_bias[:256] _A = in_proj_weight[256:512, :] _A = in_proj_bias[256:512] _A = in_proj_weight[-256:, :] _A = in_proj_bias[-256:] def a__ ( ) -> int: _A = "http://images.cocodataset.org/val2017/000000039769.jpg" _A = Image.open(requests.get(__lowercase , stream=__lowercase ).raw ) return im @torch.no_grad() def a__ ( __lowercase , __lowercase ) -> Any: _A = ConditionalDetrConfig() # set backbone and dilation attributes if "resnet101" in model_name: _A = "resnet101" if "dc5" in model_name: _A = True _A = "panoptic" in model_name if is_panoptic: _A = 250 else: _A = 91 _A = "huggingface/label-files" _A = "coco-detection-id2label.json" _A = json.load(open(hf_hub_download(__lowercase , __lowercase , repo_type="dataset" ) , "r" ) ) _A = {int(__lowercase ): v for k, v in idalabel.items()} _A = idalabel _A = {v: k for k, v in idalabel.items()} # load image processor _A = "coco_panoptic" if is_panoptic else "coco_detection" _A = ConditionalDetrImageProcessor(format=__lowercase ) # prepare image _A = prepare_img() _A = image_processor(images=__lowercase , return_tensors="pt" ) _A = encoding["pixel_values"] logger.info(f"""Converting model {model_name}...""" ) # load original model from torch hub _A = torch.hub.load("DeppMeng/ConditionalDETR" , __lowercase , pretrained=__lowercase ).eval() _A = conditional_detr.state_dict() # rename keys for src, dest in rename_keys: if is_panoptic: _A = "conditional_detr." + src rename_key(__lowercase , __lowercase , __lowercase ) _A = rename_backbone_keys(__lowercase ) # query, key and value matrices need special treatment read_in_q_k_v(__lowercase , is_panoptic=__lowercase ) # important: we need to prepend a prefix to each of the base model keys as the head models use different attributes for them _A = "conditional_detr.model." if is_panoptic else "model." for key in state_dict.copy().keys(): if is_panoptic: if ( key.startswith("conditional_detr" ) and not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ) ): _A = state_dict.pop(__lowercase ) _A = val elif "class_labels_classifier" in key or "bbox_predictor" in key: _A = state_dict.pop(__lowercase ) _A = val elif key.startswith("bbox_attention" ) or key.startswith("mask_head" ): continue else: _A = state_dict.pop(__lowercase ) _A = val else: if not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ): _A = state_dict.pop(__lowercase ) _A = val # finally, create HuggingFace model and load state dict _A = ConditionalDetrForSegmentation(__lowercase ) if is_panoptic else ConditionalDetrForObjectDetection(__lowercase ) model.load_state_dict(__lowercase ) model.eval() model.push_to_hub(repo_id=__lowercase , organization="DepuMeng" , commit_message="Add model" ) # verify our conversion _A = conditional_detr(__lowercase ) _A = model(__lowercase ) assert torch.allclose(outputs.logits , original_outputs["pred_logits"] , atol=1E-4 ) assert torch.allclose(outputs.pred_boxes , original_outputs["pred_boxes"] , atol=1E-4 ) if is_panoptic: assert torch.allclose(outputs.pred_masks , original_outputs["pred_masks"] , atol=1E-4 ) # Save model and image processor logger.info(f"""Saving PyTorch model and image processor to {pytorch_dump_folder_path}...""" ) Path(__lowercase ).mkdir(exist_ok=__lowercase ) model.save_pretrained(__lowercase ) image_processor.save_pretrained(__lowercase ) if __name__ == "__main__": a_ : int = argparse.ArgumentParser() parser.add_argument( "--model_name", default="conditional_detr_resnet50", type=str, help="Name of the CONDITIONAL_DETR model you'd like to convert.", ) parser.add_argument( "--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model." ) a_ : Tuple = parser.parse_args() convert_conditional_detr_checkpoint(args.model_name, args.pytorch_dump_folder_path)
713
"""simple docstring""" # 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 snake_case ( _UpperCamelCase): def __init__( self : List[Any] , a__ : Any ) -> Any: '''simple docstring''' _A = data def __iter__( self : List[str] ) -> str: '''simple docstring''' for element in self.data: yield element def a__ ( __lowercase=True ) -> Tuple: _A = Accelerator(even_batches=__lowercase ) assert accelerator.num_processes == 2, "this script expects that two GPUs are available" return accelerator def a__ ( __lowercase , __lowercase , __lowercase , __lowercase = False ) -> Union[str, Any]: if iterable: _A = DummyIterableDataset(torch.as_tensor(range(__lowercase ) ) ) else: _A = TensorDataset(torch.as_tensor(range(__lowercase ) ) ) _A = DataLoader(__lowercase , batch_size=__lowercase ) _A = accelerator.prepare(__lowercase ) return dl def a__ ( __lowercase , __lowercase , __lowercase , __lowercase , __lowercase , ) -> Dict: _A = create_dataloader(accelerator=__lowercase , dataset_size=__lowercase , batch_size=__lowercase ) _A = [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 a__ ( ) -> List[str]: _A = create_accelerator() # without padding, we would expect a different number of batches verify_dataloader_batch_sizes( __lowercase , 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( __lowercase , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 2] , ) def a__ ( ) -> List[Any]: _A = create_accelerator(even_batches=__lowercase ) verify_dataloader_batch_sizes( __lowercase , dataset_size=3 , batch_size=1 , process_0_expected_batch_sizes=[1, 1] , process_1_expected_batch_sizes=[1] , ) verify_dataloader_batch_sizes( __lowercase , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 1] , ) def a__ ( ) -> int: _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) _A = [] with accelerator.join_uneven_inputs([ddp_model] ): for batch_idx, batch in enumerate(__lowercase ): _A = ddp_model(batch[0].float() ) _A = output.sum() loss.backward() batch_idxs.append(__lowercase ) accelerator.wait_for_everyone() if accelerator.process_index == 0: assert batch_idxs == [0, 1] elif accelerator.process_index == 1: assert batch_idxs == [0] def a__ ( __lowercase ) -> List[str]: with warnings.catch_warnings(record=__lowercase ) as w: with accelerator.join_uneven_inputs([Mock()] ): pass assert issubclass(w[-1].category , __lowercase ) assert "only supported for multi-GPU" in str(w[-1].message ) def a__ ( ) -> Tuple: _A = True _A = False _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): _A = train_dl.batch_sampler.even_batches _A = 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 a__ ( ) -> int: _A = True _A = False _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) create_dataloader(__lowercase , dataset_size=3 , batch_size=1 , iterable=__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) with warnings.catch_warnings(): warnings.filterwarnings("ignore" ) try: with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): _A = 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 a__ ( ) -> Optional[Any]: _A = create_accelerator() _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) create_dataloader(__lowercase , dataset_size=3 , batch_size=1 , iterable=__lowercase ) with warnings.catch_warnings(record=__lowercase ) as w: with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): pass assert issubclass(w[-1].category , __lowercase ) assert "only supported for map-style datasets" in str(w[-1].message ) def a__ ( ) -> Optional[Any]: _A = 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" ) _A = accelerator.state.distributed_type _A = DistributedType.FSDP test_join_raises_warning_for_non_ddp_distributed(__lowercase ) _A = original_state if __name__ == "__main__": main()
621
0
"""simple docstring""" 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 a__ ( __lowercase , __lowercase=False ) -> List[Any]: _A = [] 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" _A = [(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 a__ ( __lowercase , __lowercase , __lowercase=False ) -> List[Any]: for i in range(config.num_hidden_layers ): if base_model: _A = "" else: _A = "vit." # read in weights + bias of input projection layer (in timm, this is a single matrix + bias) _A = state_dict.pop(f"""module.blocks.{i}.attn.qkv.weight""" ) _A = state_dict.pop(f"""module.blocks.{i}.attn.qkv.bias""" ) # next, add query, keys and values (in that order) to the state dict _A = in_proj_weight[ : config.hidden_size, : ] _A = in_proj_bias[: config.hidden_size] _A = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] _A = in_proj_bias[ config.hidden_size : config.hidden_size * 2 ] _A = in_proj_weight[ -config.hidden_size :, : ] _A = in_proj_bias[-config.hidden_size :] def a__ ( __lowercase ) -> List[Any]: _A = ["head.weight", "head.bias"] for k in ignore_keys: state_dict.pop(__lowercase , __lowercase ) def a__ ( __lowercase ) -> Union[str, Any]: # projection head is used in the self-supervised pre-training in MSN, # for downstream task it's not needed. _A = [ "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(__lowercase , __lowercase ) def a__ ( __lowercase , __lowercase , __lowercase ) -> Dict: _A = dct.pop(__lowercase ) _A = val def a__ ( __lowercase , __lowercase ) -> Optional[Any]: _A = ViTMSNConfig() _A = 1000 _A = "datasets/huggingface/label-files" _A = "imagenet-1k-id2label.json" _A = json.load(open(hf_hub_download(__lowercase , __lowercase ) , "r" ) ) _A = {int(__lowercase ): v for k, v in idalabel.items()} _A = idalabel _A = {v: k for k, v in idalabel.items()} if "s16" in checkpoint_url: _A = 384 _A = 1536 _A = 6 elif "l16" in checkpoint_url: _A = 1024 _A = 4096 _A = 24 _A = 16 _A = 0.1 elif "b4" in checkpoint_url: _A = 4 elif "l7" in checkpoint_url: _A = 7 _A = 1024 _A = 4096 _A = 24 _A = 16 _A = 0.1 _A = ViTMSNModel(__lowercase ) _A = torch.hub.load_state_dict_from_url(__lowercase , map_location="cpu" )["target_encoder"] _A = ViTImageProcessor(size=config.image_size ) remove_projection_head(__lowercase ) _A = create_rename_keys(__lowercase , base_model=__lowercase ) for src, dest in rename_keys: rename_key(__lowercase , __lowercase , __lowercase ) read_in_q_k_v(__lowercase , __lowercase , base_model=__lowercase ) model.load_state_dict(__lowercase ) model.eval() _A = "http://images.cocodataset.org/val2017/000000039769.jpg" _A = Image.open(requests.get(__lowercase , stream=__lowercase ).raw ) _A = ViTImageProcessor( size=config.image_size , image_mean=__lowercase , image_std=__lowercase ) _A = image_processor(images=__lowercase , return_tensors="pt" ) # forward pass torch.manual_seed(2 ) _A = model(**__lowercase ) _A = 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: _A = torch.tensor([[-1.0_915, -1.4_876, -1.1_809]] ) elif "b16" in checkpoint_url: _A = torch.tensor([[14.2_889, -18.9_045, 11.7_281]] ) elif "l16" in checkpoint_url: _A = torch.tensor([[41.5_028, -22.8_681, 45.6_475]] ) elif "b4" in checkpoint_url: _A = torch.tensor([[-4.3_868, 5.2_932, -0.4_137]] ) else: _A = torch.tensor([[-0.1_792, -0.6_465, 2.4_263]] ) # verify logits assert torch.allclose(last_hidden_state[:, 0, :3] , __lowercase , atol=1E-4 ) print(f"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(__lowercase ) print(f"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(__lowercase ) if __name__ == "__main__": a_ = 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_ = parser.parse_args() convert_vit_msn_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
714
"""simple docstring""" class snake_case : def __init__( self : Optional[int] , a__ : List[Any] , a__ : List[str] , a__ : Tuple ) -> Optional[Any]: '''simple docstring''' _A = None _A = None _A = graph self._normalize_graph(a__ , a__ ) _A = len(a__ ) _A = None def a_ ( self : str , a__ : List[str] , a__ : List[Any] ) -> Dict: '''simple docstring''' if sources is int: _A = [sources] if sinks is int: _A = [sinks] if len(a__ ) == 0 or len(a__ ) == 0: return _A = sources[0] _A = sinks[0] # make fake vertex if there are more # than one source or sink if len(a__ ) > 1 or len(a__ ) > 1: _A = 0 for i in sources: max_input_flow += sum(self.graph[i] ) _A = len(self.graph ) + 1 for room in self.graph: room.insert(0 , 0 ) self.graph.insert(0 , [0] * size ) for i in sources: _A = max_input_flow _A = 0 _A = len(self.graph ) + 1 for room in self.graph: room.append(0 ) self.graph.append([0] * size ) for i in sinks: _A = max_input_flow _A = size - 1 def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' if self.maximum_flow_algorithm is None: raise Exception("You need to set maximum flow algorithm before." ) if self.source_index is None or self.sink_index is None: return 0 self.maximum_flow_algorithm.execute() return self.maximum_flow_algorithm.getMaximumFlow() def a_ ( self : List[Any] , a__ : Optional[Any] ) -> str: '''simple docstring''' _A = algorithm(self ) class snake_case : def __init__( self : List[str] , a__ : List[str] ) -> Union[str, Any]: '''simple docstring''' _A = flow_network _A = flow_network.verticesCount _A = flow_network.sourceIndex _A = flow_network.sinkIndex # it's just a reference, so you shouldn't change # it in your algorithms, use deep copy before doing that _A = flow_network.graph _A = False def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' if not self.executed: self._algorithm() _A = True def a_ ( self : Any ) -> int: '''simple docstring''' pass class snake_case ( _UpperCamelCase): def __init__( self : Optional[Any] , a__ : Dict ) -> List[str]: '''simple docstring''' super().__init__(a__ ) # use this to save your result _A = -1 def a_ ( self : Any ) -> List[str]: '''simple docstring''' if not self.executed: raise Exception("You should execute algorithm before using its result!" ) return self.maximum_flow class snake_case ( _UpperCamelCase): def __init__( self : Union[str, Any] , a__ : Union[str, Any] ) -> Dict: '''simple docstring''' super().__init__(a__ ) _A = [[0] * self.verticies_count for i in range(self.verticies_count )] _A = [0] * self.verticies_count _A = [0] * self.verticies_count def a_ ( self : Any ) -> Dict: '''simple docstring''' _A = self.verticies_count # push some substance to graph for nextvertex_index, bandwidth in enumerate(self.graph[self.source_index] ): self.preflow[self.source_index][nextvertex_index] += bandwidth self.preflow[nextvertex_index][self.source_index] -= bandwidth self.excesses[nextvertex_index] += bandwidth # Relabel-to-front selection rule _A = [ i for i in range(self.verticies_count ) if i != self.source_index and i != self.sink_index ] # move through list _A = 0 while i < len(a__ ): _A = vertices_list[i] _A = self.heights[vertex_index] self.process_vertex(a__ ) if self.heights[vertex_index] > previous_height: # if it was relabeled, swap elements # and start from 0 index vertices_list.insert(0 , vertices_list.pop(a__ ) ) _A = 0 else: i += 1 _A = sum(self.preflow[self.source_index] ) def a_ ( self : Dict , a__ : Any ) -> Optional[int]: '''simple docstring''' while self.excesses[vertex_index] > 0: for neighbour_index in range(self.verticies_count ): # if it's neighbour and current vertex is higher if ( self.graph[vertex_index][neighbour_index] - self.preflow[vertex_index][neighbour_index] > 0 and self.heights[vertex_index] > self.heights[neighbour_index] ): self.push(a__ , a__ ) self.relabel(a__ ) def a_ ( self : str , a__ : Optional[int] , a__ : List[Any] ) -> Optional[int]: '''simple docstring''' _A = min( self.excesses[from_index] , self.graph[from_index][to_index] - self.preflow[from_index][to_index] , ) self.preflow[from_index][to_index] += preflow_delta self.preflow[to_index][from_index] -= preflow_delta self.excesses[from_index] -= preflow_delta self.excesses[to_index] += preflow_delta def a_ ( self : Any , a__ : Dict ) -> Any: '''simple docstring''' _A = None for to_index in range(self.verticies_count ): if ( self.graph[vertex_index][to_index] - self.preflow[vertex_index][to_index] > 0 ) and (min_height is None or self.heights[to_index] < min_height): _A = self.heights[to_index] if min_height is not None: _A = min_height + 1 if __name__ == "__main__": a_ = [0] a_ = [3] # graph = [ # [0, 0, 4, 6, 0, 0], # [0, 0, 5, 2, 0, 0], # [0, 0, 0, 0, 4, 4], # [0, 0, 0, 0, 6, 6], # [0, 0, 0, 0, 0, 0], # [0, 0, 0, 0, 0, 0], # ] a_ = [[0, 7, 0, 0], [0, 0, 6, 0], [0, 0, 0, 8], [9, 0, 0, 0]] # prepare our network a_ = FlowNetwork(graph, entrances, exits) # set algorithm flow_network.set_maximum_flow_algorithm(PushRelabelExecutor) # and calculate a_ = flow_network.find_maximum_flow() print(f'''maximum flow is {maximum_flow}''')
621
0
"""simple docstring""" import timeit import numpy as np import datasets from datasets.arrow_writer import ArrowWriter from datasets.features.features import _ArrayXD def a__ ( __lowercase ) -> Optional[Any]: def wrapper(*__lowercase , **__lowercase ): _A = timeit.default_timer() _A = func(*__lowercase , **__lowercase ) _A = timeit.default_timer() - starttime return delta _A = func.__name__ return wrapper def a__ ( __lowercase , __lowercase=100 , __lowercase=None ) -> Any: _A = [] _A = seq_shapes or {} for i in range(__lowercase ): _A = {} for col_id, (k, v) in enumerate(features.items() ): if isinstance(__lowercase , _ArrayXD ): _A = np.random.rand(*v.shape ).astype(v.dtype ) elif isinstance(__lowercase , datasets.Value ): if v.dtype == "string": _A = "The small grey turtle was surprisingly fast when challenged." else: _A = np.random.randint(10 , size=1 ).astype(v.dtype ).item() elif isinstance(__lowercase , datasets.Sequence ): while isinstance(__lowercase , datasets.Sequence ): _A = v.feature _A = seq_shapes[k] _A = np.random.rand(*__lowercase ).astype(v.dtype ) _A = data dummy_data.append((i, example) ) return dummy_data def a__ ( __lowercase , __lowercase , __lowercase=100 , __lowercase=None ) -> Union[str, Any]: _A = generate_examples(__lowercase , num_examples=__lowercase , seq_shapes=__lowercase ) with ArrowWriter(features=__lowercase , path=__lowercase ) as writer: for key, record in dummy_data: _A = features.encode_example(__lowercase ) writer.write(__lowercase ) _A , _A = writer.finalize() if not num_final_examples == num_examples: raise ValueError( f"""Error writing the dataset, wrote {num_final_examples} examples but should have written {num_examples}.""" ) _A = datasets.Dataset.from_file(filename=__lowercase , info=datasets.DatasetInfo(features=__lowercase ) ) return dataset
715
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_tf_available, is_tokenizers_available, is_torch_available, ) a_ = { "configuration_roformer": ["ROFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP", "RoFormerConfig", "RoFormerOnnxConfig"], "tokenization_roformer": ["RoFormerTokenizer"], } try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = ["RoFormerTokenizerFast"] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "RoFormerForCausalLM", "RoFormerForMaskedLM", "RoFormerForMultipleChoice", "RoFormerForQuestionAnswering", "RoFormerForSequenceClassification", "RoFormerForTokenClassification", "RoFormerLayer", "RoFormerModel", "RoFormerPreTrainedModel", "load_tf_weights_in_roformer", ] try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "TF_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "TFRoFormerForCausalLM", "TFRoFormerForMaskedLM", "TFRoFormerForMultipleChoice", "TFRoFormerForQuestionAnswering", "TFRoFormerForSequenceClassification", "TFRoFormerForTokenClassification", "TFRoFormerLayer", "TFRoFormerModel", "TFRoFormerPreTrainedModel", ] try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "FLAX_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "FlaxRoFormerForMaskedLM", "FlaxRoFormerForMultipleChoice", "FlaxRoFormerForQuestionAnswering", "FlaxRoFormerForSequenceClassification", "FlaxRoFormerForTokenClassification", "FlaxRoFormerModel", "FlaxRoFormerPreTrainedModel", ] if TYPE_CHECKING: from .configuration_roformer import ROFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, RoFormerConfig, RoFormerOnnxConfig from .tokenization_roformer import RoFormerTokenizer try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_roformer_fast import RoFormerTokenizerFast try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_roformer import ( ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, RoFormerForCausalLM, RoFormerForMaskedLM, RoFormerForMultipleChoice, RoFormerForQuestionAnswering, RoFormerForSequenceClassification, RoFormerForTokenClassification, RoFormerLayer, RoFormerModel, RoFormerPreTrainedModel, load_tf_weights_in_roformer, ) try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_tf_roformer import ( TF_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, TFRoFormerForCausalLM, TFRoFormerForMaskedLM, TFRoFormerForMultipleChoice, TFRoFormerForQuestionAnswering, TFRoFormerForSequenceClassification, TFRoFormerForTokenClassification, TFRoFormerLayer, TFRoFormerModel, TFRoFormerPreTrainedModel, ) try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_flax_roformer import ( FLAX_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, FlaxRoFormerForMaskedLM, FlaxRoFormerForMultipleChoice, FlaxRoFormerForQuestionAnswering, FlaxRoFormerForSequenceClassification, FlaxRoFormerForTokenClassification, FlaxRoFormerModel, FlaxRoFormerPreTrainedModel, ) else: import sys a_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
621
0
"""simple docstring""" import torch from diffusers import DiffusionPipeline class snake_case ( _UpperCamelCase): def __init__( self : int , a__ : Tuple , a__ : Tuple ) -> Optional[int]: '''simple docstring''' super().__init__() self.register_modules(unet=a__ , scheduler=a__ ) def __call__( self : List[str] ) -> str: '''simple docstring''' _A = torch.randn( (1, self.unet.config.in_channels, self.unet.config.sample_size, self.unet.config.sample_size) , ) _A = 1 _A = self.unet(a__ , a__ ).sample _A = self.scheduler.step(a__ , a__ , a__ ).prev_sample _A = scheduler_output - scheduler_output + torch.ones_like(a__ ) return result
716
"""simple docstring""" import warnings from ...utils import logging from .image_processing_glpn import GLPNImageProcessor a_ = logging.get_logger(__name__) class snake_case ( _UpperCamelCase): def __init__( self : str , *a__ : Dict , **a__ : Optional[int] ) -> None: '''simple docstring''' warnings.warn( "The class GLPNFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please" " use GLPNImageProcessor instead." , a__ , ) super().__init__(*a__ , **a__ )
621
0
"""simple docstring""" import argparse import json from pathlib import Path import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import BeitConfig, BeitForImageClassification, BeitForMaskedImageModeling, BeitImageProcessor from transformers.image_utils import PILImageResampling from transformers.utils import logging logging.set_verbosity_info() a_ = logging.get_logger(__name__) def a__ ( __lowercase , __lowercase=False , __lowercase=False ) -> str: _A = "backbone." if is_semantic else "" _A = [] for i in range(config.num_hidden_layers ): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append((f"""{prefix}blocks.{i}.norm1.weight""", f"""beit.encoder.layer.{i}.layernorm_before.weight""") ) rename_keys.append((f"""{prefix}blocks.{i}.norm1.bias""", f"""beit.encoder.layer.{i}.layernorm_before.bias""") ) rename_keys.append( (f"""{prefix}blocks.{i}.attn.proj.weight""", f"""beit.encoder.layer.{i}.attention.output.dense.weight""") ) rename_keys.append( (f"""{prefix}blocks.{i}.attn.proj.bias""", f"""beit.encoder.layer.{i}.attention.output.dense.bias""") ) rename_keys.append((f"""{prefix}blocks.{i}.norm2.weight""", f"""beit.encoder.layer.{i}.layernorm_after.weight""") ) rename_keys.append((f"""{prefix}blocks.{i}.norm2.bias""", f"""beit.encoder.layer.{i}.layernorm_after.bias""") ) rename_keys.append((f"""{prefix}blocks.{i}.mlp.fc1.weight""", f"""beit.encoder.layer.{i}.intermediate.dense.weight""") ) rename_keys.append((f"""{prefix}blocks.{i}.mlp.fc1.bias""", f"""beit.encoder.layer.{i}.intermediate.dense.bias""") ) rename_keys.append((f"""{prefix}blocks.{i}.mlp.fc2.weight""", f"""beit.encoder.layer.{i}.output.dense.weight""") ) rename_keys.append((f"""{prefix}blocks.{i}.mlp.fc2.bias""", f"""beit.encoder.layer.{i}.output.dense.bias""") ) # projection layer + position embeddings rename_keys.extend( [ (f"""{prefix}cls_token""", "beit.embeddings.cls_token"), (f"""{prefix}patch_embed.proj.weight""", "beit.embeddings.patch_embeddings.projection.weight"), (f"""{prefix}patch_embed.proj.bias""", "beit.embeddings.patch_embeddings.projection.bias"), (f"""{prefix}pos_embed""", "beit.embeddings.position_embeddings"), ] ) if has_lm_head: # mask token + layernorm rename_keys.extend( [ ("mask_token", "beit.embeddings.mask_token"), ("norm.weight", "layernorm.weight"), ("norm.bias", "layernorm.bias"), ] ) else: # layernorm + classification head rename_keys.extend( [ ("fc_norm.weight", "beit.pooler.layernorm.weight"), ("fc_norm.bias", "beit.pooler.layernorm.bias"), ("head.weight", "classifier.weight"), ("head.bias", "classifier.bias"), ] ) return rename_keys def a__ ( __lowercase , __lowercase , __lowercase=False , __lowercase=False ) -> Optional[int]: for i in range(config.num_hidden_layers ): _A = "backbone." if is_semantic else "" # queries, keys and values _A = state_dict.pop(f"""{prefix}blocks.{i}.attn.qkv.weight""" ) _A = state_dict.pop(f"""{prefix}blocks.{i}.attn.q_bias""" ) _A = state_dict.pop(f"""{prefix}blocks.{i}.attn.v_bias""" ) _A = in_proj_weight[ : config.hidden_size, : ] _A = q_bias _A = in_proj_weight[ config.hidden_size : config.hidden_size * 2, : ] _A = in_proj_weight[ -config.hidden_size :, : ] _A = v_bias # gamma_1 and gamma_2 # we call them lambda because otherwise they are renamed when using .from_pretrained _A = state_dict.pop(f"""{prefix}blocks.{i}.gamma_1""" ) _A = state_dict.pop(f"""{prefix}blocks.{i}.gamma_2""" ) _A = gamma_a _A = gamma_a def a__ ( __lowercase , __lowercase , __lowercase ) -> Tuple: _A = dct.pop(__lowercase ) _A = val def a__ ( ) -> List[str]: _A = "http://images.cocodataset.org/val2017/000000039769.jpg" _A = Image.open(requests.get(__lowercase , stream=__lowercase ).raw ) return im @torch.no_grad() def a__ ( __lowercase , __lowercase , __lowercase=False ) -> Tuple: _A = False if "rvlcdip" in checkpoint_url else True _A = BeitConfig(use_absolute_position_embeddings=__lowercase , use_mask_token=__lowercase ) # size of the architecture if "large" in checkpoint_url or "dit-l" in checkpoint_url: _A = 1024 _A = 4096 _A = 24 _A = 16 # labels if "rvlcdip" in checkpoint_url: _A = 16 _A = "huggingface/label-files" _A = "rvlcdip-id2label.json" _A = json.load(open(hf_hub_download(__lowercase , __lowercase , repo_type="dataset" ) , "r" ) ) _A = {int(__lowercase ): v for k, v in idalabel.items()} _A = idalabel _A = {v: k for k, v in idalabel.items()} # load state_dict of original model, remove and rename some keys _A = torch.hub.load_state_dict_from_url(__lowercase , map_location="cpu" )["model"] _A = create_rename_keys(__lowercase , has_lm_head=__lowercase ) for src, dest in rename_keys: rename_key(__lowercase , __lowercase , __lowercase ) read_in_q_k_v(__lowercase , __lowercase , has_lm_head=__lowercase ) # load HuggingFace model _A = BeitForMaskedImageModeling(__lowercase ) if has_lm_head else BeitForImageClassification(__lowercase ) model.eval() model.load_state_dict(__lowercase ) # Check outputs on an image _A = BeitImageProcessor( size=config.image_size , resample=PILImageResampling.BILINEAR , do_center_crop=__lowercase ) _A = prepare_img() _A = image_processor(images=__lowercase , return_tensors="pt" ) _A = encoding["pixel_values"] _A = model(__lowercase ) _A = outputs.logits # verify logits _A = [1, 16] if "rvlcdip" in checkpoint_url else [1, 196, 8192] assert logits.shape == torch.Size(__lowercase ), "Shape of logits not as expected" Path(__lowercase ).mkdir(exist_ok=__lowercase ) print(f"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(__lowercase ) print(f"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(__lowercase ) if push_to_hub: if has_lm_head: _A = "dit-base" if "base" in checkpoint_url else "dit-large" else: _A = "dit-base-finetuned-rvlcdip" if "dit-b" in checkpoint_url else "dit-large-finetuned-rvlcdip" image_processor.push_to_hub( repo_path_or_name=Path(__lowercase , __lowercase ) , organization="nielsr" , commit_message="Add image processor" , use_temp_dir=__lowercase , ) model.push_to_hub( repo_path_or_name=Path(__lowercase , __lowercase ) , organization="nielsr" , commit_message="Add model" , use_temp_dir=__lowercase , ) if __name__ == "__main__": a_ = argparse.ArgumentParser() parser.add_argument( "--checkpoint_url", default="https://layoutlm.blob.core.windows.net/dit/dit-pts/dit-base-224-p16-500k-62d53a.pth", type=str, help="URL to the original PyTorch checkpoint (.pth file).", ) parser.add_argument( "--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model." ) parser.add_argument( "--push_to_hub", action="store_true", ) a_ = parser.parse_args() convert_dit_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path, args.push_to_hub)
717
"""simple docstring""" import argparse import torch from torch import nn from transformers import MBartConfig, MBartForConditionalGeneration def a__ ( __lowercase ) -> Optional[int]: _A = [ "encoder.version", "decoder.version", "model.encoder.version", "model.decoder.version", "_float_tensor", "decoder.output_projection.weight", ] for k in ignore_keys: state_dict.pop(__lowercase , __lowercase ) def a__ ( __lowercase ) -> List[Any]: _A , _A = emb.weight.shape _A = nn.Linear(__lowercase , __lowercase , bias=__lowercase ) _A = emb.weight.data return lin_layer def a__ ( __lowercase , __lowercase="facebook/mbart-large-en-ro" , __lowercase=False , __lowercase=False ) -> List[str]: _A = torch.load(__lowercase , map_location="cpu" )["model"] remove_ignore_keys_(__lowercase ) _A = state_dict["encoder.embed_tokens.weight"].shape[0] _A = MBartConfig.from_pretrained(__lowercase , vocab_size=__lowercase ) if mbart_aa and finetuned: _A = "relu" _A = state_dict["decoder.embed_tokens.weight"] _A = MBartForConditionalGeneration(__lowercase ) model.model.load_state_dict(__lowercase ) if finetuned: _A = make_linear_from_emb(model.model.shared ) return model if __name__ == "__main__": a_ = argparse.ArgumentParser() # Required parameters parser.add_argument( "fairseq_path", type=str, help="bart.large, bart.large.cnn or a path to a model.pt on local filesystem." ) parser.add_argument("pytorch_dump_folder_path", default=None, type=str, help="Path to the output PyTorch model.") parser.add_argument( "--hf_config", default="facebook/mbart-large-cc25", type=str, help="Which huggingface architecture to use: mbart-large", ) parser.add_argument("--mbart_50", action="store_true", help="whether the model is mMART-50 checkpoint") parser.add_argument("--finetuned", action="store_true", help="whether the model is a fine-tuned checkpoint") a_ = parser.parse_args() a_ = convert_fairseq_mbart_checkpoint_from_disk( args.fairseq_path, hf_config_path=args.hf_config, finetuned=args.finetuned, mbart_aa=args.mbart_aa ) model.save_pretrained(args.pytorch_dump_folder_path)
621
0
"""simple docstring""" from functools import lru_cache def a__ ( __lowercase ) -> set: _A = 2 _A = set() while i * i <= n: if n % i: i += 1 else: n //= i factors.add(__lowercase ) if n > 1: factors.add(__lowercase ) return factors @lru_cache def a__ ( __lowercase ) -> int: return len(unique_prime_factors(__lowercase ) ) def a__ ( __lowercase ) -> bool: return len(set(__lowercase ) ) in (0, 1) def a__ ( __lowercase ) -> list: _A = 2 while True: # Increment each value of a generated range _A = [base + i for i in range(__lowercase )] # Run elements through out unique_prime_factors function # Append our target number to the end. _A = [upf_len(__lowercase ) for x in group] checker.append(__lowercase ) # If all numbers in the list are equal, return the group variable. if equality(__lowercase ): return group # Increment our base variable by 1 base += 1 def a__ ( __lowercase = 4 ) -> int: _A = run(__lowercase ) return results[0] if len(__lowercase ) else None if __name__ == "__main__": print(solution())
718
"""simple docstring""" import numpy as np def a__ ( __lowercase , __lowercase ) -> np.ndarray: return np.where(vector > 0 , __lowercase , (alpha * (np.exp(__lowercase ) - 1)) ) if __name__ == "__main__": import doctest doctest.testmod()
621
0
"""simple docstring""" 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_ = "src/transformers" a_ = "docs/source/en/tasks" def a__ ( __lowercase , __lowercase , __lowercase ) -> List[str]: with open(__lowercase , "r" , encoding="utf-8" , newline="\n" ) as f: _A = f.readlines() # Find the start prompt. _A = 0 while not lines[start_index].startswith(__lowercase ): start_index += 1 start_index += 1 _A = start_index while not lines[end_index].startswith(__lowercase ): 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_ = direct_transformers_import(TRANSFORMERS_PATH) a_ = { "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_ = { "summarization.md": ("nllb",), "translation.md": ("nllb",), } def a__ ( __lowercase ) -> Dict: _A = TASK_GUIDE_TO_MODELS[task_guide] _A = SPECIAL_TASK_GUIDE_TO_MODEL_TYPES.get(__lowercase , set() ) _A = { 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 a__ ( __lowercase , __lowercase=False ) -> Any: _A , _A , _A , _A = _find_text_in_file( filename=os.path.join(__lowercase , __lowercase ) , start_prompt="<!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->" , end_prompt="<!--End of the generated tip-->" , ) _A = get_model_list_for_task(__lowercase ) if current_list != new_list: if overwrite: with open(os.path.join(__lowercase , __lowercase ) , "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_ = argparse.ArgumentParser() parser.add_argument("--fix_and_overwrite", action="store_true", help="Whether to fix inconsistencies.") a_ = parser.parse_args() for task_guide in TASK_GUIDE_TO_MODELS.keys(): check_model_list_for_task(task_guide, args.fix_and_overwrite)
719
"""simple docstring""" import os import re import warnings from shutil import copyfile from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import PreTrainedTokenizer if TYPE_CHECKING: from ...tokenization_utils_base import TextInput from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"vocab_file": "spiece.model"} a_ = { "vocab_file": { "t5-small": "https://huggingface.co/t5-small/resolve/main/spiece.model", "t5-base": "https://huggingface.co/t5-base/resolve/main/spiece.model", "t5-large": "https://huggingface.co/t5-large/resolve/main/spiece.model", "t5-3b": "https://huggingface.co/t5-3b/resolve/main/spiece.model", "t5-11b": "https://huggingface.co/t5-11b/resolve/main/spiece.model", } } # TODO(PVP) - this should be removed in Transformers v5 a_ = { "t5-small": 5_12, "t5-base": 5_12, "t5-large": 5_12, "t5-3b": 5_12, "t5-11b": 5_12, } a_ = "▁" class snake_case ( _UpperCamelCase): __UpperCamelCase = VOCAB_FILES_NAMES __UpperCamelCase = PRETRAINED_VOCAB_FILES_MAP __UpperCamelCase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __UpperCamelCase = ['input_ids', 'attention_mask'] def __init__( self : List[str] , a__ : Optional[int] , a__ : Union[str, Any]="</s>" , a__ : Union[str, Any]="<unk>" , a__ : str="<pad>" , a__ : Optional[int]=1_00 , a__ : List[Any]=None , a__ : Optional[Dict[str, Any]] = None , a__ : Any=True , **a__ : Optional[int] , ) -> None: '''simple docstring''' if extra_ids > 0 and additional_special_tokens is None: _A = [F"""<extra_id_{i}>""" for i in range(a__ )] elif extra_ids > 0 and additional_special_tokens is not None: # Check that we have the right number of extra_id special tokens _A = len(set(filter(lambda a__ : bool("extra_id" in str(a__ ) ) , a__ ) ) ) if extra_tokens != extra_ids: raise ValueError( F"""Both extra_ids ({extra_ids}) and additional_special_tokens ({additional_special_tokens}) are""" " provided to T5Tokenizer. In this case the additional_special_tokens must include the extra_ids" " tokens" ) if legacy: logger.warning_once( F"""You are using the legacy behaviour of the {self.__class__}. This means that tokens that come after special tokens will not be properly handled. We recommend you to""" " read the related pull request available at https://github.com/huggingface/transformers/pull/24565" ) _A = legacy _A = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( eos_token=a__ , unk_token=a__ , pad_token=a__ , extra_ids=a__ , additional_special_tokens=a__ , sp_model_kwargs=self.sp_model_kwargs , legacy=a__ , **a__ , ) _A = vocab_file _A = extra_ids _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(a__ ) @staticmethod def a_ ( a__ : List[str] , a__ : Optional[int] , a__ : Tuple ) -> Tuple: '''simple docstring''' if pretrained_model_name_or_path in TaTokenizer.max_model_input_sizes: _A = TaTokenizer.max_model_input_sizes[pretrained_model_name_or_path] if init_max_model_length is not None and init_max_model_length != max_model_length: return init_max_model_length elif init_max_model_length is None: warnings.warn( "This tokenizer was incorrectly instantiated with a model max length of" F""" {deprecated_max_model_length} which will be corrected in Transformers v5.\nFor now, this""" " behavior is kept to avoid breaking backwards compatibility when padding/encoding with" " `truncation is True`.\n- Be aware that you SHOULD NOT rely on" F""" {pretrained_model_name_or_path} automatically truncating your input to""" F""" {deprecated_max_model_length} when padding/encoding.\n- If you want to encode/pad to sequences""" F""" longer than {deprecated_max_model_length} you can either instantiate this tokenizer with""" " `model_max_length` or pass `max_length` when encoding/padding.\n- To avoid this warning, please" " instantiate this tokenizer with `model_max_length` set to your preferred value." , a__ , ) return max_model_length @property def a_ ( self : List[Any] ) -> Dict: '''simple docstring''' return self.sp_model.get_piece_size() + self._extra_ids def a_ ( self : Dict ) -> Optional[Any]: '''simple docstring''' _A = {self.convert_ids_to_tokens(a__ ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def a_ ( self : Optional[Any] , a__ : List[int] , a__ : Optional[List[int]] = None , a__ : bool = False ) -> List[int]: '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=a__ , token_ids_a=a__ , already_has_special_tokens=a__ ) # normal case: some special tokens if token_ids_a is None: return ([0] * len(a__ )) + [1] return ([0] * len(a__ )) + [1] + ([0] * len(a__ )) + [1] def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' return list( set(filter(lambda a__ : bool(re.search(r"<extra_id_\d+>" , a__ ) ) is not None , self.additional_special_tokens ) ) ) def a_ ( self : str ) -> List[Any]: '''simple docstring''' return [self._convert_token_to_id(a__ ) for token in self.get_sentinel_tokens()] def a_ ( self : List[Any] , a__ : List[int] ) -> List[int]: '''simple docstring''' if len(a__ ) > 0 and token_ids[-1] == self.eos_token_id: warnings.warn( F"""This sequence already has {self.eos_token}. In future versions this behavior may lead to duplicated""" " eos tokens being added." ) return token_ids else: return token_ids + [self.eos_token_id] def a_ ( self : int , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = [self.eos_token_id] if token_ids_a is None: return len(token_ids_a + eos ) * [0] return len(token_ids_a + eos + token_ids_a + eos ) * [0] def a_ ( self : Union[str, Any] , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = self._add_eos_if_not_present(a__ ) if token_ids_a is None: return token_ids_a else: _A = self._add_eos_if_not_present(a__ ) return token_ids_a + token_ids_a def __getstate__( self : Dict ) -> Union[str, Any]: '''simple docstring''' _A = self.__dict__.copy() _A = None return state def __setstate__( self : int , a__ : Optional[int] ) -> Union[str, Any]: '''simple docstring''' _A = d # for backward compatibility if not hasattr(self , "sp_model_kwargs" ): _A = {} _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def a_ ( self : int , a__ : "TextInput" , **a__ : List[str] ) -> List[str]: '''simple docstring''' if not self.legacy: _A = SPIECE_UNDERLINE + text.replace(a__ , " " ) return super().tokenize(a__ , **a__ ) def a_ ( self : str , a__ : Dict , **a__ : Optional[int] ) -> Any: '''simple docstring''' if not self.legacy: _A = text.startswith(a__ ) if is_first: _A = text[1:] _A = self.sp_model.encode(a__ , out_type=a__ ) if not self.legacy and not is_first and not text.startswith(" " ) and tokens[0].startswith(a__ ): _A = ([tokens[0][1:]] if len(tokens[0] ) > 1 else []) + tokens[1:] return tokens def a_ ( self : int , a__ : List[Any] ) -> List[str]: '''simple docstring''' if token.startswith("<extra_id_" ): _A = re.match(r"<extra_id_(\d+)>" , a__ ) _A = int(match.group(1 ) ) return self.vocab_size - num - 1 return self.sp_model.piece_to_id(a__ ) def a_ ( self : Dict , a__ : Union[str, Any] ) -> Any: '''simple docstring''' if index < self.sp_model.get_piece_size(): _A = self.sp_model.IdToPiece(a__ ) else: _A = F"""<extra_id_{self.vocab_size - 1 - index}>""" return token def a_ ( self : Optional[int] , a__ : Tuple ) -> List[str]: '''simple docstring''' _A = [] _A = "" _A = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(a__ ) + token _A = True _A = [] else: current_sub_tokens.append(a__ ) _A = False out_string += self.sp_model.decode(a__ ) return out_string.strip() def a_ ( self : Dict , a__ : str , a__ : Optional[str] = None ) -> Tuple[str]: '''simple docstring''' if not os.path.isdir(a__ ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return _A = os.path.join( a__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(a__ ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , a__ ) elif not os.path.isfile(self.vocab_file ): with open(a__ , "wb" ) as fi: _A = self.sp_model.serialized_model_proto() fi.write(a__ ) return (out_vocab_file,)
621
0
"""simple docstring""" import random def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = a[left_index] _A = left_index + 1 for j in range(left_index + 1 , __lowercase ): if a[j] < pivot: _A , _A = a[i], a[j] i += 1 _A , _A = a[i - 1], a[left_index] return i - 1 def a__ ( __lowercase , __lowercase , __lowercase ) -> int: if left < right: _A = random.randint(__lowercase , right - 1 ) _A , _A = ( a[left], a[pivot], ) # switches the pivot with the left most bound _A = partition(__lowercase , __lowercase , __lowercase ) quick_sort_random( __lowercase , __lowercase , __lowercase ) # recursive quicksort to the left of the pivot point quick_sort_random( __lowercase , pivot_index + 1 , __lowercase ) # recursive quicksort to the right of the pivot point def a__ ( ) -> Dict: _A = input("Enter numbers separated by a comma:\n" ).strip() _A = [int(__lowercase ) for item in user_input.split("," )] quick_sort_random(__lowercase , 0 , len(__lowercase ) ) print(__lowercase ) if __name__ == "__main__": main()
720
"""simple docstring""" import argparse import json import os from collections import OrderedDict import numpy as np import tensorflow as tf import torch def a__ ( __lowercase ) -> List[Any]: _A = os.path.join(args.tf_model_dir , "parameters.json" ) _A = json.loads(open(__lowercase ).read() ) if not params: raise ValueError( f"""It seems that the json file at {parameter_file} is empty. Make sure you have a correct json file.""" ) if not args.output.endswith(".pt" ): _A = args.output + ".pt" _A = OrderedDict() with tf.device("/CPU:0" ): _A = tf.train.load_checkpoint(args.tf_model_dir ) _A = reader.get_variable_to_shape_map() for key_name in shapes.keys(): _A = reader.get_tensor(__lowercase ).astype(np.floataa ) if key_name.endswith("/adam_m" ) or key_name.endswith("/adam_v" ): continue if key_name.startswith("pasts/" ): if key_name.startswith("pasts/mlp" ): _A = int(key_name[9] ) elif key_name.startswith("pasts/out" ): _A = 8 _A = "model.sqout.%d.weight" % (player * 2) # enter to nn.Sequencial with Tanh, so 2 at a time _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/moe" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/switch_gating/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.router.classifier.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/softmlp/kernel" ): _A = "model.blocks.%d.feed_forward.soft_bypass_mlp.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/wo/kernel" ) or key_name.endswith("/wi/kernel" ): _A = key_name[-9:-7] for i in range(16 ): _A = "model.blocks.%d.feed_forward.mlp.experts.expert_%d.%s.weight" % (player, i, nlayer) _A = ( vnp[i].transpose([1, 0] ).copy() ) # In Mesh-Tensorflow, it is one array, so it is divided _A = torch.tensor(__lowercase ) elif key_name.startswith("model/mlp" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/p1/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wi.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p1/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wi.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wo.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wo.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/ln" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.feed_forward.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.feed_forward.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/att" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/qkv/kernel" ): _A = vnp.copy() # Compute same dimension as Mesh-tensorflow using einsum _A = state[:, 0, :, :] _A = state[:, 1, :, :] _A = state[:, 2, :, :] _A = ( state_q.reshape([state_q.shape[0], state_q.shape[1] * state_q.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_k.reshape([state_k.shape[0], state_k.shape[1] * state_k.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_v.reshape([state_v.shape[0], state_v.shape[1] * state_v.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = "model.blocks.%d.self_attn.self_attn.q_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.k_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.v_proj.weight" % player _A = torch.tensor(__lowercase ) elif key_name.endswith("/o/kernel" ): _A = "model.blocks.%d.self_attn.self_attn.out_proj.weight" % player _A = ( vnp.reshape([vnp.shape[0] * vnp.shape[1], vnp.shape[2]] ).transpose([1, 0] ).copy() ) # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/an" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.self_attn.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.self_attn.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif ( key_name.startswith("model/wte" ) or key_name.startswith("model/wpe" ) or key_name.startswith("model/ete" ) ): _A = {"wte": "embed_tokens", "wpe": "position_embeddings", "ete": "extra_position_embeddings"}[ key_name[-3:] ] _A = "model.%s.weight" % nlayer _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) if key_name.startswith("model/wte" ): _A = "lm_head.weight" _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) elif key_name.startswith("model/wob" ): _A = "final_logits_bias" _A = vnp.copy() # same in embedded _A = state.reshape((1, -1) ) _A = torch.tensor(__lowercase ) elif key_name == "model/dense/kernel": _A = "model.last_project.weight" _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name == "model/dense_1/bias": _A = "model.last_project.bias" _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) torch.save(__lowercase , args.output ) if __name__ == "__main__": a_ = argparse.ArgumentParser( description="model converter.", formatter_class=argparse.ArgumentDefaultsHelpFormatter ) parser.add_argument("--tf_model_dir", metavar="PATH", type=str, required=True, help="import model") parser.add_argument("--output", metavar="PATH", type=str, required=True, help="output model") a_ = parser.parse_args() convert_tf_gptsan_to_pt(args)
621
0
"""simple docstring""" import collections import os import re from pathlib import Path a_ = "src/transformers" # Matches is_xxx_available() a_ = re.compile(r"is\_([a-z_]*)_available()") # Catches a one-line _import_struct = {xxx} a_ = re.compile(r"^_import_structure\s+=\s+\{([^\}]+)\}") # Catches a line with a key-values pattern: "bla": ["foo", "bar"] a_ = re.compile(r"\s+\"\S*\":\s+\[([^\]]*)\]") # Catches a line if not is_foo_available a_ = re.compile(r"^\s*if\s+not\s+is\_[a-z_]*\_available\(\)") # Catches a line _import_struct["bla"].append("foo") a_ = re.compile(r"^\s*_import_structure\[\"\S*\"\]\.append\(\"(\S*)\"\)") # Catches a line _import_struct["bla"].extend(["foo", "bar"]) or _import_struct["bla"] = ["foo", "bar"] a_ = re.compile(r"^\s*_import_structure\[\S*\](?:\.extend\(|\s*=\s+)\[([^\]]*)\]") # Catches a line with an object between quotes and a comma: "MyModel", a_ = re.compile(r"^\s+\"([^\"]+)\",") # Catches a line with objects between brackets only: ["foo", "bar"], a_ = re.compile(r"^\s+\[([^\]]+)\]") # Catches a line with from foo import bar, bla, boo a_ = re.compile(r"\s+from\s+\S*\s+import\s+([^\(\s].*)\n") # Catches a line with try: a_ = re.compile(r"^\s*try:") # Catches a line with else: a_ = re.compile(r"^\s*else:") def a__ ( __lowercase ) -> List[Any]: if _re_test_backend.search(__lowercase ) is None: return None _A = [b[0] for b in _re_backend.findall(__lowercase )] backends.sort() return "_and_".join(__lowercase ) def a__ ( __lowercase ) -> List[Any]: with open(__lowercase , "r" , encoding="utf-8" , newline="\n" ) as f: _A = f.readlines() _A = 0 while line_index < len(__lowercase ) and not lines[line_index].startswith("_import_structure = {" ): line_index += 1 # If this is a traditional init, just return. if line_index >= len(__lowercase ): return None # First grab the objects without a specific backend in _import_structure _A = [] while not lines[line_index].startswith("if TYPE_CHECKING" ) and find_backend(lines[line_index] ) is None: _A = lines[line_index] # If we have everything on a single line, let's deal with it. if _re_one_line_import_struct.search(__lowercase ): _A = _re_one_line_import_struct.search(__lowercase ).groups()[0] _A = re.findall(R"\[([^\]]+)\]" , __lowercase ) for imp in imports: objects.extend([obj[1:-1] for obj in imp.split(", " )] ) line_index += 1 continue _A = _re_import_struct_key_value.search(__lowercase ) if single_line_import_search is not None: _A = [obj[1:-1] for obj in single_line_import_search.groups()[0].split(", " ) if len(__lowercase ) > 0] objects.extend(__lowercase ) elif line.startswith(" " * 8 + "\"" ): objects.append(line[9:-3] ) line_index += 1 _A = {"none": objects} # Let's continue with backend-specific objects in _import_structure while not lines[line_index].startswith("if TYPE_CHECKING" ): # If the line is an if not is_backend_available, we grab all objects associated. _A = find_backend(lines[line_index] ) # Check if the backend declaration is inside a try block: if _re_try.search(lines[line_index - 1] ) is None: _A = None if backend is not None: line_index += 1 # Scroll until we hit the else block of try-except-else while _re_else.search(lines[line_index] ) is None: line_index += 1 line_index += 1 _A = [] # Until we unindent, add backend objects to the list while len(lines[line_index] ) <= 1 or lines[line_index].startswith(" " * 4 ): _A = lines[line_index] if _re_import_struct_add_one.search(__lowercase ) is not None: objects.append(_re_import_struct_add_one.search(__lowercase ).groups()[0] ) elif _re_import_struct_add_many.search(__lowercase ) is not None: _A = _re_import_struct_add_many.search(__lowercase ).groups()[0].split(", " ) _A = [obj[1:-1] for obj in imports if len(__lowercase ) > 0] objects.extend(__lowercase ) elif _re_between_brackets.search(__lowercase ) is not None: _A = _re_between_brackets.search(__lowercase ).groups()[0].split(", " ) _A = [obj[1:-1] for obj in imports if len(__lowercase ) > 0] objects.extend(__lowercase ) elif _re_quote_object.search(__lowercase ) is not None: objects.append(_re_quote_object.search(__lowercase ).groups()[0] ) elif line.startswith(" " * 8 + "\"" ): objects.append(line[9:-3] ) elif line.startswith(" " * 12 + "\"" ): objects.append(line[13:-3] ) line_index += 1 _A = objects else: line_index += 1 # At this stage we are in the TYPE_CHECKING part, first grab the objects without a specific backend _A = [] while ( line_index < len(__lowercase ) and find_backend(lines[line_index] ) is None and not lines[line_index].startswith("else" ) ): _A = lines[line_index] _A = _re_import.search(__lowercase ) if single_line_import_search is not None: objects.extend(single_line_import_search.groups()[0].split(", " ) ) elif line.startswith(" " * 8 ): objects.append(line[8:-2] ) line_index += 1 _A = {"none": objects} # Let's continue with backend-specific objects while line_index < len(__lowercase ): # If the line is an if is_backend_available, we grab all objects associated. _A = find_backend(lines[line_index] ) # Check if the backend declaration is inside a try block: if _re_try.search(lines[line_index - 1] ) is None: _A = None if backend is not None: line_index += 1 # Scroll until we hit the else block of try-except-else while _re_else.search(lines[line_index] ) is None: line_index += 1 line_index += 1 _A = [] # Until we unindent, add backend objects to the list while len(lines[line_index] ) <= 1 or lines[line_index].startswith(" " * 8 ): _A = lines[line_index] _A = _re_import.search(__lowercase ) if single_line_import_search is not None: objects.extend(single_line_import_search.groups()[0].split(", " ) ) elif line.startswith(" " * 12 ): objects.append(line[12:-2] ) line_index += 1 _A = objects else: line_index += 1 return import_dict_objects, type_hint_objects def a__ ( __lowercase , __lowercase ) -> Dict: def find_duplicates(__lowercase ): return [k for k, v in collections.Counter(__lowercase ).items() if v > 1] if list(import_dict_objects.keys() ) != list(type_hint_objects.keys() ): return ["Both sides of the init do not have the same backends!"] _A = [] for key in import_dict_objects.keys(): _A = find_duplicates(import_dict_objects[key] ) if duplicate_imports: errors.append(f"""Duplicate _import_structure definitions for: {duplicate_imports}""" ) _A = find_duplicates(type_hint_objects[key] ) if duplicate_type_hints: errors.append(f"""Duplicate TYPE_CHECKING objects for: {duplicate_type_hints}""" ) if sorted(set(import_dict_objects[key] ) ) != sorted(set(type_hint_objects[key] ) ): _A = "base imports" if key == "none" else f"""{key} backend""" errors.append(f"""Differences for {name}:""" ) for a in type_hint_objects[key]: if a not in import_dict_objects[key]: errors.append(f""" {a} in TYPE_HINT but not in _import_structure.""" ) for a in import_dict_objects[key]: if a not in type_hint_objects[key]: errors.append(f""" {a} in _import_structure but not in TYPE_HINT.""" ) return errors def a__ ( ) -> int: _A = [] for root, _, files in os.walk(__lowercase ): if "__init__.py" in files: _A = os.path.join(__lowercase , "__init__.py" ) _A = parse_init(__lowercase ) if objects is not None: _A = analyze_results(*__lowercase ) if len(__lowercase ) > 0: _A = f"""Problem in {fname}, both halves do not define the same objects.\n{errors[0]}""" failures.append("\n".join(__lowercase ) ) if len(__lowercase ) > 0: raise ValueError("\n\n".join(__lowercase ) ) def a__ ( ) -> Union[str, Any]: _A = [] for path, directories, files in os.walk(__lowercase ): for folder in directories: # Ignore private modules if folder.startswith("_" ): directories.remove(__lowercase ) continue # Ignore leftovers from branches (empty folders apart from pycache) if len(list((Path(__lowercase ) / folder).glob("*.py" ) ) ) == 0: continue _A = str((Path(__lowercase ) / folder).relative_to(__lowercase ) ) _A = short_path.replace(os.path.sep , "." ) submodules.append(__lowercase ) for fname in files: if fname == "__init__.py": continue _A = str((Path(__lowercase ) / fname).relative_to(__lowercase ) ) _A = short_path.replace(".py" , "" ).replace(os.path.sep , "." ) if len(submodule.split("." ) ) == 1: submodules.append(__lowercase ) return submodules a_ = [ "convert_pytorch_checkpoint_to_tf2", "modeling_flax_pytorch_utils", "models.esm.openfold_utils", ] def a__ ( ) -> Optional[Any]: # This is to make sure the transformers module imported is the one in the repo. from transformers.utils import direct_transformers_import _A = direct_transformers_import(__lowercase ) _A = set(transformers._import_structure.keys() ) # This contains all the base keys of the _import_structure object defined in the init, but if the user is missing # some optional dependencies, they may not have all of them. Thus we read the init to read all additions and # (potentiall re-) add them. with open(os.path.join(__lowercase , "__init__.py" ) , "r" ) as f: _A = f.read() import_structure_keys.update(set(re.findall(R"import_structure\[\"([^\"]*)\"\]" , __lowercase ) ) ) _A = [ module for module in get_transformers_submodules() if module not in IGNORE_SUBMODULES and module not in import_structure_keys ] if len(__lowercase ) > 0: _A = "\n".join(f"""- {module}""" for module in module_not_registered ) raise ValueError( "The following submodules are not properly registed in the main init of Transformers:\n" f"""{list_of_modules}\n""" "Make sure they appear somewhere in the keys of `_import_structure` with an empty list as value." ) if __name__ == "__main__": check_all_inits() check_submodules()
721
"""simple docstring""" import argparse import torch from transformers import GPTaLMHeadModel, RobertaForMaskedLM if __name__ == "__main__": a_ = argparse.ArgumentParser( description=( "Extraction some layers of the full RobertaForMaskedLM or GPT2LMHeadModel for Transfer Learned" " Distillation" ) ) parser.add_argument("--model_type", default="roberta", choices=["roberta", "gpt2"]) parser.add_argument("--model_name", default="roberta-large", type=str) parser.add_argument("--dump_checkpoint", default="serialization_dir/tf_roberta_048131723.pth", type=str) parser.add_argument("--vocab_transform", action="store_true") a_ = parser.parse_args() if args.model_type == "roberta": a_ = RobertaForMaskedLM.from_pretrained(args.model_name) a_ = "roberta" elif args.model_type == "gpt2": a_ = GPTaLMHeadModel.from_pretrained(args.model_name) a_ = "transformer" a_ = model.state_dict() a_ = {} # Embeddings # if args.model_type == "gpt2": for param_name in ["wte.weight", "wpe.weight"]: a_ = state_dict[f'''{prefix}.{param_name}'''] else: for w in ["word_embeddings", "position_embeddings", "token_type_embeddings"]: a_ = f'''{prefix}.embeddings.{w}.weight''' a_ = state_dict[param_name] for w in ["weight", "bias"]: a_ = f'''{prefix}.embeddings.LayerNorm.{w}''' a_ = state_dict[param_name] # Transformer Blocks # a_ = 0 for teacher_idx in [0, 2, 4, 7, 9, 11]: if args.model_type == "gpt2": for layer in ["ln_1", "attn.c_attn", "attn.c_proj", "ln_2", "mlp.c_fc", "mlp.c_proj"]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.h.{teacher_idx}.{layer}.{w}''' ] a_ = state_dict[f'''{prefix}.h.{teacher_idx}.attn.bias'''] else: for layer in [ "attention.self.query", "attention.self.key", "attention.self.value", "attention.output.dense", "attention.output.LayerNorm", "intermediate.dense", "output.dense", "output.LayerNorm", ]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.encoder.layer.{teacher_idx}.{layer}.{w}''' ] std_idx += 1 # Language Modeling Head ###s if args.model_type == "roberta": for layer in ["lm_head.decoder.weight", "lm_head.bias"]: a_ = state_dict[f'''{layer}'''] if args.vocab_transform: for w in ["weight", "bias"]: a_ = state_dict[f'''lm_head.dense.{w}'''] a_ = state_dict[f'''lm_head.layer_norm.{w}'''] elif args.model_type == "gpt2": for w in ["weight", "bias"]: a_ = state_dict[f'''{prefix}.ln_f.{w}'''] a_ = state_dict["lm_head.weight"] print(f'''N layers selected for distillation: {std_idx}''') print(f'''Number of params transferred for distillation: {len(compressed_sd.keys())}''') print(f'''Save transferred checkpoint to {args.dump_checkpoint}.''') torch.save(compressed_sd, args.dump_checkpoint)
621
0
"""simple docstring""" from dataclasses import asdict, dataclass from typing import Optional from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) # TODO Update this a_ = { "facebook/esm-1b": "https://huggingface.co/facebook/esm-1b/resolve/main/config.json", # See all ESM models at https://huggingface.co/models?filter=esm } class snake_case ( _UpperCamelCase): __UpperCamelCase = 'esm' def __init__( self : Optional[Any] , a__ : List[Any]=None , a__ : int=None , a__ : List[Any]=None , a__ : Optional[int]=7_68 , a__ : str=12 , a__ : str=12 , a__ : str=30_72 , a__ : List[str]=0.1 , a__ : Tuple=0.1 , a__ : List[Any]=10_26 , a__ : Any=0.0_2 , a__ : Any=1E-1_2 , a__ : List[str]="absolute" , a__ : Optional[Any]=True , a__ : Union[str, Any]=None , a__ : Dict=False , a__ : Dict=False , a__ : Any=None , a__ : str=None , **a__ : Optional[int] , ) -> int: '''simple docstring''' super().__init__(pad_token_id=a__ , mask_token_id=a__ , **a__ ) _A = vocab_size _A = hidden_size _A = num_hidden_layers _A = num_attention_heads _A = intermediate_size _A = hidden_dropout_prob _A = attention_probs_dropout_prob _A = max_position_embeddings _A = initializer_range _A = layer_norm_eps _A = position_embedding_type _A = use_cache _A = emb_layer_norm_before _A = token_dropout _A = is_folding_model if is_folding_model: if esmfold_config is None: logger.info("No esmfold_config supplied for folding model, using default values." ) _A = EsmFoldConfig() elif isinstance(a__ , a__ ): _A = EsmFoldConfig(**a__ ) _A = esmfold_config if vocab_list is None: logger.warning("No vocab_list supplied for folding model, assuming the ESM-2 vocabulary!" ) _A = get_default_vocab_list() else: _A = vocab_list else: _A = None _A = None if self.esmfold_config is not None and getattr(self.esmfold_config , "use_esm_attn_map" , a__ ): raise ValueError("The HuggingFace port of ESMFold does not support use_esm_attn_map at this time!" ) def a_ ( self : int ) -> int: '''simple docstring''' _A = super().to_dict() if isinstance(self.esmfold_config , a__ ): _A = self.esmfold_config.to_dict() return output @dataclass class snake_case : __UpperCamelCase = None __UpperCamelCase = True __UpperCamelCase = False __UpperCamelCase = False __UpperCamelCase = False __UpperCamelCase = 0 __UpperCamelCase = True __UpperCamelCase = False __UpperCamelCase = 128 __UpperCamelCase = None def a_ ( self : str ) -> Optional[int]: '''simple docstring''' if self.trunk is None: _A = TrunkConfig() elif isinstance(self.trunk , a__ ): _A = TrunkConfig(**self.trunk ) def a_ ( self : List[str] ) -> Optional[Any]: '''simple docstring''' _A = asdict(self ) _A = self.trunk.to_dict() return output @dataclass class snake_case : __UpperCamelCase = 48 __UpperCamelCase = 1024 __UpperCamelCase = 128 __UpperCamelCase = 32 __UpperCamelCase = 32 __UpperCamelCase = 32 __UpperCamelCase = 0 __UpperCamelCase = 0 __UpperCamelCase = False __UpperCamelCase = 4 __UpperCamelCase = 128 __UpperCamelCase = None def a_ ( self : str ) -> Union[str, Any]: '''simple docstring''' if self.structure_module is None: _A = StructureModuleConfig() elif isinstance(self.structure_module , a__ ): _A = StructureModuleConfig(**self.structure_module ) if self.max_recycles <= 0: raise ValueError(F"""`max_recycles` should be positive, got {self.max_recycles}.""" ) if self.sequence_state_dim % self.sequence_state_dim != 0: raise ValueError( "`sequence_state_dim` should be a round multiple of `sequence_state_dim`, got" F""" {self.sequence_state_dim} and {self.sequence_state_dim}.""" ) if self.pairwise_state_dim % self.pairwise_state_dim != 0: raise ValueError( "`pairwise_state_dim` should be a round multiple of `pairwise_state_dim`, got" F""" {self.pairwise_state_dim} and {self.pairwise_state_dim}.""" ) _A = self.sequence_state_dim // self.sequence_head_width _A = self.pairwise_state_dim // self.pairwise_head_width if self.sequence_state_dim != sequence_num_heads * self.sequence_head_width: raise ValueError( "`sequence_state_dim` should be equal to `sequence_num_heads * sequence_head_width, got" F""" {self.sequence_state_dim} != {sequence_num_heads} * {self.sequence_head_width}.""" ) if self.pairwise_state_dim != pairwise_num_heads * self.pairwise_head_width: raise ValueError( "`pairwise_state_dim` should be equal to `pairwise_num_heads * pairwise_head_width, got" F""" {self.pairwise_state_dim} != {pairwise_num_heads} * {self.pairwise_head_width}.""" ) if self.pairwise_state_dim % 2 != 0: raise ValueError(F"""`pairwise_state_dim` should be even, got {self.pairwise_state_dim}.""" ) if self.dropout >= 0.4: raise ValueError(F"""`dropout` should not be greater than 0.4, got {self.dropout}.""" ) def a_ ( self : List[Any] ) -> Optional[Any]: '''simple docstring''' _A = asdict(self ) _A = self.structure_module.to_dict() return output @dataclass class snake_case : __UpperCamelCase = 384 __UpperCamelCase = 128 __UpperCamelCase = 16 __UpperCamelCase = 128 __UpperCamelCase = 12 __UpperCamelCase = 4 __UpperCamelCase = 8 __UpperCamelCase = 0.1 __UpperCamelCase = 8 __UpperCamelCase = 1 __UpperCamelCase = 2 __UpperCamelCase = 7 __UpperCamelCase = 10 __UpperCamelCase = 1e-8 __UpperCamelCase = 1e5 def a_ ( self : List[Any] ) -> List[Any]: '''simple docstring''' return asdict(self ) def a__ ( ) -> int: return ( "<cls>", "<pad>", "<eos>", "<unk>", "L", "A", "G", "V", "S", "E", "R", "T", "I", "D", "P", "K", "Q", "N", "F", "Y", "M", "H", "W", "C", "X", "B", "U", "Z", "O", ".", "-", "<null_1>", "<mask>", )
700
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a_ = { "configuration_upernet": ["UperNetConfig"], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "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_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
621
0
"""simple docstring""" from random import randint from tempfile import TemporaryFile import numpy as np def a__ ( __lowercase , __lowercase , __lowercase ) -> List[Any]: _A = 0 if start < end: _A = randint(__lowercase , __lowercase ) _A = a[end] _A = a[pivot] _A = temp _A , _A = _in_place_partition(__lowercase , __lowercase , __lowercase ) count += _in_place_quick_sort(__lowercase , __lowercase , p - 1 ) count += _in_place_quick_sort(__lowercase , p + 1 , __lowercase ) return count def a__ ( __lowercase , __lowercase , __lowercase ) -> List[str]: _A = 0 _A = randint(__lowercase , __lowercase ) _A = a[end] _A = a[pivot] _A = temp _A = start - 1 for index in range(__lowercase , __lowercase ): count += 1 if a[index] < a[end]: # check if current val is less than pivot value _A = new_pivot_index + 1 _A = a[new_pivot_index] _A = a[index] _A = temp _A = a[new_pivot_index + 1] _A = a[end] _A = temp return new_pivot_index + 1, count a_ = TemporaryFile() a_ = 1_00 # 1000 elements are to be sorted a_ , a_ = 0, 1 # mean and standard deviation a_ = np.random.normal(mu, sigma, p) np.save(outfile, X) print("The array is") print(X) outfile.seek(0) # using the same array a_ = np.load(outfile) a_ = len(M) - 1 a_ = _in_place_quick_sort(M, 0, r) print( "No of Comparisons for 100 elements selected from a standard normal distribution" "is :" ) print(z)
701
"""simple docstring""" import random import unittest from torch.utils.data import BatchSampler, DataLoader, IterableDataset from accelerate import Accelerator from accelerate.data_loader import ( BatchSamplerShard, DataLoaderDispatcher, DataLoaderShard, IterableDatasetShard, SkipBatchSampler, SkipDataLoader, skip_first_batches, ) class snake_case ( _UpperCamelCase): def __init__( self : Optional[int] , a__ : str=0.0_1 , a__ : str=10_00 ) -> int: '''simple docstring''' _A = p_stop _A = max_length def __iter__( self : Any ) -> Optional[Any]: '''simple docstring''' _A = 0 _A = False while not stop and count < self.max_length: yield count count += 1 _A = random.random() < self.p_stop class snake_case ( unittest.TestCase): def a_ ( self : List[Any] , a__ : Union[str, Any] , a__ : Union[str, Any] , a__ : List[str]=False , a__ : str=True ) -> Union[str, Any]: '''simple docstring''' _A = [ BatchSamplerShard(a__ , 2 , a__ , split_batches=a__ , even_batches=a__ ) for i in range(2 ) ] _A = [list(a__ ) for batch_sampler_shard in batch_sampler_shards] if not split_batches: self.assertListEqual([len(a__ ) for shard in batch_sampler_shards] , [len(a__ ) for e in expected] ) self.assertListEqual(a__ , a__ ) def a_ ( self : List[Any] ) -> str: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 22, 23]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is a round multiple of batch size but not total batch size. _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [0, 1, 2]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is not a round multiple of batch size but has a multiple of # num_processes batch. _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 0, 1]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is not a round multiple of batch size but and has not a multiple of # num_processes batch. _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 0]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [1, 2, 3]], ] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[[0, 1, 0]], [[1, 0, 1]]] self.check_batch_sampler_shards(a__ , a__ ) _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ ) def a_ ( self : int ) -> int: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size. _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [0, 1]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size or num_processes. _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 0]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [1, 2]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[[0, 1]], [[0, 1]]] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ ) def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=3 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is a round multiple of batch size but not total batch size. _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size but has a multiple of # num_processes batch. _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19, 20]], [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size but and has not a multiple of # num_processes batch. _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14], [18, 19]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(20 ) , batch_size=3 , drop_last=a__ ) _A = [ [[0, 1, 2], [6, 7, 8], [12, 13, 14]], [[3, 4, 5], [9, 10, 11], [15, 16, 17]], ] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[[0, 1]], []] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=3 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , even_batches=a__ ) def a_ ( self : List[str] ) -> str: '''simple docstring''' _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19], [22, 23]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(24 ) , batch_size=4 , drop_last=a__ ) # Expected shouldn't change self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size. _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20, 21]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(22 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is not a round multiple of batch size or num_processes. _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17], [20]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(21 ) , batch_size=4 , drop_last=a__ ) _A = [ [[0, 1], [4, 5], [8, 9], [12, 13], [16, 17]], [[2, 3], [6, 7], [10, 11], [14, 15], [18, 19]], ] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) # Check the shards when the dataset is very small. _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[[0, 1]], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) _A = BatchSampler(range(2 ) , batch_size=4 , drop_last=a__ ) _A = [[], []] self.check_batch_sampler_shards(a__ , a__ , split_batches=a__ , even_batches=a__ ) def a_ ( self : Union[str, Any] ) -> str: '''simple docstring''' _A = [[0, 1, 2], [3, 4], [5, 6, 7, 8], [9, 10, 11], [12, 13]] _A = [BatchSamplerShard(a__ , 2 , a__ , even_batches=a__ ) for i in range(2 )] self.assertEqual(len(batch_sampler_shards[0] ) , 3 ) self.assertEqual(len(batch_sampler_shards[1] ) , 2 ) self.assertListEqual(list(batch_sampler_shards[0] ) , [[0, 1, 2], [5, 6, 7, 8], [12, 13]] ) self.assertListEqual(list(batch_sampler_shards[1] ) , [[3, 4], [9, 10, 11]] ) def a_ ( self : Optional[int] , a__ : Optional[int] , a__ : Tuple , a__ : Optional[int] , a__ : Union[str, Any]=False , a__ : int=2 , a__ : List[Any]=False ) -> str: '''simple docstring''' random.seed(a__ ) _A = list(a__ ) _A = [ IterableDatasetShard( a__ , batch_size=a__ , drop_last=a__ , num_processes=a__ , process_index=a__ , split_batches=a__ , ) for i in range(a__ ) ] _A = [] for iterable_dataset_shard in iterable_dataset_shards: # Since our random iterable dataset will be... random... we need to use a seed to get reproducible results. random.seed(a__ ) iterable_dataset_lists.append(list(a__ ) ) _A = batch_size // num_processes if split_batches else batch_size # All iterable dataset shard should have the same length, a round multiple of shard_batch_size _A = iterable_dataset_lists[0] for l in iterable_dataset_lists[1:]: self.assertEqual(len(a__ ) , len(a__ ) ) self.assertTrue(len(a__ ) % shard_batch_size == 0 ) _A = [] for idx in range(0 , len(a__ ) , a__ ): for l in iterable_dataset_lists: observed += l[idx : idx + shard_batch_size] if not drop_last: while len(a__ ) < len(a__ ): reference += reference self.assertListEqual(a__ , reference[: len(a__ )] ) def a_ ( self : List[str] ) -> List[Any]: '''simple docstring''' _A = 42 _A = RandomIterableDataset() self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) # Edge case with a very small dataset _A = RandomIterableDataset(max_length=2 ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) self.check_iterable_dataset_shards(a__ , a__ , batch_size=4 , drop_last=a__ , split_batches=a__ ) def a_ ( self : List[str] ) -> Dict: '''simple docstring''' _A = BatchSampler(range(16 ) , batch_size=4 , drop_last=a__ ) _A = SkipBatchSampler(a__ , 2 ) self.assertListEqual(list(a__ ) , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : int ) -> Union[str, Any]: '''simple docstring''' _A = SkipDataLoader(list(range(16 ) ) , batch_size=4 , skip_batches=2 ) self.assertListEqual([t.tolist() for t in dataloader] , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : int ) -> Optional[int]: '''simple docstring''' _A = DataLoader(list(range(16 ) ) , batch_size=4 ) _A = skip_first_batches(a__ , num_batches=2 ) self.assertListEqual([t.tolist() for t in new_dataloader] , [[8, 9, 10, 11], [12, 13, 14, 15]] ) def a_ ( self : Optional[Any] ) -> Optional[int]: '''simple docstring''' _A = DataLoaderShard(list(range(16 ) ) , batch_size=4 ) for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) # Test it also works on the second iteration for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) def a_ ( self : int ) -> int: '''simple docstring''' Accelerator() _A = DataLoaderDispatcher(range(16 ) , batch_size=4 ) for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 ) # Test it also works on the second iteration for idx, _ in enumerate(a__ ): self.assertEqual(dataloader.end_of_dataloader , idx == 3 )
621
0
"""simple docstring""" import itertools import os from collections import Counter, defaultdict from concurrent.futures import ThreadPoolExecutor, as_completed import numpy as np import datasets from .execute import check_correctness a_ = "\\n@misc{chen2021evaluating,\n title={Evaluating Large Language Models Trained on Code},\n author={Mark Chen and Jerry Tworek and Heewoo Jun and Qiming Yuan \\nand Henrique Ponde de Oliveira Pinto and Jared Kaplan and Harri Edwards \\nand Yuri Burda and Nicholas Joseph and Greg Brockman and Alex Ray \\nand Raul Puri and Gretchen Krueger and Michael Petrov and Heidy Khlaaf \\nand Girish Sastry and Pamela Mishkin and Brooke Chan and Scott Gray \\nand Nick Ryder and Mikhail Pavlov and Alethea Power and Lukasz Kaiser \\nand Mohammad Bavarian and Clemens Winter and Philippe Tillet \\nand Felipe Petroski Such and Dave Cummings and Matthias Plappert \\nand Fotios Chantzis and Elizabeth Barnes and Ariel Herbert-Voss \\nand William Hebgen Guss and Alex Nichol and Alex Paino and Nikolas Tezak \\nand Jie Tang and Igor Babuschkin and Suchir Balaji and Shantanu Jain \\nand William Saunders and Christopher Hesse and Andrew N. Carr \\nand Jan Leike and Josh Achiam and Vedant Misra and Evan Morikawa \\nand Alec Radford and Matthew Knight and Miles Brundage and Mira Murati \\nand Katie Mayer and Peter Welinder and Bob McGrew and Dario Amodei \\nand Sam McCandlish and Ilya Sutskever and Wojciech Zaremba},\n year={2021},\n eprint={2107.03374},\n archivePrefix={arXiv},\n primaryClass={cs.LG}\n}\n" a_ = "\\nThis metric implements the evaluation harness for the HumanEval problem solving dataset\ndescribed in the paper \"Evaluating Large Language Models Trained on Code\"\n(https://arxiv.org/abs/2107.03374).\n" a_ = "\nCalculates how good are predictions given some references, using certain scores\nArgs:\n predictions: list of candidates to evaluate. Each candidates should be a list\n of strings with several code candidates to solve the problem.\n references: a list with a test for each prediction. Each test should evaluate the\n correctness of a code candidate.\n k: number of code candidates to consider in the evaluation (Default: [1, 10, 100])\n num_workers: number of workers used to evaluate the canidate programs (Default: 4).\n timeout:\nReturns:\n pass_at_k: dict with pass rates for each k\n results: dict with granular results of each unittest\nExamples:\n >>> code_eval = datasets.load_metric(\"code_eval\")\n >>> test_cases = [\"assert add(2,3)==5\"]\n >>> candidates = [[\"def add(a,b): return a*b\", \"def add(a, b): return a+b\"]]\n >>> pass_at_k, results = code_eval.compute(references=test_cases, predictions=candidates, k=[1, 2])\n >>> print(pass_at_k)\n {'pass@1': 0.5, 'pass@2': 1.0}\n" a_ = "\n################################################################################\n !!!WARNING!!!\n################################################################################\nThe \"code_eval\" metric executes untrusted model-generated code in Python.\nAlthough it is highly unlikely that model-generated code will do something\novertly malicious in response to this test suite, model-generated code may act\ndestructively due to a lack of model capability or alignment.\nUsers are strongly encouraged to sandbox this evaluation suite so that it\ndoes not perform destructive actions on their host or network. For more\ninformation on how OpenAI sandboxes its code, see the paper \"Evaluating Large\nLanguage Models Trained on Code\" (https://arxiv.org/abs/2107.03374).\n\nOnce you have read this disclaimer and taken appropriate precautions,\nset the environment variable HF_ALLOW_CODE_EVAL=\"1\". Within Python you can to this\nwith:\n\n>>> import os\n>>> os.environ[\"HF_ALLOW_CODE_EVAL\"] = \"1\"\n\n################################################################################\\n" a_ = "The MIT License\n\nCopyright (c) OpenAI (https://openai.com)\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in\nall copies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN\nTHE SOFTWARE." @datasets.utils.file_utils.add_start_docstrings(_DESCRIPTION , _KWARGS_DESCRIPTION) class snake_case ( datasets.Metric): def a_ ( self : Optional[int] ) -> int: '''simple docstring''' return datasets.MetricInfo( # This is the description that will appear on the metrics page. description=_DESCRIPTION , citation=_CITATION , inputs_description=_KWARGS_DESCRIPTION , features=datasets.Features( { "predictions": datasets.Sequence(datasets.Value("string" ) ), "references": datasets.Value("string" ), } ) , homepage="https://github.com/openai/human-eval" , codebase_urls=["https://github.com/openai/human-eval"] , reference_urls=["https://github.com/openai/human-eval"] , license=_LICENSE , ) def a_ ( self : Tuple , a__ : List[str] , a__ : int , a__ : Union[str, Any]=[1, 10, 1_00] , a__ : int=4 , a__ : Optional[Any]=3.0 ) -> str: '''simple docstring''' if os.getenv("HF_ALLOW_CODE_EVAL" , 0 ) != "1": raise ValueError(_WARNING ) if os.name == "nt": raise NotImplementedError("This metric is currently not supported on Windows." ) with ThreadPoolExecutor(max_workers=a__ ) as executor: _A = [] _A = Counter() _A = 0 _A = defaultdict(a__ ) for task_id, (candidates, test_case) in enumerate(zip(a__ , a__ ) ): for candidate in candidates: _A = candidate + "\n" + test_case _A = (test_program, timeout, task_id, completion_id[task_id]) _A = executor.submit(a__ , *a__ ) futures.append(a__ ) completion_id[task_id] += 1 n_samples += 1 for future in as_completed(a__ ): _A = future.result() results[result["task_id"]].append((result["completion_id"], result) ) _A , _A = [], [] for result in results.values(): result.sort() _A = [r[1]["passed"] for r in result] total.append(len(a__ ) ) correct.append(sum(a__ ) ) _A = np.array(a__ ) _A = np.array(a__ ) _A = k _A = {F"""pass@{k}""": estimate_pass_at_k(a__ , a__ , a__ ).mean() for k in ks if (total >= k).all()} return pass_at_k, results def a__ ( __lowercase , __lowercase , __lowercase ) -> str: def estimator(__lowercase , __lowercase , __lowercase ) -> float: if n - c < k: return 1.0 return 1.0 - np.prod(1.0 - k / np.arange(n - c + 1 , n + 1 ) ) if isinstance(__lowercase , __lowercase ): _A = itertools.repeat(__lowercase , len(__lowercase ) ) else: assert len(__lowercase ) == len(__lowercase ) _A = iter(__lowercase ) return np.array([estimator(int(__lowercase ) , int(__lowercase ) , __lowercase ) for n, c in zip(__lowercase , __lowercase )] )
702
"""simple docstring""" import gc import tempfile import unittest import numpy as np import torch from diffusers import VersatileDiffusionPipeline from diffusers.utils.testing_utils import load_image, nightly, require_torch_gpu, torch_device a_ = False class snake_case ( unittest.TestCase): pass @nightly @require_torch_gpu class snake_case ( unittest.TestCase): def a_ ( self : Optional[int] ) -> str: '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def a_ ( self : Tuple ) -> Any: '''simple docstring''' _A = VersatileDiffusionPipeline.from_pretrained("shi-labs/versatile-diffusion" , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg" ) _A = torch.manual_seed(0 ) _A = pipe.dual_guided( prompt="first prompt" , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" , ).images with tempfile.TemporaryDirectory() as tmpdirname: pipe.save_pretrained(a__ ) _A = VersatileDiffusionPipeline.from_pretrained(a__ , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = generator.manual_seed(0 ) _A = pipe.dual_guided( prompt="first prompt" , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" , ).images assert np.abs(image - new_image ).sum() < 1E-5, "Models don't have the same forward pass" def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' _A = VersatileDiffusionPipeline.from_pretrained("shi-labs/versatile-diffusion" , torch_dtype=torch.floataa ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = "cyberpunk 2077" _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg" ) _A = torch.manual_seed(0 ) _A = pipe.dual_guided( prompt=a__ , image=a__ , text_to_image_strength=0.7_5 , generator=a__ , guidance_scale=7.5 , num_inference_steps=50 , output_type="numpy" , ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.1_4_4_8, 0.1_6_1_9, 0.1_7_4_1, 0.1_0_8_6, 0.1_1_4_7, 0.1_1_2_8, 0.1_1_9_9, 0.1_1_6_5, 0.1_0_0_1] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 _A = "A painting of a squirrel eating a burger " _A = torch.manual_seed(0 ) _A = pipe.text_to_image( prompt=a__ , generator=a__ , guidance_scale=7.5 , num_inference_steps=50 , output_type="numpy" ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.3_3_6_7, 0.3_1_6_9, 0.2_6_5_6, 0.3_8_7_0, 0.4_7_9_0, 0.3_7_9_6, 0.4_0_0_9, 0.4_8_7_8, 0.4_7_7_8] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 _A = pipe.image_variation(a__ , generator=a__ , output_type="numpy" ).images _A = image[0, 2_53:2_56, 2_53:2_56, -1] assert image.shape == (1, 5_12, 5_12, 3) _A = np.array([0.3_0_7_6, 0.3_1_2_3, 0.3_2_8_4, 0.3_7_8_2, 0.3_7_7_0, 0.3_8_9_4, 0.4_2_9_7, 0.4_3_3_1, 0.4_4_5_6] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1
621
0
"""simple docstring""" from __future__ import annotations import requests a_ = set( "approved_at_utc approved_by author_flair_background_color\nauthor_flair_css_class author_flair_richtext author_flair_template_id author_fullname\nauthor_premium can_mod_post category clicked content_categories created_utc downs\nedited gilded gildings hidden hide_score is_created_from_ads_ui is_meta\nis_original_content is_reddit_media_domain is_video link_flair_css_class\nlink_flair_richtext link_flair_text link_flair_text_color media_embed mod_reason_title\nname permalink pwls quarantine saved score secure_media secure_media_embed selftext\nsubreddit subreddit_name_prefixed subreddit_type thumbnail title top_awarded_type\ntotal_awards_received ups upvote_ratio url user_reports".split() ) def a__ ( __lowercase , __lowercase = 1 , __lowercase = "new" , __lowercase = None ) -> dict: _A = wanted_data or [] if invalid_search_terms := ", ".join(sorted(set(__lowercase ) - valid_terms ) ): _A = f"""Invalid search term: {invalid_search_terms}""" raise ValueError(__lowercase ) _A = requests.get( f"""https://reddit.com/r/{subreddit}/{age}.json?limit={limit}""" , headers={"User-agent": "A random string"} , ) if response.status_code == 429: raise requests.HTTPError _A = response.json() if not wanted_data: return {id_: data["data"]["children"][id_] for id_ in range(__lowercase )} _A = {} for id_ in range(__lowercase ): _A = { item: data["data"]["children"][id_]["data"][item] for item in wanted_data } return data_dict if __name__ == "__main__": # If you get Error 429, that means you are rate limited.Try after some time print(get_subreddit_data("learnpython", wanted_data=["title", "url", "selftext"]))
703
"""simple docstring""" import os import time import warnings from dataclasses import dataclass, field from enum import Enum from typing import List, Optional, Union import torch from filelock import FileLock from torch.utils.data import Dataset from ...tokenization_utils_base import PreTrainedTokenizerBase from ...utils import logging from ..processors.glue import glue_convert_examples_to_features, glue_output_modes, glue_processors from ..processors.utils import InputFeatures a_ = logging.get_logger(__name__) @dataclass class snake_case : __UpperCamelCase = field(metadata={'help': 'The name of the task to train on: ' + ', '.join(glue_processors.keys())}) __UpperCamelCase = field( metadata={'help': 'The input data dir. Should contain the .tsv files (or other data files) for the task.'}) __UpperCamelCase = field( default=128 , metadata={ 'help': ( 'The maximum total input sequence length after tokenization. Sequences longer ' 'than this will be truncated, sequences shorter will be padded.' ) } , ) __UpperCamelCase = field( default=_UpperCamelCase , metadata={'help': 'Overwrite the cached training and evaluation sets'}) def a_ ( self : Optional[Any] ) -> Tuple: '''simple docstring''' _A = self.task_name.lower() class snake_case ( _UpperCamelCase): __UpperCamelCase = 'train' __UpperCamelCase = 'dev' __UpperCamelCase = 'test' class snake_case ( _UpperCamelCase): __UpperCamelCase = 42 __UpperCamelCase = 42 __UpperCamelCase = 42 def __init__( self : Optional[int] , a__ : GlueDataTrainingArguments , a__ : PreTrainedTokenizerBase , a__ : Optional[int] = None , a__ : Union[str, Split] = Split.train , a__ : Optional[str] = None , ) -> Tuple: '''simple docstring''' warnings.warn( "This dataset will be removed from the library soon, preprocessing should be handled with the 🤗 Datasets " "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" , a__ , ) _A = args _A = glue_processors[args.task_name]() _A = glue_output_modes[args.task_name] if isinstance(a__ , a__ ): try: _A = Split[mode] except KeyError: raise KeyError("mode is not a valid split name" ) # Load data features from cache or dataset file _A = os.path.join( cache_dir if cache_dir is not None else args.data_dir , F"""cached_{mode.value}_{tokenizer.__class__.__name__}_{args.max_seq_length}_{args.task_name}""" , ) _A = self.processor.get_labels() if args.task_name in ["mnli", "mnli-mm"] and tokenizer.__class__.__name__ in ( "RobertaTokenizer", "RobertaTokenizerFast", "XLMRobertaTokenizer", "BartTokenizer", "BartTokenizerFast", ): # HACK(label indices are swapped in RoBERTa pretrained model) _A , _A = label_list[2], label_list[1] _A = label_list # Make sure only the first process in distributed training processes the dataset, # and the others will use the cache. _A = cached_features_file + ".lock" with FileLock(a__ ): if os.path.exists(a__ ) and not args.overwrite_cache: _A = time.time() _A = torch.load(a__ ) logger.info( F"""Loading features from cached file {cached_features_file} [took %.3f s]""" , time.time() - start ) else: logger.info(F"""Creating features from dataset file at {args.data_dir}""" ) if mode == Split.dev: _A = self.processor.get_dev_examples(args.data_dir ) elif mode == Split.test: _A = self.processor.get_test_examples(args.data_dir ) else: _A = self.processor.get_train_examples(args.data_dir ) if limit_length is not None: _A = examples[:limit_length] _A = glue_convert_examples_to_features( a__ , a__ , max_length=args.max_seq_length , label_list=a__ , output_mode=self.output_mode , ) _A = time.time() torch.save(self.features , a__ ) # ^ This seems to take a lot of time so I want to investigate why and how we can improve. logger.info( F"""Saving features into cached file {cached_features_file} [took {time.time() - start:.3f} s]""" ) def __len__( self : List[Any] ) -> Any: '''simple docstring''' return len(self.features ) def __getitem__( self : Tuple , a__ : Union[str, Any] ) -> InputFeatures: '''simple docstring''' return self.features[i] def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' return self.label_list
621
0
"""simple docstring""" import math def a__ ( __lowercase ) -> bool: if 1 < number < 4: # 2 and 3 are primes return True elif number < 2 or number % 2 == 0 or number % 3 == 0: # Negatives, 0, 1, all even numbers, all multiples of 3 are not primes return False # All primes number are in format of 6k +/- 1 for i in range(5 , int(math.sqrt(__lowercase ) + 1 ) , 6 ): if number % i == 0 or number % (i + 2) == 0: return False return True def a__ ( __lowercase = 0.1 ) -> int: _A = 3 _A = 3 while primes / (2 * j - 1) >= ratio: for i in range(j * j + j + 1 , (j + 2) * (j + 2) , j + 1 ): primes += is_prime(__lowercase ) j += 2 return j if __name__ == "__main__": import doctest doctest.testmod()
704
"""simple docstring""" def a__ ( __lowercase , __lowercase , __lowercase , __lowercase ) -> str: # Return True if there is node that has not iterated. _A = [False] * len(__lowercase ) _A = [] queue.append(__lowercase ) _A = True while queue: _A = queue.pop(0 ) for ind in range(len(graph[u] ) ): if visited[ind] is False and graph[u][ind] > 0: queue.append(__lowercase ) _A = True _A = u return visited[t] def a__ ( __lowercase , __lowercase , __lowercase ) -> int: # This array is filled by BFS and to store path _A = [-1] * (len(__lowercase )) _A = 0 while bfs(__lowercase , __lowercase , __lowercase , __lowercase ): _A = float("Inf" ) _A = sink while s != source: # Find the minimum value in select path _A = min(__lowercase , graph[parent[s]][s] ) _A = parent[s] max_flow += path_flow _A = sink while v != source: _A = parent[v] graph[u][v] -= path_flow graph[v][u] += path_flow _A = parent[v] return max_flow a_ = [ [0, 16, 13, 0, 0, 0], [0, 0, 10, 12, 0, 0], [0, 4, 0, 0, 14, 0], [0, 0, 9, 0, 0, 20], [0, 0, 0, 7, 0, 4], [0, 0, 0, 0, 0, 0], ] a_ , a_ = 0, 5 print(ford_fulkerson(graph, source, sink))
621
0
"""simple docstring""" import asyncio import os import shutil import subprocess import sys import tempfile import unittest from distutils.util import strtobool from functools import partial from pathlib import Path from typing import List, Union from unittest import mock import torch from ..state import AcceleratorState, PartialState from ..utils import ( gather, is_bnb_available, is_comet_ml_available, is_datasets_available, is_deepspeed_available, is_mps_available, is_safetensors_available, is_tensorboard_available, is_torch_version, is_tpu_available, is_transformers_available, is_wandb_available, is_xpu_available, ) def a__ ( __lowercase , __lowercase=False ) -> List[str]: try: _A = os.environ[key] except KeyError: # KEY isn't set, default to `default`. _A = default else: # KEY is set, convert it to True or False. try: _A = strtobool(__lowercase ) except ValueError: # More values are supported, but let's keep the message simple. raise ValueError(f"""If set, {key} must be yes or no.""" ) return _value a_ = parse_flag_from_env("RUN_SLOW", default=False) def a__ ( __lowercase ) -> List[str]: return unittest.skip("Test was skipped" )(__lowercase ) def a__ ( __lowercase ) -> Dict: return unittest.skipUnless(_run_slow_tests , "test is slow" )(__lowercase ) def a__ ( __lowercase ) -> Union[str, Any]: return unittest.skipUnless(not torch.cuda.is_available() , "test requires only a CPU" )(__lowercase ) def a__ ( __lowercase ) -> List[Any]: return unittest.skipUnless(torch.cuda.is_available() , "test requires a GPU" )(__lowercase ) def a__ ( __lowercase ) -> Tuple: return unittest.skipUnless(is_xpu_available() , "test requires a XPU" )(__lowercase ) def a__ ( __lowercase ) -> List[str]: return unittest.skipUnless(is_mps_available() , "test requires a `mps` backend support in `torch`" )(__lowercase ) def a__ ( __lowercase ) -> Optional[int]: return unittest.skipUnless( is_transformers_available() and is_datasets_available() , "test requires the Hugging Face suite" )(__lowercase ) def a__ ( __lowercase ) -> Any: return unittest.skipUnless(is_bnb_available() , "test requires the bitsandbytes library" )(__lowercase ) def a__ ( __lowercase ) -> Optional[Any]: return unittest.skipUnless(is_tpu_available() , "test requires TPU" )(__lowercase ) def a__ ( __lowercase ) -> List[str]: return unittest.skipUnless(torch.cuda.device_count() == 1 , "test requires a GPU" )(__lowercase ) def a__ ( __lowercase ) -> List[Any]: return unittest.skipUnless(torch.xpu.device_count() == 1 , "test requires a XPU" )(__lowercase ) def a__ ( __lowercase ) -> Union[str, Any]: return unittest.skipUnless(torch.cuda.device_count() > 1 , "test requires multiple GPUs" )(__lowercase ) def a__ ( __lowercase ) -> Optional[Any]: return unittest.skipUnless(torch.xpu.device_count() > 1 , "test requires multiple XPUs" )(__lowercase ) def a__ ( __lowercase ) -> Optional[int]: return unittest.skipUnless(is_safetensors_available() , "test requires safetensors" )(__lowercase ) def a__ ( __lowercase ) -> str: return unittest.skipUnless(is_deepspeed_available() , "test requires DeepSpeed" )(__lowercase ) def a__ ( __lowercase ) -> Any: return unittest.skipUnless(is_torch_version(">=" , "1.12.0" ) , "test requires torch version >= 1.12.0" )(__lowercase ) def a__ ( __lowercase=None , __lowercase=None ) -> List[Any]: if test_case is None: return partial(__lowercase , version=__lowercase ) return unittest.skipUnless(is_torch_version(">=" , __lowercase ) , f"""test requires torch version >= {version}""" )(__lowercase ) def a__ ( __lowercase ) -> int: return unittest.skipUnless(is_tensorboard_available() , "test requires Tensorboard" )(__lowercase ) def a__ ( __lowercase ) -> Tuple: return unittest.skipUnless(is_wandb_available() , "test requires wandb" )(__lowercase ) def a__ ( __lowercase ) -> Tuple: return unittest.skipUnless(is_comet_ml_available() , "test requires comet_ml" )(__lowercase ) a_ = ( any([is_wandb_available(), is_tensorboard_available()]) and not is_comet_ml_available() ) def a__ ( __lowercase ) -> Any: return unittest.skipUnless( _atleast_one_tracker_available , "test requires at least one tracker to be available and for `comet_ml` to not be installed" , )(__lowercase ) class snake_case ( unittest.TestCase): __UpperCamelCase = True @classmethod def a_ ( cls : List[Any] ) -> Any: '''simple docstring''' _A = tempfile.mkdtemp() @classmethod def a_ ( cls : Dict ) -> List[str]: '''simple docstring''' if os.path.exists(cls.tmpdir ): shutil.rmtree(cls.tmpdir ) def a_ ( self : List[Any] ) -> Union[str, Any]: '''simple docstring''' if self.clear_on_setup: for path in Path(self.tmpdir ).glob("**/*" ): if path.is_file(): path.unlink() elif path.is_dir(): shutil.rmtree(a__ ) class snake_case ( unittest.TestCase): def a_ ( self : int ) -> Union[str, Any]: '''simple docstring''' super().tearDown() # Reset the state of the AcceleratorState singleton. AcceleratorState._reset_state() PartialState._reset_state() class snake_case ( unittest.TestCase): def a_ ( self : str , a__ : Union[mock.Mock, List[mock.Mock]] ) -> List[str]: '''simple docstring''' _A = mocks if isinstance(a__ , (tuple, list) ) else [mocks] for m in self.mocks: m.start() self.addCleanup(m.stop ) def a__ ( __lowercase ) -> str: _A = AcceleratorState() _A = tensor[None].clone().to(state.device ) _A = gather(__lowercase ).cpu() _A = tensor[0].cpu() for i in range(tensors.shape[0] ): if not torch.equal(tensors[i] , __lowercase ): return False return True class snake_case : def __init__( self : int , a__ : Optional[Any] , a__ : List[str] , a__ : Optional[int] ) -> Union[str, Any]: '''simple docstring''' _A = returncode _A = stdout _A = stderr async def a__ ( __lowercase , __lowercase ) -> Union[str, Any]: while True: _A = await stream.readline() if line: callback(__lowercase ) else: break async def a__ ( __lowercase , __lowercase=None , __lowercase=None , __lowercase=None , __lowercase=False , __lowercase=False ) -> _RunOutput: if echo: print("\nRunning: " , " ".join(__lowercase ) ) _A = await asyncio.create_subprocess_exec( cmd[0] , *cmd[1:] , stdin=__lowercase , stdout=asyncio.subprocess.PIPE , stderr=asyncio.subprocess.PIPE , env=__lowercase , ) # note: there is a warning for a possible deadlock when using `wait` with huge amounts of data in the pipe # https://docs.python.org/3/library/asyncio-subprocess.html#asyncio.asyncio.subprocess.Process.wait # # If it starts hanging, will need to switch to the following code. The problem is that no data # will be seen until it's done and if it hangs for example there will be no debug info. # out, err = await p.communicate() # return _RunOutput(p.returncode, out, err) _A = [] _A = [] def tee(__lowercase , __lowercase , __lowercase , __lowercase="" ): _A = line.decode("utf-8" ).rstrip() sink.append(__lowercase ) if not quiet: print(__lowercase , __lowercase , file=__lowercase ) # XXX: the timeout doesn't seem to make any difference here await asyncio.wait( [ asyncio.create_task(_read_stream(p.stdout , lambda __lowercase : tee(__lowercase , __lowercase , sys.stdout , label="stdout:" ) ) ), asyncio.create_task(_read_stream(p.stderr , lambda __lowercase : tee(__lowercase , __lowercase , sys.stderr , label="stderr:" ) ) ), ] , timeout=__lowercase , ) return _RunOutput(await p.wait() , __lowercase , __lowercase ) def a__ ( __lowercase , __lowercase=None , __lowercase=None , __lowercase=180 , __lowercase=False , __lowercase=True ) -> _RunOutput: _A = asyncio.get_event_loop() _A = loop.run_until_complete( _stream_subprocess(__lowercase , env=__lowercase , stdin=__lowercase , timeout=__lowercase , quiet=__lowercase , echo=__lowercase ) ) _A = " ".join(__lowercase ) if result.returncode > 0: _A = "\n".join(result.stderr ) raise RuntimeError( f"""'{cmd_str}' failed with returncode {result.returncode}\n\n""" f"""The combined stderr from workers follows:\n{stderr}""" ) return result class snake_case ( _UpperCamelCase): pass def a__ ( __lowercase , __lowercase=False ) -> Tuple: try: _A = subprocess.check_output(__lowercase , stderr=subprocess.STDOUT ) if return_stdout: if hasattr(__lowercase , "decode" ): _A = output.decode("utf-8" ) return output except subprocess.CalledProcessError as e: raise SubprocessCallException( f"""Command `{' '.join(__lowercase )}` failed with the following error:\n\n{e.output.decode()}""" ) from e
705
"""simple docstring""" import argparse import json from collections import OrderedDict from pathlib import Path import requests import torch from huggingface_hub import hf_hub_download from PIL import Image from transformers import ( ConditionalDetrConfig, ConditionalDetrForObjectDetection, ConditionalDetrForSegmentation, ConditionalDetrImageProcessor, ) from transformers.utils import logging logging.set_verbosity_info() a_ = logging.get_logger(__name__) # here we list all keys to be renamed (original name on the left, our name on the right) a_ = [] for i in range(6): # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.weight''', f'''encoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.encoder.layers.{i}.self_attn.out_proj.bias''', f'''encoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.weight''', f'''encoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear1.bias''', f'''encoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.weight''', f'''encoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.linear2.bias''', f'''encoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.encoder.layers.{i}.norm1.weight''', f'''encoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.encoder.layers.{i}.norm1.bias''', f'''encoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.weight''', f'''encoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.encoder.layers.{i}.norm2.bias''', f'''encoder.layers.{i}.final_layer_norm.bias''')) # decoder layers: 2 times output projection, 2 feedforward neural networks and 3 layernorms rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.weight''', f'''decoder.layers.{i}.self_attn.out_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.self_attn.out_proj.bias''', f'''decoder.layers.{i}.self_attn.out_proj.bias''') ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.weight''', f'''decoder.layers.{i}.encoder_attn.out_proj.weight''', ) ) rename_keys.append( ( f'''transformer.decoder.layers.{i}.cross_attn.out_proj.bias''', f'''decoder.layers.{i}.encoder_attn.out_proj.bias''', ) ) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.weight''', f'''decoder.layers.{i}.fc1.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear1.bias''', f'''decoder.layers.{i}.fc1.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.weight''', f'''decoder.layers.{i}.fc2.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.linear2.bias''', f'''decoder.layers.{i}.fc2.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm1.weight''', f'''decoder.layers.{i}.self_attn_layer_norm.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm1.bias''', f'''decoder.layers.{i}.self_attn_layer_norm.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.weight''', f'''decoder.layers.{i}.encoder_attn_layer_norm.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.norm2.bias''', f'''decoder.layers.{i}.encoder_attn_layer_norm.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.weight''', f'''decoder.layers.{i}.final_layer_norm.weight''')) rename_keys.append((f'''transformer.decoder.layers.{i}.norm3.bias''', f'''decoder.layers.{i}.final_layer_norm.bias''')) # q, k, v projections in self/cross-attention in decoder for conditional DETR rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.weight''', f'''decoder.layers.{i}.sa_qcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.weight''', f'''decoder.layers.{i}.sa_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qpos_proj.weight''', f'''decoder.layers.{i}.sa_qpos_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kpos_proj.weight''', f'''decoder.layers.{i}.sa_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.weight''', f'''decoder.layers.{i}.sa_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.weight''', f'''decoder.layers.{i}.ca_qcontent_proj.weight''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.weight", f"decoder.layers.{i}.ca_qpos_proj.weight")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.weight''', f'''decoder.layers.{i}.ca_kcontent_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kpos_proj.weight''', f'''decoder.layers.{i}.ca_kpos_proj.weight''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.weight''', f'''decoder.layers.{i}.ca_v_proj.weight''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.weight''', f'''decoder.layers.{i}.ca_qpos_sine_proj.weight''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_qcontent_proj.bias''', f'''decoder.layers.{i}.sa_qcontent_proj.bias''') ) rename_keys.append( (f'''transformer.decoder.layers.{i}.sa_kcontent_proj.bias''', f'''decoder.layers.{i}.sa_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_qpos_proj.bias''', f'''decoder.layers.{i}.sa_qpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_kpos_proj.bias''', f'''decoder.layers.{i}.sa_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.sa_v_proj.bias''', f'''decoder.layers.{i}.sa_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qcontent_proj.bias''', f'''decoder.layers.{i}.ca_qcontent_proj.bias''') ) # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.bias", f"decoder.layers.{i}.ca_qpos_proj.bias")) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_kcontent_proj.bias''', f'''decoder.layers.{i}.ca_kcontent_proj.bias''') ) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_kpos_proj.bias''', f'''decoder.layers.{i}.ca_kpos_proj.bias''')) rename_keys.append((f'''transformer.decoder.layers.{i}.ca_v_proj.bias''', f'''decoder.layers.{i}.ca_v_proj.bias''')) rename_keys.append( (f'''transformer.decoder.layers.{i}.ca_qpos_sine_proj.bias''', f'''decoder.layers.{i}.ca_qpos_sine_proj.bias''') ) # convolutional projection + query embeddings + layernorm of decoder + class and bounding box heads # for conditional DETR, also convert reference point head and query scale MLP rename_keys.extend( [ ("input_proj.weight", "input_projection.weight"), ("input_proj.bias", "input_projection.bias"), ("query_embed.weight", "query_position_embeddings.weight"), ("transformer.decoder.norm.weight", "decoder.layernorm.weight"), ("transformer.decoder.norm.bias", "decoder.layernorm.bias"), ("class_embed.weight", "class_labels_classifier.weight"), ("class_embed.bias", "class_labels_classifier.bias"), ("bbox_embed.layers.0.weight", "bbox_predictor.layers.0.weight"), ("bbox_embed.layers.0.bias", "bbox_predictor.layers.0.bias"), ("bbox_embed.layers.1.weight", "bbox_predictor.layers.1.weight"), ("bbox_embed.layers.1.bias", "bbox_predictor.layers.1.bias"), ("bbox_embed.layers.2.weight", "bbox_predictor.layers.2.weight"), ("bbox_embed.layers.2.bias", "bbox_predictor.layers.2.bias"), ("transformer.decoder.ref_point_head.layers.0.weight", "decoder.ref_point_head.layers.0.weight"), ("transformer.decoder.ref_point_head.layers.0.bias", "decoder.ref_point_head.layers.0.bias"), ("transformer.decoder.ref_point_head.layers.1.weight", "decoder.ref_point_head.layers.1.weight"), ("transformer.decoder.ref_point_head.layers.1.bias", "decoder.ref_point_head.layers.1.bias"), ("transformer.decoder.query_scale.layers.0.weight", "decoder.query_scale.layers.0.weight"), ("transformer.decoder.query_scale.layers.0.bias", "decoder.query_scale.layers.0.bias"), ("transformer.decoder.query_scale.layers.1.weight", "decoder.query_scale.layers.1.weight"), ("transformer.decoder.query_scale.layers.1.bias", "decoder.query_scale.layers.1.bias"), ("transformer.decoder.layers.0.ca_qpos_proj.weight", "decoder.layers.0.ca_qpos_proj.weight"), ("transformer.decoder.layers.0.ca_qpos_proj.bias", "decoder.layers.0.ca_qpos_proj.bias"), ] ) def a__ ( __lowercase , __lowercase , __lowercase ) -> List[str]: _A = state_dict.pop(__lowercase ) _A = val def a__ ( __lowercase ) -> List[str]: _A = OrderedDict() for key, value in state_dict.items(): if "backbone.0.body" in key: _A = key.replace("backbone.0.body" , "backbone.conv_encoder.model" ) _A = value else: _A = value return new_state_dict def a__ ( __lowercase , __lowercase=False ) -> Any: _A = "" if is_panoptic: _A = "conditional_detr." # first: transformer encoder for i in range(6 ): # read in weights + bias of input projection layer (in PyTorch's MultiHeadAttention, this is a single matrix + bias) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_weight""" ) _A = state_dict.pop(f"""{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_bias""" ) # next, add query, keys and values (in that order) to the state dict _A = in_proj_weight[:256, :] _A = in_proj_bias[:256] _A = in_proj_weight[256:512, :] _A = in_proj_bias[256:512] _A = in_proj_weight[-256:, :] _A = in_proj_bias[-256:] def a__ ( ) -> int: _A = "http://images.cocodataset.org/val2017/000000039769.jpg" _A = Image.open(requests.get(__lowercase , stream=__lowercase ).raw ) return im @torch.no_grad() def a__ ( __lowercase , __lowercase ) -> Any: _A = ConditionalDetrConfig() # set backbone and dilation attributes if "resnet101" in model_name: _A = "resnet101" if "dc5" in model_name: _A = True _A = "panoptic" in model_name if is_panoptic: _A = 250 else: _A = 91 _A = "huggingface/label-files" _A = "coco-detection-id2label.json" _A = json.load(open(hf_hub_download(__lowercase , __lowercase , repo_type="dataset" ) , "r" ) ) _A = {int(__lowercase ): v for k, v in idalabel.items()} _A = idalabel _A = {v: k for k, v in idalabel.items()} # load image processor _A = "coco_panoptic" if is_panoptic else "coco_detection" _A = ConditionalDetrImageProcessor(format=__lowercase ) # prepare image _A = prepare_img() _A = image_processor(images=__lowercase , return_tensors="pt" ) _A = encoding["pixel_values"] logger.info(f"""Converting model {model_name}...""" ) # load original model from torch hub _A = torch.hub.load("DeppMeng/ConditionalDETR" , __lowercase , pretrained=__lowercase ).eval() _A = conditional_detr.state_dict() # rename keys for src, dest in rename_keys: if is_panoptic: _A = "conditional_detr." + src rename_key(__lowercase , __lowercase , __lowercase ) _A = rename_backbone_keys(__lowercase ) # query, key and value matrices need special treatment read_in_q_k_v(__lowercase , is_panoptic=__lowercase ) # important: we need to prepend a prefix to each of the base model keys as the head models use different attributes for them _A = "conditional_detr.model." if is_panoptic else "model." for key in state_dict.copy().keys(): if is_panoptic: if ( key.startswith("conditional_detr" ) and not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ) ): _A = state_dict.pop(__lowercase ) _A = val elif "class_labels_classifier" in key or "bbox_predictor" in key: _A = state_dict.pop(__lowercase ) _A = val elif key.startswith("bbox_attention" ) or key.startswith("mask_head" ): continue else: _A = state_dict.pop(__lowercase ) _A = val else: if not key.startswith("class_labels_classifier" ) and not key.startswith("bbox_predictor" ): _A = state_dict.pop(__lowercase ) _A = val # finally, create HuggingFace model and load state dict _A = ConditionalDetrForSegmentation(__lowercase ) if is_panoptic else ConditionalDetrForObjectDetection(__lowercase ) model.load_state_dict(__lowercase ) model.eval() model.push_to_hub(repo_id=__lowercase , organization="DepuMeng" , commit_message="Add model" ) # verify our conversion _A = conditional_detr(__lowercase ) _A = model(__lowercase ) assert torch.allclose(outputs.logits , original_outputs["pred_logits"] , atol=1E-4 ) assert torch.allclose(outputs.pred_boxes , original_outputs["pred_boxes"] , atol=1E-4 ) if is_panoptic: assert torch.allclose(outputs.pred_masks , original_outputs["pred_masks"] , atol=1E-4 ) # Save model and image processor logger.info(f"""Saving PyTorch model and image processor to {pytorch_dump_folder_path}...""" ) Path(__lowercase ).mkdir(exist_ok=__lowercase ) model.save_pretrained(__lowercase ) image_processor.save_pretrained(__lowercase ) if __name__ == "__main__": a_ = argparse.ArgumentParser() parser.add_argument( "--model_name", default="conditional_detr_resnet50", type=str, help="Name of the CONDITIONAL_DETR model you'd like to convert.", ) parser.add_argument( "--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model." ) a_ = parser.parse_args() convert_conditional_detr_checkpoint(args.model_name, args.pytorch_dump_folder_path)
621
0
"""simple docstring""" import os import time import warnings from dataclasses import dataclass, field from enum import Enum from typing import List, Optional, Union import torch from filelock import FileLock from torch.utils.data import Dataset from ...tokenization_utils_base import PreTrainedTokenizerBase from ...utils import logging from ..processors.glue import glue_convert_examples_to_features, glue_output_modes, glue_processors from ..processors.utils import InputFeatures a_ = logging.get_logger(__name__) @dataclass class snake_case : __UpperCamelCase = field(metadata={'help': 'The name of the task to train on: ' + ', '.join(glue_processors.keys())}) __UpperCamelCase = field( metadata={'help': 'The input data dir. Should contain the .tsv files (or other data files) for the task.'}) __UpperCamelCase = field( default=128 , metadata={ 'help': ( 'The maximum total input sequence length after tokenization. Sequences longer ' 'than this will be truncated, sequences shorter will be padded.' ) } , ) __UpperCamelCase = field( default=_UpperCamelCase , metadata={'help': 'Overwrite the cached training and evaluation sets'}) def a_ ( self : Optional[Any] ) -> Tuple: '''simple docstring''' _A = self.task_name.lower() class snake_case ( _UpperCamelCase): __UpperCamelCase = 'train' __UpperCamelCase = 'dev' __UpperCamelCase = 'test' class snake_case ( _UpperCamelCase): __UpperCamelCase = 42 __UpperCamelCase = 42 __UpperCamelCase = 42 def __init__( self : Optional[int] , a__ : GlueDataTrainingArguments , a__ : PreTrainedTokenizerBase , a__ : Optional[int] = None , a__ : Union[str, Split] = Split.train , a__ : Optional[str] = None , ) -> Tuple: '''simple docstring''' warnings.warn( "This dataset will be removed from the library soon, preprocessing should be handled with the 🤗 Datasets " "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" , a__ , ) _A = args _A = glue_processors[args.task_name]() _A = glue_output_modes[args.task_name] if isinstance(a__ , a__ ): try: _A = Split[mode] except KeyError: raise KeyError("mode is not a valid split name" ) # Load data features from cache or dataset file _A = os.path.join( cache_dir if cache_dir is not None else args.data_dir , F"""cached_{mode.value}_{tokenizer.__class__.__name__}_{args.max_seq_length}_{args.task_name}""" , ) _A = self.processor.get_labels() if args.task_name in ["mnli", "mnli-mm"] and tokenizer.__class__.__name__ in ( "RobertaTokenizer", "RobertaTokenizerFast", "XLMRobertaTokenizer", "BartTokenizer", "BartTokenizerFast", ): # HACK(label indices are swapped in RoBERTa pretrained model) _A , _A = label_list[2], label_list[1] _A = label_list # Make sure only the first process in distributed training processes the dataset, # and the others will use the cache. _A = cached_features_file + ".lock" with FileLock(a__ ): if os.path.exists(a__ ) and not args.overwrite_cache: _A = time.time() _A = torch.load(a__ ) logger.info( F"""Loading features from cached file {cached_features_file} [took %.3f s]""" , time.time() - start ) else: logger.info(F"""Creating features from dataset file at {args.data_dir}""" ) if mode == Split.dev: _A = self.processor.get_dev_examples(args.data_dir ) elif mode == Split.test: _A = self.processor.get_test_examples(args.data_dir ) else: _A = self.processor.get_train_examples(args.data_dir ) if limit_length is not None: _A = examples[:limit_length] _A = glue_convert_examples_to_features( a__ , a__ , max_length=args.max_seq_length , label_list=a__ , output_mode=self.output_mode , ) _A = time.time() torch.save(self.features , a__ ) # ^ This seems to take a lot of time so I want to investigate why and how we can improve. logger.info( F"""Saving features into cached file {cached_features_file} [took {time.time() - start:.3f} s]""" ) def __len__( self : List[Any] ) -> Any: '''simple docstring''' return len(self.features ) def __getitem__( self : Tuple , a__ : Union[str, Any] ) -> InputFeatures: '''simple docstring''' return self.features[i] def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' return self.label_list
706
"""simple docstring""" import random def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = a[left_index] _A = left_index + 1 for j in range(left_index + 1 , __lowercase ): if a[j] < pivot: _A , _A = a[i], a[j] i += 1 _A , _A = a[i - 1], a[left_index] return i - 1 def a__ ( __lowercase , __lowercase , __lowercase ) -> int: if left < right: _A = random.randint(__lowercase , right - 1 ) _A , _A = ( a[left], a[pivot], ) # switches the pivot with the left most bound _A = partition(__lowercase , __lowercase , __lowercase ) quick_sort_random( __lowercase , __lowercase , __lowercase ) # recursive quicksort to the left of the pivot point quick_sort_random( __lowercase , pivot_index + 1 , __lowercase ) # recursive quicksort to the right of the pivot point def a__ ( ) -> Dict: _A = input("Enter numbers separated by a comma:\n" ).strip() _A = [int(__lowercase ) for item in user_input.split("," )] quick_sort_random(__lowercase , 0 , len(__lowercase ) ) print(__lowercase ) if __name__ == "__main__": main()
621
0
"""simple docstring""" from collections.abc import Callable from math import pi, sqrt from random import uniform from statistics import mean def a__ ( __lowercase ) -> Tuple: # A local function to see if a dot lands in the circle. def is_in_circle(__lowercase , __lowercase ) -> bool: _A = sqrt((x**2) + (y**2) ) # Our circle has a radius of 1, so a distance # greater than 1 would land outside the circle. return distance_from_centre <= 1 # The proportion of guesses that landed in the circle _A = mean( int(is_in_circle(uniform(-1.0 , 1.0 ) , uniform(-1.0 , 1.0 ) ) ) for _ in range(__lowercase ) ) # The ratio of the area for circle to square is pi/4. _A = proportion * 4 print(f"""The estimated value of pi is {pi_estimate}""" ) print(f"""The numpy value of pi is {pi}""" ) print(f"""The total error is {abs(pi - pi_estimate )}""" ) def a__ ( __lowercase , __lowercase , __lowercase = 0.0 , __lowercase = 1.0 , ) -> float: return mean( function_to_integrate(uniform(__lowercase , __lowercase ) ) for _ in range(__lowercase ) ) * (max_value - min_value) def a__ ( __lowercase , __lowercase = 0.0 , __lowercase = 1.0 ) -> None: def identity_function(__lowercase ) -> float: return x _A = area_under_curve_estimator( __lowercase , __lowercase , __lowercase , __lowercase ) _A = (max_value * max_value - min_value * min_value) / 2 print("******************" ) print(f"""Estimating area under y=x where x varies from {min_value} to {max_value}""" ) print(f"""Estimated value is {estimated_value}""" ) print(f"""Expected value is {expected_value}""" ) print(f"""Total error is {abs(estimated_value - expected_value )}""" ) print("******************" ) def a__ ( __lowercase ) -> None: def function_to_integrate(__lowercase ) -> float: return sqrt(4.0 - x * x ) _A = area_under_curve_estimator( __lowercase , __lowercase , 0.0 , 2.0 ) print("******************" ) print("Estimating pi using area_under_curve_estimator" ) print(f"""Estimated value is {estimated_value}""" ) print(f"""Expected value is {pi}""" ) print(f"""Total error is {abs(estimated_value - pi )}""" ) print("******************" ) if __name__ == "__main__": import doctest doctest.testmod()
707
"""simple docstring""" import copy from typing import Any, Dict, List, Optional, Union import numpy as np 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_ = logging.get_logger(__name__) class snake_case ( _UpperCamelCase): __UpperCamelCase = ['input_features'] def __init__( self : int , a__ : Optional[Any]=80 , a__ : Optional[int]=1_60_00 , a__ : int=1_60 , a__ : Union[str, Any]=30 , a__ : Tuple=4_00 , a__ : List[Any]=0.0 , a__ : Optional[Any]=False , **a__ : List[Any] , ) -> str: '''simple docstring''' super().__init__( feature_size=a__ , sampling_rate=a__ , padding_value=a__ , return_attention_mask=a__ , **a__ , ) _A = n_fft _A = hop_length _A = chunk_length _A = chunk_length * sampling_rate _A = self.n_samples // hop_length _A = sampling_rate _A = mel_filter_bank( num_frequency_bins=1 + n_fft // 2 , num_mel_filters=a__ , min_frequency=0.0 , max_frequency=8_0_0_0.0 , sampling_rate=a__ , norm="slaney" , mel_scale="slaney" , ) def a_ ( self : int , a__ : np.array ) -> np.ndarray: '''simple docstring''' _A = spectrogram( a__ , window_function(self.n_fft , "hann" ) , frame_length=self.n_fft , hop_length=self.hop_length , power=2.0 , mel_filters=self.mel_filters , log_mel="log10" , ) _A = log_spec[:, :-1] _A = np.maximum(a__ , log_spec.max() - 8.0 ) _A = (log_spec + 4.0) / 4.0 return log_spec @staticmethod # Copied from transformers.models.wav2vec2.feature_extraction_wav2vec2.Wav2Vec2FeatureExtractor.zero_mean_unit_var_norm def a_ ( a__ : List[np.ndarray] , a__ : List[np.ndarray] , a__ : float = 0.0 ) -> List[np.ndarray]: '''simple docstring''' if attention_mask is not None: _A = np.array(a__ , np.intaa ) _A = [] for vector, length in zip(a__ , attention_mask.sum(-1 ) ): _A = (vector - vector[:length].mean()) / np.sqrt(vector[:length].var() + 1E-7 ) if length < normed_slice.shape[0]: _A = padding_value normed_input_values.append(a__ ) else: _A = [(x - x.mean()) / np.sqrt(x.var() + 1E-7 ) for x in input_values] return normed_input_values def __call__( self : Optional[int] , a__ : Union[np.ndarray, List[float], List[np.ndarray], List[List[float]]] , a__ : bool = True , a__ : Optional[int] = None , a__ : Optional[Union[str, TensorType]] = None , a__ : Optional[bool] = None , a__ : Optional[str] = "max_length" , a__ : Optional[int] = None , a__ : Optional[int] = None , a__ : Optional[bool] = None , **a__ : Dict , ) -> BatchFeature: '''simple docstring''' 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." ) _A = isinstance(a__ , 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}""" ) _A = is_batched_numpy or ( isinstance(a__ , (list, tuple) ) and (isinstance(raw_speech[0] , (np.ndarray, tuple, list) )) ) if is_batched: _A = [np.asarray([speech] , dtype=np.floataa ).T for speech in raw_speech] elif not is_batched and not isinstance(a__ , np.ndarray ): _A = np.asarray(a__ , dtype=np.floataa ) elif isinstance(a__ , np.ndarray ) and raw_speech.dtype is np.dtype(np.floataa ): _A = raw_speech.astype(np.floataa ) # always return batch if not is_batched: _A = [np.asarray([raw_speech] ).T] _A = BatchFeature({"input_features": raw_speech} ) # convert into correct format for padding _A = self.pad( a__ , padding=a__ , max_length=max_length if max_length else self.n_samples , truncation=a__ , pad_to_multiple_of=a__ , return_attention_mask=return_attention_mask or do_normalize , ) # zero-mean and unit-variance normalization if do_normalize: _A = self.zero_mean_unit_var_norm( padded_inputs["input_features"] , attention_mask=padded_inputs["attention_mask"] , padding_value=self.padding_value , ) _A = np.stack(padded_inputs["input_features"] , axis=0 ) # make sure list is in array format _A = padded_inputs.get("input_features" ).transpose(2 , 0 , 1 ) _A = [self._np_extract_fbank_features(a__ ) for waveform in input_features[0]] if isinstance(input_features[0] , a__ ): _A = [np.asarray(a__ , dtype=np.floataa ) for feature in input_features] else: _A = input_features if return_attention_mask: # rescale from sample (48000) to feature (3000) _A = padded_inputs["attention_mask"][:, :: self.hop_length] if return_tensors is not None: _A = padded_inputs.convert_to_tensors(a__ ) return padded_inputs def a_ ( self : Dict ) -> Dict[str, Any]: '''simple docstring''' _A = copy.deepcopy(self.__dict__ ) _A = self.__class__.__name__ if "mel_filters" in output: del output["mel_filters"] return output
621
0
"""simple docstring""" a_ = [ 9_99, 8_00, 7_99, 6_00, 5_99, 5_00, 4_00, 3_99, 3_77, 3_55, 3_33, 3_11, 2_88, 2_66, 2_44, 2_22, 2_00, 1_99, 1_77, 1_55, 1_33, 1_11, 88, 66, 44, 22, 0, ] a_ = [ 9_99, 9_76, 9_52, 9_28, 9_05, 8_82, 8_58, 8_57, 8_10, 7_62, 7_15, 7_14, 5_72, 4_29, 4_28, 2_86, 2_85, 2_38, 1_90, 1_43, 1_42, 1_18, 95, 71, 47, 24, 0, ] a_ = [ 9_99, 9_88, 9_77, 9_66, 9_55, 9_44, 9_33, 9_22, 9_11, 9_00, 8_99, 8_79, 8_59, 8_40, 8_20, 8_00, 7_99, 7_66, 7_33, 7_00, 6_99, 6_50, 6_00, 5_99, 5_00, 4_99, 4_00, 3_99, 3_50, 3_00, 2_99, 2_66, 2_33, 2_00, 1_99, 1_79, 1_59, 1_40, 1_20, 1_00, 99, 88, 77, 66, 55, 44, 33, 22, 11, 0, ] a_ = [ 9_99, 9_95, 9_92, 9_89, 9_85, 9_81, 9_78, 9_75, 9_71, 9_67, 9_64, 9_61, 9_57, 9_56, 9_51, 9_47, 9_42, 9_37, 9_33, 9_28, 9_23, 9_19, 9_14, 9_13, 9_08, 9_03, 8_97, 8_92, 8_87, 8_81, 8_76, 8_71, 8_70, 8_64, 8_58, 8_52, 8_46, 8_40, 8_34, 8_28, 8_27, 8_20, 8_13, 8_06, 7_99, 7_92, 7_85, 7_84, 7_77, 7_70, 7_63, 7_56, 7_49, 7_42, 7_41, 7_33, 7_24, 7_16, 7_07, 6_99, 6_98, 6_88, 6_77, 6_66, 6_56, 6_55, 6_45, 6_34, 6_23, 6_13, 6_12, 5_98, 5_84, 5_70, 5_69, 5_55, 5_41, 5_27, 5_26, 5_05, 4_84, 4_83, 4_62, 4_40, 4_39, 3_96, 3_95, 3_52, 3_51, 3_08, 3_07, 2_64, 2_63, 2_20, 2_19, 1_76, 1_32, 88, 44, 0, ] a_ = [ 9_99, 9_97, 9_95, 9_92, 9_90, 9_88, 9_86, 9_84, 9_81, 9_79, 9_77, 9_75, 9_72, 9_70, 9_68, 9_66, 9_64, 9_61, 9_59, 9_57, 9_56, 9_54, 9_51, 9_49, 9_46, 9_44, 9_41, 9_39, 9_36, 9_34, 9_31, 9_29, 9_26, 9_24, 9_21, 9_19, 9_16, 9_14, 9_13, 9_10, 9_07, 9_05, 9_02, 8_99, 8_96, 8_93, 8_91, 8_88, 8_85, 8_82, 8_79, 8_77, 8_74, 8_71, 8_70, 8_67, 8_64, 8_61, 8_58, 8_55, 8_52, 8_49, 8_46, 8_43, 8_40, 8_37, 8_34, 8_31, 8_28, 8_27, 8_24, 8_21, 8_17, 8_14, 8_11, 8_08, 8_04, 8_01, 7_98, 7_95, 7_91, 7_88, 7_85, 7_84, 7_80, 7_77, 7_74, 7_70, 7_66, 7_63, 7_60, 7_56, 7_52, 7_49, 7_46, 7_42, 7_41, 7_37, 7_33, 7_30, 7_26, 7_22, 7_18, 7_14, 7_10, 7_07, 7_03, 6_99, 6_98, 6_94, 6_90, 6_85, 6_81, 6_77, 6_73, 6_69, 6_64, 6_60, 6_56, 6_55, 6_50, 6_46, 6_41, 6_36, 6_32, 6_27, 6_22, 6_18, 6_13, 6_12, 6_07, 6_02, 5_96, 5_91, 5_86, 5_80, 5_75, 5_70, 5_69, 5_63, 5_57, 5_51, 5_45, 5_39, 5_33, 5_27, 5_26, 5_19, 5_12, 5_05, 4_98, 4_91, 4_84, 4_83, 4_74, 4_66, 4_57, 4_49, 4_40, 4_39, 4_28, 4_18, 4_07, 3_96, 3_95, 3_81, 3_66, 3_52, 3_51, 3_30, 3_08, 3_07, 2_86, 2_64, 2_63, 2_42, 2_20, 2_19, 1_76, 1_75, 1_32, 1_31, 88, 44, 0, ] a_ = [ 9_99, 9_91, 9_82, 9_74, 9_66, 9_58, 9_50, 9_41, 9_33, 9_25, 9_16, 9_08, 9_00, 8_99, 8_74, 8_50, 8_25, 8_00, 7_99, 7_00, 6_00, 5_00, 4_00, 3_00, 2_00, 1_00, 0, ] a_ = [ 9_99, 9_92, 9_85, 9_78, 9_71, 9_64, 9_57, 9_49, 9_42, 9_35, 9_28, 9_21, 9_14, 9_07, 9_00, 8_99, 8_79, 8_59, 8_40, 8_20, 8_00, 7_99, 7_66, 7_33, 7_00, 6_99, 6_50, 6_00, 5_99, 5_00, 4_99, 4_00, 3_99, 3_00, 2_99, 2_00, 1_99, 1_00, 99, 0, ] a_ = [ 9_99, 9_96, 9_92, 9_89, 9_85, 9_82, 9_79, 9_75, 9_72, 9_68, 9_65, 9_61, 9_58, 9_55, 9_51, 9_48, 9_44, 9_41, 9_38, 9_34, 9_31, 9_27, 9_24, 9_20, 9_17, 9_14, 9_10, 9_07, 9_03, 9_00, 8_99, 8_91, 8_84, 8_76, 8_69, 8_61, 8_53, 8_46, 8_38, 8_30, 8_23, 8_15, 8_08, 8_00, 7_99, 7_88, 7_77, 7_66, 7_55, 7_44, 7_33, 7_22, 7_11, 7_00, 6_99, 6_88, 6_77, 6_66, 6_55, 6_44, 6_33, 6_22, 6_11, 6_00, 5_99, 5_85, 5_71, 5_57, 5_42, 5_28, 5_14, 5_00, 4_99, 4_85, 4_71, 4_57, 4_42, 4_28, 4_14, 4_00, 3_99, 3_79, 3_59, 3_40, 3_20, 3_00, 2_99, 2_79, 2_59, 2_40, 2_20, 2_00, 1_99, 1_66, 1_33, 1_00, 99, 66, 33, 0, ]
708
"""simple docstring""" from __future__ import annotations def a__ ( __lowercase , __lowercase ) -> float: _A = sorted(numsa + numsa ) _A , _A = divmod(len(__lowercase ) , 2 ) if mod == 1: return all_numbers[div] else: return (all_numbers[div] + all_numbers[div - 1]) / 2 if __name__ == "__main__": import doctest doctest.testmod() a_ = [float(x) for x in input("Enter the elements of first array: ").split()] a_ = [float(x) for x in input("Enter the elements of second array: ").split()] print(f'''The median of two arrays is: {median_of_two_arrays(array_a, array_a)}''')
621
0
"""simple docstring""" def a__ ( __lowercase , __lowercase , __lowercase , __lowercase , __lowercase , __lowercase ) -> int: if index == r: for j in range(__lowercase ): print(data[j] , end=" " ) print(" " ) return # When no more elements are there to put in data[] if i >= n: return # current is included, put next at next location _A = arr[i] combination_util(__lowercase , __lowercase , __lowercase , index + 1 , __lowercase , i + 1 ) # current is excluded, replace it with # next (Note that i+1 is passed, but # index is not changed) combination_util(__lowercase , __lowercase , __lowercase , __lowercase , __lowercase , i + 1 ) # The main function that prints all combinations # of size r in arr[] of size n. This function # mainly uses combinationUtil() def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[Any]: # A temporary array to store all combination one by one _A = [0] * r # Print all combination using temporary array 'data[]' combination_util(__lowercase , __lowercase , __lowercase , 0 , __lowercase , 0 ) if __name__ == "__main__": # Driver code to check the function above a_ = [10, 20, 30, 40, 50] print_combination(arr, len(arr), 3) # This code is contributed by Ambuj sahu
709
"""simple docstring""" import copy import os from typing import Union from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { "Salesforce/blip-vqa-base": "https://huggingface.co/Salesforce/blip-vqa-base/resolve/main/config.json", "Salesforce/blip-vqa-capfit-large": ( "https://huggingface.co/Salesforce/blip-vqa-base-capfit/resolve/main/config.json" ), "Salesforce/blip-image-captioning-base": ( "https://huggingface.co/Salesforce/blip-image-captioning-base/resolve/main/config.json" ), "Salesforce/blip-image-captioning-large": ( "https://huggingface.co/Salesforce/blip-image-captioning-large/resolve/main/config.json" ), "Salesforce/blip-itm-base-coco": "https://huggingface.co/Salesforce/blip-itm-base-coco/resolve/main/config.json", "Salesforce/blip-itm-large-coco": "https://huggingface.co/Salesforce/blip-itm-large-coco/resolve/main/config.json", "Salesforce/blip-itm-base-flikr": "https://huggingface.co/Salesforce/blip-itm-base-flikr/resolve/main/config.json", "Salesforce/blip-itm-large-flikr": ( "https://huggingface.co/Salesforce/blip-itm-large-flikr/resolve/main/config.json" ), } class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip_text_model' def __init__( self : int , a__ : List[str]=3_05_24 , a__ : List[str]=7_68 , a__ : List[Any]=7_68 , a__ : int=30_72 , a__ : List[str]=7_68 , a__ : Dict=12 , a__ : Optional[int]=8 , a__ : Optional[Any]=5_12 , a__ : List[Any]="gelu" , a__ : Optional[Any]=1E-1_2 , a__ : Any=0.0 , a__ : int=0.0 , a__ : Dict=0.0_2 , a__ : Optional[Any]=3_05_22 , a__ : Any=2 , a__ : int=0 , a__ : Union[str, Any]=1_02 , a__ : Tuple=True , a__ : Optional[int]=True , **a__ : Any , ) -> List[Any]: '''simple docstring''' super().__init__( pad_token_id=a__ , bos_token_id=a__ , eos_token_id=a__ , sep_token_id=a__ , **a__ , ) _A = vocab_size _A = hidden_size _A = encoder_hidden_size _A = intermediate_size _A = projection_dim _A = hidden_dropout_prob _A = num_hidden_layers _A = num_attention_heads _A = max_position_embeddings _A = layer_norm_eps _A = hidden_act _A = initializer_range _A = attention_probs_dropout_prob _A = is_decoder _A = use_cache @classmethod def a_ ( cls : Optional[Any] , a__ : Union[str, os.PathLike] , **a__ : Optional[Any] ) -> "PretrainedConfig": '''simple docstring''' cls._set_token_in_kwargs(a__ ) _A , _A = cls.get_config_dict(a__ , **a__ ) # get the text config dict if we are loading from BlipConfig if config_dict.get("model_type" ) == "blip": _A = config_dict["text_config"] if "model_type" in config_dict and hasattr(cls , "model_type" ) and config_dict["model_type"] != cls.model_type: logger.warning( F"""You are using a model of type {config_dict['model_type']} to instantiate a model of type """ F"""{cls.model_type}. This is not supported for all configurations of models and can yield errors.""" ) return cls.from_dict(a__ , **a__ ) class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip_vision_model' def __init__( self : Optional[Any] , a__ : Any=7_68 , a__ : List[str]=30_72 , a__ : str=5_12 , a__ : Any=12 , a__ : int=12 , a__ : int=3_84 , a__ : Tuple=16 , a__ : str="gelu" , a__ : Tuple=1E-5 , a__ : List[str]=0.0 , a__ : List[Any]=1E-1_0 , **a__ : int , ) -> List[str]: '''simple docstring''' super().__init__(**a__ ) _A = hidden_size _A = intermediate_size _A = projection_dim _A = num_hidden_layers _A = num_attention_heads _A = patch_size _A = image_size _A = initializer_range _A = attention_dropout _A = layer_norm_eps _A = hidden_act @classmethod def a_ ( cls : Any , a__ : Union[str, os.PathLike] , **a__ : int ) -> "PretrainedConfig": '''simple docstring''' cls._set_token_in_kwargs(a__ ) _A , _A = cls.get_config_dict(a__ , **a__ ) # get the vision config dict if we are loading from BlipConfig if config_dict.get("model_type" ) == "blip": _A = config_dict["vision_config"] if "model_type" in config_dict and hasattr(cls , "model_type" ) and config_dict["model_type"] != cls.model_type: logger.warning( F"""You are using a model of type {config_dict['model_type']} to instantiate a model of type """ F"""{cls.model_type}. This is not supported for all configurations of models and can yield errors.""" ) return cls.from_dict(a__ , **a__ ) class snake_case ( _UpperCamelCase): __UpperCamelCase = 'blip' __UpperCamelCase = True def __init__( self : List[Any] , a__ : Optional[int]=None , a__ : str=None , a__ : List[str]=5_12 , a__ : Any=2.6_5_9_2 , a__ : str=2_56 , **a__ : Optional[int] , ) -> Dict: '''simple docstring''' super().__init__(**a__ ) if text_config is None: _A = {} logger.info("`text_config` is `None`. Initializing the `BlipTextConfig` with default values." ) if vision_config is None: _A = {} logger.info("`vision_config` is `None`. Initializing the `BlipVisionConfig` with default values." ) _A = BlipTextConfig(**a__ ) _A = BlipVisionConfig(**a__ ) _A = self.vision_config.hidden_size _A = projection_dim _A = logit_scale_init_value _A = 1.0 _A = 0.0_2 _A = image_text_hidden_size @classmethod def a_ ( cls : Tuple , a__ : BlipTextConfig , a__ : BlipVisionConfig , **a__ : Optional[int] ) -> str: '''simple docstring''' return cls(text_config=text_config.to_dict() , vision_config=vision_config.to_dict() , **a__ ) def a_ ( self : Union[str, Any] ) -> List[Any]: '''simple docstring''' _A = copy.deepcopy(self.__dict__ ) _A = self.text_config.to_dict() _A = self.vision_config.to_dict() _A = self.__class__.model_type return output
621
0
"""simple docstring""" import argparse from pathlib import Path import fairseq import torch from fairseq.models.xmod import XMODModel as FairseqXmodModel from packaging import version from transformers import XmodConfig, XmodForMaskedLM, XmodForSequenceClassification from transformers.utils import logging if version.parse(fairseq.__version__) < version.parse("0.12.2"): raise Exception("requires fairseq >= 0.12.2") if version.parse(fairseq.__version__) > version.parse("2"): raise Exception("requires fairseq < v2") logging.set_verbosity_info() a_ = logging.get_logger(__name__) a_ = "Hello, World!" a_ = "en_XX" def a__ ( __lowercase , __lowercase , __lowercase ) -> Optional[Any]: _A = Path("data_bin" ) _A = FairseqXmodModel.from_pretrained( model_name_or_path=str(Path(__lowercase ).parent ) , checkpoint_file=Path(__lowercase ).name , _name="xmod_base" , arch="xmod_base" , task="multilingual_masked_lm" , data_name_or_path=str(__lowercase ) , bpe="sentencepiece" , sentencepiece_model=str(Path(__lowercase ).parent / "sentencepiece.bpe.model" ) , src_dict=str(data_dir / "dict.txt" ) , ) xmod.eval() # disable dropout print(__lowercase ) _A = xmod.model.encoder.sentence_encoder _A = XmodConfig( vocab_size=xmod_sent_encoder.embed_tokens.num_embeddings , hidden_size=xmod.cfg.model.encoder_embed_dim , num_hidden_layers=xmod.cfg.model.encoder_layers , num_attention_heads=xmod.cfg.model.encoder_attention_heads , intermediate_size=xmod.cfg.model.encoder_ffn_embed_dim , max_position_embeddings=514 , type_vocab_size=1 , layer_norm_eps=1E-5 , pre_norm=xmod.cfg.model.encoder_normalize_before , adapter_reduction_factor=getattr(xmod.cfg.model , "bottleneck" , 2 ) , adapter_layer_norm=xmod.cfg.model.adapter_layer_norm , adapter_reuse_layer_norm=xmod.cfg.model.adapter_reuse_layer_norm , ln_before_adapter=xmod.cfg.model.ln_before_adapter , languages=xmod.cfg.model.languages , ) if classification_head: _A = xmod.model.classification_heads["mnli"].out_proj.weight.shape[0] print("Our X-MOD config:" , __lowercase ) _A = XmodForSequenceClassification(__lowercase ) if classification_head else XmodForMaskedLM(__lowercase ) model.eval() # Now let's copy all the weights. # Embeddings _A = xmod_sent_encoder.embed_tokens.weight _A = xmod_sent_encoder.embed_positions.weight _A = torch.zeros_like( model.roberta.embeddings.token_type_embeddings.weight ) # just zero them out b/c xmod doesn't use them. _A = xmod_sent_encoder.layernorm_embedding.weight _A = xmod_sent_encoder.layernorm_embedding.bias for i in range(config.num_hidden_layers ): # Encoder: start of layer _A = model.roberta.encoder.layer[i] _A = xmod_sent_encoder.layers[i] # self attention _A = layer.attention.self if not ( xmod_layer.self_attn.k_proj.weight.data.shape == xmod_layer.self_attn.q_proj.weight.data.shape == xmod_layer.self_attn.v_proj.weight.data.shape == torch.Size((config.hidden_size, config.hidden_size) ) ): raise AssertionError("Dimensions of self-attention weights do not match." ) _A = xmod_layer.self_attn.q_proj.weight _A = xmod_layer.self_attn.q_proj.bias _A = xmod_layer.self_attn.k_proj.weight _A = xmod_layer.self_attn.k_proj.bias _A = xmod_layer.self_attn.v_proj.weight _A = xmod_layer.self_attn.v_proj.bias # self-attention output _A = layer.attention.output if self_output.dense.weight.shape != xmod_layer.self_attn.out_proj.weight.shape: raise AssertionError("Dimensions of self-attention output weights do not match." ) _A = xmod_layer.self_attn.out_proj.weight _A = xmod_layer.self_attn.out_proj.bias _A = xmod_layer.self_attn_layer_norm.weight _A = xmod_layer.self_attn_layer_norm.bias # intermediate _A = layer.intermediate if intermediate.dense.weight.shape != xmod_layer.fca.weight.shape: raise AssertionError("Dimensions of intermediate weights do not match." ) _A = xmod_layer.fca.weight _A = xmod_layer.fca.bias # output _A = layer.output if bert_output.dense.weight.shape != xmod_layer.fca.weight.shape: raise AssertionError("Dimensions of feed-forward weights do not match." ) _A = xmod_layer.fca.weight _A = xmod_layer.fca.bias _A = xmod_layer.final_layer_norm.weight _A = xmod_layer.final_layer_norm.bias if bert_output.adapter_layer_norm is not None: _A = xmod_layer.adapter_layer_norm.weight _A = xmod_layer.adapter_layer_norm.bias if sorted(bert_output.adapter_modules.keys() ) != sorted(xmod_layer.adapter_modules.keys() ): raise AssertionError("Lists of language adapters do not match." ) for lang_code, adapter in xmod_layer.adapter_modules.items(): _A = bert_output.adapter_modules[lang_code] _A = xmod_layer.adapter_modules[lang_code] _A = from_adapter.fca.weight _A = from_adapter.fca.bias _A = from_adapter.fca.weight _A = from_adapter.fca.bias # end of layer if xmod_sent_encoder.layer_norm is not None: _A = xmod_sent_encoder.layer_norm.weight _A = xmod_sent_encoder.layer_norm.bias if classification_head: _A = xmod.model.classification_heads["mnli"].dense.weight _A = xmod.model.classification_heads["mnli"].dense.bias _A = xmod.model.classification_heads["mnli"].out_proj.weight _A = xmod.model.classification_heads["mnli"].out_proj.bias else: # LM Head _A = xmod.model.encoder.lm_head.dense.weight _A = xmod.model.encoder.lm_head.dense.bias _A = xmod.model.encoder.lm_head.layer_norm.weight _A = xmod.model.encoder.lm_head.layer_norm.bias _A = xmod.model.encoder.lm_head.weight _A = xmod.model.encoder.lm_head.bias # Let's check that we get the same results. _A = xmod.encode(__lowercase ).unsqueeze(0 ) # batch of size 1 model.roberta.set_default_language(__lowercase ) _A = model(__lowercase )[0] if classification_head: _A = xmod.model.classification_heads["mnli"](xmod.extract_features(__lowercase ) ) else: _A = xmod.model(__lowercase , lang_id=[SAMPLE_LANGUAGE] )[0] print(our_output.shape , their_output.shape ) _A = torch.max(torch.abs(our_output - their_output ) ).item() print(f"""max_absolute_diff = {max_absolute_diff}""" ) # ~ 1e-7 _A = torch.allclose(__lowercase , __lowercase , atol=1E-3 ) print("Do both models output the same tensors?" , "🔥" if success else "💩" ) if not success: raise Exception("Something went wRoNg" ) Path(__lowercase ).mkdir(parents=__lowercase , exist_ok=__lowercase ) print(f"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(__lowercase ) if __name__ == "__main__": a_ = argparse.ArgumentParser() # Required parameters parser.add_argument( "--xmod_checkpoint_path", default=None, type=str, required=True, help="Path the official PyTorch dump." ) parser.add_argument( "--pytorch_dump_folder_path", default=None, type=str, required=True, help="Path to the output PyTorch model." ) parser.add_argument( "--classification_head", action="store_true", help="Whether to convert a final classification head." ) a_ = parser.parse_args() convert_xmod_checkpoint_to_pytorch( args.xmod_checkpoint_path, args.pytorch_dump_folder_path, args.classification_head )
710
"""simple docstring""" import unittest from transformers import load_tool from .test_tools_common import ToolTesterMixin class snake_case ( unittest.TestCase , _UpperCamelCase): def a_ ( self : Optional[Any] ) -> List[str]: '''simple docstring''' _A = load_tool("text-classification" ) self.tool.setup() _A = load_tool("text-classification" , remote=a__ ) def a_ ( self : Optional[int] ) -> Dict: '''simple docstring''' _A = self.tool("That's quite cool" , ["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Optional[Any] ) -> Dict: '''simple docstring''' _A = self.remote_tool("That's quite cool" , ["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Dict ) -> Optional[int]: '''simple docstring''' _A = self.tool(text="That's quite cool" , labels=["positive", "negative"] ) self.assertEqual(a__ , "positive" ) def a_ ( self : Dict ) -> Any: '''simple docstring''' _A = self.remote_tool(text="That's quite cool" , labels=["positive", "negative"] ) self.assertEqual(a__ , "positive" )
621
0
"""simple docstring""" import math from typing import Dict, Iterable, List, Optional, Tuple, Union import numpy as np from ...image_processing_utils import BaseImageProcessor, BatchFeature, get_size_dict from ...image_transforms import normalize, rescale, resize, to_channel_dimension_format from ...image_utils import ( IMAGENET_STANDARD_MEAN, IMAGENET_STANDARD_STD, ChannelDimension, ImageInput, PILImageResampling, get_image_size, is_torch_available, is_torch_tensor, make_list_of_images, to_numpy_array, valid_images, ) from ...utils import TensorType, is_vision_available, logging if is_torch_available(): import torch if is_vision_available(): import PIL a_ = logging.get_logger(__name__) def a__ ( __lowercase , __lowercase , __lowercase , __lowercase ) -> Tuple[int, int]: def constraint_to_multiple_of(__lowercase , __lowercase , __lowercase=0 , __lowercase=None ): _A = round(val / multiple ) * multiple if max_val is not None and x > max_val: _A = math.floor(val / multiple ) * multiple if x < min_val: _A = math.ceil(val / multiple ) * multiple return x _A = (output_size, output_size) if isinstance(__lowercase , __lowercase ) else output_size _A , _A = get_image_size(__lowercase ) _A , _A = output_size # determine new height and width _A = output_height / input_height _A = output_width / input_width if keep_aspect_ratio: # scale as little as possible if abs(1 - scale_width ) < abs(1 - scale_height ): # fit width _A = scale_width else: # fit height _A = scale_height _A = constraint_to_multiple_of(scale_height * input_height , multiple=__lowercase ) _A = constraint_to_multiple_of(scale_width * input_width , multiple=__lowercase ) return (new_height, new_width) class snake_case ( _UpperCamelCase): __UpperCamelCase = ['pixel_values'] def __init__( self : Any , a__ : bool = True , a__ : Dict[str, int] = None , a__ : PILImageResampling = PILImageResampling.BILINEAR , a__ : bool = False , a__ : int = 1 , a__ : bool = True , a__ : Union[int, float] = 1 / 2_55 , a__ : bool = True , a__ : Optional[Union[float, List[float]]] = None , a__ : Optional[Union[float, List[float]]] = None , **a__ : Optional[Any] , ) -> None: '''simple docstring''' super().__init__(**a__ ) _A = size if size is not None else {"height": 3_84, "width": 3_84} _A = get_size_dict(a__ ) _A = do_resize _A = size _A = keep_aspect_ratio _A = ensure_multiple_of _A = resample _A = do_rescale _A = rescale_factor _A = do_normalize _A = image_mean if image_mean is not None else IMAGENET_STANDARD_MEAN _A = image_std if image_std is not None else IMAGENET_STANDARD_STD def a_ ( self : int , a__ : np.ndarray , a__ : Dict[str, int] , a__ : bool = False , a__ : int = 1 , a__ : PILImageResampling = PILImageResampling.BICUBIC , a__ : Optional[Union[str, ChannelDimension]] = None , **a__ : List[str] , ) -> np.ndarray: '''simple docstring''' _A = get_size_dict(a__ ) if "height" not in size or "width" not in size: raise ValueError(F"""The size dictionary must contain the keys 'height' and 'width'. Got {size.keys()}""" ) _A = get_resize_output_image_size( a__ , output_size=(size["height"], size["width"]) , keep_aspect_ratio=a__ , multiple=a__ , ) return resize(a__ , size=a__ , resample=a__ , data_format=a__ , **a__ ) def a_ ( self : int , a__ : np.ndarray , a__ : Union[int, float] , a__ : Optional[Union[str, ChannelDimension]] = None , **a__ : int , ) -> Any: '''simple docstring''' return rescale(a__ , scale=a__ , data_format=a__ , **a__ ) def a_ ( self : List[Any] , a__ : np.ndarray , a__ : Union[float, List[float]] , a__ : Union[float, List[float]] , a__ : Optional[Union[str, ChannelDimension]] = None , **a__ : Dict , ) -> np.ndarray: '''simple docstring''' return normalize(a__ , mean=a__ , std=a__ , data_format=a__ , **a__ ) def a_ ( self : Dict , a__ : ImageInput , a__ : bool = None , a__ : int = None , a__ : bool = None , a__ : int = None , a__ : PILImageResampling = None , a__ : bool = None , a__ : float = None , a__ : bool = None , a__ : Optional[Union[float, List[float]]] = None , a__ : Optional[Union[float, List[float]]] = None , a__ : Optional[Union[str, TensorType]] = None , a__ : ChannelDimension = ChannelDimension.FIRST , **a__ : List[str] , ) -> PIL.Image.Image: '''simple docstring''' _A = do_resize if do_resize is not None else self.do_resize _A = size if size is not None else self.size _A = get_size_dict(a__ ) _A = keep_aspect_ratio if keep_aspect_ratio is not None else self.keep_aspect_ratio _A = ensure_multiple_of if ensure_multiple_of is not None else self.ensure_multiple_of _A = resample if resample is not None else self.resample _A = do_rescale if do_rescale is not None else self.do_rescale _A = rescale_factor if rescale_factor is not None else self.rescale_factor _A = do_normalize if do_normalize is not None else self.do_normalize _A = image_mean if image_mean is not None else self.image_mean _A = image_std if image_std is not None else self.image_std _A = make_list_of_images(a__ ) if not valid_images(a__ ): raise ValueError( "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, " "torch.Tensor, tf.Tensor or jax.ndarray." ) if do_resize and size is None or resample is None: raise ValueError("Size and resample must be specified if do_resize is True." ) if do_rescale and rescale_factor is None: raise ValueError("Rescale factor must be specified if do_rescale is True." ) if do_normalize and (image_mean is None or image_std is None): raise ValueError("Image mean and std must be specified if do_normalize is True." ) # All transformations expect numpy arrays. _A = [to_numpy_array(a__ ) for image in images] if do_resize: _A = [self.resize(image=a__ , size=a__ , resample=a__ ) for image in images] if do_rescale: _A = [self.rescale(image=a__ , scale=a__ ) for image in images] if do_normalize: _A = [self.normalize(image=a__ , mean=a__ , std=a__ ) for image in images] _A = [to_channel_dimension_format(a__ , a__ ) for image in images] _A = {"pixel_values": images} return BatchFeature(data=a__ , tensor_type=a__ ) def a_ ( self : int , a__ : str , a__ : List[Tuple] = None ) -> str: '''simple docstring''' _A = outputs.logits # Resize logits and compute semantic segmentation maps if target_sizes is not None: if len(a__ ) != len(a__ ): raise ValueError( "Make sure that you pass in as many target sizes as the batch dimension of the logits" ) if is_torch_tensor(a__ ): _A = target_sizes.numpy() _A = [] for idx in range(len(a__ ) ): _A = torch.nn.functional.interpolate( logits[idx].unsqueeze(dim=0 ) , size=target_sizes[idx] , mode="bilinear" , align_corners=a__ ) _A = resized_logits[0].argmax(dim=0 ) semantic_segmentation.append(a__ ) else: _A = logits.argmax(dim=1 ) _A = [semantic_segmentation[i] for i in range(semantic_segmentation.shape[0] )] return semantic_segmentation
711
"""simple docstring""" import gc import random import unittest import numpy as np import torch from PIL import Image from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer from diffusers import AutoencoderKL, PNDMScheduler, StableDiffusionInpaintPipeline, UNetaDConditionModel from diffusers.utils import floats_tensor, load_image, load_numpy, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu, slow from ..pipeline_params import TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS, TEXT_GUIDED_IMAGE_INPAINTING_PARAMS from ..test_pipelines_common import PipelineKarrasSchedulerTesterMixin, PipelineLatentTesterMixin, PipelineTesterMixin enable_full_determinism() class snake_case ( _UpperCamelCase , _UpperCamelCase , _UpperCamelCase , unittest.TestCase): __UpperCamelCase = StableDiffusionInpaintPipeline __UpperCamelCase = TEXT_GUIDED_IMAGE_INPAINTING_PARAMS __UpperCamelCase = TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS __UpperCamelCase = frozenset( []) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess __UpperCamelCase = frozenset([]) def a_ ( self : Dict ) -> Optional[int]: '''simple docstring''' torch.manual_seed(0 ) _A = UNetaDConditionModel( block_out_channels=(32, 64) , layers_per_block=2 , sample_size=32 , in_channels=9 , out_channels=4 , down_block_types=("DownBlock2D", "CrossAttnDownBlock2D") , up_block_types=("CrossAttnUpBlock2D", "UpBlock2D") , cross_attention_dim=32 , attention_head_dim=(2, 4) , use_linear_projection=a__ , ) _A = PNDMScheduler(skip_prk_steps=a__ ) torch.manual_seed(0 ) _A = AutoencoderKL( block_out_channels=[32, 64] , in_channels=3 , out_channels=3 , down_block_types=["DownEncoderBlock2D", "DownEncoderBlock2D"] , up_block_types=["UpDecoderBlock2D", "UpDecoderBlock2D"] , latent_channels=4 , sample_size=1_28 , ) torch.manual_seed(0 ) _A = CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=32 , intermediate_size=37 , layer_norm_eps=1E-0_5 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=10_00 , hidden_act="gelu" , projection_dim=5_12 , ) _A = CLIPTextModel(a__ ) _A = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip" ) _A = { "unet": unet, "scheduler": scheduler, "vae": vae, "text_encoder": text_encoder, "tokenizer": tokenizer, "safety_checker": None, "feature_extractor": None, } return components def a_ ( self : Optional[Any] , a__ : List[str] , a__ : Tuple=0 ) -> int: '''simple docstring''' _A = floats_tensor((1, 3, 32, 32) , rng=random.Random(a__ ) ).to(a__ ) _A = image.cpu().permute(0 , 2 , 3 , 1 )[0] _A = Image.fromarray(np.uinta(a__ ) ).convert("RGB" ).resize((64, 64) ) _A = Image.fromarray(np.uinta(image + 4 ) ).convert("RGB" ).resize((64, 64) ) if str(a__ ).startswith("mps" ): _A = torch.manual_seed(a__ ) else: _A = torch.Generator(device=a__ ).manual_seed(a__ ) _A = { "prompt": "A painting of a squirrel eating a burger", "image": init_image, "mask_image": mask_image, "generator": generator, "num_inference_steps": 2, "guidance_scale": 6.0, "output_type": "numpy", } return inputs def a_ ( self : Optional[Any] ) -> Optional[Any]: '''simple docstring''' _A = "cpu" # ensure determinism for the device-dependent torch.Generator _A = self.get_dummy_components() _A = StableDiffusionInpaintPipeline(**a__ ) _A = sd_pipe.to(a__ ) sd_pipe.set_progress_bar_config(disable=a__ ) _A = self.get_dummy_inputs(a__ ) _A = sd_pipe(**a__ ).images _A = image[0, -3:, -3:, -1] assert image.shape == (1, 64, 64, 3) _A = np.array([0.4_7_2_7, 0.5_7_3_5, 0.3_9_4_1, 0.5_4_4_6, 0.5_9_2_6, 0.4_3_9_4, 0.5_0_6_2, 0.4_6_5_4, 0.4_4_7_6] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-2 def a_ ( self : str ) -> Union[str, Any]: '''simple docstring''' super().test_inference_batch_single_identical(expected_max_diff=3E-3 ) @slow @require_torch_gpu class snake_case ( unittest.TestCase): def a_ ( self : List[Any] ) -> Any: '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def a_ ( self : Union[str, Any] ) -> Tuple: '''simple docstring''' _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = load_numpy( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint" "/yellow_cat_sitting_on_a_park_bench.npy" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = StableDiffusionInpaintPipeline.from_pretrained(a__ , safety_checker=a__ ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , output_type="np" , ) _A = output.images[0] assert image.shape == (5_12, 5_12, 3) assert np.abs(expected_image - image ).max() < 9E-3 def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = load_numpy( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint" "/yellow_cat_sitting_on_a_park_bench_fp16.npy" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = StableDiffusionInpaintPipeline.from_pretrained( a__ , torch_dtype=torch.floataa , safety_checker=a__ , ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , output_type="np" , ) _A = output.images[0] assert image.shape == (5_12, 5_12, 3) assert np.abs(expected_image - image ).max() < 5E-1 def a_ ( self : List[Any] ) -> Union[str, Any]: '''simple docstring''' torch.cuda.empty_cache() torch.cuda.reset_max_memory_allocated() torch.cuda.reset_peak_memory_stats() _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/sd2-inpaint/init_image.png" ) _A = load_image( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd2-inpaint/mask.png" ) _A = "stabilityai/stable-diffusion-2-inpainting" _A = PNDMScheduler.from_pretrained(a__ , subfolder="scheduler" ) _A = StableDiffusionInpaintPipeline.from_pretrained( a__ , safety_checker=a__ , scheduler=a__ , torch_dtype=torch.floataa , ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) pipe.enable_attention_slicing(1 ) pipe.enable_sequential_cpu_offload() _A = "Face of a yellow cat, high resolution, sitting on a park bench" _A = torch.manual_seed(0 ) _A = pipe( prompt=a__ , image=a__ , mask_image=a__ , generator=a__ , num_inference_steps=2 , output_type="np" , ) _A = torch.cuda.max_memory_allocated() # make sure that less than 2.65 GB is allocated assert mem_bytes < 2.6_5 * 10**9
621
0
"""simple docstring""" import gc import unittest import numpy as np import torch from diffusers import AutoencoderKL, DDIMScheduler, DiTPipeline, DPMSolverMultistepScheduler, TransformeraDModel from diffusers.utils import is_xformers_available, load_numpy, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..pipeline_params import ( CLASS_CONDITIONED_IMAGE_GENERATION_BATCH_PARAMS, CLASS_CONDITIONED_IMAGE_GENERATION_PARAMS, ) from ..test_pipelines_common import PipelineTesterMixin enable_full_determinism() class snake_case ( _UpperCamelCase , unittest.TestCase): __UpperCamelCase = DiTPipeline __UpperCamelCase = CLASS_CONDITIONED_IMAGE_GENERATION_PARAMS __UpperCamelCase = PipelineTesterMixin.required_optional_params - { 'latents', 'num_images_per_prompt', 'callback', 'callback_steps', } __UpperCamelCase = CLASS_CONDITIONED_IMAGE_GENERATION_BATCH_PARAMS __UpperCamelCase = False def a_ ( self : Tuple ) -> Union[str, Any]: '''simple docstring''' torch.manual_seed(0 ) _A = TransformeraDModel( sample_size=16 , num_layers=2 , patch_size=4 , attention_head_dim=8 , num_attention_heads=2 , in_channels=4 , out_channels=8 , attention_bias=a__ , activation_fn="gelu-approximate" , num_embeds_ada_norm=10_00 , norm_type="ada_norm_zero" , norm_elementwise_affine=a__ , ) _A = AutoencoderKL() _A = DDIMScheduler() _A = {"transformer": transformer.eval(), "vae": vae.eval(), "scheduler": scheduler} return components def a_ ( self : Optional[Any] , a__ : Any , a__ : Dict=0 ) -> Optional[Any]: '''simple docstring''' if str(a__ ).startswith("mps" ): _A = torch.manual_seed(a__ ) else: _A = torch.Generator(device=a__ ).manual_seed(a__ ) _A = { "class_labels": [1], "generator": generator, "num_inference_steps": 2, "output_type": "numpy", } return inputs def a_ ( self : List[Any] ) -> List[Any]: '''simple docstring''' _A = "cpu" _A = self.get_dummy_components() _A = self.pipeline_class(**a__ ) pipe.to(a__ ) pipe.set_progress_bar_config(disable=a__ ) _A = self.get_dummy_inputs(a__ ) _A = pipe(**a__ ).images _A = image[0, -3:, -3:, -1] self.assertEqual(image.shape , (1, 16, 16, 3) ) _A = np.array([0.2_9_4_6, 0.6_6_0_1, 0.4_3_2_9, 0.3_2_9_6, 0.4_1_4_4, 0.5_3_1_9, 0.7_2_7_3, 0.5_0_1_3, 0.4_4_5_7] ) _A = np.abs(image_slice.flatten() - expected_slice ).max() self.assertLessEqual(a__ , 1E-3 ) def a_ ( self : str ) -> int: '''simple docstring''' self._test_inference_batch_single_identical(relax_max_difference=a__ , expected_max_diff=1E-3 ) @unittest.skipIf( torch_device != "cuda" or not is_xformers_available() , reason="XFormers attention is only available with CUDA and `xformers` installed" , ) def a_ ( self : str ) -> Optional[int]: '''simple docstring''' self._test_xformers_attention_forwardGenerator_pass(expected_max_diff=1E-3 ) @require_torch_gpu @slow class snake_case ( unittest.TestCase): def a_ ( self : Optional[Any] ) -> Any: '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def a_ ( self : Any ) -> Any: '''simple docstring''' _A = torch.manual_seed(0 ) _A = DiTPipeline.from_pretrained("facebook/DiT-XL-2-256" ) pipe.to("cuda" ) _A = ["vase", "umbrella", "white shark", "white wolf"] _A = pipe.get_label_ids(a__ ) _A = pipe(a__ , generator=a__ , num_inference_steps=40 , output_type="np" ).images for word, image in zip(a__ , a__ ): _A = load_numpy( F"""https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/dit/{word}.npy""" ) assert np.abs((expected_image - image).max() ) < 1E-2 def a_ ( self : str ) -> str: '''simple docstring''' _A = DiTPipeline.from_pretrained("facebook/DiT-XL-2-512" ) _A = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config ) pipe.to("cuda" ) _A = ["vase", "umbrella"] _A = pipe.get_label_ids(a__ ) _A = torch.manual_seed(0 ) _A = pipe(a__ , generator=a__ , num_inference_steps=25 , output_type="np" ).images for word, image in zip(a__ , a__ ): _A = load_numpy( "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" F"""/dit/{word}_512.npy""" ) assert np.abs((expected_image - image).max() ) < 1E-1
712
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> int: while a != 0: _A , _A = b % a, a return b def a__ ( __lowercase , __lowercase ) -> int: if gcd(__lowercase , __lowercase ) != 1: _A = f"""mod inverse of {a!r} and {m!r} does not exist""" raise ValueError(__lowercase ) _A , _A , _A = 1, 0, a _A , _A , _A = 0, 1, m while va != 0: _A = ua // va _A , _A , _A , _A , _A , _A = (ua - q * va), (ua - q * va), (ua - q * va), va, va, va return ua % m
621
0
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> bool: return numa ^ numa < 0 if __name__ == "__main__": import doctest doctest.testmod()
713
"""simple docstring""" # 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 snake_case ( _UpperCamelCase): def __init__( self : List[Any] , a__ : Any ) -> Any: '''simple docstring''' _A = data def __iter__( self : List[str] ) -> str: '''simple docstring''' for element in self.data: yield element def a__ ( __lowercase=True ) -> Tuple: _A = Accelerator(even_batches=__lowercase ) assert accelerator.num_processes == 2, "this script expects that two GPUs are available" return accelerator def a__ ( __lowercase , __lowercase , __lowercase , __lowercase = False ) -> Union[str, Any]: if iterable: _A = DummyIterableDataset(torch.as_tensor(range(__lowercase ) ) ) else: _A = TensorDataset(torch.as_tensor(range(__lowercase ) ) ) _A = DataLoader(__lowercase , batch_size=__lowercase ) _A = accelerator.prepare(__lowercase ) return dl def a__ ( __lowercase , __lowercase , __lowercase , __lowercase , __lowercase , ) -> Dict: _A = create_dataloader(accelerator=__lowercase , dataset_size=__lowercase , batch_size=__lowercase ) _A = [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 a__ ( ) -> List[str]: _A = create_accelerator() # without padding, we would expect a different number of batches verify_dataloader_batch_sizes( __lowercase , 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( __lowercase , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 2] , ) def a__ ( ) -> List[Any]: _A = create_accelerator(even_batches=__lowercase ) verify_dataloader_batch_sizes( __lowercase , dataset_size=3 , batch_size=1 , process_0_expected_batch_sizes=[1, 1] , process_1_expected_batch_sizes=[1] , ) verify_dataloader_batch_sizes( __lowercase , dataset_size=7 , batch_size=2 , process_0_expected_batch_sizes=[2, 2] , process_1_expected_batch_sizes=[2, 1] , ) def a__ ( ) -> int: _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) _A = [] with accelerator.join_uneven_inputs([ddp_model] ): for batch_idx, batch in enumerate(__lowercase ): _A = ddp_model(batch[0].float() ) _A = output.sum() loss.backward() batch_idxs.append(__lowercase ) accelerator.wait_for_everyone() if accelerator.process_index == 0: assert batch_idxs == [0, 1] elif accelerator.process_index == 1: assert batch_idxs == [0] def a__ ( __lowercase ) -> List[str]: with warnings.catch_warnings(record=__lowercase ) as w: with accelerator.join_uneven_inputs([Mock()] ): pass assert issubclass(w[-1].category , __lowercase ) assert "only supported for multi-GPU" in str(w[-1].message ) def a__ ( ) -> Tuple: _A = True _A = False _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): _A = train_dl.batch_sampler.even_batches _A = 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 a__ ( ) -> int: _A = True _A = False _A = create_accelerator(even_batches=__lowercase ) _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) create_dataloader(__lowercase , dataset_size=3 , batch_size=1 , iterable=__lowercase ) _A = create_dataloader(__lowercase , dataset_size=3 , batch_size=1 ) with warnings.catch_warnings(): warnings.filterwarnings("ignore" ) try: with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): _A = 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 a__ ( ) -> Optional[Any]: _A = create_accelerator() _A = torch.nn.Linear(1 , 1 ) _A = accelerator.prepare(__lowercase ) create_dataloader(__lowercase , dataset_size=3 , batch_size=1 , iterable=__lowercase ) with warnings.catch_warnings(record=__lowercase ) as w: with accelerator.join_uneven_inputs([ddp_model] , even_batches=__lowercase ): pass assert issubclass(w[-1].category , __lowercase ) assert "only supported for map-style datasets" in str(w[-1].message ) def a__ ( ) -> Optional[Any]: _A = 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" ) _A = accelerator.state.distributed_type _A = DistributedType.FSDP test_join_raises_warning_for_non_ddp_distributed(__lowercase ) _A = original_state if __name__ == "__main__": main()
621
0
"""simple docstring""" 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_ = { "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_ = ["BertTokenizerFast"] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "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_ = [ "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_ = ["TFBertTokenizer"] try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "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_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
714
"""simple docstring""" class snake_case : def __init__( self : Optional[int] , a__ : List[Any] , a__ : List[str] , a__ : Tuple ) -> Optional[Any]: '''simple docstring''' _A = None _A = None _A = graph self._normalize_graph(a__ , a__ ) _A = len(a__ ) _A = None def a_ ( self : str , a__ : List[str] , a__ : List[Any] ) -> Dict: '''simple docstring''' if sources is int: _A = [sources] if sinks is int: _A = [sinks] if len(a__ ) == 0 or len(a__ ) == 0: return _A = sources[0] _A = sinks[0] # make fake vertex if there are more # than one source or sink if len(a__ ) > 1 or len(a__ ) > 1: _A = 0 for i in sources: max_input_flow += sum(self.graph[i] ) _A = len(self.graph ) + 1 for room in self.graph: room.insert(0 , 0 ) self.graph.insert(0 , [0] * size ) for i in sources: _A = max_input_flow _A = 0 _A = len(self.graph ) + 1 for room in self.graph: room.append(0 ) self.graph.append([0] * size ) for i in sinks: _A = max_input_flow _A = size - 1 def a_ ( self : Optional[int] ) -> List[Any]: '''simple docstring''' if self.maximum_flow_algorithm is None: raise Exception("You need to set maximum flow algorithm before." ) if self.source_index is None or self.sink_index is None: return 0 self.maximum_flow_algorithm.execute() return self.maximum_flow_algorithm.getMaximumFlow() def a_ ( self : List[Any] , a__ : Optional[Any] ) -> str: '''simple docstring''' _A = algorithm(self ) class snake_case : def __init__( self : List[str] , a__ : List[str] ) -> Union[str, Any]: '''simple docstring''' _A = flow_network _A = flow_network.verticesCount _A = flow_network.sourceIndex _A = flow_network.sinkIndex # it's just a reference, so you shouldn't change # it in your algorithms, use deep copy before doing that _A = flow_network.graph _A = False def a_ ( self : Optional[Any] ) -> List[Any]: '''simple docstring''' if not self.executed: self._algorithm() _A = True def a_ ( self : Any ) -> int: '''simple docstring''' pass class snake_case ( _UpperCamelCase): def __init__( self : Optional[Any] , a__ : Dict ) -> List[str]: '''simple docstring''' super().__init__(a__ ) # use this to save your result _A = -1 def a_ ( self : Any ) -> List[str]: '''simple docstring''' if not self.executed: raise Exception("You should execute algorithm before using its result!" ) return self.maximum_flow class snake_case ( _UpperCamelCase): def __init__( self : Union[str, Any] , a__ : Union[str, Any] ) -> Dict: '''simple docstring''' super().__init__(a__ ) _A = [[0] * self.verticies_count for i in range(self.verticies_count )] _A = [0] * self.verticies_count _A = [0] * self.verticies_count def a_ ( self : Any ) -> Dict: '''simple docstring''' _A = self.verticies_count # push some substance to graph for nextvertex_index, bandwidth in enumerate(self.graph[self.source_index] ): self.preflow[self.source_index][nextvertex_index] += bandwidth self.preflow[nextvertex_index][self.source_index] -= bandwidth self.excesses[nextvertex_index] += bandwidth # Relabel-to-front selection rule _A = [ i for i in range(self.verticies_count ) if i != self.source_index and i != self.sink_index ] # move through list _A = 0 while i < len(a__ ): _A = vertices_list[i] _A = self.heights[vertex_index] self.process_vertex(a__ ) if self.heights[vertex_index] > previous_height: # if it was relabeled, swap elements # and start from 0 index vertices_list.insert(0 , vertices_list.pop(a__ ) ) _A = 0 else: i += 1 _A = sum(self.preflow[self.source_index] ) def a_ ( self : Dict , a__ : Any ) -> Optional[int]: '''simple docstring''' while self.excesses[vertex_index] > 0: for neighbour_index in range(self.verticies_count ): # if it's neighbour and current vertex is higher if ( self.graph[vertex_index][neighbour_index] - self.preflow[vertex_index][neighbour_index] > 0 and self.heights[vertex_index] > self.heights[neighbour_index] ): self.push(a__ , a__ ) self.relabel(a__ ) def a_ ( self : str , a__ : Optional[int] , a__ : List[Any] ) -> Optional[int]: '''simple docstring''' _A = min( self.excesses[from_index] , self.graph[from_index][to_index] - self.preflow[from_index][to_index] , ) self.preflow[from_index][to_index] += preflow_delta self.preflow[to_index][from_index] -= preflow_delta self.excesses[from_index] -= preflow_delta self.excesses[to_index] += preflow_delta def a_ ( self : Any , a__ : Dict ) -> Any: '''simple docstring''' _A = None for to_index in range(self.verticies_count ): if ( self.graph[vertex_index][to_index] - self.preflow[vertex_index][to_index] > 0 ) and (min_height is None or self.heights[to_index] < min_height): _A = self.heights[to_index] if min_height is not None: _A = min_height + 1 if __name__ == "__main__": a_ = [0] a_ = [3] # graph = [ # [0, 0, 4, 6, 0, 0], # [0, 0, 5, 2, 0, 0], # [0, 0, 0, 0, 4, 4], # [0, 0, 0, 0, 6, 6], # [0, 0, 0, 0, 0, 0], # [0, 0, 0, 0, 0, 0], # ] a_ = [[0, 7, 0, 0], [0, 0, 6, 0], [0, 0, 0, 8], [9, 0, 0, 0]] # prepare our network a_ = FlowNetwork(graph, entrances, exits) # set algorithm flow_network.set_maximum_flow_algorithm(PushRelabelExecutor) # and calculate a_ = flow_network.find_maximum_flow() print(f'''maximum flow is {maximum_flow}''')
621
0
"""simple docstring""" import json from typing import List, Optional, Tuple from tokenizers import pre_tokenizers, processors from ...tokenization_utils_base import AddedToken, BatchEncoding from ...tokenization_utils_fast import PreTrainedTokenizerFast from ...utils import logging from .tokenization_mvp import MvpTokenizer a_ = logging.get_logger(__name__) a_ = {"vocab_file": "vocab.json", "merges_file": "merges.txt", "tokenizer_file": "tokenizer.json"} # See all MVP models at https://huggingface.co/models?filter=mvp a_ = { "vocab_file": { "RUCAIBox/mvp": "https://huggingface.co/RUCAIBox/mvp/resolve/main/vocab.json", }, "added_tokens.json": { "RUCAIBox/mvp": "https://huggingface.co/RUCAIBox/mvp/resolve/main/added_tokens.json", }, "merges_file": { "RUCAIBox/mvp": "https://huggingface.co/RUCAIBox/mvp/resolve/main/merges.txt", }, "tokenizer_file": { "RUCAIBox/mvp": "https://huggingface.co/RUCAIBox/mvp/resolve/main/tokenizer.json", }, } a_ = { "RUCAIBox/mvp": 10_24, } class snake_case ( _UpperCamelCase): __UpperCamelCase = VOCAB_FILES_NAMES __UpperCamelCase = PRETRAINED_VOCAB_FILES_MAP __UpperCamelCase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __UpperCamelCase = ['input_ids', 'attention_mask'] __UpperCamelCase = MvpTokenizer def __init__( self : List[Any] , a__ : Any=None , a__ : Optional[int]=None , a__ : int=None , a__ : Dict="replace" , a__ : List[Any]="<s>" , a__ : Dict="</s>" , a__ : int="</s>" , a__ : int="<s>" , a__ : Optional[int]="<unk>" , a__ : Tuple="<pad>" , a__ : List[Any]="<mask>" , a__ : Optional[Any]=False , a__ : Union[str, Any]=True , **a__ : Union[str, Any] , ) -> Optional[int]: '''simple docstring''' super().__init__( a__ , a__ , tokenizer_file=a__ , errors=a__ , bos_token=a__ , eos_token=a__ , sep_token=a__ , cls_token=a__ , unk_token=a__ , pad_token=a__ , mask_token=a__ , add_prefix_space=a__ , trim_offsets=a__ , **a__ , ) _A = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() ) if pre_tok_state.get("add_prefix_space" , a__ ) != add_prefix_space: _A = getattr(a__ , pre_tok_state.pop("type" ) ) _A = add_prefix_space _A = pre_tok_class(**a__ ) _A = add_prefix_space # the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__` _A = "post_processor" _A = getattr(self.backend_tokenizer , a__ , a__ ) if tokenizer_component_instance: _A = 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: _A = tuple(state["sep"] ) if "cls" in state: _A = tuple(state["cls"] ) _A = False if state.get("add_prefix_space" , a__ ) != add_prefix_space: _A = add_prefix_space _A = True if state.get("trim_offsets" , a__ ) != trim_offsets: _A = trim_offsets _A = True if changes_to_apply: _A = getattr(a__ , state.pop("type" ) ) _A = component_class(**a__ ) setattr(self.backend_tokenizer , a__ , a__ ) @property def a_ ( self : Union[str, Any] ) -> str: '''simple docstring''' 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 a_ ( self : List[Any] , a__ : List[str] ) -> Optional[Any]: '''simple docstring''' _A = AddedToken(a__ , lstrip=a__ , rstrip=a__ ) if isinstance(a__ , a__ ) else value _A = value def a_ ( self : Optional[int] , *a__ : Any , **a__ : Dict ) -> BatchEncoding: '''simple docstring''' _A = kwargs.get("is_split_into_words" , a__ ) 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(*a__ , **a__ ) def a_ ( self : Optional[Any] , *a__ : Tuple , **a__ : List[Any] ) -> BatchEncoding: '''simple docstring''' _A = kwargs.get("is_split_into_words" , a__ ) 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(*a__ , **a__ ) def a_ ( self : Tuple , a__ : str , a__ : Optional[str] = None ) -> Tuple[str]: '''simple docstring''' _A = self._tokenizer.model.save(a__ , name=a__ ) return tuple(a__ ) def a_ ( self : int , a__ : List[Any] , a__ : int=None ) -> Any: '''simple docstring''' _A = [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 a_ ( self : Any , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = [self.sep_token_id] _A = [self.cls_token_id] if token_ids_a is None: return len(cls + token_ids_a + sep ) * [0] return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
715
"""simple docstring""" from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_tf_available, is_tokenizers_available, is_torch_available, ) a_ = { "configuration_roformer": ["ROFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP", "RoFormerConfig", "RoFormerOnnxConfig"], "tokenization_roformer": ["RoFormerTokenizer"], } try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = ["RoFormerTokenizerFast"] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "RoFormerForCausalLM", "RoFormerForMaskedLM", "RoFormerForMultipleChoice", "RoFormerForQuestionAnswering", "RoFormerForSequenceClassification", "RoFormerForTokenClassification", "RoFormerLayer", "RoFormerModel", "RoFormerPreTrainedModel", "load_tf_weights_in_roformer", ] try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "TF_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "TFRoFormerForCausalLM", "TFRoFormerForMaskedLM", "TFRoFormerForMultipleChoice", "TFRoFormerForQuestionAnswering", "TFRoFormerForSequenceClassification", "TFRoFormerForTokenClassification", "TFRoFormerLayer", "TFRoFormerModel", "TFRoFormerPreTrainedModel", ] try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ "FLAX_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", "FlaxRoFormerForMaskedLM", "FlaxRoFormerForMultipleChoice", "FlaxRoFormerForQuestionAnswering", "FlaxRoFormerForSequenceClassification", "FlaxRoFormerForTokenClassification", "FlaxRoFormerModel", "FlaxRoFormerPreTrainedModel", ] if TYPE_CHECKING: from .configuration_roformer import ROFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, RoFormerConfig, RoFormerOnnxConfig from .tokenization_roformer import RoFormerTokenizer try: if not is_tokenizers_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .tokenization_roformer_fast import RoFormerTokenizerFast try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_roformer import ( ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, RoFormerForCausalLM, RoFormerForMaskedLM, RoFormerForMultipleChoice, RoFormerForQuestionAnswering, RoFormerForSequenceClassification, RoFormerForTokenClassification, RoFormerLayer, RoFormerModel, RoFormerPreTrainedModel, load_tf_weights_in_roformer, ) try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_tf_roformer import ( TF_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, TFRoFormerForCausalLM, TFRoFormerForMaskedLM, TFRoFormerForMultipleChoice, TFRoFormerForQuestionAnswering, TFRoFormerForSequenceClassification, TFRoFormerForTokenClassification, TFRoFormerLayer, TFRoFormerModel, TFRoFormerPreTrainedModel, ) try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_flax_roformer import ( FLAX_ROFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, FlaxRoFormerForMaskedLM, FlaxRoFormerForMultipleChoice, FlaxRoFormerForQuestionAnswering, FlaxRoFormerForSequenceClassification, FlaxRoFormerForTokenClassification, FlaxRoFormerModel, FlaxRoFormerPreTrainedModel, ) else: import sys a_ = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
621
0
"""simple docstring""" import argparse import os # New Code # import evaluate import torch from datasets import load_dataset from torch.optim import AdamW from torch.utils.data import DataLoader from transformers import AutoModelForSequenceClassification, AutoTokenizer, get_linear_schedule_with_warmup, set_seed from accelerate import Accelerator, DistributedType from accelerate.utils import find_executable_batch_size ######################################################################## # This is a fully working simple example to use Accelerate, # specifically showcasing how to ensure out-of-memory errors never # interrupt training, and builds off the `nlp_example.py` script. # # This example trains a Bert base model on GLUE MRPC # in any of the following settings (with the same script): # - single CPU or single GPU # - multi GPUS (using PyTorch distributed mode) # - (multi) TPUs # - fp16 (mixed-precision) or fp32 (normal precision) # # New additions from the base script can be found quickly by # looking for the # New Code # tags # # To run it in each of these various modes, follow the instructions # in the readme for examples: # https://github.com/huggingface/accelerate/tree/main/examples # ######################################################################## a_ = 16 a_ = 32 def a__ ( __lowercase , __lowercase = 16 ) -> Tuple: _A = AutoTokenizer.from_pretrained("bert-base-cased" ) _A = load_dataset("glue" , "mrpc" ) def tokenize_function(__lowercase ): # max_length=None => use the model max length (it's actually the default) _A = tokenizer(examples["sentence1"] , examples["sentence2"] , truncation=__lowercase , max_length=__lowercase ) return outputs # Apply the method we just defined to all the examples in all the splits of the dataset # starting with the main process first: with accelerator.main_process_first(): _A = datasets.map( __lowercase , batched=__lowercase , remove_columns=["idx", "sentence1", "sentence2"] , ) # We also rename the 'label' column to 'labels' which is the expected name for labels by the models of the # transformers library _A = tokenized_datasets.rename_column("label" , "labels" ) def collate_fn(__lowercase ): # On TPU it's best to pad everything to the same length or training will be very slow. _A = 128 if accelerator.distributed_type == DistributedType.TPU else None # When using mixed precision we want round multiples of 8/16 if accelerator.mixed_precision == "fp8": _A = 16 elif accelerator.mixed_precision != "no": _A = 8 else: _A = None return tokenizer.pad( __lowercase , padding="longest" , max_length=__lowercase , pad_to_multiple_of=__lowercase , return_tensors="pt" , ) # Instantiate dataloaders. _A = DataLoader( tokenized_datasets["train"] , shuffle=__lowercase , collate_fn=__lowercase , batch_size=__lowercase ) _A = DataLoader( tokenized_datasets["validation"] , shuffle=__lowercase , collate_fn=__lowercase , batch_size=__lowercase ) return train_dataloader, eval_dataloader # For testing only if os.environ.get("TESTING_MOCKED_DATALOADERS", None) == "1": from accelerate.test_utils.training import mocked_dataloaders a_ = mocked_dataloaders # noqa: F811 def a__ ( __lowercase , __lowercase ) -> str: # For testing only if os.environ.get("TESTING_MOCKED_DATALOADERS" , __lowercase ) == "1": _A = 2 # Initialize accelerator _A = Accelerator(cpu=args.cpu , mixed_precision=args.mixed_precision ) # Sample hyper-parameters for learning rate, batch size, seed and a few other HPs _A = config["lr"] _A = int(config["num_epochs"] ) _A = int(config["seed"] ) _A = int(config["batch_size"] ) _A = evaluate.load("glue" , "mrpc" ) # New Code # # We now can define an inner training loop function. It should take a batch size as the only parameter, # and build the dataloaders in there. # It also gets our decorator @find_executable_batch_size(starting_batch_size=__lowercase ) def inner_training_loop(__lowercase ): # And now just move everything below under this function # We need to bring in the Accelerator object from earlier nonlocal accelerator # And reset all of its attributes that could hold onto any memory: accelerator.free_memory() # Then we can declare the model, optimizer, and everything else: set_seed(__lowercase ) # Instantiate the model (we build the model here so that the seed also control new weights initialization) _A = AutoModelForSequenceClassification.from_pretrained("bert-base-cased" , return_dict=__lowercase ) # We could avoid this line since the accelerator is set with `device_placement=True` (default value). # Note that if you are placing tensors on devices manually, this line absolutely needs to be before the optimizer # creation otherwise training will not work on TPU (`accelerate` will kindly throw an error to make us aware of that). _A = model.to(accelerator.device ) # Instantiate optimizer _A = AdamW(params=model.parameters() , lr=__lowercase ) _A , _A = get_dataloaders(__lowercase , __lowercase ) # Instantiate scheduler _A = get_linear_schedule_with_warmup( optimizer=__lowercase , num_warmup_steps=100 , num_training_steps=(len(__lowercase ) * num_epochs) , ) # Prepare everything # There is no specific order to remember, we just need to unpack the objects in the same order we gave them to the # prepare method. _A , _A , _A , _A , _A = accelerator.prepare( __lowercase , __lowercase , __lowercase , __lowercase , __lowercase ) # Now we train the model for epoch in range(__lowercase ): model.train() for step, batch in enumerate(__lowercase ): # We could avoid this line since we set the accelerator with `device_placement=True`. batch.to(accelerator.device ) _A = model(**__lowercase ) _A = outputs.loss accelerator.backward(__lowercase ) optimizer.step() lr_scheduler.step() optimizer.zero_grad() model.eval() for step, batch in enumerate(__lowercase ): # We could avoid this line since we set the accelerator with `device_placement=True`. batch.to(accelerator.device ) with torch.no_grad(): _A = model(**__lowercase ) _A = outputs.logits.argmax(dim=-1 ) _A , _A = accelerator.gather_for_metrics((predictions, batch["labels"]) ) metric.add_batch( predictions=__lowercase , references=__lowercase , ) _A = metric.compute() # Use accelerator.print to print only on the main process. accelerator.print(f"""epoch {epoch}:""" , __lowercase ) # New Code # # And call it at the end with no arguments # Note: You could also refactor this outside of your training loop function inner_training_loop() def a__ ( ) -> int: _A = argparse.ArgumentParser(description="Simple example of training script." ) parser.add_argument( "--mixed_precision" , type=__lowercase , default=__lowercase , choices=["no", "fp16", "bf16", "fp8"] , help="Whether to use mixed precision. Choose" "between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >= 1.10." "and an Nvidia Ampere GPU." , ) parser.add_argument("--cpu" , action="store_true" , help="If passed, will train on the CPU." ) _A = parser.parse_args() _A = {"lr": 2E-5, "num_epochs": 3, "seed": 42, "batch_size": 16} training_function(__lowercase , __lowercase ) if __name__ == "__main__": main()
716
"""simple docstring""" import warnings from ...utils import logging from .image_processing_glpn import GLPNImageProcessor a_ = logging.get_logger(__name__) class snake_case ( _UpperCamelCase): def __init__( self : str , *a__ : Dict , **a__ : Optional[int] ) -> None: '''simple docstring''' warnings.warn( "The class GLPNFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please" " use GLPNImageProcessor instead." , a__ , ) super().__init__(*a__ , **a__ )
621
0
"""simple docstring""" import os import sys import unittest a_ = os.path.abspath(os.path.dirname(os.path.dirname(os.path.dirname(__file__)))) sys.path.append(os.path.join(git_repo_path, "utils")) import get_test_info # noqa: E402 from get_test_info import ( # noqa: E402 get_model_to_test_mapping, get_model_to_tester_mapping, get_test_to_tester_mapping, ) a_ = os.path.join("tests", "models", "bert", "test_modeling_bert.py") a_ = os.path.join("tests", "models", "blip", "test_modeling_blip.py") class snake_case ( unittest.TestCase): def a_ ( self : str ) -> List[str]: '''simple docstring''' _A = get_test_to_tester_mapping(a__ ) _A = get_test_to_tester_mapping(a__ ) _A = {"BertModelTest": "BertModelTester"} _A = { "BlipModelTest": "BlipModelTester", "BlipTextImageModelTest": "BlipTextImageModelsModelTester", "BlipTextModelTest": "BlipTextModelTester", "BlipTextRetrievalModelTest": "BlipTextRetrievalModelTester", "BlipVQAModelTest": "BlipVQAModelTester", "BlipVisionModelTest": "BlipVisionModelTester", } self.assertEqual(get_test_info.to_json(a__ ) , a__ ) self.assertEqual(get_test_info.to_json(a__ ) , a__ ) def a_ ( self : Union[str, Any] ) -> int: '''simple docstring''' _A = get_model_to_test_mapping(a__ ) _A = get_model_to_test_mapping(a__ ) _A = { "BertForMaskedLM": ["BertModelTest"], "BertForMultipleChoice": ["BertModelTest"], "BertForNextSentencePrediction": ["BertModelTest"], "BertForPreTraining": ["BertModelTest"], "BertForQuestionAnswering": ["BertModelTest"], "BertForSequenceClassification": ["BertModelTest"], "BertForTokenClassification": ["BertModelTest"], "BertLMHeadModel": ["BertModelTest"], "BertModel": ["BertModelTest"], } _A = { "BlipForConditionalGeneration": ["BlipTextImageModelTest"], "BlipForImageTextRetrieval": ["BlipTextRetrievalModelTest"], "BlipForQuestionAnswering": ["BlipVQAModelTest"], "BlipModel": ["BlipModelTest"], "BlipTextModel": ["BlipTextModelTest"], "BlipVisionModel": ["BlipVisionModelTest"], } self.assertEqual(get_test_info.to_json(a__ ) , a__ ) self.assertEqual(get_test_info.to_json(a__ ) , a__ ) def a_ ( self : str ) -> Tuple: '''simple docstring''' _A = get_model_to_tester_mapping(a__ ) _A = get_model_to_tester_mapping(a__ ) _A = { "BertForMaskedLM": ["BertModelTester"], "BertForMultipleChoice": ["BertModelTester"], "BertForNextSentencePrediction": ["BertModelTester"], "BertForPreTraining": ["BertModelTester"], "BertForQuestionAnswering": ["BertModelTester"], "BertForSequenceClassification": ["BertModelTester"], "BertForTokenClassification": ["BertModelTester"], "BertLMHeadModel": ["BertModelTester"], "BertModel": ["BertModelTester"], } _A = { "BlipForConditionalGeneration": ["BlipTextImageModelsModelTester"], "BlipForImageTextRetrieval": ["BlipTextRetrievalModelTester"], "BlipForQuestionAnswering": ["BlipVQAModelTester"], "BlipModel": ["BlipModelTester"], "BlipTextModel": ["BlipTextModelTester"], "BlipVisionModel": ["BlipVisionModelTester"], } self.assertEqual(get_test_info.to_json(a__ ) , a__ ) self.assertEqual(get_test_info.to_json(a__ ) , a__ )
717
"""simple docstring""" import argparse import torch from torch import nn from transformers import MBartConfig, MBartForConditionalGeneration def a__ ( __lowercase ) -> Optional[int]: _A = [ "encoder.version", "decoder.version", "model.encoder.version", "model.decoder.version", "_float_tensor", "decoder.output_projection.weight", ] for k in ignore_keys: state_dict.pop(__lowercase , __lowercase ) def a__ ( __lowercase ) -> List[Any]: _A , _A = emb.weight.shape _A = nn.Linear(__lowercase , __lowercase , bias=__lowercase ) _A = emb.weight.data return lin_layer def a__ ( __lowercase , __lowercase="facebook/mbart-large-en-ro" , __lowercase=False , __lowercase=False ) -> List[str]: _A = torch.load(__lowercase , map_location="cpu" )["model"] remove_ignore_keys_(__lowercase ) _A = state_dict["encoder.embed_tokens.weight"].shape[0] _A = MBartConfig.from_pretrained(__lowercase , vocab_size=__lowercase ) if mbart_aa and finetuned: _A = "relu" _A = state_dict["decoder.embed_tokens.weight"] _A = MBartForConditionalGeneration(__lowercase ) model.model.load_state_dict(__lowercase ) if finetuned: _A = make_linear_from_emb(model.model.shared ) return model if __name__ == "__main__": a_ = argparse.ArgumentParser() # Required parameters parser.add_argument( "fairseq_path", type=str, help="bart.large, bart.large.cnn or a path to a model.pt on local filesystem." ) parser.add_argument("pytorch_dump_folder_path", default=None, type=str, help="Path to the output PyTorch model.") parser.add_argument( "--hf_config", default="facebook/mbart-large-cc25", type=str, help="Which huggingface architecture to use: mbart-large", ) parser.add_argument("--mbart_50", action="store_true", help="whether the model is mMART-50 checkpoint") parser.add_argument("--finetuned", action="store_true", help="whether the model is a fine-tuned checkpoint") a_ = parser.parse_args() a_ = convert_fairseq_mbart_checkpoint_from_disk( args.fairseq_path, hf_config_path=args.hf_config, finetuned=args.finetuned, mbart_aa=args.mbart_aa ) model.save_pretrained(args.pytorch_dump_folder_path)
621
0
"""simple docstring""" import argparse import csv import logging import os import random import numpy as np import torch from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset from tqdm import tqdm, trange from transformers import ( CONFIG_NAME, WEIGHTS_NAME, AdamW, OpenAIGPTDoubleHeadsModel, OpenAIGPTTokenizer, get_linear_schedule_with_warmup, ) logging.basicConfig( format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO ) a_ = logging.getLogger(__name__) def a__ ( __lowercase , __lowercase ) -> int: _A = np.argmax(__lowercase , axis=1 ) return np.sum(outputs == labels ) def a__ ( __lowercase ) -> List[Any]: with open(__lowercase , encoding="utf_8" ) as f: _A = csv.reader(__lowercase ) _A = [] next(__lowercase ) # skip the first line for line in tqdm(__lowercase ): output.append((" ".join(line[1:5] ), line[5], line[6], int(line[-1] ) - 1) ) return output def a__ ( __lowercase , __lowercase , __lowercase , __lowercase , __lowercase , __lowercase ) -> int: _A = [] for dataset in encoded_datasets: _A = len(__lowercase ) _A = np.zeros((n_batch, 2, input_len) , dtype=np.intaa ) _A = np.zeros((n_batch, 2) , dtype=np.intaa ) _A = np.full((n_batch, 2, input_len) , fill_value=-100 , dtype=np.intaa ) _A = np.zeros((n_batch,) , dtype=np.intaa ) for ( i, (story, conta, conta, mc_label), ) in enumerate(__lowercase ): _A = [start_token] + story[:cap_length] + [delimiter_token] + conta[:cap_length] + [clf_token] _A = [start_token] + story[:cap_length] + [delimiter_token] + conta[:cap_length] + [clf_token] _A = with_conta _A = with_conta _A = len(__lowercase ) - 1 _A = len(__lowercase ) - 1 _A = with_conta _A = with_conta _A = mc_label _A = (input_ids, mc_token_ids, lm_labels, mc_labels) tensor_datasets.append(tuple(torch.tensor(__lowercase ) for t in all_inputs ) ) return tensor_datasets def a__ ( ) -> List[str]: _A = argparse.ArgumentParser() parser.add_argument("--model_name" , type=__lowercase , default="openai-gpt" , help="pretrained model name" ) parser.add_argument("--do_train" , action="store_true" , help="Whether to run training." ) parser.add_argument("--do_eval" , action="store_true" , help="Whether to run eval on the dev set." ) parser.add_argument( "--output_dir" , default=__lowercase , type=__lowercase , required=__lowercase , help="The output directory where the model predictions and checkpoints will be written." , ) parser.add_argument("--train_dataset" , type=__lowercase , default="" ) parser.add_argument("--eval_dataset" , type=__lowercase , default="" ) parser.add_argument("--seed" , type=__lowercase , default=42 ) parser.add_argument("--num_train_epochs" , type=__lowercase , default=3 ) parser.add_argument("--train_batch_size" , type=__lowercase , default=8 ) parser.add_argument("--eval_batch_size" , type=__lowercase , default=16 ) parser.add_argument("--adam_epsilon" , default=1E-8 , type=__lowercase , help="Epsilon for Adam optimizer." ) parser.add_argument("--max_grad_norm" , type=__lowercase , default=1 ) parser.add_argument( "--max_steps" , default=-1 , type=__lowercase , help=( "If > 0: set total number of training steps to perform. Override num_train_epochs." ) , ) parser.add_argument( "--gradient_accumulation_steps" , type=__lowercase , default=1 , help="Number of updates steps to accumulate before performing a backward/update pass." , ) parser.add_argument("--learning_rate" , type=__lowercase , default=6.25E-5 ) parser.add_argument("--warmup_steps" , default=0 , type=__lowercase , help="Linear warmup over warmup_steps." ) parser.add_argument("--lr_schedule" , type=__lowercase , default="warmup_linear" ) parser.add_argument("--weight_decay" , type=__lowercase , default=0.01 ) parser.add_argument("--lm_coef" , type=__lowercase , default=0.9 ) parser.add_argument("--n_valid" , type=__lowercase , default=374 ) parser.add_argument("--server_ip" , type=__lowercase , default="" , help="Can be used for distant debugging." ) parser.add_argument("--server_port" , type=__lowercase , default="" , help="Can be used for distant debugging." ) _A = parser.parse_args() print(__lowercase ) if args.server_ip and args.server_port: # Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script import ptvsd print("Waiting for debugger attach" ) ptvsd.enable_attach(address=(args.server_ip, args.server_port) , redirect_output=__lowercase ) ptvsd.wait_for_attach() random.seed(args.seed ) np.random.seed(args.seed ) torch.manual_seed(args.seed ) torch.cuda.manual_seed_all(args.seed ) _A = torch.device("cuda" if torch.cuda.is_available() else "cpu" ) _A = torch.cuda.device_count() logger.info("device: {}, n_gpu {}".format(__lowercase , __lowercase ) ) if not args.do_train and not args.do_eval: raise ValueError("At least one of `do_train` or `do_eval` must be True." ) if not os.path.exists(args.output_dir ): os.makedirs(args.output_dir ) # Load tokenizer and model # This loading functions also add new tokens and embeddings called `special tokens` # These new embeddings will be fine-tuned on the RocStories dataset _A = ["_start_", "_delimiter_", "_classify_"] _A = OpenAIGPTTokenizer.from_pretrained(args.model_name ) tokenizer.add_tokens(__lowercase ) _A = tokenizer.convert_tokens_to_ids(__lowercase ) _A = OpenAIGPTDoubleHeadsModel.from_pretrained(args.model_name ) model.resize_token_embeddings(len(__lowercase ) ) model.to(__lowercase ) # Load and encode the datasets def tokenize_and_encode(__lowercase ): if isinstance(__lowercase , __lowercase ): return tokenizer.convert_tokens_to_ids(tokenizer.tokenize(__lowercase ) ) elif isinstance(__lowercase , __lowercase ): return obj return [tokenize_and_encode(__lowercase ) for o in obj] logger.info("Encoding dataset..." ) _A = load_rocstories_dataset(args.train_dataset ) _A = load_rocstories_dataset(args.eval_dataset ) _A = (train_dataset, eval_dataset) _A = tokenize_and_encode(__lowercase ) # Compute the max input length for the Transformer _A = model.config.n_positions // 2 - 2 _A = max( len(story[:max_length] ) + max(len(conta[:max_length] ) , len(conta[:max_length] ) ) + 3 for dataset in encoded_datasets for story, conta, conta, _ in dataset ) _A = min(__lowercase , model.config.n_positions ) # Max size of input for the pre-trained model # Prepare inputs tensors and dataloaders _A = pre_process_datasets(__lowercase , __lowercase , __lowercase , *__lowercase ) _A , _A = tensor_datasets[0], tensor_datasets[1] _A = TensorDataset(*__lowercase ) _A = RandomSampler(__lowercase ) _A = DataLoader(__lowercase , sampler=__lowercase , batch_size=args.train_batch_size ) _A = TensorDataset(*__lowercase ) _A = SequentialSampler(__lowercase ) _A = DataLoader(__lowercase , sampler=__lowercase , batch_size=args.eval_batch_size ) # Prepare optimizer if args.do_train: if args.max_steps > 0: _A = args.max_steps _A = args.max_steps // (len(__lowercase ) // args.gradient_accumulation_steps) + 1 else: _A = len(__lowercase ) // args.gradient_accumulation_steps * args.num_train_epochs _A = list(model.named_parameters() ) _A = ["bias", "LayerNorm.bias", "LayerNorm.weight"] _A = [ { "params": [p for n, p in param_optimizer if not any(nd in n for nd in no_decay )], "weight_decay": args.weight_decay, }, {"params": [p for n, p in param_optimizer if any(nd in n for nd in no_decay )], "weight_decay": 0.0}, ] _A = AdamW(__lowercase , lr=args.learning_rate , eps=args.adam_epsilon ) _A = get_linear_schedule_with_warmup( __lowercase , num_warmup_steps=args.warmup_steps , num_training_steps=__lowercase ) if args.do_train: _A , _A , _A = 0, 0, None model.train() for _ in trange(int(args.num_train_epochs ) , desc="Epoch" ): _A = 0 _A = 0 _A = tqdm(__lowercase , desc="Training" ) for step, batch in enumerate(__lowercase ): _A = tuple(t.to(__lowercase ) for t in batch ) _A , _A , _A , _A = batch _A = model(__lowercase , mc_token_ids=__lowercase , lm_labels=__lowercase , mc_labels=__lowercase ) _A = args.lm_coef * losses[0] + losses[1] loss.backward() optimizer.step() scheduler.step() optimizer.zero_grad() tr_loss += loss.item() _A = ( loss.item() if exp_average_loss is None else 0.7 * exp_average_loss + 0.3 * loss.item() ) nb_tr_steps += 1 _A = "Training loss: {:.2e} lr: {:.2e}".format(__lowercase , scheduler.get_lr()[0] ) # Save a trained model if args.do_train: # Save a trained model, configuration and tokenizer _A = model.module if hasattr(__lowercase , "module" ) else model # Only save the model itself # If we save using the predefined names, we can load using `from_pretrained` _A = os.path.join(args.output_dir , __lowercase ) _A = os.path.join(args.output_dir , __lowercase ) torch.save(model_to_save.state_dict() , __lowercase ) model_to_save.config.to_json_file(__lowercase ) tokenizer.save_vocabulary(args.output_dir ) # Load a trained model and vocabulary that you have fine-tuned _A = OpenAIGPTDoubleHeadsModel.from_pretrained(args.output_dir ) _A = OpenAIGPTTokenizer.from_pretrained(args.output_dir ) model.to(__lowercase ) if args.do_eval: model.eval() _A , _A = 0, 0 _A , _A = 0, 0 for batch in tqdm(__lowercase , desc="Evaluating" ): _A = tuple(t.to(__lowercase ) for t in batch ) _A , _A , _A , _A = batch with torch.no_grad(): _A , _A , _A , _A = model( __lowercase , mc_token_ids=__lowercase , lm_labels=__lowercase , mc_labels=__lowercase ) _A = mc_logits.detach().cpu().numpy() _A = mc_labels.to("cpu" ).numpy() _A = accuracy(__lowercase , __lowercase ) eval_loss += mc_loss.mean().item() eval_accuracy += tmp_eval_accuracy nb_eval_examples += input_ids.size(0 ) nb_eval_steps += 1 _A = eval_loss / nb_eval_steps _A = eval_accuracy / nb_eval_examples _A = tr_loss / nb_tr_steps if args.do_train else None _A = {"eval_loss": eval_loss, "eval_accuracy": eval_accuracy, "train_loss": train_loss} _A = os.path.join(args.output_dir , "eval_results.txt" ) with open(__lowercase , "w" ) as writer: logger.info("***** Eval results *****" ) for key in sorted(result.keys() ): logger.info(" %s = %s" , __lowercase , str(result[key] ) ) writer.write("%s = %s\n" % (key, str(result[key] )) ) if __name__ == "__main__": main()
718
"""simple docstring""" import numpy as np def a__ ( __lowercase , __lowercase ) -> np.ndarray: return np.where(vector > 0 , __lowercase , (alpha * (np.exp(__lowercase ) - 1)) ) if __name__ == "__main__": import doctest doctest.testmod()
621
0
"""simple docstring""" from collections import OrderedDict from typing import Mapping from packaging import version from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { "microsoft/beit-base-patch16-224-pt22k": ( "https://huggingface.co/microsoft/beit-base-patch16-224-pt22k/resolve/main/config.json" ), # See all BEiT models at https://huggingface.co/models?filter=beit } class snake_case ( _UpperCamelCase): '''simple docstring''' __UpperCamelCase = 'beit' def __init__( self : Tuple , a__ : List[Any]=81_92 , a__ : Tuple=7_68 , a__ : Dict=12 , a__ : List[Any]=12 , a__ : Dict=30_72 , a__ : List[Any]="gelu" , a__ : Optional[Any]=0.0 , a__ : Optional[int]=0.0 , a__ : Dict=0.0_2 , a__ : str=1E-1_2 , a__ : Union[str, Any]=2_24 , a__ : Dict=16 , a__ : Any=3 , a__ : Optional[int]=False , a__ : Union[str, Any]=False , a__ : Any=False , a__ : Union[str, Any]=False , a__ : Any=0.1 , a__ : Dict=0.1 , a__ : int=True , a__ : str=[3, 5, 7, 11] , a__ : Union[str, Any]=[1, 2, 3, 6] , a__ : Any=True , a__ : Tuple=0.4 , a__ : str=2_56 , a__ : int=1 , a__ : Tuple=False , a__ : int=2_55 , **a__ : Optional[Any] , ) -> Optional[Any]: '''simple docstring''' super().__init__(**a__ ) _A = vocab_size _A = hidden_size _A = num_hidden_layers _A = num_attention_heads _A = intermediate_size _A = hidden_act _A = hidden_dropout_prob _A = attention_probs_dropout_prob _A = initializer_range _A = layer_norm_eps _A = image_size _A = patch_size _A = num_channels _A = use_mask_token _A = use_absolute_position_embeddings _A = use_relative_position_bias _A = use_shared_relative_position_bias _A = layer_scale_init_value _A = drop_path_rate _A = use_mean_pooling # decode head attributes (semantic segmentation) _A = out_indices _A = pool_scales # auxiliary head attributes (semantic segmentation) _A = use_auxiliary_head _A = auxiliary_loss_weight _A = auxiliary_channels _A = auxiliary_num_convs _A = auxiliary_concat_input _A = semantic_loss_ignore_index class snake_case ( _UpperCamelCase): '''simple docstring''' __UpperCamelCase = version.parse('1.11') @property def a_ ( self : Dict ) -> Mapping[str, Mapping[int, str]]: '''simple docstring''' return OrderedDict( [ ("pixel_values", {0: "batch", 1: "num_channels", 2: "height", 3: "width"}), ] ) @property def a_ ( self : Dict ) -> float: '''simple docstring''' return 1E-4
719
"""simple docstring""" import os import re import warnings from shutil import copyfile from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import PreTrainedTokenizer if TYPE_CHECKING: from ...tokenization_utils_base import TextInput from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"vocab_file": "spiece.model"} a_ = { "vocab_file": { "t5-small": "https://huggingface.co/t5-small/resolve/main/spiece.model", "t5-base": "https://huggingface.co/t5-base/resolve/main/spiece.model", "t5-large": "https://huggingface.co/t5-large/resolve/main/spiece.model", "t5-3b": "https://huggingface.co/t5-3b/resolve/main/spiece.model", "t5-11b": "https://huggingface.co/t5-11b/resolve/main/spiece.model", } } # TODO(PVP) - this should be removed in Transformers v5 a_ = { "t5-small": 5_12, "t5-base": 5_12, "t5-large": 5_12, "t5-3b": 5_12, "t5-11b": 5_12, } a_ = "▁" class snake_case ( _UpperCamelCase): __UpperCamelCase = VOCAB_FILES_NAMES __UpperCamelCase = PRETRAINED_VOCAB_FILES_MAP __UpperCamelCase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES __UpperCamelCase = ['input_ids', 'attention_mask'] def __init__( self : List[str] , a__ : Optional[int] , a__ : Union[str, Any]="</s>" , a__ : Union[str, Any]="<unk>" , a__ : str="<pad>" , a__ : Optional[int]=1_00 , a__ : List[Any]=None , a__ : Optional[Dict[str, Any]] = None , a__ : Any=True , **a__ : Optional[int] , ) -> None: '''simple docstring''' if extra_ids > 0 and additional_special_tokens is None: _A = [F"""<extra_id_{i}>""" for i in range(a__ )] elif extra_ids > 0 and additional_special_tokens is not None: # Check that we have the right number of extra_id special tokens _A = len(set(filter(lambda a__ : bool("extra_id" in str(a__ ) ) , a__ ) ) ) if extra_tokens != extra_ids: raise ValueError( F"""Both extra_ids ({extra_ids}) and additional_special_tokens ({additional_special_tokens}) are""" " provided to T5Tokenizer. In this case the additional_special_tokens must include the extra_ids" " tokens" ) if legacy: logger.warning_once( F"""You are using the legacy behaviour of the {self.__class__}. This means that tokens that come after special tokens will not be properly handled. We recommend you to""" " read the related pull request available at https://github.com/huggingface/transformers/pull/24565" ) _A = legacy _A = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( eos_token=a__ , unk_token=a__ , pad_token=a__ , extra_ids=a__ , additional_special_tokens=a__ , sp_model_kwargs=self.sp_model_kwargs , legacy=a__ , **a__ , ) _A = vocab_file _A = extra_ids _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(a__ ) @staticmethod def a_ ( a__ : List[str] , a__ : Optional[int] , a__ : Tuple ) -> Tuple: '''simple docstring''' if pretrained_model_name_or_path in TaTokenizer.max_model_input_sizes: _A = TaTokenizer.max_model_input_sizes[pretrained_model_name_or_path] if init_max_model_length is not None and init_max_model_length != max_model_length: return init_max_model_length elif init_max_model_length is None: warnings.warn( "This tokenizer was incorrectly instantiated with a model max length of" F""" {deprecated_max_model_length} which will be corrected in Transformers v5.\nFor now, this""" " behavior is kept to avoid breaking backwards compatibility when padding/encoding with" " `truncation is True`.\n- Be aware that you SHOULD NOT rely on" F""" {pretrained_model_name_or_path} automatically truncating your input to""" F""" {deprecated_max_model_length} when padding/encoding.\n- If you want to encode/pad to sequences""" F""" longer than {deprecated_max_model_length} you can either instantiate this tokenizer with""" " `model_max_length` or pass `max_length` when encoding/padding.\n- To avoid this warning, please" " instantiate this tokenizer with `model_max_length` set to your preferred value." , a__ , ) return max_model_length @property def a_ ( self : List[Any] ) -> Dict: '''simple docstring''' return self.sp_model.get_piece_size() + self._extra_ids def a_ ( self : Dict ) -> Optional[Any]: '''simple docstring''' _A = {self.convert_ids_to_tokens(a__ ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def a_ ( self : Optional[Any] , a__ : List[int] , a__ : Optional[List[int]] = None , a__ : bool = False ) -> List[int]: '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=a__ , token_ids_a=a__ , already_has_special_tokens=a__ ) # normal case: some special tokens if token_ids_a is None: return ([0] * len(a__ )) + [1] return ([0] * len(a__ )) + [1] + ([0] * len(a__ )) + [1] def a_ ( self : List[str] ) -> List[str]: '''simple docstring''' return list( set(filter(lambda a__ : bool(re.search(r"<extra_id_\d+>" , a__ ) ) is not None , self.additional_special_tokens ) ) ) def a_ ( self : str ) -> List[Any]: '''simple docstring''' return [self._convert_token_to_id(a__ ) for token in self.get_sentinel_tokens()] def a_ ( self : List[Any] , a__ : List[int] ) -> List[int]: '''simple docstring''' if len(a__ ) > 0 and token_ids[-1] == self.eos_token_id: warnings.warn( F"""This sequence already has {self.eos_token}. In future versions this behavior may lead to duplicated""" " eos tokens being added." ) return token_ids else: return token_ids + [self.eos_token_id] def a_ ( self : int , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = [self.eos_token_id] if token_ids_a is None: return len(token_ids_a + eos ) * [0] return len(token_ids_a + eos + token_ids_a + eos ) * [0] def a_ ( self : Union[str, Any] , a__ : List[int] , a__ : Optional[List[int]] = None ) -> List[int]: '''simple docstring''' _A = self._add_eos_if_not_present(a__ ) if token_ids_a is None: return token_ids_a else: _A = self._add_eos_if_not_present(a__ ) return token_ids_a + token_ids_a def __getstate__( self : Dict ) -> Union[str, Any]: '''simple docstring''' _A = self.__dict__.copy() _A = None return state def __setstate__( self : int , a__ : Optional[int] ) -> Union[str, Any]: '''simple docstring''' _A = d # for backward compatibility if not hasattr(self , "sp_model_kwargs" ): _A = {} _A = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def a_ ( self : int , a__ : "TextInput" , **a__ : List[str] ) -> List[str]: '''simple docstring''' if not self.legacy: _A = SPIECE_UNDERLINE + text.replace(a__ , " " ) return super().tokenize(a__ , **a__ ) def a_ ( self : str , a__ : Dict , **a__ : Optional[int] ) -> Any: '''simple docstring''' if not self.legacy: _A = text.startswith(a__ ) if is_first: _A = text[1:] _A = self.sp_model.encode(a__ , out_type=a__ ) if not self.legacy and not is_first and not text.startswith(" " ) and tokens[0].startswith(a__ ): _A = ([tokens[0][1:]] if len(tokens[0] ) > 1 else []) + tokens[1:] return tokens def a_ ( self : int , a__ : List[Any] ) -> List[str]: '''simple docstring''' if token.startswith("<extra_id_" ): _A = re.match(r"<extra_id_(\d+)>" , a__ ) _A = int(match.group(1 ) ) return self.vocab_size - num - 1 return self.sp_model.piece_to_id(a__ ) def a_ ( self : Dict , a__ : Union[str, Any] ) -> Any: '''simple docstring''' if index < self.sp_model.get_piece_size(): _A = self.sp_model.IdToPiece(a__ ) else: _A = F"""<extra_id_{self.vocab_size - 1 - index}>""" return token def a_ ( self : Optional[int] , a__ : Tuple ) -> List[str]: '''simple docstring''' _A = [] _A = "" _A = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(a__ ) + token _A = True _A = [] else: current_sub_tokens.append(a__ ) _A = False out_string += self.sp_model.decode(a__ ) return out_string.strip() def a_ ( self : Dict , a__ : str , a__ : Optional[str] = None ) -> Tuple[str]: '''simple docstring''' if not os.path.isdir(a__ ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return _A = os.path.join( a__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(a__ ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , a__ ) elif not os.path.isfile(self.vocab_file ): with open(a__ , "wb" ) as fi: _A = self.sp_model.serialized_model_proto() fi.write(a__ ) return (out_vocab_file,)
621
0
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> list: _A = word.split() def justify(__lowercase , __lowercase , __lowercase ) -> str: _A = max_width - width _A = len(__lowercase ) if len(__lowercase ) == 1: # if there is only word in line # just insert overall_spaces_count for the remainder of line return line[0] + " " * overall_spaces_count else: _A = words_count - 1 # num_spaces_between_words_list[i] : tells you to insert # num_spaces_between_words_list[i] spaces # after word on line[i] _A = spaces_to_insert_between_words * [ overall_spaces_count // spaces_to_insert_between_words ] _A = ( overall_spaces_count % spaces_to_insert_between_words ) # distribute spaces via round robin to the left words for i in range(__lowercase ): num_spaces_between_words_list[i] += 1 _A = [] for i in range(__lowercase ): # add the word aligned_words_list.append(line[i] ) # add the spaces to insert aligned_words_list.append(num_spaces_between_words_list[i] * " " ) # just add the last word to the sentence aligned_words_list.append(line[-1] ) # join the aligned words list to form a justified line return "".join(__lowercase ) _A = [] _A = [] _A = 0 for word in words: if width + len(__lowercase ) + len(__lowercase ) <= max_width: # keep adding words until we can fill out max_width # width = sum of length of all words (without overall_spaces_count) # len(word) = length of current word # len(line) = number of overall_spaces_count to insert between words line.append(__lowercase ) width += len(__lowercase ) else: # justify the line and add it to result answer.append(justify(__lowercase , __lowercase , __lowercase ) ) # reset new line and new width _A , _A = [word], len(__lowercase ) _A = max_width - width - len(__lowercase ) answer.append(" ".join(__lowercase ) + (remaining_spaces + 1) * " " ) return answer if __name__ == "__main__": from doctest import testmod testmod()
720
"""simple docstring""" import argparse import json import os from collections import OrderedDict import numpy as np import tensorflow as tf import torch def a__ ( __lowercase ) -> List[Any]: _A = os.path.join(args.tf_model_dir , "parameters.json" ) _A = json.loads(open(__lowercase ).read() ) if not params: raise ValueError( f"""It seems that the json file at {parameter_file} is empty. Make sure you have a correct json file.""" ) if not args.output.endswith(".pt" ): _A = args.output + ".pt" _A = OrderedDict() with tf.device("/CPU:0" ): _A = tf.train.load_checkpoint(args.tf_model_dir ) _A = reader.get_variable_to_shape_map() for key_name in shapes.keys(): _A = reader.get_tensor(__lowercase ).astype(np.floataa ) if key_name.endswith("/adam_m" ) or key_name.endswith("/adam_v" ): continue if key_name.startswith("pasts/" ): if key_name.startswith("pasts/mlp" ): _A = int(key_name[9] ) elif key_name.startswith("pasts/out" ): _A = 8 _A = "model.sqout.%d.weight" % (player * 2) # enter to nn.Sequencial with Tanh, so 2 at a time _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/moe" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/switch_gating/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.router.classifier.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/softmlp/kernel" ): _A = "model.blocks.%d.feed_forward.soft_bypass_mlp.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/wo/kernel" ) or key_name.endswith("/wi/kernel" ): _A = key_name[-9:-7] for i in range(16 ): _A = "model.blocks.%d.feed_forward.mlp.experts.expert_%d.%s.weight" % (player, i, nlayer) _A = ( vnp[i].transpose([1, 0] ).copy() ) # In Mesh-Tensorflow, it is one array, so it is divided _A = torch.tensor(__lowercase ) elif key_name.startswith("model/mlp" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/p1/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wi.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p1/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wi.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/kernel" ): _A = "model.blocks.%d.feed_forward.mlp.wo.weight" % player _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.endswith("/p2/bias" ): _A = "model.blocks.%d.feed_forward.mlp.wo.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/ln" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.feed_forward.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.feed_forward.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.startswith("model/att" ): _A = int(key_name[9:].split("/" )[0] ) if key_name.endswith("/qkv/kernel" ): _A = vnp.copy() # Compute same dimension as Mesh-tensorflow using einsum _A = state[:, 0, :, :] _A = state[:, 1, :, :] _A = state[:, 2, :, :] _A = ( state_q.reshape([state_q.shape[0], state_q.shape[1] * state_q.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_k.reshape([state_k.shape[0], state_k.shape[1] * state_k.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = ( state_v.reshape([state_v.shape[0], state_v.shape[1] * state_v.shape[2]] ) .transpose([1, 0] ) .copy() ) # Mesh-Tensorflow is a diagonal matrix _A = "model.blocks.%d.self_attn.self_attn.q_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.k_proj.weight" % player _A = torch.tensor(__lowercase ) _A = "model.blocks.%d.self_attn.self_attn.v_proj.weight" % player _A = torch.tensor(__lowercase ) elif key_name.endswith("/o/kernel" ): _A = "model.blocks.%d.self_attn.self_attn.out_proj.weight" % player _A = ( vnp.reshape([vnp.shape[0] * vnp.shape[1], vnp.shape[2]] ).transpose([1, 0] ).copy() ) # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name.startswith("model/an" ): _A = int(key_name[8:].split("/" )[0] ) if key_name.endswith("/b" ): _A = "model.blocks.%d.self_attn.norm.bias" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif key_name.endswith("/g" ): _A = "model.blocks.%d.self_attn.norm.weight" % player _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) elif ( key_name.startswith("model/wte" ) or key_name.startswith("model/wpe" ) or key_name.startswith("model/ete" ) ): _A = {"wte": "embed_tokens", "wpe": "position_embeddings", "ete": "extra_position_embeddings"}[ key_name[-3:] ] _A = "model.%s.weight" % nlayer _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) if key_name.startswith("model/wte" ): _A = "lm_head.weight" _A = vnp.copy() # same in embedded _A = torch.tensor(__lowercase ) elif key_name.startswith("model/wob" ): _A = "final_logits_bias" _A = vnp.copy() # same in embedded _A = state.reshape((1, -1) ) _A = torch.tensor(__lowercase ) elif key_name == "model/dense/kernel": _A = "model.last_project.weight" _A = vnp.transpose([1, 0] ).copy() # Mesh-Tensorflow is a diagonal matrix _A = torch.tensor(__lowercase ) elif key_name == "model/dense_1/bias": _A = "model.last_project.bias" _A = vnp.copy() # same because it is one dimensional _A = torch.tensor(__lowercase ) torch.save(__lowercase , args.output ) if __name__ == "__main__": a_ = argparse.ArgumentParser( description="model converter.", formatter_class=argparse.ArgumentDefaultsHelpFormatter ) parser.add_argument("--tf_model_dir", metavar="PATH", type=str, required=True, help="import model") parser.add_argument("--output", metavar="PATH", type=str, required=True, help="output model") a_ = parser.parse_args() convert_tf_gptsan_to_pt(args)
621
0
"""simple docstring""" def a__ ( __lowercase , __lowercase ) -> int: return abs(__lowercase ) if a == 0 else greatest_common_divisor(b % a , __lowercase ) def a__ ( __lowercase , __lowercase ) -> int: while y: # --> when y=0 then loop will terminate and return x as final GCD. _A , _A = y, x % y return abs(__lowercase ) def a__ ( ) -> Tuple: try: _A = input("Enter two integers separated by comma (,): " ).split("," ) _A = int(nums[0] ) _A = int(nums[1] ) print( f"""greatest_common_divisor({num_a}, {num_a}) = """ f"""{greatest_common_divisor(__lowercase , __lowercase )}""" ) print(f"""By iterative gcd({num_a}, {num_a}) = {gcd_by_iterative(__lowercase , __lowercase )}""" ) except (IndexError, UnboundLocalError, ValueError): print("Wrong input" ) if __name__ == "__main__": main()
721
"""simple docstring""" import argparse import torch from transformers import GPTaLMHeadModel, RobertaForMaskedLM if __name__ == "__main__": a_ = argparse.ArgumentParser( description=( "Extraction some layers of the full RobertaForMaskedLM or GPT2LMHeadModel for Transfer Learned" " Distillation" ) ) parser.add_argument("--model_type", default="roberta", choices=["roberta", "gpt2"]) parser.add_argument("--model_name", default="roberta-large", type=str) parser.add_argument("--dump_checkpoint", default="serialization_dir/tf_roberta_048131723.pth", type=str) parser.add_argument("--vocab_transform", action="store_true") a_ = parser.parse_args() if args.model_type == "roberta": a_ = RobertaForMaskedLM.from_pretrained(args.model_name) a_ = "roberta" elif args.model_type == "gpt2": a_ = GPTaLMHeadModel.from_pretrained(args.model_name) a_ = "transformer" a_ = model.state_dict() a_ = {} # Embeddings # if args.model_type == "gpt2": for param_name in ["wte.weight", "wpe.weight"]: a_ = state_dict[f'''{prefix}.{param_name}'''] else: for w in ["word_embeddings", "position_embeddings", "token_type_embeddings"]: a_ = f'''{prefix}.embeddings.{w}.weight''' a_ = state_dict[param_name] for w in ["weight", "bias"]: a_ = f'''{prefix}.embeddings.LayerNorm.{w}''' a_ = state_dict[param_name] # Transformer Blocks # a_ = 0 for teacher_idx in [0, 2, 4, 7, 9, 11]: if args.model_type == "gpt2": for layer in ["ln_1", "attn.c_attn", "attn.c_proj", "ln_2", "mlp.c_fc", "mlp.c_proj"]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.h.{teacher_idx}.{layer}.{w}''' ] a_ = state_dict[f'''{prefix}.h.{teacher_idx}.attn.bias'''] else: for layer in [ "attention.self.query", "attention.self.key", "attention.self.value", "attention.output.dense", "attention.output.LayerNorm", "intermediate.dense", "output.dense", "output.LayerNorm", ]: for w in ["weight", "bias"]: a_ = state_dict[ f'''{prefix}.encoder.layer.{teacher_idx}.{layer}.{w}''' ] std_idx += 1 # Language Modeling Head ###s if args.model_type == "roberta": for layer in ["lm_head.decoder.weight", "lm_head.bias"]: a_ = state_dict[f'''{layer}'''] if args.vocab_transform: for w in ["weight", "bias"]: a_ = state_dict[f'''lm_head.dense.{w}'''] a_ = state_dict[f'''lm_head.layer_norm.{w}'''] elif args.model_type == "gpt2": for w in ["weight", "bias"]: a_ = state_dict[f'''{prefix}.ln_f.{w}'''] a_ = state_dict["lm_head.weight"] print(f'''N layers selected for distillation: {std_idx}''') print(f'''Number of params transferred for distillation: {len(compressed_sd.keys())}''') print(f'''Save transferred checkpoint to {args.dump_checkpoint}.''') torch.save(compressed_sd, args.dump_checkpoint)
621
0
from typing import Dict, Optional, Union import numpy as np from ...image_processing_utils import BaseImageProcessor, BatchFeature, get_size_dict from ...image_transforms import flip_channel_order, resize, to_channel_dimension_format, to_pil_image from ...image_utils import ( ChannelDimension, ImageInput, PILImageResampling, make_list_of_images, to_numpy_array, valid_images, ) from ...utils import TensorType, is_pytesseract_available, is_vision_available, logging, requires_backends if is_vision_available(): import PIL # soft dependency if is_pytesseract_available(): import pytesseract a_ = logging.get_logger(__name__) def a__ ( _UpperCamelCase : Tuple ,_UpperCamelCase : Dict ,_UpperCamelCase : Optional[Any] ): return [ int(10_00 * (box[0] / width) ), int(10_00 * (box[1] / height) ), int(10_00 * (box[2] / width) ), int(10_00 * (box[3] / height) ), ] def a__ ( _UpperCamelCase : np.ndarray ,_UpperCamelCase : Optional[str] ,_UpperCamelCase : Optional[str] = None ): __lowerCamelCase = tesseract_config if tesseract_config is not None else '''''' # apply OCR __lowerCamelCase = to_pil_image(_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = pil_image.size __lowerCamelCase = pytesseract.image_to_data(_UpperCamelCase ,lang=_UpperCamelCase ,output_type='''dict''' ,config=_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = data['''text'''], data['''left'''], data['''top'''], data['''width'''], data['''height'''] # filter empty words and corresponding coordinates __lowerCamelCase = [idx for idx, word in enumerate(_UpperCamelCase ) if not word.strip()] __lowerCamelCase = [word for idx, word in enumerate(_UpperCamelCase ) if idx not in irrelevant_indices] __lowerCamelCase = [coord for idx, coord in enumerate(_UpperCamelCase ) if idx not in irrelevant_indices] __lowerCamelCase = [coord for idx, coord in enumerate(_UpperCamelCase ) if idx not in irrelevant_indices] __lowerCamelCase = [coord for idx, coord in enumerate(_UpperCamelCase ) if idx not in irrelevant_indices] __lowerCamelCase = [coord for idx, coord in enumerate(_UpperCamelCase ) if idx not in irrelevant_indices] # turn coordinates into (left, top, left+width, top+height) format __lowerCamelCase = [] for x, y, w, h in zip(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = [x, y, x + w, y + h] actual_boxes.append(_UpperCamelCase ) # finally, normalize the bounding boxes __lowerCamelCase = [] for box in actual_boxes: normalized_boxes.append(normalize_box(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ) ) assert len(_UpperCamelCase ) == len(_UpperCamelCase ), "Not as many words as there are bounding boxes" return words, normalized_boxes class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = ["""pixel_values"""] def __init__( self , __UpperCAmelCase = True , __UpperCAmelCase = None , __UpperCAmelCase = PILImageResampling.BILINEAR , __UpperCAmelCase = True , __UpperCAmelCase = None , __UpperCAmelCase = "" , **__UpperCAmelCase , ): '''simple docstring''' super().__init__(**__UpperCAmelCase ) __lowerCamelCase = size if size is not None else {'''height''': 224, '''width''': 224} __lowerCamelCase = get_size_dict(__UpperCAmelCase ) __lowerCamelCase = do_resize __lowerCamelCase = size __lowerCamelCase = resample __lowerCamelCase = apply_ocr __lowerCamelCase = ocr_lang __lowerCamelCase = tesseract_config def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase = PILImageResampling.BILINEAR , __UpperCAmelCase = None , **__UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = get_size_dict(__UpperCAmelCase ) if "height" not in size or "width" not in size: raise ValueError(F"""The size dictionary must contain the keys 'height' and 'width'. Got {size.keys()}""" ) __lowerCamelCase = (size['''height'''], size['''width''']) return resize(__UpperCAmelCase , size=__UpperCAmelCase , resample=__UpperCAmelCase , data_format=__UpperCAmelCase , **__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = ChannelDimension.FIRST , **__UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = do_resize if do_resize is not None else self.do_resize __lowerCamelCase = size if size is not None else self.size __lowerCamelCase = get_size_dict(__UpperCAmelCase ) __lowerCamelCase = resample if resample is not None else self.resample __lowerCamelCase = apply_ocr if apply_ocr is not None else self.apply_ocr __lowerCamelCase = ocr_lang if ocr_lang is not None else self.ocr_lang __lowerCamelCase = tesseract_config if tesseract_config is not None else self.tesseract_config __lowerCamelCase = make_list_of_images(__UpperCAmelCase ) if not valid_images(__UpperCAmelCase ): raise ValueError( '''Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, ''' '''torch.Tensor, tf.Tensor or jax.ndarray.''' ) if do_resize and size is None: raise ValueError('''Size must be specified if do_resize is True.''' ) # All transformations expect numpy arrays. __lowerCamelCase = [to_numpy_array(__UpperCAmelCase ) for image in images] if apply_ocr: requires_backends(self , '''pytesseract''' ) __lowerCamelCase = [] __lowerCamelCase = [] for image in images: __lowerCamelCase ,__lowerCamelCase = apply_tesseract(__UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase ) words_batch.append(__UpperCAmelCase ) boxes_batch.append(__UpperCAmelCase ) if do_resize: __lowerCamelCase = [self.resize(image=__UpperCAmelCase , size=__UpperCAmelCase , resample=__UpperCAmelCase ) for image in images] # flip color channels from RGB to BGR (as Detectron2 requires this) __lowerCamelCase = [flip_channel_order(__UpperCAmelCase ) for image in images] __lowerCamelCase = [to_channel_dimension_format(__UpperCAmelCase , __UpperCAmelCase ) for image in images] __lowerCamelCase = BatchFeature(data={'''pixel_values''': images} , tensor_type=__UpperCAmelCase ) if apply_ocr: __lowerCamelCase = words_batch __lowerCamelCase = boxes_batch return data
622
import os import unicodedata from shutil import copyfile from typing import Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import AddedToken, PreTrainedTokenizer from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"""vocab_file""": """spiece.model"""} a_ = { """vocab_file""": { """albert-base-v1""": """https://huggingface.co/albert-base-v1/resolve/main/spiece.model""", """albert-large-v1""": """https://huggingface.co/albert-large-v1/resolve/main/spiece.model""", """albert-xlarge-v1""": """https://huggingface.co/albert-xlarge-v1/resolve/main/spiece.model""", """albert-xxlarge-v1""": """https://huggingface.co/albert-xxlarge-v1/resolve/main/spiece.model""", """albert-base-v2""": """https://huggingface.co/albert-base-v2/resolve/main/spiece.model""", """albert-large-v2""": """https://huggingface.co/albert-large-v2/resolve/main/spiece.model""", """albert-xlarge-v2""": """https://huggingface.co/albert-xlarge-v2/resolve/main/spiece.model""", """albert-xxlarge-v2""": """https://huggingface.co/albert-xxlarge-v2/resolve/main/spiece.model""", } } a_ = { """albert-base-v1""": 512, """albert-large-v1""": 512, """albert-xlarge-v1""": 512, """albert-xxlarge-v1""": 512, """albert-base-v2""": 512, """albert-large-v2""": 512, """albert-xlarge-v2""": 512, """albert-xxlarge-v2""": 512, } a_ = """▁""" class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = VOCAB_FILES_NAMES lowerCAmelCase__ = PRETRAINED_VOCAB_FILES_MAP lowerCAmelCase__ = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES def __init__( self , __UpperCAmelCase , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=False , __UpperCAmelCase="[CLS]" , __UpperCAmelCase="[SEP]" , __UpperCAmelCase="<unk>" , __UpperCAmelCase="[SEP]" , __UpperCAmelCase="<pad>" , __UpperCAmelCase="[CLS]" , __UpperCAmelCase="[MASK]" , __UpperCAmelCase = None , **__UpperCAmelCase , ): '''simple docstring''' # Mask token behave like a normal word, i.e. include the space before it and # is included in the raw text, there should be a match in a non-normalized sentence. __lowerCamelCase = ( AddedToken(__UpperCAmelCase , lstrip=__UpperCAmelCase , rstrip=__UpperCAmelCase , normalized=__UpperCAmelCase ) if isinstance(__UpperCAmelCase , __UpperCAmelCase ) else mask_token ) __lowerCamelCase = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( do_lower_case=__UpperCAmelCase , remove_space=__UpperCAmelCase , keep_accents=__UpperCAmelCase , bos_token=__UpperCAmelCase , eos_token=__UpperCAmelCase , unk_token=__UpperCAmelCase , sep_token=__UpperCAmelCase , pad_token=__UpperCAmelCase , cls_token=__UpperCAmelCase , mask_token=__UpperCAmelCase , sp_model_kwargs=self.sp_model_kwargs , **__UpperCAmelCase , ) __lowerCamelCase = do_lower_case __lowerCamelCase = remove_space __lowerCamelCase = keep_accents __lowerCamelCase = vocab_file __lowerCamelCase = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(__UpperCAmelCase ) @property def lowerCamelCase ( self ): '''simple docstring''' return len(self.sp_model ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = {self.convert_ids_to_tokens(__UpperCAmelCase ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def __getstate__( self ): '''simple docstring''' __lowerCamelCase = self.__dict__.copy() __lowerCamelCase = None return state def __setstate__( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = d # for backward compatibility if not hasattr(self , '''sp_model_kwargs''' ): __lowerCamelCase = {} __lowerCamelCase = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' if self.remove_space: __lowerCamelCase = ''' '''.join(inputs.strip().split() ) else: __lowerCamelCase = inputs __lowerCamelCase = outputs.replace('''``''' , '''"''' ).replace('''\'\'''' , '''"''' ) if not self.keep_accents: __lowerCamelCase = unicodedata.normalize('''NFKD''' , __UpperCAmelCase ) __lowerCamelCase = ''''''.join([c for c in outputs if not unicodedata.combining(__UpperCAmelCase )] ) if self.do_lower_case: __lowerCamelCase = outputs.lower() return outputs def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.preprocess_text(__UpperCAmelCase ) __lowerCamelCase = self.sp_model.encode(__UpperCAmelCase , out_type=__UpperCAmelCase ) __lowerCamelCase = [] for piece in pieces: if len(__UpperCAmelCase ) > 1 and piece[-1] == str(''',''' ) and piece[-2].isdigit(): __lowerCamelCase = self.sp_model.EncodeAsPieces(piece[:-1].replace(__UpperCAmelCase , '''''' ) ) if piece[0] != SPIECE_UNDERLINE and cur_pieces[0][0] == SPIECE_UNDERLINE: if len(cur_pieces[0] ) == 1: __lowerCamelCase = cur_pieces[1:] else: __lowerCamelCase = cur_pieces[0][1:] cur_pieces.append(piece[-1] ) new_pieces.extend(__UpperCAmelCase ) else: new_pieces.append(__UpperCAmelCase ) return new_pieces def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.sp_model.PieceToId(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.sp_model.IdToPiece(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = [] __lowerCamelCase = '''''' __lowerCamelCase = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(__UpperCAmelCase ) + token __lowerCamelCase = True __lowerCamelCase = [] else: current_sub_tokens.append(__UpperCAmelCase ) __lowerCamelCase = False out_string += self.sp_model.decode(__UpperCAmelCase ) return out_string.strip() def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' __lowerCamelCase = [self.sep_token_id] __lowerCamelCase = [self.cls_token_id] if token_ids_a is None: return cls + token_ids_a + sep return cls + token_ids_a + sep + token_ids_a + sep def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None , __UpperCAmelCase = False ): '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=__UpperCAmelCase , token_ids_a=__UpperCAmelCase , already_has_special_tokens=__UpperCAmelCase ) if token_ids_a is not None: return [1] + ([0] * len(__UpperCAmelCase )) + [1] + ([0] * len(__UpperCAmelCase )) + [1] return [1] + ([0] * len(__UpperCAmelCase )) + [1] def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' __lowerCamelCase = [self.sep_token_id] __lowerCamelCase = [self.cls_token_id] if token_ids_a is None: return len(cls + token_ids_a + sep ) * [0] return len(cls + token_ids_a + sep ) * [0] + len(token_ids_a + sep ) * [1] def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' if not os.path.isdir(__UpperCAmelCase ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return __lowerCamelCase = os.path.join( __UpperCAmelCase , (filename_prefix + '''-''' if filename_prefix else '''''') + VOCAB_FILES_NAMES['''vocab_file'''] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(__UpperCAmelCase ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , __UpperCAmelCase ) elif not os.path.isfile(self.vocab_file ): with open(__UpperCAmelCase , '''wb''' ) as fi: __lowerCamelCase = self.sp_model.serialized_model_proto() fi.write(__UpperCAmelCase ) return (out_vocab_file,)
622
1
import os import pytest from attr import dataclass a_ = """us-east-1""" # defaults region @dataclass class __lowerCAmelCase : lowerCAmelCase__ = 42 lowerCAmelCase__ = """arn:aws:iam::558105141721:role/sagemaker_execution_role""" lowerCAmelCase__ = { """task_name""": """mnli""", """per_device_train_batch_size""": 1_6, """per_device_eval_batch_size""": 1_6, """do_train""": True, """do_eval""": True, """do_predict""": True, """output_dir""": """/opt/ml/model""", """overwrite_output_dir""": True, """max_steps""": 5_0_0, """save_steps""": 5_5_0_0, } lowerCAmelCase__ = {**hyperparameters, """max_steps""": 1_0_0_0} @property def lowerCamelCase ( self ): '''simple docstring''' if self.framework == "pytorch": return [ {"Name": "train_runtime", "Regex": r"train_runtime.*=\D*(.*?)$"}, {"Name": "eval_accuracy", "Regex": r"eval_accuracy.*=\D*(.*?)$"}, {"Name": "eval_loss", "Regex": r"eval_loss.*=\D*(.*?)$"}, ] else: return [ {"Name": "train_runtime", "Regex": r"train_runtime.*=\D*(.*?)$"}, {"Name": "eval_accuracy", "Regex": r"loss.*=\D*(.*?)]?$"}, {"Name": "eval_loss", "Regex": r"sparse_categorical_accuracy.*=\D*(.*?)]?$"}, ] @property def lowerCamelCase ( self ): '''simple docstring''' return F"""{self.framework}-transfromers-test""" @property def lowerCamelCase ( self ): '''simple docstring''' return F"""./tests/sagemaker/scripts/{self.framework}""" @property def lowerCamelCase ( self ): '''simple docstring''' if self.framework == "pytorch": return "763104351884.dkr.ecr.us-east-1.amazonaws.com/huggingface-pytorch-training:1.7.1-transformers4.6.1-gpu-py36-cu110-ubuntu18.04" else: return "763104351884.dkr.ecr.us-east-1.amazonaws.com/huggingface-tensorflow-training:2.4.1-transformers4.6.1-gpu-py37-cu110-ubuntu18.04" @pytest.fixture(scope='''class''' ) def a__ ( _UpperCamelCase : str ): __lowerCamelCase = SageMakerTestEnvironment(framework=request.cls.framework )
622
import math import os from copy import deepcopy import datasets import evaluate import torch import transformers from datasets import load_dataset from torch.utils.data import DataLoader from transformers import AutoModelForSequenceClassification, AutoTokenizer from accelerate import Accelerator from accelerate.test_utils import RegressionDataset, RegressionModel from accelerate.utils import is_tpu_available, set_seed a_ = """true""" def a__ ( _UpperCamelCase : Union[str, Any] ,_UpperCamelCase : List[str]=82 ,_UpperCamelCase : Optional[Any]=16 ): set_seed(42 ) __lowerCamelCase = RegressionModel() __lowerCamelCase = deepcopy(_UpperCamelCase ) __lowerCamelCase = RegressionDataset(length=_UpperCamelCase ) __lowerCamelCase = DataLoader(_UpperCamelCase ,batch_size=_UpperCamelCase ) model.to(accelerator.device ) __lowerCamelCase ,__lowerCamelCase = accelerator.prepare(_UpperCamelCase ,_UpperCamelCase ) return model, ddp_model, dataloader def a__ ( _UpperCamelCase : Accelerator ,_UpperCamelCase : str=False ): __lowerCamelCase = AutoTokenizer.from_pretrained('''hf-internal-testing/mrpc-bert-base-cased''' ) __lowerCamelCase = load_dataset('''glue''' ,'''mrpc''' ,split='''validation''' ) def tokenize_function(_UpperCamelCase : int ): __lowerCamelCase = tokenizer(examples['''sentence1'''] ,examples['''sentence2'''] ,truncation=_UpperCamelCase ,max_length=_UpperCamelCase ) return outputs with accelerator.main_process_first(): __lowerCamelCase = dataset.map( _UpperCamelCase ,batched=_UpperCamelCase ,remove_columns=['''idx''', '''sentence1''', '''sentence2'''] ,) __lowerCamelCase = tokenized_datasets.rename_column('''label''' ,'''labels''' ) def collate_fn(_UpperCamelCase : Any ): if use_longest: return tokenizer.pad(_UpperCamelCase ,padding='''longest''' ,return_tensors='''pt''' ) return tokenizer.pad(_UpperCamelCase ,padding='''max_length''' ,max_length=1_28 ,return_tensors='''pt''' ) return DataLoader(_UpperCamelCase ,shuffle=_UpperCamelCase ,collate_fn=_UpperCamelCase ,batch_size=16 ) def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : List[str] ): __lowerCamelCase = Accelerator(dispatch_batches=_UpperCamelCase ,split_batches=_UpperCamelCase ) __lowerCamelCase = get_dataloader(_UpperCamelCase ,not dispatch_batches ) __lowerCamelCase = AutoModelForSequenceClassification.from_pretrained( '''hf-internal-testing/mrpc-bert-base-cased''' ,return_dict=_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = accelerator.prepare(_UpperCamelCase ,_UpperCamelCase ) return {"ddp": [ddp_model, ddp_dataloader, "cuda:0"], "no": [model, dataloader, accelerator.device]}, accelerator def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : Optional[Any] ,_UpperCamelCase : Union[str, Any] ): __lowerCamelCase = [] for batch in dataloader: __lowerCamelCase ,__lowerCamelCase = batch.values() with torch.no_grad(): __lowerCamelCase = model(_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = accelerator.gather_for_metrics((logit, target) ) logits_and_targets.append((logit, target) ) __lowerCamelCase ,__lowerCamelCase = [], [] for logit, targ in logits_and_targets: logits.append(_UpperCamelCase ) targs.append(_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = torch.cat(_UpperCamelCase ), torch.cat(_UpperCamelCase ) return logits, targs def a__ ( _UpperCamelCase : Accelerator ,_UpperCamelCase : List[Any]=82 ,_UpperCamelCase : str=False ,_UpperCamelCase : List[str]=False ,_UpperCamelCase : Optional[int]=16 ): __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = get_basic_setup(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = generate_predictions(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ) assert ( len(_UpperCamelCase ) == num_samples ), F"""Unexpected number of inputs:\n Expected: {num_samples}\n Actual: {len(_UpperCamelCase )}""" def a__ ( _UpperCamelCase : bool = False ,_UpperCamelCase : bool = False ): __lowerCamelCase = evaluate.load('''glue''' ,'''mrpc''' ) __lowerCamelCase ,__lowerCamelCase = get_mrpc_setup(_UpperCamelCase ,_UpperCamelCase ) # First do baseline __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = setup['''no'''] model.to(_UpperCamelCase ) model.eval() for batch in dataloader: batch.to(_UpperCamelCase ) with torch.inference_mode(): __lowerCamelCase = model(**_UpperCamelCase ) __lowerCamelCase = outputs.logits.argmax(dim=-1 ) metric.add_batch(predictions=_UpperCamelCase ,references=batch['''labels'''] ) __lowerCamelCase = metric.compute() # Then do distributed __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = setup['''ddp'''] model.eval() for batch in dataloader: with torch.inference_mode(): __lowerCamelCase = model(**_UpperCamelCase ) __lowerCamelCase = outputs.logits.argmax(dim=-1 ) __lowerCamelCase = batch['''labels'''] __lowerCamelCase ,__lowerCamelCase = accelerator.gather_for_metrics((preds, references) ) metric.add_batch(predictions=_UpperCamelCase ,references=_UpperCamelCase ) __lowerCamelCase = metric.compute() for key in "accuracy f1".split(): assert math.isclose( baseline[key] ,distributed[key] ), F"""Baseline and Distributed are not the same for key {key}:\n\tBaseline: {baseline[key]}\n\tDistributed: {distributed[key]}\n""" def a__ ( ): __lowerCamelCase = Accelerator(split_batches=_UpperCamelCase ,dispatch_batches=_UpperCamelCase ) if accelerator.is_local_main_process: datasets.utils.logging.set_verbosity_warning() transformers.utils.logging.set_verbosity_warning() else: datasets.utils.logging.set_verbosity_error() transformers.utils.logging.set_verbosity_error() # These are a bit slower so they should only be ran on the GPU or TPU if torch.cuda.is_available() or is_tpu_available(): if accelerator.is_local_main_process: print('''**Testing gather_for_metrics**''' ) for split_batches in [True, False]: for dispatch_batches in [True, False]: if accelerator.is_local_main_process: print(F"""With: `split_batches={split_batches}`, `dispatch_batches={dispatch_batches}`""" ) test_mrpc(_UpperCamelCase ,_UpperCamelCase ) accelerator.state._reset_state() if accelerator.is_local_main_process: print('''**Test torch metrics**''' ) for split_batches in [True, False]: for dispatch_batches in [True, False]: __lowerCamelCase = Accelerator(split_batches=_UpperCamelCase ,dispatch_batches=_UpperCamelCase ) if accelerator.is_local_main_process: print(F"""With: `split_batches={split_batches}`, `dispatch_batches={dispatch_batches}`, length=99""" ) test_torch_metrics(_UpperCamelCase ,99 ) accelerator.state._reset_state() if accelerator.is_local_main_process: print('''**Test last batch is not dropped when perfectly divisible**''' ) __lowerCamelCase = Accelerator() test_torch_metrics(_UpperCamelCase ,5_12 ) accelerator.state._reset_state() def a__ ( _UpperCamelCase : Optional[int] ): # For xla_spawn (TPUs) main() if __name__ == "__main__": main()
622
1
from typing import List, Optional, Union from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { """huggingface/informer-tourism-monthly""": ( """https://huggingface.co/huggingface/informer-tourism-monthly/resolve/main/config.json""" ), # See all Informer models at https://huggingface.co/models?filter=informer } class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = """informer""" lowerCAmelCase__ = { """hidden_size""": """d_model""", """num_attention_heads""": """encoder_attention_heads""", """num_hidden_layers""": """encoder_layers""", } def __init__( self , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = "student_t" , __UpperCAmelCase = "nll" , __UpperCAmelCase = 1 , __UpperCAmelCase = None , __UpperCAmelCase = "mean" , __UpperCAmelCase = 0 , __UpperCAmelCase = 0 , __UpperCAmelCase = 0 , __UpperCAmelCase = 0 , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = 64 , __UpperCAmelCase = 32 , __UpperCAmelCase = 32 , __UpperCAmelCase = 2 , __UpperCAmelCase = 2 , __UpperCAmelCase = 2 , __UpperCAmelCase = 2 , __UpperCAmelCase = True , __UpperCAmelCase = "gelu" , __UpperCAmelCase = 0.05 , __UpperCAmelCase = 0.1 , __UpperCAmelCase = 0.1 , __UpperCAmelCase = 0.1 , __UpperCAmelCase = 0.1 , __UpperCAmelCase = 100 , __UpperCAmelCase = 0.02 , __UpperCAmelCase=True , __UpperCAmelCase = "prob" , __UpperCAmelCase = 5 , __UpperCAmelCase = True , **__UpperCAmelCase , ): '''simple docstring''' # time series specific configuration __lowerCamelCase = prediction_length __lowerCamelCase = context_length or prediction_length __lowerCamelCase = distribution_output __lowerCamelCase = loss __lowerCamelCase = input_size __lowerCamelCase = num_time_features __lowerCamelCase = lags_sequence if lags_sequence is not None else [1, 2, 3, 4, 5, 6, 7] __lowerCamelCase = scaling __lowerCamelCase = num_dynamic_real_features __lowerCamelCase = num_static_real_features __lowerCamelCase = num_static_categorical_features # set cardinality if cardinality and num_static_categorical_features > 0: if len(__UpperCAmelCase ) != num_static_categorical_features: raise ValueError( '''The cardinality should be a list of the same length as `num_static_categorical_features`''' ) __lowerCamelCase = cardinality else: __lowerCamelCase = [0] # set embedding_dimension if embedding_dimension and num_static_categorical_features > 0: if len(__UpperCAmelCase ) != num_static_categorical_features: raise ValueError( '''The embedding dimension should be a list of the same length as `num_static_categorical_features`''' ) __lowerCamelCase = embedding_dimension else: __lowerCamelCase = [min(50 , (cat + 1) // 2 ) for cat in self.cardinality] __lowerCamelCase = num_parallel_samples # Transformer architecture configuration __lowerCamelCase = input_size * len(self.lags_sequence ) + self._number_of_features __lowerCamelCase = d_model __lowerCamelCase = encoder_attention_heads __lowerCamelCase = decoder_attention_heads __lowerCamelCase = encoder_ffn_dim __lowerCamelCase = decoder_ffn_dim __lowerCamelCase = encoder_layers __lowerCamelCase = decoder_layers __lowerCamelCase = dropout __lowerCamelCase = attention_dropout __lowerCamelCase = activation_dropout __lowerCamelCase = encoder_layerdrop __lowerCamelCase = decoder_layerdrop __lowerCamelCase = activation_function __lowerCamelCase = init_std __lowerCamelCase = use_cache # Informer __lowerCamelCase = attention_type __lowerCamelCase = sampling_factor __lowerCamelCase = distil super().__init__(is_encoder_decoder=__UpperCAmelCase , **__UpperCAmelCase ) @property def lowerCamelCase ( self ): '''simple docstring''' return ( sum(self.embedding_dimension ) + self.num_dynamic_real_features + self.num_time_features + self.num_static_real_features + self.input_size * 2 # the log1p(abs(loc)) and log(scale) features )
622
import gc import random import unittest import numpy as np import torch from transformers import CLIPTextConfig, CLIPTextModel, CLIPTextModelWithProjection, CLIPTokenizer from diffusers import ( AutoencoderKL, DiffusionPipeline, EulerDiscreteScheduler, StableDiffusionXLImgaImgPipeline, UNetaDConditionModel, ) from diffusers.utils import floats_tensor, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..pipeline_params import ( IMAGE_TO_IMAGE_IMAGE_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS, ) from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin enable_full_determinism() class __lowerCAmelCase ( lowerCAmelCase__ , lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = StableDiffusionXLImgaImgPipeline lowerCAmelCase__ = TEXT_GUIDED_IMAGE_VARIATION_PARAMS - {"""height""", """width"""} lowerCAmelCase__ = PipelineTesterMixin.required_optional_params - {"""latents"""} lowerCAmelCase__ = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS lowerCAmelCase__ = IMAGE_TO_IMAGE_IMAGE_PARAMS lowerCAmelCase__ = IMAGE_TO_IMAGE_IMAGE_PARAMS def lowerCamelCase ( self ): '''simple docstring''' torch.manual_seed(0 ) __lowerCamelCase = 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''') , attention_head_dim=(2, 4) , use_linear_projection=__UpperCAmelCase , addition_embed_type='''text_time''' , addition_time_embed_dim=8 , transformer_layers_per_block=(1, 2) , projection_class_embeddings_input_dim=80 , cross_attention_dim=64 , ) __lowerCamelCase = EulerDiscreteScheduler( beta_start=0.00_085 , beta_end=0.012 , steps_offset=1 , beta_schedule='''scaled_linear''' , timestep_spacing='''leading''' , ) torch.manual_seed(0 ) __lowerCamelCase = AutoencoderKL( block_out_channels=[32, 64] , in_channels=3 , out_channels=3 , down_block_types=['''DownEncoderBlock2D''', '''DownEncoderBlock2D'''] , up_block_types=['''UpDecoderBlock2D''', '''UpDecoderBlock2D'''] , latent_channels=4 , sample_size=128 , ) torch.manual_seed(0 ) __lowerCamelCase = CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=32 , intermediate_size=37 , layer_norm_eps=1E-0_5 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , hidden_act='''gelu''' , projection_dim=32 , ) __lowerCamelCase = CLIPTextModel(__UpperCAmelCase ) __lowerCamelCase = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' , local_files_only=__UpperCAmelCase ) __lowerCamelCase = CLIPTextModelWithProjection(__UpperCAmelCase ) __lowerCamelCase = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' , local_files_only=__UpperCAmelCase ) __lowerCamelCase = { '''unet''': unet, '''scheduler''': scheduler, '''vae''': vae, '''text_encoder''': text_encoder, '''tokenizer''': tokenizer, '''text_encoder_2''': text_encoder_a, '''tokenizer_2''': tokenizer_a, # "safety_checker": None, # "feature_extractor": None, } return components def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase=0 ): '''simple docstring''' __lowerCamelCase = floats_tensor((1, 3, 32, 32) , rng=random.Random(__UpperCAmelCase ) ).to(__UpperCAmelCase ) __lowerCamelCase = image / 2 + 0.5 if str(__UpperCAmelCase ).startswith('''mps''' ): __lowerCamelCase = torch.manual_seed(__UpperCAmelCase ) else: __lowerCamelCase = torch.Generator(device=__UpperCAmelCase ).manual_seed(__UpperCAmelCase ) __lowerCamelCase = { '''prompt''': '''A painting of a squirrel eating a burger''', '''image''': image, '''generator''': generator, '''num_inference_steps''': 2, '''guidance_scale''': 5.0, '''output_type''': '''numpy''', '''strength''': 0.75, } return inputs def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''cpu''' # ensure determinism for the device-dependent torch.Generator __lowerCamelCase = self.get_dummy_components() __lowerCamelCase = StableDiffusionXLImgaImgPipeline(**__UpperCAmelCase ) __lowerCamelCase = sd_pipe.to(__UpperCAmelCase ) sd_pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = self.get_dummy_inputs(__UpperCAmelCase ) __lowerCamelCase = sd_pipe(**__UpperCAmelCase ).images __lowerCamelCase = image[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) __lowerCamelCase = np.array([0.4_656, 0.4_840, 0.4_439, 0.6_698, 0.5_574, 0.4_524, 0.5_799, 0.5_943, 0.5_165] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-2 def lowerCamelCase ( self ): '''simple docstring''' super().test_attention_slicing_forward_pass(expected_max_diff=3E-3 ) def lowerCamelCase ( self ): '''simple docstring''' super().test_inference_batch_single_identical(expected_max_diff=3E-3 ) def lowerCamelCase ( self ): '''simple docstring''' pass def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_components() __lowerCamelCase = StableDiffusionXLImgaImgPipeline(**__UpperCAmelCase ) __lowerCamelCase = sd_pipe.to(__UpperCAmelCase ) __lowerCamelCase = sd_pipe.to(__UpperCAmelCase ) sd_pipe.set_progress_bar_config(disable=__UpperCAmelCase ) # forward without prompt embeds __lowerCamelCase = self.get_dummy_inputs(__UpperCAmelCase ) __lowerCamelCase = 3 * ['''this is a negative prompt'''] __lowerCamelCase = negative_prompt __lowerCamelCase = 3 * [inputs['''prompt''']] __lowerCamelCase = sd_pipe(**__UpperCAmelCase ) __lowerCamelCase = output.images[0, -3:, -3:, -1] # forward with prompt embeds __lowerCamelCase = self.get_dummy_inputs(__UpperCAmelCase ) __lowerCamelCase = 3 * ['''this is a negative prompt'''] __lowerCamelCase = 3 * [inputs.pop('''prompt''' )] ( ( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) , ) = sd_pipe.encode_prompt(__UpperCAmelCase , negative_prompt=__UpperCAmelCase ) __lowerCamelCase = sd_pipe( **__UpperCAmelCase , prompt_embeds=__UpperCAmelCase , negative_prompt_embeds=__UpperCAmelCase , pooled_prompt_embeds=__UpperCAmelCase , negative_pooled_prompt_embeds=__UpperCAmelCase , ) __lowerCamelCase = output.images[0, -3:, -3:, -1] # make sure that it's equal assert np.abs(image_slice_a.flatten() - image_slice_a.flatten() ).max() < 1E-4 @slow @require_torch_gpu class __lowerCAmelCase ( unittest.TestCase ): def lowerCamelCase ( self ): '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase="cpu" , __UpperCAmelCase=torch.floataa , __UpperCAmelCase=0 ): '''simple docstring''' __lowerCamelCase = torch.Generator(device=__UpperCAmelCase ).manual_seed(__UpperCAmelCase ) __lowerCamelCase = np.random.RandomState(__UpperCAmelCase ).standard_normal((1, 4, 64, 64) ) __lowerCamelCase = torch.from_numpy(__UpperCAmelCase ).to(device=__UpperCAmelCase , dtype=__UpperCAmelCase ) __lowerCamelCase = { '''prompt''': '''a photograph of an astronaut riding a horse''', '''latents''': latents, '''generator''': generator, '''num_inference_steps''': 3, '''guidance_scale''': 7.5, '''output_type''': '''numpy''', } return inputs def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = DiffusionPipeline.from_pretrained('''stabilityai/stable-diffusion-2-base''' ) pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = self.get_inputs(__UpperCAmelCase ) __lowerCamelCase = pipe(**__UpperCAmelCase ).images __lowerCamelCase = image[0, -3:, -3:, -1].flatten() assert image.shape == (1, 512, 512, 3) __lowerCamelCase = np.array([0.49_493, 0.47_896, 0.40_798, 0.54_214, 0.53_212, 0.48_202, 0.47_656, 0.46_329, 0.48_506] ) assert np.abs(image_slice - expected_slice ).max() < 7E-3
622
1
import json import os import unittest from transformers.models.ctrl.tokenization_ctrl import VOCAB_FILES_NAMES, CTRLTokenizer from ...test_tokenization_common import TokenizerTesterMixin class __lowerCAmelCase ( lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = CTRLTokenizer lowerCAmelCase__ = False lowerCAmelCase__ = False def lowerCamelCase ( self ): '''simple docstring''' super().setUp() # Adapted from Sennrich et al. 2015 and https://github.com/rsennrich/subword-nmt __lowerCamelCase = ['''adapt''', '''re@@''', '''a@@''', '''apt''', '''c@@''', '''t''', '''<unk>'''] __lowerCamelCase = dict(zip(__UpperCAmelCase , range(len(__UpperCAmelCase ) ) ) ) __lowerCamelCase = ['''#version: 0.2''', '''a p''', '''ap t</w>''', '''r e''', '''a d''', '''ad apt</w>''', ''''''] __lowerCamelCase = {'''unk_token''': '''<unk>'''} __lowerCamelCase = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES['''vocab_file'''] ) __lowerCamelCase = 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(__UpperCAmelCase ) + '''\n''' ) with open(self.merges_file , '''w''' , encoding='''utf-8''' ) as fp: fp.write('''\n'''.join(__UpperCAmelCase ) ) def lowerCamelCase ( self , **__UpperCAmelCase ): '''simple docstring''' kwargs.update(self.special_tokens_map ) return CTRLTokenizer.from_pretrained(self.tmpdirname , **__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = '''adapt react readapt apt''' __lowerCamelCase = '''adapt react readapt apt''' return input_text, output_text def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = CTRLTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map ) __lowerCamelCase = '''adapt react readapt apt''' __lowerCamelCase = '''adapt re@@ a@@ c@@ t re@@ adapt apt'''.split() __lowerCamelCase = tokenizer.tokenize(__UpperCAmelCase ) self.assertListEqual(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = tokens + [tokenizer.unk_token] __lowerCamelCase = [0, 1, 2, 4, 5, 1, 0, 3, 6] self.assertListEqual(tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) , __UpperCAmelCase )
622
import torch from diffusers import StableDiffusionPipeline a_ = """path-to-your-trained-model""" a_ = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.floataa).to("""cuda""") a_ = """A photo of sks dog in a bucket""" a_ = pipe(prompt, num_inference_steps=50, guidance_scale=7.5).images[0] image.save("""dog-bucket.png""")
622
1
import sys from collections import defaultdict class __lowerCAmelCase : def __init__( self ): '''simple docstring''' __lowerCamelCase = [] def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.node_position[vertex] def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = pos def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' if start > size // 2 - 1: return else: if 2 * start + 2 >= size: __lowerCamelCase = 2 * start + 1 else: if heap[2 * start + 1] < heap[2 * start + 2]: __lowerCamelCase = 2 * start + 1 else: __lowerCamelCase = 2 * start + 2 if heap[smallest_child] < heap[start]: __lowerCamelCase ,__lowerCamelCase = heap[smallest_child], positions[smallest_child] __lowerCamelCase ,__lowerCamelCase = ( heap[start], positions[start], ) __lowerCamelCase ,__lowerCamelCase = temp, tempa __lowerCamelCase = self.get_position(positions[smallest_child] ) self.set_position( positions[smallest_child] , self.get_position(positions[start] ) ) self.set_position(positions[start] , __UpperCAmelCase ) self.top_to_bottom(__UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = position[index] while index != 0: __lowerCamelCase = int((index - 2) / 2 ) if index % 2 == 0 else int((index - 1) / 2 ) if val < heap[parent]: __lowerCamelCase = heap[parent] __lowerCamelCase = position[parent] self.set_position(position[parent] , __UpperCAmelCase ) else: __lowerCamelCase = val __lowerCamelCase = temp self.set_position(__UpperCAmelCase , __UpperCAmelCase ) break __lowerCamelCase = parent else: __lowerCamelCase = val __lowerCamelCase = temp self.set_position(__UpperCAmelCase , 0 ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = len(__UpperCAmelCase ) // 2 - 1 for i in range(__UpperCAmelCase , -1 , -1 ): self.top_to_bottom(__UpperCAmelCase , __UpperCAmelCase , len(__UpperCAmelCase ) , __UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = positions[0] __lowerCamelCase = sys.maxsize self.top_to_bottom(__UpperCAmelCase , 0 , len(__UpperCAmelCase ) , __UpperCAmelCase ) return temp def a__ ( _UpperCamelCase : str ): __lowerCamelCase = Heap() __lowerCamelCase = [0] * len(_UpperCamelCase ) __lowerCamelCase = [-1] * len(_UpperCamelCase ) # Neighboring Tree Vertex of selected vertex # Minimum Distance of explored vertex with neighboring vertex of partial tree # formed in graph __lowerCamelCase = [] # Heap of Distance of vertices from their neighboring vertex __lowerCamelCase = [] for vertex in range(len(_UpperCamelCase ) ): distance_tv.append(sys.maxsize ) positions.append(_UpperCamelCase ) heap.node_position.append(_UpperCamelCase ) __lowerCamelCase = [] __lowerCamelCase = 1 __lowerCamelCase = sys.maxsize for neighbor, distance in adjacency_list[0]: __lowerCamelCase = 0 __lowerCamelCase = distance heap.heapify(_UpperCamelCase ,_UpperCamelCase ) for _ in range(1 ,len(_UpperCamelCase ) ): __lowerCamelCase = heap.delete_minimum(_UpperCamelCase ,_UpperCamelCase ) if visited[vertex] == 0: tree_edges.append((nbr_tv[vertex], vertex) ) __lowerCamelCase = 1 for neighbor, distance in adjacency_list[vertex]: if ( visited[neighbor] == 0 and distance < distance_tv[heap.get_position(_UpperCamelCase )] ): __lowerCamelCase = distance heap.bottom_to_top( _UpperCamelCase ,heap.get_position(_UpperCamelCase ) ,_UpperCamelCase ,_UpperCamelCase ) __lowerCamelCase = vertex return tree_edges if __name__ == "__main__": # pragma: no cover # < --------- Prims Algorithm --------- > a_ = int(input("""Enter number of edges: """).strip()) a_ = defaultdict(list) for _ in range(edges_number): a_ = [int(x) for x in input().strip().split()] adjacency_list[edge[0]].append([edge[1], edge[2]]) adjacency_list[edge[1]].append([edge[0], edge[2]]) print(prisms_algorithm(adjacency_list))
622
import unittest from transformers import MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING, AutoTokenizer, is_vision_available from transformers.pipelines import pipeline from transformers.pipelines.document_question_answering import apply_tesseract from transformers.testing_utils import ( is_pipeline_test, nested_simplify, require_detectrona, require_pytesseract, require_tf, require_torch, require_vision, slow, ) from .test_pipelines_common import ANY if is_vision_available(): from PIL import Image from transformers.image_utils import load_image else: class __lowerCAmelCase : @staticmethod def lowerCamelCase ( *__UpperCAmelCase , **__UpperCAmelCase ): '''simple docstring''' pass def a__ ( _UpperCamelCase : List[str] ): return None # This is a pinned image from a specific revision of a document question answering space, hosted by HuggingFace, # so we can expect it to be available. a_ = ( """https://huggingface.co/spaces/impira/docquery/resolve/2f6c96314dc84dfda62d40de9da55f2f5165d403/invoice.png""" ) @is_pipeline_test @require_torch @require_vision class __lowerCAmelCase ( unittest.TestCase ): lowerCAmelCase__ = MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING @require_pytesseract @require_vision def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model=__UpperCAmelCase , tokenizer=__UpperCAmelCase , image_processor=__UpperCAmelCase ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = list(zip(*apply_tesseract(load_image(__UpperCAmelCase ) , __UpperCAmelCase , '''''' ) ) ) __lowerCamelCase = '''What is the placebo?''' __lowerCamelCase = [ { '''image''': load_image(__UpperCAmelCase ), '''question''': question, }, { '''image''': image, '''question''': question, }, { '''image''': image, '''question''': question, '''word_boxes''': word_boxes, }, ] return dqa_pipeline, examples def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = dqa_pipeline(__UpperCAmelCase , top_k=2 ) self.assertEqual( __UpperCAmelCase , [ [ {'''score''': ANY(__UpperCAmelCase ), '''answer''': ANY(__UpperCAmelCase ), '''start''': ANY(__UpperCAmelCase ), '''end''': ANY(__UpperCAmelCase )}, {'''score''': ANY(__UpperCAmelCase ), '''answer''': ANY(__UpperCAmelCase ), '''start''': ANY(__UpperCAmelCase ), '''end''': ANY(__UpperCAmelCase )}, ] ] * 3 , ) @require_torch @require_detectrona @require_pytesseract def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline('''document-question-answering''' , model='''hf-internal-testing/tiny-random-layoutlmv2''' ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''How many cats are there?''' __lowerCamelCase = [ {'''score''': 0.0_001, '''answer''': '''oy 2312/2019''', '''start''': 38, '''end''': 39}, {'''score''': 0.0_001, '''answer''': '''oy 2312/2019 DUE''', '''start''': 38, '''end''': 40}, ] __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual(nested_simplify(__UpperCAmelCase , decimals=4 ) , __UpperCAmelCase ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual(nested_simplify(__UpperCAmelCase , decimals=4 ) , __UpperCAmelCase ) # This image does not detect ANY text in it, meaning layoutlmv2 should fail. # Empty answer probably __lowerCamelCase = '''./tests/fixtures/tests_samples/COCO/000000039769.png''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual(__UpperCAmelCase , [] ) # We can optionnally pass directly the words and bounding boxes __lowerCamelCase = '''./tests/fixtures/tests_samples/COCO/000000039769.png''' __lowerCamelCase = [] __lowerCamelCase = [] __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , words=__UpperCAmelCase , boxes=__UpperCAmelCase , top_k=2 ) self.assertEqual(__UpperCAmelCase , [] ) @slow @require_torch @require_detectrona @require_pytesseract def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model='''tiennvcs/layoutlmv2-base-uncased-finetuned-docvqa''' , revision='''9977165''' , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_944, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_009, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_944, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_009, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.9_944, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_009, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ], ] * 2 , ) @slow @require_torch @require_detectrona @require_pytesseract def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model='''tiennvcs/layoutlmv2-base-uncased-finetuned-docvqa''' , revision='''9977165''' , max_seq_len=50 , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_974, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, {'''score''': 0.9_948, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_974, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, {'''score''': 0.9_948, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.9_974, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, {'''score''': 0.9_948, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] ] * 2 , ) @slow @require_torch @require_pytesseract @require_vision def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = AutoTokenizer.from_pretrained( '''impira/layoutlm-document-qa''' , revision='''3dc6de3''' , add_prefix_space=__UpperCAmelCase ) __lowerCamelCase = pipeline( '''document-question-answering''' , model='''impira/layoutlm-document-qa''' , tokenizer=__UpperCAmelCase , revision='''3dc6de3''' , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] , ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] ] * 2 , ) __lowerCamelCase = list(zip(*apply_tesseract(load_image(__UpperCAmelCase ) , __UpperCAmelCase , '''''' ) ) ) # This model should also work if `image` is set to None __lowerCamelCase = dqa_pipeline({'''image''': None, '''word_boxes''': word_boxes, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] , ) @slow @require_torch @require_pytesseract @require_vision def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = AutoTokenizer.from_pretrained( '''impira/layoutlm-document-qa''' , revision='''3dc6de3''' , add_prefix_space=__UpperCAmelCase ) __lowerCamelCase = pipeline( '''document-question-answering''' , model='''impira/layoutlm-document-qa''' , tokenizer=__UpperCAmelCase , revision='''3dc6de3''' , max_seq_len=50 , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_999, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.9_998, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.9_999, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.9_998, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] ] * 2 , ) __lowerCamelCase = list(zip(*apply_tesseract(load_image(__UpperCAmelCase ) , __UpperCAmelCase , '''''' ) ) ) # This model should also work if `image` is set to None __lowerCamelCase = dqa_pipeline({'''image''': None, '''word_boxes''': word_boxes, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_999, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.9_998, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) @slow @require_torch def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model='''naver-clova-ix/donut-base-finetuned-docvqa''' , tokenizer=AutoTokenizer.from_pretrained('''naver-clova-ix/donut-base-finetuned-docvqa''' ) , feature_extractor='''naver-clova-ix/donut-base-finetuned-docvqa''' , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual(nested_simplify(__UpperCAmelCase , decimals=4 ) , [{'''answer''': '''us-001'''}] ) @require_tf @unittest.skip('''Document question answering not implemented in TF''' ) def lowerCamelCase ( self ): '''simple docstring''' pass
622
1
def a__ ( _UpperCamelCase : str ,_UpperCamelCase : str = " " ): __lowerCamelCase = [] __lowerCamelCase = 0 for index, char in enumerate(_UpperCamelCase ): if char == separator: split_words.append(string[last_index:index] ) __lowerCamelCase = index + 1 elif index + 1 == len(_UpperCamelCase ): split_words.append(string[last_index : index + 1] ) return split_words if __name__ == "__main__": from doctest import testmod testmod()
622
import unittest from transformers.models.xlm_prophetnet.tokenization_xlm_prophetnet import SPIECE_UNDERLINE, XLMProphetNetTokenizer from transformers.testing_utils import get_tests_dir, require_sentencepiece, slow from transformers.utils import cached_property from ...test_tokenization_common import TokenizerTesterMixin a_ = get_tests_dir("""fixtures/test_sentencepiece.model""") @require_sentencepiece class __lowerCAmelCase ( lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = XLMProphetNetTokenizer lowerCAmelCase__ = False lowerCAmelCase__ = True def lowerCamelCase ( self ): '''simple docstring''' super().setUp() # We have a SentencePiece fixture for testing __lowerCamelCase = XLMProphetNetTokenizer(__UpperCAmelCase , keep_accents=__UpperCAmelCase ) tokenizer.save_pretrained(self.tmpdirname ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''[PAD]''' __lowerCamelCase = 0 self.assertEqual(self.get_tokenizer()._convert_token_to_id(__UpperCAmelCase ) , __UpperCAmelCase ) self.assertEqual(self.get_tokenizer()._convert_id_to_token(__UpperCAmelCase ) , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = list(self.get_tokenizer().get_vocab().keys() ) self.assertEqual(vocab_keys[0] , '''[PAD]''' ) self.assertEqual(vocab_keys[1] , '''[CLS]''' ) self.assertEqual(vocab_keys[-1] , '''j''' ) self.assertEqual(len(__UpperCAmelCase ) , 1012 ) def lowerCamelCase ( self ): '''simple docstring''' self.assertEqual(self.get_tokenizer().vocab_size , 1012 ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = XLMProphetNetTokenizer(__UpperCAmelCase , keep_accents=__UpperCAmelCase ) __lowerCamelCase = tokenizer.tokenize('''This is a test''' ) self.assertListEqual(__UpperCAmelCase , ['''▁This''', '''▁is''', '''▁a''', '''▁t''', '''est'''] ) self.assertListEqual( tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , ) __lowerCamelCase = tokenizer.tokenize('''I was born in 92000, and this is falsé.''' ) self.assertListEqual( __UpperCAmelCase , [ SPIECE_UNDERLINE + '''I''', SPIECE_UNDERLINE + '''was''', SPIECE_UNDERLINE + '''b''', '''or''', '''n''', SPIECE_UNDERLINE + '''in''', SPIECE_UNDERLINE + '''''', '''9''', '''2''', '''0''', '''0''', '''0''', ''',''', SPIECE_UNDERLINE + '''and''', SPIECE_UNDERLINE + '''this''', SPIECE_UNDERLINE + '''is''', SPIECE_UNDERLINE + '''f''', '''al''', '''s''', '''é''', '''.''', ] , ) __lowerCamelCase = tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) self.assertListEqual( __UpperCAmelCase , [ value + tokenizer.fairseq_offset for value in [8, 21, 84, 55, 24, 19, 7, -9, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, -9, 4] ] , ) __lowerCamelCase = tokenizer.convert_ids_to_tokens(__UpperCAmelCase ) self.assertListEqual( __UpperCAmelCase , [ SPIECE_UNDERLINE + '''I''', SPIECE_UNDERLINE + '''was''', SPIECE_UNDERLINE + '''b''', '''or''', '''n''', SPIECE_UNDERLINE + '''in''', SPIECE_UNDERLINE + '''''', '''[UNK]''', '''2''', '''0''', '''0''', '''0''', ''',''', SPIECE_UNDERLINE + '''and''', SPIECE_UNDERLINE + '''this''', SPIECE_UNDERLINE + '''is''', SPIECE_UNDERLINE + '''f''', '''al''', '''s''', '''[UNK]''', '''.''', ] , ) @cached_property def lowerCamelCase ( self ): '''simple docstring''' return XLMProphetNetTokenizer.from_pretrained('''microsoft/xprophetnet-large-wiki100-cased''' ) @slow def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''Hello World!''' __lowerCamelCase = [35389, 6672, 49, 2] self.assertListEqual(__UpperCAmelCase , self.big_tokenizer.encode(__UpperCAmelCase ) ) @slow def lowerCamelCase ( self ): '''simple docstring''' # fmt: off __lowerCamelCase = {'''input_ids''': [[11073, 82783, 18, 26, 82783, 549, 51540, 248, 17209, 1301, 217, 20, 215186, 1325, 147, 17209, 1301, 217, 20, 56370, 53, 122020, 20, 16477, 27, 87355, 4548, 20, 4728, 78392, 17, 159969, 18, 26, 24491, 629, 15, 538, 22704, 5439, 15, 2788, 24491, 9885, 15, 43534, 605, 15, 814, 18403, 33200, 29, 15, 43534, 24458, 12410, 111, 24966, 83669, 9637, 144068, 26, 850, 22346, 27, 147, 24966, 83669, 83490, 26, 39113, 735, 27, 689, 656, 2800, 1339, 4600, 53, 122020, 115785, 34, 816, 1339, 46887, 18, 147, 53905, 1951, 42238, 41170, 17732, 834, 436, 15, 27523, 98733, 217, 147, 5542, 4981, 930, 17347, 16, 2], [20091, 629, 94, 82786, 58, 490, 20, 1528, 84, 53905, 344, 80592, 110128, 18822, 5267, 1306, 62, 152537, 308, 7997, 401, 124427, 549, 35442, 225, 109, 15055, 25748, 147, 7119, 43712, 34, 767, 135366, 18, 16, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [592, 63784, 119466, 17, 147808, 88214, 18, 656, 81, 32, 3296, 10280, 16, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], '''attention_mask''': [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]} # noqa: E501 # fmt: on self.tokenizer_integration_test_util( expected_encoding=__UpperCAmelCase , model_name='''microsoft/xprophetnet-large-wiki100-cased''' , revision='''1acad1643ddd54a44df6a1b797ada8373685d90e''' , )
622
1
from pathlib import Path import fire def a__ ( _UpperCamelCase : str ,_UpperCamelCase : str ,_UpperCamelCase : int ): __lowerCamelCase = Path(_UpperCamelCase ) __lowerCamelCase = Path(_UpperCamelCase ) dest_dir.mkdir(exist_ok=_UpperCamelCase ) for path in src_dir.iterdir(): __lowerCamelCase = [x.rstrip() for x in list(path.open().readlines() )][:n] __lowerCamelCase = dest_dir.joinpath(path.name ) print(_UpperCamelCase ) dest_path.open('''w''' ).write('''\n'''.join(_UpperCamelCase ) ) if __name__ == "__main__": fire.Fire(minify)
622
import inspect import re 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_config_docstrings.py a_ = """src/transformers""" # This is to make sure the transformers module imported is the one in the repo. a_ = direct_transformers_import(PATH_TO_TRANSFORMERS) a_ = transformers.models.auto.configuration_auto.CONFIG_MAPPING # Regex pattern used to find the checkpoint mentioned in the docstring of `config_class`. # For example, `[bert-base-uncased](https://huggingface.co/bert-base-uncased)` a_ = re.compile(R"""\[(.+?)\]\((https://huggingface\.co/.+?)\)""") a_ = { """DecisionTransformerConfig""", """EncoderDecoderConfig""", """MusicgenConfig""", """RagConfig""", """SpeechEncoderDecoderConfig""", """TimmBackboneConfig""", """VisionEncoderDecoderConfig""", """VisionTextDualEncoderConfig""", """LlamaConfig""", } def a__ ( _UpperCamelCase : Union[str, Any] ): __lowerCamelCase = None # source code of `config_class` __lowerCamelCase = inspect.getsource(_UpperCamelCase ) __lowerCamelCase = _re_checkpoint.findall(_UpperCamelCase ) # Each `checkpoint` is a tuple of a checkpoint name and a checkpoint link. # For example, `('bert-base-uncased', 'https://huggingface.co/bert-base-uncased')` for ckpt_name, ckpt_link in checkpoints: # allow the link to end with `/` if ckpt_link.endswith('''/''' ): __lowerCamelCase = ckpt_link[:-1] # verify the checkpoint name corresponds to the checkpoint link __lowerCamelCase = F"""https://huggingface.co/{ckpt_name}""" if ckpt_link == ckpt_link_from_name: __lowerCamelCase = ckpt_name break return checkpoint def a__ ( ): __lowerCamelCase = [] for config_class in list(CONFIG_MAPPING.values() ): # Skip deprecated models if "models.deprecated" in config_class.__module__: continue __lowerCamelCase = get_checkpoint_from_config_class(_UpperCamelCase ) __lowerCamelCase = config_class.__name__ if checkpoint is None and name not in CONFIG_CLASSES_TO_IGNORE_FOR_DOCSTRING_CHECKPOINT_CHECK: configs_without_checkpoint.append(_UpperCamelCase ) if len(_UpperCamelCase ) > 0: __lowerCamelCase = '''\n'''.join(sorted(_UpperCamelCase ) ) raise ValueError(F"""The following configurations don't contain any valid checkpoint:\n{message}""" ) if __name__ == "__main__": check_config_docstrings_have_checkpoints()
622
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_ = logging.get_logger(__name__) # General docstring a_ = """MobileNetV1Config""" # Base docstring a_ = """google/mobilenet_v1_1.0_224""" a_ = [1, 1_024, 7, 7] # Image classification docstring a_ = """google/mobilenet_v1_1.0_224""" a_ = """tabby, tabby cat""" a_ = [ """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 a__ ( _UpperCamelCase : Any ,_UpperCamelCase : Union[str, Any] ,_UpperCamelCase : List[Any]=None ): __lowerCamelCase = {} if isinstance(_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = model.mobilenet_va else: __lowerCamelCase = model __lowerCamelCase = '''MobilenetV1/Conv2d_0/''' __lowerCamelCase = backbone.conv_stem.convolution.weight __lowerCamelCase = backbone.conv_stem.normalization.bias __lowerCamelCase = backbone.conv_stem.normalization.weight __lowerCamelCase = backbone.conv_stem.normalization.running_mean __lowerCamelCase = backbone.conv_stem.normalization.running_var for i in range(13 ): __lowerCamelCase = i + 1 __lowerCamelCase = i * 2 __lowerCamelCase = backbone.layer[pt_index] __lowerCamelCase = F"""MobilenetV1/Conv2d_{tf_index}_depthwise/""" __lowerCamelCase = pointer.convolution.weight __lowerCamelCase = pointer.normalization.bias __lowerCamelCase = pointer.normalization.weight __lowerCamelCase = pointer.normalization.running_mean __lowerCamelCase = pointer.normalization.running_var __lowerCamelCase = backbone.layer[pt_index + 1] __lowerCamelCase = F"""MobilenetV1/Conv2d_{tf_index}_pointwise/""" __lowerCamelCase = pointer.convolution.weight __lowerCamelCase = pointer.normalization.bias __lowerCamelCase = pointer.normalization.weight __lowerCamelCase = pointer.normalization.running_mean __lowerCamelCase = pointer.normalization.running_var if isinstance(_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = '''MobilenetV1/Logits/Conv2d_1c_1x1/''' __lowerCamelCase = model.classifier.weight __lowerCamelCase = model.classifier.bias return tf_to_pt_map def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : Dict ,_UpperCamelCase : Any ): 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 __lowerCamelCase = tf.train.list_variables(_UpperCamelCase ) __lowerCamelCase = {} for name, shape in init_vars: logger.info(F"""Loading TF weight {name} with shape {shape}""" ) __lowerCamelCase = tf.train.load_variable(_UpperCamelCase ,_UpperCamelCase ) __lowerCamelCase = array # Build TF to PyTorch weights loading map __lowerCamelCase = _build_tf_to_pytorch_map(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ) 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 __lowerCamelCase = tf_weights[name] if "depthwise_weights" in name: logger.info('''Transposing depthwise''' ) __lowerCamelCase = np.transpose(_UpperCamelCase ,(2, 3, 0, 1) ) elif "weights" in name: logger.info('''Transposing''' ) if len(pointer.shape ) == 2: # copying into linear layer __lowerCamelCase = array.squeeze().transpose() else: __lowerCamelCase = np.transpose(_UpperCamelCase ,(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}""" ) __lowerCamelCase = torch.from_numpy(_UpperCamelCase ) tf_weights.pop(_UpperCamelCase ,_UpperCamelCase ) tf_weights.pop(name + '''/RMSProp''' ,_UpperCamelCase ) tf_weights.pop(name + '''/RMSProp_1''' ,_UpperCamelCase ) tf_weights.pop(name + '''/ExponentialMovingAverage''' ,_UpperCamelCase ) logger.info(F"""Weights not copied to PyTorch model: {", ".join(tf_weights.keys() )}""" ) return model def a__ ( _UpperCamelCase : torch.Tensor ,_UpperCamelCase : nn.Convad ): __lowerCamelCase ,__lowerCamelCase = features.shape[-2:] __lowerCamelCase ,__lowerCamelCase = conv_layer.stride __lowerCamelCase ,__lowerCamelCase = conv_layer.kernel_size if in_height % stride_height == 0: __lowerCamelCase = max(kernel_height - stride_height ,0 ) else: __lowerCamelCase = max(kernel_height - (in_height % stride_height) ,0 ) if in_width % stride_width == 0: __lowerCamelCase = max(kernel_width - stride_width ,0 ) else: __lowerCamelCase = max(kernel_width - (in_width % stride_width) ,0 ) __lowerCamelCase = pad_along_width // 2 __lowerCamelCase = pad_along_width - pad_left __lowerCamelCase = pad_along_height // 2 __lowerCamelCase = pad_along_height - pad_top __lowerCamelCase = (pad_left, pad_right, pad_top, pad_bottom) return nn.functional.pad(_UpperCamelCase ,_UpperCamelCase ,'''constant''' ,0.0 ) class __lowerCAmelCase ( nn.Module ): def __init__( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase = 1 , __UpperCAmelCase = 1 , __UpperCAmelCase = False , __UpperCAmelCase = True , __UpperCAmelCase = True , ): '''simple docstring''' super().__init__() __lowerCamelCase = 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.""" ) __lowerCamelCase = 0 if config.tf_padding else int((kernel_size - 1) / 2 ) __lowerCamelCase = nn.Convad( in_channels=__UpperCAmelCase , out_channels=__UpperCAmelCase , kernel_size=__UpperCAmelCase , stride=__UpperCAmelCase , padding=__UpperCAmelCase , groups=__UpperCAmelCase , bias=__UpperCAmelCase , padding_mode='''zeros''' , ) if use_normalization: __lowerCamelCase = nn.BatchNormad( num_features=__UpperCAmelCase , eps=config.layer_norm_eps , momentum=0.9_997 , affine=__UpperCAmelCase , track_running_stats=__UpperCAmelCase , ) else: __lowerCamelCase = None if use_activation: if isinstance(__UpperCAmelCase , __UpperCAmelCase ): __lowerCamelCase = ACTaFN[use_activation] elif isinstance(config.hidden_act , __UpperCAmelCase ): __lowerCamelCase = ACTaFN[config.hidden_act] else: __lowerCamelCase = config.hidden_act else: __lowerCamelCase = None def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' if self.config.tf_padding: __lowerCamelCase = apply_tf_padding(__UpperCAmelCase , self.convolution ) __lowerCamelCase = self.convolution(__UpperCAmelCase ) if self.normalization is not None: __lowerCamelCase = self.normalization(__UpperCAmelCase ) if self.activation is not None: __lowerCamelCase = self.activation(__UpperCAmelCase ) return features class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = MobileNetVaConfig lowerCAmelCase__ = load_tf_weights_in_mobilenet_va lowerCAmelCase__ = """mobilenet_v1""" lowerCAmelCase__ = """pixel_values""" lowerCAmelCase__ = False def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' if isinstance(__UpperCAmelCase , (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(__UpperCAmelCase , nn.BatchNormad ): module.bias.data.zero_() module.weight.data.fill_(1.0 ) a_ = 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_ = 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.""" , lowerCAmelCase__ , ) class __lowerCAmelCase ( lowerCAmelCase__ ): def __init__( self , __UpperCAmelCase , __UpperCAmelCase = True ): '''simple docstring''' super().__init__(__UpperCAmelCase ) __lowerCamelCase = config __lowerCamelCase = 32 __lowerCamelCase = max(int(depth * config.depth_multiplier ) , config.min_depth ) __lowerCamelCase = MobileNetVaConvLayer( __UpperCAmelCase , in_channels=config.num_channels , out_channels=__UpperCAmelCase , kernel_size=3 , stride=2 , ) __lowerCamelCase = [1, 2, 1, 2, 1, 2, 1, 1, 1, 1, 1, 2, 1] __lowerCamelCase = nn.ModuleList() for i in range(13 ): __lowerCamelCase = out_channels if strides[i] == 2 or i == 0: depth *= 2 __lowerCamelCase = max(int(depth * config.depth_multiplier ) , config.min_depth ) self.layer.append( MobileNetVaConvLayer( __UpperCAmelCase , in_channels=__UpperCAmelCase , out_channels=__UpperCAmelCase , kernel_size=3 , stride=strides[i] , groups=__UpperCAmelCase , ) ) self.layer.append( MobileNetVaConvLayer( __UpperCAmelCase , in_channels=__UpperCAmelCase , out_channels=__UpperCAmelCase , kernel_size=1 , ) ) __lowerCamelCase = nn.AdaptiveAvgPoolad((1, 1) ) if add_pooling_layer else None # Initialize weights and apply final processing self.post_init() def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' raise NotImplementedError @add_start_docstrings_to_model_forward(__UpperCAmelCase ) @add_code_sample_docstrings( checkpoint=_CHECKPOINT_FOR_DOC , output_type=__UpperCAmelCase , config_class=_CONFIG_FOR_DOC , modality='''vision''' , expected_output=_EXPECTED_OUTPUT_SHAPE , ) def lowerCamelCase ( self , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , ): '''simple docstring''' __lowerCamelCase = ( output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states ) __lowerCamelCase = 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''' ) __lowerCamelCase = self.conv_stem(__UpperCAmelCase ) __lowerCamelCase = () if output_hidden_states else None for i, layer_module in enumerate(self.layer ): __lowerCamelCase = layer_module(__UpperCAmelCase ) if output_hidden_states: __lowerCamelCase = all_hidden_states + (hidden_states,) __lowerCamelCase = hidden_states if self.pooler is not None: __lowerCamelCase = torch.flatten(self.pooler(__UpperCAmelCase ) , start_dim=1 ) else: __lowerCamelCase = 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=__UpperCAmelCase , pooler_output=__UpperCAmelCase , hidden_states=__UpperCAmelCase , ) @add_start_docstrings( """ MobileNetV1 model with an image classification head on top (a linear layer on top of the pooled features), e.g. for ImageNet. """ , lowerCAmelCase__ , ) class __lowerCAmelCase ( lowerCAmelCase__ ): def __init__( self , __UpperCAmelCase ): '''simple docstring''' super().__init__(__UpperCAmelCase ) __lowerCamelCase = config.num_labels __lowerCamelCase = MobileNetVaModel(__UpperCAmelCase ) __lowerCamelCase = self.mobilenet_va.layer[-1].convolution.out_channels # Classifier head __lowerCamelCase = nn.Dropout(config.classifier_dropout_prob , inplace=__UpperCAmelCase ) __lowerCamelCase = nn.Linear(__UpperCAmelCase , 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(__UpperCAmelCase ) @add_code_sample_docstrings( checkpoint=_IMAGE_CLASS_CHECKPOINT , output_type=__UpperCAmelCase , config_class=_CONFIG_FOR_DOC , expected_output=_IMAGE_CLASS_EXPECTED_OUTPUT , ) def lowerCamelCase ( self , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , __UpperCAmelCase = None , ): '''simple docstring''' __lowerCamelCase = return_dict if return_dict is not None else self.config.use_return_dict __lowerCamelCase = self.mobilenet_va(__UpperCAmelCase , output_hidden_states=__UpperCAmelCase , return_dict=__UpperCAmelCase ) __lowerCamelCase = outputs.pooler_output if return_dict else outputs[1] __lowerCamelCase = self.classifier(self.dropout(__UpperCAmelCase ) ) __lowerCamelCase = None if labels is not None: if self.config.problem_type is None: if self.num_labels == 1: __lowerCamelCase = '''regression''' elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int): __lowerCamelCase = '''single_label_classification''' else: __lowerCamelCase = '''multi_label_classification''' if self.config.problem_type == "regression": __lowerCamelCase = MSELoss() if self.num_labels == 1: __lowerCamelCase = loss_fct(logits.squeeze() , labels.squeeze() ) else: __lowerCamelCase = loss_fct(__UpperCAmelCase , __UpperCAmelCase ) elif self.config.problem_type == "single_label_classification": __lowerCamelCase = CrossEntropyLoss() __lowerCamelCase = loss_fct(logits.view(-1 , self.num_labels ) , labels.view(-1 ) ) elif self.config.problem_type == "multi_label_classification": __lowerCamelCase = BCEWithLogitsLoss() __lowerCamelCase = loss_fct(__UpperCAmelCase , __UpperCAmelCase ) if not return_dict: __lowerCamelCase = (logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return ImageClassifierOutputWithNoAttention( loss=__UpperCAmelCase , logits=__UpperCAmelCase , hidden_states=outputs.hidden_states , )
622
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a_ = { """configuration_clipseg""": [ """CLIPSEG_PRETRAINED_CONFIG_ARCHIVE_MAP""", """CLIPSegConfig""", """CLIPSegTextConfig""", """CLIPSegVisionConfig""", ], """processing_clipseg""": ["""CLIPSegProcessor"""], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ """CLIPSEG_PRETRAINED_MODEL_ARCHIVE_LIST""", """CLIPSegModel""", """CLIPSegPreTrainedModel""", """CLIPSegTextModel""", """CLIPSegVisionModel""", """CLIPSegForImageSegmentation""", ] if TYPE_CHECKING: from .configuration_clipseg import ( CLIPSEG_PRETRAINED_CONFIG_ARCHIVE_MAP, CLIPSegConfig, CLIPSegTextConfig, CLIPSegVisionConfig, ) from .processing_clipseg import CLIPSegProcessor try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_clipseg import ( CLIPSEG_PRETRAINED_MODEL_ARCHIVE_LIST, CLIPSegForImageSegmentation, CLIPSegModel, CLIPSegPreTrainedModel, CLIPSegTextModel, CLIPSegVisionModel, ) else: import sys a_ = _LazyModule(__name__, globals()["""__file__"""], _import_structure, module_spec=__spec__)
622
1
import json import os import pickle import shutil import tempfile from unittest import TestCase from unittest.mock import patch import numpy as np from datasets import Dataset from transformers import is_faiss_available from transformers.models.bart.configuration_bart import BartConfig from transformers.models.bart.tokenization_bart import BartTokenizer from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES from transformers.models.dpr.configuration_dpr import DPRConfig from transformers.models.dpr.tokenization_dpr import DPRContextEncoderTokenizer, DPRQuestionEncoderTokenizer from transformers.models.rag.configuration_rag import RagConfig from transformers.models.rag.retrieval_rag import CustomHFIndex, RagRetriever from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES from transformers.testing_utils import require_faiss, require_sentencepiece, require_tokenizers, require_torch if is_faiss_available(): import faiss @require_faiss class __lowerCAmelCase ( lowerCAmelCase__ ): def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = tempfile.mkdtemp() __lowerCamelCase = 8 # DPR tok __lowerCamelCase = [ '''[UNK]''', '''[CLS]''', '''[SEP]''', '''[PAD]''', '''[MASK]''', '''want''', '''##want''', '''##ed''', '''wa''', '''un''', '''runn''', '''##ing''', ''',''', '''low''', '''lowest''', ] __lowerCamelCase = os.path.join(self.tmpdirname , '''dpr_tokenizer''' ) os.makedirs(__UpperCAmelCase , exist_ok=__UpperCAmelCase ) __lowerCamelCase = os.path.join(__UpperCAmelCase , DPR_VOCAB_FILES_NAMES['''vocab_file'''] ) with open(self.vocab_file , '''w''' , encoding='''utf-8''' ) as vocab_writer: vocab_writer.write(''''''.join([x + '''\n''' for x in vocab_tokens] ) ) # BART tok __lowerCamelCase = [ '''l''', '''o''', '''w''', '''e''', '''r''', '''s''', '''t''', '''i''', '''d''', '''n''', '''\u0120''', '''\u0120l''', '''\u0120n''', '''\u0120lo''', '''\u0120low''', '''er''', '''\u0120lowest''', '''\u0120newer''', '''\u0120wider''', '''<unk>''', ] __lowerCamelCase = dict(zip(__UpperCAmelCase , range(len(__UpperCAmelCase ) ) ) ) __lowerCamelCase = ['''#version: 0.2''', '''\u0120 l''', '''\u0120l o''', '''\u0120lo w''', '''e r''', ''''''] __lowerCamelCase = {'''unk_token''': '''<unk>'''} __lowerCamelCase = os.path.join(self.tmpdirname , '''bart_tokenizer''' ) os.makedirs(__UpperCAmelCase , exist_ok=__UpperCAmelCase ) __lowerCamelCase = os.path.join(__UpperCAmelCase , BART_VOCAB_FILES_NAMES['''vocab_file'''] ) __lowerCamelCase = os.path.join(__UpperCAmelCase , BART_VOCAB_FILES_NAMES['''merges_file'''] ) with open(self.vocab_file , '''w''' , encoding='''utf-8''' ) as fp: fp.write(json.dumps(__UpperCAmelCase ) + '''\n''' ) with open(self.merges_file , '''w''' , encoding='''utf-8''' ) as fp: fp.write('''\n'''.join(__UpperCAmelCase ) ) def lowerCamelCase ( self ): '''simple docstring''' return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , '''dpr_tokenizer''' ) ) def lowerCamelCase ( self ): '''simple docstring''' return DPRContextEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , '''dpr_tokenizer''' ) ) def lowerCamelCase ( self ): '''simple docstring''' return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , '''bart_tokenizer''' ) ) def lowerCamelCase ( self ): '''simple docstring''' shutil.rmtree(self.tmpdirname ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = Dataset.from_dict( { '''id''': ['''0''', '''1'''], '''text''': ['''foo''', '''bar'''], '''title''': ['''Foo''', '''Bar'''], '''embeddings''': [np.ones(self.retrieval_vector_size ), 2 * np.ones(self.retrieval_vector_size )], } ) dataset.add_faiss_index('''embeddings''' , string_factory='''Flat''' , metric_type=faiss.METRIC_INNER_PRODUCT ) return dataset def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_dataset() __lowerCamelCase = RagConfig( retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , ) with patch('''transformers.models.rag.retrieval_rag.load_dataset''' ) as mock_load_dataset: __lowerCamelCase = dataset __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , ) return retriever def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.get_dummy_dataset() __lowerCamelCase = RagConfig( retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , index_name='''custom''' , ) if from_disk: __lowerCamelCase = os.path.join(self.tmpdirname , '''dataset''' ) __lowerCamelCase = os.path.join(self.tmpdirname , '''index.faiss''' ) dataset.get_index('''embeddings''' ).save(os.path.join(self.tmpdirname , '''index.faiss''' ) ) dataset.drop_index('''embeddings''' ) dataset.save_to_disk(os.path.join(self.tmpdirname , '''dataset''' ) ) del dataset __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , ) else: __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , index=CustomHFIndex(config.retrieval_vector_size , __UpperCAmelCase ) , ) return retriever def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = Dataset.from_dict( { '''id''': ['''0''', '''1'''], '''text''': ['''foo''', '''bar'''], '''title''': ['''Foo''', '''Bar'''], '''embeddings''': [np.ones(self.retrieval_vector_size + 1 ), 2 * np.ones(self.retrieval_vector_size + 1 )], } ) dataset.add_faiss_index('''embeddings''' , string_factory='''Flat''' , metric_type=faiss.METRIC_INNER_PRODUCT ) __lowerCamelCase = os.path.join(self.tmpdirname , '''hf_bert_base.hnswSQ8_correct_phi_128.c_index''' ) dataset.save_faiss_index('''embeddings''' , index_file_name + '''.index.dpr''' ) pickle.dump(dataset['''id'''] , open(index_file_name + '''.index_meta.dpr''' , '''wb''' ) ) __lowerCamelCase = os.path.join(self.tmpdirname , '''psgs_w100.tsv.pkl''' ) __lowerCamelCase = {sample['''id''']: [sample['''text'''], sample['''title''']] for sample in dataset} pickle.dump(__UpperCAmelCase , open(__UpperCAmelCase , '''wb''' ) ) __lowerCamelCase = RagConfig( retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , index_name='''legacy''' , index_path=self.tmpdirname , ) __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() ) return retriever def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_canonical_hf_index_retriever() __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''embeddings''', '''id''', '''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''id'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''id'''][0] , '''1''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''id'''][0] , '''0''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_canonical_hf_index_retriever() with tempfile.TemporaryDirectory() as tmp_dirname: with patch('''transformers.models.rag.retrieval_rag.load_dataset''' ) as mock_load_dataset: __lowerCamelCase = self.get_dummy_dataset() retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''embeddings''', '''id''', '''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''id'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''id'''][0] , '''1''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''id'''][0] , '''0''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) with tempfile.TemporaryDirectory() as tmp_dirname: retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''embeddings''', '''id''', '''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''id'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''id'''][0] , '''1''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''id'''][0] , '''0''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) with tempfile.TemporaryDirectory() as tmp_dirname: retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_legacy_index_retriever() __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''text'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''text'''][0] , '''bar''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''text'''][0] , '''foo''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_legacy_index_retriever() with tempfile.TemporaryDirectory() as tmp_dirname: retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) @require_torch @require_tokenizers @require_sentencepiece def lowerCamelCase ( self ): '''simple docstring''' import torch __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_canonical_hf_index_retriever() __lowerCamelCase = [[5, 7], [10, 11]] __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever(__UpperCAmelCase , __UpperCAmelCase , prefix=retriever.config.generator.prefix , n_docs=__UpperCAmelCase ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = ( out['''context_input_ids'''], out['''context_attention_mask'''], out['''retrieved_doc_embeds'''], ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , np.ndarray ) __lowerCamelCase = retriever( __UpperCAmelCase , __UpperCAmelCase , prefix=retriever.config.generator.prefix , n_docs=__UpperCAmelCase , return_tensors='''pt''' , ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = ( # noqa: F841 out['''context_input_ids'''], out['''context_attention_mask'''], out['''retrieved_doc_embeds'''], out['''doc_ids'''], ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertIsInstance(__UpperCAmelCase , torch.Tensor ) self.assertIsInstance(__UpperCAmelCase , torch.Tensor ) self.assertIsInstance(__UpperCAmelCase , torch.Tensor ) @require_torch @require_tokenizers @require_sentencepiece def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dpr_ctx_encoder_tokenizer() __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) retriever.set_ctx_encoder_tokenizer(__UpperCAmelCase ) __lowerCamelCase = [[5, 7], [10, 11]] __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever(__UpperCAmelCase , __UpperCAmelCase , prefix=retriever.config.generator.prefix , n_docs=__UpperCAmelCase ) self.assertEqual( len(__UpperCAmelCase ) , 6 ) # check whether the retriever output consist of 6 attributes including tokenized docs self.assertEqual( all(k in out for k in ('''tokenized_doc_ids''', '''tokenized_doc_attention_mask''') ) , __UpperCAmelCase ) # check for doc token related keys in dictionary.
622
import unittest from transformers import RoFormerTokenizer, RoFormerTokenizerFast from transformers.testing_utils import require_rjieba, require_tokenizers from ...test_tokenization_common import TokenizerTesterMixin @require_rjieba @require_tokenizers class __lowerCAmelCase ( lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = RoFormerTokenizer lowerCAmelCase__ = RoFormerTokenizerFast lowerCAmelCase__ = True lowerCAmelCase__ = True def lowerCamelCase ( self ): '''simple docstring''' super().setUp() def lowerCamelCase ( self , **__UpperCAmelCase ): '''simple docstring''' return self.tokenizer_class.from_pretrained('''junnyu/roformer_chinese_base''' , **__UpperCAmelCase ) def lowerCamelCase ( self , **__UpperCAmelCase ): '''simple docstring''' return self.rust_tokenizer_class.from_pretrained('''junnyu/roformer_chinese_base''' , **__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''永和服装饰品有限公司,今天天气非常好''' __lowerCamelCase = '''永和 服装 饰品 有限公司 , 今 天 天 气 非常 好''' return input_text, output_text def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_tokenizer() __lowerCamelCase ,__lowerCamelCase = self.get_chinese_input_output_texts() __lowerCamelCase = tokenizer.tokenize(__UpperCAmelCase ) self.assertListEqual(__UpperCAmelCase , output_text.split() ) __lowerCamelCase = tokens + [tokenizer.unk_token] __lowerCamelCase = [22943, 21332, 34431, 45904, 117, 306, 1231, 1231, 2653, 33994, 1266, 100] self.assertListEqual(tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_rust_tokenizer() __lowerCamelCase ,__lowerCamelCase = self.get_chinese_input_output_texts() __lowerCamelCase = tokenizer.tokenize(__UpperCAmelCase ) self.assertListEqual(__UpperCAmelCase , output_text.split() ) __lowerCamelCase = tokens + [tokenizer.unk_token] __lowerCamelCase = [22943, 21332, 34431, 45904, 117, 306, 1231, 1231, 2653, 33994, 1266, 100] self.assertListEqual(tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' pass def lowerCamelCase ( self ): '''simple docstring''' pass def lowerCamelCase ( self ): '''simple docstring''' pass
622
1
from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { """facebook/timesformer""": """https://huggingface.co/facebook/timesformer/resolve/main/config.json""", } class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = """timesformer""" def __init__( self , __UpperCAmelCase=224 , __UpperCAmelCase=16 , __UpperCAmelCase=3 , __UpperCAmelCase=8 , __UpperCAmelCase=768 , __UpperCAmelCase=12 , __UpperCAmelCase=12 , __UpperCAmelCase=3072 , __UpperCAmelCase="gelu" , __UpperCAmelCase=0.0 , __UpperCAmelCase=0.0 , __UpperCAmelCase=0.02 , __UpperCAmelCase=1E-6 , __UpperCAmelCase=True , __UpperCAmelCase="divided_space_time" , __UpperCAmelCase=0 , **__UpperCAmelCase , ): '''simple docstring''' super().__init__(**__UpperCAmelCase ) __lowerCamelCase = image_size __lowerCamelCase = patch_size __lowerCamelCase = num_channels __lowerCamelCase = num_frames __lowerCamelCase = hidden_size __lowerCamelCase = num_hidden_layers __lowerCamelCase = num_attention_heads __lowerCamelCase = intermediate_size __lowerCamelCase = hidden_act __lowerCamelCase = hidden_dropout_prob __lowerCamelCase = attention_probs_dropout_prob __lowerCamelCase = initializer_range __lowerCamelCase = layer_norm_eps __lowerCamelCase = qkv_bias __lowerCamelCase = attention_type __lowerCamelCase = drop_path_rate
622
import gc import tempfile import unittest import numpy as np import torch from diffusers import VersatileDiffusionPipeline from diffusers.utils.testing_utils import load_image, nightly, require_torch_gpu, torch_device a_ = False class __lowerCAmelCase ( unittest.TestCase ): pass @nightly @require_torch_gpu class __lowerCAmelCase ( unittest.TestCase ): def lowerCamelCase ( self ): '''simple docstring''' # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = VersatileDiffusionPipeline.from_pretrained('''shi-labs/versatile-diffusion''' , torch_dtype=torch.floataa ) pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg''' ) __lowerCamelCase = torch.manual_seed(0 ) __lowerCamelCase = pipe.dual_guided( prompt='''first prompt''' , image=__UpperCAmelCase , text_to_image_strength=0.75 , generator=__UpperCAmelCase , guidance_scale=7.5 , num_inference_steps=2 , output_type='''numpy''' , ).images with tempfile.TemporaryDirectory() as tmpdirname: pipe.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = VersatileDiffusionPipeline.from_pretrained(__UpperCAmelCase , torch_dtype=torch.floataa ) pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = generator.manual_seed(0 ) __lowerCamelCase = pipe.dual_guided( prompt='''first prompt''' , image=__UpperCAmelCase , text_to_image_strength=0.75 , generator=__UpperCAmelCase , guidance_scale=7.5 , num_inference_steps=2 , output_type='''numpy''' , ).images assert np.abs(image - new_image ).sum() < 1E-5, "Models don't have the same forward pass" def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = VersatileDiffusionPipeline.from_pretrained('''shi-labs/versatile-diffusion''' , torch_dtype=torch.floataa ) pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = '''cyberpunk 2077''' __lowerCamelCase = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/versatile_diffusion/benz.jpg''' ) __lowerCamelCase = torch.manual_seed(0 ) __lowerCamelCase = pipe.dual_guided( prompt=__UpperCAmelCase , image=__UpperCAmelCase , text_to_image_strength=0.75 , generator=__UpperCAmelCase , guidance_scale=7.5 , num_inference_steps=50 , output_type='''numpy''' , ).images __lowerCamelCase = image[0, 253:256, 253:256, -1] assert image.shape == (1, 512, 512, 3) __lowerCamelCase = np.array([0.1_448, 0.1_619, 0.1_741, 0.1_086, 0.1_147, 0.1_128, 0.1_199, 0.1_165, 0.1_001] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 __lowerCamelCase = '''A painting of a squirrel eating a burger ''' __lowerCamelCase = torch.manual_seed(0 ) __lowerCamelCase = pipe.text_to_image( prompt=__UpperCAmelCase , generator=__UpperCAmelCase , guidance_scale=7.5 , num_inference_steps=50 , output_type='''numpy''' ).images __lowerCamelCase = image[0, 253:256, 253:256, -1] assert image.shape == (1, 512, 512, 3) __lowerCamelCase = np.array([0.3_367, 0.3_169, 0.2_656, 0.3_870, 0.4_790, 0.3_796, 0.4_009, 0.4_878, 0.4_778] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1 __lowerCamelCase = pipe.image_variation(__UpperCAmelCase , generator=__UpperCAmelCase , output_type='''numpy''' ).images __lowerCamelCase = image[0, 253:256, 253:256, -1] assert image.shape == (1, 512, 512, 3) __lowerCamelCase = np.array([0.3_076, 0.3_123, 0.3_284, 0.3_782, 0.3_770, 0.3_894, 0.4_297, 0.4_331, 0.4_456] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-1
622
1
import math def a__ ( _UpperCamelCase : str ,_UpperCamelCase : List[str] ): if 0 not in (x, y): # We use the relation x^y = y*log10(x), where 10 is the base. return y * math.logaa(_UpperCamelCase ) else: if x == 0: # 0 raised to any number is 0 return 0 elif y == 0: return 1 # any number raised to 0 is 1 raise AssertionError('''This should never happen''' ) if __name__ == "__main__": # Main function # Read two numbers from input and typecast them to int using map function. # Here x is the base and y is the power. a_ = """Enter the base and the power separated by a comma: """ a_ , a_ = map(int, input(prompt).split(""",""")) a_ , a_ = map(int, input(prompt).split(""",""")) # We find the log of each number, using the function res(), which takes two # arguments. a_ = res(xa, ya) a_ = res(xa, ya) # We check for the largest number if resa > resa: print("""Largest number is""", xa, """^""", ya) elif resa > resa: print("""Largest number is""", xa, """^""", ya) else: print("""Both are equal""")
622
import argparse import datetime import json import time import warnings from logging import getLogger from pathlib import Path from typing import Dict, List import torch from tqdm import tqdm from transformers import AutoModelForSeqaSeqLM, AutoTokenizer from utils import calculate_bleu, calculate_rouge, chunks, parse_numeric_n_bool_cl_kwargs, use_task_specific_params a_ = getLogger(__name__) a_ = """cuda""" if torch.cuda.is_available() else """cpu""" def a__ ( _UpperCamelCase : List[str] ,_UpperCamelCase : str ,_UpperCamelCase : str ,_UpperCamelCase : int = 8 ,_UpperCamelCase : str = DEFAULT_DEVICE ,_UpperCamelCase : Dict=False ,_UpperCamelCase : Dict="summarization" ,_UpperCamelCase : Optional[int]=None ,**_UpperCamelCase : Dict ,): __lowerCamelCase = Path(_UpperCamelCase ).open('''w''' ,encoding='''utf-8''' ) __lowerCamelCase = str(_UpperCamelCase ) __lowerCamelCase = AutoModelForSeqaSeqLM.from_pretrained(_UpperCamelCase ).to(_UpperCamelCase ) if fpaa: __lowerCamelCase = model.half() __lowerCamelCase = AutoTokenizer.from_pretrained(_UpperCamelCase ) logger.info(F"""Inferred tokenizer type: {tokenizer.__class__}""" ) # if this is wrong, check config.model_type. __lowerCamelCase = time.time() # update config with task specific params use_task_specific_params(_UpperCamelCase ,_UpperCamelCase ) if prefix is None: __lowerCamelCase = prefix or getattr(model.config ,'''prefix''' ,'''''' ) or '''''' for examples_chunk in tqdm(list(chunks(_UpperCamelCase ,_UpperCamelCase ) ) ): __lowerCamelCase = [prefix + text for text in examples_chunk] __lowerCamelCase = tokenizer(_UpperCamelCase ,return_tensors='''pt''' ,truncation=_UpperCamelCase ,padding='''longest''' ).to(_UpperCamelCase ) __lowerCamelCase = model.generate( input_ids=batch.input_ids ,attention_mask=batch.attention_mask ,**_UpperCamelCase ,) __lowerCamelCase = tokenizer.batch_decode(_UpperCamelCase ,skip_special_tokens=_UpperCamelCase ,clean_up_tokenization_spaces=_UpperCamelCase ) for hypothesis in dec: fout.write(hypothesis + '''\n''' ) fout.flush() fout.close() __lowerCamelCase = int(time.time() - start_time ) # seconds __lowerCamelCase = len(_UpperCamelCase ) return {"n_obs": n_obs, "runtime": runtime, "seconds_per_sample": round(runtime / n_obs ,4 )} def a__ ( ): return datetime.datetime.now().strftime('''%Y-%m-%d %H:%M:%S''' ) def a__ ( _UpperCamelCase : Union[str, Any]=True ): __lowerCamelCase = argparse.ArgumentParser() parser.add_argument('''model_name''' ,type=_UpperCamelCase ,help='''like facebook/bart-large-cnn,t5-base, etc.''' ) parser.add_argument('''input_path''' ,type=_UpperCamelCase ,help='''like cnn_dm/test.source''' ) parser.add_argument('''save_path''' ,type=_UpperCamelCase ,help='''where to save summaries''' ) parser.add_argument('''--reference_path''' ,type=_UpperCamelCase ,required=_UpperCamelCase ,help='''like cnn_dm/test.target''' ) parser.add_argument('''--score_path''' ,type=_UpperCamelCase ,required=_UpperCamelCase ,default='''metrics.json''' ,help='''where to save metrics''' ) parser.add_argument('''--device''' ,type=_UpperCamelCase ,required=_UpperCamelCase ,default=_UpperCamelCase ,help='''cuda, cuda:1, cpu etc.''' ) parser.add_argument( '''--prefix''' ,type=_UpperCamelCase ,required=_UpperCamelCase ,default=_UpperCamelCase ,help='''will be added to the begininng of src examples''' ) parser.add_argument('''--task''' ,type=_UpperCamelCase ,default='''summarization''' ,help='''used for task_specific_params + metrics''' ) parser.add_argument('''--bs''' ,type=_UpperCamelCase ,default=8 ,required=_UpperCamelCase ,help='''batch size''' ) parser.add_argument( '''--n_obs''' ,type=_UpperCamelCase ,default=-1 ,required=_UpperCamelCase ,help='''How many observations. Defaults to all.''' ) parser.add_argument('''--fp16''' ,action='''store_true''' ) parser.add_argument('''--dump-args''' ,action='''store_true''' ,help='''print the custom hparams with the results''' ) parser.add_argument( '''--info''' ,nargs='''?''' ,type=_UpperCamelCase ,const=datetime_now() ,help=( '''use in conjunction w/ --dump-args to print with the results whatever other info you\'d like, e.g.''' ''' lang=en-ru. If no value is passed, the current datetime string will be used.''' ) ,) # Unspecified args like --num_beams=2 --decoder_start_token_id=4 are passed to model.generate __lowerCamelCase ,__lowerCamelCase = parser.parse_known_args() __lowerCamelCase = parse_numeric_n_bool_cl_kwargs(_UpperCamelCase ) if parsed_args and verbose: print(F"""parsed the following generate kwargs: {parsed_args}""" ) __lowerCamelCase = [''' ''' + x.rstrip() if '''t5''' in args.model_name else x.rstrip() for x in open(args.input_path ).readlines()] if args.n_obs > 0: __lowerCamelCase = examples[: args.n_obs] Path(args.save_path ).parent.mkdir(exist_ok=_UpperCamelCase ) if args.reference_path is None and Path(args.score_path ).exists(): warnings.warn(F"""score_path {args.score_path} will be overwritten unless you type ctrl-c.""" ) if args.device == "cpu" and args.fpaa: # this mix leads to RuntimeError: "threshold_cpu" not implemented for 'Half' raise ValueError('''Can\'t mix --fp16 and --device cpu''' ) __lowerCamelCase = generate_summaries_or_translations( _UpperCamelCase ,args.save_path ,args.model_name ,batch_size=args.bs ,device=args.device ,fpaa=args.fpaa ,task=args.task ,prefix=args.prefix ,**_UpperCamelCase ,) if args.reference_path is None: return {} # Compute scores __lowerCamelCase = calculate_bleu if '''translation''' in args.task else calculate_rouge __lowerCamelCase = [x.rstrip() for x in open(args.save_path ).readlines()] __lowerCamelCase = [x.rstrip() for x in open(args.reference_path ).readlines()][: len(_UpperCamelCase )] __lowerCamelCase = score_fn(_UpperCamelCase ,_UpperCamelCase ) scores.update(_UpperCamelCase ) if args.dump_args: scores.update(_UpperCamelCase ) if args.info: __lowerCamelCase = args.info if verbose: print(_UpperCamelCase ) if args.score_path is not None: json.dump(_UpperCamelCase ,open(args.score_path ,'''w''' ) ) return scores if __name__ == "__main__": # Usage for MT: # python run_eval.py MODEL_NAME $DATA_DIR/test.source $save_dir/test_translations.txt --reference_path $DATA_DIR/test.target --score_path $save_dir/test_bleu.json --task translation $@ run_generate(verbose=True)
622
1
import unittest import numpy as np from transformers.testing_utils import require_torch, require_vision from transformers.utils import is_torch_available, is_vision_available from ...test_image_processing_common import ImageProcessingSavingTestMixin, prepare_image_inputs if is_torch_available(): import torch if is_vision_available(): from PIL import Image from transformers import LevitImageProcessor class __lowerCAmelCase ( unittest.TestCase ): def __init__( self , __UpperCAmelCase , __UpperCAmelCase=7 , __UpperCAmelCase=3 , __UpperCAmelCase=18 , __UpperCAmelCase=30 , __UpperCAmelCase=400 , __UpperCAmelCase=True , __UpperCAmelCase=None , __UpperCAmelCase=True , __UpperCAmelCase=None , __UpperCAmelCase=True , __UpperCAmelCase=[0.5, 0.5, 0.5] , __UpperCAmelCase=[0.5, 0.5, 0.5] , ): '''simple docstring''' __lowerCamelCase = size if size is not None else {'''shortest_edge''': 18} __lowerCamelCase = crop_size if crop_size is not None else {'''height''': 18, '''width''': 18} __lowerCamelCase = parent __lowerCamelCase = batch_size __lowerCamelCase = num_channels __lowerCamelCase = image_size __lowerCamelCase = min_resolution __lowerCamelCase = max_resolution __lowerCamelCase = do_resize __lowerCamelCase = size __lowerCamelCase = do_center_crop __lowerCamelCase = crop_size __lowerCamelCase = do_normalize __lowerCamelCase = image_mean __lowerCamelCase = image_std def lowerCamelCase ( self ): '''simple docstring''' return { "image_mean": self.image_mean, "image_std": self.image_std, "do_normalize": self.do_normalize, "do_resize": self.do_resize, "do_center_crop": self.do_center_crop, "size": self.size, "crop_size": self.crop_size, } @require_torch @require_vision class __lowerCAmelCase ( lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = LevitImageProcessor if is_vision_available() else None def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = LevitImageProcessingTester(self ) @property def lowerCamelCase ( self ): '''simple docstring''' return self.image_processor_tester.prepare_image_processor_dict() def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.image_processing_class(**self.image_processor_dict ) self.assertTrue(hasattr(__UpperCAmelCase , '''image_mean''' ) ) self.assertTrue(hasattr(__UpperCAmelCase , '''image_std''' ) ) self.assertTrue(hasattr(__UpperCAmelCase , '''do_normalize''' ) ) self.assertTrue(hasattr(__UpperCAmelCase , '''do_resize''' ) ) self.assertTrue(hasattr(__UpperCAmelCase , '''do_center_crop''' ) ) self.assertTrue(hasattr(__UpperCAmelCase , '''size''' ) ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.image_processing_class.from_dict(self.image_processor_dict ) self.assertEqual(image_processor.size , {'''shortest_edge''': 18} ) self.assertEqual(image_processor.crop_size , {'''height''': 18, '''width''': 18} ) __lowerCamelCase = self.image_processing_class.from_dict(self.image_processor_dict , size=42 , crop_size=84 ) self.assertEqual(image_processor.size , {'''shortest_edge''': 42} ) self.assertEqual(image_processor.crop_size , {'''height''': 84, '''width''': 84} ) def lowerCamelCase ( self ): '''simple docstring''' pass def lowerCamelCase ( self ): '''simple docstring''' # Initialize image_processing __lowerCamelCase = self.image_processing_class(**self.image_processor_dict ) # create random PIL images __lowerCamelCase = prepare_image_inputs(self.image_processor_tester , equal_resolution=__UpperCAmelCase ) for image in image_inputs: self.assertIsInstance(__UpperCAmelCase , Image.Image ) # Test not batched input __lowerCamelCase = image_processing(image_inputs[0] , return_tensors='''pt''' ).pixel_values self.assertEqual( encoded_images.shape , ( 1, self.image_processor_tester.num_channels, self.image_processor_tester.crop_size['''height'''], self.image_processor_tester.crop_size['''width'''], ) , ) # Test batched __lowerCamelCase = image_processing(__UpperCAmelCase , return_tensors='''pt''' ).pixel_values self.assertEqual( encoded_images.shape , ( self.image_processor_tester.batch_size, self.image_processor_tester.num_channels, self.image_processor_tester.crop_size['''height'''], self.image_processor_tester.crop_size['''width'''], ) , ) def lowerCamelCase ( self ): '''simple docstring''' # Initialize image_processing __lowerCamelCase = self.image_processing_class(**self.image_processor_dict ) # create random numpy tensors __lowerCamelCase = prepare_image_inputs(self.image_processor_tester , equal_resolution=__UpperCAmelCase , numpify=__UpperCAmelCase ) for image in image_inputs: self.assertIsInstance(__UpperCAmelCase , np.ndarray ) # Test not batched input __lowerCamelCase = image_processing(image_inputs[0] , return_tensors='''pt''' ).pixel_values self.assertEqual( encoded_images.shape , ( 1, self.image_processor_tester.num_channels, self.image_processor_tester.crop_size['''height'''], self.image_processor_tester.crop_size['''width'''], ) , ) # Test batched __lowerCamelCase = image_processing(__UpperCAmelCase , return_tensors='''pt''' ).pixel_values self.assertEqual( encoded_images.shape , ( self.image_processor_tester.batch_size, self.image_processor_tester.num_channels, self.image_processor_tester.crop_size['''height'''], self.image_processor_tester.crop_size['''width'''], ) , ) def lowerCamelCase ( self ): '''simple docstring''' # Initialize image_processing __lowerCamelCase = self.image_processing_class(**self.image_processor_dict ) # create random PyTorch tensors __lowerCamelCase = prepare_image_inputs(self.image_processor_tester , equal_resolution=__UpperCAmelCase , torchify=__UpperCAmelCase ) for image in image_inputs: self.assertIsInstance(__UpperCAmelCase , torch.Tensor ) # Test not batched input __lowerCamelCase = image_processing(image_inputs[0] , return_tensors='''pt''' ).pixel_values self.assertEqual( encoded_images.shape , ( 1, self.image_processor_tester.num_channels, self.image_processor_tester.crop_size['''height'''], self.image_processor_tester.crop_size['''width'''], ) , ) # Test batched __lowerCamelCase = image_processing(__UpperCAmelCase , return_tensors='''pt''' ).pixel_values self.assertEqual( encoded_images.shape , ( self.image_processor_tester.batch_size, self.image_processor_tester.num_channels, self.image_processor_tester.crop_size['''height'''], self.image_processor_tester.crop_size['''width'''], ) , )
622
from __future__ import annotations import unittest import numpy as np from transformers import OPTConfig, is_tf_available from transformers.testing_utils import require_sentencepiece, require_tf, slow from ...test_configuration_common import ConfigTester from ...test_modeling_tf_common import TFModelTesterMixin, ids_tensor from ...test_pipeline_mixin import PipelineTesterMixin if is_tf_available(): import tensorflow as tf from transformers import GPTaTokenizer, TFOPTForCausalLM, TFOPTModel def a__ ( _UpperCamelCase : Optional[int] ,_UpperCamelCase : List[str] ,_UpperCamelCase : List[Any]=None ,_UpperCamelCase : Any=None ): if attention_mask is None: __lowerCamelCase = tf.cast(tf.math.not_equal(_UpperCamelCase ,config.pad_token_id ) ,tf.inta ) return {"input_ids": input_ids, "attention_mask": attention_mask} @require_tf class __lowerCAmelCase : lowerCAmelCase__ = OPTConfig lowerCAmelCase__ = {} lowerCAmelCase__ = """gelu""" def __init__( self , __UpperCAmelCase , __UpperCAmelCase=13 , __UpperCAmelCase=7 , __UpperCAmelCase=True , __UpperCAmelCase=False , __UpperCAmelCase=99 , __UpperCAmelCase=16 , __UpperCAmelCase=2 , __UpperCAmelCase=4 , __UpperCAmelCase=4 , __UpperCAmelCase="gelu" , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=20 , __UpperCAmelCase=2 , __UpperCAmelCase=1 , __UpperCAmelCase=0 , __UpperCAmelCase=16 , __UpperCAmelCase=16 , ): '''simple docstring''' __lowerCamelCase = parent __lowerCamelCase = batch_size __lowerCamelCase = seq_length __lowerCamelCase = is_training __lowerCamelCase = use_labels __lowerCamelCase = vocab_size __lowerCamelCase = hidden_size __lowerCamelCase = num_hidden_layers __lowerCamelCase = num_attention_heads __lowerCamelCase = intermediate_size __lowerCamelCase = hidden_act __lowerCamelCase = hidden_dropout_prob __lowerCamelCase = attention_probs_dropout_prob __lowerCamelCase = max_position_embeddings __lowerCamelCase = eos_token_id __lowerCamelCase = pad_token_id __lowerCamelCase = bos_token_id __lowerCamelCase = embed_dim __lowerCamelCase = word_embed_proj_dim __lowerCamelCase = False def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = ids_tensor([self.batch_size, self.seq_length - 1] , self.vocab_size ) __lowerCamelCase = tf.expand_dims(tf.constant([self.eos_token_id] * self.batch_size ) , 1 ) __lowerCamelCase = tf.concat([input_ids, eos_tensor] , axis=1 ) __lowerCamelCase = self.config_cls( vocab_size=self.vocab_size , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , 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 , embed_dim=self.embed_dim , word_embed_proj_dim=self.word_embed_proj_dim , is_encoder_decoder=__UpperCAmelCase , **self.config_updates , ) __lowerCamelCase = prepare_opt_inputs_dict(__UpperCAmelCase , __UpperCAmelCase ) return config, inputs_dict def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = TFOPTModel(config=__UpperCAmelCase ) __lowerCamelCase = inputs_dict['''input_ids'''] __lowerCamelCase = input_ids[:1, :] __lowerCamelCase = inputs_dict['''attention_mask'''][:1, :] __lowerCamelCase = 1 # first forward pass __lowerCamelCase = model(__UpperCAmelCase , attention_mask=__UpperCAmelCase , use_cache=__UpperCAmelCase ) __lowerCamelCase ,__lowerCamelCase = outputs.to_tuple() # create hypothetical next token and extent to next_input_ids __lowerCamelCase = ids_tensor((self.batch_size, 3) , config.vocab_size ) __lowerCamelCase = tf.cast(ids_tensor((self.batch_size, 3) , 2 ) , tf.inta ) # append to next input_ids and __lowerCamelCase = tf.concat([input_ids, next_tokens] , axis=-1 ) __lowerCamelCase = tf.concat([attention_mask, next_attn_mask] , axis=-1 ) __lowerCamelCase = model(__UpperCAmelCase , attention_mask=__UpperCAmelCase )[0] __lowerCamelCase = model(__UpperCAmelCase , attention_mask=__UpperCAmelCase , past_key_values=__UpperCAmelCase )[0] self.parent.assertEqual(next_tokens.shape[1] , output_from_past.shape[1] ) # select random slice __lowerCamelCase = int(ids_tensor((1,) , output_from_past.shape[-1] ) ) __lowerCamelCase = output_from_no_past[:, -3:, random_slice_idx] __lowerCamelCase = output_from_past[:, :, random_slice_idx] # test that outputs are equal for slice tf.debugging.assert_near(__UpperCAmelCase , __UpperCAmelCase , rtol=1E-3 ) @require_tf class __lowerCAmelCase ( lowerCAmelCase__ , lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = (TFOPTModel, TFOPTForCausalLM) if is_tf_available() else () lowerCAmelCase__ = (TFOPTForCausalLM,) if is_tf_available() else () lowerCAmelCase__ = ( {"""feature-extraction""": TFOPTModel, """text-generation""": TFOPTForCausalLM} if is_tf_available() else {} ) lowerCAmelCase__ = False lowerCAmelCase__ = False lowerCAmelCase__ = False lowerCAmelCase__ = 1_0 def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = TFOPTModelTester(self ) __lowerCamelCase = ConfigTester(self , config_class=__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' self.config_tester.run_common_tests() def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs_for_common() self.model_tester.check_decoder_model_past_large_inputs(*__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase ,__lowerCamelCase = self.model_tester.prepare_config_and_inputs_for_common() def _get_word_embedding_weight(__UpperCAmelCase , __UpperCAmelCase ): if hasattr(__UpperCAmelCase , '''weight''' ): return embedding_layer.weight else: # Here we build the word embeddings weights if not exists. # And then we retry to get the attribute once built. model.build() if hasattr(__UpperCAmelCase , '''weight''' ): return embedding_layer.weight else: return None for model_class in self.all_model_classes: for size in [config.vocab_size - 10, config.vocab_size + 10]: # build the embeddings __lowerCamelCase = model_class(config=__UpperCAmelCase ) __lowerCamelCase = _get_word_embedding_weight(__UpperCAmelCase , model.get_input_embeddings() ) __lowerCamelCase = _get_word_embedding_weight(__UpperCAmelCase , model.get_output_embeddings() ) # reshape the embeddings model.resize_token_embeddings(__UpperCAmelCase ) __lowerCamelCase = _get_word_embedding_weight(__UpperCAmelCase , model.get_input_embeddings() ) __lowerCamelCase = _get_word_embedding_weight(__UpperCAmelCase , model.get_output_embeddings() ) # check that the resized embeddings size matches the desired size. __lowerCamelCase = size if size is not None else config.vocab_size self.assertEqual(new_input_embeddings.shape[0] , __UpperCAmelCase ) # check that weights remain the same after resizing __lowerCamelCase = True for pa, pa in zip(old_input_embeddings.value() , new_input_embeddings.value() ): if tf.math.reduce_sum(tf.math.abs(pa - pa ) ) > 0: __lowerCamelCase = False self.assertTrue(__UpperCAmelCase ) if old_output_embeddings is not None and new_output_embeddings is not None: self.assertEqual(new_output_embeddings.shape[0] , __UpperCAmelCase ) __lowerCamelCase = True for pa, pa in zip(old_output_embeddings.value() , new_output_embeddings.value() ): if tf.math.reduce_sum(tf.math.abs(pa - pa ) ) > 0: __lowerCamelCase = False self.assertTrue(__UpperCAmelCase ) def a__ ( _UpperCamelCase : Optional[Any] ): return tf.constant(_UpperCamelCase ,dtype=tf.intaa ) @require_tf class __lowerCAmelCase ( unittest.TestCase ): lowerCAmelCase__ = 9_9 def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = tf.ones((4, 1) , dtype=tf.intaa ) * 2 __lowerCamelCase = tf.concat([ids_tensor((4, 6) , self.vocab_size - 3 ) + 3, eos_column_vector] , axis=1 ) __lowerCamelCase = input_ids.shape[0] __lowerCamelCase = OPTConfig( vocab_size=self.vocab_size , hidden_size=24 , num_hidden_layers=2 , num_attention_heads=2 , 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 @require_sentencepiece @require_tf class __lowerCAmelCase ( unittest.TestCase ): @slow def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = TFOPTModel.from_pretrained('''facebook/opt-350m''' ) __lowerCamelCase = _long_tensor([[0, 31414, 232, 328, 740, 1140, 12695, 69, 46078, 1588, 2]] ) __lowerCamelCase = tf.not_equal(__UpperCAmelCase , model.config.pad_token_id ) with tf.GradientTape(): __lowerCamelCase = model(input_ids=__UpperCAmelCase , attention_mask=__UpperCAmelCase ).last_hidden_state __lowerCamelCase = (1, 11, 512) self.assertEqual(output.shape , __UpperCAmelCase ) __lowerCamelCase = tf.constant( [[-0.2_873, -1.9_218, -0.3_033], [-1.2_710, -0.1_338, -0.1_902], [0.4_095, 0.1_214, -1.3_121]] ) self.assertTrue(np.allclose(output[:, :3, :3] , __UpperCAmelCase , atol=4E-3 ) ) __lowerCamelCase = tf.function(__UpperCAmelCase , jit_compile=__UpperCAmelCase ) __lowerCamelCase = xla_generate(__UpperCAmelCase , __UpperCAmelCase )[0] self.assertTrue(np.allclose(output[:, :3, :3] , __UpperCAmelCase , atol=4E-2 ) ) @require_tf @slow class __lowerCAmelCase ( unittest.TestCase ): def lowerCamelCase ( self ): '''simple docstring''' super().setUp() __lowerCamelCase = '''facebook/opt-350m''' def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = TFOPTForCausalLM.from_pretrained(self.path_model ) __lowerCamelCase = GPTaTokenizer.from_pretrained(self.path_model ) __lowerCamelCase = [ '''Today is a beautiful day and I want to''', '''In the city of''', '''Paris is the capital of France and''', '''Computers and mobile phones have taken''', ] # verify that prompt without BOS token is identical to Metaseq -> add_special_tokens=False __lowerCamelCase = tokenizer(__UpperCAmelCase , return_tensors='''tf''' , padding=__UpperCAmelCase , add_special_tokens=__UpperCAmelCase ) __lowerCamelCase = tf.math.reduce_mean(model(inputs.input_ids , attention_mask=inputs.attention_mask )[0] , axis=-1 ) __lowerCamelCase = tf.constant( [ [1.3_851, -13.8_923, -10.5_229, -10.7_533, -0.2_309, -10.2_384, -0.5_365, -9.0_947, -5.1_670], [-4.7_073, -10.6_276, -3.9_415, -21.5_242, -0.2_822, -0.2_822, -0.2_822, -0.2_822, -0.2_822], [0.6_247, -3.4_229, -8.9_179, -1.4_297, -14.1_650, 1.4_146, -9.0_218, -0.2_703, -0.2_703], [6.4_783, -1.9_913, -10.7_926, -2.3_336, 1.5_092, -0.9_974, -6.8_213, 1.3_477, 1.3_477], ] ) self.assertTrue(np.allclose(__UpperCAmelCase , __UpperCAmelCase , atol=1E-4 ) ) __lowerCamelCase = tf.function(__UpperCAmelCase , jit_compile=__UpperCAmelCase ) __lowerCamelCase = tf.math.reduce_mean(xla_generate(inputs.input_ids , attention_mask=inputs.attention_mask )[0] , axis=-1 ) self.assertTrue(np.allclose(__UpperCAmelCase , __UpperCAmelCase , atol=1E-4 ) ) @require_tf @slow class __lowerCAmelCase ( unittest.TestCase ): @property def lowerCamelCase ( self ): '''simple docstring''' return [ "Today is a beautiful day and I want", "In the city of", "Paris is the capital of France and", "Computers and mobile phones have taken", ] def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''facebook/opt-125m''' __lowerCamelCase = [ '''Today is a beautiful day and I want to''', '''In the city of New York, the city''', '''Paris is the capital of France and the capital''', '''Computers and mobile phones have taken over the''', ] __lowerCamelCase = [] __lowerCamelCase = GPTaTokenizer.from_pretrained(__UpperCAmelCase ) __lowerCamelCase = TFOPTForCausalLM.from_pretrained(__UpperCAmelCase ) for prompt in self.prompts: __lowerCamelCase = tokenizer(__UpperCAmelCase , return_tensors='''tf''' ).input_ids __lowerCamelCase = model.generate(__UpperCAmelCase , max_length=10 ) __lowerCamelCase = tokenizer.batch_decode(__UpperCAmelCase , skip_special_tokens=__UpperCAmelCase ) predicted_outputs += generated_string self.assertListEqual(__UpperCAmelCase , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''facebook/opt-350m''' __lowerCamelCase = GPTaTokenizer.from_pretrained(__UpperCAmelCase ) __lowerCamelCase = TFOPTForCausalLM.from_pretrained(__UpperCAmelCase ) __lowerCamelCase = '''left''' # use different length sentences to test batching __lowerCamelCase = [ '''Hello, my dog is a little''', '''Today, I''', ] __lowerCamelCase = tokenizer(__UpperCAmelCase , return_tensors='''tf''' , padding=__UpperCAmelCase ) __lowerCamelCase = inputs['''input_ids'''] __lowerCamelCase = model.generate(input_ids=__UpperCAmelCase , attention_mask=inputs['''attention_mask'''] ) __lowerCamelCase = tokenizer(sentences[0] , return_tensors='''tf''' ).input_ids __lowerCamelCase = model.generate(input_ids=__UpperCAmelCase ) __lowerCamelCase = inputs_non_padded.shape[-1] - tf.math.reduce_sum( tf.cast(inputs['''attention_mask'''][-1] , tf.intaa ) ) __lowerCamelCase = tokenizer(sentences[1] , return_tensors='''tf''' ).input_ids __lowerCamelCase = model.generate(input_ids=__UpperCAmelCase , max_length=model.config.max_length - num_paddings ) __lowerCamelCase = tokenizer.batch_decode(__UpperCAmelCase , skip_special_tokens=__UpperCAmelCase ) __lowerCamelCase = tokenizer.decode(output_non_padded[0] , skip_special_tokens=__UpperCAmelCase ) __lowerCamelCase = tokenizer.decode(output_padded[0] , skip_special_tokens=__UpperCAmelCase ) __lowerCamelCase = [ '''Hello, my dog is a little bit of a dork.\nI\'m a little bit''', '''Today, I was in the middle of a conversation with a friend about the''', ] self.assertListEqual(__UpperCAmelCase , __UpperCAmelCase ) self.assertListEqual(__UpperCAmelCase , [non_padded_sentence, padded_sentence] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''facebook/opt-350m''' __lowerCamelCase = [ '''Today is a beautiful day and I want to''', '''In the city of San Francisco, the city''', '''Paris is the capital of France and the capital''', '''Computers and mobile phones have taken over the''', ] __lowerCamelCase = [] __lowerCamelCase = GPTaTokenizer.from_pretrained(__UpperCAmelCase ) __lowerCamelCase = TFOPTForCausalLM.from_pretrained(__UpperCAmelCase ) for prompt in self.prompts: __lowerCamelCase = tokenizer(__UpperCAmelCase , return_tensors='''tf''' ).input_ids __lowerCamelCase = model.generate(__UpperCAmelCase , max_length=10 ) __lowerCamelCase = tokenizer.batch_decode(__UpperCAmelCase , skip_special_tokens=__UpperCAmelCase ) predicted_outputs += generated_string self.assertListEqual(__UpperCAmelCase , __UpperCAmelCase )
622
1
from math import sqrt def a__ ( _UpperCamelCase : int ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and ( number >= 0 ), "'number' must been an int and positive" __lowerCamelCase = True # 0 and 1 are none primes. if number <= 1: __lowerCamelCase = False for divisor in range(2 ,int(round(sqrt(_UpperCamelCase ) ) ) + 1 ): # if 'number' divisible by 'divisor' then sets 'status' # of false and break up the loop. if number % divisor == 0: __lowerCamelCase = False break # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'status' must been from type bool" return status def a__ ( _UpperCamelCase : Any ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and (n > 2), "'N' must been an int and > 2" # beginList: contains all natural numbers from 2 up to N __lowerCamelCase = list(range(2 ,n + 1 ) ) __lowerCamelCase = [] # this list will be returns. # actual sieve of erathostenes for i in range(len(_UpperCamelCase ) ): for j in range(i + 1 ,len(_UpperCamelCase ) ): if (begin_list[i] != 0) and (begin_list[j] % begin_list[i] == 0): __lowerCamelCase = 0 # filters actual prime numbers. __lowerCamelCase = [x for x in begin_list if x != 0] # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'ans' must been from type list" return ans def a__ ( _UpperCamelCase : List[Any] ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and (n > 2), "'N' must been an int and > 2" __lowerCamelCase = [] # iterates over all numbers between 2 up to N+1 # if a number is prime then appends to list 'ans' for number in range(2 ,n + 1 ): if is_prime(_UpperCamelCase ): ans.append(_UpperCamelCase ) # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'ans' must been from type list" return ans def a__ ( _UpperCamelCase : Dict ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and number >= 0, "'number' must been an int and >= 0" __lowerCamelCase = [] # this list will be returns of the function. # potential prime number factors. __lowerCamelCase = 2 __lowerCamelCase = number if number == 0 or number == 1: ans.append(_UpperCamelCase ) # if 'number' not prime then builds the prime factorization of 'number' elif not is_prime(_UpperCamelCase ): while quotient != 1: if is_prime(_UpperCamelCase ) and (quotient % factor == 0): ans.append(_UpperCamelCase ) quotient /= factor else: factor += 1 else: ans.append(_UpperCamelCase ) # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'ans' must been from type list" return ans def a__ ( _UpperCamelCase : Tuple ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and ( number >= 0 ), "'number' bust been an int and >= 0" __lowerCamelCase = 0 # prime factorization of 'number' __lowerCamelCase = prime_factorization(_UpperCamelCase ) __lowerCamelCase = max(_UpperCamelCase ) # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'ans' must been from type int" return ans def a__ ( _UpperCamelCase : int ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and ( number >= 0 ), "'number' bust been an int and >= 0" __lowerCamelCase = 0 # prime factorization of 'number' __lowerCamelCase = prime_factorization(_UpperCamelCase ) __lowerCamelCase = min(_UpperCamelCase ) # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'ans' must been from type int" return ans def a__ ( _UpperCamelCase : int ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'number' must been an int" assert isinstance(number % 2 == 0 ,_UpperCamelCase ), "compare bust been from type bool" return number % 2 == 0 def a__ ( _UpperCamelCase : str ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ), "'number' must been an int" assert isinstance(number % 2 != 0 ,_UpperCamelCase ), "compare bust been from type bool" return number % 2 != 0 def a__ ( _UpperCamelCase : List[Any] ): assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and (number > 2) and is_even(_UpperCamelCase ) ), "'number' must been an int, even and > 2" __lowerCamelCase = [] # this list will returned # creates a list of prime numbers between 2 up to 'number' __lowerCamelCase = get_prime_numbers(_UpperCamelCase ) __lowerCamelCase = len(_UpperCamelCase ) # run variable for while-loops. __lowerCamelCase = 0 __lowerCamelCase = None # exit variable. for break up the loops __lowerCamelCase = True while i < len_pn and loop: __lowerCamelCase = i + 1 while j < len_pn and loop: if prime_numbers[i] + prime_numbers[j] == number: __lowerCamelCase = False ans.append(prime_numbers[i] ) ans.append(prime_numbers[j] ) j += 1 i += 1 # precondition assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and (len(_UpperCamelCase ) == 2) and (ans[0] + ans[1] == number) and is_prime(ans[0] ) and is_prime(ans[1] ) ), "'ans' must contains two primes. And sum of elements must been eq 'number'" return ans def a__ ( _UpperCamelCase : Tuple ,_UpperCamelCase : Optional[Any] ): assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and isinstance(_UpperCamelCase ,_UpperCamelCase ) and (numbera >= 0) and (numbera >= 0) ), "'number1' and 'number2' must been positive integer." __lowerCamelCase = 0 while numbera != 0: __lowerCamelCase = numbera % numbera __lowerCamelCase = numbera __lowerCamelCase = rest # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and ( numbera >= 0 ), "'number' must been from type int and positive" return numbera def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : Union[str, Any] ): assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and isinstance(_UpperCamelCase ,_UpperCamelCase ) and (numbera >= 1) and (numbera >= 1) ), "'number1' and 'number2' must been positive integer." __lowerCamelCase = 1 # actual answer that will be return. # for kgV (x,1) if numbera > 1 and numbera > 1: # builds the prime factorization of 'number1' and 'number2' __lowerCamelCase = prime_factorization(_UpperCamelCase ) __lowerCamelCase = prime_factorization(_UpperCamelCase ) elif numbera == 1 or numbera == 1: __lowerCamelCase = [] __lowerCamelCase = [] __lowerCamelCase = max(_UpperCamelCase ,_UpperCamelCase ) __lowerCamelCase = 0 __lowerCamelCase = 0 __lowerCamelCase = [] # captured numbers int both 'primeFac1' and 'primeFac2' # iterates through primeFac1 for n in prime_fac_a: if n not in done: if n in prime_fac_a: __lowerCamelCase = prime_fac_a.count(_UpperCamelCase ) __lowerCamelCase = prime_fac_a.count(_UpperCamelCase ) for _ in range(max(_UpperCamelCase ,_UpperCamelCase ) ): ans *= n else: __lowerCamelCase = prime_fac_a.count(_UpperCamelCase ) for _ in range(_UpperCamelCase ): ans *= n done.append(_UpperCamelCase ) # iterates through primeFac2 for n in prime_fac_a: if n not in done: __lowerCamelCase = prime_fac_a.count(_UpperCamelCase ) for _ in range(_UpperCamelCase ): ans *= n done.append(_UpperCamelCase ) # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and ( ans >= 0 ), "'ans' must been from type int and positive" return ans def a__ ( _UpperCamelCase : Optional[int] ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and (n >= 0), "'number' must been a positive int" __lowerCamelCase = 0 __lowerCamelCase = 2 # this variable holds the answer while index < n: index += 1 ans += 1 # counts to the next number # if ans not prime then # runs to the next prime number. while not is_prime(_UpperCamelCase ): ans += 1 # precondition assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and is_prime( _UpperCamelCase ), "'ans' must been a prime number and from type int" return ans def a__ ( _UpperCamelCase : str ,_UpperCamelCase : Union[str, Any] ): assert ( is_prime(_UpperCamelCase ) and is_prime(_UpperCamelCase ) and (p_number_a < p_number_a) ), "The arguments must been prime numbers and 'pNumber1' < 'pNumber2'" __lowerCamelCase = p_number_a + 1 # jump to the next number __lowerCamelCase = [] # this list will be returns. # if number is not prime then # fetch the next prime number. while not is_prime(_UpperCamelCase ): number += 1 while number < p_number_a: ans.append(_UpperCamelCase ) number += 1 # fetch the next prime number. while not is_prime(_UpperCamelCase ): number += 1 # precondition assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and ans[0] != p_number_a and ans[len(_UpperCamelCase ) - 1] != p_number_a ), "'ans' must been a list without the arguments" # 'ans' contains not 'pNumber1' and 'pNumber2' ! return ans def a__ ( _UpperCamelCase : str ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and (n >= 1), "'n' must been int and >= 1" __lowerCamelCase = [] # will be returned. for divisor in range(1 ,n + 1 ): if n % divisor == 0: ans.append(_UpperCamelCase ) # precondition assert ans[0] == 1 and ans[len(_UpperCamelCase ) - 1] == n, "Error in function getDivisiors(...)" return ans def a__ ( _UpperCamelCase : str ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and ( number > 1 ), "'number' must been an int and >= 1" __lowerCamelCase = get_divisors(_UpperCamelCase ) # precondition assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and (divisors[0] == 1) and (divisors[len(_UpperCamelCase ) - 1] == number) ), "Error in help-function getDivisiors(...)" # summed all divisors up to 'number' (exclusive), hence [:-1] return sum(divisors[:-1] ) == number def a__ ( _UpperCamelCase : List[str] ,_UpperCamelCase : Optional[int] ): assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and isinstance(_UpperCamelCase ,_UpperCamelCase ) and (denominator != 0) ), "The arguments must been from type int and 'denominator' != 0" # build the greatest common divisor of numerator and denominator. __lowerCamelCase = gcd(abs(_UpperCamelCase ) ,abs(_UpperCamelCase ) ) # precondition assert ( isinstance(_UpperCamelCase ,_UpperCamelCase ) and (numerator % gcd_of_fraction == 0) and (denominator % gcd_of_fraction == 0) ), "Error in function gcd(...,...)" return (numerator // gcd_of_fraction, denominator // gcd_of_fraction) def a__ ( _UpperCamelCase : Dict ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and (n >= 0), "'n' must been a int and >= 0" __lowerCamelCase = 1 # this will be return. for factor in range(1 ,n + 1 ): ans *= factor return ans def a__ ( _UpperCamelCase : Optional[Any] ): assert isinstance(_UpperCamelCase ,_UpperCamelCase ) and (n >= 0), "'n' must been an int and >= 0" __lowerCamelCase = 0 __lowerCamelCase = 1 __lowerCamelCase = 1 # this will be return for _ in range(n - 1 ): __lowerCamelCase = ans ans += fiba __lowerCamelCase = tmp return ans
622
import argparse import csv import logging import os import random import numpy as np import torch from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset from tqdm import tqdm, trange from transformers import ( CONFIG_NAME, WEIGHTS_NAME, AdamW, OpenAIGPTDoubleHeadsModel, OpenAIGPTTokenizer, get_linear_schedule_with_warmup, ) logging.basicConfig( format="""%(asctime)s - %(levelname)s - %(name)s - %(message)s""", datefmt="""%m/%d/%Y %H:%M:%S""", level=logging.INFO ) a_ = logging.getLogger(__name__) def a__ ( _UpperCamelCase : str ,_UpperCamelCase : List[Any] ): __lowerCamelCase = np.argmax(_UpperCamelCase ,axis=1 ) return np.sum(outputs == labels ) def a__ ( _UpperCamelCase : Optional[int] ): with open(_UpperCamelCase ,encoding='''utf_8''' ) as f: __lowerCamelCase = csv.reader(_UpperCamelCase ) __lowerCamelCase = [] next(_UpperCamelCase ) # skip the first line for line in tqdm(_UpperCamelCase ): output.append((''' '''.join(line[1:5] ), line[5], line[6], int(line[-1] ) - 1) ) return output def a__ ( _UpperCamelCase : Any ,_UpperCamelCase : Dict ,_UpperCamelCase : str ,_UpperCamelCase : Union[str, Any] ,_UpperCamelCase : Any ,_UpperCamelCase : Dict ): __lowerCamelCase = [] for dataset in encoded_datasets: __lowerCamelCase = len(_UpperCamelCase ) __lowerCamelCase = np.zeros((n_batch, 2, input_len) ,dtype=np.intaa ) __lowerCamelCase = np.zeros((n_batch, 2) ,dtype=np.intaa ) __lowerCamelCase = np.full((n_batch, 2, input_len) ,fill_value=-1_00 ,dtype=np.intaa ) __lowerCamelCase = np.zeros((n_batch,) ,dtype=np.intaa ) for ( i, (story, conta, conta, mc_label), ) in enumerate(_UpperCamelCase ): __lowerCamelCase = [start_token] + story[:cap_length] + [delimiter_token] + conta[:cap_length] + [clf_token] __lowerCamelCase = [start_token] + story[:cap_length] + [delimiter_token] + conta[:cap_length] + [clf_token] __lowerCamelCase = with_conta __lowerCamelCase = with_conta __lowerCamelCase = len(_UpperCamelCase ) - 1 __lowerCamelCase = len(_UpperCamelCase ) - 1 __lowerCamelCase = with_conta __lowerCamelCase = with_conta __lowerCamelCase = mc_label __lowerCamelCase = (input_ids, mc_token_ids, lm_labels, mc_labels) tensor_datasets.append(tuple(torch.tensor(_UpperCamelCase ) for t in all_inputs ) ) return tensor_datasets def a__ ( ): __lowerCamelCase = argparse.ArgumentParser() parser.add_argument('''--model_name''' ,type=_UpperCamelCase ,default='''openai-gpt''' ,help='''pretrained model name''' ) parser.add_argument('''--do_train''' ,action='''store_true''' ,help='''Whether to run training.''' ) parser.add_argument('''--do_eval''' ,action='''store_true''' ,help='''Whether to run eval on the dev set.''' ) parser.add_argument( '''--output_dir''' ,default=_UpperCamelCase ,type=_UpperCamelCase ,required=_UpperCamelCase ,help='''The output directory where the model predictions and checkpoints will be written.''' ,) parser.add_argument('''--train_dataset''' ,type=_UpperCamelCase ,default='''''' ) parser.add_argument('''--eval_dataset''' ,type=_UpperCamelCase ,default='''''' ) parser.add_argument('''--seed''' ,type=_UpperCamelCase ,default=42 ) parser.add_argument('''--num_train_epochs''' ,type=_UpperCamelCase ,default=3 ) parser.add_argument('''--train_batch_size''' ,type=_UpperCamelCase ,default=8 ) parser.add_argument('''--eval_batch_size''' ,type=_UpperCamelCase ,default=16 ) parser.add_argument('''--adam_epsilon''' ,default=1e-8 ,type=_UpperCamelCase ,help='''Epsilon for Adam optimizer.''' ) parser.add_argument('''--max_grad_norm''' ,type=_UpperCamelCase ,default=1 ) parser.add_argument( '''--max_steps''' ,default=-1 ,type=_UpperCamelCase ,help=( '''If > 0: set total number of training steps to perform. Override num_train_epochs.''' ) ,) parser.add_argument( '''--gradient_accumulation_steps''' ,type=_UpperCamelCase ,default=1 ,help='''Number of updates steps to accumulate before performing a backward/update pass.''' ,) parser.add_argument('''--learning_rate''' ,type=_UpperCamelCase ,default=6.25e-5 ) parser.add_argument('''--warmup_steps''' ,default=0 ,type=_UpperCamelCase ,help='''Linear warmup over warmup_steps.''' ) parser.add_argument('''--lr_schedule''' ,type=_UpperCamelCase ,default='''warmup_linear''' ) parser.add_argument('''--weight_decay''' ,type=_UpperCamelCase ,default=0.01 ) parser.add_argument('''--lm_coef''' ,type=_UpperCamelCase ,default=0.9 ) parser.add_argument('''--n_valid''' ,type=_UpperCamelCase ,default=3_74 ) parser.add_argument('''--server_ip''' ,type=_UpperCamelCase ,default='''''' ,help='''Can be used for distant debugging.''' ) parser.add_argument('''--server_port''' ,type=_UpperCamelCase ,default='''''' ,help='''Can be used for distant debugging.''' ) __lowerCamelCase = parser.parse_args() print(_UpperCamelCase ) if args.server_ip and args.server_port: # Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script import ptvsd print('''Waiting for debugger attach''' ) ptvsd.enable_attach(address=(args.server_ip, args.server_port) ,redirect_output=_UpperCamelCase ) ptvsd.wait_for_attach() random.seed(args.seed ) np.random.seed(args.seed ) torch.manual_seed(args.seed ) torch.cuda.manual_seed_all(args.seed ) __lowerCamelCase = torch.device('''cuda''' if torch.cuda.is_available() else '''cpu''' ) __lowerCamelCase = torch.cuda.device_count() logger.info('''device: {}, n_gpu {}'''.format(_UpperCamelCase ,_UpperCamelCase ) ) if not args.do_train and not args.do_eval: raise ValueError('''At least one of `do_train` or `do_eval` must be True.''' ) if not os.path.exists(args.output_dir ): os.makedirs(args.output_dir ) # Load tokenizer and model # This loading functions also add new tokens and embeddings called `special tokens` # These new embeddings will be fine-tuned on the RocStories dataset __lowerCamelCase = ['''_start_''', '''_delimiter_''', '''_classify_'''] __lowerCamelCase = OpenAIGPTTokenizer.from_pretrained(args.model_name ) tokenizer.add_tokens(_UpperCamelCase ) __lowerCamelCase = tokenizer.convert_tokens_to_ids(_UpperCamelCase ) __lowerCamelCase = OpenAIGPTDoubleHeadsModel.from_pretrained(args.model_name ) model.resize_token_embeddings(len(_UpperCamelCase ) ) model.to(_UpperCamelCase ) # Load and encode the datasets def tokenize_and_encode(_UpperCamelCase : Dict ): if isinstance(_UpperCamelCase ,_UpperCamelCase ): return tokenizer.convert_tokens_to_ids(tokenizer.tokenize(_UpperCamelCase ) ) elif isinstance(_UpperCamelCase ,_UpperCamelCase ): return obj return [tokenize_and_encode(_UpperCamelCase ) for o in obj] logger.info('''Encoding dataset...''' ) __lowerCamelCase = load_rocstories_dataset(args.train_dataset ) __lowerCamelCase = load_rocstories_dataset(args.eval_dataset ) __lowerCamelCase = (train_dataset, eval_dataset) __lowerCamelCase = tokenize_and_encode(_UpperCamelCase ) # Compute the max input length for the Transformer __lowerCamelCase = model.config.n_positions // 2 - 2 __lowerCamelCase = max( len(story[:max_length] ) + max(len(conta[:max_length] ) ,len(conta[:max_length] ) ) + 3 for dataset in encoded_datasets for story, conta, conta, _ in dataset ) __lowerCamelCase = min(_UpperCamelCase ,model.config.n_positions ) # Max size of input for the pre-trained model # Prepare inputs tensors and dataloaders __lowerCamelCase = pre_process_datasets(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ,*_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = tensor_datasets[0], tensor_datasets[1] __lowerCamelCase = TensorDataset(*_UpperCamelCase ) __lowerCamelCase = RandomSampler(_UpperCamelCase ) __lowerCamelCase = DataLoader(_UpperCamelCase ,sampler=_UpperCamelCase ,batch_size=args.train_batch_size ) __lowerCamelCase = TensorDataset(*_UpperCamelCase ) __lowerCamelCase = SequentialSampler(_UpperCamelCase ) __lowerCamelCase = DataLoader(_UpperCamelCase ,sampler=_UpperCamelCase ,batch_size=args.eval_batch_size ) # Prepare optimizer if args.do_train: if args.max_steps > 0: __lowerCamelCase = args.max_steps __lowerCamelCase = args.max_steps // (len(_UpperCamelCase ) // args.gradient_accumulation_steps) + 1 else: __lowerCamelCase = len(_UpperCamelCase ) // args.gradient_accumulation_steps * args.num_train_epochs __lowerCamelCase = list(model.named_parameters() ) __lowerCamelCase = ['''bias''', '''LayerNorm.bias''', '''LayerNorm.weight'''] __lowerCamelCase = [ { '''params''': [p for n, p in param_optimizer if not any(nd in n for nd in no_decay )], '''weight_decay''': args.weight_decay, }, {'''params''': [p for n, p in param_optimizer if any(nd in n for nd in no_decay )], '''weight_decay''': 0.0}, ] __lowerCamelCase = AdamW(_UpperCamelCase ,lr=args.learning_rate ,eps=args.adam_epsilon ) __lowerCamelCase = get_linear_schedule_with_warmup( _UpperCamelCase ,num_warmup_steps=args.warmup_steps ,num_training_steps=_UpperCamelCase ) if args.do_train: __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = 0, 0, None model.train() for _ in trange(int(args.num_train_epochs ) ,desc='''Epoch''' ): __lowerCamelCase = 0 __lowerCamelCase = 0 __lowerCamelCase = tqdm(_UpperCamelCase ,desc='''Training''' ) for step, batch in enumerate(_UpperCamelCase ): __lowerCamelCase = tuple(t.to(_UpperCamelCase ) for t in batch ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = batch __lowerCamelCase = model(_UpperCamelCase ,mc_token_ids=_UpperCamelCase ,lm_labels=_UpperCamelCase ,mc_labels=_UpperCamelCase ) __lowerCamelCase = args.lm_coef * losses[0] + losses[1] loss.backward() optimizer.step() scheduler.step() optimizer.zero_grad() tr_loss += loss.item() __lowerCamelCase = ( loss.item() if exp_average_loss is None else 0.7 * exp_average_loss + 0.3 * loss.item() ) nb_tr_steps += 1 __lowerCamelCase = '''Training loss: {:.2e} lr: {:.2e}'''.format(_UpperCamelCase ,scheduler.get_lr()[0] ) # Save a trained model if args.do_train: # Save a trained model, configuration and tokenizer __lowerCamelCase = model.module if hasattr(_UpperCamelCase ,'''module''' ) else model # Only save the model itself # If we save using the predefined names, we can load using `from_pretrained` __lowerCamelCase = os.path.join(args.output_dir ,_UpperCamelCase ) __lowerCamelCase = os.path.join(args.output_dir ,_UpperCamelCase ) torch.save(model_to_save.state_dict() ,_UpperCamelCase ) model_to_save.config.to_json_file(_UpperCamelCase ) tokenizer.save_vocabulary(args.output_dir ) # Load a trained model and vocabulary that you have fine-tuned __lowerCamelCase = OpenAIGPTDoubleHeadsModel.from_pretrained(args.output_dir ) __lowerCamelCase = OpenAIGPTTokenizer.from_pretrained(args.output_dir ) model.to(_UpperCamelCase ) if args.do_eval: model.eval() __lowerCamelCase ,__lowerCamelCase = 0, 0 __lowerCamelCase ,__lowerCamelCase = 0, 0 for batch in tqdm(_UpperCamelCase ,desc='''Evaluating''' ): __lowerCamelCase = tuple(t.to(_UpperCamelCase ) for t in batch ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = batch with torch.no_grad(): __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = model( _UpperCamelCase ,mc_token_ids=_UpperCamelCase ,lm_labels=_UpperCamelCase ,mc_labels=_UpperCamelCase ) __lowerCamelCase = mc_logits.detach().cpu().numpy() __lowerCamelCase = mc_labels.to('''cpu''' ).numpy() __lowerCamelCase = accuracy(_UpperCamelCase ,_UpperCamelCase ) eval_loss += mc_loss.mean().item() eval_accuracy += tmp_eval_accuracy nb_eval_examples += input_ids.size(0 ) nb_eval_steps += 1 __lowerCamelCase = eval_loss / nb_eval_steps __lowerCamelCase = eval_accuracy / nb_eval_examples __lowerCamelCase = tr_loss / nb_tr_steps if args.do_train else None __lowerCamelCase = {'''eval_loss''': eval_loss, '''eval_accuracy''': eval_accuracy, '''train_loss''': train_loss} __lowerCamelCase = os.path.join(args.output_dir ,'''eval_results.txt''' ) with open(_UpperCamelCase ,'''w''' ) as writer: logger.info('''***** Eval results *****''' ) for key in sorted(result.keys() ): logger.info(''' %s = %s''' ,_UpperCamelCase ,str(result[key] ) ) writer.write('''%s = %s\n''' % (key, str(result[key] )) ) if __name__ == "__main__": main()
622
1
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 __lowerCAmelCase ( lowerCAmelCase__ , lowerCAmelCase__ , lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = StableUnCLIPImgaImgPipeline lowerCAmelCase__ = TEXT_GUIDED_IMAGE_VARIATION_PARAMS lowerCAmelCase__ = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS lowerCAmelCase__ = frozenset( [] ) # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess lowerCAmelCase__ = frozenset([] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 32 __lowerCamelCase = embedder_hidden_size # image encoding components __lowerCamelCase = CLIPImageProcessor(crop_size=32 , size=32 ) torch.manual_seed(0 ) __lowerCamelCase = CLIPVisionModelWithProjection( CLIPVisionConfig( hidden_size=__UpperCAmelCase , projection_dim=__UpperCAmelCase , num_hidden_layers=5 , num_attention_heads=4 , image_size=32 , intermediate_size=37 , patch_size=1 , ) ) # regular denoising components torch.manual_seed(0 ) __lowerCamelCase = StableUnCLIPImageNormalizer(embedding_dim=__UpperCAmelCase ) __lowerCamelCase = DDPMScheduler(beta_schedule='''squaredcos_cap_v2''' ) torch.manual_seed(0 ) __lowerCamelCase = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' ) torch.manual_seed(0 ) __lowerCamelCase = CLIPTextModel( CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=__UpperCAmelCase , projection_dim=32 , intermediate_size=37 , layer_norm_eps=1E-0_5 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , ) ) torch.manual_seed(0 ) __lowerCamelCase = 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=__UpperCAmelCase , layers_per_block=1 , upcast_attention=__UpperCAmelCase , use_linear_projection=__UpperCAmelCase , ) torch.manual_seed(0 ) __lowerCamelCase = DDIMScheduler( beta_schedule='''scaled_linear''' , beta_start=0.00_085 , beta_end=0.012 , prediction_type='''v_prediction''' , set_alpha_to_one=__UpperCAmelCase , steps_offset=1 , ) torch.manual_seed(0 ) __lowerCamelCase = AutoencoderKL() __lowerCamelCase = { # 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 , __UpperCAmelCase , __UpperCAmelCase=0 , __UpperCAmelCase=True ): '''simple docstring''' if str(__UpperCAmelCase ).startswith('''mps''' ): __lowerCamelCase = torch.manual_seed(__UpperCAmelCase ) else: __lowerCamelCase = torch.Generator(device=__UpperCAmelCase ).manual_seed(__UpperCAmelCase ) __lowerCamelCase = floats_tensor((1, 3, 32, 32) , rng=random.Random(__UpperCAmelCase ) ).to(__UpperCAmelCase ) if pil_image: __lowerCamelCase = input_image * 0.5 + 0.5 __lowerCamelCase = input_image.clamp(0 , 1 ) __lowerCamelCase = input_image.cpu().permute(0 , 2 , 3 , 1 ).float().numpy() __lowerCamelCase = DiffusionPipeline.numpy_to_pil(__UpperCAmelCase )[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 ): '''simple docstring''' __lowerCamelCase = '''cpu''' # ensure determinism for the device-dependent torch.Generator __lowerCamelCase = self.get_dummy_components() __lowerCamelCase = StableUnCLIPImgaImgPipeline(**__UpperCAmelCase ) __lowerCamelCase = sd_pipe.to(__UpperCAmelCase ) sd_pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = self.get_dummy_inputs(__UpperCAmelCase ) inputs.update({'''image_embeds''': None} ) __lowerCamelCase = sd_pipe(**__UpperCAmelCase ).images __lowerCamelCase = image[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) __lowerCamelCase = np.array([0.3_872, 0.7_224, 0.5_601, 0.4_741, 0.6_872, 0.5_814, 0.4_636, 0.3_867, 0.5_078] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-3 def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = torch_device in ['''cpu''', '''mps'''] self._test_attention_slicing_forward_pass(test_max_difference=__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = torch_device in ['''cpu''', '''mps'''] self._test_inference_batch_single_identical(test_max_difference=__UpperCAmelCase ) @unittest.skipIf( torch_device != '''cuda''' or not is_xformers_available() , reason='''XFormers attention is only available with CUDA and `xformers` installed''' , ) def lowerCamelCase ( self ): '''simple docstring''' self._test_xformers_attention_forwardGenerator_pass(test_max_difference=__UpperCAmelCase ) @slow @require_torch_gpu class __lowerCAmelCase ( unittest.TestCase ): def lowerCamelCase ( self ): '''simple docstring''' # clean up the VRAM after each test super().tearDown() gc.collect() torch.cuda.empty_cache() def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) __lowerCamelCase = 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''' ) __lowerCamelCase = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-l-img2img''' , torch_dtype=torch.floataa ) pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) # 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() __lowerCamelCase = torch.Generator(device='''cpu''' ).manual_seed(0 ) __lowerCamelCase = pipe(__UpperCAmelCase , '''anime turle''' , generator=__UpperCAmelCase , output_type='''np''' ) __lowerCamelCase = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(__UpperCAmelCase , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = load_image( '''https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/stable_unclip/turtle.png''' ) __lowerCamelCase = 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''' ) __lowerCamelCase = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) # 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() __lowerCamelCase = torch.Generator(device='''cpu''' ).manual_seed(0 ) __lowerCamelCase = pipe(__UpperCAmelCase , '''anime turle''' , generator=__UpperCAmelCase , output_type='''np''' ) __lowerCamelCase = output.images[0] assert image.shape == (768, 768, 3) assert_mean_pixel_difference(__UpperCAmelCase , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 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() __lowerCamelCase = StableUnCLIPImgaImgPipeline.from_pretrained( '''fusing/stable-unclip-2-1-h-img2img''' , torch_dtype=torch.floataa ) __lowerCamelCase = pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) pipe.enable_attention_slicing() pipe.enable_sequential_cpu_offload() __lowerCamelCase = pipe( __UpperCAmelCase , '''anime turtle''' , num_inference_steps=2 , output_type='''np''' , ) __lowerCamelCase = torch.cuda.max_memory_allocated() # make sure that less than 7 GB is allocated assert mem_bytes < 7 * 10**9
622
import logging import torch from accelerate import Accelerator from arguments import EvaluationArguments from datasets import load_dataset from torch.utils.data import IterableDataset from torch.utils.data.dataloader import DataLoader from transformers import AutoModelForCausalLM, AutoTokenizer, HfArgumentParser, set_seed class __lowerCAmelCase ( lowerCAmelCase__ ): def __init__( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase=1024 , __UpperCAmelCase=1024 , __UpperCAmelCase=3.6 ): '''simple docstring''' __lowerCamelCase = tokenizer __lowerCamelCase = tokenizer.bos_token_id __lowerCamelCase = dataset __lowerCamelCase = seq_length __lowerCamelCase = seq_length * chars_per_token * num_of_sequences def __iter__( self ): '''simple docstring''' __lowerCamelCase = iter(self.dataset ) __lowerCamelCase = True while more_examples: __lowerCamelCase ,__lowerCamelCase = [], 0 while True: if buffer_len >= self.input_characters: break try: buffer.append(next(__UpperCAmelCase )['''content'''] ) buffer_len += len(buffer[-1] ) except StopIteration: __lowerCamelCase = False break __lowerCamelCase = tokenizer(__UpperCAmelCase , truncation=__UpperCAmelCase )['''input_ids'''] __lowerCamelCase = [] for tokenized_input in tokenized_inputs: all_token_ids.extend(tokenized_input + [self.concat_token_id] ) for i in range(0 , len(__UpperCAmelCase ) , self.seq_length ): __lowerCamelCase = all_token_ids[i : i + self.seq_length] if len(__UpperCAmelCase ) == self.seq_length: yield torch.tensor(__UpperCAmelCase ) def a__ ( _UpperCamelCase : List[Any] ): __lowerCamelCase = {'''streaming''': True} __lowerCamelCase = load_dataset(args.dataset_name ,split='''train''' ,**_UpperCamelCase ) __lowerCamelCase = ConstantLengthDataset(_UpperCamelCase ,_UpperCamelCase ,seq_length=args.seq_length ) __lowerCamelCase = DataLoader(_UpperCamelCase ,batch_size=args.batch_size ) return eval_dataloader def a__ ( _UpperCamelCase : str ): model.eval() __lowerCamelCase = [] for step, batch in enumerate(_UpperCamelCase ): with torch.no_grad(): __lowerCamelCase = model(_UpperCamelCase ,labels=_UpperCamelCase ) __lowerCamelCase = outputs.loss.repeat(args.batch_size ) losses.append(accelerator.gather(_UpperCamelCase ) ) if args.max_eval_steps > 0 and step >= args.max_eval_steps: break __lowerCamelCase = torch.mean(torch.cat(_UpperCamelCase ) ) try: __lowerCamelCase = torch.exp(_UpperCamelCase ) except OverflowError: __lowerCamelCase = float('''inf''' ) return loss.item(), perplexity.item() # Setup Accelerator a_ = Accelerator() # Parse configuration a_ = HfArgumentParser(EvaluationArguments) a_ = parser.parse_args() set_seed(args.seed) # Logging a_ = logging.getLogger(__name__) logging.basicConfig( format="""%(asctime)s - %(levelname)s - %(name)s - %(message)s""", datefmt="""%m/%d/%Y %H:%M:%S""", level=logging.INFO ) # Load model and tokenizer a_ = AutoModelForCausalLM.from_pretrained(args.model_ckpt) a_ = AutoTokenizer.from_pretrained(args.model_ckpt) # Load dataset and dataloader a_ = create_dataloader(args) # Prepare everything with our `accelerator`. a_ , a_ = accelerator.prepare(model, eval_dataloader) # Evaluate and save the last checkpoint logger.info("""Evaluating and saving model after training""") a_ , a_ = evaluate(args) logger.info(f"loss/eval: {eval_loss}, perplexity: {perplexity}")
622
1
from typing import TYPE_CHECKING from ...utils import ( OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_tf_available, is_torch_available, is_vision_available, ) a_ = {"""configuration_vit""": ["""VIT_PRETRAINED_CONFIG_ARCHIVE_MAP""", """ViTConfig""", """ViTOnnxConfig"""]} try: if not is_vision_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = ["""ViTFeatureExtractor"""] a_ = ["""ViTImageProcessor"""] try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ """VIT_PRETRAINED_MODEL_ARCHIVE_LIST""", """ViTForImageClassification""", """ViTForMaskedImageModeling""", """ViTModel""", """ViTPreTrainedModel""", ] try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ """TFViTForImageClassification""", """TFViTModel""", """TFViTPreTrainedModel""", ] try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ """FlaxViTForImageClassification""", """FlaxViTModel""", """FlaxViTPreTrainedModel""", ] if TYPE_CHECKING: from .configuration_vit import VIT_PRETRAINED_CONFIG_ARCHIVE_MAP, ViTConfig, ViTOnnxConfig try: if not is_vision_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .feature_extraction_vit import ViTFeatureExtractor from .image_processing_vit import ViTImageProcessor try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_vit import ( VIT_PRETRAINED_MODEL_ARCHIVE_LIST, ViTForImageClassification, ViTForMaskedImageModeling, ViTModel, ViTPreTrainedModel, ) try: if not is_tf_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_tf_vit import TFViTForImageClassification, TFViTModel, TFViTPreTrainedModel try: if not is_flax_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_flax_vit import FlaxViTForImageClassification, FlaxViTModel, FlaxViTPreTrainedModel else: import sys a_ = _LazyModule(__name__, globals()["""__file__"""], _import_structure, module_spec=__spec__)
622
from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { """unc-nlp/lxmert-base-uncased""": """https://huggingface.co/unc-nlp/lxmert-base-uncased/resolve/main/config.json""", } class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = """lxmert""" lowerCAmelCase__ = {} def __init__( self , __UpperCAmelCase=30522 , __UpperCAmelCase=768 , __UpperCAmelCase=12 , __UpperCAmelCase=9500 , __UpperCAmelCase=1600 , __UpperCAmelCase=400 , __UpperCAmelCase=3072 , __UpperCAmelCase="gelu" , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=512 , __UpperCAmelCase=2 , __UpperCAmelCase=0.02 , __UpperCAmelCase=1E-1_2 , __UpperCAmelCase=9 , __UpperCAmelCase=5 , __UpperCAmelCase=5 , __UpperCAmelCase=2048 , __UpperCAmelCase=4 , __UpperCAmelCase=6.67 , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=True , **__UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = vocab_size __lowerCamelCase = hidden_size __lowerCamelCase = num_attention_heads __lowerCamelCase = hidden_act __lowerCamelCase = intermediate_size __lowerCamelCase = hidden_dropout_prob __lowerCamelCase = attention_probs_dropout_prob __lowerCamelCase = max_position_embeddings __lowerCamelCase = type_vocab_size __lowerCamelCase = initializer_range __lowerCamelCase = layer_norm_eps __lowerCamelCase = num_qa_labels __lowerCamelCase = num_object_labels __lowerCamelCase = num_attr_labels __lowerCamelCase = l_layers __lowerCamelCase = x_layers __lowerCamelCase = r_layers __lowerCamelCase = visual_feat_dim __lowerCamelCase = visual_pos_dim __lowerCamelCase = visual_loss_normalizer __lowerCamelCase = task_matched __lowerCamelCase = task_mask_lm __lowerCamelCase = task_obj_predict __lowerCamelCase = task_qa __lowerCamelCase = visual_obj_loss __lowerCamelCase = visual_attr_loss __lowerCamelCase = visual_feat_loss __lowerCamelCase = {'''vision''': r_layers, '''cross_encoder''': x_layers, '''language''': l_layers} super().__init__(**__UpperCAmelCase )
622
1
import argparse import os import re import packaging.version a_ = """examples/""" a_ = { """examples""": (re.compile(R"""^check_min_version\(\"[^\"]+\"\)\s*$""", re.MULTILINE), """check_min_version(\"VERSION\")\n"""), """init""": (re.compile(R"""^__version__\s+=\s+\"([^\"]+)\"\s*$""", re.MULTILINE), """__version__ = \"VERSION\"\n"""), """setup""": (re.compile(R"""^(\s*)version\s*=\s*\"[^\"]+\",""", re.MULTILINE), R"""\1version=\"VERSION\","""), """doc""": (re.compile(R"""^(\s*)release\s*=\s*\"[^\"]+\"$""", re.MULTILINE), """release = \"VERSION\"\n"""), } a_ = { """init""": """src/transformers/__init__.py""", """setup""": """setup.py""", } a_ = """README.md""" def a__ ( _UpperCamelCase : Optional[Any] ,_UpperCamelCase : Tuple ,_UpperCamelCase : Any ): with open(_UpperCamelCase ,'''r''' ,encoding='''utf-8''' ,newline='''\n''' ) as f: __lowerCamelCase = f.read() __lowerCamelCase ,__lowerCamelCase = REPLACE_PATTERNS[pattern] __lowerCamelCase = replace.replace('''VERSION''' ,_UpperCamelCase ) __lowerCamelCase = re_pattern.sub(_UpperCamelCase ,_UpperCamelCase ) with open(_UpperCamelCase ,'''w''' ,encoding='''utf-8''' ,newline='''\n''' ) as f: f.write(_UpperCamelCase ) def a__ ( _UpperCamelCase : Union[str, Any] ): for folder, directories, fnames in os.walk(_UpperCamelCase ): # Removing some of the folders with non-actively maintained examples from the walk if "research_projects" in directories: directories.remove('''research_projects''' ) if "legacy" in directories: directories.remove('''legacy''' ) for fname in fnames: if fname.endswith('''.py''' ): update_version_in_file(os.path.join(_UpperCamelCase ,_UpperCamelCase ) ,_UpperCamelCase ,pattern='''examples''' ) def a__ ( _UpperCamelCase : Union[str, Any] ,_UpperCamelCase : List[Any]=False ): for pattern, fname in REPLACE_FILES.items(): update_version_in_file(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ) if not patch: update_version_in_examples(_UpperCamelCase ) def a__ ( ): __lowerCamelCase = '''🤗 Transformers currently provides the following architectures''' __lowerCamelCase = '''1. Want to contribute a new model?''' with open(_UpperCamelCase ,'''r''' ,encoding='''utf-8''' ,newline='''\n''' ) as f: __lowerCamelCase = f.readlines() # Find the start of the list. __lowerCamelCase = 0 while not lines[start_index].startswith(_start_prompt ): start_index += 1 start_index += 1 __lowerCamelCase = start_index # Update the lines in the model list. while not lines[index].startswith(_end_prompt ): if lines[index].startswith('''1.''' ): __lowerCamelCase = lines[index].replace( '''https://huggingface.co/docs/transformers/main/model_doc''' ,'''https://huggingface.co/docs/transformers/model_doc''' ,) index += 1 with open(_UpperCamelCase ,'''w''' ,encoding='''utf-8''' ,newline='''\n''' ) as f: f.writelines(_UpperCamelCase ) def a__ ( ): with open(REPLACE_FILES['''init'''] ,'''r''' ) as f: __lowerCamelCase = f.read() __lowerCamelCase = REPLACE_PATTERNS['''init'''][0].search(_UpperCamelCase ).groups()[0] return packaging.version.parse(_UpperCamelCase ) def a__ ( _UpperCamelCase : Optional[Any]=False ): __lowerCamelCase = get_version() if patch and default_version.is_devrelease: raise ValueError('''Can\'t create a patch version from the dev branch, checkout a released version!''' ) if default_version.is_devrelease: __lowerCamelCase = default_version.base_version elif patch: __lowerCamelCase = F"""{default_version.major}.{default_version.minor}.{default_version.micro + 1}""" else: __lowerCamelCase = F"""{default_version.major}.{default_version.minor + 1}.0""" # Now let's ask nicely if that's the right one. __lowerCamelCase = input(F"""Which version are you releasing? [{default_version}]""" ) if len(_UpperCamelCase ) == 0: __lowerCamelCase = default_version print(F"""Updating version to {version}.""" ) global_version_update(_UpperCamelCase ,patch=_UpperCamelCase ) if not patch: print('''Cleaning main README, don\'t forget to run `make fix-copies`.''' ) clean_main_ref_in_model_list() def a__ ( ): __lowerCamelCase = get_version() __lowerCamelCase = F"""{current_version.major}.{current_version.minor + 1}.0.dev0""" __lowerCamelCase = current_version.base_version # Check with the user we got that right. __lowerCamelCase = input(F"""Which version are we developing now? [{dev_version}]""" ) if len(_UpperCamelCase ) == 0: __lowerCamelCase = dev_version print(F"""Updating version to {version}.""" ) global_version_update(_UpperCamelCase ) print('''Cleaning main README, don\'t forget to run `make fix-copies`.''' ) clean_main_ref_in_model_list() if __name__ == "__main__": a_ = argparse.ArgumentParser() parser.add_argument("""--post_release""", action="""store_true""", help="""Whether this is pre or post release.""") parser.add_argument("""--patch""", action="""store_true""", help="""Whether or not this is a patch release.""") a_ = parser.parse_args() if not args.post_release: pre_release_work(patch=args.patch) elif args.patch: print("""Nothing to do after a patch :-)""") else: post_release_work()
622
import unicodedata from dataclasses import dataclass from typing import Optional, Union import numpy as np from transformers.data.data_collator import DataCollatorMixin from transformers.file_utils import PaddingStrategy from transformers.tokenization_utils_base import PreTrainedTokenizerBase def a__ ( _UpperCamelCase : Tuple ,_UpperCamelCase : Any ,_UpperCamelCase : Union[str, Any] ,_UpperCamelCase : Any ): if isinstance(_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = np.full((len(_UpperCamelCase ), sequence_length, 2) ,_UpperCamelCase ) else: __lowerCamelCase = np.full((len(_UpperCamelCase ), sequence_length) ,_UpperCamelCase ) for i, tensor in enumerate(_UpperCamelCase ): if padding_side == "right": if isinstance(_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = tensor[:sequence_length] else: __lowerCamelCase = tensor[:sequence_length] else: if isinstance(_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = tensor[:sequence_length] else: __lowerCamelCase = tensor[:sequence_length] return out_tensor.tolist() def a__ ( _UpperCamelCase : Dict ): __lowerCamelCase = ord(_UpperCamelCase ) if (cp >= 33 and cp <= 47) or (cp >= 58 and cp <= 64) or (cp >= 91 and cp <= 96) or (cp >= 1_23 and cp <= 1_26): return True __lowerCamelCase = unicodedata.category(_UpperCamelCase ) if cat.startswith('''P''' ): return True return False @dataclass class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = 42 lowerCAmelCase__ = True lowerCAmelCase__ = None lowerCAmelCase__ = None lowerCAmelCase__ = -1_0_0 lowerCAmelCase__ = "pt" def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' import torch __lowerCamelCase = '''label''' if '''label''' in features[0].keys() else '''labels''' __lowerCamelCase = [feature[label_name] for feature in features] if label_name in features[0].keys() else None __lowerCamelCase = self.tokenizer.pad( __UpperCAmelCase , padding=self.padding , max_length=self.max_length , pad_to_multiple_of=self.pad_to_multiple_of , return_tensors='''pt''' if labels is None else None , ) if labels is None: return batch __lowerCamelCase = torch.tensor(batch['''entity_ids'''] ).shape[1] __lowerCamelCase = self.tokenizer.padding_side if padding_side == "right": __lowerCamelCase = [ list(__UpperCAmelCase ) + [self.label_pad_token_id] * (sequence_length - len(__UpperCAmelCase )) for label in labels ] else: __lowerCamelCase = [ [self.label_pad_token_id] * (sequence_length - len(__UpperCAmelCase )) + list(__UpperCAmelCase ) for label in labels ] __lowerCamelCase = [feature['''ner_tags'''] for feature in features] __lowerCamelCase = padding_tensor(__UpperCAmelCase , -1 , __UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = [feature['''original_entity_spans'''] for feature in features] __lowerCamelCase = padding_tensor(__UpperCAmelCase , (-1, -1) , __UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = {k: torch.tensor(__UpperCAmelCase , dtype=torch.intaa ) for k, v in batch.items()} return batch
622
1
def a__ ( _UpperCamelCase : list ): if not grid or not grid[0]: raise TypeError('''The grid does not contain the appropriate information''' ) for cell_n in range(1 ,len(grid[0] ) ): grid[0][cell_n] += grid[0][cell_n - 1] __lowerCamelCase = grid[0] for row_n in range(1 ,len(_UpperCamelCase ) ): __lowerCamelCase = grid[row_n] __lowerCamelCase = fill_row(_UpperCamelCase ,_UpperCamelCase ) __lowerCamelCase = grid[row_n] return grid[-1][-1] def a__ ( _UpperCamelCase : list ,_UpperCamelCase : list ): current_row[0] += row_above[0] for cell_n in range(1 ,len(_UpperCamelCase ) ): current_row[cell_n] += min(current_row[cell_n - 1] ,row_above[cell_n] ) return current_row if __name__ == "__main__": import doctest doctest.testmod()
622
from __future__ import annotations import unittest from transformers import FunnelConfig, is_tf_available from transformers.testing_utils import require_tf from ...test_configuration_common import ConfigTester from ...test_modeling_tf_common import TFModelTesterMixin, ids_tensor, random_attention_mask from ...test_pipeline_mixin import PipelineTesterMixin if is_tf_available(): import tensorflow as tf from transformers import ( TFFunnelBaseModel, TFFunnelForMaskedLM, TFFunnelForMultipleChoice, TFFunnelForPreTraining, TFFunnelForQuestionAnswering, TFFunnelForSequenceClassification, TFFunnelForTokenClassification, TFFunnelModel, ) class __lowerCAmelCase : def __init__( self , __UpperCAmelCase , __UpperCAmelCase=13 , __UpperCAmelCase=7 , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=99 , __UpperCAmelCase=[1, 1, 2] , __UpperCAmelCase=1 , __UpperCAmelCase=32 , __UpperCAmelCase=4 , __UpperCAmelCase=8 , __UpperCAmelCase=37 , __UpperCAmelCase="gelu_new" , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.0 , __UpperCAmelCase=512 , __UpperCAmelCase=3 , __UpperCAmelCase=0.02 , __UpperCAmelCase=3 , __UpperCAmelCase=4 , __UpperCAmelCase=None , __UpperCAmelCase=False , ): '''simple docstring''' __lowerCamelCase = parent __lowerCamelCase = batch_size __lowerCamelCase = seq_length __lowerCamelCase = is_training __lowerCamelCase = use_input_mask __lowerCamelCase = use_token_type_ids __lowerCamelCase = use_labels __lowerCamelCase = vocab_size __lowerCamelCase = block_sizes __lowerCamelCase = num_decoder_layers __lowerCamelCase = d_model __lowerCamelCase = n_head __lowerCamelCase = d_head __lowerCamelCase = d_inner __lowerCamelCase = hidden_act __lowerCamelCase = hidden_dropout __lowerCamelCase = attention_dropout __lowerCamelCase = activation_dropout __lowerCamelCase = max_position_embeddings __lowerCamelCase = type_vocab_size __lowerCamelCase = 2 __lowerCamelCase = num_labels __lowerCamelCase = num_choices __lowerCamelCase = scope __lowerCamelCase = initializer_std # Used in the tests to check the size of the first attention layer __lowerCamelCase = n_head # Used in the tests to check the size of the first hidden state __lowerCamelCase = self.d_model # Used in the tests to check the number of output hidden states/attentions __lowerCamelCase = sum(self.block_sizes ) + (0 if base else self.num_decoder_layers) # FunnelModel adds two hidden layers: input embeddings and the sum of the upsampled encoder hidden state with # the last hidden state of the first block (which is the first hidden state of the decoder). if not base: __lowerCamelCase = self.num_hidden_layers + 2 def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size ) __lowerCamelCase = None if self.use_input_mask: __lowerCamelCase = random_attention_mask([self.batch_size, self.seq_length] ) __lowerCamelCase = None if self.use_token_type_ids: __lowerCamelCase = ids_tensor([self.batch_size, self.seq_length] , self.type_vocab_size ) __lowerCamelCase = None __lowerCamelCase = None __lowerCamelCase = None if self.use_labels: __lowerCamelCase = ids_tensor([self.batch_size] , self.type_sequence_label_size ) __lowerCamelCase = ids_tensor([self.batch_size, self.seq_length] , self.num_labels ) __lowerCamelCase = ids_tensor([self.batch_size] , self.num_choices ) __lowerCamelCase = FunnelConfig( vocab_size=self.vocab_size , block_sizes=self.block_sizes , num_decoder_layers=self.num_decoder_layers , d_model=self.d_model , n_head=self.n_head , d_head=self.d_head , d_inner=self.d_inner , hidden_act=self.hidden_act , hidden_dropout=self.hidden_dropout , attention_dropout=self.attention_dropout , activation_dropout=self.activation_dropout , max_position_embeddings=self.max_position_embeddings , type_vocab_size=self.type_vocab_size , initializer_std=self.initializer_std , ) return ( config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels, ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = TFFunnelModel(config=__UpperCAmelCase ) __lowerCamelCase = {'''input_ids''': input_ids, '''attention_mask''': input_mask, '''token_type_ids''': token_type_ids} __lowerCamelCase = model(__UpperCAmelCase ) __lowerCamelCase = [input_ids, input_mask] __lowerCamelCase = model(__UpperCAmelCase ) __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.d_model) ) __lowerCamelCase = False __lowerCamelCase = TFFunnelModel(config=__UpperCAmelCase ) __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.d_model) ) __lowerCamelCase = False __lowerCamelCase = TFFunnelModel(config=__UpperCAmelCase ) __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.d_model) ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = TFFunnelBaseModel(config=__UpperCAmelCase ) __lowerCamelCase = {'''input_ids''': input_ids, '''attention_mask''': input_mask, '''token_type_ids''': token_type_ids} __lowerCamelCase = model(__UpperCAmelCase ) __lowerCamelCase = [input_ids, input_mask] __lowerCamelCase = model(__UpperCAmelCase ) __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, 2, self.d_model) ) __lowerCamelCase = False __lowerCamelCase = TFFunnelBaseModel(config=__UpperCAmelCase ) __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, 3, self.d_model) ) __lowerCamelCase = False __lowerCamelCase = TFFunnelBaseModel(config=__UpperCAmelCase ) __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, 2, self.d_model) ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = TFFunnelForPreTraining(config=__UpperCAmelCase ) __lowerCamelCase = {'''input_ids''': input_ids, '''attention_mask''': input_mask, '''token_type_ids''': token_type_ids} __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length) ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = TFFunnelForMaskedLM(config=__UpperCAmelCase ) __lowerCamelCase = {'''input_ids''': input_ids, '''attention_mask''': input_mask, '''token_type_ids''': token_type_ids} __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = self.num_labels __lowerCamelCase = TFFunnelForSequenceClassification(config=__UpperCAmelCase ) __lowerCamelCase = {'''input_ids''': input_ids, '''attention_mask''': input_mask, '''token_type_ids''': token_type_ids} __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = self.num_choices __lowerCamelCase = TFFunnelForMultipleChoice(config=__UpperCAmelCase ) __lowerCamelCase = tf.tile(tf.expand_dims(__UpperCAmelCase , 1 ) , (1, self.num_choices, 1) ) __lowerCamelCase = tf.tile(tf.expand_dims(__UpperCAmelCase , 1 ) , (1, self.num_choices, 1) ) __lowerCamelCase = tf.tile(tf.expand_dims(__UpperCAmelCase , 1 ) , (1, self.num_choices, 1) ) __lowerCamelCase = { '''input_ids''': multiple_choice_inputs_ids, '''attention_mask''': multiple_choice_input_mask, '''token_type_ids''': multiple_choice_token_type_ids, } __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = self.num_labels __lowerCamelCase = TFFunnelForTokenClassification(config=__UpperCAmelCase ) __lowerCamelCase = {'''input_ids''': input_ids, '''attention_mask''': input_mask, '''token_type_ids''': token_type_ids} __lowerCamelCase = model(__UpperCAmelCase ) self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) ) def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = TFFunnelForQuestionAnswering(config=__UpperCAmelCase ) __lowerCamelCase = {'''input_ids''': input_ids, '''attention_mask''': input_mask, '''token_type_ids''': token_type_ids} __lowerCamelCase = model(__UpperCAmelCase ) 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 ): '''simple docstring''' __lowerCamelCase = self.prepare_config_and_inputs() ( ( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) , ) = config_and_inputs __lowerCamelCase = {'''input_ids''': input_ids, '''token_type_ids''': token_type_ids, '''attention_mask''': input_mask} return config, inputs_dict @require_tf class __lowerCAmelCase ( lowerCAmelCase__ , lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = ( ( TFFunnelModel, TFFunnelForMaskedLM, TFFunnelForPreTraining, TFFunnelForQuestionAnswering, TFFunnelForTokenClassification, ) if is_tf_available() else () ) lowerCAmelCase__ = ( { """feature-extraction""": (TFFunnelBaseModel, TFFunnelModel), """fill-mask""": TFFunnelForMaskedLM, """question-answering""": TFFunnelForQuestionAnswering, """text-classification""": TFFunnelForSequenceClassification, """token-classification""": TFFunnelForTokenClassification, """zero-shot""": TFFunnelForSequenceClassification, } if is_tf_available() else {} ) lowerCAmelCase__ = False lowerCAmelCase__ = False def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = TFFunnelModelTester(self ) __lowerCamelCase = ConfigTester(self , config_class=__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' self.config_tester.run_common_tests() def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_model(*__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_pretraining(*__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_masked_lm(*__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_token_classification(*__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_question_answering(*__UpperCAmelCase ) @require_tf class __lowerCAmelCase ( lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = ( (TFFunnelBaseModel, TFFunnelForMultipleChoice, TFFunnelForSequenceClassification) if is_tf_available() else () ) lowerCAmelCase__ = False lowerCAmelCase__ = False def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = TFFunnelModelTester(self , base=__UpperCAmelCase ) __lowerCamelCase = ConfigTester(self , config_class=__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' self.config_tester.run_common_tests() def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_base_model(*__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_sequence_classification(*__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.model_tester.prepare_config_and_inputs() self.model_tester.create_and_check_for_multiple_choice(*__UpperCAmelCase )
622
1
import pickle import unittest import torch from accelerate import Accelerator from accelerate.state import AcceleratorState from accelerate.test_utils import require_cpu @require_cpu class __lowerCAmelCase ( unittest.TestCase ): def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = torch.nn.Linear(10 , 10 ) __lowerCamelCase = torch.optim.SGD(model.parameters() , 0.1 ) __lowerCamelCase = Accelerator() __lowerCamelCase = accelerator.prepare(__UpperCAmelCase ) try: pickle.loads(pickle.dumps(__UpperCAmelCase ) ) except Exception as e: self.fail(F"""Accelerated optimizer pickling failed with {e}""" ) AcceleratorState._reset_state()
622
from collections import namedtuple import requests from lxml import html # type: ignore a_ = namedtuple("""covid_data""", """cases deaths recovered""") def a__ ( _UpperCamelCase : str = "https://www.worldometers.info/coronavirus/" ): __lowerCamelCase = '''//div[@class = "maincounter-number"]/span/text()''' return covid_data(*html.fromstring(requests.get(_UpperCamelCase ).content ).xpath(_UpperCamelCase ) ) a_ = """Total COVID-19 cases in the world: {} Total deaths due to COVID-19 in the world: {} Total COVID-19 patients recovered in the world: {}""" print(fmt.format(*covid_stats()))
622
1
import logging import sys from dataclasses import dataclass, field from typing import Any, Dict, List, Optional, Union import librosa import torch from datasets import DatasetDict, load_dataset from packaging import version from torch import nn from transformers import ( HfArgumentParser, Trainer, TrainingArguments, WavaVecaConfig, WavaVecaFeatureExtractor, WavaVecaForPreTraining, is_apex_available, trainer_utils, ) from transformers.models.wavaveca.modeling_wavaveca import _compute_mask_indices if is_apex_available(): from apex import amp if version.parse(version.parse(torch.__version__).base_version) >= version.parse("""1.6"""): a_ = True from torch.cuda.amp import autocast a_ = logging.getLogger(__name__) @dataclass class __lowerCAmelCase : lowerCAmelCase__ = field( metadata={"""help""": """Path to pretrained model or model identifier from huggingface.co/models"""} ) lowerCAmelCase__ = field( default=lowerCAmelCase__ , metadata={"""help""": """Where do you want to store the pretrained models downloaded from huggingface.co"""} , ) lowerCAmelCase__ = field( default=lowerCAmelCase__ , metadata={"""help""": """Whether to freeze the feature extractor layers of the model."""} ) lowerCAmelCase__ = field( default=lowerCAmelCase__ , metadata={"""help""": """Whether to log verbose messages or not."""} , ) lowerCAmelCase__ = field( default=2.0 , metadata={"""help""": """Maximum temperature for gumbel softmax."""} ) lowerCAmelCase__ = field( default=0.5 , metadata={"""help""": """Minimum temperature for gumbel softmax."""} ) lowerCAmelCase__ = field( default=0.99_99_95 , metadata={"""help""": """Decay of gumbel temperature during training."""} ) def a__ ( _UpperCamelCase : ModelArguments ,_UpperCamelCase : TrainingArguments ): logging.basicConfig( format='''%(asctime)s - %(levelname)s - %(name)s - %(message)s''' ,datefmt='''%m/%d/%Y %H:%M:%S''' ,handlers=[logging.StreamHandler(sys.stdout )] ,) __lowerCamelCase = logging.WARNING if model_args.verbose_logging: __lowerCamelCase = logging.DEBUG elif trainer_utils.is_main_process(training_args.local_rank ): __lowerCamelCase = logging.INFO logger.setLevel(_UpperCamelCase ) @dataclass class __lowerCAmelCase : lowerCAmelCase__ = field( default=lowerCAmelCase__ , metadata={"""help""": """The name of the dataset to use (via the datasets library)."""} ) lowerCAmelCase__ = field( default=lowerCAmelCase__ , metadata={"""help""": """The configuration name of the dataset to use (via the datasets library)."""} ) lowerCAmelCase__ = field( default="""train""" , metadata={ """help""": """The name of the training data set split to use (via the datasets library). Defaults to 'train'""" } , ) lowerCAmelCase__ = field( default="""validation""" , metadata={ """help""": ( """The name of the validation data set split to use (via the datasets library). Defaults to 'validation'""" ) } , ) lowerCAmelCase__ = field( default="""file""" , metadata={"""help""": """Column in the dataset that contains speech file path. Defaults to 'file'"""} , ) lowerCAmelCase__ = field( default=lowerCAmelCase__ , metadata={"""help""": """Overwrite the cached preprocessed datasets or not."""} ) lowerCAmelCase__ = field( default=1 , metadata={ """help""": """The percentage of the train set used as validation set in case there's no validation split""" } , ) lowerCAmelCase__ = field( default=lowerCAmelCase__ , metadata={"""help""": """The number of processes to use for the preprocessing."""} , ) lowerCAmelCase__ = field( default=20.0 , metadata={"""help""": """Filter audio files that are longer than `max_duration_in_seconds` seconds"""} ) @dataclass class __lowerCAmelCase : lowerCAmelCase__ = 42 lowerCAmelCase__ = 42 lowerCAmelCase__ = "longest" lowerCAmelCase__ = None lowerCAmelCase__ = None def __call__( self , __UpperCAmelCase ): '''simple docstring''' # reformat list to dict and set to pytorch format __lowerCamelCase = self.feature_extractor.pad( __UpperCAmelCase , max_length=self.max_length , padding=self.padding , pad_to_multiple_of=self.pad_to_multiple_of , return_tensors='''pt''' , ) __lowerCamelCase = self.model._get_feat_extract_output_lengths(batch['''input_values'''].shape[-1] ) __lowerCamelCase = batch['''input_values'''].shape[0] # make sure that no loss is computed on padded inputs if batch["attention_mask"] is not None: # compute real output lengths according to convolution formula __lowerCamelCase = self.model._get_feat_extract_output_lengths(batch['''attention_mask'''].sum(-1 ) ).to( torch.long ) __lowerCamelCase = torch.zeros( (batch_size, mask_indices_seq_length) , dtype=torch.long , device=batch['''input_values'''].device ) # these two operations makes sure that all values # before the output lengths indices are attended to __lowerCamelCase = 1 __lowerCamelCase = attention_mask.flip([-1] ).cumsum(-1 ).flip([-1] ).bool() # sample randomly masked indices __lowerCamelCase = _compute_mask_indices( (batch_size, mask_indices_seq_length) , self.model.config.mask_time_prob , self.model.config.mask_time_length , attention_mask=__UpperCAmelCase , min_masks=2 , ) return batch class __lowerCAmelCase ( lowerCAmelCase__ ): def __init__( self , *__UpperCAmelCase , __UpperCAmelCase=1 , __UpperCAmelCase=0 , __UpperCAmelCase=1.0 , **__UpperCAmelCase ): '''simple docstring''' super().__init__(*__UpperCAmelCase , **__UpperCAmelCase ) __lowerCamelCase = 0 __lowerCamelCase = max_gumbel_temp __lowerCamelCase = min_gumbel_temp __lowerCamelCase = gumbel_temp_decay def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' model.train() __lowerCamelCase = self._prepare_inputs(__UpperCAmelCase ) if self.use_amp: with autocast(): __lowerCamelCase = self.compute_loss(__UpperCAmelCase , __UpperCAmelCase ) else: __lowerCamelCase = self.compute_loss(__UpperCAmelCase , __UpperCAmelCase ) if self.args.n_gpu > 1 or self.deepspeed: if model.module.config.ctc_loss_reduction == "mean": __lowerCamelCase = loss.mean() elif model.module.config.ctc_loss_reduction == "sum": __lowerCamelCase = loss.sum() / (inputs['''mask_time_indices''']).sum() else: raise ValueError(F"""{model.config.ctc_loss_reduction} is not valid. Choose one of ['mean', 'sum']""" ) if self.args.gradient_accumulation_steps > 1: __lowerCamelCase = loss / self.args.gradient_accumulation_steps if self.use_amp: self.scaler.scale(__UpperCAmelCase ).backward() elif self.use_apex: with amp.scale_loss(__UpperCAmelCase , self.optimizer ) as scaled_loss: scaled_loss.backward() elif self.deepspeed: self.deepspeed.backward(__UpperCAmelCase ) else: loss.backward() self.num_update_step += 1 # make sure gumbel softmax temperature is decayed if self.args.n_gpu > 1 or self.deepspeed: model.module.set_gumbel_temperature( max(self.max_gumbel_temp * self.gumbel_temp_decay**self.num_update_step , self.min_gumbel_temp ) ) else: model.set_gumbel_temperature( max(self.max_gumbel_temp * self.gumbel_temp_decay**self.num_update_step , self.min_gumbel_temp ) ) return loss.detach() def a__ ( ): # See all possible arguments in src/transformers/training_args.py # or by passing the --help flag to this script. # We now keep distinct sets of args, for a cleaner separation of concerns. __lowerCamelCase = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments) ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = parser.parse_args_into_dataclasses() configure_logger(_UpperCamelCase ,_UpperCamelCase ) # Downloading and loading a dataset from the hub. __lowerCamelCase = load_dataset(data_args.dataset_name ,data_args.dataset_config_name ,cache_dir=model_args.cache_dir ) if "validation" not in datasets.keys(): # make sure only "validation" and "train" keys remain" __lowerCamelCase = DatasetDict() __lowerCamelCase = load_dataset( data_args.dataset_name ,data_args.dataset_config_name ,split=F"""{data_args.train_split_name}[:{data_args.validation_split_percentage}%]""" ,cache_dir=model_args.cache_dir ,) __lowerCamelCase = load_dataset( data_args.dataset_name ,data_args.dataset_config_name ,split=F"""{data_args.train_split_name}[{data_args.validation_split_percentage}%:]""" ,cache_dir=model_args.cache_dir ,) else: # make sure only "validation" and "train" keys remain" __lowerCamelCase = DatasetDict() __lowerCamelCase = load_dataset( data_args.dataset_name ,data_args.dataset_config_name ,split='''validation''' ,cache_dir=model_args.cache_dir ,) __lowerCamelCase = load_dataset( data_args.dataset_name ,data_args.dataset_config_name ,split=F"""{data_args.train_split_name}""" ,cache_dir=model_args.cache_dir ,) # only normalized-inputs-training is supported __lowerCamelCase = WavaVecaFeatureExtractor.from_pretrained( model_args.model_name_or_path ,cache_dir=model_args.cache_dir ,do_normalize=_UpperCamelCase ) def prepare_dataset(_UpperCamelCase : Optional[Any] ): # check that all files have the correct sampling rate __lowerCamelCase ,__lowerCamelCase = librosa.load(batch[data_args.speech_file_column] ,sr=feature_extractor.sampling_rate ) return batch # load audio files into numpy arrays __lowerCamelCase = datasets.map( _UpperCamelCase ,num_proc=data_args.preprocessing_num_workers ,remove_columns=datasets['''train'''].column_names ) # filter audio files that are too long __lowerCamelCase = vectorized_datasets.filter( lambda _UpperCamelCase : len(data['''speech'''] ) < int(data_args.max_duration_in_seconds * feature_extractor.sampling_rate ) ) def normalize(_UpperCamelCase : List[Any] ): return feature_extractor(batch['''speech'''] ,sampling_rate=feature_extractor.sampling_rate ) # normalize and transform to `BatchFeatures` __lowerCamelCase = vectorized_datasets.map( _UpperCamelCase ,batched=_UpperCamelCase ,num_proc=data_args.preprocessing_num_workers ,load_from_cache_file=not data_args.overwrite_cache ,remove_columns=vectorized_datasets['''train'''].column_names ,) # pretraining is only supported for "newer" stable layer norm architecture # apply_spec_augment has to be True, mask_feature_prob has to be 0.0 __lowerCamelCase = WavaVecaConfig.from_pretrained( model_args.model_name_or_path ,cache_dir=model_args.cache_dir ,gradient_checkpointing=training_args.gradient_checkpointing ,) if not config.do_stable_layer_norm or config.feat_extract_norm != "layer": raise ValueError( '''PreTraining is only supported for ``config.do_stable_layer_norm=True`` and''' ''' ``config.feat_extract_norm=\'layer\'''' ) __lowerCamelCase = WavaVecaForPreTraining(_UpperCamelCase ) __lowerCamelCase = DataCollatorForWavaVecaPretraining(model=_UpperCamelCase ,feature_extractor=_UpperCamelCase ) __lowerCamelCase = WavaVecaPreTrainer( model=_UpperCamelCase ,data_collator=_UpperCamelCase ,args=_UpperCamelCase ,train_dataset=vectorized_datasets['''train'''] ,eval_dataset=vectorized_datasets['''validation'''] ,tokenizer=_UpperCamelCase ,max_gumbel_temp=model_args.max_gumbel_temperature ,min_gumbel_temp=model_args.min_gumbel_temperature ,gumbel_temp_decay=model_args.gumbel_temperature_decay ,) trainer.train() if __name__ == "__main__": main()
622
def a__ ( _UpperCamelCase : str ,_UpperCamelCase : str = " " ): __lowerCamelCase = [] __lowerCamelCase = 0 for index, char in enumerate(_UpperCamelCase ): if char == separator: split_words.append(string[last_index:index] ) __lowerCamelCase = index + 1 elif index + 1 == len(_UpperCamelCase ): split_words.append(string[last_index : index + 1] ) return split_words if __name__ == "__main__": from doctest import testmod testmod()
622
1
import os import unicodedata from shutil import copyfile from typing import Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import AddedToken, PreTrainedTokenizer from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"""vocab_file""": """spiece.model"""} a_ = { """vocab_file""": { """albert-base-v1""": """https://huggingface.co/albert-base-v1/resolve/main/spiece.model""", """albert-large-v1""": """https://huggingface.co/albert-large-v1/resolve/main/spiece.model""", """albert-xlarge-v1""": """https://huggingface.co/albert-xlarge-v1/resolve/main/spiece.model""", """albert-xxlarge-v1""": """https://huggingface.co/albert-xxlarge-v1/resolve/main/spiece.model""", """albert-base-v2""": """https://huggingface.co/albert-base-v2/resolve/main/spiece.model""", """albert-large-v2""": """https://huggingface.co/albert-large-v2/resolve/main/spiece.model""", """albert-xlarge-v2""": """https://huggingface.co/albert-xlarge-v2/resolve/main/spiece.model""", """albert-xxlarge-v2""": """https://huggingface.co/albert-xxlarge-v2/resolve/main/spiece.model""", } } a_ = { """albert-base-v1""": 512, """albert-large-v1""": 512, """albert-xlarge-v1""": 512, """albert-xxlarge-v1""": 512, """albert-base-v2""": 512, """albert-large-v2""": 512, """albert-xlarge-v2""": 512, """albert-xxlarge-v2""": 512, } a_ = """▁""" class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = VOCAB_FILES_NAMES lowerCAmelCase__ = PRETRAINED_VOCAB_FILES_MAP lowerCAmelCase__ = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES def __init__( self , __UpperCAmelCase , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=False , __UpperCAmelCase="[CLS]" , __UpperCAmelCase="[SEP]" , __UpperCAmelCase="<unk>" , __UpperCAmelCase="[SEP]" , __UpperCAmelCase="<pad>" , __UpperCAmelCase="[CLS]" , __UpperCAmelCase="[MASK]" , __UpperCAmelCase = None , **__UpperCAmelCase , ): '''simple docstring''' # Mask token behave like a normal word, i.e. include the space before it and # is included in the raw text, there should be a match in a non-normalized sentence. __lowerCamelCase = ( AddedToken(__UpperCAmelCase , lstrip=__UpperCAmelCase , rstrip=__UpperCAmelCase , normalized=__UpperCAmelCase ) if isinstance(__UpperCAmelCase , __UpperCAmelCase ) else mask_token ) __lowerCamelCase = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( do_lower_case=__UpperCAmelCase , remove_space=__UpperCAmelCase , keep_accents=__UpperCAmelCase , bos_token=__UpperCAmelCase , eos_token=__UpperCAmelCase , unk_token=__UpperCAmelCase , sep_token=__UpperCAmelCase , pad_token=__UpperCAmelCase , cls_token=__UpperCAmelCase , mask_token=__UpperCAmelCase , sp_model_kwargs=self.sp_model_kwargs , **__UpperCAmelCase , ) __lowerCamelCase = do_lower_case __lowerCamelCase = remove_space __lowerCamelCase = keep_accents __lowerCamelCase = vocab_file __lowerCamelCase = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(__UpperCAmelCase ) @property def lowerCamelCase ( self ): '''simple docstring''' return len(self.sp_model ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = {self.convert_ids_to_tokens(__UpperCAmelCase ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def __getstate__( self ): '''simple docstring''' __lowerCamelCase = self.__dict__.copy() __lowerCamelCase = None return state def __setstate__( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = d # for backward compatibility if not hasattr(self , '''sp_model_kwargs''' ): __lowerCamelCase = {} __lowerCamelCase = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' if self.remove_space: __lowerCamelCase = ''' '''.join(inputs.strip().split() ) else: __lowerCamelCase = inputs __lowerCamelCase = outputs.replace('''``''' , '''"''' ).replace('''\'\'''' , '''"''' ) if not self.keep_accents: __lowerCamelCase = unicodedata.normalize('''NFKD''' , __UpperCAmelCase ) __lowerCamelCase = ''''''.join([c for c in outputs if not unicodedata.combining(__UpperCAmelCase )] ) if self.do_lower_case: __lowerCamelCase = outputs.lower() return outputs def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.preprocess_text(__UpperCAmelCase ) __lowerCamelCase = self.sp_model.encode(__UpperCAmelCase , out_type=__UpperCAmelCase ) __lowerCamelCase = [] for piece in pieces: if len(__UpperCAmelCase ) > 1 and piece[-1] == str(''',''' ) and piece[-2].isdigit(): __lowerCamelCase = self.sp_model.EncodeAsPieces(piece[:-1].replace(__UpperCAmelCase , '''''' ) ) if piece[0] != SPIECE_UNDERLINE and cur_pieces[0][0] == SPIECE_UNDERLINE: if len(cur_pieces[0] ) == 1: __lowerCamelCase = cur_pieces[1:] else: __lowerCamelCase = cur_pieces[0][1:] cur_pieces.append(piece[-1] ) new_pieces.extend(__UpperCAmelCase ) else: new_pieces.append(__UpperCAmelCase ) return new_pieces def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.sp_model.PieceToId(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.sp_model.IdToPiece(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = [] __lowerCamelCase = '''''' __lowerCamelCase = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(__UpperCAmelCase ) + token __lowerCamelCase = True __lowerCamelCase = [] else: current_sub_tokens.append(__UpperCAmelCase ) __lowerCamelCase = False out_string += self.sp_model.decode(__UpperCAmelCase ) return out_string.strip() def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' __lowerCamelCase = [self.sep_token_id] __lowerCamelCase = [self.cls_token_id] if token_ids_a is None: return cls + token_ids_a + sep return cls + token_ids_a + sep + token_ids_a + sep def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None , __UpperCAmelCase = False ): '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=__UpperCAmelCase , token_ids_a=__UpperCAmelCase , already_has_special_tokens=__UpperCAmelCase ) if token_ids_a is not None: return [1] + ([0] * len(__UpperCAmelCase )) + [1] + ([0] * len(__UpperCAmelCase )) + [1] return [1] + ([0] * len(__UpperCAmelCase )) + [1] def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' __lowerCamelCase = [self.sep_token_id] __lowerCamelCase = [self.cls_token_id] if token_ids_a is None: return len(cls + token_ids_a + sep ) * [0] return len(cls + token_ids_a + sep ) * [0] + len(token_ids_a + sep ) * [1] def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' if not os.path.isdir(__UpperCAmelCase ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return __lowerCamelCase = os.path.join( __UpperCAmelCase , (filename_prefix + '''-''' if filename_prefix else '''''') + VOCAB_FILES_NAMES['''vocab_file'''] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(__UpperCAmelCase ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , __UpperCAmelCase ) elif not os.path.isfile(self.vocab_file ): with open(__UpperCAmelCase , '''wb''' ) as fi: __lowerCamelCase = self.sp_model.serialized_model_proto() fi.write(__UpperCAmelCase ) return (out_vocab_file,)
622
import json import os import pickle import shutil import tempfile from unittest import TestCase from unittest.mock import patch import numpy as np from datasets import Dataset from transformers import is_faiss_available from transformers.models.bart.configuration_bart import BartConfig from transformers.models.bart.tokenization_bart import BartTokenizer from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES from transformers.models.dpr.configuration_dpr import DPRConfig from transformers.models.dpr.tokenization_dpr import DPRContextEncoderTokenizer, DPRQuestionEncoderTokenizer from transformers.models.rag.configuration_rag import RagConfig from transformers.models.rag.retrieval_rag import CustomHFIndex, RagRetriever from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES from transformers.testing_utils import require_faiss, require_sentencepiece, require_tokenizers, require_torch if is_faiss_available(): import faiss @require_faiss class __lowerCAmelCase ( lowerCAmelCase__ ): def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = tempfile.mkdtemp() __lowerCamelCase = 8 # DPR tok __lowerCamelCase = [ '''[UNK]''', '''[CLS]''', '''[SEP]''', '''[PAD]''', '''[MASK]''', '''want''', '''##want''', '''##ed''', '''wa''', '''un''', '''runn''', '''##ing''', ''',''', '''low''', '''lowest''', ] __lowerCamelCase = os.path.join(self.tmpdirname , '''dpr_tokenizer''' ) os.makedirs(__UpperCAmelCase , exist_ok=__UpperCAmelCase ) __lowerCamelCase = os.path.join(__UpperCAmelCase , DPR_VOCAB_FILES_NAMES['''vocab_file'''] ) with open(self.vocab_file , '''w''' , encoding='''utf-8''' ) as vocab_writer: vocab_writer.write(''''''.join([x + '''\n''' for x in vocab_tokens] ) ) # BART tok __lowerCamelCase = [ '''l''', '''o''', '''w''', '''e''', '''r''', '''s''', '''t''', '''i''', '''d''', '''n''', '''\u0120''', '''\u0120l''', '''\u0120n''', '''\u0120lo''', '''\u0120low''', '''er''', '''\u0120lowest''', '''\u0120newer''', '''\u0120wider''', '''<unk>''', ] __lowerCamelCase = dict(zip(__UpperCAmelCase , range(len(__UpperCAmelCase ) ) ) ) __lowerCamelCase = ['''#version: 0.2''', '''\u0120 l''', '''\u0120l o''', '''\u0120lo w''', '''e r''', ''''''] __lowerCamelCase = {'''unk_token''': '''<unk>'''} __lowerCamelCase = os.path.join(self.tmpdirname , '''bart_tokenizer''' ) os.makedirs(__UpperCAmelCase , exist_ok=__UpperCAmelCase ) __lowerCamelCase = os.path.join(__UpperCAmelCase , BART_VOCAB_FILES_NAMES['''vocab_file'''] ) __lowerCamelCase = os.path.join(__UpperCAmelCase , BART_VOCAB_FILES_NAMES['''merges_file'''] ) with open(self.vocab_file , '''w''' , encoding='''utf-8''' ) as fp: fp.write(json.dumps(__UpperCAmelCase ) + '''\n''' ) with open(self.merges_file , '''w''' , encoding='''utf-8''' ) as fp: fp.write('''\n'''.join(__UpperCAmelCase ) ) def lowerCamelCase ( self ): '''simple docstring''' return DPRQuestionEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , '''dpr_tokenizer''' ) ) def lowerCamelCase ( self ): '''simple docstring''' return DPRContextEncoderTokenizer.from_pretrained(os.path.join(self.tmpdirname , '''dpr_tokenizer''' ) ) def lowerCamelCase ( self ): '''simple docstring''' return BartTokenizer.from_pretrained(os.path.join(self.tmpdirname , '''bart_tokenizer''' ) ) def lowerCamelCase ( self ): '''simple docstring''' shutil.rmtree(self.tmpdirname ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = Dataset.from_dict( { '''id''': ['''0''', '''1'''], '''text''': ['''foo''', '''bar'''], '''title''': ['''Foo''', '''Bar'''], '''embeddings''': [np.ones(self.retrieval_vector_size ), 2 * np.ones(self.retrieval_vector_size )], } ) dataset.add_faiss_index('''embeddings''' , string_factory='''Flat''' , metric_type=faiss.METRIC_INNER_PRODUCT ) return dataset def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_dataset() __lowerCamelCase = RagConfig( retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , ) with patch('''transformers.models.rag.retrieval_rag.load_dataset''' ) as mock_load_dataset: __lowerCamelCase = dataset __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , ) return retriever def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.get_dummy_dataset() __lowerCamelCase = RagConfig( retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , index_name='''custom''' , ) if from_disk: __lowerCamelCase = os.path.join(self.tmpdirname , '''dataset''' ) __lowerCamelCase = os.path.join(self.tmpdirname , '''index.faiss''' ) dataset.get_index('''embeddings''' ).save(os.path.join(self.tmpdirname , '''index.faiss''' ) ) dataset.drop_index('''embeddings''' ) dataset.save_to_disk(os.path.join(self.tmpdirname , '''dataset''' ) ) del dataset __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , ) else: __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() , index=CustomHFIndex(config.retrieval_vector_size , __UpperCAmelCase ) , ) return retriever def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = Dataset.from_dict( { '''id''': ['''0''', '''1'''], '''text''': ['''foo''', '''bar'''], '''title''': ['''Foo''', '''Bar'''], '''embeddings''': [np.ones(self.retrieval_vector_size + 1 ), 2 * np.ones(self.retrieval_vector_size + 1 )], } ) dataset.add_faiss_index('''embeddings''' , string_factory='''Flat''' , metric_type=faiss.METRIC_INNER_PRODUCT ) __lowerCamelCase = os.path.join(self.tmpdirname , '''hf_bert_base.hnswSQ8_correct_phi_128.c_index''' ) dataset.save_faiss_index('''embeddings''' , index_file_name + '''.index.dpr''' ) pickle.dump(dataset['''id'''] , open(index_file_name + '''.index_meta.dpr''' , '''wb''' ) ) __lowerCamelCase = os.path.join(self.tmpdirname , '''psgs_w100.tsv.pkl''' ) __lowerCamelCase = {sample['''id''']: [sample['''text'''], sample['''title''']] for sample in dataset} pickle.dump(__UpperCAmelCase , open(__UpperCAmelCase , '''wb''' ) ) __lowerCamelCase = RagConfig( retrieval_vector_size=self.retrieval_vector_size , question_encoder=DPRConfig().to_dict() , generator=BartConfig().to_dict() , index_name='''legacy''' , index_path=self.tmpdirname , ) __lowerCamelCase = RagRetriever( __UpperCAmelCase , question_encoder_tokenizer=self.get_dpr_tokenizer() , generator_tokenizer=self.get_bart_tokenizer() ) return retriever def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_canonical_hf_index_retriever() __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''embeddings''', '''id''', '''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''id'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''id'''][0] , '''1''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''id'''][0] , '''0''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_canonical_hf_index_retriever() with tempfile.TemporaryDirectory() as tmp_dirname: with patch('''transformers.models.rag.retrieval_rag.load_dataset''' ) as mock_load_dataset: __lowerCamelCase = self.get_dummy_dataset() retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''embeddings''', '''id''', '''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''id'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''id'''][0] , '''1''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''id'''][0] , '''0''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) with tempfile.TemporaryDirectory() as tmp_dirname: retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''embeddings''', '''id''', '''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''id'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''id'''][0] , '''1''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''id'''][0] , '''0''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) with tempfile.TemporaryDirectory() as tmp_dirname: retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_legacy_index_retriever() __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=__UpperCAmelCase ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertEqual(len(__UpperCAmelCase ) , 2 ) self.assertEqual(sorted(doc_dicts[0] ) , ['''text''', '''title'''] ) self.assertEqual(len(doc_dicts[0]['''text'''] ) , __UpperCAmelCase ) self.assertEqual(doc_dicts[0]['''text'''][0] , '''bar''' ) # max inner product is reached with second doc self.assertEqual(doc_dicts[1]['''text'''][0] , '''foo''' ) # max inner product is reached with first doc self.assertListEqual(doc_ids.tolist() , [[1], [0]] ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_legacy_index_retriever() with tempfile.TemporaryDirectory() as tmp_dirname: retriever.save_pretrained(__UpperCAmelCase ) __lowerCamelCase = RagRetriever.from_pretrained(__UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever.retrieve(__UpperCAmelCase , n_docs=1 ) self.assertTrue(out is not None ) @require_torch @require_tokenizers @require_sentencepiece def lowerCamelCase ( self ): '''simple docstring''' import torch __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_canonical_hf_index_retriever() __lowerCamelCase = [[5, 7], [10, 11]] __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever(__UpperCAmelCase , __UpperCAmelCase , prefix=retriever.config.generator.prefix , n_docs=__UpperCAmelCase ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = ( out['''context_input_ids'''], out['''context_attention_mask'''], out['''retrieved_doc_embeds'''], ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , __UpperCAmelCase ) self.assertIsInstance(__UpperCAmelCase , np.ndarray ) __lowerCamelCase = retriever( __UpperCAmelCase , __UpperCAmelCase , prefix=retriever.config.generator.prefix , n_docs=__UpperCAmelCase , return_tensors='''pt''' , ) __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = ( # noqa: F841 out['''context_input_ids'''], out['''context_attention_mask'''], out['''retrieved_doc_embeds'''], out['''doc_ids'''], ) self.assertEqual(retrieved_doc_embeds.shape , (2, n_docs, self.retrieval_vector_size) ) self.assertIsInstance(__UpperCAmelCase , torch.Tensor ) self.assertIsInstance(__UpperCAmelCase , torch.Tensor ) self.assertIsInstance(__UpperCAmelCase , torch.Tensor ) @require_torch @require_tokenizers @require_sentencepiece def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dpr_ctx_encoder_tokenizer() __lowerCamelCase = 1 __lowerCamelCase = self.get_dummy_custom_hf_index_retriever(from_disk=__UpperCAmelCase ) retriever.set_ctx_encoder_tokenizer(__UpperCAmelCase ) __lowerCamelCase = [[5, 7], [10, 11]] __lowerCamelCase = np.array( [np.ones(self.retrieval_vector_size ), -np.ones(self.retrieval_vector_size )] , dtype=np.floataa ) __lowerCamelCase = retriever(__UpperCAmelCase , __UpperCAmelCase , prefix=retriever.config.generator.prefix , n_docs=__UpperCAmelCase ) self.assertEqual( len(__UpperCAmelCase ) , 6 ) # check whether the retriever output consist of 6 attributes including tokenized docs self.assertEqual( all(k in out for k in ('''tokenized_doc_ids''', '''tokenized_doc_attention_mask''') ) , __UpperCAmelCase ) # check for doc token related keys in dictionary.
622
1
import math import numpy as np import qiskit from qiskit import Aer, ClassicalRegister, QuantumCircuit, QuantumRegister, execute def a__ ( _UpperCamelCase : int = 3 ): if isinstance(_UpperCamelCase ,_UpperCamelCase ): raise TypeError('''number of qubits must be a integer.''' ) if number_of_qubits <= 0: raise ValueError('''number of qubits must be > 0.''' ) if math.floor(_UpperCamelCase ) != number_of_qubits: raise ValueError('''number of qubits must be exact integer.''' ) if number_of_qubits > 10: raise ValueError('''number of qubits too large to simulate(>10).''' ) __lowerCamelCase = QuantumRegister(_UpperCamelCase ,'''qr''' ) __lowerCamelCase = ClassicalRegister(_UpperCamelCase ,'''cr''' ) __lowerCamelCase = QuantumCircuit(_UpperCamelCase ,_UpperCamelCase ) __lowerCamelCase = number_of_qubits for i in range(_UpperCamelCase ): quantum_circuit.h(number_of_qubits - i - 1 ) counter -= 1 for j in range(_UpperCamelCase ): quantum_circuit.cp(np.pi / 2 ** (counter - j) ,_UpperCamelCase ,_UpperCamelCase ) for k in range(number_of_qubits // 2 ): quantum_circuit.swap(_UpperCamelCase ,number_of_qubits - k - 1 ) # measure all the qubits quantum_circuit.measure(_UpperCamelCase ,_UpperCamelCase ) # simulate with 10000 shots __lowerCamelCase = Aer.get_backend('''qasm_simulator''' ) __lowerCamelCase = execute(_UpperCamelCase ,_UpperCamelCase ,shots=1_00_00 ) return job.result().get_counts(_UpperCamelCase ) if __name__ == "__main__": print( f"Total count for quantum fourier transform state is: \ {quantum_fourier_transform(3)}" )
622
from collections import OrderedDict from typing import Mapping from packaging import version from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { """sail/poolformer_s12""": """https://huggingface.co/sail/poolformer_s12/resolve/main/config.json""", # See all PoolFormer models at https://huggingface.co/models?filter=poolformer } class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = """poolformer""" def __init__( self , __UpperCAmelCase=3 , __UpperCAmelCase=16 , __UpperCAmelCase=16 , __UpperCAmelCase=3 , __UpperCAmelCase=4.0 , __UpperCAmelCase=[2, 2, 6, 2] , __UpperCAmelCase=[64, 128, 320, 512] , __UpperCAmelCase=[7, 3, 3, 3] , __UpperCAmelCase=[4, 2, 2, 2] , __UpperCAmelCase=[2, 1, 1, 1] , __UpperCAmelCase=4 , __UpperCAmelCase=0.0 , __UpperCAmelCase="gelu" , __UpperCAmelCase=True , __UpperCAmelCase=1E-5 , __UpperCAmelCase=0.02 , **__UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = num_channels __lowerCamelCase = patch_size __lowerCamelCase = stride __lowerCamelCase = padding __lowerCamelCase = pool_size __lowerCamelCase = hidden_sizes __lowerCamelCase = mlp_ratio __lowerCamelCase = depths __lowerCamelCase = patch_sizes __lowerCamelCase = strides __lowerCamelCase = num_encoder_blocks __lowerCamelCase = drop_path_rate __lowerCamelCase = hidden_act __lowerCamelCase = use_layer_scale __lowerCamelCase = layer_scale_init_value __lowerCamelCase = initializer_range super().__init__(**__UpperCAmelCase ) class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = version.parse("""1.11""" ) @property def lowerCamelCase ( self ): '''simple docstring''' return OrderedDict( [ ('''pixel_values''', {0: '''batch''', 1: '''num_channels''', 2: '''height''', 3: '''width'''}), ] ) @property def lowerCamelCase ( self ): '''simple docstring''' return 2E-3
622
1
import inspect import re 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_config_docstrings.py a_ = """src/transformers""" # This is to make sure the transformers module imported is the one in the repo. a_ = direct_transformers_import(PATH_TO_TRANSFORMERS) a_ = transformers.models.auto.configuration_auto.CONFIG_MAPPING # Regex pattern used to find the checkpoint mentioned in the docstring of `config_class`. # For example, `[bert-base-uncased](https://huggingface.co/bert-base-uncased)` a_ = re.compile(R"""\[(.+?)\]\((https://huggingface\.co/.+?)\)""") a_ = { """DecisionTransformerConfig""", """EncoderDecoderConfig""", """MusicgenConfig""", """RagConfig""", """SpeechEncoderDecoderConfig""", """TimmBackboneConfig""", """VisionEncoderDecoderConfig""", """VisionTextDualEncoderConfig""", """LlamaConfig""", } def a__ ( _UpperCamelCase : Union[str, Any] ): __lowerCamelCase = None # source code of `config_class` __lowerCamelCase = inspect.getsource(_UpperCamelCase ) __lowerCamelCase = _re_checkpoint.findall(_UpperCamelCase ) # Each `checkpoint` is a tuple of a checkpoint name and a checkpoint link. # For example, `('bert-base-uncased', 'https://huggingface.co/bert-base-uncased')` for ckpt_name, ckpt_link in checkpoints: # allow the link to end with `/` if ckpt_link.endswith('''/''' ): __lowerCamelCase = ckpt_link[:-1] # verify the checkpoint name corresponds to the checkpoint link __lowerCamelCase = F"""https://huggingface.co/{ckpt_name}""" if ckpt_link == ckpt_link_from_name: __lowerCamelCase = ckpt_name break return checkpoint def a__ ( ): __lowerCamelCase = [] for config_class in list(CONFIG_MAPPING.values() ): # Skip deprecated models if "models.deprecated" in config_class.__module__: continue __lowerCamelCase = get_checkpoint_from_config_class(_UpperCamelCase ) __lowerCamelCase = config_class.__name__ if checkpoint is None and name not in CONFIG_CLASSES_TO_IGNORE_FOR_DOCSTRING_CHECKPOINT_CHECK: configs_without_checkpoint.append(_UpperCamelCase ) if len(_UpperCamelCase ) > 0: __lowerCamelCase = '''\n'''.join(sorted(_UpperCamelCase ) ) raise ValueError(F"""The following configurations don't contain any valid checkpoint:\n{message}""" ) if __name__ == "__main__": check_config_docstrings_have_checkpoints()
622
from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { """uclanlp/visualbert-vqa""": """https://huggingface.co/uclanlp/visualbert-vqa/resolve/main/config.json""", """uclanlp/visualbert-vqa-pre""": """https://huggingface.co/uclanlp/visualbert-vqa-pre/resolve/main/config.json""", """uclanlp/visualbert-vqa-coco-pre""": ( """https://huggingface.co/uclanlp/visualbert-vqa-coco-pre/resolve/main/config.json""" ), """uclanlp/visualbert-vcr""": """https://huggingface.co/uclanlp/visualbert-vcr/resolve/main/config.json""", """uclanlp/visualbert-vcr-pre""": """https://huggingface.co/uclanlp/visualbert-vcr-pre/resolve/main/config.json""", """uclanlp/visualbert-vcr-coco-pre""": ( """https://huggingface.co/uclanlp/visualbert-vcr-coco-pre/resolve/main/config.json""" ), """uclanlp/visualbert-nlvr2""": """https://huggingface.co/uclanlp/visualbert-nlvr2/resolve/main/config.json""", """uclanlp/visualbert-nlvr2-pre""": """https://huggingface.co/uclanlp/visualbert-nlvr2-pre/resolve/main/config.json""", """uclanlp/visualbert-nlvr2-coco-pre""": ( """https://huggingface.co/uclanlp/visualbert-nlvr2-coco-pre/resolve/main/config.json""" ) # See all VisualBERT models at https://huggingface.co/models?filter=visual_bert } class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = """visual_bert""" def __init__( self , __UpperCAmelCase=30522 , __UpperCAmelCase=768 , __UpperCAmelCase=512 , __UpperCAmelCase=12 , __UpperCAmelCase=12 , __UpperCAmelCase=3072 , __UpperCAmelCase="gelu" , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=512 , __UpperCAmelCase=2 , __UpperCAmelCase=0.02 , __UpperCAmelCase=1E-1_2 , __UpperCAmelCase=False , __UpperCAmelCase=True , __UpperCAmelCase=1 , __UpperCAmelCase=0 , __UpperCAmelCase=2 , **__UpperCAmelCase , ): '''simple docstring''' super().__init__(pad_token_id=__UpperCAmelCase , bos_token_id=__UpperCAmelCase , eos_token_id=__UpperCAmelCase , **__UpperCAmelCase ) __lowerCamelCase = vocab_size __lowerCamelCase = max_position_embeddings __lowerCamelCase = hidden_size __lowerCamelCase = visual_embedding_dim __lowerCamelCase = num_hidden_layers __lowerCamelCase = num_attention_heads __lowerCamelCase = intermediate_size __lowerCamelCase = hidden_act __lowerCamelCase = hidden_dropout_prob __lowerCamelCase = attention_probs_dropout_prob __lowerCamelCase = initializer_range __lowerCamelCase = type_vocab_size __lowerCamelCase = layer_norm_eps __lowerCamelCase = bypass_transformer __lowerCamelCase = special_visual_initialize
622
1
import string import numpy def a__ ( _UpperCamelCase : int ,_UpperCamelCase : int ): return b if a == 0 else greatest_common_divisor(b % a ,_UpperCamelCase ) class __lowerCAmelCase : lowerCAmelCase__ = 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) lowerCAmelCase__ = numpy.vectorize(lambda lowerCAmelCase__ : x % 3_6 ) lowerCAmelCase__ = numpy.vectorize(lowerCAmelCase__ ) def __init__( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.modulus(__UpperCAmelCase ) # mod36 calc's on the encrypt key self.check_determinant() # validate the determinant of the encryption key __lowerCamelCase = encrypt_key.shape[0] def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.key_string.index(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.key_string[round(__UpperCAmelCase )] def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = round(numpy.linalg.det(self.encrypt_key ) ) if det < 0: __lowerCamelCase = det % len(self.key_string ) __lowerCamelCase = len(self.key_string ) if greatest_common_divisor(__UpperCAmelCase , len(self.key_string ) ) != 1: __lowerCamelCase = ( F"""determinant modular {req_l} of encryption key({det}) """ F"""is not co prime w.r.t {req_l}.\nTry another key.""" ) raise ValueError(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = [char for char in text.upper() if char in self.key_string] __lowerCamelCase = chars[-1] while len(__UpperCAmelCase ) % self.break_key != 0: chars.append(__UpperCAmelCase ) return "".join(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.process_text(text.upper() ) __lowerCamelCase = '''''' for i in range(0 , len(__UpperCAmelCase ) - self.break_key + 1 , self.break_key ): __lowerCamelCase = text[i : i + self.break_key] __lowerCamelCase = [self.replace_letters(__UpperCAmelCase ) for char in batch] __lowerCamelCase = numpy.array([vec] ).T __lowerCamelCase = self.modulus(self.encrypt_key.dot(__UpperCAmelCase ) ).T.tolist()[ 0 ] __lowerCamelCase = ''''''.join( self.replace_digits(__UpperCAmelCase ) for num in batch_encrypted ) encrypted += encrypted_batch return encrypted def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = round(numpy.linalg.det(self.encrypt_key ) ) if det < 0: __lowerCamelCase = det % len(self.key_string ) __lowerCamelCase = None for i in range(len(self.key_string ) ): if (det * i) % len(self.key_string ) == 1: __lowerCamelCase = i break __lowerCamelCase = ( det_inv * numpy.linalg.det(self.encrypt_key ) * numpy.linalg.inv(self.encrypt_key ) ) return self.to_int(self.modulus(__UpperCAmelCase ) ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.make_decrypt_key() __lowerCamelCase = self.process_text(text.upper() ) __lowerCamelCase = '''''' for i in range(0 , len(__UpperCAmelCase ) - self.break_key + 1 , self.break_key ): __lowerCamelCase = text[i : i + self.break_key] __lowerCamelCase = [self.replace_letters(__UpperCAmelCase ) for char in batch] __lowerCamelCase = numpy.array([vec] ).T __lowerCamelCase = self.modulus(decrypt_key.dot(__UpperCAmelCase ) ).T.tolist()[0] __lowerCamelCase = ''''''.join( self.replace_digits(__UpperCAmelCase ) for num in batch_decrypted ) decrypted += decrypted_batch return decrypted def a__ ( ): __lowerCamelCase = int(input('''Enter the order of the encryption key: ''' ) ) __lowerCamelCase = [] print('''Enter each row of the encryption key with space separated integers''' ) for _ in range(_UpperCamelCase ): __lowerCamelCase = [int(_UpperCamelCase ) for x in input().split()] hill_matrix.append(_UpperCamelCase ) __lowerCamelCase = HillCipher(numpy.array(_UpperCamelCase ) ) print('''Would you like to encrypt or decrypt some text? (1 or 2)''' ) __lowerCamelCase = input('''\n1. Encrypt\n2. Decrypt\n''' ) if option == "1": __lowerCamelCase = input('''What text would you like to encrypt?: ''' ) print('''Your encrypted text is:''' ) print(hc.encrypt(_UpperCamelCase ) ) elif option == "2": __lowerCamelCase = input('''What text would you like to decrypt?: ''' ) print('''Your decrypted text is:''' ) print(hc.decrypt(_UpperCamelCase ) ) if __name__ == "__main__": import doctest doctest.testmod() main()
622
import os import unicodedata from shutil import copyfile from typing import Any, Dict, List, Optional, Tuple import sentencepiece as spm from ...tokenization_utils import AddedToken, PreTrainedTokenizer from ...utils import logging a_ = logging.get_logger(__name__) a_ = {"""vocab_file""": """spiece.model"""} a_ = { """vocab_file""": { """albert-base-v1""": """https://huggingface.co/albert-base-v1/resolve/main/spiece.model""", """albert-large-v1""": """https://huggingface.co/albert-large-v1/resolve/main/spiece.model""", """albert-xlarge-v1""": """https://huggingface.co/albert-xlarge-v1/resolve/main/spiece.model""", """albert-xxlarge-v1""": """https://huggingface.co/albert-xxlarge-v1/resolve/main/spiece.model""", """albert-base-v2""": """https://huggingface.co/albert-base-v2/resolve/main/spiece.model""", """albert-large-v2""": """https://huggingface.co/albert-large-v2/resolve/main/spiece.model""", """albert-xlarge-v2""": """https://huggingface.co/albert-xlarge-v2/resolve/main/spiece.model""", """albert-xxlarge-v2""": """https://huggingface.co/albert-xxlarge-v2/resolve/main/spiece.model""", } } a_ = { """albert-base-v1""": 512, """albert-large-v1""": 512, """albert-xlarge-v1""": 512, """albert-xxlarge-v1""": 512, """albert-base-v2""": 512, """albert-large-v2""": 512, """albert-xlarge-v2""": 512, """albert-xxlarge-v2""": 512, } a_ = """▁""" class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = VOCAB_FILES_NAMES lowerCAmelCase__ = PRETRAINED_VOCAB_FILES_MAP lowerCAmelCase__ = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES def __init__( self , __UpperCAmelCase , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase=False , __UpperCAmelCase="[CLS]" , __UpperCAmelCase="[SEP]" , __UpperCAmelCase="<unk>" , __UpperCAmelCase="[SEP]" , __UpperCAmelCase="<pad>" , __UpperCAmelCase="[CLS]" , __UpperCAmelCase="[MASK]" , __UpperCAmelCase = None , **__UpperCAmelCase , ): '''simple docstring''' # Mask token behave like a normal word, i.e. include the space before it and # is included in the raw text, there should be a match in a non-normalized sentence. __lowerCamelCase = ( AddedToken(__UpperCAmelCase , lstrip=__UpperCAmelCase , rstrip=__UpperCAmelCase , normalized=__UpperCAmelCase ) if isinstance(__UpperCAmelCase , __UpperCAmelCase ) else mask_token ) __lowerCamelCase = {} if sp_model_kwargs is None else sp_model_kwargs super().__init__( do_lower_case=__UpperCAmelCase , remove_space=__UpperCAmelCase , keep_accents=__UpperCAmelCase , bos_token=__UpperCAmelCase , eos_token=__UpperCAmelCase , unk_token=__UpperCAmelCase , sep_token=__UpperCAmelCase , pad_token=__UpperCAmelCase , cls_token=__UpperCAmelCase , mask_token=__UpperCAmelCase , sp_model_kwargs=self.sp_model_kwargs , **__UpperCAmelCase , ) __lowerCamelCase = do_lower_case __lowerCamelCase = remove_space __lowerCamelCase = keep_accents __lowerCamelCase = vocab_file __lowerCamelCase = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(__UpperCAmelCase ) @property def lowerCamelCase ( self ): '''simple docstring''' return len(self.sp_model ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = {self.convert_ids_to_tokens(__UpperCAmelCase ): i for i in range(self.vocab_size )} vocab.update(self.added_tokens_encoder ) return vocab def __getstate__( self ): '''simple docstring''' __lowerCamelCase = self.__dict__.copy() __lowerCamelCase = None return state def __setstate__( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = d # for backward compatibility if not hasattr(self , '''sp_model_kwargs''' ): __lowerCamelCase = {} __lowerCamelCase = spm.SentencePieceProcessor(**self.sp_model_kwargs ) self.sp_model.Load(self.vocab_file ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' if self.remove_space: __lowerCamelCase = ''' '''.join(inputs.strip().split() ) else: __lowerCamelCase = inputs __lowerCamelCase = outputs.replace('''``''' , '''"''' ).replace('''\'\'''' , '''"''' ) if not self.keep_accents: __lowerCamelCase = unicodedata.normalize('''NFKD''' , __UpperCAmelCase ) __lowerCamelCase = ''''''.join([c for c in outputs if not unicodedata.combining(__UpperCAmelCase )] ) if self.do_lower_case: __lowerCamelCase = outputs.lower() return outputs def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = self.preprocess_text(__UpperCAmelCase ) __lowerCamelCase = self.sp_model.encode(__UpperCAmelCase , out_type=__UpperCAmelCase ) __lowerCamelCase = [] for piece in pieces: if len(__UpperCAmelCase ) > 1 and piece[-1] == str(''',''' ) and piece[-2].isdigit(): __lowerCamelCase = self.sp_model.EncodeAsPieces(piece[:-1].replace(__UpperCAmelCase , '''''' ) ) if piece[0] != SPIECE_UNDERLINE and cur_pieces[0][0] == SPIECE_UNDERLINE: if len(cur_pieces[0] ) == 1: __lowerCamelCase = cur_pieces[1:] else: __lowerCamelCase = cur_pieces[0][1:] cur_pieces.append(piece[-1] ) new_pieces.extend(__UpperCAmelCase ) else: new_pieces.append(__UpperCAmelCase ) return new_pieces def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.sp_model.PieceToId(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' return self.sp_model.IdToPiece(__UpperCAmelCase ) def lowerCamelCase ( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = [] __lowerCamelCase = '''''' __lowerCamelCase = False for token in tokens: # make sure that special tokens are not decoded using sentencepiece model if token in self.all_special_tokens: if not prev_is_special: out_string += " " out_string += self.sp_model.decode(__UpperCAmelCase ) + token __lowerCamelCase = True __lowerCamelCase = [] else: current_sub_tokens.append(__UpperCAmelCase ) __lowerCamelCase = False out_string += self.sp_model.decode(__UpperCAmelCase ) return out_string.strip() def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' __lowerCamelCase = [self.sep_token_id] __lowerCamelCase = [self.cls_token_id] if token_ids_a is None: return cls + token_ids_a + sep return cls + token_ids_a + sep + token_ids_a + sep def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None , __UpperCAmelCase = False ): '''simple docstring''' if already_has_special_tokens: return super().get_special_tokens_mask( token_ids_a=__UpperCAmelCase , token_ids_a=__UpperCAmelCase , already_has_special_tokens=__UpperCAmelCase ) if token_ids_a is not None: return [1] + ([0] * len(__UpperCAmelCase )) + [1] + ([0] * len(__UpperCAmelCase )) + [1] return [1] + ([0] * len(__UpperCAmelCase )) + [1] def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' __lowerCamelCase = [self.sep_token_id] __lowerCamelCase = [self.cls_token_id] if token_ids_a is None: return len(cls + token_ids_a + sep ) * [0] return len(cls + token_ids_a + sep ) * [0] + len(token_ids_a + sep ) * [1] def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase = None ): '''simple docstring''' if not os.path.isdir(__UpperCAmelCase ): logger.error(F"""Vocabulary path ({save_directory}) should be a directory""" ) return __lowerCamelCase = os.path.join( __UpperCAmelCase , (filename_prefix + '''-''' if filename_prefix else '''''') + VOCAB_FILES_NAMES['''vocab_file'''] ) if os.path.abspath(self.vocab_file ) != os.path.abspath(__UpperCAmelCase ) and os.path.isfile(self.vocab_file ): copyfile(self.vocab_file , __UpperCAmelCase ) elif not os.path.isfile(self.vocab_file ): with open(__UpperCAmelCase , '''wb''' ) as fi: __lowerCamelCase = self.sp_model.serialized_model_proto() fi.write(__UpperCAmelCase ) return (out_vocab_file,)
622
1
import argparse import requests import torch from PIL import Image from transformers import ViTMAEConfig, ViTMAEForPreTraining, ViTMAEImageProcessor def a__ ( _UpperCamelCase : int ): if "cls_token" in name: __lowerCamelCase = name.replace('''cls_token''' ,'''vit.embeddings.cls_token''' ) if "mask_token" in name: __lowerCamelCase = name.replace('''mask_token''' ,'''decoder.mask_token''' ) if "decoder_pos_embed" in name: __lowerCamelCase = name.replace('''decoder_pos_embed''' ,'''decoder.decoder_pos_embed''' ) if "pos_embed" in name and "decoder" not in name: __lowerCamelCase = name.replace('''pos_embed''' ,'''vit.embeddings.position_embeddings''' ) if "patch_embed.proj" in name: __lowerCamelCase = name.replace('''patch_embed.proj''' ,'''vit.embeddings.patch_embeddings.projection''' ) if "patch_embed.norm" in name: __lowerCamelCase = name.replace('''patch_embed.norm''' ,'''vit.embeddings.norm''' ) if "decoder_blocks" in name: __lowerCamelCase = name.replace('''decoder_blocks''' ,'''decoder.decoder_layers''' ) if "blocks" in name: __lowerCamelCase = name.replace('''blocks''' ,'''vit.encoder.layer''' ) if "attn.proj" in name: __lowerCamelCase = name.replace('''attn.proj''' ,'''attention.output.dense''' ) if "attn" in name: __lowerCamelCase = name.replace('''attn''' ,'''attention.self''' ) if "norm1" in name: __lowerCamelCase = name.replace('''norm1''' ,'''layernorm_before''' ) if "norm2" in name: __lowerCamelCase = name.replace('''norm2''' ,'''layernorm_after''' ) if "mlp.fc1" in name: __lowerCamelCase = name.replace('''mlp.fc1''' ,'''intermediate.dense''' ) if "mlp.fc2" in name: __lowerCamelCase = name.replace('''mlp.fc2''' ,'''output.dense''' ) if "decoder_embed" in name: __lowerCamelCase = name.replace('''decoder_embed''' ,'''decoder.decoder_embed''' ) if "decoder_norm" in name: __lowerCamelCase = name.replace('''decoder_norm''' ,'''decoder.decoder_norm''' ) if "decoder_pred" in name: __lowerCamelCase = name.replace('''decoder_pred''' ,'''decoder.decoder_pred''' ) if "norm.weight" in name and "decoder" not in name: __lowerCamelCase = name.replace('''norm.weight''' ,'''vit.layernorm.weight''' ) if "norm.bias" in name and "decoder" not in name: __lowerCamelCase = name.replace('''norm.bias''' ,'''vit.layernorm.bias''' ) return name def a__ ( _UpperCamelCase : str ,_UpperCamelCase : Optional[Any] ): for key in orig_state_dict.copy().keys(): __lowerCamelCase = orig_state_dict.pop(_UpperCamelCase ) if "qkv" in key: __lowerCamelCase = key.split('''.''' ) __lowerCamelCase = int(key_split[1] ) if "decoder_blocks" in key: __lowerCamelCase = config.decoder_hidden_size __lowerCamelCase = '''decoder.decoder_layers.''' if "weight" in key: __lowerCamelCase = val[:dim, :] __lowerCamelCase = val[dim : dim * 2, :] __lowerCamelCase = val[-dim:, :] elif "bias" in key: __lowerCamelCase = val[:dim] __lowerCamelCase = val[dim : dim * 2] __lowerCamelCase = val[-dim:] else: __lowerCamelCase = config.hidden_size __lowerCamelCase = '''vit.encoder.layer.''' if "weight" in key: __lowerCamelCase = val[:dim, :] __lowerCamelCase = val[dim : dim * 2, :] __lowerCamelCase = val[-dim:, :] elif "bias" in key: __lowerCamelCase = val[:dim] __lowerCamelCase = val[dim : dim * 2] __lowerCamelCase = val[-dim:] else: __lowerCamelCase = val return orig_state_dict def a__ ( _UpperCamelCase : Tuple ,_UpperCamelCase : Dict ): __lowerCamelCase = ViTMAEConfig() if "large" in checkpoint_url: __lowerCamelCase = 10_24 __lowerCamelCase = 40_96 __lowerCamelCase = 24 __lowerCamelCase = 16 elif "huge" in checkpoint_url: __lowerCamelCase = 14 __lowerCamelCase = 12_80 __lowerCamelCase = 51_20 __lowerCamelCase = 32 __lowerCamelCase = 16 __lowerCamelCase = ViTMAEForPreTraining(_UpperCamelCase ) __lowerCamelCase = torch.hub.load_state_dict_from_url(_UpperCamelCase ,map_location='''cpu''' )['''model'''] __lowerCamelCase = ViTMAEImageProcessor(size=config.image_size ) __lowerCamelCase = convert_state_dict(_UpperCamelCase ,_UpperCamelCase ) model.load_state_dict(_UpperCamelCase ) model.eval() __lowerCamelCase = '''https://user-images.githubusercontent.com/11435359/147738734-196fd92f-9260-48d5-ba7e-bf103d29364d.jpg''' __lowerCamelCase = Image.open(requests.get(_UpperCamelCase ,stream=_UpperCamelCase ).raw ) __lowerCamelCase = ViTMAEImageProcessor(size=config.image_size ) __lowerCamelCase = image_processor(images=_UpperCamelCase ,return_tensors='''pt''' ) # forward pass torch.manual_seed(2 ) __lowerCamelCase = model(**_UpperCamelCase ) __lowerCamelCase = outputs.logits if "large" in checkpoint_url: __lowerCamelCase = torch.tensor( [[-0.7_309, -0.7_128, -1.0_169], [-1.0_161, -0.9_058, -1.1_878], [-1.0_478, -0.9_411, -1.1_911]] ) elif "huge" in checkpoint_url: __lowerCamelCase = torch.tensor( [[-1.1_599, -0.9_199, -1.2_221], [-1.1_952, -0.9_269, -1.2_307], [-1.2_143, -0.9_337, -1.2_262]] ) else: __lowerCamelCase = torch.tensor( [[-0.9_192, -0.8_481, -1.1_259], [-1.1_349, -1.0_034, -1.2_599], [-1.1_757, -1.0_429, -1.2_726]] ) # verify logits assert torch.allclose(logits[0, :3, :3] ,_UpperCamelCase ,atol=1e-4 ) print(F"""Saving model to {pytorch_dump_folder_path}""" ) model.save_pretrained(_UpperCamelCase ) print(F"""Saving image processor to {pytorch_dump_folder_path}""" ) image_processor.save_pretrained(_UpperCamelCase ) if __name__ == "__main__": a_ = argparse.ArgumentParser() # Required parameters parser.add_argument( """--checkpoint_url""", default="""https://dl.fbaipublicfiles.com/mae/visualize/mae_visualize_vit_base.pth""", 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_ = parser.parse_args() convert_vit_mae_checkpoint(args.checkpoint_url, args.pytorch_dump_folder_path)
622
import math import os from copy import deepcopy import datasets import evaluate import torch import transformers from datasets import load_dataset from torch.utils.data import DataLoader from transformers import AutoModelForSequenceClassification, AutoTokenizer from accelerate import Accelerator from accelerate.test_utils import RegressionDataset, RegressionModel from accelerate.utils import is_tpu_available, set_seed a_ = """true""" def a__ ( _UpperCamelCase : Union[str, Any] ,_UpperCamelCase : List[str]=82 ,_UpperCamelCase : Optional[Any]=16 ): set_seed(42 ) __lowerCamelCase = RegressionModel() __lowerCamelCase = deepcopy(_UpperCamelCase ) __lowerCamelCase = RegressionDataset(length=_UpperCamelCase ) __lowerCamelCase = DataLoader(_UpperCamelCase ,batch_size=_UpperCamelCase ) model.to(accelerator.device ) __lowerCamelCase ,__lowerCamelCase = accelerator.prepare(_UpperCamelCase ,_UpperCamelCase ) return model, ddp_model, dataloader def a__ ( _UpperCamelCase : Accelerator ,_UpperCamelCase : str=False ): __lowerCamelCase = AutoTokenizer.from_pretrained('''hf-internal-testing/mrpc-bert-base-cased''' ) __lowerCamelCase = load_dataset('''glue''' ,'''mrpc''' ,split='''validation''' ) def tokenize_function(_UpperCamelCase : int ): __lowerCamelCase = tokenizer(examples['''sentence1'''] ,examples['''sentence2'''] ,truncation=_UpperCamelCase ,max_length=_UpperCamelCase ) return outputs with accelerator.main_process_first(): __lowerCamelCase = dataset.map( _UpperCamelCase ,batched=_UpperCamelCase ,remove_columns=['''idx''', '''sentence1''', '''sentence2'''] ,) __lowerCamelCase = tokenized_datasets.rename_column('''label''' ,'''labels''' ) def collate_fn(_UpperCamelCase : Any ): if use_longest: return tokenizer.pad(_UpperCamelCase ,padding='''longest''' ,return_tensors='''pt''' ) return tokenizer.pad(_UpperCamelCase ,padding='''max_length''' ,max_length=1_28 ,return_tensors='''pt''' ) return DataLoader(_UpperCamelCase ,shuffle=_UpperCamelCase ,collate_fn=_UpperCamelCase ,batch_size=16 ) def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : List[str] ): __lowerCamelCase = Accelerator(dispatch_batches=_UpperCamelCase ,split_batches=_UpperCamelCase ) __lowerCamelCase = get_dataloader(_UpperCamelCase ,not dispatch_batches ) __lowerCamelCase = AutoModelForSequenceClassification.from_pretrained( '''hf-internal-testing/mrpc-bert-base-cased''' ,return_dict=_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = accelerator.prepare(_UpperCamelCase ,_UpperCamelCase ) return {"ddp": [ddp_model, ddp_dataloader, "cuda:0"], "no": [model, dataloader, accelerator.device]}, accelerator def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : Optional[Any] ,_UpperCamelCase : Union[str, Any] ): __lowerCamelCase = [] for batch in dataloader: __lowerCamelCase ,__lowerCamelCase = batch.values() with torch.no_grad(): __lowerCamelCase = model(_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = accelerator.gather_for_metrics((logit, target) ) logits_and_targets.append((logit, target) ) __lowerCamelCase ,__lowerCamelCase = [], [] for logit, targ in logits_and_targets: logits.append(_UpperCamelCase ) targs.append(_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = torch.cat(_UpperCamelCase ), torch.cat(_UpperCamelCase ) return logits, targs def a__ ( _UpperCamelCase : Accelerator ,_UpperCamelCase : List[Any]=82 ,_UpperCamelCase : str=False ,_UpperCamelCase : List[str]=False ,_UpperCamelCase : Optional[int]=16 ): __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = get_basic_setup(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ) __lowerCamelCase ,__lowerCamelCase = generate_predictions(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ) assert ( len(_UpperCamelCase ) == num_samples ), F"""Unexpected number of inputs:\n Expected: {num_samples}\n Actual: {len(_UpperCamelCase )}""" def a__ ( _UpperCamelCase : bool = False ,_UpperCamelCase : bool = False ): __lowerCamelCase = evaluate.load('''glue''' ,'''mrpc''' ) __lowerCamelCase ,__lowerCamelCase = get_mrpc_setup(_UpperCamelCase ,_UpperCamelCase ) # First do baseline __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = setup['''no'''] model.to(_UpperCamelCase ) model.eval() for batch in dataloader: batch.to(_UpperCamelCase ) with torch.inference_mode(): __lowerCamelCase = model(**_UpperCamelCase ) __lowerCamelCase = outputs.logits.argmax(dim=-1 ) metric.add_batch(predictions=_UpperCamelCase ,references=batch['''labels'''] ) __lowerCamelCase = metric.compute() # Then do distributed __lowerCamelCase ,__lowerCamelCase ,__lowerCamelCase = setup['''ddp'''] model.eval() for batch in dataloader: with torch.inference_mode(): __lowerCamelCase = model(**_UpperCamelCase ) __lowerCamelCase = outputs.logits.argmax(dim=-1 ) __lowerCamelCase = batch['''labels'''] __lowerCamelCase ,__lowerCamelCase = accelerator.gather_for_metrics((preds, references) ) metric.add_batch(predictions=_UpperCamelCase ,references=_UpperCamelCase ) __lowerCamelCase = metric.compute() for key in "accuracy f1".split(): assert math.isclose( baseline[key] ,distributed[key] ), F"""Baseline and Distributed are not the same for key {key}:\n\tBaseline: {baseline[key]}\n\tDistributed: {distributed[key]}\n""" def a__ ( ): __lowerCamelCase = Accelerator(split_batches=_UpperCamelCase ,dispatch_batches=_UpperCamelCase ) if accelerator.is_local_main_process: datasets.utils.logging.set_verbosity_warning() transformers.utils.logging.set_verbosity_warning() else: datasets.utils.logging.set_verbosity_error() transformers.utils.logging.set_verbosity_error() # These are a bit slower so they should only be ran on the GPU or TPU if torch.cuda.is_available() or is_tpu_available(): if accelerator.is_local_main_process: print('''**Testing gather_for_metrics**''' ) for split_batches in [True, False]: for dispatch_batches in [True, False]: if accelerator.is_local_main_process: print(F"""With: `split_batches={split_batches}`, `dispatch_batches={dispatch_batches}`""" ) test_mrpc(_UpperCamelCase ,_UpperCamelCase ) accelerator.state._reset_state() if accelerator.is_local_main_process: print('''**Test torch metrics**''' ) for split_batches in [True, False]: for dispatch_batches in [True, False]: __lowerCamelCase = Accelerator(split_batches=_UpperCamelCase ,dispatch_batches=_UpperCamelCase ) if accelerator.is_local_main_process: print(F"""With: `split_batches={split_batches}`, `dispatch_batches={dispatch_batches}`, length=99""" ) test_torch_metrics(_UpperCamelCase ,99 ) accelerator.state._reset_state() if accelerator.is_local_main_process: print('''**Test last batch is not dropped when perfectly divisible**''' ) __lowerCamelCase = Accelerator() test_torch_metrics(_UpperCamelCase ,5_12 ) accelerator.state._reset_state() def a__ ( _UpperCamelCase : Optional[int] ): # For xla_spawn (TPUs) main() if __name__ == "__main__": main()
622
1
def a__ ( _UpperCamelCase : str ,_UpperCamelCase : bool = False ): if not isinstance(_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = F"""Expected string as input, found {type(_UpperCamelCase )}""" raise ValueError(_UpperCamelCase ) if not isinstance(_UpperCamelCase ,_UpperCamelCase ): __lowerCamelCase = F"""Expected boolean as use_pascal parameter, found {type(_UpperCamelCase )}""" raise ValueError(_UpperCamelCase ) __lowerCamelCase = input_str.split('''_''' ) __lowerCamelCase = 0 if use_pascal else 1 __lowerCamelCase = words[start_index:] __lowerCamelCase = [word[0].upper() + word[1:] for word in words_to_capitalize] __lowerCamelCase = '''''' if use_pascal else words[0] return "".join([initial_word, *capitalized_words] ) if __name__ == "__main__": from doctest import testmod testmod()
622
import gc import random import unittest import numpy as np import torch from transformers import CLIPTextConfig, CLIPTextModel, CLIPTextModelWithProjection, CLIPTokenizer from diffusers import ( AutoencoderKL, DiffusionPipeline, EulerDiscreteScheduler, StableDiffusionXLImgaImgPipeline, UNetaDConditionModel, ) from diffusers.utils import floats_tensor, slow, torch_device from diffusers.utils.testing_utils import enable_full_determinism, require_torch_gpu from ..pipeline_params import ( IMAGE_TO_IMAGE_IMAGE_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS, ) from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin enable_full_determinism() class __lowerCAmelCase ( lowerCAmelCase__ , lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = StableDiffusionXLImgaImgPipeline lowerCAmelCase__ = TEXT_GUIDED_IMAGE_VARIATION_PARAMS - {"""height""", """width"""} lowerCAmelCase__ = PipelineTesterMixin.required_optional_params - {"""latents"""} lowerCAmelCase__ = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS lowerCAmelCase__ = IMAGE_TO_IMAGE_IMAGE_PARAMS lowerCAmelCase__ = IMAGE_TO_IMAGE_IMAGE_PARAMS def lowerCamelCase ( self ): '''simple docstring''' torch.manual_seed(0 ) __lowerCamelCase = 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''') , attention_head_dim=(2, 4) , use_linear_projection=__UpperCAmelCase , addition_embed_type='''text_time''' , addition_time_embed_dim=8 , transformer_layers_per_block=(1, 2) , projection_class_embeddings_input_dim=80 , cross_attention_dim=64 , ) __lowerCamelCase = EulerDiscreteScheduler( beta_start=0.00_085 , beta_end=0.012 , steps_offset=1 , beta_schedule='''scaled_linear''' , timestep_spacing='''leading''' , ) torch.manual_seed(0 ) __lowerCamelCase = AutoencoderKL( block_out_channels=[32, 64] , in_channels=3 , out_channels=3 , down_block_types=['''DownEncoderBlock2D''', '''DownEncoderBlock2D'''] , up_block_types=['''UpDecoderBlock2D''', '''UpDecoderBlock2D'''] , latent_channels=4 , sample_size=128 , ) torch.manual_seed(0 ) __lowerCamelCase = CLIPTextConfig( bos_token_id=0 , eos_token_id=2 , hidden_size=32 , intermediate_size=37 , layer_norm_eps=1E-0_5 , num_attention_heads=4 , num_hidden_layers=5 , pad_token_id=1 , vocab_size=1000 , hidden_act='''gelu''' , projection_dim=32 , ) __lowerCamelCase = CLIPTextModel(__UpperCAmelCase ) __lowerCamelCase = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' , local_files_only=__UpperCAmelCase ) __lowerCamelCase = CLIPTextModelWithProjection(__UpperCAmelCase ) __lowerCamelCase = CLIPTokenizer.from_pretrained('''hf-internal-testing/tiny-random-clip''' , local_files_only=__UpperCAmelCase ) __lowerCamelCase = { '''unet''': unet, '''scheduler''': scheduler, '''vae''': vae, '''text_encoder''': text_encoder, '''tokenizer''': tokenizer, '''text_encoder_2''': text_encoder_a, '''tokenizer_2''': tokenizer_a, # "safety_checker": None, # "feature_extractor": None, } return components def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase=0 ): '''simple docstring''' __lowerCamelCase = floats_tensor((1, 3, 32, 32) , rng=random.Random(__UpperCAmelCase ) ).to(__UpperCAmelCase ) __lowerCamelCase = image / 2 + 0.5 if str(__UpperCAmelCase ).startswith('''mps''' ): __lowerCamelCase = torch.manual_seed(__UpperCAmelCase ) else: __lowerCamelCase = torch.Generator(device=__UpperCAmelCase ).manual_seed(__UpperCAmelCase ) __lowerCamelCase = { '''prompt''': '''A painting of a squirrel eating a burger''', '''image''': image, '''generator''': generator, '''num_inference_steps''': 2, '''guidance_scale''': 5.0, '''output_type''': '''numpy''', '''strength''': 0.75, } return inputs def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''cpu''' # ensure determinism for the device-dependent torch.Generator __lowerCamelCase = self.get_dummy_components() __lowerCamelCase = StableDiffusionXLImgaImgPipeline(**__UpperCAmelCase ) __lowerCamelCase = sd_pipe.to(__UpperCAmelCase ) sd_pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = self.get_dummy_inputs(__UpperCAmelCase ) __lowerCamelCase = sd_pipe(**__UpperCAmelCase ).images __lowerCamelCase = image[0, -3:, -3:, -1] assert image.shape == (1, 32, 32, 3) __lowerCamelCase = np.array([0.4_656, 0.4_840, 0.4_439, 0.6_698, 0.5_574, 0.4_524, 0.5_799, 0.5_943, 0.5_165] ) assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-2 def lowerCamelCase ( self ): '''simple docstring''' super().test_attention_slicing_forward_pass(expected_max_diff=3E-3 ) def lowerCamelCase ( self ): '''simple docstring''' super().test_inference_batch_single_identical(expected_max_diff=3E-3 ) def lowerCamelCase ( self ): '''simple docstring''' pass def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_dummy_components() __lowerCamelCase = StableDiffusionXLImgaImgPipeline(**__UpperCAmelCase ) __lowerCamelCase = sd_pipe.to(__UpperCAmelCase ) __lowerCamelCase = sd_pipe.to(__UpperCAmelCase ) sd_pipe.set_progress_bar_config(disable=__UpperCAmelCase ) # forward without prompt embeds __lowerCamelCase = self.get_dummy_inputs(__UpperCAmelCase ) __lowerCamelCase = 3 * ['''this is a negative prompt'''] __lowerCamelCase = negative_prompt __lowerCamelCase = 3 * [inputs['''prompt''']] __lowerCamelCase = sd_pipe(**__UpperCAmelCase ) __lowerCamelCase = output.images[0, -3:, -3:, -1] # forward with prompt embeds __lowerCamelCase = self.get_dummy_inputs(__UpperCAmelCase ) __lowerCamelCase = 3 * ['''this is a negative prompt'''] __lowerCamelCase = 3 * [inputs.pop('''prompt''' )] ( ( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) ,( __lowerCamelCase ) , ) = sd_pipe.encode_prompt(__UpperCAmelCase , negative_prompt=__UpperCAmelCase ) __lowerCamelCase = sd_pipe( **__UpperCAmelCase , prompt_embeds=__UpperCAmelCase , negative_prompt_embeds=__UpperCAmelCase , pooled_prompt_embeds=__UpperCAmelCase , negative_pooled_prompt_embeds=__UpperCAmelCase , ) __lowerCamelCase = output.images[0, -3:, -3:, -1] # make sure that it's equal assert np.abs(image_slice_a.flatten() - image_slice_a.flatten() ).max() < 1E-4 @slow @require_torch_gpu class __lowerCAmelCase ( unittest.TestCase ): def lowerCamelCase ( self ): '''simple docstring''' super().tearDown() gc.collect() torch.cuda.empty_cache() def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase="cpu" , __UpperCAmelCase=torch.floataa , __UpperCAmelCase=0 ): '''simple docstring''' __lowerCamelCase = torch.Generator(device=__UpperCAmelCase ).manual_seed(__UpperCAmelCase ) __lowerCamelCase = np.random.RandomState(__UpperCAmelCase ).standard_normal((1, 4, 64, 64) ) __lowerCamelCase = torch.from_numpy(__UpperCAmelCase ).to(device=__UpperCAmelCase , dtype=__UpperCAmelCase ) __lowerCamelCase = { '''prompt''': '''a photograph of an astronaut riding a horse''', '''latents''': latents, '''generator''': generator, '''num_inference_steps''': 3, '''guidance_scale''': 7.5, '''output_type''': '''numpy''', } return inputs def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = DiffusionPipeline.from_pretrained('''stabilityai/stable-diffusion-2-base''' ) pipe.to(__UpperCAmelCase ) pipe.set_progress_bar_config(disable=__UpperCAmelCase ) __lowerCamelCase = self.get_inputs(__UpperCAmelCase ) __lowerCamelCase = pipe(**__UpperCAmelCase ).images __lowerCamelCase = image[0, -3:, -3:, -1].flatten() assert image.shape == (1, 512, 512, 3) __lowerCamelCase = np.array([0.49_493, 0.47_896, 0.40_798, 0.54_214, 0.53_212, 0.48_202, 0.47_656, 0.46_329, 0.48_506] ) assert np.abs(image_slice - expected_slice ).max() < 7E-3
622
1
def a__ ( _UpperCamelCase : list ,_UpperCamelCase : int = 0 ): __lowerCamelCase = length or len(_UpperCamelCase ) __lowerCamelCase = False for i in range(length - 1 ): if list_data[i] > list_data[i + 1]: __lowerCamelCase ,__lowerCamelCase = list_data[i + 1], list_data[i] __lowerCamelCase = True return list_data if not swapped else bubble_sort(_UpperCamelCase ,length - 1 ) if __name__ == "__main__": import doctest doctest.testmod()
622
import torch from diffusers import StableDiffusionPipeline a_ = """path-to-your-trained-model""" a_ = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.floataa).to("""cuda""") a_ = """A photo of sks dog in a bucket""" a_ = pipe(prompt, num_inference_steps=50, guidance_scale=7.5).images[0] image.save("""dog-bucket.png""")
622
1
from __future__ import annotations def a__ ( _UpperCamelCase : str ,_UpperCamelCase : list[str] | None = None ): __lowerCamelCase = word_bank or [] # create a table __lowerCamelCase = len(_UpperCamelCase ) + 1 __lowerCamelCase = [] for _ in range(_UpperCamelCase ): table.append([] ) # seed value __lowerCamelCase = [[]] # because empty string has empty combination # iterate through the indices for i in range(_UpperCamelCase ): # condition if table[i] != []: for word in word_bank: # slice condition if target[i : i + len(_UpperCamelCase )] == word: __lowerCamelCase = [ [word, *way] for way in table[i] ] # adds the word to every combination the current position holds # now,push that combination to the table[i+len(word)] table[i + len(_UpperCamelCase )] += new_combinations # combinations are in reverse order so reverse for better output for combination in table[len(_UpperCamelCase )]: combination.reverse() return table[len(_UpperCamelCase )] if __name__ == "__main__": print(all_construct("""jwajalapa""", ["""jwa""", """j""", """w""", """a""", """la""", """lapa"""])) print(all_construct("""rajamati""", ["""s""", """raj""", """amat""", """raja""", """ma""", """i""", """t"""])) print( all_construct( """hexagonosaurus""", ["""h""", """ex""", """hex""", """ag""", """ago""", """ru""", """auru""", """rus""", """go""", """no""", """o""", """s"""], ) )
622
import unittest from transformers import MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING, AutoTokenizer, is_vision_available from transformers.pipelines import pipeline from transformers.pipelines.document_question_answering import apply_tesseract from transformers.testing_utils import ( is_pipeline_test, nested_simplify, require_detectrona, require_pytesseract, require_tf, require_torch, require_vision, slow, ) from .test_pipelines_common import ANY if is_vision_available(): from PIL import Image from transformers.image_utils import load_image else: class __lowerCAmelCase : @staticmethod def lowerCamelCase ( *__UpperCAmelCase , **__UpperCAmelCase ): '''simple docstring''' pass def a__ ( _UpperCamelCase : List[str] ): return None # This is a pinned image from a specific revision of a document question answering space, hosted by HuggingFace, # so we can expect it to be available. a_ = ( """https://huggingface.co/spaces/impira/docquery/resolve/2f6c96314dc84dfda62d40de9da55f2f5165d403/invoice.png""" ) @is_pipeline_test @require_torch @require_vision class __lowerCAmelCase ( unittest.TestCase ): lowerCAmelCase__ = MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING @require_pytesseract @require_vision def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model=__UpperCAmelCase , tokenizer=__UpperCAmelCase , image_processor=__UpperCAmelCase ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = list(zip(*apply_tesseract(load_image(__UpperCAmelCase ) , __UpperCAmelCase , '''''' ) ) ) __lowerCamelCase = '''What is the placebo?''' __lowerCamelCase = [ { '''image''': load_image(__UpperCAmelCase ), '''question''': question, }, { '''image''': image, '''question''': question, }, { '''image''': image, '''question''': question, '''word_boxes''': word_boxes, }, ] return dqa_pipeline, examples def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = dqa_pipeline(__UpperCAmelCase , top_k=2 ) self.assertEqual( __UpperCAmelCase , [ [ {'''score''': ANY(__UpperCAmelCase ), '''answer''': ANY(__UpperCAmelCase ), '''start''': ANY(__UpperCAmelCase ), '''end''': ANY(__UpperCAmelCase )}, {'''score''': ANY(__UpperCAmelCase ), '''answer''': ANY(__UpperCAmelCase ), '''start''': ANY(__UpperCAmelCase ), '''end''': ANY(__UpperCAmelCase )}, ] ] * 3 , ) @require_torch @require_detectrona @require_pytesseract def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline('''document-question-answering''' , model='''hf-internal-testing/tiny-random-layoutlmv2''' ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''How many cats are there?''' __lowerCamelCase = [ {'''score''': 0.0_001, '''answer''': '''oy 2312/2019''', '''start''': 38, '''end''': 39}, {'''score''': 0.0_001, '''answer''': '''oy 2312/2019 DUE''', '''start''': 38, '''end''': 40}, ] __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual(nested_simplify(__UpperCAmelCase , decimals=4 ) , __UpperCAmelCase ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual(nested_simplify(__UpperCAmelCase , decimals=4 ) , __UpperCAmelCase ) # This image does not detect ANY text in it, meaning layoutlmv2 should fail. # Empty answer probably __lowerCamelCase = '''./tests/fixtures/tests_samples/COCO/000000039769.png''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual(__UpperCAmelCase , [] ) # We can optionnally pass directly the words and bounding boxes __lowerCamelCase = '''./tests/fixtures/tests_samples/COCO/000000039769.png''' __lowerCamelCase = [] __lowerCamelCase = [] __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , words=__UpperCAmelCase , boxes=__UpperCAmelCase , top_k=2 ) self.assertEqual(__UpperCAmelCase , [] ) @slow @require_torch @require_detectrona @require_pytesseract def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model='''tiennvcs/layoutlmv2-base-uncased-finetuned-docvqa''' , revision='''9977165''' , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_944, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_009, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_944, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_009, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.9_944, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_009, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ], ] * 2 , ) @slow @require_torch @require_detectrona @require_pytesseract def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model='''tiennvcs/layoutlmv2-base-uncased-finetuned-docvqa''' , revision='''9977165''' , max_seq_len=50 , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_974, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, {'''score''': 0.9_948, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_974, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, {'''score''': 0.9_948, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.9_974, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, {'''score''': 0.9_948, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] ] * 2 , ) @slow @require_torch @require_pytesseract @require_vision def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = AutoTokenizer.from_pretrained( '''impira/layoutlm-document-qa''' , revision='''3dc6de3''' , add_prefix_space=__UpperCAmelCase ) __lowerCamelCase = pipeline( '''document-question-answering''' , model='''impira/layoutlm-document-qa''' , tokenizer=__UpperCAmelCase , revision='''3dc6de3''' , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] , ) __lowerCamelCase = dqa_pipeline({'''image''': image, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] ] * 2 , ) __lowerCamelCase = list(zip(*apply_tesseract(load_image(__UpperCAmelCase ) , __UpperCAmelCase , '''''' ) ) ) # This model should also work if `image` is set to None __lowerCamelCase = dqa_pipeline({'''image''': None, '''word_boxes''': word_boxes, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.4_251, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.0_819, '''answer''': '''1110212019''', '''start''': 23, '''end''': 23}, ] , ) @slow @require_torch @require_pytesseract @require_vision def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = AutoTokenizer.from_pretrained( '''impira/layoutlm-document-qa''' , revision='''3dc6de3''' , add_prefix_space=__UpperCAmelCase ) __lowerCamelCase = pipeline( '''document-question-answering''' , model='''impira/layoutlm-document-qa''' , tokenizer=__UpperCAmelCase , revision='''3dc6de3''' , max_seq_len=50 , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_999, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.9_998, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) __lowerCamelCase = dqa_pipeline( [{'''image''': image, '''question''': question}, {'''image''': image, '''question''': question}] , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ [ {'''score''': 0.9_999, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.9_998, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] ] * 2 , ) __lowerCamelCase = list(zip(*apply_tesseract(load_image(__UpperCAmelCase ) , __UpperCAmelCase , '''''' ) ) ) # This model should also work if `image` is set to None __lowerCamelCase = dqa_pipeline({'''image''': None, '''word_boxes''': word_boxes, '''question''': question} , top_k=2 ) self.assertEqual( nested_simplify(__UpperCAmelCase , decimals=4 ) , [ {'''score''': 0.9_999, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, {'''score''': 0.9_998, '''answer''': '''us-001''', '''start''': 16, '''end''': 16}, ] , ) @slow @require_torch def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = pipeline( '''document-question-answering''' , model='''naver-clova-ix/donut-base-finetuned-docvqa''' , tokenizer=AutoTokenizer.from_pretrained('''naver-clova-ix/donut-base-finetuned-docvqa''' ) , feature_extractor='''naver-clova-ix/donut-base-finetuned-docvqa''' , ) __lowerCamelCase = INVOICE_URL __lowerCamelCase = '''What is the invoice number?''' __lowerCamelCase = dqa_pipeline(image=__UpperCAmelCase , question=__UpperCAmelCase , top_k=2 ) self.assertEqual(nested_simplify(__UpperCAmelCase , decimals=4 ) , [{'''answer''': '''us-001'''}] ) @require_tf @unittest.skip('''Document question answering not implemented in TF''' ) def lowerCamelCase ( self ): '''simple docstring''' pass
622
1
from __future__ import annotations from typing import Any class __lowerCAmelCase ( lowerCAmelCase__ ): pass class __lowerCAmelCase : def __init__( self , __UpperCAmelCase ): '''simple docstring''' __lowerCamelCase = data __lowerCamelCase = None def __iter__( self ): '''simple docstring''' __lowerCamelCase = self __lowerCamelCase = [] while node: if node in visited: raise ContainsLoopError visited.append(__UpperCAmelCase ) yield node.data __lowerCamelCase = node.next_node @property def lowerCamelCase ( self ): '''simple docstring''' try: list(self ) return False except ContainsLoopError: return True if __name__ == "__main__": a_ = Node(1) a_ = Node(2) a_ = Node(3) a_ = Node(4) print(root_node.has_loop) # False a_ = root_node.next_node print(root_node.has_loop) # True a_ = Node(5) a_ = Node(6) a_ = Node(5) a_ = Node(6) print(root_node.has_loop) # False a_ = Node(1) print(root_node.has_loop) # False
622
import unittest from transformers.models.xlm_prophetnet.tokenization_xlm_prophetnet import SPIECE_UNDERLINE, XLMProphetNetTokenizer from transformers.testing_utils import get_tests_dir, require_sentencepiece, slow from transformers.utils import cached_property from ...test_tokenization_common import TokenizerTesterMixin a_ = get_tests_dir("""fixtures/test_sentencepiece.model""") @require_sentencepiece class __lowerCAmelCase ( lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = XLMProphetNetTokenizer lowerCAmelCase__ = False lowerCAmelCase__ = True def lowerCamelCase ( self ): '''simple docstring''' super().setUp() # We have a SentencePiece fixture for testing __lowerCamelCase = XLMProphetNetTokenizer(__UpperCAmelCase , keep_accents=__UpperCAmelCase ) tokenizer.save_pretrained(self.tmpdirname ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''[PAD]''' __lowerCamelCase = 0 self.assertEqual(self.get_tokenizer()._convert_token_to_id(__UpperCAmelCase ) , __UpperCAmelCase ) self.assertEqual(self.get_tokenizer()._convert_id_to_token(__UpperCAmelCase ) , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = list(self.get_tokenizer().get_vocab().keys() ) self.assertEqual(vocab_keys[0] , '''[PAD]''' ) self.assertEqual(vocab_keys[1] , '''[CLS]''' ) self.assertEqual(vocab_keys[-1] , '''j''' ) self.assertEqual(len(__UpperCAmelCase ) , 1012 ) def lowerCamelCase ( self ): '''simple docstring''' self.assertEqual(self.get_tokenizer().vocab_size , 1012 ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = XLMProphetNetTokenizer(__UpperCAmelCase , keep_accents=__UpperCAmelCase ) __lowerCamelCase = tokenizer.tokenize('''This is a test''' ) self.assertListEqual(__UpperCAmelCase , ['''▁This''', '''▁is''', '''▁a''', '''▁t''', '''est'''] ) self.assertListEqual( tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , ) __lowerCamelCase = tokenizer.tokenize('''I was born in 92000, and this is falsé.''' ) self.assertListEqual( __UpperCAmelCase , [ SPIECE_UNDERLINE + '''I''', SPIECE_UNDERLINE + '''was''', SPIECE_UNDERLINE + '''b''', '''or''', '''n''', SPIECE_UNDERLINE + '''in''', SPIECE_UNDERLINE + '''''', '''9''', '''2''', '''0''', '''0''', '''0''', ''',''', SPIECE_UNDERLINE + '''and''', SPIECE_UNDERLINE + '''this''', SPIECE_UNDERLINE + '''is''', SPIECE_UNDERLINE + '''f''', '''al''', '''s''', '''é''', '''.''', ] , ) __lowerCamelCase = tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) self.assertListEqual( __UpperCAmelCase , [ value + tokenizer.fairseq_offset for value in [8, 21, 84, 55, 24, 19, 7, -9, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, -9, 4] ] , ) __lowerCamelCase = tokenizer.convert_ids_to_tokens(__UpperCAmelCase ) self.assertListEqual( __UpperCAmelCase , [ SPIECE_UNDERLINE + '''I''', SPIECE_UNDERLINE + '''was''', SPIECE_UNDERLINE + '''b''', '''or''', '''n''', SPIECE_UNDERLINE + '''in''', SPIECE_UNDERLINE + '''''', '''[UNK]''', '''2''', '''0''', '''0''', '''0''', ''',''', SPIECE_UNDERLINE + '''and''', SPIECE_UNDERLINE + '''this''', SPIECE_UNDERLINE + '''is''', SPIECE_UNDERLINE + '''f''', '''al''', '''s''', '''[UNK]''', '''.''', ] , ) @cached_property def lowerCamelCase ( self ): '''simple docstring''' return XLMProphetNetTokenizer.from_pretrained('''microsoft/xprophetnet-large-wiki100-cased''' ) @slow def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''Hello World!''' __lowerCamelCase = [35389, 6672, 49, 2] self.assertListEqual(__UpperCAmelCase , self.big_tokenizer.encode(__UpperCAmelCase ) ) @slow def lowerCamelCase ( self ): '''simple docstring''' # fmt: off __lowerCamelCase = {'''input_ids''': [[11073, 82783, 18, 26, 82783, 549, 51540, 248, 17209, 1301, 217, 20, 215186, 1325, 147, 17209, 1301, 217, 20, 56370, 53, 122020, 20, 16477, 27, 87355, 4548, 20, 4728, 78392, 17, 159969, 18, 26, 24491, 629, 15, 538, 22704, 5439, 15, 2788, 24491, 9885, 15, 43534, 605, 15, 814, 18403, 33200, 29, 15, 43534, 24458, 12410, 111, 24966, 83669, 9637, 144068, 26, 850, 22346, 27, 147, 24966, 83669, 83490, 26, 39113, 735, 27, 689, 656, 2800, 1339, 4600, 53, 122020, 115785, 34, 816, 1339, 46887, 18, 147, 53905, 1951, 42238, 41170, 17732, 834, 436, 15, 27523, 98733, 217, 147, 5542, 4981, 930, 17347, 16, 2], [20091, 629, 94, 82786, 58, 490, 20, 1528, 84, 53905, 344, 80592, 110128, 18822, 5267, 1306, 62, 152537, 308, 7997, 401, 124427, 549, 35442, 225, 109, 15055, 25748, 147, 7119, 43712, 34, 767, 135366, 18, 16, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [592, 63784, 119466, 17, 147808, 88214, 18, 656, 81, 32, 3296, 10280, 16, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], '''attention_mask''': [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]} # noqa: E501 # fmt: on self.tokenizer_integration_test_util( expected_encoding=__UpperCAmelCase , model_name='''microsoft/xprophetnet-large-wiki100-cased''' , revision='''1acad1643ddd54a44df6a1b797ada8373685d90e''' , )
622
1
from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { """facebook/nllb-moe-54B""": """https://huggingface.co/facebook/nllb-moe-54b/resolve/main/config.json""", } class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = """nllb-moe""" lowerCAmelCase__ = ["""past_key_values"""] lowerCAmelCase__ = {"""num_attention_heads""": """encoder_attention_heads""", """hidden_size""": """d_model"""} def __init__( self , __UpperCAmelCase=128112 , __UpperCAmelCase=1024 , __UpperCAmelCase=12 , __UpperCAmelCase=4096 , __UpperCAmelCase=16 , __UpperCAmelCase=12 , __UpperCAmelCase=4096 , __UpperCAmelCase=16 , __UpperCAmelCase=0.05 , __UpperCAmelCase=0.05 , __UpperCAmelCase=True , __UpperCAmelCase=True , __UpperCAmelCase="relu" , __UpperCAmelCase=1024 , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.0 , __UpperCAmelCase=0.02 , __UpperCAmelCase=2 , __UpperCAmelCase=True , __UpperCAmelCase=False , __UpperCAmelCase="float32" , __UpperCAmelCase=False , __UpperCAmelCase=128 , __UpperCAmelCase=64 , __UpperCAmelCase=4 , __UpperCAmelCase=4 , __UpperCAmelCase=0.001 , __UpperCAmelCase=0.001 , __UpperCAmelCase="all" , __UpperCAmelCase=False , __UpperCAmelCase=False , __UpperCAmelCase=1.0 , __UpperCAmelCase=0.2 , __UpperCAmelCase=1 , __UpperCAmelCase=0 , __UpperCAmelCase=2 , __UpperCAmelCase=False , **__UpperCAmelCase , ): '''simple docstring''' __lowerCamelCase = vocab_size __lowerCamelCase = max_position_embeddings __lowerCamelCase = d_model __lowerCamelCase = encoder_ffn_dim __lowerCamelCase = encoder_layers __lowerCamelCase = encoder_attention_heads __lowerCamelCase = decoder_ffn_dim __lowerCamelCase = decoder_layers __lowerCamelCase = decoder_attention_heads __lowerCamelCase = dropout __lowerCamelCase = attention_dropout __lowerCamelCase = activation_dropout __lowerCamelCase = activation_function __lowerCamelCase = init_std __lowerCamelCase = encoder_layerdrop __lowerCamelCase = decoder_layerdrop __lowerCamelCase = use_cache __lowerCamelCase = encoder_layers __lowerCamelCase = scale_embedding # scale factor will be sqrt(d_model) if True __lowerCamelCase = router_z_loss_coef __lowerCamelCase = router_aux_loss_coef __lowerCamelCase = decoder_sparse_step __lowerCamelCase = encoder_sparse_step __lowerCamelCase = num_experts __lowerCamelCase = expert_capacity __lowerCamelCase = router_bias if router_dtype not in ["float32", "float16", "bfloat16"]: raise ValueError(F"""`router_dtype` must be one of 'float32', 'float16' or 'bfloat16', got {router_dtype}""" ) __lowerCamelCase = router_dtype __lowerCamelCase = router_ignore_padding_tokens __lowerCamelCase = batch_prioritized_routing __lowerCamelCase = second_expert_policy __lowerCamelCase = normalize_router_prob_before_dropping __lowerCamelCase = moe_eval_capacity_token_fraction __lowerCamelCase = moe_token_dropout __lowerCamelCase = output_router_logits super().__init__( pad_token_id=__UpperCAmelCase , bos_token_id=__UpperCAmelCase , eos_token_id=__UpperCAmelCase , is_encoder_decoder=__UpperCAmelCase , decoder_start_token_id=__UpperCAmelCase , **__UpperCAmelCase , )
622
import inspect import re 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_config_docstrings.py a_ = """src/transformers""" # This is to make sure the transformers module imported is the one in the repo. a_ = direct_transformers_import(PATH_TO_TRANSFORMERS) a_ = transformers.models.auto.configuration_auto.CONFIG_MAPPING # Regex pattern used to find the checkpoint mentioned in the docstring of `config_class`. # For example, `[bert-base-uncased](https://huggingface.co/bert-base-uncased)` a_ = re.compile(R"""\[(.+?)\]\((https://huggingface\.co/.+?)\)""") a_ = { """DecisionTransformerConfig""", """EncoderDecoderConfig""", """MusicgenConfig""", """RagConfig""", """SpeechEncoderDecoderConfig""", """TimmBackboneConfig""", """VisionEncoderDecoderConfig""", """VisionTextDualEncoderConfig""", """LlamaConfig""", } def a__ ( _UpperCamelCase : Union[str, Any] ): __lowerCamelCase = None # source code of `config_class` __lowerCamelCase = inspect.getsource(_UpperCamelCase ) __lowerCamelCase = _re_checkpoint.findall(_UpperCamelCase ) # Each `checkpoint` is a tuple of a checkpoint name and a checkpoint link. # For example, `('bert-base-uncased', 'https://huggingface.co/bert-base-uncased')` for ckpt_name, ckpt_link in checkpoints: # allow the link to end with `/` if ckpt_link.endswith('''/''' ): __lowerCamelCase = ckpt_link[:-1] # verify the checkpoint name corresponds to the checkpoint link __lowerCamelCase = F"""https://huggingface.co/{ckpt_name}""" if ckpt_link == ckpt_link_from_name: __lowerCamelCase = ckpt_name break return checkpoint def a__ ( ): __lowerCamelCase = [] for config_class in list(CONFIG_MAPPING.values() ): # Skip deprecated models if "models.deprecated" in config_class.__module__: continue __lowerCamelCase = get_checkpoint_from_config_class(_UpperCamelCase ) __lowerCamelCase = config_class.__name__ if checkpoint is None and name not in CONFIG_CLASSES_TO_IGNORE_FOR_DOCSTRING_CHECKPOINT_CHECK: configs_without_checkpoint.append(_UpperCamelCase ) if len(_UpperCamelCase ) > 0: __lowerCamelCase = '''\n'''.join(sorted(_UpperCamelCase ) ) raise ValueError(F"""The following configurations don't contain any valid checkpoint:\n{message}""" ) if __name__ == "__main__": check_config_docstrings_have_checkpoints()
622
1
import collections import gzip import os import urllib import numpy from tensorflow.python.framework import dtypes, random_seed from tensorflow.python.platform import gfile from tensorflow.python.util.deprecation import deprecated a_ = collections.namedtuple("""_Datasets""", ["""train""", """validation""", """test"""]) # CVDF mirror of http://yann.lecun.com/exdb/mnist/ a_ = """https://storage.googleapis.com/cvdf-datasets/mnist/""" def a__ ( _UpperCamelCase : Optional[Any] ): __lowerCamelCase = numpy.dtype(numpy.uintaa ).newbyteorder('''>''' ) return numpy.frombuffer(bytestream.read(4 ) ,dtype=_UpperCamelCase )[0] @deprecated(_UpperCamelCase ,'''Please use tf.data to implement this functionality.''' ) def a__ ( _UpperCamelCase : str ): print('''Extracting''' ,f.name ) with gzip.GzipFile(fileobj=_UpperCamelCase ) as bytestream: __lowerCamelCase = _readaa(_UpperCamelCase ) if magic != 20_51: raise ValueError( '''Invalid magic number %d in MNIST image file: %s''' % (magic, f.name) ) __lowerCamelCase = _readaa(_UpperCamelCase ) __lowerCamelCase = _readaa(_UpperCamelCase ) __lowerCamelCase = _readaa(_UpperCamelCase ) __lowerCamelCase = bytestream.read(rows * cols * num_images ) __lowerCamelCase = numpy.frombuffer(_UpperCamelCase ,dtype=numpy.uinta ) __lowerCamelCase = data.reshape(_UpperCamelCase ,_UpperCamelCase ,_UpperCamelCase ,1 ) return data @deprecated(_UpperCamelCase ,'''Please use tf.one_hot on tensors.''' ) def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : Union[str, Any] ): __lowerCamelCase = labels_dense.shape[0] __lowerCamelCase = numpy.arange(_UpperCamelCase ) * num_classes __lowerCamelCase = numpy.zeros((num_labels, num_classes) ) __lowerCamelCase = 1 return labels_one_hot @deprecated(_UpperCamelCase ,'''Please use tf.data to implement this functionality.''' ) def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : Dict=False ,_UpperCamelCase : Any=10 ): print('''Extracting''' ,f.name ) with gzip.GzipFile(fileobj=_UpperCamelCase ) as bytestream: __lowerCamelCase = _readaa(_UpperCamelCase ) if magic != 20_49: raise ValueError( '''Invalid magic number %d in MNIST label file: %s''' % (magic, f.name) ) __lowerCamelCase = _readaa(_UpperCamelCase ) __lowerCamelCase = bytestream.read(_UpperCamelCase ) __lowerCamelCase = numpy.frombuffer(_UpperCamelCase ,dtype=numpy.uinta ) if one_hot: return _dense_to_one_hot(_UpperCamelCase ,_UpperCamelCase ) return labels class __lowerCAmelCase : @deprecated( __UpperCAmelCase , '''Please use alternatives such as official/mnist/_DataSet.py''' ''' from tensorflow/models.''' , ) def __init__( self , __UpperCAmelCase , __UpperCAmelCase , __UpperCAmelCase=False , __UpperCAmelCase=False , __UpperCAmelCase=dtypes.floataa , __UpperCAmelCase=True , __UpperCAmelCase=None , ): '''simple docstring''' __lowerCamelCase ,__lowerCamelCase = random_seed.get_seed(__UpperCAmelCase ) # If op level seed is not set, use whatever graph level seed is returned numpy.random.seed(seeda if seed is None else seeda ) __lowerCamelCase = dtypes.as_dtype(__UpperCAmelCase ).base_dtype if dtype not in (dtypes.uinta, dtypes.floataa): raise TypeError('''Invalid image dtype %r, expected uint8 or float32''' % dtype ) if fake_data: __lowerCamelCase = 10000 __lowerCamelCase = one_hot else: assert ( images.shape[0] == labels.shape[0] ), F"""images.shape: {images.shape} labels.shape: {labels.shape}""" __lowerCamelCase = images.shape[0] # Convert shape from [num examples, rows, columns, depth] # to [num examples, rows*columns] (assuming depth == 1) if reshape: assert images.shape[3] == 1 __lowerCamelCase = images.reshape( images.shape[0] , images.shape[1] * images.shape[2] ) if dtype == dtypes.floataa: # Convert from [0, 255] -> [0.0, 1.0]. __lowerCamelCase = images.astype(numpy.floataa ) __lowerCamelCase = numpy.multiply(__UpperCAmelCase , 1.0 / 255.0 ) __lowerCamelCase = images __lowerCamelCase = labels __lowerCamelCase = 0 __lowerCamelCase = 0 @property def lowerCamelCase ( self ): '''simple docstring''' return self._images @property def lowerCamelCase ( self ): '''simple docstring''' return self._labels @property def lowerCamelCase ( self ): '''simple docstring''' return self._num_examples @property def lowerCamelCase ( self ): '''simple docstring''' return self._epochs_completed def lowerCamelCase ( self , __UpperCAmelCase , __UpperCAmelCase=False , __UpperCAmelCase=True ): '''simple docstring''' if fake_data: __lowerCamelCase = [1] * 784 __lowerCamelCase = [1] + [0] * 9 if self.one_hot else 0 return ( [fake_image for _ in range(__UpperCAmelCase )], [fake_label for _ in range(__UpperCAmelCase )], ) __lowerCamelCase = self._index_in_epoch # Shuffle for the first epoch if self._epochs_completed == 0 and start == 0 and shuffle: __lowerCamelCase = numpy.arange(self._num_examples ) numpy.random.shuffle(__UpperCAmelCase ) __lowerCamelCase = self.images[perma] __lowerCamelCase = self.labels[perma] # Go to the next epoch if start + batch_size > self._num_examples: # Finished epoch self._epochs_completed += 1 # Get the rest examples in this epoch __lowerCamelCase = self._num_examples - start __lowerCamelCase = self._images[start : self._num_examples] __lowerCamelCase = self._labels[start : self._num_examples] # Shuffle the data if shuffle: __lowerCamelCase = numpy.arange(self._num_examples ) numpy.random.shuffle(__UpperCAmelCase ) __lowerCamelCase = self.images[perm] __lowerCamelCase = self.labels[perm] # Start next epoch __lowerCamelCase = 0 __lowerCamelCase = batch_size - rest_num_examples __lowerCamelCase = self._index_in_epoch __lowerCamelCase = self._images[start:end] __lowerCamelCase = self._labels[start:end] return ( numpy.concatenate((images_rest_part, images_new_part) , axis=0 ), numpy.concatenate((labels_rest_part, labels_new_part) , axis=0 ), ) else: self._index_in_epoch += batch_size __lowerCamelCase = self._index_in_epoch return self._images[start:end], self._labels[start:end] @deprecated(_UpperCamelCase ,'''Please write your own downloading logic.''' ) def a__ ( _UpperCamelCase : Dict ,_UpperCamelCase : Union[str, Any] ,_UpperCamelCase : Optional[Any] ): if not gfile.Exists(_UpperCamelCase ): gfile.MakeDirs(_UpperCamelCase ) __lowerCamelCase = os.path.join(_UpperCamelCase ,_UpperCamelCase ) if not gfile.Exists(_UpperCamelCase ): urllib.request.urlretrieve(_UpperCamelCase ,_UpperCamelCase ) # noqa: S310 with gfile.GFile(_UpperCamelCase ) as f: __lowerCamelCase = f.size() print('''Successfully downloaded''' ,_UpperCamelCase ,_UpperCamelCase ,'''bytes.''' ) return filepath @deprecated( _UpperCamelCase ,'''Please use alternatives such as:''' ''' tensorflow_datasets.load(\'mnist\')''' ) def a__ ( _UpperCamelCase : List[str] ,_UpperCamelCase : List[Any]=False ,_UpperCamelCase : Tuple=False ,_UpperCamelCase : List[str]=dtypes.floataa ,_UpperCamelCase : Tuple=True ,_UpperCamelCase : Tuple=50_00 ,_UpperCamelCase : str=None ,_UpperCamelCase : Optional[Any]=DEFAULT_SOURCE_URL ,): if fake_data: def fake(): return _DataSet( [] ,[] ,fake_data=_UpperCamelCase ,one_hot=_UpperCamelCase ,dtype=_UpperCamelCase ,seed=_UpperCamelCase ) __lowerCamelCase = fake() __lowerCamelCase = fake() __lowerCamelCase = fake() return _Datasets(train=_UpperCamelCase ,validation=_UpperCamelCase ,test=_UpperCamelCase ) if not source_url: # empty string check __lowerCamelCase = DEFAULT_SOURCE_URL __lowerCamelCase = '''train-images-idx3-ubyte.gz''' __lowerCamelCase = '''train-labels-idx1-ubyte.gz''' __lowerCamelCase = '''t10k-images-idx3-ubyte.gz''' __lowerCamelCase = '''t10k-labels-idx1-ubyte.gz''' __lowerCamelCase = _maybe_download( _UpperCamelCase ,_UpperCamelCase ,source_url + train_images_file ) with gfile.Open(_UpperCamelCase ,'''rb''' ) as f: __lowerCamelCase = _extract_images(_UpperCamelCase ) __lowerCamelCase = _maybe_download( _UpperCamelCase ,_UpperCamelCase ,source_url + train_labels_file ) with gfile.Open(_UpperCamelCase ,'''rb''' ) as f: __lowerCamelCase = _extract_labels(_UpperCamelCase ,one_hot=_UpperCamelCase ) __lowerCamelCase = _maybe_download( _UpperCamelCase ,_UpperCamelCase ,source_url + test_images_file ) with gfile.Open(_UpperCamelCase ,'''rb''' ) as f: __lowerCamelCase = _extract_images(_UpperCamelCase ) __lowerCamelCase = _maybe_download( _UpperCamelCase ,_UpperCamelCase ,source_url + test_labels_file ) with gfile.Open(_UpperCamelCase ,'''rb''' ) as f: __lowerCamelCase = _extract_labels(_UpperCamelCase ,one_hot=_UpperCamelCase ) if not 0 <= validation_size <= len(_UpperCamelCase ): __lowerCamelCase = ( '''Validation size should be between 0 and ''' F"""{len(_UpperCamelCase )}. Received: {validation_size}.""" ) raise ValueError(_UpperCamelCase ) __lowerCamelCase = train_images[:validation_size] __lowerCamelCase = train_labels[:validation_size] __lowerCamelCase = train_images[validation_size:] __lowerCamelCase = train_labels[validation_size:] __lowerCamelCase = {'''dtype''': dtype, '''reshape''': reshape, '''seed''': seed} __lowerCamelCase = _DataSet(_UpperCamelCase ,_UpperCamelCase ,**_UpperCamelCase ) __lowerCamelCase = _DataSet(_UpperCamelCase ,_UpperCamelCase ,**_UpperCamelCase ) __lowerCamelCase = _DataSet(_UpperCamelCase ,_UpperCamelCase ,**_UpperCamelCase ) return _Datasets(train=_UpperCamelCase ,validation=_UpperCamelCase ,test=_UpperCamelCase )
622
from typing import TYPE_CHECKING from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available a_ = { """configuration_clipseg""": [ """CLIPSEG_PRETRAINED_CONFIG_ARCHIVE_MAP""", """CLIPSegConfig""", """CLIPSegTextConfig""", """CLIPSegVisionConfig""", ], """processing_clipseg""": ["""CLIPSegProcessor"""], } try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: a_ = [ """CLIPSEG_PRETRAINED_MODEL_ARCHIVE_LIST""", """CLIPSegModel""", """CLIPSegPreTrainedModel""", """CLIPSegTextModel""", """CLIPSegVisionModel""", """CLIPSegForImageSegmentation""", ] if TYPE_CHECKING: from .configuration_clipseg import ( CLIPSEG_PRETRAINED_CONFIG_ARCHIVE_MAP, CLIPSegConfig, CLIPSegTextConfig, CLIPSegVisionConfig, ) from .processing_clipseg import CLIPSegProcessor try: if not is_torch_available(): raise OptionalDependencyNotAvailable() except OptionalDependencyNotAvailable: pass else: from .modeling_clipseg import ( CLIPSEG_PRETRAINED_MODEL_ARCHIVE_LIST, CLIPSegForImageSegmentation, CLIPSegModel, CLIPSegPreTrainedModel, CLIPSegTextModel, CLIPSegVisionModel, ) else: import sys a_ = _LazyModule(__name__, globals()["""__file__"""], _import_structure, module_spec=__spec__)
622
1
import math from ...configuration_utils import PretrainedConfig from ...utils import logging a_ = logging.get_logger(__name__) a_ = { """facebook/data2vec-base-960h""": """https://huggingface.co/facebook/data2vec-audio-base-960h/resolve/main/config.json""", # See all Data2VecAudio models at https://huggingface.co/models?filter=data2vec-audio } class __lowerCAmelCase ( lowerCAmelCase__ ): lowerCAmelCase__ = """data2vec-audio""" def __init__( self , __UpperCAmelCase=32 , __UpperCAmelCase=768 , __UpperCAmelCase=12 , __UpperCAmelCase=12 , __UpperCAmelCase=3072 , __UpperCAmelCase="gelu" , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.0 , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.1 , __UpperCAmelCase=0.02 , __UpperCAmelCase=1E-5 , __UpperCAmelCase="gelu" , __UpperCAmelCase=(512, 512, 512, 512, 512, 512, 512) , __UpperCAmelCase=(5, 2, 2, 2, 2, 2, 2) , __UpperCAmelCase=(10, 3, 3, 3, 3, 2, 2) , __UpperCAmelCase=False , __UpperCAmelCase=16 , __UpperCAmelCase=19 , __UpperCAmelCase=5 , __UpperCAmelCase=0.05 , __UpperCAmelCase=10 , __UpperCAmelCase=2 , __UpperCAmelCase=0.0 , __UpperCAmelCase=10 , __UpperCAmelCase=0 , __UpperCAmelCase="sum" , __UpperCAmelCase=False , __UpperCAmelCase=False , __UpperCAmelCase=256 , __UpperCAmelCase=(512, 512, 512, 512, 1500) , __UpperCAmelCase=(5, 3, 3, 1, 1) , __UpperCAmelCase=(1, 2, 3, 1, 1) , __UpperCAmelCase=512 , __UpperCAmelCase=0 , __UpperCAmelCase=1 , __UpperCAmelCase=2 , __UpperCAmelCase=False , __UpperCAmelCase=3 , __UpperCAmelCase=2 , __UpperCAmelCase=3 , __UpperCAmelCase=None , **__UpperCAmelCase , ): '''simple docstring''' super().__init__(**__UpperCAmelCase , pad_token_id=__UpperCAmelCase , bos_token_id=__UpperCAmelCase , eos_token_id=__UpperCAmelCase ) __lowerCamelCase = hidden_size __lowerCamelCase = feat_extract_activation __lowerCamelCase = list(__UpperCAmelCase ) __lowerCamelCase = list(__UpperCAmelCase ) __lowerCamelCase = list(__UpperCAmelCase ) __lowerCamelCase = conv_bias __lowerCamelCase = num_conv_pos_embeddings __lowerCamelCase = num_conv_pos_embedding_groups __lowerCamelCase = conv_pos_kernel_size __lowerCamelCase = len(self.conv_dim ) __lowerCamelCase = num_hidden_layers __lowerCamelCase = intermediate_size __lowerCamelCase = hidden_act __lowerCamelCase = num_attention_heads __lowerCamelCase = hidden_dropout __lowerCamelCase = attention_dropout __lowerCamelCase = activation_dropout __lowerCamelCase = feat_proj_dropout __lowerCamelCase = final_dropout __lowerCamelCase = layerdrop __lowerCamelCase = layer_norm_eps __lowerCamelCase = initializer_range __lowerCamelCase = vocab_size __lowerCamelCase = use_weighted_layer_sum if ( (len(self.conv_stride ) != self.num_feat_extract_layers) or (len(self.conv_kernel ) != self.num_feat_extract_layers) or (len(self.conv_dim ) != self.num_feat_extract_layers) ): raise ValueError( '''Configuration for convolutional layers is incorrect. It is required that `len(config.conv_dim)` ==''' ''' `len(config.conv_stride)` == `len(config.conv_kernel)`, but is `len(config.conv_dim) =''' F""" {len(self.conv_dim )}`, `len(config.conv_stride) = {len(self.conv_stride )}`,""" F""" `len(config.conv_kernel) = {len(self.conv_kernel )}`.""" ) # fine-tuning config parameters for SpecAugment: https://arxiv.org/abs/1904.08779 __lowerCamelCase = mask_time_prob __lowerCamelCase = mask_time_length __lowerCamelCase = mask_time_min_masks __lowerCamelCase = mask_feature_prob __lowerCamelCase = mask_feature_length __lowerCamelCase = mask_feature_min_masks # ctc loss __lowerCamelCase = ctc_loss_reduction __lowerCamelCase = ctc_zero_infinity # adapter __lowerCamelCase = add_adapter __lowerCamelCase = adapter_kernel_size __lowerCamelCase = adapter_stride __lowerCamelCase = num_adapter_layers __lowerCamelCase = output_hidden_size or hidden_size # SequenceClassification-specific parameter. Feel free to ignore for other classes. __lowerCamelCase = classifier_proj_size # XVector-specific parameters. Feel free to ignore for other classes. __lowerCamelCase = list(__UpperCAmelCase ) __lowerCamelCase = list(__UpperCAmelCase ) __lowerCamelCase = list(__UpperCAmelCase ) __lowerCamelCase = xvector_output_dim @property def lowerCamelCase ( self ): '''simple docstring''' return math.prod(self.conv_stride )
622
import unittest from transformers import RoFormerTokenizer, RoFormerTokenizerFast from transformers.testing_utils import require_rjieba, require_tokenizers from ...test_tokenization_common import TokenizerTesterMixin @require_rjieba @require_tokenizers class __lowerCAmelCase ( lowerCAmelCase__ , unittest.TestCase ): lowerCAmelCase__ = RoFormerTokenizer lowerCAmelCase__ = RoFormerTokenizerFast lowerCAmelCase__ = True lowerCAmelCase__ = True def lowerCamelCase ( self ): '''simple docstring''' super().setUp() def lowerCamelCase ( self , **__UpperCAmelCase ): '''simple docstring''' return self.tokenizer_class.from_pretrained('''junnyu/roformer_chinese_base''' , **__UpperCAmelCase ) def lowerCamelCase ( self , **__UpperCAmelCase ): '''simple docstring''' return self.rust_tokenizer_class.from_pretrained('''junnyu/roformer_chinese_base''' , **__UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = '''永和服装饰品有限公司,今天天气非常好''' __lowerCamelCase = '''永和 服装 饰品 有限公司 , 今 天 天 气 非常 好''' return input_text, output_text def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_tokenizer() __lowerCamelCase ,__lowerCamelCase = self.get_chinese_input_output_texts() __lowerCamelCase = tokenizer.tokenize(__UpperCAmelCase ) self.assertListEqual(__UpperCAmelCase , output_text.split() ) __lowerCamelCase = tokens + [tokenizer.unk_token] __lowerCamelCase = [22943, 21332, 34431, 45904, 117, 306, 1231, 1231, 2653, 33994, 1266, 100] self.assertListEqual(tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' __lowerCamelCase = self.get_rust_tokenizer() __lowerCamelCase ,__lowerCamelCase = self.get_chinese_input_output_texts() __lowerCamelCase = tokenizer.tokenize(__UpperCAmelCase ) self.assertListEqual(__UpperCAmelCase , output_text.split() ) __lowerCamelCase = tokens + [tokenizer.unk_token] __lowerCamelCase = [22943, 21332, 34431, 45904, 117, 306, 1231, 1231, 2653, 33994, 1266, 100] self.assertListEqual(tokenizer.convert_tokens_to_ids(__UpperCAmelCase ) , __UpperCAmelCase ) def lowerCamelCase ( self ): '''simple docstring''' pass def lowerCamelCase ( self ): '''simple docstring''' pass def lowerCamelCase ( self ): '''simple docstring''' pass
622
1