Spaces:
Runtime error
Runtime error
File size: 23,721 Bytes
f8f5cdf |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 |
# Copyright 2020 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 json
import os
import pickle
import random
import time
import warnings
from typing import Dict, List, Optional
import torch
from filelock import FileLock
from torch.utils.data import Dataset
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
logger = logging.get_logger(__name__)
DEPRECATION_WARNING = (
"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: {0}"
)
class TextDataset(Dataset):
"""
This will be superseded by a framework-agnostic approach soon.
"""
def __init__(
self,
tokenizer: PreTrainedTokenizer,
file_path: str,
block_size: int,
overwrite_cache=False,
cache_dir: Optional[str] = None,
):
warnings.warn(
DEPRECATION_WARNING.format(
"https://github.com/huggingface/transformers/blob/main/examples/pytorch/language-modeling/run_mlm.py"
),
FutureWarning,
)
if os.path.isfile(file_path) is False:
raise ValueError(f"Input file path {file_path} not found")
block_size = block_size - tokenizer.num_special_tokens_to_add(pair=False)
directory, filename = os.path.split(file_path)
cached_features_file = os.path.join(
cache_dir if cache_dir is not None else directory,
f"cached_lm_{tokenizer.__class__.__name__}_{block_size}_{filename}",
)
# Make sure only the first process in distributed training processes the dataset,
# and the others will use the cache.
lock_path = cached_features_file + ".lock"
with FileLock(lock_path):
if os.path.exists(cached_features_file) and not overwrite_cache:
start = time.time()
with open(cached_features_file, "rb") as handle:
self.examples = pickle.load(handle)
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 {directory}")
self.examples = []
with open(file_path, encoding="utf-8") as f:
text = f.read()
tokenized_text = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(text))
for i in range(0, len(tokenized_text) - block_size + 1, block_size): # Truncate in block of block_size
self.examples.append(
tokenizer.build_inputs_with_special_tokens(tokenized_text[i : i + block_size])
)
# Note that we are losing the last truncated example here for the sake of simplicity (no padding)
# If your dataset is small, first you should look for a bigger one :-) and second you
# can change this behavior by adding (model specific) padding.
start = time.time()
with open(cached_features_file, "wb") as handle:
pickle.dump(self.examples, handle, protocol=pickle.HIGHEST_PROTOCOL)
logger.info(
f"Saving features into cached file {cached_features_file} [took {time.time() - start:.3f} s]"
)
def __len__(self):
return len(self.examples)
def __getitem__(self, i) -> torch.Tensor:
return torch.tensor(self.examples[i], dtype=torch.long)
class LineByLineTextDataset(Dataset):
"""
This will be superseded by a framework-agnostic approach soon.
"""
def __init__(self, tokenizer: PreTrainedTokenizer, file_path: str, block_size: int):
warnings.warn(
DEPRECATION_WARNING.format(
"https://github.com/huggingface/transformers/blob/main/examples/pytorch/language-modeling/run_mlm.py"
),
FutureWarning,
)
if os.path.isfile(file_path) is False:
raise ValueError(f"Input file path {file_path} not found")
# Here, we do not cache the features, operating under the assumption
# that we will soon use fast multithreaded tokenizers from the
# `tokenizers` repo everywhere =)
logger.info(f"Creating features from dataset file at {file_path}")
with open(file_path, encoding="utf-8") as f:
lines = [line for line in f.read().splitlines() if (len(line) > 0 and not line.isspace())]
batch_encoding = tokenizer(lines, add_special_tokens=True, truncation=True, max_length=block_size)
self.examples = batch_encoding["input_ids"]
self.examples = [{"input_ids": torch.tensor(e, dtype=torch.long)} for e in self.examples]
def __len__(self):
return len(self.examples)
def __getitem__(self, i) -> Dict[str, torch.tensor]:
return self.examples[i]
class LineByLineWithRefDataset(Dataset):
"""
This will be superseded by a framework-agnostic approach soon.
"""
def __init__(self, tokenizer: PreTrainedTokenizer, file_path: str, block_size: int, ref_path: str):
warnings.warn(
DEPRECATION_WARNING.format(
"https://github.com/huggingface/transformers/blob/main/examples/pytorch/language-modeling/run_mlm_wwm.py"
),
FutureWarning,
)
if os.path.isfile(file_path) is False:
raise ValueError(f"Input file path {file_path} not found")
if os.path.isfile(ref_path) is False:
raise ValueError(f"Ref file path {file_path} not found")
# Here, we do not cache the features, operating under the assumption
# that we will soon use fast multithreaded tokenizers from the
# `tokenizers` repo everywhere =)
logger.info(f"Creating features from dataset file at {file_path}")
logger.info(f"Use ref segment results at {ref_path}")
with open(file_path, encoding="utf-8") as f:
data = f.readlines() # use this method to avoid delimiter '\u2029' to split a line
data = [line.strip() for line in data if len(line) > 0 and not line.isspace()]
# Get ref inf from file
with open(ref_path, encoding="utf-8") as f:
ref = [json.loads(line) for line in f.read().splitlines() if (len(line) > 0 and not line.isspace())]
if len(data) != len(ref):
raise ValueError(
f"Length of Input file should be equal to Ref file. But the length of {file_path} is {len(data)} "
f"while length of {ref_path} is {len(ref)}"
)
batch_encoding = tokenizer(data, add_special_tokens=True, truncation=True, max_length=block_size)
self.examples = batch_encoding["input_ids"]
self.examples = [{"input_ids": torch.tensor(e, dtype=torch.long)} for e in self.examples]
n = len(self.examples)
for i in range(n):
self.examples[i]["chinese_ref"] = torch.tensor(ref[i], dtype=torch.long)
def __len__(self):
return len(self.examples)
def __getitem__(self, i) -> Dict[str, torch.tensor]:
return self.examples[i]
class LineByLineWithSOPTextDataset(Dataset):
"""
Dataset for sentence order prediction task, prepare sentence pairs for SOP task
"""
def __init__(self, tokenizer: PreTrainedTokenizer, file_dir: str, block_size: int):
warnings.warn(
DEPRECATION_WARNING.format(
"https://github.com/huggingface/transformers/blob/main/examples/pytorch/language-modeling/run_mlm.py"
),
FutureWarning,
)
if os.path.isdir(file_dir) is False:
raise ValueError(f"{file_dir} is not a directory")
logger.info(f"Creating features from dataset file folder at {file_dir}")
self.examples = []
# TODO: randomness could apply a random seed, ex. rng = random.Random(random_seed)
# file path looks like ./dataset/wiki_1, ./dataset/wiki_2
for file_name in os.listdir(file_dir):
file_path = os.path.join(file_dir, file_name)
if os.path.isfile(file_path) is False:
raise ValueError(f"{file_path} is not a file")
article_open = False
with open(file_path, encoding="utf-8") as f:
original_lines = f.readlines()
article_lines = []
for line in original_lines:
if "<doc id=" in line:
article_open = True
elif "</doc>" in line:
article_open = False
document = [
tokenizer.convert_tokens_to_ids(tokenizer.tokenize(line))
for line in article_lines[1:]
if (len(line) > 0 and not line.isspace())
]
examples = self.create_examples_from_document(document, block_size, tokenizer)
self.examples.extend(examples)
article_lines = []
else:
if article_open:
article_lines.append(line)
logger.info("Dataset parse finished.")
def create_examples_from_document(self, document, block_size, tokenizer, short_seq_prob=0.1):
"""Creates examples for a single document."""
# Account for special tokens
max_num_tokens = block_size - tokenizer.num_special_tokens_to_add(pair=True)
# We *usually* want to fill up the entire sequence since we are padding
# to `block_size` anyways, so short sequences are generally wasted
# computation. However, we *sometimes*
# (i.e., short_seq_prob == 0.1 == 10% of the time) want to use shorter
# sequences to minimize the mismatch between pretraining and fine-tuning.
# The `target_seq_length` is just a rough target however, whereas
# `block_size` is a hard limit.
target_seq_length = max_num_tokens
if random.random() < short_seq_prob:
target_seq_length = random.randint(2, max_num_tokens)
# We DON'T just concatenate all of the tokens from a document into a long
# sequence and choose an arbitrary split point because this would make the
# next sentence prediction task too easy. Instead, we split the input into
# segments "A" and "B" based on the actual "sentences" provided by the user
# input.
examples = []
current_chunk = [] # a buffer stored current working segments
current_length = 0
i = 0
while i < len(document):
segment = document[i] # get a segment
if not segment:
i += 1
continue
current_chunk.append(segment) # add a segment to current chunk
current_length += len(segment) # overall token length
# if current length goes to the target length or reaches the end of file, start building token a and b
if i == len(document) - 1 or current_length >= target_seq_length:
if current_chunk:
# `a_end` is how many segments from `current_chunk` go into the `A` (first) sentence.
a_end = 1
# if current chunk has more than 2 sentences, pick part of it `A` (first) sentence
if len(current_chunk) >= 2:
a_end = random.randint(1, len(current_chunk) - 1)
# token a
tokens_a = []
for j in range(a_end):
tokens_a.extend(current_chunk[j])
# token b
tokens_b = []
for j in range(a_end, len(current_chunk)):
tokens_b.extend(current_chunk[j])
if len(tokens_a) == 0 or len(tokens_b) == 0:
continue
# switch tokens_a and tokens_b randomly
if random.random() < 0.5:
is_next = False
tokens_a, tokens_b = tokens_b, tokens_a
else:
is_next = True
def truncate_seq_pair(tokens_a, tokens_b, max_num_tokens):
"""Truncates a pair of sequences to a maximum sequence length."""
while True:
total_length = len(tokens_a) + len(tokens_b)
if total_length <= max_num_tokens:
break
trunc_tokens = tokens_a if len(tokens_a) > len(tokens_b) else tokens_b
if not (len(trunc_tokens) >= 1):
raise ValueError("Sequence length to be truncated must be no less than one")
# We want to sometimes truncate from the front and sometimes from the
# back to add more randomness and avoid biases.
if random.random() < 0.5:
del trunc_tokens[0]
else:
trunc_tokens.pop()
truncate_seq_pair(tokens_a, tokens_b, max_num_tokens)
if not (len(tokens_a) >= 1):
raise ValueError(f"Length of sequence a is {len(tokens_a)} which must be no less than 1")
if not (len(tokens_b) >= 1):
raise ValueError(f"Length of sequence b is {len(tokens_b)} which must be no less than 1")
# add special tokens
input_ids = tokenizer.build_inputs_with_special_tokens(tokens_a, tokens_b)
# add token type ids, 0 for sentence a, 1 for sentence b
token_type_ids = tokenizer.create_token_type_ids_from_sequences(tokens_a, tokens_b)
example = {
"input_ids": torch.tensor(input_ids, dtype=torch.long),
"token_type_ids": torch.tensor(token_type_ids, dtype=torch.long),
"sentence_order_label": torch.tensor(0 if is_next else 1, dtype=torch.long),
}
examples.append(example)
current_chunk = [] # clear current chunk
current_length = 0 # reset current text length
i += 1 # go to next line
return examples
def __len__(self):
return len(self.examples)
def __getitem__(self, i) -> Dict[str, torch.tensor]:
return self.examples[i]
class TextDatasetForNextSentencePrediction(Dataset):
"""
This will be superseded by a framework-agnostic approach soon.
"""
def __init__(
self,
tokenizer: PreTrainedTokenizer,
file_path: str,
block_size: int,
overwrite_cache=False,
short_seq_probability=0.1,
nsp_probability=0.5,
):
warnings.warn(
DEPRECATION_WARNING.format(
"https://github.com/huggingface/transformers/blob/main/examples/pytorch/language-modeling/run_mlm.py"
),
FutureWarning,
)
if not os.path.isfile(file_path):
raise ValueError(f"Input file path {file_path} not found")
self.short_seq_probability = short_seq_probability
self.nsp_probability = nsp_probability
directory, filename = os.path.split(file_path)
cached_features_file = os.path.join(
directory,
f"cached_nsp_{tokenizer.__class__.__name__}_{block_size}_{filename}",
)
self.tokenizer = tokenizer
# Make sure only the first process in distributed training processes the dataset,
# and the others will use the cache.
lock_path = cached_features_file + ".lock"
# Input file format:
# (1) One sentence per line. These should ideally be actual sentences, not
# entire paragraphs or arbitrary spans of text. (Because we use the
# sentence boundaries for the "next sentence prediction" task).
# (2) Blank lines between documents. Document boundaries are needed so
# that the "next sentence prediction" task doesn't span between documents.
#
# Example:
# I am very happy.
# Here is the second sentence.
#
# A new document.
with FileLock(lock_path):
if os.path.exists(cached_features_file) and not overwrite_cache:
start = time.time()
with open(cached_features_file, "rb") as handle:
self.examples = pickle.load(handle)
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 {directory}")
self.documents = [[]]
with open(file_path, encoding="utf-8") as f:
while True:
line = f.readline()
if not line:
break
line = line.strip()
# Empty lines are used as document delimiters
if not line and len(self.documents[-1]) != 0:
self.documents.append([])
tokens = tokenizer.tokenize(line)
tokens = tokenizer.convert_tokens_to_ids(tokens)
if tokens:
self.documents[-1].append(tokens)
logger.info(f"Creating examples from {len(self.documents)} documents.")
self.examples = []
for doc_index, document in enumerate(self.documents):
self.create_examples_from_document(document, doc_index, block_size)
start = time.time()
with open(cached_features_file, "wb") as handle:
pickle.dump(self.examples, handle, protocol=pickle.HIGHEST_PROTOCOL)
logger.info(
f"Saving features into cached file {cached_features_file} [took {time.time() - start:.3f} s]"
)
def create_examples_from_document(self, document: List[List[int]], doc_index: int, block_size: int):
"""Creates examples for a single document."""
max_num_tokens = block_size - self.tokenizer.num_special_tokens_to_add(pair=True)
# We *usually* want to fill up the entire sequence since we are padding
# to `block_size` anyways, so short sequences are generally wasted
# computation. However, we *sometimes*
# (i.e., short_seq_prob == 0.1 == 10% of the time) want to use shorter
# sequences to minimize the mismatch between pretraining and fine-tuning.
# The `target_seq_length` is just a rough target however, whereas
# `block_size` is a hard limit.
target_seq_length = max_num_tokens
if random.random() < self.short_seq_probability:
target_seq_length = random.randint(2, max_num_tokens)
current_chunk = [] # a buffer stored current working segments
current_length = 0
i = 0
while i < len(document):
segment = document[i]
current_chunk.append(segment)
current_length += len(segment)
if i == len(document) - 1 or current_length >= target_seq_length:
if current_chunk:
# `a_end` is how many segments from `current_chunk` go into the `A`
# (first) sentence.
a_end = 1
if len(current_chunk) >= 2:
a_end = random.randint(1, len(current_chunk) - 1)
tokens_a = []
for j in range(a_end):
tokens_a.extend(current_chunk[j])
tokens_b = []
if len(current_chunk) == 1 or random.random() < self.nsp_probability:
is_random_next = True
target_b_length = target_seq_length - len(tokens_a)
# This should rarely go for more than one iteration for large
# corpora. However, just to be careful, we try to make sure that
# the random document is not the same as the document
# we're processing.
for _ in range(10):
random_document_index = random.randint(0, len(self.documents) - 1)
if random_document_index != doc_index:
break
random_document = self.documents[random_document_index]
random_start = random.randint(0, len(random_document) - 1)
for j in range(random_start, len(random_document)):
tokens_b.extend(random_document[j])
if len(tokens_b) >= target_b_length:
break
# We didn't actually use these segments so we "put them back" so
# they don't go to waste.
num_unused_segments = len(current_chunk) - a_end
i -= num_unused_segments
# Actual next
else:
is_random_next = False
for j in range(a_end, len(current_chunk)):
tokens_b.extend(current_chunk[j])
if not (len(tokens_a) >= 1):
raise ValueError(f"Length of sequence a is {len(tokens_a)} which must be no less than 1")
if not (len(tokens_b) >= 1):
raise ValueError(f"Length of sequence b is {len(tokens_b)} which must be no less than 1")
# add special tokens
input_ids = self.tokenizer.build_inputs_with_special_tokens(tokens_a, tokens_b)
# add token type ids, 0 for sentence a, 1 for sentence b
token_type_ids = self.tokenizer.create_token_type_ids_from_sequences(tokens_a, tokens_b)
example = {
"input_ids": torch.tensor(input_ids, dtype=torch.long),
"token_type_ids": torch.tensor(token_type_ids, dtype=torch.long),
"next_sentence_label": torch.tensor(1 if is_random_next else 0, dtype=torch.long),
}
self.examples.append(example)
current_chunk = []
current_length = 0
i += 1
def __len__(self):
return len(self.examples)
def __getitem__(self, i):
return self.examples[i]
|