File size: 19,197 Bytes
0a7c458 8ddbc24 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 0a7c458 8b460b7 |
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 |
#!/usr/bin/env python
# coding=utf-8
#
# 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
""" Fine-tuning a 🤗 Transformers Whisper model for automatic speech recognition"""
import functools
import json
import logging
import os
import re
import sys
import warnings
from dataclasses import dataclass, field
from typing import Any, Dict, List, Optional, Union
import evaluate
import numpy as np
import torch
from pprint import pprint
import evaluate
from datasets import DatasetDict, load_dataset
from datasets import Audio
from transformers import (
HfArgumentParser,
TrainingArguments,
set_seed,
WhisperFeatureExtractor,
WhisperTokenizer,
WhisperForConditionalGeneration,
WhisperProcessor,
Seq2SeqTrainer,
Seq2SeqTrainingArguments,
)
from transformers.trainer_utils import get_last_checkpoint, is_main_process
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
def list_field(default=None, metadata=None):
return field(default_factory=lambda: default, metadata=metadata)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
"""
model_name_or_path: str = field(
metadata={
"help": "Path to pretrained model or model identifier from huggingface.co/models"}
)
language: str = field(
metadata={"help": "Whisper specific language"}
)
task: str = field(
metadata={
"help": "Whisper specific task, i.e., 'transcribe' or 'translate'"}
)
tokenizer_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": "Path to pretrained tokenizer or tokenizer identifier from huggingface.co/models"},
)
cache_dir: Optional[str] = field(
default=None,
metadata={
"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
)
freeze_feature_encoder: bool = field(
default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
)
attention_dropout: float = field(
default=0.0, metadata={"help": "The dropout ratio for the attention probabilities."}
)
activation_dropout: float = field(
default=0.0, metadata={"help": "The dropout ratio for activations inside the fully connected layer."}
)
feat_proj_dropout: float = field(default=0.0, metadata={
"help": "The dropout ratio for the projected features."})
hidden_dropout: float = field(
default=0.0,
metadata={
"help": "The dropout probability for all fully connected layers in the embeddings, encoder, and pooler."
},
)
final_dropout: float = field(
default=0.0,
metadata={
"help": "The dropout probability for the final projection layer."},
)
mask_time_prob: float = field(
default=0.05,
metadata={
"help": "Probability of each feature vector along the time axis to be chosen as the start of the vector"
"span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
"vectors will be masked along the time axis."
},
)
mask_time_length: int = field(
default=10,
metadata={"help": "Length of vector span to mask along the time axis."},
)
mask_feature_prob: float = field(
default=0.0,
metadata={
"help": "Probability of each feature vector along the feature axis to be chosen as the start of the vector"
"span to be masked. Approximately ``mask_feature_prob * sequence_length // mask_feature_length`` feature bins will be masked along the time axis."
},
)
mask_feature_length: int = field(
default=10,
metadata={"help": "Length of vector span to mask along the feature axis."},
)
layerdrop: float = field(default=0.0, metadata={
"help": "The LayerDrop probability."})
ctc_loss_reduction: Optional[str] = field(
default="mean", metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."}
)
ctc_zero_infinity: Optional[bool] = field(
default=False, metadata={"help": "If True, will try yo aboud the CTC loss goinf to infinity."}
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
Using `HfArgumentParser` we can turn this class
into argparse arguments to be able to specify them on
the command line.
"""
dataset_name: str = field(
metadata={
"help": "The configuration name of the dataset to use (via the datasets library)."}
)
dataset_config_name: str = field(
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
train_split_name: str = field(
default="train",
metadata={
"help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
},
)
eval_split_name: str = field(
default="test",
metadata={
"help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
},
)
audio_column_name: str = field(
default="audio",
metadata={
"help": "The name of the dataset column containing the audio data. Defaults to 'audio'"},
)
text_column_name: str = field(
default="text",
metadata={
"help": "The name of the dataset column containing the text data. Defaults to 'text'"},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
max_train_samples: Optional[int] = field(
default=None,
metadata={
"help": "For debugging purposes or quicker training, truncate the number of training examples to this "
"value if set."
},
)
max_eval_samples: Optional[int] = field(
default=None,
metadata={
"help": "For debugging purposes or quicker training, truncate the number of validation examples to this "
"value if set."
},
)
chars_to_ignore: Optional[List[str]] = list_field(
default=None,
metadata={"help": "A list of characters to remove from the transcripts."},
)
eval_metrics: List[str] = list_field(
default=["wer"],
metadata={
"help": "A list of metrics the model should be evaluated on. E.g. `'wer cer'`"},
)
max_duration_in_seconds: float = field(
default=20.0,
metadata={
"help": "Filter audio files that are longer than `max_duration_in_seconds` seconds to 'max_duration_in_seconds`"
},
)
min_duration_in_seconds: float = field(
default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
)
preprocessing_only: bool = field(
default=False,
metadata={
"help": "Whether to only do data preprocessing and skip training. "
"This is especially useful when data preprocessing errors out in distributed training due to timeout. "
"In this case, one should run the preprocessing in a non-distributed setup with `preprocessing_only=True` "
"so that the cached datasets can consequently be loaded in distributed training"
},
)
use_auth_token: bool = field(
default=False,
metadata={
"help": "If :obj:`True`, will use the token generated when running"
":obj:`transformers-cli login` as HTTP bearer authorization for remote files."
},
)
unk_token: str = field(
default="[UNK]",
metadata={"help": "The unk token for the tokenizer"},
)
pad_token: str = field(
default="[PAD]",
metadata={"help": "The padding token for the tokenizer"},
)
word_delimiter_token: str = field(
default="|",
metadata={"help": "The word delimiter token for the tokenizer"},
)
predict_with_generate: bool = field(
default=True,
metadata={
"help": "Output tokens in addition to loss and digits for calculating metrics"},
)
generation_max_length: int = field(
default=225,
metadata={"help": "Maximum number of tokens generated"},
)
phoneme_language: Optional[str] = field(
default=None,
metadata={
"help": "The target language that should be used be"
" passed to the tokenizer for tokenization. Note that"
" this is only relevant if the model classifies the"
" input audio to a sequence of phoneme sequences."
},
)
print_training_arguments: bool = field(
default=True,
metadata={
"help": "Prints the training arguments. For debugging"
},
)
@dataclass
class DataCollatorSpeechSeq2SeqWithPadding:
processor: Any
def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
# split inputs and labels since they have to be of different lengths and need different padding methods
# first treat the audio inputs by simply returning torch tensors
input_features = [{"input_features": feature["input_features"]}
for feature in features]
batch = self.processor.feature_extractor.pad(
input_features, return_tensors="pt")
# get the tokenized label sequences
label_features = [{"input_ids": feature["labels"]}
for feature in features]
# pad the labels to max length
labels_batch = self.processor.tokenizer.pad(
label_features, return_tensors="pt")
# replace padding with -100 to ignore loss correctly
labels = labels_batch["input_ids"].masked_fill(
labels_batch.attention_mask.ne(1), -100)
# if bos token is appended in previous tokenization step,
# cut bos token here as it's append later anyways
if (labels[:, 0] == self.processor.tokenizer.bos_token_id).all().cpu().item():
labels = labels[:, 1:]
batch["labels"] = labels
return batch
def main():
# 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.
parser = HfArgumentParser(
(ModelArguments, DataTrainingArguments, TrainingArguments))
model_args, data_args, training_args = parser.parse_args_into_dataclasses()
# Metrics
def compute_metrics(pred):
pred_ids = pred.predictions
label_ids = pred.label_ids
# replace -100 with the pad_token_id
label_ids[label_ids == -100] = tokenizer.pad_token_id
# we do not want to group tokens when computing the metrics
pred_str = tokenizer.batch_decode(pred_ids, skip_special_tokens=True)
label_str = tokenizer.batch_decode(label_ids, skip_special_tokens=True)
wer = 100 * metric.compute(predictions=pred_str, references=label_str)
return {"wer": wer}
# Prepare dataset
def prepare_dataset(batch):
# load and resample audio data from 48 to 16kHz
audio = batch["audio"]
# compute log-Mel input features from input audio array
batch["input_features"] = feature_extractor(
audio["array"], sampling_rate=audio["sampling_rate"]).input_features[0]
# encode target text to label ids
batch["labels"] = tokenizer(batch["sentence"]).input_ids
return batch
def print_training_arguments(model_args, data_args, training_args):
print("Starting with the following parameters:")
print("\n* Model arguments:")
pprint(vars(model_args), indent=2)
print("\n* Data arguments")
pprint(vars(data_args), indent=2)
print("\n* Training arguments")
pprint(vars(training_args), indent=2)
# Print training arguments
if data_args.print_training_arguments:
print_training_arguments(model_args, data_args, training_args)
# Load dataset
train_dataset = load_dataset(data_args.dataset_name, data_args.dataset_config_name, split="train", streaming=True, use_auth_token=True)
eval_dataset = load_dataset(data_args.dataset_name, data_args.dataset_config_name, split="validation", streaming=True, use_auth_token=True)
# Rename columns
if data_args.audio_column_name != "audio":
train_dataset = train_dataset.rename_column(data_args.audio_column_name, "audio")
eval_dataset = eval_dataset.rename_column(data_args.audio_column_name, "audio")
if data_args.text_column_name != "sentence":
train_dataset = train_dataset.rename_column(data_args.text_column_name, "sentence")
eval_dataset = eval_dataset.rename_column(data_args.text_column_name, "sentence")
# Initialise
feature_extractor = WhisperFeatureExtractor.from_pretrained(
model_args.model_name_or_path)
tokenizer = WhisperTokenizer.from_pretrained(
model_args.model_name_or_path, language=model_args.language, task=model_args.task)
processor = WhisperProcessor.from_pretrained(
model_args.model_name_or_path, language=model_args.language, task=model_args.task)
data_collator = DataCollatorSpeechSeq2SeqWithPadding(processor=processor)
# Prepare data
train_dataset = train_dataset.cast_column("audio", Audio(sampling_rate=16000))
eval_dataset = eval_dataset.cast_column("audio", Audio(sampling_rate=16000))
# TODO Not able to implement in Streaming mode. Can not find a way to list columns. But is is necessary?
# train_data = train_data.map(prepare_dataset, remove_columns=train_data.column_names, num_proc=1)
train_dataset = train_dataset.map(prepare_dataset)
# Metrics
metric = evaluate.load("wer")
# Detecting last checkpoint.
last_checkpoint = None
if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir:
last_checkpoint = get_last_checkpoint(training_args.output_dir)
if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0:
raise ValueError(
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
)
# Training
if training_args.do_train:
# use last checkpoint if exist
if last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
# We need to set use_cache=False here if we want to use gradient accumulation
model = WhisperForConditionalGeneration.from_pretrained(
"openai/whisper-small", use_cache=False)
# Overriding generation arguments - no tokens are forced as decoder outputs (see [`forced_decoder_ids`](https://huggingface.co/docs/transformers/main_classes/text_generation#transformers.generation_utils.GenerationMixin.generate.forced_decoder_ids)), no tokens are suppressed during generation (see [`suppress_tokens`](https://huggingface.co/docs/transformers/main_classes/text_generation#transformers.generation_utils.GenerationMixin.generate.suppress_tokens)):
model.config.forced_decoder_ids = None
model.config.suppress_tokens = []
# Set seed before initializing model.
set_seed(training_args.seed)
trainer = Seq2SeqTrainer(
args=training_args,
model=model,
train_dataset=train_dataset.with_format("torch"),
eval_dataset=eval_dataset.with_format("torch"),
data_collator=data_collator,
compute_metrics=compute_metrics,
tokenizer=processor.feature_extractor,
)
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model()
metrics = train_result.metrics
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
if training_args.push_to_hub:
trainer.push_to_hub(**kwargs)
else:
trainer.create_model_card(**kwargs)
# TODO - Look closer into the evaluation and the model card writing.
# Evaluation
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
metrics = trainer.evaluate()
max_eval_samples = (
data_args.max_eval_samples if data_args.max_eval_samples is not None else len(
vectorized_datasets["eval"])
)
metrics["eval_samples"] = min(
max_eval_samples, len(vectorized_datasets["eval"]))
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
# Write model card and (optionally) push to hub
config_name = data_args.dataset_config_name if data_args.dataset_config_name is not None else "na"
kwargs = {
"finetuned_from": model_args.model_name_or_path,
"tasks": "automatic-speech-recognition",
"tags": ["hf-asr-leaderboard", "automatic-speech-recognition", data_args.dataset_name],
"dataset_args": f"Config: {config_name}, Training split: {data_args.train_split_name}, Eval split: {data_args.eval_split_name}",
"dataset": f"{data_args.dataset_name.upper()} - {config_name.upper()}",
"language": model_args.language,
}
return results
# XLA hook
def _mp_fn(index):
# For xla_spawn (TPUs)
print("The XLA is initiated")
main()
if __name__ == "__main__":
main()
|