import json import random import re from dataclasses import dataclass, field from typing import Any, Dict, List, Optional, Union import numpy as np import pandas as pd import torch import torchaudio import transformers import datasets from datasets import ClassLabel, load_dataset, load_metric from transformers import (Trainer, TrainingArguments, Wav2Vec2CTCTokenizer, Wav2Vec2FeatureExtractor, Wav2Vec2ForCTC, Wav2Vec2Processor) import argparse parser = argparse.ArgumentParser() parser.add_argument('--model', type=str, default="facebook/wav2vec2-xls-r-300m") parser.add_argument('--unfreeze', action='store_true') parser.add_argument('--lr', type=float, default=3e-4) parser.add_argument('--warmup', type=float, default=500) args = parser.parse_args() print(f"args: {args}") common_voice_train = datasets.load_dataset("mozilla-foundation/common_voice_8_0", "zh-HK", split="train+validation", use_auth_token=True) common_voice_test = datasets.load_dataset("mozilla-foundation/common_voice_8_0", "zh-HK", split="test[:10%]", use_auth_token=True) # common_voice_train = datasets.load_dataset("common_voice", "zh-HK", split="train+validation", use_auth_token=True) # common_voice_test = datasets.load_dataset("common_voice", "zh-HK", split="test[:10%]", use_auth_token=True) unused_cols = ["accent", "age", "client_id", "down_votes", "gender", "locale", "segment", "up_votes"] common_voice_train = common_voice_train.remove_columns(unused_cols) common_voice_test = common_voice_test.remove_columns(unused_cols) chars_to_ignore_regex = '[\丶\,\?\.\!\-\;\:"\“\%\‘\”\�\．\⋯\！\－\：\–\。\》\,\）\,\？\；\～\~\…\︰\，\（\」\‧\《\﹔\、\—\／\,\「\﹖\·\']' import string def remove_special_characters(batch): sen = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower() + " " # convert 'D' and 'd' to '啲' if there a 'D' in sentence # hacky stuff, wont work on 'D', 'd' co-occure with normal english words # wont work on multiple 'D' if "d" in sen: if len([c for c in sen if c in string.ascii_lowercase]) == 1: sen = sen.replace("d", "啲") batch["sentence"] = sen return batch common_voice_train = common_voice_train.map(remove_special_characters) common_voice_test = common_voice_test.map(remove_special_characters) def extract_all_chars(batch): all_text = " ".join(batch["sentence"]) vocab = list(set(all_text)) return {"vocab": [vocab], "all_text": [all_text]} vocab_train = common_voice_train.map(extract_all_chars, batched=True, batch_size=-1, keep_in_memory=True, remove_columns=common_voice_train.column_names,) vocab_test = common_voice_test.map(extract_all_chars, batched=True, batch_size=-1, keep_in_memory=True, remove_columns=common_voice_test.column_names,) vocab_list = list(set(vocab_train["vocab"][0]) | set(vocab_test["vocab"][0])) vocab_list = [char for char in vocab_list if not char.isascii()] # remove english char from vocab_list, so tokenizer will replace english with [UNK] vocab_list.append(" ") # previous will remove " " from vocab_list vocab_dict = {v: k for k, v in enumerate(vocab_list)} vocab_dict["|"] = vocab_dict[" "] del vocab_dict[" "] vocab_dict["[UNK]"] = len(vocab_dict) vocab_dict["[PAD]"] = len(vocab_dict) with open("vocab.json", "w") as vocab_file: json.dump(vocab_dict, vocab_file) tokenizer = Wav2Vec2CTCTokenizer("./vocab.json", unk_token="[UNK]", pad_token="[PAD]", word_delimiter_token="|") feature_extractor = Wav2Vec2FeatureExtractor(feature_size=1, sampling_rate=16000, padding_value=0.0, do_normalize=True, return_attention_mask=True,) processor = Wav2Vec2Processor(feature_extractor=feature_extractor, tokenizer=tokenizer) processor.save_pretrained("./finetuned-wav2vec2-xls-r-300m-cantonese") # resamplers = { # 48000: torchaudio.transforms.Resample(48000, 16000), # 44100: torchaudio.transforms.Resample(44100, 16000), # } # def load_and_resample(batch): # speech_array, sampling_rate = torchaudio.load(batch["path"]) # batch["array"] = resamplers[sampling_rate](speech_array).squeeze().numpy() # batch["sampling_rate"] = 16_000 # batch["target_text"] = batch["sentence"] # return batch # common_voice_train = common_voice_train.map(load_and_resample, remove_columns=common_voice_train.column_names,) # common_voice_test = common_voice_test.map(load_and_resample, remove_columns=common_voice_test.column_names,) common_voice_train = common_voice_train.cast_column('audio', datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)) common_voice_test = common_voice_test.cast_column('audio', datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)) def prepare_dataset(batch): batch["input_values"] = processor(batch["array"], sampling_rate=batch["sampling_rate"][0]).input_values with processor.as_target_processor(): batch["labels"] = processor(batch["target_text"]).input_ids return batch print(common_voice_train[0]['audio']) common_voice_train = common_voice_train.map(prepare_dataset, remove_columns=common_voice_train.column_names, batched=True,) common_voice_test = common_voice_test.map(prepare_dataset, remove_columns=common_voice_test.column_names, batched=True,) @dataclass class DataCollatorCTCWithPadding: """ Data collator that will dynamically pad the inputs received. Args: processor (:class:`~transformers.Wav2Vec2Processor`) The processor used for proccessing the data. padding (:obj:`bool`, :obj:`str` or :class:`~transformers.tokenization_utils_base.PaddingStrategy`, `optional`, defaults to :obj:`True`): Select a strategy to pad the returned sequences (according to the model's padding side and padding index) among: * :obj:`True` or :obj:`'longest'`: Pad to the longest sequence in the batch (or no padding if only a single sequence if provided). * :obj:`'max_length'`: Pad to a maximum length specified with the argument :obj:`max_length` or to the maximum acceptable input length for the model if that argument is not provided. * :obj:`False` or :obj:`'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of different lengths). max_length (:obj:`int`, `optional`): Maximum length of the ``input_values`` of the returned list and optionally padding length (see above). max_length_labels (:obj:`int`, `optional`): Maximum length of the ``labels`` returned list and optionally padding length (see above). pad_to_multiple_of (:obj:`int`, `optional`): If set will pad the sequence to a multiple of the provided value. This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >= 7.5 (Volta). """ processor: Wav2Vec2Processor padding: Union[bool, str] = True max_length: Optional[int] = None max_length_labels: Optional[int] = None pad_to_multiple_of: Optional[int] = None pad_to_multiple_of_labels: Optional[int] = None 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 lenghts and need # different padding methods input_features = [ {"input_values": feature["input_values"]} for feature in features ] label_features = [{"input_ids": feature["labels"]} for feature in features] batch = self.processor.pad( input_features, padding=self.padding, max_length=self.max_length, pad_to_multiple_of=self.pad_to_multiple_of, return_tensors="pt", ) with self.processor.as_target_processor(): labels_batch = self.processor.pad( label_features, padding=self.padding, max_length=self.max_length_labels, pad_to_multiple_of=self.pad_to_multiple_of_labels, 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 ) batch["labels"] = labels return batch data_collator = DataCollatorCTCWithPadding(processor=processor, padding=True) # cer_metric = load_metric("./cer") # def compute_metrics(pred): # pred_logits = pred.predictions # pred_ids = np.argmax(pred_logits, axis=-1) # pred.label_ids[pred.label_ids == -100] = processor.tokenizer.pad_token_id # pred_str = processor.batch_decode(pred_ids) # # we do not want to group tokens when computing the metrics # label_str = processor.batch_decode(pred.label_ids, group_tokens=False) # cer = cer_metric.compute(predictions=pred_str, references=label_str) # return {"cer": cer} def compute_metrics(pred): pred_logits = pred.predictions pred_ids = np.argmax(pred_logits, axis=-1) pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id pred_str = tokenizer.batch_decode(pred_ids) # we do not want to group tokens when computing the metrics label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False) metrics = {k: v.compute(predictions=pred_str, references=label_str) for k, v in eval_metrics.items()} return metrics model = Wav2Vec2ForCTC.from_pretrained( args.model, attention_dropout=0.1, hidden_dropout=0.1, feat_proj_dropout=0.0, mask_time_prob=0.05, layerdrop=0.1, gradient_checkpointing=True, ctc_loss_reduction="mean", pad_token_id=processor.tokenizer.pad_token_id, vocab_size=len(processor.tokenizer), ) if not args.unfreeze: model.freeze_feature_extractor() training_args = TrainingArguments( output_dir="./finetuned-wav2vec2-xls-r-300m-cantonese/wav2vec2-xls-r-300m-cantonese", group_by_length=True, per_device_train_batch_size=8, gradient_accumulation_steps=2, #evaluation_strategy="no", evaluation_strategy="steps", #evaluation_strategy="epoch", eval_steps=400, #eval_accumulation_steps=60, num_train_epochs=1, fp16=True, fp16_backend="amp", logging_strategy="steps", logging_steps=400, #logging_strategy="epoch", learning_rate=args.lr, warmup_steps=100, save_steps=2376, # every 3 epoch with batch_size 8 #save_strategy="epoch", save_total_limit=3, ################### # fp16_full_eval=True, dataloader_num_workers=20, ) trainer = Trainer( model=model, data_collator=data_collator, args=training_args, compute_metrics=compute_metrics, train_dataset=common_voice_train, eval_dataset=common_voice_test, tokenizer=processor.feature_extractor, ) trainer.train()