run_speech_recognition_seq2seq.py

#!/usr/bin/env python
# coding=utf-8
# 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.
"""
Fine-tuning the library models for sequence to sequence speech recognition.
"""
# You can also adapt this script on your own sequence to sequence speech
# recognition task. Pointers for this are left as comments.

import logging
import os
import sys
from dataclasses import dataclass, field
from typing import Any, Dict, List, Optional, Union

import datasets
import torch
from datasets import DatasetDict, load_dataset, load_metric

import bitsandbytes as bnb
import transformers
from transformers import (
    AutoConfig,
    AutoFeatureExtractor,
    AutoModelForSpeechSeq2Seq,
    AutoProcessor,
    AutoTokenizer,
    HfArgumentParser,
    Seq2SeqTrainer,
    Seq2SeqTrainingArguments,
    set_seed,
)
from transformers.trainer_pt_utils import get_parameter_names
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
from transformers.optimization import Adafactor


# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")

require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/summarization/requirements.txt")

logger = logging.getLogger(__name__)


@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"}
    )
    config_name: Optional[str] = field(
        default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
    )
    tokenizer_name: Optional[str] = field(
        default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
    )
    feature_extractor_name: Optional[str] = field(
        default=None, metadata={"help": "feature extractor name or path if not the same as model_name"}
    )
    cache_dir: Optional[str] = field(
        default=None,
        metadata={"help": "Where to store the pretrained models downloaded from huggingface.co"},
    )
    use_fast_tokenizer: bool = field(
        default=True,
        metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
    )
    model_revision: str = field(
        default="main",
        metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
    )
    use_auth_token: bool = field(
        default=False,
        metadata={
            "help": "Will use the token generated when running `transformers-cli login` (necessary to use this script "
            "with private models)."
        },
    )
    freeze_feature_encoder: bool = field(
        default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
    )


@dataclass
class DataTrainingArguments:
    """
    Arguments pertaining to what data we are going to input our model for training and eval.
    """

    dataset_name: str = field(
        default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
    )
    dataset_config_name: Optional[str] = field(
        default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
    )
    text_column: Optional[str] = field(
        default=None,
        metadata={"help": "The name of the column in the datasets containing the full texts (for summarization)."},
    )
    overwrite_cache: bool = field(
        default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
    )
    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 evaluation examples to this "
            "value if set."
        },
    )
    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'"},
    )
    max_duration_in_seconds: float = field(
        default=20.0,
        metadata={
            "help": "Truncate 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"
        },
    )
    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'"
        },
    )
    do_lower_case: bool = field(
        default=True,
        metadata={"help": "Whether the target text should be lower cased."},
    )


@dataclass
class DataCollatorSpeechSeq2SeqWithPadding:
    """
    Data collator that will dynamically pad the inputs received.
    Args:
        processor ([`Wav2Vec2Processor`])
            The processor used for proccessing the data.
        decoder_start_token_id (`int`)
            The begin-of-sentence of the decoder.
    """

    processor: Any
    decoder_start_token_id: int

    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.feature_extractor.pad(input_features, return_tensors="pt")

        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.decoder_start_token_id).all().cpu().item():
            labels = labels[:, 1:]

        batch["labels"] = labels

        return batch


def main():
    # 1. Parse input arguments
    # 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, Seq2SeqTrainingArguments))

    if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
        # If we pass only one argument to the script and it's the path to a json file,
        # let's parse it to get our arguments.
        model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
    else:
        model_args, data_args, training_args = parser.parse_args_into_dataclasses()

    # 2. Setup logging
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        handlers=[logging.StreamHandler(sys.stdout)],
    )
    log_level = training_args.get_process_log_level()
    logger.setLevel(log_level)
    datasets.utils.logging.set_verbosity(log_level)
    transformers.utils.logging.set_verbosity(log_level)
    transformers.utils.logging.enable_default_handler()
    transformers.utils.logging.enable_explicit_format()

    logger.setLevel(logging.INFO if is_main_process(training_args.local_rank) else logging.WARN)

    # Log on each process the small summary:
    logger.warning(
        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
        f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
    )
    logger.info(f"Training/evaluation parameters {training_args}")

    # Set the verbosity to info of the Transformers logger (on main process only):
    if is_main_process(training_args.local_rank):
        transformers.utils.logging.set_verbosity_info()
    logger.info("Training/evaluation parameters %s", training_args)

    # 3. Detecting last checkpoint and eventualy continue from 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 and training_args.resume_from_checkpoint is None:
            logger.info(
                f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
                "the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
            )

    # Set seed before initializing model.
    set_seed(training_args.seed)

    # 4. Load dataset
    raw_datasets = DatasetDict()

    if training_args.do_train:
        raw_datasets["train"] = load_dataset(
            data_args.dataset_name, data_args.dataset_config_name, split=data_args.train_split_name
        )

    if training_args.do_eval:
        raw_datasets["eval"] = load_dataset(
            data_args.dataset_name, data_args.dataset_config_name, split=data_args.eval_split_name
        )

    if data_args.audio_column_name not in next(iter(raw_datasets.values())).column_names:
        raise ValueError(
            f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_name}'. "
            "Make sure to set `--audio_column_name` to the correct audio column - one of "
            f"{', '.join(next(iter(raw_datasets.values())).column_names)}."
        )

    if data_args.text_column_name not in next(iter(raw_datasets.values())).column_names:
        raise ValueError(
            f"--text_column_name {data_args.text_column_name} not found in dataset '{data_args.dataset_name}'. "
            "Make sure to set `--text_column_name` to the correct text column - one of "
            f"{', '.join(next(iter(raw_datasets.values())).column_names)}."
        )

    # 5. Load pretrained model, tokenizer, and feature extractor
    #
    # Distributed training:
    # The .from_pretrained methods guarantee that only one local process can concurrently
    config = AutoConfig.from_pretrained(
        model_args.config_name if model_args.config_name else model_args.model_name_or_path,
        cache_dir=model_args.cache_dir,
        revision=model_args.model_revision,
        use_auth_token=True if model_args.use_auth_token else None,
    )

    feature_extractor = AutoFeatureExtractor.from_pretrained(
        model_args.feature_extractor_name if model_args.feature_extractor_name else model_args.model_name_or_path,
        cache_dir=model_args.cache_dir,
        revision=model_args.model_revision,
        use_auth_token=True if model_args.use_auth_token else None,
    )
    tokenizer = AutoTokenizer.from_pretrained(
        model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path,
        cache_dir=model_args.cache_dir,
        use_fast=model_args.use_fast_tokenizer,
        revision=model_args.model_revision,
        use_auth_token=True if model_args.use_auth_token else None,
    )
    model = AutoModelForSpeechSeq2Seq.from_pretrained(
        model_args.model_name_or_path,
        config=config,
        cache_dir=model_args.cache_dir,
        revision=model_args.model_revision,
        use_auth_token=True if model_args.use_auth_token else None,
    )

    if model.config.decoder_start_token_id is None:
        raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined")

    if model_args.freeze_feature_encoder:
        model.freeze_feature_encoder()

    # 6. Resample speech dataset if necassary
    dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
    if dataset_sampling_rate != feature_extractor.sampling_rate:
        raw_datasets = raw_datasets.cast_column(
            data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
        )

    # 7. Preprocessing the datasets.
    # We need to read the audio files as arrays and tokenize the targets.
    max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
    min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
    audio_column_name = data_args.audio_column_name
    num_workers = data_args.preprocessing_num_workers
    text_column_name = data_args.text_column_name
    model_input_name = feature_extractor.model_input_names[0]
    do_lower_case = data_args.do_lower_case

    if data_args.max_train_samples is not None:
        raw_datasets["train"] = raw_datasets["train"].select(range(data_args.max_train_samples))

    if data_args.max_eval_samples is not None:
        raw_datasets["eval"] = raw_datasets["eval"].select(range(data_args.max_eval_samples))

    def prepare_dataset(batch):
        # process audio
        sample = batch[audio_column_name]
        inputs = feature_extractor(sample["array"], sampling_rate=sample["sampling_rate"])
        # process audio length
        batch[model_input_name] = inputs.input_values[0]
        batch["input_length"] = len(batch["input_values"])

        # process targets
        input_str = batch[text_column_name].lower() if do_lower_case else batch[text_column_name]
        batch["labels"] = tokenizer(input_str).input_ids
        return batch

    with training_args.main_process_first(desc="dataset map pre-processing"):
        vectorized_datasets = raw_datasets.map(
            prepare_dataset,
            remove_columns=next(iter(raw_datasets.values())).column_names,
            num_proc=data_args.preprocessing_num_workers,
            desc="preprocess train dataset",
        )

    # filter data that is shorter than min_input_length or longer than
    # max_input_length
    def is_audio_in_length_range(length):
        return length > min_input_length and length < max_input_length

    vectorized_datasets = vectorized_datasets.filter(
        is_audio_in_length_range,
        num_proc=num_workers,
        input_columns=["input_length"],
    )

    # for large datasets it is advised to run the preprocessing on a
    # single machine first with `args.preprocessing_only` since there will mostly likely
    # be a timeout when running the script in distributed mode.
    # In a second step `args.preprocessing_only` can then be set to `False` to load the
    # cached dataset
    if data_args.preprocessing_only:
        cache = {k: v.cache_files for k, v in vectorized_datasets.items()}
        logger.info(f"Data preprocessing finished. Files cached at {cache}.")
        return

    # 8. Load Metric
    metric = load_metric("wer")

    def compute_metrics(pred):
        pred_ids = pred.predictions

        pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id

        pred_str = tokenizer.batch_decode(pred_ids, skip_special_tokens=True)
        # we do not want to group tokens when computing the metrics
        label_str = tokenizer.batch_decode(pred.label_ids, skip_special_tokens=True)

        wer = metric.compute(predictions=pred_str, references=label_str)

        return {"wer": wer}

    # 9. Create a single speech processor
    if is_main_process(training_args.local_rank):
        # save feature extractor, tokenizer and config
        feature_extractor.save_pretrained(training_args.output_dir)
        tokenizer.save_pretrained(training_args.output_dir)
        config.save_pretrained(training_args.output_dir)

    processor = AutoProcessor.from_pretrained(training_args.output_dir)

    # 10. Define data collator
    data_collator = DataCollatorSpeechSeq2SeqWithPadding(
        processor=processor, decoder_start_token_id=model.config.decoder_start_token_id
    )
	
    decay_parameters = get_parameter_names(model, [torch.nn.LayerNorm])
    decay_parameters = [name for name in decay_parameters if "bias" not in name]
    optimizer_grouped_parameters = [
        {
            "params": [p for n, p in model.named_parameters() if n in decay_parameters],
            "weight_decay": training_args.weight_decay,
        },
        {
            "params": [p for n, p in model.named_parameters() if n not in decay_parameters],
            "weight_decay": 0.0,
        },
    ]    

    optimizer = bnb.optim.Adam8bit(
        params=optimizer_grouped_parameters,
        lr=training_args.learning_rate,
        betas=(training_args.adam_beta1, training_args.adam_beta2),
        eps=training_args.adam_epsilon,
    )

    """optimizer = Adafactor(
        params=optimizer_grouped_parameters,
        lr=training_args.learning_rate,
        beta1=training_args.adam_beta1,
        eps=training_args.adam_epsilon,
        relative_step=False,
    )"""


    optimizers = (optimizer, None)
	

    #11. Initialize Trainer

    trainer = Seq2SeqTrainer(
        model=model,
        args=training_args,
        train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
        eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
        tokenizer=feature_extractor,
        data_collator=data_collator,
        compute_metrics=compute_metrics if training_args.predict_with_generate else None,
		optimizers=optimizers,
    )

    # 12. Training
    if training_args.do_train:
        checkpoint = None
        if training_args.resume_from_checkpoint is not None:
            checkpoint = training_args.resume_from_checkpoint
        elif last_checkpoint is not None:
            checkpoint = last_checkpoint
        train_result = trainer.train(resume_from_checkpoint=checkpoint)
        trainer.save_model()  # Saves the feature extractor too for easy upload

        metrics = train_result.metrics
        max_train_samples = (
            data_args.max_train_samples
            if data_args.max_train_samples is not None
            else len(vectorized_datasets["train"])
        )
        metrics["train_samples"] = min(max_train_samples, len(vectorized_datasets["train"]))
        trainer.log_metrics("train", metrics)
        trainer.save_metrics("train", metrics)
        trainer.save_state()

    # 13. Evaluation
    results = {}
    if training_args.do_eval:
        logger.info("*** Evaluate ***")
        metrics = trainer.evaluate(
            metric_key_prefix="eval", max_length=model.config.max_length, num_beams=model.config.num_beams
        )
        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)

    # 14. Write Training Stats
    kwargs = {"finetuned_from": model_args.model_name_or_path, "tasks": "speech recognition"}
    if data_args.dataset_name is not None:
        kwargs["dataset_tags"] = data_args.dataset_name
        if data_args.dataset_config_name is not None:
            kwargs["dataset_args"] = data_args.dataset_config_name
            kwargs["dataset"] = f"{data_args.dataset_name} {data_args.dataset_config_name}"
        else:
            kwargs["dataset"] = data_args.dataset_name

    if training_args.push_to_hub:
        trainer.push_to_hub(**kwargs)
    else:
        trainer.create_model_card(**kwargs)

    return results


if __name__ == "__main__":
    main()