run_summarization-bhc-data-preprocessed-bart-guidance.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.
"""
# You can also adapt this script on your own sequence to sequence task. Pointers for this are left as comments.

"""
Adapted script for clinical BHC fine-tuning, for Abstractive encoder/decoder models such as T5 and BART as well as 
Encoder/Decoder models such BERT-2-BERT. 
"""

import json
import logging
import os
import sys
from dataclasses import dataclass, field
from typing import Optional

import datasets
import nltk  # Here to have a nice missing dependency error message early on
import numpy as np
import pandas as pd
from datasets import load_dataset, load_metric, load_from_disk, Dataset

import transformers
from filelock import FileLock
from transformers import (
    AutoConfig,
    AutoModelForSeq2SeqLM,
    AutoTokenizer,
    EncoderDecoderConfig,
    EncoderDecoderModel,
    HfArgumentParser,
    Seq2SeqTrainer,
    Seq2SeqTrainingArguments,
    set_seed,
)
from transformers.file_utils import is_offline_mode
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version


# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
from guidance_models.BARTModelWithGuidance import BartWithGuidanceForConditionalGeneration, \
    adapted_state_dict, DataCollatorForBartGuidanceSeq2Seq, BartGuidanceSeq2SeqTrainer

check_min_version("4.10.0.dev0")

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

logger = logging.getLogger(__name__)

try:
    nltk.data.find("tokenizers/punkt")
except (LookupError, OSError):
    if is_offline_mode():
        raise LookupError(
            "Offline mode: run this script without TRANSFORMERS_OFFLINE first to download nltk data files"
        )
    with FileLock(".lock") as lock:
        nltk.download("punkt", quiet=True)


@dataclass
class ModelArguments:
    """
    Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
    """

    model_name_or_path: str = field(
        default=None,
        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"}
    )
    encoder_name_or_path: str = field(
        default=None,
        metadata={"help": "Path to pretrained encoder model or identifier from huggingface.co/models"}
    )
    decoder_name_or_path: str = field(
        default=None,
        metadata={"help": "Path to pretrained decoder model or identifier from huggingface.co/models"}
    )
    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)."
        },
    )


@dataclass
class DataTrainingArguments:
    """
    Arguments pertaining to what data we are going to input our model for training and eval.
    """
    ds_path: Optional[str] = field(
        default=None, metadata={"help": "The name of the dataset to load via load_dataset_from_disk."}
    )
    text_column: Optional[str] = field(
        default=None,
        metadata={"help": "The name of the column in the datasets containing the full texts (for summarization)."},
    )
    summary_column: Optional[str] = field(
        default=None,
        metadata={"help": "The name of the column in the datasets containing the summaries (for summarization)."},
    ),
    guidance_summ_column: Optional[str] = field(
        default=None,
        metadata={"help": "The name of the column in the datasets containing the summary guidance signal. Used during training for teacher forcing"},
    ),
    guidance_text_column: Optional[str] = field(
        default=None,
        metadata={"help": "The name of the column in the datasets containing the full texts guidance signal, used for eval and testing"},
    ),
    train_file: Optional[str] = field(
        default=None, metadata={"help": "The input training data file (a jsonlines or csv file)."}
    )
    validation_file: Optional[str] = field(
        default=None,
        metadata={
            "help": "An optional input evaluation data file to evaluate the metrics (rouge) on "
            "(a jsonlines or csv file)."
        },
    )
    test_file: Optional[str] = field(
        default=None,
        metadata={
            "help": "An optional input test data file to evaluate the metrics (rouge) on " "(a jsonlines or csv file)."
        },
    )
    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_source_length: Optional[int] = field(
        default=1024,
        metadata={
            "help": "The maximum total input sequence length after tokenization. Sequences longer "
            "than this will be truncated, sequences shorter will be padded."
        },
    )
    max_target_length: Optional[int] = field(
        default=128,
        metadata={
            "help": "The maximum total sequence length for target text after tokenization. Sequences longer "
            "than this will be truncated, sequences shorter will be padded."
        },
    )
    val_max_target_length: Optional[int] = field(
        default=None,
        metadata={
            "help": "The maximum total sequence length for validation target text after tokenization. Sequences longer "
            "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
            "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
            "during ``evaluate`` and ``predict``."
        },
    )
    pad_to_max_length: bool = field(
        default=False,
        metadata={
            "help": "Whether to pad all samples to model maximum sentence length. "
            "If False, will pad the samples dynamically when batching to the maximum length in the batch. More "
            "efficient on GPU but very bad for TPU."
        },
    )
    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."
        },
    )
    max_predict_samples: Optional[int] = field(
        default=None,
        metadata={
            "help": "For debugging purposes or quicker training, truncate the number of prediction examples to this "
            "value if set."
        },
    )
    save_all_preds: bool = field(
        default=False,
        metadata={
            "help": "save all preds during evaluation or prediction, ignores num_rand_pred_samples if this is True"
        }
    ),
    num_rand_pred_samples: Optional[int] = field(
        default=None,
        metadata={
            "help": "Randomly sample a number of predictions, associated labels and their index, during validation, testing"
        }
    )
    num_beams: Optional[int] = field(
        default=None,
        metadata={
            "help": "Number of beams to use for evaluation. This argument will be passed to ``model.generate``, "
            "which is used during ``evaluate`` and ``predict``."
        },
    )
    ignore_pad_token_for_loss: bool = field(
        default=True,
        metadata={
            "help": "Whether to ignore the tokens corresponding to padded labels in the loss computation or not."
        },
    )
    source_prefix: Optional[str] = field(
        default=None, metadata={"help": "A prefix to add before every source text (useful for T5 models)."}
    )

    def __post_init__(self):
        """if self.dataset_name is None and self.train_file is None and self.validation_file is None:
            raise ValueError("Need either a dataset name or a training/validation file.")
        else:
            if self.train_file is not None:
                extension = self.train_file.split(".")[-1]
                assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
            if self.validation_file is not None:
                extension = self.validation_file.split(".")[-1]
                assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
        """
        if self.val_max_target_length is None:
            self.val_max_target_length = self.max_target_length


def main(path=None):
    # 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.

    # expect a .json cfg file.
    if path is None:
        path = os.path.abspath(sys.argv[1])
    parser = HfArgumentParser((ModelArguments, DataTrainingArguments, Seq2SeqTrainingArguments))
    model_args, data_args, training_args = parser.parse_json_file(json_file=path)

    # 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()

    # 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}")

    # 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 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)

    # Load pretrained model and tokenizer
    #
    # Distributed training:
    # The .from_pretrained methods guarantee that only one local process can concurrently
    # download model & vocab.
    
    # train test splits
    ds_path = data_args.ds_path
    ds = load_from_disk(ds_path)

    src_col = data_args.text_column
    ds = ds.train_test_split(train_size=0.8, test_size=0.2, shuffle=False)
    ds['train'].sort(src_col, reverse=True)
    val_test_ds = ds['test'].train_test_split(train_size=0.5, test_size=0.5, shuffle=False)
    val_ds = val_test_ds['train'].sort(src_col, reverse=True)
    test_ds = val_test_ds['test'].sort(src_col, reverse=True)
    raw_datasets = {
        'train': ds['train'],
        'validation': val_ds,
        'test': test_ds,
    }

    column_names = [n for n in ds['train'].features.keys() if n not in ('hadm_id', 'text_limd', 'clean_sum')]

    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,
    )

    # use the custom BART with guidance model
    model = BartWithGuidanceForConditionalGeneration.from_pretrained(model_args.model_name_or_path,
                                                                     state_dict=adapted_state_dict(model_args.model_name_or_path,
                                                                                                   cache_dir=model_args.cache_dir),
                                                                     cache_dir=model_args.cache_dir)
    model.resize_token_embeddings(len(tokenizer))

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

    # Temporarily set max_target_length for training.
    max_target_length = data_args.max_target_length
    padding = "max_length" if data_args.pad_to_max_length else False
    
    text_column = data_args.text_column
    summary_column = data_args.summary_column
    guidance_text_column = data_args.guidance_text_column
    guidance_summ_column = data_args.guidance_summ_column

    if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
        logger.warning(
            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for"
            f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
        )

    def preprocess_train_function(examples):
        """ passes extracted guidance from target texts """
        return preprocess_function(examples, training=True)

    def preprocess_eval_test_function(examples):
        """ passes extracted guidance from the training texts """
        return preprocess_function(examples, training=False)

    def preprocess_function(examples, training=True):
        inputs = examples[text_column]
        targets = examples[summary_column]

        # guidance inferred from the training only
        guidance_inputs = examples[guidance_text_column]
        # guidance inferred from the the target summary.
        guidance_targets = examples[guidance_summ_column]
        model_inputs = tokenizer(inputs, max_length=data_args.max_source_length, padding=padding, truncation=True)
        # Setup the tokenizer for targets
        with tokenizer.as_target_tokenizer():
            labels = tokenizer(targets, max_length=max_target_length, padding=padding, truncation=True)

        # If we are padding here, replace all tokenizer.pad_token_id in the labels by -100 when we want to ignore
        # padding in the loss.
        if padding == "max_length" and data_args.ignore_pad_token_for_loss:
            labels["input_ids"] = [
                [(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"]
            ]
        model_inputs["labels"] = labels["input_ids"]

        # pre-process-train
        if training:
            guidance_decoder_labels = tokenizer(guidance_targets, max_length=max_target_length, padding=padding,
                                                truncation=True)
            model_inputs["input_guidance_ids"] = guidance_decoder_labels['input_ids']
            model_inputs["attention_guidance_mask"] = guidance_decoder_labels['attention_mask']
        else:  # vs pre-process-val / pre-process-test
            guidance_encoder_inputs = tokenizer(guidance_inputs, max_length=data_args.max_source_length, padding=padding,
                                                truncation=True)
            model_inputs["input_guidance_ids"] = guidance_encoder_inputs['input_ids']
            model_inputs["attention_guidance_mask"] = guidance_encoder_inputs['attention_mask']
        return model_inputs

    if training_args.do_train:
        if "train" not in raw_datasets:
            raise ValueError("--do_train requires a train dataset")
        train_dataset = raw_datasets["train"]
        if data_args.max_train_samples is not None:
            train_dataset = train_dataset.select(range(data_args.max_train_samples))
        with training_args.main_process_first(desc="train dataset map pre-processing"):
            train_dataset = train_dataset.map(
                preprocess_train_function,
                batched=True,
                num_proc=data_args.preprocessing_num_workers,
                remove_columns=column_names,
                load_from_cache_file=not data_args.overwrite_cache,
                desc="Running tokenizer on train dataset",
            )

    if training_args.do_eval:
        max_target_length = data_args.val_max_target_length
        if "validation" not in raw_datasets:
            raise ValueError("--do_eval requires a validation dataset")
        eval_dataset = raw_datasets["validation"]
        if data_args.max_eval_samples is not None:
            eval_dataset = eval_dataset.select(range(data_args.max_eval_samples))
        with training_args.main_process_first(desc="validation dataset map pre-processing"):
            eval_dataset = eval_dataset.map(
                preprocess_eval_test_function,
                batched=True,
                num_proc=data_args.preprocessing_num_workers,
                remove_columns=column_names,
                load_from_cache_file=not data_args.overwrite_cache,
                desc="Running tokenizer on validation dataset",
            )

    if training_args.do_predict:
        max_target_length = data_args.val_max_target_length
        if "test" not in raw_datasets:
            raise ValueError("--do_predict requires a test dataset")
        predict_dataset = raw_datasets["test"]
        if data_args.max_predict_samples is not None:
            predict_dataset = predict_dataset.select(range(data_args.max_predict_samples))
        with training_args.main_process_first(desc="prediction dataset map pre-processing"):
            predict_dataset = predict_dataset.map(
                preprocess_eval_test_function,
                batched=True,
                num_proc=data_args.preprocessing_num_workers,
                remove_columns=column_names,
                load_from_cache_file=not data_args.overwrite_cache,
                desc="Running tokenizer on prediction dataset",
            )

    # Data collator
    label_pad_token_id = -100 if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id
    data_collator = DataCollatorForBartGuidanceSeq2Seq(
        tokenizer,
        model=model,
        label_pad_token_id=label_pad_token_id,
        pad_to_multiple_of=8 if training_args.fp16 else None,
    )

    # Metric
    metric = load_metric("rouge")

    def postprocess_text(preds, labels):
        preds = [pred.strip() for pred in preds]
        labels = [label.strip() for label in labels]

        # rougeLSum expects newline after each sentence
        preds = ["\n".join(nltk.sent_tokenize(pred)) for pred in preds]
        labels = ["\n".join(nltk.sent_tokenize(label)) for label in labels]

        return preds, labels

    def compute_metrics(eval_preds):
        preds, labels = eval_preds
        if isinstance(preds, tuple):
            preds = preds[0]
        decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True)
        if data_args.ignore_pad_token_for_loss:
            # Replace -100 in the labels as we can't decode them.
            labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
        decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)

        # Some simple post-processing
        decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels)

        # save a random sample of preds / labels
        if data_args.save_all_preds and eval_file_name:
            preds_labels_indices = [(i, decoded_preds[i], decoded_labels[i]) for i in range(len(decoded_preds))]
            with open(os.path.join(training_args.output_dir, eval_file_name), 'w') as f:
                json.dump(preds_labels_indices, f, indent=2)
        elif data_args.num_rand_pred_samples:
            rnd_preds = np.random.choice(range(len(decoded_preds)), data_args.num_rand_pred_samples, replace=False)
            preds_labels_indices = [(int(i), decoded_preds[i], decoded_labels[i]) for i in rnd_preds]
            with open(os.path.join(training_args.output_dir, 'rand_sample_preds.json'), 'w') as f:
                json.dump(preds_labels_indices, f, indent=2)
            logger.info('Saved a random sample of preds / labels / indices to disk in rand_sample_preds.json')

        result = metric.compute(predictions=decoded_preds, references=decoded_labels, use_stemmer=True)
        # Extract a few results from ROUGE - rouge reported in % points vs <1.0
        result = {key: value.mid.fmeasure * 100 for key, value in result.items()}

        prediction_lens = [np.count_nonzero(pred != tokenizer.pad_token_id) for pred in preds]
        result["gen_len"] = np.mean(prediction_lens)
        result = {k: round(v, 4) for k, v in result.items()}
        return result

    # Initialize our Trainer
    trainer = BartGuidanceSeq2SeqTrainer(
        model=model,
        args=training_args,
        train_dataset=train_dataset if training_args.do_train else None,
        eval_dataset=eval_dataset if training_args.do_eval else None,
        tokenizer=tokenizer,
        data_collator=data_collator,
        compute_metrics=compute_metrics if training_args.predict_with_generate else None,
    )

    eval_file_name = None

    # 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 tokenizer 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(train_dataset)
        )
        metrics["train_samples"] = min(max_train_samples, len(train_dataset))

        trainer.log_metrics("train", metrics)
        trainer.save_metrics("train", metrics)
        trainer.save_state()

    # Evaluation
    results = {}
    if training_args.do_eval:
        logger.info("*** Evaluate ***")

        if data_args.save_all_preds:
            eval_file_name = "all_eval_set.json"

        metrics = trainer.evaluate(
            max_length=data_args.val_max_target_length, num_beams=data_args.num_beams, metric_key_prefix="eval"
        )
        max_eval_samples = data_args.max_eval_samples if data_args.max_eval_samples is not None else len(eval_dataset)
        metrics["eval_samples"] = min(max_eval_samples, len(eval_dataset))

        trainer.log_metrics("eval", metrics)
        trainer.save_metrics("eval", metrics)

    if training_args.do_predict:
        logger.info("*** Predict ***")

        if data_args.save_all_preds:
            eval_file_name = "all_preds_set.json"

        predict_results = trainer.predict(
            predict_dataset,
            metric_key_prefix="predict",
            max_length=data_args.val_max_target_length,
            num_beams=data_args.num_beams,
        )
        metrics = predict_results.metrics
        max_predict_samples = (
            data_args.max_predict_samples if data_args.max_predict_samples is not None else len(predict_dataset)
        )
        metrics["predict_samples"] = min(max_predict_samples, len(predict_dataset))

        trainer.log_metrics("predict", metrics)
        trainer.save_metrics("predict", metrics)

        if trainer.is_world_process_zero():
            if training_args.predict_with_generate:
                predictions = tokenizer.batch_decode(
                    predict_results.predictions, skip_special_tokens=True, clean_up_tokenization_spaces=True
                )
                predictions = [pred.strip() for pred in predictions]
                output_prediction_file = os.path.join(training_args.output_dir, "generated_predictions.txt")
                with open(output_prediction_file, "w") as writer:
                    writer.write("\n".join(predictions))

    if training_args.push_to_hub:
        kwargs = {"finetuned_from": model_args.model_name_or_path, "tasks": "summarization"}
        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

        trainer.push_to_hub(**kwargs)

    return results


def _mp_fn(index):
    # For xla_spawn (TPUs)
    main()


if __name__ == "__main__":
    main()