run_ner.py

#!/usr/bin/env python
# coding=utf-8
# 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.
"""
Fine-tuning the library models for token classification.
"""
# You can also adapt this script on your own token classification task and datasets. Pointers for this are left as
# comments.
import datetime
import json
import logging
import os
import sys
import time
from dataclasses import dataclass, field
from typing import Optional

import numpy as np
from datasets import ClassLabel, load_dataset, load_metric

import transformers
from transformers import (
    AutoConfig,
    AutoModelForTokenClassification,
    AutoTokenizer,
    DataCollatorForTokenClassification,
    HfArgumentParser,
    PreTrainedTokenizerFast,
    Trainer,
    TrainingArguments,
    set_seed,
)
from transformers.trainer_utils import get_last_checkpoint, is_main_process
from transformers.utils import check_min_version

os.environ["WANDB_DISABLED"] = "true"


# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
# check_min_version("4.5.0.dev0")
from hps_ner import hyperparameter_tune

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"}
    )
    cache_dir: Optional[str] = field(
        default=None,
        metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
    )
    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)."
        },
    )
    add_prefix_space: bool = field(
        default=False,
        metadata={
            "help": "Some tokenizers require an additional prefix space when handling pre-tokenized inputs."
        }
    )


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

    task_name: Optional[str] = field(default="ner", metadata={"help": "The name of the task (ner, pos...)."})
    dataset_name: Optional[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)."}
    )
    train_file: Optional[str] = field(
        default=None, metadata={"help": "The input training data file (a csv or JSON file)."}
    )
    validation_file: Optional[str] = field(
        default=None,
        metadata={"help": "An optional input evaluation data file to evaluate on (a csv or JSON file)."},
    )
    test_file: Optional[str] = field(
        default=None,
        metadata={"help": "An optional input test data file to predict on (a csv or JSON 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."},
    )
    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_val_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."
        },
    )
    max_test_samples: Optional[int] = field(
        default=None,
        metadata={
            "help": "For debugging purposes or quicker training, truncate the number of test examples to this "
            "value if set."
        },
    )
    label_all_tokens: bool = field(
        default=False,
        metadata={
            "help": "Whether to put the label for one word on all tokens of generated by that word or just on the "
            "one (in which case the other tokens will have a padding index)."
        },
    )
    return_entity_level_metrics: bool = field(
        default=False,
        metadata={"help": "Whether to return all the entity levels during evaluation or just the overall ones."},
    )
    tune: str = field(
        default=None, metadata={"help": "tune with ray tune: give name of the tuning experiment"}
    )
    tune_alg: str = field(
        default='BOHB', metadata={"help": "The HPO algorithm that ray tune will use, e.g. BOHB, PBT, or ASHA"}
    )
    tune_trials: int = field(
        default=5, metadata={"help": "The number of trials that will be explored in the HPO."}
    )
    tune_local_dir: str = field(
        default="~/ray_results/", metadata={"help": "The directory in which the ray tune experiment folder will be."}
    )
    eval_all: bool = field(
        default=False,
        metadata={"help": "Evaluate the model on all dev & test sets."}
    )

    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", "jsonl"], "`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", "jsonl"], "`validation_file` should be a csv or a json file."
        self.task_name = self.task_name.lower()


EVAL_LOGS_DIR = "./eval_logs"
EVAL_LOGS_FILE = ""
def args_wrapper_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))
    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()

    if data_args.eval_all:
        # TAG_FORMATS = ["original_tags", "really_simple_tags"]
        CASE_FORMATS = ["cased", "lower", "lower_mix", "lower_both", "ne_lower", "ne_lower_mix", "ne_lower_both"]
        base_data_path_dev = "./data/{}/{}/dev.jsonl"
        base_data_path_test = "./data/{}/{}/test.jsonl"
        from pathlib import Path
        Path(EVAL_LOGS_DIR).mkdir(parents=True, exist_ok=True)

        global EVAL_LOGS_FILE
        now = datetime.datetime.now()
        dt_string = now.strftime("%Y-%m-%d_%H:%M:%S")
        EVAL_LOGS_FILE = f"{EVAL_LOGS_DIR}/{model_args.model_name_or_path.split('/')[-1]}" + "_" + dt_string

        # for tag_format in TAG_FORMATS:
        tag_format = "original_tags" if "original_tags" in data_args.tune else "really_simple_tags"
        for case_format in CASE_FORMATS:
            dev_path = base_data_path_dev.format(tag_format, case_format)
            test_path = base_data_path_test.format(tag_format, case_format)
            if case_format == "cased":
                dev_path = dev_path.replace(case_format + "/", "")
                test_path = test_path.replace(case_format + "/", "")

            data_args.validation_file = dev_path
            data_args.test_file = test_path
            with open(EVAL_LOGS_FILE, 'a') as f:
                f.write(f"\n************\n{tag_format}/{case_format}")
            main(model_args, data_args, training_args)
        return

    main(model_args, data_args, training_args)


def write_res_to_file(metrics):
    assert EVAL_LOGS_FILE != ""
    with open(EVAL_LOGS_FILE, 'a') as f:
        metrics_str = json.dumps(metrics)
        f.write("\n" + metrics_str)


def main(model_args, data_args, training_args):
    # 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))
    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()
    """

    # 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."
            )

    # 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)],
    )
    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}"
    )
    # 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()
        transformers.utils.logging.enable_default_handler()
        transformers.utils.logging.enable_explicit_format()
    logger.info("Training/evaluation parameters %s", training_args)

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

    # Get the datasets: you can either provide your own CSV/JSON/TXT training and evaluation files (see below)
    # or just provide the name of one of the public datasets available on the hub at https://huggingface.co/datasets/
    # (the dataset will be downloaded automatically from the datasets Hub).
    #
    # For CSV/JSON files, this script will use the column called 'text' or the first column if no column called
    # 'text' is found. You can easily tweak this behavior (see below).
    #
    # In distributed training, the load_dataset function guarantee that only one local process can concurrently
    # download the dataset.
    if data_args.dataset_name is not None:
        # Downloading and loading a dataset from the hub.
        datasets = load_dataset(data_args.dataset_name, data_args.dataset_config_name)
    else:
        data_files = {}
        if data_args.train_file is not None:
            data_files["train"] = data_args.train_file
        if data_args.validation_file is not None:
            data_files["validation"] = data_args.validation_file
        if data_args.test_file is not None:
            data_files["test"] = data_args.test_file
        extension = data_args.train_file.split(".")[-1]
        if extension == "jsonl":
            extension = "json"
        datasets = load_dataset(extension, data_files=data_files)
    # See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at
    # https://huggingface.co/docs/datasets/loading_datasets.html.

    if training_args.do_train:
        column_names = datasets["train"].column_names
        features = datasets["train"].features
    else:
        column_names = datasets["validation"].column_names
        features = datasets["validation"].features
    text_column_name = "tokens" if "tokens" in column_names else column_names[0]
    label_column_name = (
        f"{data_args.task_name}_tags" if f"{data_args.task_name}_tags" in column_names else column_names[1]
    )

    # In the event the labels are not a `Sequence[ClassLabel]`, we will need to go through the dataset to get the
    # unique labels.
    def get_label_list(labels):
        unique_labels = set()
        for label in labels:
            unique_labels = unique_labels | set(label)
        label_list = list(unique_labels)
        label_list.sort()
        return label_list

    if isinstance(features[label_column_name].feature, ClassLabel):
        label_list = features[label_column_name].feature.names
        # No need to convert the labels since they are already ints.
        label_to_id = {i: i for i in range(len(label_list))}
    else:
        label_list = get_label_list(datasets["train"][label_column_name])
        label_to_id = {l: i for i, l in enumerate(label_list)}
    num_labels = len(label_list)
    logger.info(label_list)
    logger.info(num_labels)

    # Load pretrained model and tokenizer
    #
    # Distributed training:
    # The .from_pretrained methods guarantee that only one local process can concurrently
    # download model & vocab.
    config = AutoConfig.from_pretrained(
        model_args.config_name if model_args.config_name else model_args.model_name_or_path,
        num_labels=num_labels,
        finetuning_task=data_args.task_name,
        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=True,
        revision=model_args.model_revision,
        use_auth_token=True if model_args.use_auth_token else None,
        add_prefix_space=True if model_args.add_prefix_space else False,
        truncation=True,
        max_length=512,
        model_max_length=512,
    )
    """
    model = AutoModelForTokenClassification.from_pretrained(
        model_args.model_name_or_path,
        from_tf=bool(".ckpt" in 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,
    )
    """

    # Tokenizer check: this script requires a fast tokenizer.
    if not isinstance(tokenizer, PreTrainedTokenizerFast):
        raise ValueError(
            "This example script only works for models that have a fast tokenizer. Checkout the big table of models "
            "at https://huggingface.co/transformers/index.html#bigtable to find the model types that meet this "
            "requirement"
        )

    # Preprocessing the dataset
    # Padding strategy
    padding = "max_length" if data_args.pad_to_max_length else False

    # Tokenize all texts and align the labels with them.
    def tokenize_and_align_labels(examples):
        tokenized_inputs = tokenizer(
            examples[text_column_name],
            padding=padding,
            truncation=True,
            # We use this argument because the texts in our dataset are lists of words (with a label for each word).
            is_split_into_words=True,
        )
        labels = []
        for i, label in enumerate(examples[label_column_name]):
            word_ids = tokenized_inputs.word_ids(batch_index=i)
            previous_word_idx = None
            label_ids = []
            for word_idx in word_ids:
                # Special tokens have a word id that is None. We set the label to -100 so they are automatically
                # ignored in the loss function.
                if word_idx is None:
                    label_ids.append(-100)
                # We set the label for the first token of each word.
                elif word_idx != previous_word_idx:
                    try:
                        label_ids.append(label_to_id[label[word_idx]])
                    except KeyError:
                        logger.info("Did not see that label before. Will be ignored...")
                        logger.info(word_idx, label[word_idx])
                        label_ids.append(-100)
                # For the other tokens in a word, we set the label to either the current label or -100, depending on
                # the label_all_tokens flag.
                else:
                    label_ids.append(label_to_id[label[word_idx]] if data_args.label_all_tokens else -100)
                previous_word_idx = word_idx

            labels.append(label_ids)
        tokenized_inputs["labels"] = labels
        return tokenized_inputs

    if training_args.do_train:
        if "train" not in datasets:
            raise ValueError("--do_train requires a train dataset")
        train_dataset = datasets["train"]
        if data_args.max_train_samples is not None:
            train_dataset = train_dataset.select(range(data_args.max_train_samples))
        train_dataset = train_dataset.map(
            tokenize_and_align_labels,
            batched=True,
            num_proc=data_args.preprocessing_num_workers,
            load_from_cache_file=not data_args.overwrite_cache,
        )

    if training_args.do_eval:
        if "validation" not in datasets:
            raise ValueError("--do_eval requires a validation dataset")
        eval_dataset = datasets["validation"]
        if data_args.max_val_samples is not None:
            eval_dataset = eval_dataset.select(range(data_args.max_val_samples))
        eval_dataset = eval_dataset.map(
            tokenize_and_align_labels,
            batched=True,
            num_proc=data_args.preprocessing_num_workers,
            load_from_cache_file=not data_args.overwrite_cache,
        )

    if training_args.do_predict:
        if "test" not in datasets:
            raise ValueError("--do_predict requires a test dataset")
        test_dataset = datasets["test"]
        if data_args.max_test_samples is not None:
            test_dataset = test_dataset.select(range(data_args.max_test_samples))
        test_dataset = test_dataset.map(
            tokenize_and_align_labels,
            batched=True,
            num_proc=data_args.preprocessing_num_workers,
            load_from_cache_file=not data_args.overwrite_cache,
        )

    # Data collator
    data_collator = DataCollatorForTokenClassification(tokenizer, pad_to_multiple_of=8 if training_args.fp16 else None)

    # Metrics
    metric = load_metric("seqeval")

    def compute_metrics(p):
        predictions, labels = p
        predictions = np.argmax(predictions, axis=2)

        # Remove ignored index (special tokens)
        true_predictions = [
            [label_list[p] for (p, l) in zip(prediction, label) if l != -100]
            for prediction, label in zip(predictions, labels)
        ]
        true_labels = [
            [label_list[l] for (p, l) in zip(prediction, label) if l != -100]
            for prediction, label in zip(predictions, labels)
        ]

        results = metric.compute(predictions=true_predictions, references=true_labels)
        if data_args.return_entity_level_metrics:
            # Unpack nested dictionaries
            final_results = {}
            for key, value in results.items():
                if isinstance(value, dict):
                    for n, v in value.items():
                        final_results[f"{key}_{n}"] = v
                else:
                    final_results[key] = value
            return final_results
        else:
            return {
                "precision": results["overall_precision"],
                "recall": results["overall_recall"],
                "f1": results["overall_f1"],
                "accuracy": results["overall_accuracy"],
            }

    def model_init(trial=None):
        conf = config
        if trial is not None:
            conf.update({
                "hidden_dropout_prob": trial['hidden_dropout_prob'],
                "attention_probs_dropout_prob": trial['attention_probs_dropout_prob']
            })

        model = AutoModelForTokenClassification.from_pretrained(
            model_args.model_name_or_path,
            from_tf=bool(".ckpt" in model_args.model_name_or_path),
            config=conf,
            # cache_dir=model_args.cache_dir,
            revision=model_args.model_revision,
            use_auth_token=True if model_args.use_auth_token else None,
            # ignore_mismatched_sizes=True
        )

        return model
        # return AutoModelForTokenClassification.from_pretrained(
        # model_args.model_name_or_path, return_dict=True)

    # Uncomment to shard and use small subset of dataset for testing purposes
    # train_dataset = train_dataset.shard(index=1, num_shards=50)
    # eval_dataset = eval_dataset.shard(index=1, num_shards=30)

    # Initialize our Trainer
    trainer = Trainer(
        #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,
        model_init=model_init
    )

    # Training
    if training_args.do_train:
        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
###
        if data_args.tune:
            # See hps_ner.py
            hps_start_time = time.time()
            best_run = hyperparameter_tune(trainer, training_args, data_args)
            # Configure trainer object to use best hyperparameters found
            logger.info("***** Best Hyperparameters found *****")
            for n, v in best_run.hyperparameters.items():
                if n != 'max_steps':  # Use original max_steps
                    setattr(trainer.args, n, v)
                    logger.info(f"{n}: {v}")

            time_str_hps = datetime.timedelta(seconds=round(time.time() - hps_start_time))
            logger.info(f"Time Elapsed HPS (h:mm:ss): {time_str_hps}")
            logger.info("*" * 10)
            return

        train_result = trainer.train(resume_from_checkpoint=checkpoint)
        metrics = train_result.metrics
        trainer.save_model()  # Saves the tokenizer too for easy upload

        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
    if training_args.do_eval:
        logger.info("*** Evaluate ***")

        metrics = trainer.evaluate()
        if data_args.eval_all:
            write_res_to_file(metrics)

        max_val_samples = data_args.max_val_samples if data_args.max_val_samples is not None else len(eval_dataset)
        metrics["eval_samples"] = min(max_val_samples, len(eval_dataset))

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

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

        predictions, labels, metrics = trainer.predict(test_dataset)
        if data_args.eval_all:
            write_res_to_file(metrics)
        predictions = np.argmax(predictions, axis=2)

        # Remove ignored index (special tokens)
        true_predictions = [
            [label_list[p] for (p, l) in zip(prediction, label) if l != -100]
            for prediction, label in zip(predictions, labels)
        ]

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

        # Save predictions
        output_test_predictions_file = os.path.join(training_args.output_dir, "test_predictions.txt")
        if trainer.is_world_process_zero():
            with open(output_test_predictions_file, "w") as writer:
                for prediction in true_predictions:
                    writer.write(" ".join(prediction) + "\n")

        # Print evaluation dataset names
        logger.info("Model name or path:")
        logger.info(model_args.model_name_or_path)
        logger.info("Validation file:")
        logger.info(data_args.validation_file)
        logger.info("Test file:")
        logger.info(data_args.test_file)


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


def mainTimeWrapper():
    start_time = time.time()
    args_wrapper_main()
    time_str = datetime.timedelta(seconds=round(time.time() - start_time))
    logger.info(f"Time Elapsed Total (h:mm:ss): {time_str}")


if __name__ == "__main__":
    mainTimeWrapper()