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

import numpy as np
from datasets import load_dataset, load_metric, ClassLabel
from transformers import (
    AutoConfig,
    AutoTokenizer,
    HfArgumentParser,
    TrainingArguments,
    set_seed,
    DataCollatorForTokenClassification,
    AutoModelForTokenClassification,
    PreTrainedTokenizerFast,
)
from datamux_pretraining.models.multiplexing_legacy import (
    RobertaTokenClassificationMuxed,
)
from datamux_pretraining.models.finetune_trainer import FinetuneTrainer
from datamux_pretraining.models.multiplexing_pretraining_electra import (
    MuxedElectraForTokenClassification,
)
from datamux_pretraining.models.multiplexing_pretraining_bert import (
    MuxedBertForTokenClassification,
)
from transformers.trainer_utils import get_last_checkpoint

logger = logging.getLogger(__name__)

version_2_modelcls = {
    "electra": MuxedElectraForTokenClassification,
    "datamux_legacy": RobertaTokenClassificationMuxed,
    "bert": MuxedBertForTokenClassification,
}


@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.
    """

    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)."
        },
    )
    max_seq_length: int = field(
        default=128,
        metadata={
            "help": "The maximum total input sequence length after tokenization. Sequences longer "
            "than this will be truncated, sequences shorter will be padded."
        },
    )
    overwrite_cache: bool = field(
        default=False,
        metadata={"help": "Overwrite the cached preprocessed datasets or not."},
    )
    pad_to_max_length: bool = field(
        default=True,
        metadata={
            "help": "Whether to pad all samples to `max_seq_length`. "
            "If False, will pad the samples dynamically when batching to the maximum length in the batch."
        },
    )
    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."
        },
    )
    train_file: Optional[str] = field(
        default=None,
        metadata={"help": "A csv or a json file containing the training data."},
    )
    validation_file: Optional[str] = field(
        default=None,
        metadata={"help": "A csv or a json file containing the validation data."},
    )
    test_file: Optional[str] = field(
        default=None,
        metadata={"help": "A csv or a json file containing the test data."},
    )
    text_column_name: Optional[str] = field(
        default=None,
        metadata={
            "help": "The column name of text to input in the file (a csv or JSON file)."
        },
    )
    label_column_name: Optional[str] = field(
        default=None,
        metadata={
            "help": "The column name of label to input in the file (a csv or JSON file)."
        },
    )

    preprocessing_num_workers: Optional[int] = field(
        default=None,
        metadata={"help": "The number of processes to use for the preprocessing."},
    )
    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."
        },
    )

    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."
        self.task_name = self.task_name.lower()


@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"
        },
    )
    cache_dir: Optional[str] = field(
        default=None,
        metadata={
            "help": "Where do you want 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)."
        },
    )
    # multi instance arguments
    num_instances: Optional[int] = field(
        default=5,
        metadata={"help": "Number of instances i.e. N"},
    )
    muxing_variant: Optional[str] = field(
        default="gaussian_hadamard",
        metadata={
            "help": "muxing variant; choose from gaussian_hadamard or random_ortho or binary_hadamard"
        },
    )
    demuxing_variant: Optional[str] = field(
        default="index",
        metadata={"help": "demuxing variant, choose from  'index' or 'mlp'"},
    )
    should_mux: Optional[int] = field(
        default=1,
        metadata={"help": "whether to mux, turn off for non-multiplexed baselines"},
    )
    retrieval_percentage: Optional[float] = field(
        default=1.0,
        metadata={"help": "percentage of tokens to retrieve during inference"},
    )
    retrieval_pretraining: Optional[int] = field(
        default=0,
        metadata={"help": "Retrieval Pretraining"},
    )
    gaussian_hadamard_norm: Optional[float] = field(
        default=1,
        metadata={"help": "Norm of sentence embeddings if we use random projections"},
    )
    binary_hadamard_epsilon: Optional[float] = field(
        default=0,
        metadata={
            "help": "Percentage intersection among binary vectors, default is no intersection"
        },
    )
    retrieval_loss_coeff: Optional[float] = field(
        default=0.1,
        metadata={"help": "Coefficient for retrieval loss"},
    )
    task_loss_coeff: Optional[float] = field(
        default=0.9,
        metadata={"help": "Coefficient for task loss"},
    )
    learn_muxing: Optional[int] = field(
        default=0,
        metadata={"help": "whether instance embeddings are learnt or not"},
    )
    model_version: Optional[str] = field(
        default="bert",
        metadata={
            "help": "pretraining architecture, choose from  'roberta' or 'electra'"
        },
    )
    num_hidden_demux_layers: Optional[int] = field(
        default=3,
        metadata={"help": "number of hidden layers for demuxing"},
    )


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)
    )
    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()

    # 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 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)

    # 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.
        raw_datasets = load_dataset(
            data_args.dataset_name,
            data_args.dataset_config_name,
            cache_dir=model_args.cache_dir,
        )
    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]
        raw_datasets = load_dataset(
            extension, data_files=data_files, cache_dir=model_args.cache_dir
        )
    # 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 = raw_datasets["train"].column_names
        features = raw_datasets["train"].features
    else:
        column_names = raw_datasets["validation"].column_names
        features = raw_datasets["validation"].features

    if data_args.text_column_name is not None:
        text_column_name = data_args.text_column_name
    elif "tokens" in column_names:
        text_column_name = "tokens"
    else:
        text_column_name = column_names[0]

    if data_args.label_column_name is not None:
        label_column_name = data_args.label_column_name
    elif f"{data_args.task_name}_tags" in column_names:
        label_column_name = f"{data_args.task_name}_tags"
    else:
        label_column_name = 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(raw_datasets["train"][label_column_name])
        label_to_id = {l: i for i, l in enumerate(label_list)}
    num_labels = len(label_list)

    # Map that sends B-Xxx label to its I-Xxx counterpart
    b_to_i_label = []
    for idx, label in enumerate(label_list):
        if label.startswith("B-") and label.replace("B-", "I-") in label_list:
            b_to_i_label.append(label_list.index(label.replace("B-", "I-")))
        else:
            b_to_i_label.append(idx)

    # 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,
        label2id=label_to_id,
        id2label={i: l for l, i in label_to_id.items()},
        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,
    )

    config.num_instances = model_args.num_instances
    config.muxing_variant = model_args.muxing_variant
    config.demuxing_variant = model_args.demuxing_variant
    config.retrieval_percentage = model_args.retrieval_percentage
    config.gaussian_hadamard_norm = model_args.gaussian_hadamard_norm
    config.binary_hadamard_epsilon = model_args.binary_hadamard_epsilon
    config.retrieval_loss_coeff = model_args.retrieval_loss_coeff
    config.task_loss_coeff = model_args.task_loss_coeff
    config.learn_muxing = model_args.learn_muxing
    config.num_hidden_demux_layers = model_args.num_hidden_demux_layers
    tokenizer_name_or_path = (
        model_args.tokenizer_name
        if model_args.tokenizer_name
        else model_args.model_name_or_path
    )

    if config.model_type in {"gpt2", "roberta"}:
        tokenizer = AutoTokenizer.from_pretrained(
            tokenizer_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,
        )
    else:
        tokenizer = AutoTokenizer.from_pretrained(
            tokenizer_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,
        )

    model_path_supplied = model_args.model_name_or_path is not None
    if model_args.should_mux:
        model_cls = version_2_modelcls[model_args.model_version]
        if model_path_supplied:
            model = model_cls.from_pretrained(
                model_args.model_name_or_path,
                config=config,
            )
        else:
            model = model_cls(config=config)
    else:
        # non-multiplexed baseline
        if model_path_supplied:
            model = AutoModelForTokenClassification.from_pretrained(
                model_args.model_name_or_path,
                config=config,
            )
        else:
            model = AutoModelForTokenClassification.from_config(config)

    # 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#supported-frameworks 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,
            max_length=data_args.max_seq_length,
            # 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:
                    label_ids.append(label_to_id[label[word_idx]])
                # 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:
                    if data_args.label_all_tokens:
                        label_ids.append(b_to_i_label[label_to_id[label[word_idx]]])
                    else:
                        label_ids.append(-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 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))

        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 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))
        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 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(
                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"],
            }

    # Initialize our Trainer
    trainer = FinetuneTrainer(
        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,
    )

    # Training
    if training_args.do_train:
        logger.info("*** Train ***")
        checkpoint = None
        if last_checkpoint is not None:
            checkpoint = last_checkpoint

        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 ***")
        output_dir = training_args.output_dir
        last_checkpoint = get_last_checkpoint(output_dir)
        logger.info(f"Loading model from {last_checkpoint}")
        checkpoint = None
        if last_checkpoint is not None:
            checkpoint = last_checkpoint
        all_metrics = {}
        for seed in range(1, 6):
            set_seed(seed)
            metrics = trainer.evaluate(
                eval_dataset=eval_dataset, resume_from_checkpoint=checkpoint
            )
            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))
            for key, value in metrics.items():
                all_metrics[f"{key}_{seed}"] = value

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

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

        predictions, labels, metrics = trainer.predict(
            predict_dataset, metric_key_prefix="predict"
        )
        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("predict", metrics)
        trainer.save_metrics("predict", metrics)

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

    kwargs = {
        "finetuned_from": model_args.model_name_or_path,
        "tasks": "token-classification",
    }
    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


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


if __name__ == "__main__":
    main()