llm2vec/experiments/run_word_task.py

"""
The script is adapted from https://huggingface.co/docs/transformers/en/tasks/token_classification
"""

import logging
import os
import sys
import warnings
from dataclasses import dataclass, field
import numpy as np
from typing import List, Optional, Tuple, Union

import datasets
import evaluate
from datasets import load_dataset

import torch
from torch import nn
from torch.nn import CrossEntropyLoss
import transformers
from transformers import (
    PreTrainedModel,
    MODEL_FOR_MASKED_LM_MAPPING,
    AutoConfig,
    AutoTokenizer,
    HfArgumentParser,
    Trainer,
    TrainingArguments,
    TrainerCallback,
    set_seed,
    AutoModelForTokenClassification,
    DataCollatorForTokenClassification,
)

from transformers.modeling_outputs import TokenClassifierOutput
from transformers.utils import send_example_telemetry
from transformers.utils.versions import require_version

from llm2vec import LLM2Vec

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


class ModelForWordTask(PreTrainedModel):
    def __init__(self, config, model, merge_subwords=False, **model_args):
        PreTrainedModel.__init__(self, config)
        self.model = model
        self.merge_subwords = merge_subwords

        if (
            hasattr(config, "classifier_dropout")
            and config.classifier_dropout is not None
        ):
            classifier_dropout = config.classifier_dropout
        elif hasattr(config, "hidden_dropout") and config.hidden_dropout is not None:
            classifier_dropout = config.hidden_dropout
        else:
            classifier_dropout = 0.1

        self.dropout = nn.Dropout(classifier_dropout)
        self.num_labels = config.num_labels
        self.classifier = nn.Linear(config.hidden_size, config.num_labels).to(
            model_args.get("torch_dtype")
        )

        # Initialize weights and apply final processing
        self.post_init()

    def _merge_subwords(self, hidden_states, token_type_ids, attention_mask):
        new_hidden_states = hidden_states.clone()
        for b in range(hidden_states.shape[0]):
            for w in torch.arange(0, token_type_ids[b].max() + 1):
                words_w = (token_type_ids[b] == w) * (attention_mask[b] > 0)
                new_hidden_states[b][words_w] = torch.mean(
                    hidden_states[b][words_w], dim=0
                ).repeat(sum(words_w), 1)
        return new_hidden_states

    def forward(
        self,
        input_ids: torch.LongTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple, TokenClassifierOutput]:
        output_attentions = (
            output_attentions
            if output_attentions is not None
            else self.config.output_attentions
        )
        output_hidden_states = (
            output_hidden_states
            if output_hidden_states is not None
            else self.config.output_hidden_states
        )

        return_dict = (
            return_dict if return_dict is not None else self.config.use_return_dict
        )

        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
        outputs = self.model(
            input_ids=input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        hidden_states = outputs[0]

        if self.merge_subwords:
            hidden_states = self._merge_subwords(
                hidden_states, token_type_ids, attention_mask
            )

        hidden_states = self.dropout(hidden_states)
        logits = self.classifier(hidden_states)

        loss = None
        if labels is not None:
            labels = labels.to(logits.device)
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

        if not return_dict:
            output = (logits,) + outputs.hidden_states
            return ((loss,) + output) if loss is not None else output

        return TokenClassifierOutput(
            loss=loss,
            logits=logits,
            hidden_states=hidden_states,
            attentions=outputs.attentions,
        )


logger = logging.getLogger(__name__)
MODEL_CONFIG_CLASSES = list(MODEL_FOR_MASKED_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
LABELS = {
    "conll2003": {
        "pos_tags": {
            '"': 0,
            "''": 1,
            "#": 2,
            "$": 3,
            "(": 4,
            ")": 5,
            ",": 6,
            ".": 7,
            ":": 8,
            "``": 9,
            "CC": 10,
            "CD": 11,
            "DT": 12,
            "EX": 13,
            "FW": 14,
            "IN": 15,
            "JJ": 16,
            "JJR": 17,
            "JJS": 18,
            "LS": 19,
            "MD": 20,
            "NN": 21,
            "NNP": 22,
            "NNPS": 23,
            "NNS": 24,
            "NN|SYM": 25,
            "PDT": 26,
            "POS": 27,
            "PRP": 28,
            "PRP$": 29,
            "RB": 30,
            "RBR": 31,
            "RBS": 32,
            "RP": 33,
            "SYM": 34,
            "TO": 35,
            "UH": 36,
            "VB": 37,
            "VBD": 38,
            "VBG": 39,
            "VBN": 40,
            "VBP": 41,
            "VBZ": 42,
            "WDT": 43,
            "WP": 44,
            "WP$": 45,
            "WRB": 46,
        },
        "chunk_tags": {
            "O": 0,
            "B-ADJP": 1,
            "I-ADJP": 2,
            "B-ADVP": 3,
            "I-ADVP": 4,
            "B-CONJP": 5,
            "I-CONJP": 6,
            "B-INTJ": 7,
            "I-INTJ": 8,
            "B-LST": 9,
            "I-LST": 10,
            "B-NP": 11,
            "I-NP": 12,
            "B-PP": 13,
            "I-PP": 14,
            "B-PRT": 15,
            "I-PRT": 16,
            "B-SBAR": 17,
            "I-SBAR": 18,
            "B-UCP": 19,
            "I-UCP": 20,
            "B-VP": 21,
            "I-VP": 22,
        },
        "ner_tags": {
            "O": 0,
            "B-PER": 1,
            "I-PER": 2,
            "B-ORG": 3,
            "I-ORG": 4,
            "B-LOC": 5,
            "I-LOC": 6,
            "B-MISC": 7,
            "I-MISC": 8,
        },
    }
}


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

    model_name_or_path: Optional[str] = field(
        default=None,
        metadata={},
    )
    config_overrides: Optional[str] = field(
        default=None,
        metadata={
            "help": (
                "Override some existing default config settings when a model is trained from scratch. Example: "
                "n_embd=10,resid_pdrop=0.2,scale_attn_weights=false,summary_type=cls_index"
            )
        },
    )
    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)."
        },
    )
    token: str = field(
        default=None,
        metadata={
            "help": (
                "The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
                "generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
            )
        },
    )
    use_auth_token: bool = field(
        default=None,
        metadata={
            "help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead."
        },
    )
    trust_remote_code: bool = field(
        default=False,
        metadata={
            "help": (
                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                "execute code present on the Hub on your local machine."
            )
        },
    )
    low_cpu_mem_usage: bool = field(
        default=False,
        metadata={
            "help": (
                "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded. "
                "set True will benefit LLM loading time and RAM consumption."
            )
        },
    )
    torch_dtype: Optional[str] = field(
        default=None,
        metadata={
            "help": (
                "Override the default `torch.dtype` and load the model under this dtype. If `auto` is passed, the "
                "dtype will be automatically derived from the model's weights."
            ),
            "choices": ["auto", "bfloat16", "float16", "float32"],
        },
    )
    attn_implementation: Optional[str] = field(
        default="sdpa",
        metadata={
            "help": ("The attention implementation to use in the model."),
            "choices": ["eager", "sdpa", "flash_attention_2"],
        },
    )
    classifier_dropout: Optional[float] = field(
        default=0.1, metadata={"help": "The dropout rate for models"}
    )
    peft_addr: Optional[str] = field(
        default=None, metadata={"help": "addr of lora adapter weights"}
    )
    model_class: str = field(
        default="custom",
        metadata={
            "help": "One of the items 'custom' or 'auto'. 'custom' for LLM2Vec models and 'auto' for pretrained encoders such as BERT.",
            "choices": ["custom", "auto"],
        },
    )
    merge_subwords: bool = field(
        default=True,
        metadata={"help": "Whether the representations of the subtokens get averaged."},
    )
    bidirectional: bool = field(
        default=True, metadata={"help": "Whether to use bidirectional attention."}
    )

    def __post_init__(self):
        if self.config_overrides is not None and (
            self.config_name is not None or self.model_name_or_path is not None
        ):
            raise ValueError(
                "--config_overrides can't be used in combination with --config_name or --model_name_or_path"
            )


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

    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 text file)."}
    )
    validation_file: Optional[str] = field(
        default=None,
        metadata={
            "help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."
        },
    )
    overwrite_cache: bool = field(
        default=True,
        metadata={"help": "Overwrite the cached training and evaluation sets"},
    )
    validation_split_percentage: Optional[int] = field(
        default=5,
        metadata={
            "help": "The percentage of the train set used as validation set in case there's no validation split"
        },
    )
    max_seq_length: Optional[int] = field(
        default=None,
        metadata={
            "help": (
                "The maximum total input sequence length after tokenization. Sequences longer "
                "than this will be truncated."
            )
        },
    )
    preprocessing_num_workers: Optional[int] = field(
        default=None,
        metadata={"help": "The number of processes to use for the preprocessing."},
    )
    mlm_probability: float = field(
        default=0.15,
        metadata={"help": "Ratio of tokens to mask for masked language modeling loss"},
    )
    line_by_line: bool = field(
        default=False,
        metadata={
            "help": "Whether distinct lines of text in the dataset are to be handled as distinct sequences."
        },
    )
    pad_to_max_length: bool = field(
        default=False,
        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."
            )
        },
    )
    streaming: bool = field(default=False, metadata={"help": "Enable streaming mode"})

    def __post_init__(self):
        if self.streaming:
            require_version(
                "datasets>=2.0.0", "The streaming feature requires `datasets>=2.0.0`"
            )

        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]
                if extension not in ["csv", "json", "txt"]:
                    raise ValueError(
                        "`train_file` should be a csv, a json or a txt file."
                    )
            if self.validation_file is not None:
                extension = self.validation_file.split(".")[-1]
                if extension not in ["csv", "json", "txt"]:
                    raise ValueError(
                        "`validation_file` should be a csv, a json or a txt file."
                    )


# add more arguments
@dataclass
class CustomArguments:
    """
    Custom arguments for the script
    """

    stop_after_n_steps: int = field(
        default=10000, metadata={"help": "Stop training after n steps"}
    )
    data_collator_type: str = field(
        default="custom",
        metadata={
            "help": "The type of data collator. Options: custom, default, custom_no_random"
        },
    )
    task: Optional[str] = field(
        default="pos_tags",
        metadata={
            "help": "One of the 'pos_tags', 'chunk_tags', and 'ner_tags' choices",
            "choices": ["pos_tags", "ner_tags", "chunk_tags"],
        },
    )
    retroactive_labels: str = field(
        default="next_token",
        metadata={
            "help": "Whether the tokens representations are used to predict the next token's labels. Options: same_token, next_word, next_token.",
            "choices": ["next_token", "same_token"],
        },
    )


class StopTrainingCallback(TrainerCallback):
    def __init__(self, stop_after_n_steps: int):
        self.stop_after_n_steps = stop_after_n_steps

    def on_step_end(self, args, state, control, **kwargs):
        if state.global_step >= self.stop_after_n_steps:
            control.should_training_stop = True


class WordTaskTrainer(Trainer):
    def _save(self, output_dir: Optional[str] = None, state_dict=None):
        # If we are executing this function, we are the process zero, so we don't check for that.
        output_dir = output_dir if output_dir is not None else self.args.output_dir
        os.makedirs(output_dir, exist_ok=True)
        logger.info(f"Saving model checkpoint to {output_dir}")

        torch.save(self.model.classifier, os.path.join(output_dir, "classifier.pt"))
        self.tokenizer.save_pretrained(output_dir)

        # Good practice: save your training arguments together with the trained model
        torch.save(self.args, os.path.join(output_dir, "training_args.bin"))


def main():
    parser = HfArgumentParser(
        (ModelArguments, DataTrainingArguments, TrainingArguments, CustomArguments)
    )
    # model_args, data_args, training_args, custom_args = parser.parse_args_into_dataclasses()
    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, custom_args = parser.parse_json_file(
            json_file=os.path.abspath(sys.argv[1])
        )
    else:
        (
            model_args,
            data_args,
            training_args,
            custom_args,
        ) = parser.parse_args_into_dataclasses()

    if training_args.gradient_checkpointing:
        training_args.gradient_checkpointing_kwargs = {"use_reentrant": False}

    if model_args.use_auth_token is not None:
        warnings.warn(
            "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead.",
            FutureWarning,
        )
        if model_args.token is not None:
            raise ValueError(
                "`token` and `use_auth_token` are both specified. Please set only the argument `token`."
            )
        model_args.token = model_args.use_auth_token

    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
    # information sent is the one passed as arguments along with your Python/PyTorch versions.
    send_example_telemetry("run_word_task", model_args, data_args)

    # 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)],
    )

    if training_args.should_log:
        # The default of training_args.log_level is passive, so we set log level at info here to have that default.
        transformers.utils.logging.set_verbosity_info()

    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: {training_args.parallel_mode.value == 'distributed'}, 16-bits training: {training_args.fp16}"
    )
    # Set the verbosity to info of the Transformers logger (on main process only):
    logger.info(f"Training/evaluation parameters {training_args}")

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

    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,
            token=model_args.token,
            streaming=data_args.streaming,
        )
        if "validation" not in raw_datasets.keys():
            raw_datasets["validation"] = load_dataset(
                data_args.dataset_name,
                data_args.dataset_config_name,
                split=f"train[:{data_args.validation_split_percentage}%]",
                cache_dir=model_args.cache_dir,
                token=model_args.token,
                streaming=data_args.streaming,
            )
            raw_datasets["train"] = load_dataset(
                data_args.dataset_name,
                data_args.dataset_config_name,
                split=f"train[{data_args.validation_split_percentage}%:]",
                cache_dir=model_args.cache_dir,
                token=model_args.token,
                streaming=data_args.streaming,
            )
    else:
        data_files = {}
        if data_args.train_file is not None:
            data_files["train"] = data_args.train_file
            extension = data_args.train_file.split(".")[-1]
        if data_args.validation_file is not None:
            data_files["validation"] = data_args.validation_file
            extension = data_args.validation_file.split(".")[-1]
        if extension == "txt":
            extension = "text"
        raw_datasets = load_dataset(
            extension,
            data_files=data_files,
            cache_dir=model_args.cache_dir,
            token=model_args.token,
        )

        # If no validation data is there, validation_split_percentage will be used to divide the dataset.
        if "validation" not in raw_datasets.keys():
            raw_datasets["validation"] = load_dataset(
                extension,
                data_files=data_files,
                split=f"train[:{data_args.validation_split_percentage}%]",
                cache_dir=model_args.cache_dir,
                token=model_args.token,
            )
            raw_datasets["train"] = load_dataset(
                extension,
                data_files=data_files,
                split=f"train[{data_args.validation_split_percentage}%:]",
                cache_dir=model_args.cache_dir,
                token=model_args.token,
            )

    assert (
        data_args.dataset_name in LABELS
        and custom_args.task in LABELS[data_args.dataset_name]
    ), f"LABELS[{data_args.dataset_name}][{custom_args.task}] is not defined."

    config_kwargs = {
        "num_labels": len(LABELS[data_args.dataset_name][custom_args.task]),
        "id2label": {
            i: lab
            for (lab, i) in LABELS[data_args.dataset_name][custom_args.task].items()
        },
        "label2id": LABELS[data_args.dataset_name][custom_args.task],
        "classifier_dropout": model_args.classifier_dropout,
    }

    tokenizer_kwargs = {
        "cache_dir": model_args.cache_dir,
        "use_fast": model_args.use_fast_tokenizer,
        "revision": model_args.model_revision,
        "token": model_args.token,
        "trust_remote_code": model_args.trust_remote_code,
    }
    if model_args.tokenizer_name:
        if "gpt" in model_args.tokenizer_name:
            tokenizer_kwargs["add_prefix_space"] = True
        tokenizer = AutoTokenizer.from_pretrained(
            model_args.tokenizer_name, **tokenizer_kwargs
        )
    elif model_args.model_name_or_path:
        if "gpt" in model_args.model_name_or_path:
            tokenizer_kwargs["add_prefix_space"] = True
        tokenizer = AutoTokenizer.from_pretrained(
            model_args.model_name_or_path, **tokenizer_kwargs
        )
    else:
        raise ValueError(
            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
            "You can do it from another script, save it, and load it from here, using --tokenizer_name."
        )

    if tokenizer.pad_token is None:
        tokenizer.pad_token = tokenizer.eos_token
    if model_args.model_class == "custom":
        tokenizer.model_input_names.append("token_type_ids")
    if model_args.model_class == "auto":
        assert not model_args.merge_subwords

    if model_args.model_class == "custom":
        if model_args.config_name:
            config = AutoConfig.from_pretrained(model_args.config_name, **config_kwargs)
        elif model_args.model_name_or_path:
            config = AutoConfig.from_pretrained(
                model_args.model_name_or_path, **config_kwargs
            )
        else:
            raise ValueError("Invalid config loading")

        for k, v in config_kwargs.items():
            config.__setattr__(k, v)

        torch_dtype = (
            model_args.torch_dtype
            if model_args.torch_dtype in ["auto", None]
            else getattr(torch, model_args.torch_dtype)
        )
        l2v = LLM2Vec.from_pretrained(
            base_model_name_or_path=model_args.model_name_or_path,
            enable_bidirectional=model_args.bidirectional,
            peft_model_name_or_path=model_args.peft_addr,
            merge_peft=False,
            torch_dtype=torch_dtype,
            attn_implementation=model_args.attn_implementation,
        )

        model = ModelForWordTask(
            model=l2v.model,
            merge_subwords=model_args.merge_subwords,
            config=config,
            torch_dtype=torch_dtype,
        )

        MyTrainer = WordTaskTrainer

    elif model_args.model_class == "auto":
        model = AutoModelForTokenClassification.from_pretrained(
            model_args.model_name_or_path,
            num_labels=config_kwargs["num_labels"],
            id2label=config_kwargs["id2label"],
            label2id=config_kwargs["label2id"],
        )
        MyTrainer = Trainer

    else:
        raise ValueError(
            f"{model_args.model_class} is not implemented. Only 'auto' and 'custom' model_class options are valid."
        )

    # only train classifier
    for n, p in list(model.named_parameters()):
        if "classifier" in n:
            p.requires_grad = True
        else:
            p.requires_grad = False

    if data_args.max_seq_length is None:
        max_seq_length = tokenizer.model_max_length
        if max_seq_length > 1024:
            logger.warning(
                "The chosen tokenizer supports a `model_max_length` that is longer than the default `block_size` value"
                " of 1024. If you would like to use a longer `block_size` up to `tokenizer.model_max_length` you can"
                " override this default with `--block_size xxx`."
            )
            max_seq_length = 1024
    else:
        if data_args.max_seq_length > tokenizer.model_max_length:
            logger.warning(
                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
                f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
            )
        max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

    def tokenize_and_align_labels(examples):
        task = custom_args.task
        padding = "max_length" if data_args.pad_to_max_length else False
        tokenized_inputs = tokenizer(
            examples["tokens"],
            truncation=True,
            is_split_into_words=True,
            padding=padding,
            max_length=max_seq_length,
        )

        labels = []
        words = []
        for i, label in enumerate(examples[task]):
            if custom_args.retroactive_labels in ["same_token"]:
                word_ids = tokenized_inputs.word_ids(batch_index=i)
                previous_word_idx = None
                label_ids = []
                for word_idx in word_ids:
                    if word_idx is None:
                        label_ids.append(-100)
                    elif word_idx != previous_word_idx:
                        label_ids.append(label[word_idx])
                    else:
                        label_ids.append(-100)
                    previous_word_idx = word_idx
                labels.append(label_ids)
                word_ids = [-1 if w is None else w for w in word_ids]
                words.append(word_ids)
            elif custom_args.retroactive_labels == "next_token":
                word_ids = tokenized_inputs.word_ids(batch_index=i)
                previous_word_idx = None
                label_ids = []
                for word_idx in word_ids:
                    if word_idx is None:
                        label_ids.append(-100)
                    elif word_idx != previous_word_idx:
                        label_ids.append(label[word_idx])
                    else:
                        label_ids.append(-100)
                    previous_word_idx = word_idx
                label_ids.append(-100)
                labels.append(label_ids[1:])
                word_ids = word_ids[1:] + [None]
                word_ids = [-1 if w is None else w for w in word_ids]
                words.append(word_ids)
            else:
                raise ValueError(
                    f"retroactive_labels {custom_args.retroactive_labels} is not implemented."
                )

        tokenized_inputs["labels"] = labels
        if model_args.model_class == "custom":
            tokenized_inputs["token_type_ids"] = words
        return tokenized_inputs

    tokenized_dataset = raw_datasets.map(
        tokenize_and_align_labels,
        batched=True,
        remove_columns=list(LABELS[data_args.dataset_name].keys()) + ["tokens", "id"],
        load_from_cache_file=not data_args.overwrite_cache,
    )
    data_collator = DataCollatorForTokenClassification(tokenizer=tokenizer)
    seqeval = evaluate.load("seqeval")

    def compute_metrics(p):
        predictions, labels = p
        predictions = predictions[0]
        predictions = np.argmax(predictions, axis=2)

        true_predictions = [
            [
                config_kwargs["id2label"][p]
                for (p, l) in zip(prediction, label)
                if l != -100
            ]
            for prediction, label in zip(predictions, labels)
        ]
        true_labels = [
            [
                config_kwargs["id2label"][l]
                for (p, l) in zip(prediction, label)
                if l != -100
            ]
            for prediction, label in zip(predictions, labels)
        ]

        results = seqeval.compute(predictions=true_predictions, references=true_labels)
        return {
            "precision": results["overall_precision"],
            "recall": results["overall_recall"],
            "f1": results["overall_f1"],
            "accuracy": results["overall_accuracy"],
        }

    trainer = MyTrainer(
        model=model,
        args=training_args,
        train_dataset=tokenized_dataset["train"],
        eval_dataset=tokenized_dataset["validation"],
        tokenizer=tokenizer,
        data_collator=data_collator,
        compute_metrics=compute_metrics,
    )
    trainer.add_callback(StopTrainingCallback(custom_args.stop_after_n_steps))

    trainer.train()


if __name__ == "__main__":
    main()