autogen/flaml/nlp/utils.py

import argparse
from dataclasses import dataclass, field
from ..data import SEQCLASSIFICATION, SEQREGRESSION


def _is_nlp_task(task):
    if task in [SEQCLASSIFICATION, SEQREGRESSION]:
        return True
    else:
        return False


global tokenized_column_names


def tokenize_text(X, task, custom_hpo_task):
    from ..data import SEQCLASSIFICATION

    if task in (SEQCLASSIFICATION, SEQREGRESSION):
        return tokenize_text_seqclassification(X, custom_hpo_task)


def tokenize_text_seqclassification(X, custom_hpo_args):
    from transformers import AutoTokenizer
    import pandas

    global tokenized_column_names

    this_tokenizer = AutoTokenizer.from_pretrained(
        custom_hpo_args.model_path, use_fast=True
    )
    d = X.apply(
        lambda x: tokenize_glue(x, this_tokenizer, custom_hpo_args),
        axis=1,
        result_type="expand",
    )
    X_tokenized = pandas.DataFrame(columns=tokenized_column_names)
    X_tokenized[tokenized_column_names] = d
    return X_tokenized


def tokenize_glue(this_row, this_tokenizer, custom_hpo_args):
    global tokenized_column_names
    assert (
        "max_seq_length" in custom_hpo_args.__dict__
    ), "max_seq_length must be provided for glue"

    tokenized_example = this_tokenizer(
        *tuple(this_row),
        padding="max_length",
        max_length=custom_hpo_args.max_seq_length,
        truncation=True,
    )
    tokenized_column_names = sorted(tokenized_example.keys())
    return [tokenized_example[x] for x in tokenized_column_names]


def separate_config(config):
    from transformers import TrainingArguments

    training_args_config = {}
    per_model_config = {}

    for key, val in config.items():
        if key in TrainingArguments.__dict__:
            training_args_config[key] = val
        else:
            per_model_config[key] = val

    return training_args_config, per_model_config


def get_num_labels(task, y_train):
    if task == SEQREGRESSION:
        return 1
    elif task == SEQCLASSIFICATION:
        return len(set(y_train))


def load_model(checkpoint_path, task, num_labels, per_model_config=None):
    from transformers import AutoConfig
    from .huggingface.switch_head_auto import (
        AutoSeqClassificationHead,
        MODEL_CLASSIFICATION_HEAD_MAPPING,
    )

    this_model_type = AutoConfig.from_pretrained(checkpoint_path).model_type
    this_vocab_size = AutoConfig.from_pretrained(checkpoint_path).vocab_size

    def get_this_model():
        from transformers import AutoModelForSequenceClassification

        return AutoModelForSequenceClassification.from_pretrained(
            checkpoint_path, config=model_config
        )

    def is_pretrained_model_in_classification_head_list(model_type):
        return model_type in MODEL_CLASSIFICATION_HEAD_MAPPING

    def _set_model_config(checkpoint_path):
        if per_model_config and len(per_model_config) > 0:
            model_config = AutoConfig.from_pretrained(
                checkpoint_path,
                num_labels=model_config_num_labels,
                **per_model_config,
            )
        else:
            model_config = AutoConfig.from_pretrained(
                checkpoint_path, num_labels=model_config_num_labels
            )
        return model_config

    if task == SEQCLASSIFICATION:
        num_labels_old = AutoConfig.from_pretrained(checkpoint_path).num_labels
        if is_pretrained_model_in_classification_head_list(this_model_type):
            model_config_num_labels = num_labels_old
        else:
            model_config_num_labels = num_labels
        model_config = _set_model_config(checkpoint_path)

        if is_pretrained_model_in_classification_head_list(this_model_type):
            if num_labels != num_labels_old:
                this_model = get_this_model()
                model_config.num_labels = num_labels
                this_model.num_labels = num_labels
                this_model.classifier = (
                    AutoSeqClassificationHead.from_model_type_and_config(
                        this_model_type, model_config
                    )
                )
            else:
                this_model = get_this_model()
        else:
            this_model = get_this_model()
        this_model.resize_token_embeddings(this_vocab_size)
        return this_model
    elif task == SEQREGRESSION:
        model_config_num_labels = 1
        model_config = _set_model_config(checkpoint_path)
        this_model = get_this_model()
        return this_model


def compute_checkpoint_freq(
    train_data_size,
    custom_hpo_args,
    num_train_epochs,
    batch_size,
):
    ckpt_step_freq = (
        int(
            min(num_train_epochs, 1)
            * train_data_size
            / batch_size
            / custom_hpo_args.ckpt_per_epoch
        )
        + 1
    )
    return ckpt_step_freq


@dataclass
class HPOArgs:
    """The HPO setting

    Args:
        output_dir (:obj:`str`):
            data root directory for outputing the log, etc.
        model_path (:obj:`str`, `optional`, defaults to :obj:`facebook/muppet-roberta-base`):
            A string, the path of the language model file, either a path from huggingface
            model card huggingface.co/models, or a local path for the model
        fp16 (:obj:`bool`, `optional`, defaults to :obj:`False`):
            A bool, whether to use FP16
        max_seq_length (:obj:`int`, `optional`, defaults to :obj:`128`):
            An integer, the max length of the sequence
        ckpt_per_epoch (:obj:`int`, `optional`, defaults to :obj:`1`):
            An integer, the number of checkpoints per epoch

    """

    output_dir: str = field(
        default="data/output/", metadata={"help": "data dir", "required": True}
    )

    model_path: str = field(
        default="facebook/muppet-roberta-base",
        metadata={"help": "model path model for HPO"},
    )

    fp16: bool = field(default=True, metadata={"help": "whether to use the FP16 mode"})

    max_seq_length: int = field(default=128, metadata={"help": "max seq length"})

    ckpt_per_epoch: int = field(default=1, metadata={"help": "checkpoint per epoch"})

    @staticmethod
    def load_args():
        from dataclasses import fields

        arg_parser = argparse.ArgumentParser()
        for each_field in fields(HPOArgs):
            print(each_field)
            arg_parser.add_argument(
                "--" + each_field.name,
                type=each_field.type,
                help=each_field.metadata["help"],
                required=each_field.metadata["required"]
                if "required" in each_field.metadata
                else False,
                choices=each_field.metadata["choices"]
                if "choices" in each_field.metadata
                else None,
                default=each_field.default,
            )
        console_args, unknown = arg_parser.parse_known_args()
        return console_args