model_utils.py

# coding=utf-8
# coding=utf-8
# Copyright 2023 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.

from transformers import AutoTokenizer, PreTrainedTokenizer

from .configs import DataArguments, ModelArguments
from .data import DEFAULT_CHAT_TEMPLATE


def get_tokenizer(model_args: ModelArguments, data_args: DataArguments) -> PreTrainedTokenizer:
    """Get the tokenizer for the model."""
    tokenizer = AutoTokenizer.from_pretrained(
        model_args.model_name_or_path,
        revision=model_args.model_revision,
    )
    if tokenizer.pad_token_id is None:
        tokenizer.pad_token_id = tokenizer.eos_token_id

    if data_args.truncation_side is not None:
        tokenizer.truncation_side = data_args.truncation_side

    # Set reasonable default for models without max length
    if tokenizer.model_max_length > 100_000:
        tokenizer.model_max_length = 2048

    if data_args.chat_template is not None:
        tokenizer.chat_template = data_args.chat_template
    elif tokenizer.chat_template is None:
        tokenizer.chat_template = DEFAULT_CHAT_TEMPLATE

    return tokenizer