# Copyright 2024 the LlamaFactory team.
#
# 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 types import MethodType
from typing import TYPE_CHECKING, Dict, List, Optional, Tuple, Union

import torch
from transformers import Trainer
from typing_extensions import override

from ...extras.packages import is_transformers_version_equal_to_4_46, is_transformers_version_greater_than
from ..callbacks import PissaConvertCallback, SaveProcessorCallback
from ..trainer_utils import create_custom_optimizer, create_custom_scheduler


if TYPE_CHECKING:
    from transformers import PreTrainedModel, ProcessorMixin

    from ...hparams import FinetuningArguments


class CustomTrainer(Trainer):
    r"""
    Inherits Trainer for custom optimizer.
    """

    def __init__(
        self, finetuning_args: "FinetuningArguments", processor: Optional["ProcessorMixin"], **kwargs
    ) -> None:
        if is_transformers_version_greater_than("4.46"):
            kwargs["processing_class"] = kwargs.pop("tokenizer")

        super().__init__(**kwargs)
        self.finetuning_args = finetuning_args

        if processor is not None:
            self.add_callback(SaveProcessorCallback(processor))

        if finetuning_args.pissa_convert:
            self.add_callback(PissaConvertCallback)

        if finetuning_args.use_badam:
            from badam import BAdamCallback, clip_grad_norm_old_version  # type: ignore

            self.accelerator.clip_grad_norm_ = MethodType(clip_grad_norm_old_version, self.accelerator)
            self.add_callback(BAdamCallback)

    @override
    def create_optimizer(self) -> "torch.optim.Optimizer":
        if self.optimizer is None:
            self.optimizer = create_custom_optimizer(self.model, self.args, self.finetuning_args)
        return super().create_optimizer()

    @override
    def create_scheduler(
        self, num_training_steps: int, optimizer: Optional["torch.optim.Optimizer"] = None
    ) -> "torch.optim.lr_scheduler.LRScheduler":
        create_custom_scheduler(self.args, num_training_steps, optimizer)
        return super().create_scheduler(num_training_steps, optimizer)

    @override
    def _get_train_sampler(self) -> Optional["torch.utils.data.Sampler"]:
        if self.finetuning_args.disable_shuffling:
            return torch.utils.data.SequentialSampler(self.train_dataset)

        return super()._get_train_sampler()

    @override
    def compute_loss(
        self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs
    ) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]:
        r"""
        Fixes the loss value for transformers 4.46.0.
        https://github.com/huggingface/transformers/blob/v4.46.0/src/transformers/trainer.py#L3605
        """
        loss = super().compute_loss(model, inputs, return_outputs, **kwargs)
        if is_transformers_version_equal_to_4_46() and not getattr(self, "model_accepts_loss_kwargs", False):
            # other model should not scale the loss
            if return_outputs:
                return (loss[0] / self.args.gradient_accumulation_steps, *loss[1:])
            else:
                return loss / self.args.gradient_accumulation_steps

        return loss