def __init__()

in src/open-r1-multimodal/src/open_r1/trainer/vllm_grpo_trainer.py [0:0]

• 21 lines of code
• 1 McCabe index (conditional complexity)

    def __init__(
        self,
        model: Union[str, PreTrainedModel],
        reward_funcs: Union[RewardFunc, list[RewardFunc]],
        args: GRPOConfig = None,
        train_dataset: Optional[Union[Dataset, IterableDataset]] = None,
        eval_dataset: Optional[
            Union[Dataset, IterableDataset, dict[str, Union[Dataset, IterableDataset]]]
        ] = None,
        processing_class: Optional[PreTrainedTokenizerBase] = None,
        reward_processing_classes: Optional[
            Union[PreTrainedTokenizerBase, list[PreTrainedTokenizerBase]]
        ] = None,
        callbacks: Optional[list[TrainerCallback]] = None,
        optimizers: tuple[
            Optional[torch.optim.Optimizer], Optional[torch.optim.lr_scheduler.LambdaLR]
        ] = (None, None),
        peft_config: Optional["PeftConfig"] = None,
        # qwen2-vl related params
        max_pixels: Optional[int] = 12845056,
        min_pixels: Optional[int] = 3136,
        attn_implementation: str = "flash_attention_2",