src/llmtuner/hparams/finetuning_args.py

import json
from dataclasses import asdict, dataclass, field
from typing import Literal, Optional


@dataclass
class FreezeArguments:
    r"""
    Arguments pertaining to the freeze (partial-parameter) training.
    """
    name_module_trainable: Optional[str] = field(
        default="mlp",
        metadata={
            "help": 'Name of trainable modules for partial-parameter (freeze) fine-tuning. \
                  Use commas to separate multiple modules. \
                  LLaMA choices: ["mlp", "self_attn"], \
                  BLOOM & Falcon & ChatGLM choices: ["mlp", "self_attention"], \
                  Qwen choices: ["mlp", "attn"], \
                  Phi choices: ["mlp", "mixer"], \
                  Others choices: the same as LLaMA.'
        },
    )
    num_layer_trainable: Optional[int] = field(
        default=3, metadata={"help": "The number of trainable layers for partial-parameter (freeze) fine-tuning."}
    )


@dataclass
class LoraArguments:
    r"""
    Arguments pertaining to the LoRA training.
    """
    additional_target: Optional[str] = field(
        default=None,
        metadata={
            "help": "Name(s) of modules apart from LoRA layers to be set as trainable and saved in the final checkpoint."
        },
    )
    lora_alpha: Optional[int] = field(
        default=None, metadata={"help": "The scale factor for LoRA fine-tuning (default: lora_rank * 2)."}
    )
    lora_dropout: Optional[float] = field(default=0.0, metadata={"help": "Dropout rate for the LoRA fine-tuning."})
    lora_rank: Optional[int] = field(default=8, metadata={"help": "The intrinsic dimension for LoRA fine-tuning."})
    lora_target: Optional[str] = field(
        default=None,
        metadata={
            "help": 'Name(s) of target modules to apply LoRA. Use commas to separate multiple modules. \
                  LLaMA choices: ["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"], \
                  BLOOM & Falcon & ChatGLM choices: ["query_key_value", "dense", "dense_h_to_4h", "dense_4h_to_h"], \
                  Baichuan choices: ["W_pack", "o_proj", "gate_proj", "up_proj", "down_proj"], \
                  Qwen choices: ["c_attn", "attn.c_proj", "w1", "w2", "mlp.c_proj"], \
                  Phi choices: ["Wqkv", "out_proj", "fc1", "fc2"], \
                  Others choices: the same as LLaMA.'
        },
    )
    lora_bf16_mode: Optional[bool] = field(
        default=False, metadata={"help": "Whether or not to train lora adapters in bf16 precision."}
    )
    create_new_adapter: Optional[bool] = field(
        default=False, metadata={"help": "Whether or not to create a new adapter with randomly initialized weight."}
    )


@dataclass
class RLHFArguments:
    r"""
    Arguments pertaining to the PPO and DPO training.
    """
    dpo_beta: Optional[float] = field(default=0.1, metadata={"help": "The beta parameter for the DPO loss."})
    dpo_loss: Optional[Literal["sigmoid", "hinge", "ipo", "kto"]] = field(
        default="sigmoid", metadata={"help": "The type of DPO loss to use."}
    )
    dpo_ftx: Optional[float] = field(
        default=0, metadata={"help": "The supervised fine-tuning loss coefficient in DPO training."}
    )
    ppo_buffer_size: Optional[int] = field(
        default=1,
        metadata={"help": "The number of mini-batches to make experience buffer in a PPO optimization step."},
    )
    ppo_epochs: Optional[int] = field(
        default=4, metadata={"help": "The number of epochs to perform in a PPO optimization step."}
    )
    ppo_logger: Optional[str] = field(
        default=None, metadata={"help": 'Log with either "wandb" or "tensorboard" in PPO training.'}
    )
    ppo_score_norm: Optional[bool] = field(
        default=False, metadata={"help": "Use score normalization in PPO training."}
    )
    ppo_target: Optional[float] = field(
        default=6.0, metadata={"help": "Target KL value for adaptive KL control in PPO training."}
    )
    ppo_whiten_rewards: Optional[bool] = field(
        default=False, metadata={"help": "Whiten the rewards before compute advantages in PPO training."}
    )
    ref_model: Optional[str] = field(
        default=None, metadata={"help": "Path to the reference model used for the PPO or DPO training."}
    )
    ref_model_adapters: Optional[str] = field(
        default=None, metadata={"help": "Path to the adapters of the reference model."}
    )
    ref_model_quantization_bit: Optional[int] = field(
        default=None, metadata={"help": "The number of bits to quantize the reference model."}
    )
    reward_model: Optional[str] = field(
        default=None, metadata={"help": "Path to the reward model used for the PPO training."}
    )
    reward_model_adapters: Optional[str] = field(
        default=None, metadata={"help": "Path to the adapters of the reward model."}
    )
    reward_model_quantization_bit: Optional[int] = field(
        default=None, metadata={"help": "The number of bits to quantize the reward model."}
    )
    reward_model_type: Optional[Literal["lora", "full", "api"]] = field(
        default="lora",
        metadata={"help": "The type of the reward model in PPO training. Lora model only supports lora training."},
    )


@dataclass
class FinetuningArguments(FreezeArguments, LoraArguments, RLHFArguments):
    r"""
    Arguments pertaining to which techniques we are going to fine-tuning with.
    """
    stage: Optional[Literal["pt", "sft", "rm", "ppo", "dpo"]] = field(
        default="sft", metadata={"help": "Which stage will be performed in training."}
    )
    finetuning_type: Optional[Literal["lora", "freeze", "full"]] = field(
        default="lora", metadata={"help": "Which fine-tuning method to use."}
    )
    plot_loss: Optional[bool] = field(
        default=False, metadata={"help": "Whether or not to save the training loss curves."}
    )

    def __post_init__(self):
        def split_arg(arg):
            if isinstance(arg, str):
                return [item.strip() for item in arg.split(",")]
            return arg

        self.name_module_trainable = split_arg(self.name_module_trainable)
        self.lora_alpha = self.lora_alpha or self.lora_rank * 2
        self.lora_target = split_arg(self.lora_target)
        self.additional_target = split_arg(self.additional_target)

        assert self.finetuning_type in ["lora", "freeze", "full"], "Invalid fine-tuning method."
        assert self.ref_model_quantization_bit in [None, 8, 4], "We only accept 4-bit or 8-bit quantization."
        assert self.reward_model_quantization_bit in [None, 8, 4], "We only accept 4-bit or 8-bit quantization."

        if self.stage == "ppo" and self.reward_model is None:
            raise ValueError("Reward model is necessary for PPO training.")

        if self.stage == "ppo" and self.reward_model_type == "lora" and self.finetuning_type != "lora":
            raise ValueError("Freeze/Full PPO training needs `reward_model_type=full`.")

    def save_to_json(self, json_path: str):
        r"""Saves the content of this instance in JSON format inside `json_path`."""
        json_string = json.dumps(asdict(self), indent=2, sort_keys=True) + "\n"
        with open(json_path, "w", encoding="utf-8") as f:
            f.write(json_string)

    @classmethod
    def load_from_json(cls, json_path: str):
        r"""Creates an instance from the content of `json_path`."""
        with open(json_path, "r", encoding="utf-8") as f:
            text = f.read()

        return cls(**json.loads(text))