From 927ff702ffcdb565363c126f33e3a3538ede76ac Mon Sep 17 00:00:00 2001 From: hiyouga Date: Wed, 27 Sep 2023 22:28:06 +0800 Subject: [PATCH] refactor finetuning Args Former-commit-id: 620efe1d8d2429f4bc3fa8009900ec43e1b5ef4b --- src/llmtuner/hparams/finetuning_args.py | 29 ++++++++----------------- src/llmtuner/tuner/core/adapter.py | 11 ++++++++-- 2 files changed, 18 insertions(+), 22 deletions(-) diff --git a/src/llmtuner/hparams/finetuning_args.py b/src/llmtuner/hparams/finetuning_args.py index bda0adaf..09ed25d3 100644 --- a/src/llmtuner/hparams/finetuning_args.py +++ b/src/llmtuner/hparams/finetuning_args.py @@ -12,18 +12,6 @@ class FinetuningArguments: default="lora", metadata={"help": "Which fine-tuning method to use."} ) - num_hidden_layers: Optional[int] = field( - default=32, - metadata={"help": "Number of decoder blocks in the model for partial-parameter (freeze) fine-tuning. \ - LLaMA choices: [\"32\", \"40\", \"60\", \"80\"], \ - LLaMA-2 choices: [\"32\", \"40\", \"80\"], \ - BLOOM choices: [\"24\", \"30\", \"70\"], \ - Falcon choices: [\"32\", \"60\"], \ - Baichuan choices: [\"32\", \"40\"] \ - Qwen choices: [\"32\"], \ - XVERSE choices: [\"40\"], \ - ChatGLM2 choices: [\"28\"]"} - ) num_layer_trainable: Optional[int] = field( default=3, metadata={"help": "Number of trainable layers for partial-parameter (freeze) fine-tuning."} @@ -33,9 +21,9 @@ class FinetuningArguments: metadata={"help": "Name of trainable modules for partial-parameter (freeze) fine-tuning. \ LLaMA choices: [\"mlp\", \"self_attn\"], \ BLOOM & Falcon & ChatGLM2 choices: [\"mlp\", \"self_attention\"], \ - Baichuan choices: [\"mlp\", \"self_attn\"], \ Qwen choices: [\"mlp\", \"attn\"], \ - LLaMA-2, InternLM, XVERSE choices: the same as LLaMA."} + Phi-1.5 choices: [\"mlp\", \"mixer\"], \ + LLaMA-2, Baichuan, InternLM, XVERSE choices: the same as LLaMA."} ) lora_rank: Optional[int] = field( default=8, @@ -56,8 +44,13 @@ class FinetuningArguments: BLOOM & Falcon & ChatGLM2 choices: [\"query_key_value\", \"self_attention.dense\", \"mlp.dense\"], \ 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-1.5 choices: [\"Wqkv\", \"out_proj\", \"fc1\", \"fc2\"], \ LLaMA-2, InternLM, XVERSE choices: the same as LLaMA."} ) + 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."} + ) resume_lora_training: Optional[bool] = field( default=True, metadata={"help": "Whether to resume training from the last LoRA weights or create new weights after merging them."} @@ -75,12 +68,8 @@ class FinetuningArguments: if isinstance(self.lora_target, str): # support custom target modules/layers of LoRA self.lora_target = [target.strip() for target in self.lora_target.split(",")] - if self.num_layer_trainable > 0: # fine-tuning the last n layers if num_layer_trainable > 0 - trainable_layer_ids = [self.num_hidden_layers - k - 1 for k in range(self.num_layer_trainable)] - else: # fine-tuning the first n layers if num_layer_trainable < 0 - trainable_layer_ids = [k for k in range(-self.num_layer_trainable)] - - self.trainable_layers = ["{:d}.{}".format(idx, self.name_module_trainable) for idx in trainable_layer_ids] + if isinstance(self.additional_target, str): + self.additional_target = [target.strip() for target in self.additional_target.split(",")] assert self.finetuning_type in ["lora", "freeze", "full", "none"], "Invalid fine-tuning method." diff --git a/src/llmtuner/tuner/core/adapter.py b/src/llmtuner/tuner/core/adapter.py index 6a9e454e..cba9b10f 100644 --- a/src/llmtuner/tuner/core/adapter.py +++ b/src/llmtuner/tuner/core/adapter.py @@ -45,9 +45,15 @@ def init_adapter( if finetuning_args.finetuning_type == "freeze": logger.info("Fine-tuning method: Freeze") + num_layers = getattr(model.config, "num_layers") + if finetuning_args.num_layer_trainable > 0: # fine-tuning the last n layers if num_layer_trainable > 0 + trainable_layer_ids = [num_layers - k - 1 for k in range(finetuning_args.num_layer_trainable)] + else: # fine-tuning the first n layers if num_layer_trainable < 0 + trainable_layer_ids = [k for k in range(-finetuning_args.num_layer_trainable)] + trainable_layers = ["{:d}.{}".format(idx, finetuning_args.name_module_trainable) for idx in trainable_layer_ids] for name, param in model.named_parameters(): - if not any(trainable_layer in name for trainable_layer in finetuning_args.trainable_layers): + if not any(trainable_layer in name for trainable_layer in trainable_layers): param.requires_grad_(False) else: param.data = param.data.to(torch.float32) @@ -89,7 +95,8 @@ def init_adapter( r=finetuning_args.lora_rank, lora_alpha=finetuning_args.lora_alpha, lora_dropout=finetuning_args.lora_dropout, - target_modules=target_modules + target_modules=target_modules, + modules_to_save=finetuning_args.additional_target ) model = get_peft_model(model, lora_config) if id(model.peft_config) != id(model.base_model.peft_config): # https://github.com/huggingface/peft/issues/923