p12475398
/
LLaMA-Factory-Mirror
forked from p04798526/LLaMA-Factory-Mirror
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix PPO trainer #551 , update readme

This commit is contained in:
hiyouga 2023-08-18 11:43:10 +08:00
parent e4eec9ddfd
commit 9020524418
6 changed files with 42 additions and 53 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
README.md
View File

@ -171,11 +171,12 @@ Currently the web UI only supports training on **a single GPU**.
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage pt \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset wiki_demo \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--output_dir path_to_pt_checkpoint \
--overwrite_cache \
--per_device_train_batch_size 4 \
@ -194,11 +195,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage sft \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset alpaca_gpt4_en \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--output_dir path_to_sft_checkpoint \
--overwrite_cache \
--per_device_train_batch_size 4 \
@ -217,11 +219,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage rm \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset comparison_gpt4_en \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--resume_lora_training False \
--checkpoint_dir path_to_sft_checkpoint \
--output_dir path_to_rm_checkpoint \
@ -230,7 +233,7 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--lr_scheduler_type cosine \
--logging_steps 10 \
--save_steps 1000 \
--learning_rate 1e-5 \
--learning_rate 1e-6 \
--num_train_epochs 1.0 \
--plot_loss \
--fp16
@ -241,11 +244,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage ppo \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset alpaca_gpt4_en \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--resume_lora_training False \
--checkpoint_dir path_to_sft_checkpoint \
--reward_model path_to_rm_checkpoint \
@ -266,11 +270,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage dpo \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset comparison_gpt4_en \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--resume_lora_training False \
--checkpoint_dir path_to_sft_checkpoint \
--output_dir path_to_dpo_checkpoint \
@ -364,7 +369,7 @@ deepspeed --num_gpus 8 --master_port=9901 src/train_bash.py \
```bash
python src/export_model.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint \
@ -375,7 +380,7 @@ python src/export_model.py \
```bash
python src/api_demo.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint
@ -387,7 +392,7 @@ Visit `http://localhost:8000/docs` for API documentation.
```bash
python src/cli_demo.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint
@ -397,7 +402,7 @@ python src/cli_demo.py \
```bash
python src/web_demo.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint
@ -408,7 +413,7 @@ python src/web_demo.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage sft \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_eval \
--dataset alpaca_gpt4_en \
--template default \
@ -427,7 +432,7 @@ We recommend using `--per_device_eval_batch_size=1` and `--max_target_length 128
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage sft \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_predict \
--dataset alpaca_gpt4_en \
--template default \

29
README_zh.md
View File

@ -171,11 +171,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_web.py
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage pt \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset wiki_demo \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--output_dir path_to_pt_checkpoint \
--overwrite_cache \
--per_device_train_batch_size 4 \
@ -194,11 +195,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage sft \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset alpaca_gpt4_zh \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--output_dir path_to_sft_checkpoint \
--overwrite_cache \
--per_device_train_batch_size 4 \
@ -217,11 +219,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage rm \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset comparison_gpt4_zh \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--resume_lora_training False \
--checkpoint_dir path_to_sft_checkpoint \
--output_dir path_to_rm_checkpoint \
@ -230,7 +233,7 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--lr_scheduler_type cosine \
--logging_steps 10 \
--save_steps 1000 \
--learning_rate 1e-5 \
--learning_rate 1e-6 \
--num_train_epochs 1.0 \
--plot_loss \
--fp16
@ -241,11 +244,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage ppo \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset alpaca_gpt4_zh \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--resume_lora_training False \
--checkpoint_dir path_to_sft_checkpoint \
--reward_model path_to_rm_checkpoint \
@ -265,11 +269,12 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage dpo \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_train \
--dataset comparison_gpt4_zh \
--template default \
--finetuning_type lora \
--lora_target q_proj,v_proj \
--resume_lora_training False \
--checkpoint_dir path_to_sft_checkpoint \
--output_dir path_to_dpo_checkpoint \
@ -363,7 +368,7 @@ deepspeed --num_gpus 8 --master_port=9901 src/train_bash.py \
```bash
python src/export_model.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint \
@ -374,7 +379,7 @@ python src/export_model.py \
```bash
python src/api_demo.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint
@ -386,7 +391,7 @@ python src/api_demo.py \
```bash
python src/cli_demo.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint
@ -396,7 +401,7 @@ python src/cli_demo.py \
```bash
python src/web_demo.py \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--template default \
--finetuning_type lora \
--checkpoint_dir path_to_checkpoint
@ -407,7 +412,7 @@ python src/web_demo.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage sft \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_eval \
--dataset alpaca_gpt4_zh \
--template default \
@ -426,7 +431,7 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
--stage sft \
--model_name_or_path path_to_your_model \
--model_name_or_path path_to_llama_model \
--do_predict \
--dataset alpaca_gpt4_zh \
--template default \

6
src/llmtuner/hparams/finetuning_args.py
View File

@ -8,7 +8,7 @@ class FinetuningArguments:
r"""
Arguments pertaining to which techniques we are going to fine-tuning with.
"""
finetuning_type: Optional[Literal["none", "freeze", "lora", "full"]] = field(
finetuning_type: Optional[Literal["lora", "freeze", "full", "none"]] = field(
default="lora",
metadata={"help": "Which fine-tuning method to use."}
)
@ -49,7 +49,7 @@ class FinetuningArguments:
metadata={"help": "Dropout rate for the LoRA fine-tuning."}
)
lora_target: Optional[str] = field(
default="q_proj,v_proj",
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 choices: [\"query_key_value\", \"self_attention.dense\", \"mlp.dense\"], \
@ -77,7 +77,7 @@ class FinetuningArguments:
self.trainable_layers = ["{:d}.{}".format(idx, self.name_module_trainable) for idx in trainable_layer_ids]
assert self.finetuning_type in ["none", "freeze", "lora", "full"], "Invalid fine-tuning method."
assert self.finetuning_type in ["lora", "freeze", "full", "none"], "Invalid fine-tuning method."
def save_to_json(self, json_path: str):
r"""Saves the content of this instance in JSON format inside `json_path`."""

3
src/llmtuner/tuner/core/parser.py
View File

@ -125,6 +125,9 @@ def get_train_args(
if training_args.do_train and training_args.predict_with_generate:
raise ValueError("`predict_with_generate` cannot be set as True while training.")
if training_args.do_train and finetuning_args.finetuning_type == "lora" and finetuning_args.lora_target is None:
raise ValueError("Please specify `lora_target` in LoRA training.")
if model_args.quantization_bit is not None and finetuning_args.finetuning_type != "lora":
raise ValueError("Quantization is only compatible with the LoRA method.")

4
src/llmtuner/tuner/ppo/trainer.py
View File

@ -12,7 +12,7 @@ from trl.core import LengthSampler, PPODecorators, logprobs_from_logits
from llmtuner.extras.logging import get_logger
from llmtuner.extras.misc import AverageMeter, count_parameters, get_logits_processor
from llmtuner.tuner.core.trainer import PeftTrainer
from llmtuner.tuner.ppo.utils import cast_layernorm_dtype, replace_model
from llmtuner.tuner.ppo.utils import replace_model
if TYPE_CHECKING:
from transformers import Seq2SeqTrainingArguments
@ -152,10 +152,8 @@ class PPOPeftTrainer(PPOTrainer, PeftTrainer):
if length_sampler is not None:
generation_kwargs["max_new_tokens"] = length_sampler()
self.model, layer_norm_params = cast_layernorm_dtype(self.model)
unwrapped_model: "AutoModelForCausalLMWithValueHead" = self.accelerator.unwrap_model(self.model)
response: torch.Tensor = unwrapped_model.generate(**batch, **generation_kwargs)
self.model, _ = cast_layernorm_dtype(self.model, layer_norm_params)
# Temporary hack to ensure the generation config is not initialized for each iteration of the evaluation loop
# Inspired by: https://github.com/huggingface/transformers/blob/v4.28.1/src/transformers/trainer_seq2seq.py#L273

24
src/llmtuner/tuner/ppo/utils.py
View File

@ -1,7 +1,4 @@
import torch
from typing import TYPE_CHECKING, Dict, List, Literal, Optional, Tuple
from llmtuner.extras.constants import LAYERNORM_NAMES
from typing import TYPE_CHECKING, Literal
if TYPE_CHECKING:
from trl import AutoModelForCausalLMWithValueHead
@ -18,22 +15,3 @@ def replace_model(model: "AutoModelForCausalLMWithValueHead", target: Literal["d
"summary.weight": getattr(model, "{}_head_weight".format(target)),
"summary.bias": getattr(model, "{}_head_bias".format(target))
})
def cast_layernorm_dtype(
model: "AutoModelForCausalLMWithValueHead",
layer_norm_names: List[str] = LAYERNORM_NAMES,
layer_norm_params: Optional[Dict[str, torch.Tensor]] = None
) -> Tuple["AutoModelForCausalLMWithValueHead", Dict[str, torch.Tensor]]:
layer_norm_state_dict = {}
for name, param in model.named_parameters():
if param.ndim == 1 and any(layer_norm_name in name for layer_norm_name in layer_norm_names):
if layer_norm_params is not None:
param.data = layer_norm_params[name] # restore float32 weights
else:
layer_norm_state_dict[name] = param.data.detach().clone() # store float32 weights for stability
param.data = param.data.to(torch.float16)
return model, layer_norm_state_dict