423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-16 11:50:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix #2161

Browse Source
This commit is contained in:
hiyouga
2024-01-11 17:04:13 +08:00
parent 1653c22438
commit 898ec3696a
4 changed files with 75 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/llmtuner/train/ppo/workflow.py
View File

@@ -9,6 +9,7 @@ from transformers.optimization import get_scheduler
from llmtuner.data import get_dataset, preprocess_dataset
from llmtuner.extras.callbacks import FixValueHeadModelCallback
from llmtuner.extras.misc import fix_valuehead_checkpoint
from llmtuner.extras.ploting import plot_loss
from llmtuner.model import load_model_and_tokenizer
from llmtuner.train.utils import create_ref_model, create_reward_model
@@ -95,6 +96,8 @@ def run_ppo(
if training_args.do_train:
ppo_trainer.ppo_train(resume_from_checkpoint=training_args.resume_from_checkpoint)
ppo_trainer.save_model()
if training_args.should_save:
fix_valuehead_checkpoint(model, training_args.output_dir, training_args.save_safetensors)
ppo_trainer.save_state() # must be called after save_model to have a folder
if ppo_trainer.is_world_process_zero() and finetuning_args.plot_loss:
plot_loss(training_args.output_dir, keys=["loss", "reward"])