ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #1553 from hannlp/hans

Browse Source 
Change the default argument settings for PPO training

Former-commit-id: 1b64678fa4979485f67c3bb1420dfdff6fcbc6e7
This commit is contained in:
hoshi-hiyouga
2023-11-20 20:32:55 +08:00
committed by GitHub
parent bcd661afa6 992be39f90
commit e585950c54
4 changed files with 9 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/llmtuner/train/ppo/workflow.py
View File

@@ -45,7 +45,7 @@ def run_ppo(
mini_batch_size=training_args.per_device_train_batch_size,
batch_size=training_args.per_device_train_batch_size * training_args.gradient_accumulation_steps,
gradient_accumulation_steps=training_args.gradient_accumulation_steps,
ppo_epochs=1,
ppo_epochs=finetuning_args.ppo_epochs,
max_grad_norm=training_args.max_grad_norm,
seed=training_args.seed,
optimize_device_cache=True,