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

improve KTO impl., replace datasets

Browse Source 
Former-commit-id: e56a57ddcf061de6e4acc8679f7dbf0b68364986
This commit is contained in:
hiyouga
2024-05-18 03:44:56 +08:00
parent e4570e28a8
commit 2bff90719b
53 changed files with 448 additions and 330 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -29,7 +29,7 @@ def run_ppo(
):
tokenizer_module = load_tokenizer(model_args)
tokenizer = tokenizer_module["tokenizer"]
dataset = get_dataset(model_args, data_args, training_args, stage="ppo", **tokenizer_module)
dataset = get_dataset(model_args, data_args, training_args, stage="pt", **tokenizer_module)
model = load_model(tokenizer, model_args, finetuning_args, training_args.do_train, add_valuehead=True)
tokenizer.padding_side = "left" # use left-padding in generation while using right-padding in training