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

update get template

Browse Source 
Former-commit-id: 21ea0d0786f91c0bce79630963e66b815a6792a0
This commit is contained in:
hiyouga
2024-09-04 22:36:20 +08:00
parent 5d85be31ca
commit af178cbcd1
17 changed files with 57 additions and 56 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/llamafactory/train/ppo/workflow.py
View File

@@ -17,7 +17,7 @@
from typing import TYPE_CHECKING, List, Optional
from ...data import MultiModalDataCollatorForSeq2Seq, get_dataset
from ...data import MultiModalDataCollatorForSeq2Seq, get_dataset, get_template_and_fix_tokenizer
from ...extras.ploting import plot_loss
from ...model import load_model, load_tokenizer
from ..callbacks import fix_valuehead_checkpoint
@@ -41,7 +41,8 @@ def run_ppo(
):
tokenizer_module = load_tokenizer(model_args)
tokenizer = tokenizer_module["tokenizer"]
dataset_module, template = get_dataset(model_args, data_args, training_args, stage="ppo", **tokenizer_module)
template = get_template_and_fix_tokenizer(tokenizer, data_args)
dataset_module = get_dataset(template, model_args, data_args, training_args, stage="ppo", **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