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

fix mixed mm inputs and rlhf-v

Browse Source 
Former-commit-id: 7c248fac20bf85d57a91132ce7a793c7f84e9218
This commit is contained in:
hiyouga
2024-09-01 20:52:47 +08:00
parent 1d8e9c7897
commit 7e4c5d4bb3
20 changed files with 306 additions and 277 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llamafactory/train/ppo/workflow.py
View File

@@ -17,7 +17,7 @@
from typing import TYPE_CHECKING, List, Optional
from ...data import CustomDataCollatorForSeq2Seq, get_dataset
from ...data import MultiModalDataCollatorForSeq2Seq, get_dataset
from ...extras.ploting import plot_loss
from ...model import load_model, load_tokenizer
from ..callbacks import fix_valuehead_checkpoint
@@ -45,7 +45,7 @@ def run_ppo(
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
data_collator = CustomDataCollatorForSeq2Seq(tokenizer=tokenizer)
data_collator = MultiModalDataCollatorForSeq2Seq(tokenizer=tokenizer)
# Create reference model and reward model
ref_model = create_ref_model(model_args, finetuning_args, add_valuehead=True)