423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-03 20:22:49 +08:00
Code Issues Packages Projects Releases Wiki Activity

add dataset stage check

Browse Source 
Former-commit-id: f7fdc088d49564f7d436fd445e7e1987a9a00a0b
This commit is contained in:
codemayq 2023-08-30 16:23:08 +08:00
parent 00c3cd5454
commit c955d9267c
1 changed files with 12 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/llmtuner/tuner/core/parser.py
View File

@ -186,6 +186,18 @@ def get_train_args(
else:
model_args.compute_dtype = torch.float16
# transfer training stage to dataset stage
dataset_stage = general_args.stage
if general_args.stage == "ppo":
dataset_stage = "sft"
elif general_args.stage == "dpo":
dataset_stage = "rm"
for dataset_attr in data_args.dataset_list:
if dataset_attr.stage and dataset_attr.stage != dataset_stage:
raise ValueError("Dataset {} is not supported for the stage {}"
.format(dataset_attr.dataset_name, general_args.stage))
model_args.model_max_length = data_args.max_source_length + data_args.max_target_length
# Log on each process the small summary: