p04798526
/
LLaMA-Factory-310P3
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #3057 from marko1616/bugfix/lora-model-merge 

Fix Llama model save for full param train
This commit is contained in:
hoshi-hiyouga 2024-03-31 00:07:20 +08:00 committed by GitHub
parent 7a086ed333 d9a5134617
commit de3564ff70
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 9 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
src/llmtuner/model/patcher.py
View File

@ -312,6 +312,15 @@ def patch_config(
def patch_model( def patch_model(
model: "PreTrainedModel", tokenizer: "PreTrainedTokenizer", model_args: "ModelArguments", is_trainable: bool model: "PreTrainedModel", tokenizer: "PreTrainedTokenizer", model_args: "ModelArguments", is_trainable: bool
) -> None: ) -> None:
#Config check and fix
gen_config = model.generation_config
if not gen_config.do_sample and (
(gen_config.temperature is not None and gen_config.temperature != 1.0)
or (gen_config.top_p is not None and gen_config.top_p != 1.0)
or (gen_config.typical_p is not None and gen_config.typical_p != 1.0)
):
gen_config.do_sample = True
if "GenerationMixin" not in str(model.generate.__func__): if "GenerationMixin" not in str(model.generate.__func__):
model.generate = MethodType(PreTrainedModel.generate, model) model.generate = MethodType(PreTrainedModel.generate, model)

8
src/llmtuner/train/tuner.py
View File

@ -64,14 +64,6 @@ def export_model(args: Optional[Dict[str, Any]] = None):
for param in model.parameters(): for param in model.parameters():
param.data = param.data.to(output_dtype) param.data = param.data.to(output_dtype)
gen_config = model.generation_config # check and fix generation config
if not gen_config.do_sample and (
(gen_config.temperature is not None and gen_config.temperature != 1.0)
or (gen_config.top_p is not None and gen_config.top_p != 1.0)
or (gen_config.typical_p is not None and gen_config.typical_p != 1.0)
):
gen_config.do_sample = True
model.save_pretrained( model.save_pretrained(
save_directory=model_args.export_dir, save_directory=model_args.export_dir,
max_shard_size="{}GB".format(model_args.export_size), max_shard_size="{}GB".format(model_args.export_size),