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

add some

Browse Source 
Former-commit-id: b4790c66c126567bd193de52a564e3ce11c94769
This commit is contained in:
fzc8578
2025-01-06 19:32:39 +08:00
parent 08729dbefc
commit 8c2a712247
4 changed files with 15 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llamafactory/model/model_utils/visual.py
View File

@@ -142,6 +142,10 @@ def get_forbidden_modules(config: "PretrainedConfig", finetuning_args: "Finetuni
forbidden_modules.update({"visual.patch_embed", "visual.blocks", "model", "lm_head"})
elif finetuning_args.freeze_vision_tower:
forbidden_modules.add("visual")
elif model_type == "minicpmv":
if finetuning_args.freeze_vision_tower:
forbidden_modules.add("vpm")
return forbidden_modules
@@ -196,6 +200,8 @@ def patch_target_modules(
return "^(?!.*vision_model).*(?:{}).*".format("|".join(target_modules))
elif model_type == "qwen2_vl":
return "^(?!.*visual).*(?:{}).*".format("|".join(target_modules))
elif model_type == "minicpmv":
return "^(?!.*vpm).*(?:{}).*".format("|".join(target_modules))
else:
return target_modules
else:

13
src/llamafactory/model/patcher.py
View File

@@ -138,12 +138,13 @@ def patch_model(
add_valuehead: bool,
) -> None:
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
if gen_config is not None:
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__):
model.generate = MethodType(PreTrainedModel.generate, model)