423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-18 12:50:38 +08:00
Code Issues Packages Projects Releases Wiki Activity

add forbidden modules

Browse Source
This commit is contained in:
hiyouga
2024-11-23 18:34:15 +00:00
parent 446441fdb0
commit df477370dc
2 changed files with 6 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -34,13 +34,15 @@ def find_all_linear_modules(model: "PreTrainedModel", freeze_vision_tower: bool)
forbidden_modules.add("output_layer")
elif model_type == "internlm2":
forbidden_modules.add("output")
elif model_type in ["llava", "llava_next", "llava_next_video", "paligemma", "video_llava"]:
elif model_type in ["llava", "llava_next", "llava_next_video", "mllama", "paligemma", "video_llava"]:
forbidden_modules.add("multi_modal_projector")
elif model_type == "qwen2_vl":
forbidden_modules.add("merger")
if freeze_vision_tower:
if model_type == "qwen2_vl":
if model_type == "mllama":
forbidden_modules.add("vision_model")
elif model_type == "qwen2_vl":
forbidden_modules.add("visual")
else:
forbidden_modules.add("vision_tower")