ros/LlamaFactory
1
0
Fork 0
mirror of https://github.com/hiyouga/LlamaFactory.git synced 2026-02-03 08:53:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

[fix] add visual.pos_embed to Qwen3-VL visual model keys (#10139)

Browse Source
This commit is contained in:
Jewon Lee
2026-01-27 17:33:01 +09:00
committed by GitHub
parent 7ef19eea00
commit 9640f79ae5
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -356,7 +356,7 @@ _register_composite_model(
_register_composite_model( _register_composite_model(
model_type="qwen3_vl", model_type="qwen3_vl",
projector_key="visual.merger", projector_key="visual.merger",
vision_model_keys=["visual.patch_embed", "visual.blocks", "visual.deepstack_merger_list"], vision_model_keys=["visual.pos_embed", "visual.patch_embed", "visual.blocks", "visual.deepstack_merger_list"],
language_model_keys=["language_model", "lm_head"], language_model_keys=["language_model", "lm_head"],
lora_conflict_keys=["patch_embed"], lora_conflict_keys=["patch_embed"],
) )
@@ -365,7 +365,7 @@ _register_composite_model(
_register_composite_model( _register_composite_model(
model_type="qwen3_vl_moe", model_type="qwen3_vl_moe",
projector_key="visual.merger", projector_key="visual.merger",
vision_model_keys=["visual.patch_embed", "visual.blocks", "visual.deepstack_merger_list"], vision_model_keys=["visual.pos_embed", "visual.patch_embed", "visual.blocks", "visual.deepstack_merger_list"],
language_model_keys=["language_model", "lm_head"], language_model_keys=["language_model", "lm_head"],
lora_conflict_keys=["patch_embed"], lora_conflict_keys=["patch_embed"],
) )
@@ -374,7 +374,7 @@ _register_composite_model(
_register_composite_model( _register_composite_model(
model_type="qwen3_omni_moe_thinker", model_type="qwen3_omni_moe_thinker",
projector_key="visual.merger", projector_key="visual.merger",
vision_model_keys=["visual.patch_embed", "visual.blocks", "visual.deepstack_merger_list", "audio_tower"], vision_model_keys=["visual.pos_embed", "visual.patch_embed", "visual.blocks", "visual.deepstack_merger_list", "audio_tower"],
language_model_keys=["model", "lm_head"], language_model_keys=["model", "lm_head"],
lora_conflict_keys=["patch_embed"], lora_conflict_keys=["patch_embed"],
) )