initial-commit

Former-commit-id: b6a39847a10b417b09db4b5512dd835e9e4ce928
2024-08-28 16:51:35 +08:00
parent 7272792f65
commit 0f3d54d8a0
8 changed files with 183 additions and 5 deletions
--- a/src/llamafactory/model/model_utils/misc.py
+++ b/src/llamafactory/model/model_utils/misc.py
@@ -36,6 +36,8 @@ def find_all_linear_modules(model: "PreTrainedModel", freeze_vision_tower: bool)
        forbidden_modules.add("output")
    elif model.config.model_type in ["llava", "paligemma"]:
        forbidden_modules.add("multi_modal_projector")
+    elif model.config.model_type in ["qwen2_vl"]:
+        forbidden_modules.add("merger")

    if freeze_vision_tower:
        forbidden_modules.add("vision_tower")