Update subproject commit and clean up training configuration

- Updated subproject commit to indicate a dirty state.
- Removed the `create_new_adapter` field from `TrainSftArgs` class in `config_models.py` to streamline training configuration.

weclone/utils/config_models.py

@@ -206,10 +206,6 @@ class TrainSftArgs(BaseConfigModel):
         None,
         description="Existing LoRA adapter path to continue SFT from. Output still uses common_args.adapter_name_or_path.",
     )
-    create_new_adapter: bool = Field(
-        False,
-        description="Whether to create a new LoRA adapter instead of continuing resume_adapter_name_or_path.",
-    )
     freeze_multi_modal_projector: bool = Field(
         False, description="Whether to freeze multimodal projector during MLLM training"
     )