[Misc] Reduce LoRA-related static variable (#13166)

2025-02-22 16:21:30 +08:00
parent 2cb8c1540e
commit 105b8ce4c0
41 changed files with 120 additions and 395 deletions
--- a/vllm/model_executor/models/glm4v.py
+++ b/vllm/model_executor/models/glm4v.py
@@ -534,21 +534,6 @@ class GLM4VForCausalLM(ChatGLMBaseModel, SupportsLoRA, SupportsPP,
        "dense_h_to_4h": ["dense_h_to_4h"],
        "merged_proj": ["gate_proj", "dense_h_to_4h"]
    }
-    # LoRA specific attributes
-    supported_lora_modules = [
-        "query_key_value",
-        "dense",
-        "dense_h_to_4h",
-        "dense_4h_to_h",
-        # vision
-        "fc1",
-        "fc2",
-        "merged_proj",
-        "linear_proj"
-    ]
-
-    embedding_modules = {}
-    embedding_padding_modules = []

    def get_mm_mapping(self) -> MultiModelKeys:
        """