[Quant][GLM] Adapt glm quant. (#3147)

adapt glm quant - vLLM version: v0.10.2 - vLLM main: f225ea7dd9 Signed-off-by: whx-sjtu <2952154980@qq.com>
2025-09-25 11:13:29 +08:00
parent a055183821
commit c814b32b90
1 changed files with 14 additions and 1 deletions
--- a/vllm_ascend/quantization/quant_config.py
+++ b/vllm_ascend/quantization/quant_config.py
@@ -210,7 +210,20 @@ packed_modules_model_mapping = {
            "gate_proj",
            "up_proj",
        ],
-    }
+    },
+    "glm4_moe": {
+        "qkv_proj": [
+            "q_proj",
+            "k_proj",
+            "v_proj",
+        ],
+        "gate_up_proj": [
+            "gate_proj",
+            "up_proj",
+        ],
+        "experts":
+        ["experts.0.gate_proj", "experts.0.up_proj", "experts.0.down_proj"]
+    },
 }