diff --git a/python/sglang/srt/layers/quantization/modelslim/modelslim.py b/python/sglang/srt/layers/quantization/modelslim/modelslim.py
index 2106d6568d68..1212a40d48d3 100644
--- a/python/sglang/srt/layers/quantization/modelslim/modelslim.py
+++ b/python/sglang/srt/layers/quantization/modelslim/modelslim.py
@@ -239,15 +239,6 @@ def is_layer_skipped(
     ):
         # adapted from vllm.model_executor.layers.quantization.utils.quant_utils.is_layer_skipped
         proj_name = prefix.split(".")[-1]
-        if not hasattr(self, "_quant_description_normalized"):
-            quant_description = {}
-            for prefix_, value in self.quant_description.items():
-                prefix_ = prefix_.replace("language_model.", "")
-                if "visual" in prefix_:
-                    prefix_ = prefix_.replace("model.", "")
-                quant_description[prefix_] = value
-            self.quant_description = quant_description
-            self._quant_description_normalized = True
         if proj_name in fused_mapping:
             shard_prefixes = [
                 prefix.replace(proj_name, shard_proj_name)