Enable RadixCache for Mamba2 models (#13584)

python/sglang/srt/model_executor/model_runner.py

@@ -441,11 +441,6 @@ class ModelRunner:
             if architectures and not any("Llama4" in arch for arch in architectures):
                 self.is_hybrid_swa = self.model_config.is_hybrid_swa = True
 
-        if config := self.mamba2_config:
-            class_name = config.__class__.__name__
-            logger.warning(f"{class_name} model detected, disable radix cache")
-            self.server_args.disable_radix_cache = True
-
         # For MTP models like DeepSeek-V3 or GLM-4.5, the MTP layer(s) are used separately as draft
         # models for speculative decoding. In those cases, `num_nextn_predict_layers` is used to
         # determine the number of layers.