diff --git a/python/sglang/srt/layers/activation.py b/python/sglang/srt/layers/activation.py
index 5dc48821adc..d00c34b7c55 100644
--- a/python/sglang/srt/layers/activation.py
+++ b/python/sglang/srt/layers/activation.py
@@ -380,4 +380,14 @@ def get_cross_encoder_activation_function(config: PretrainedConfig):
     logger.info(
         "sgl-kernel is not available on Non-NV, Non-AMD platforms or Non-AMX CPUs. Fallback to other kernel libraries."
     )
-    from vllm.model_executor.layers.activation import GeluAndMul, SiluAndMul
+    try:
+        from vllm.model_executor.layers.activation import GeluAndMul, SiluAndMul
+    except ImportError:
+
+        class GeluAndMul:
+            def __init__(self, *args, **kwargs):
+                pass
+
+        class SiluAndMul:
+            def __init__(self, *args, **kwargs):
+                pass
diff --git a/python/sglang/utils.py b/python/sglang/utils.py
index 1d62c5df854..f34c6fa1b9a 100644
--- a/python/sglang/utils.py
+++ b/python/sglang/utils.py
@@ -14,7 +14,7 @@
 import urllib.request
 import weakref
 from concurrent.futures import ThreadPoolExecutor
-from functools import wraps
+from functools import lru_cache, wraps
 from io import BytesIO
 from json import dumps
 from typing import Any, Callable, List, Optional, Tuple, Type, Union
@@ -542,6 +542,7 @@ async def async_stream_and_merge(llm, prompt, sampling_params):
         yield cleaned_chunk  # yield the non-overlapping portion
 
 
+@lru_cache(maxsize=128)
 def resolve_obj_by_qualname(qualname: str) -> Any:
     """
     Resolve an object by its fully qualified name.