diff --git a/python/sglang/srt/disaggregation/fake/conn.py b/python/sglang/srt/disaggregation/fake/conn.py
index d59641c3c..e44cbb7b3 100644
--- a/python/sglang/srt/disaggregation/fake/conn.py
+++ b/python/sglang/srt/disaggregation/fake/conn.py
@@ -28,6 +28,7 @@ class FakeKVManager(BaseKVManager):
         is_mla_backend: Optional[bool] = False,
     ):
         super().__init__(args, disaggregation_mode, server_args, is_mla_backend)
+        self.kv_args = args
         self.req_to_decode_prefix_len = {}
 
     def register_to_bootstrap(self):
diff --git a/python/sglang/srt/layers/moe/mega_moe.py b/python/sglang/srt/layers/moe/mega_moe.py
index 93f4d9a15..1c13f7be9 100644
--- a/python/sglang/srt/layers/moe/mega_moe.py
+++ b/python/sglang/srt/layers/moe/mega_moe.py
@@ -206,8 +206,8 @@ def _run_mega_routed(
     )
 
     if num_tokens > 0:
-        topk_ids_in = topk_ids
-        topk_weights_in = topk_weights
+        topk_ids_in = topk_ids.to(torch.int32)
+        topk_weights_in = topk_weights.to(torch.float32)
     else:
         topk_ids_in = hidden_states.new_empty((0, top_k), dtype=torch.int32)
         topk_weights_in = hidden_states.new_empty((0, top_k), dtype=torch.float32)