使用groupgemm完成高吞吐模式适配

python/sglang/srt/layers/moe/token_dispatcher/deepep.py

@@ -380,7 +380,7 @@ class _DeepEPDispatcherImplNormal(_DeepEPDispatcherImplBase):
             num_recv_tokens_per_expert,
             event,
         ) = self._dispatch_core(hidden_states, topk_ids, topk_weights, previous_event)
-        event.current_stream_wait() if self.async_finish else ()
+        # event.current_stream_wait() if self.async_finish else ()
 
         if isinstance(hidden_states, tuple):
             hidden_states, hidden_states_scale = hidden_states
@@ -500,7 +500,7 @@ class _DeepEPDispatcherImplNormal(_DeepEPDispatcherImplBase):
 
     def combine_b(self, output, previous_event):
         hidden_states, event = self._combine_core(output, previous_event)
-        event.current_stream_wait() if self.async_finish else ()
+        # event.current_stream_wait() if self.async_finish else ()
         self.handle = None
         self.src2dst = None
         return hidden_states