Tiny assert no running requests when releasing memory to avoid IMA (#12341)

python/sglang/srt/managers/scheduler_update_weights_mixin.py

@@ -106,6 +106,10 @@ class SchedulerUpdateWeightsMixin:
     def release_memory_occupation(
         self: Scheduler, recv_req: ReleaseMemoryOccupationReqInput
     ):
+        assert (
+            self._is_no_request()
+        ), "release_memory_occupation should be called only when no ongoing request."
+
         tags = recv_req.tags
 
         if tags is None or len(tags) == 0: