[Fix] Add --disable-draft-model-update to control draft model updates(especially in RL) (#15726)

Co-authored-by: leihaodi <haodilei@gmail.com>

python/sglang/srt/managers/io_struct.py

@@ -1337,6 +1337,8 @@ class UpdateWeightsFromTensorReqInput(BaseReq):
     abort_all_requests: bool = False
     # Optional: Update weight version along with weights
     weight_version: Optional[str] = None
+    # Optional: Determine whether to disable updating the draft model
+    disable_draft_model: Optional[bool] = None
 
 
 @dataclass

python/sglang/srt/managers/scheduler_update_weights_mixin.py

@@ -88,7 +88,10 @@ class SchedulerUpdateWeightsMixin:
         self: Scheduler, recv_req: UpdateWeightsFromTensorReqInput
     ):
         """Update the online model parameter from tensors."""
-        worker = self.draft_worker or self.tp_worker
+        if recv_req.disable_draft_model:
+            worker = self.tp_worker
+        else:
+            worker = self.draft_worker or self.tp_worker
         success, message = worker.update_weights_from_tensor(recv_req)
         # TODO extract common code b/t update_weights_from_distributed and update_weights_from_tensor later
         if success: