Revert "Support Multi Process Tokenizer Manager" (#8960)

python/sglang/srt/managers/scheduler.py

@@ -79,7 +79,6 @@ from sglang.srt.managers.io_struct import (
     InitWeightsUpdateGroupReqInput,
     LoadLoRAAdapterReqInput,
     LoadLoRAAdapterReqOutput,
-    MultiTokenizerRegisterReq,
     OpenSessionReqInput,
     OpenSessionReqOutput,
     ProfileReq,
@@ -248,6 +247,7 @@ class Scheduler(
         # Init inter-process communication
         context = zmq.Context(2)
         self.idle_sleeper = None
+
         if self.pp_rank == 0 and self.attn_tp_rank == 0:
             self.recv_from_tokenizer = get_zmq_socket(
                 context, zmq.PULL, port_args.scheduler_input_ipc_name, False
@@ -522,7 +522,6 @@ class Scheduler(
                 (ExpertDistributionReq, self.expert_distribution_handle),
                 (LoadLoRAAdapterReqInput, self.load_lora_adapter),
                 (UnloadLoRAAdapterReqInput, self.unload_lora_adapter),
-                (MultiTokenizerRegisterReq, self.register_multi_tokenizer),
             ]
         )
 
@@ -1065,8 +1064,6 @@ class Scheduler(
                     if self.recv_from_rpc is not None:
                         self.recv_from_rpc.send_pyobj(output)
                 else:
-                    if recv_req.rids is not None:
-                        output.rids = recv_req.rids
                     self.send_to_tokenizer.send_pyobj(output)
 
     def handle_generate_request(
@@ -2407,10 +2404,6 @@ class Scheduler(
         result = self.tp_worker.unload_lora_adapter(recv_req)
         return result
 
-    def register_multi_tokenizer(self, recv_req: MultiTokenizerRegisterReq):
-        self.send_to_detokenizer.send_pyobj(recv_req)
-        return recv_req
-
     def slow_down(self, recv_req: SlowDownReqInput):
         t = recv_req.forward_sleep_time
         if t is not None and t <= 0: