fix MultiTokenizerWrapper name (#10049)

Signed-off-by: huanglong <huanglong@linux.alibaba.com>

python/sglang/srt/managers/io_struct.py

@@ -1195,7 +1195,7 @@ class MultiTokenizerRegisterReq:
 
 
 @dataclass
-class MultiTokenizerWarpper:
+class MultiTokenizerWrapper:
     worker_id: int
     obj: Optional[Any] = None
 

python/sglang/srt/managers/multi_tokenizer_mixin.py

@@ -34,7 +34,7 @@ from sglang.srt.managers.io_struct import (
     BatchStrOut,
     BatchTokenIDOut,
     MultiTokenizerRegisterReq,
-    MultiTokenizerWarpper,
+    MultiTokenizerWrapper,
 )
 from sglang.srt.managers.tokenizer_manager import TokenizerManager, _Communicator
 from sglang.srt.server_args import PortArgs, ServerArgs
@@ -441,7 +441,7 @@ class MultiTokenizerRouter(TokenizerManager, MultiTokenizerMixin):
 
     async def _distribute_result_to_workers(self, recv_obj):
         """Distribute result to corresponding workers based on rid"""
-        if isinstance(recv_obj, MultiTokenizerWarpper):
+        if isinstance(recv_obj, MultiTokenizerWrapper):
             worker_ids = [recv_obj.worker_id]
             recv_obj = recv_obj.obj
         else:

python/sglang/srt/managers/scheduler.py

@@ -85,7 +85,7 @@ from sglang.srt.managers.io_struct import (
     LoadLoRAAdapterReqInput,
     LoadLoRAAdapterReqOutput,
     MultiTokenizerRegisterReq,
-    MultiTokenizerWarpper,
+    MultiTokenizerWrapper,
     OpenSessionReqInput,
     OpenSessionReqOutput,
     ProfileReq,
@@ -1096,13 +1096,13 @@ class Scheduler(
                     self.send_to_tokenizer.send_pyobj(abort_req)
                     continue
 
-            # If it is a MultiTokenizerWarpper, unwrap it and handle the inner request.
-            if isinstance(recv_req, MultiTokenizerWarpper):
+            # If it is a MultiTokenizerWrapper, unwrap it and handle the inner request.
+            if isinstance(recv_req, MultiTokenizerWrapper):
                 worker_id = recv_req.worker_id
                 recv_req = recv_req.obj
                 output = self._request_dispatcher(recv_req)
                 if output is not None:
-                    output = MultiTokenizerWarpper(worker_id, output)
+                    output = MultiTokenizerWrapper(worker_id, output)
                     self.send_to_tokenizer.send_pyobj(output)
                 continue
 

python/sglang/srt/managers/tokenizer_manager.py

@@ -94,7 +94,7 @@ from sglang.srt.managers.io_struct import (
     LoadLoRAAdapterReqInput,
     LoadLoRAAdapterReqOutput,
     LoRAUpdateResult,
-    MultiTokenizerWarpper,
+    MultiTokenizerWrapper,
     OpenSessionReqInput,
     OpenSessionReqOutput,
     ProfileReq,
@@ -1118,7 +1118,7 @@ class TokenizerManager:
         self, obj: UpdateWeightFromDiskReqInput
     ) -> Tuple[bool, str]:
         if self.server_args.tokenizer_worker_num > 1:
-            obj = MultiTokenizerWarpper(self.worker_id, obj)
+            obj = MultiTokenizerWrapper(self.worker_id, obj)
         self.send_to_scheduler.send_pyobj(obj)
         self.model_update_result = asyncio.Future()
         if self.server_args.dp_size == 1:
@@ -1339,7 +1339,7 @@ class TokenizerManager:
             return None
 
         if self.server_args.tokenizer_worker_num > 1:
-            obj = MultiTokenizerWarpper(self.worker_id, obj)
+            obj = MultiTokenizerWrapper(self.worker_id, obj)
         self.send_to_scheduler.send_pyobj(obj)
 
         self.session_futures[obj.session_id] = asyncio.Future()
@@ -2165,7 +2165,7 @@ class _Communicator(Generic[T]):
 
         if obj:
             if _Communicator.enable_multi_tokenizer:
-                obj = MultiTokenizerWarpper(worker_id=os.getpid(), obj=obj)
+                obj = MultiTokenizerWrapper(worker_id=os.getpid(), obj=obj)
             self._sender.send_pyobj(obj)
 
         self._result_event = asyncio.Event()