feat: add direct routing strategy to DP worker (#6884)

python/sglang/srt/disaggregation/common/conn.py

@@ -109,10 +109,12 @@ class CommonKVReceiver(BaseKVReceiver):
         mgr: BaseKVManager,
         bootstrap_addr: str,
         bootstrap_room: Optional[int] = None,
+        data_parallel_rank: Optional[int] = None,
     ):
         self.bootstrap_room = bootstrap_room
         self.bootstrap_addr = bootstrap_addr
         self.kv_mgr = mgr
+        self.data_parallel_rank = data_parallel_rank
 
         if self.bootstrap_addr not in self.kv_mgr.prefill_dp_size_table:
             self.prefill_tp_size, self.prefill_dp_size = (
@@ -180,6 +182,10 @@ class CommonKVReceiver(BaseKVReceiver):
             self.target_tp_rank = self.target_tp_ranks[0]
             self.required_dst_info_num = 1
 
+        if self.data_parallel_rank is not None:
+            logger.debug(f"Targeting DP rank: {self.data_parallel_rank}")
+            self.target_dp_group = self.data_parallel_rank
+        else:
             self.target_dp_group = bootstrap_room % self.prefill_dp_size
 
         # NOTE: key distinguished by bootstrap_addr, target_dp_group, and target_tp_rank

python/sglang/srt/disaggregation/decode.py

@@ -156,6 +156,7 @@ class DecodePreallocQueue:
             mgr=self.kv_manager,
             bootstrap_addr=f"{req.bootstrap_host}:{req.bootstrap_port}",
             bootstrap_room=req.bootstrap_room,
+            data_parallel_rank=req.data_parallel_rank,
         )
         self.queue.append(DecodeRequest(req=req, kv_receiver=kv_receiver))
 

python/sglang/srt/disaggregation/fake/conn.py

@@ -56,6 +56,7 @@ class FakeKVReceiver(BaseKVReceiver):
         mgr: BaseKVManager,
         bootstrap_addr: str,
         bootstrap_room: Optional[int] = None,
+        data_parallel_rank: Optional[int] = None,
     ):
         self.has_init = False
 

python/sglang/srt/disaggregation/mooncake/conn.py

@@ -765,6 +765,7 @@ class MooncakeKVReceiver(BaseKVReceiver):
         mgr: MooncakeKVManager,
         bootstrap_addr: str,
         bootstrap_room: Optional[int] = None,
+        data_parallel_rank: Optional[int] = None,
     ):
         self.bootstrap_room = bootstrap_room
         self.bootstrap_addr = bootstrap_addr
@@ -772,6 +773,7 @@ class MooncakeKVReceiver(BaseKVReceiver):
         self.session_id = self.kv_mgr.get_session_id()
         self.kv_mgr.update_status(self.bootstrap_room, KVPoll.Bootstrapping)
         self.conclude_state = None
+        self.data_parallel_rank = data_parallel_rank
 
         if self.bootstrap_addr not in self.kv_mgr.prefill_dp_size_table:
             self.prefill_tp_size, self.prefill_dp_size = (
@@ -845,7 +847,11 @@ class MooncakeKVReceiver(BaseKVReceiver):
             self.target_tp_rank = self.target_tp_ranks[0]
             self.required_dst_info_num = 1
 
-        self.target_dp_group = self.bootstrap_room % self.prefill_dp_size
+        if self.data_parallel_rank is not None:
+            logger.debug(f"Targeting DP rank: {self.data_parallel_rank}")
+            self.target_dp_group = self.data_parallel_rank
+        else:
+            self.target_dp_group = bootstrap_room % self.prefill_dp_size
 
         # NOTE: key distinguished by bootstrap_addr, target_dp_group, and target_tp_rank
         bootstrap_key = (

python/sglang/srt/disaggregation/nixl/conn.py

@@ -407,9 +407,10 @@ class NixlKVReceiver(CommonKVReceiver):
         mgr: NixlKVManager,
         bootstrap_addr: str,
         bootstrap_room: Optional[int] = None,
+        data_parallel_rank: Optional[int] = None,
     ):
         self.started_transfer = False
-        super().__init__(mgr, bootstrap_addr, bootstrap_room)
+        super().__init__(mgr, bootstrap_addr, bootstrap_room, data_parallel_rank)
 
     def init(self, kv_indices: npt.NDArray[np.int64], aux_index: Optional[int] = None):
         for bootstrap_info in self.bootstrap_infos:

python/sglang/srt/entrypoints/EngineBase.py

@@ -23,6 +23,12 @@ class EngineBase(ABC):
         token_ids_logprob: Optional[Union[List[List[int]], List[int]]] = None,
         lora_path: Optional[Union[List[Optional[str]], Optional[str]]] = None,
         custom_logit_processor: Optional[Union[List[str], str]] = None,
+        return_hidden_states: Optional[bool] = None,
+        stream: Optional[bool] = None,
+        bootstrap_host: Optional[Union[List[str], str]] = None,
+        bootstrap_port: Optional[Union[List[int], int]] = None,
+        bootstrap_room: Optional[Union[List[int], int]] = None,
+        data_parallel_rank: Optional[int] = None,
     ) -> Union[Dict, Iterator[Dict]]:
         """Generate outputs based on given inputs."""
         pass

python/sglang/srt/entrypoints/engine.py

@@ -167,11 +167,22 @@ class Engine(EngineBase):
         bootstrap_host: Optional[Union[List[str], str]] = None,
         bootstrap_port: Optional[Union[List[int], int]] = None,
         bootstrap_room: Optional[Union[List[int], int]] = None,
+        data_parallel_rank: Optional[int] = None,
     ) -> Union[Dict, Iterator[Dict]]:
         """
         The arguments of this function is the same as `sglang/srt/managers/io_struct.py::GenerateReqInput`.
         Please refer to `GenerateReqInput` for the documentation.
         """
+        if self.server_args.enable_dp_attention:
+            if data_parallel_rank is None:
+                logger.info("data_parallel_rank not provided, using default dispatch")
+            elif data_parallel_rank < 0:
+                raise ValueError("data_parallel_rank must be non-negative")
+            elif data_parallel_rank >= self.server_args.dp_size:
+                raise ValueError(
+                    f"data_parallel_rank must be less than dp_size: {self.server_args.dp_size}"
+                )
+
         obj = GenerateReqInput(
             text=prompt,
             input_ids=input_ids,
@@ -188,6 +199,7 @@ class Engine(EngineBase):
             bootstrap_host=bootstrap_host,
             bootstrap_port=bootstrap_port,
             bootstrap_room=bootstrap_room,
+            data_parallel_rank=data_parallel_rank,
         )
         loop = asyncio.get_event_loop()
         generator = self.tokenizer_manager.generate_request(obj, None)
@@ -237,11 +249,24 @@ class Engine(EngineBase):
         bootstrap_host: Optional[Union[List[str], str]] = None,
         bootstrap_port: Optional[Union[List[int], int]] = None,
         bootstrap_room: Optional[Union[List[int], int]] = None,
+        data_parallel_rank: Optional[int] = None,
     ) -> Union[Dict, AsyncIterator[Dict]]:
         """
         The arguments of this function is the same as `sglang/srt/managers/io_struct.py::GenerateReqInput`.
         Please refer to `GenerateReqInput` for the documentation.
         """
+
+        if self.server_args.enable_dp_attention:
+            if data_parallel_rank is None:
+                logger.info("data_parallel_rank not provided, using default dispatch")
+            elif data_parallel_rank < 0:
+                raise ValueError("data_parallel_rank must be non-negative")
+            elif data_parallel_rank >= self.server_args.dp_size:
+                raise ValueError(
+                    f"data_parallel_rank must be in range [0, {self.server_args.dp_size-1}]"
+                )
+
+        logger.info(f"data_parallel_rank: {data_parallel_rank}")
         obj = GenerateReqInput(
             text=prompt,
             input_ids=input_ids,
@@ -257,6 +282,7 @@ class Engine(EngineBase):
             bootstrap_host=bootstrap_host,
             bootstrap_port=bootstrap_port,
             bootstrap_room=bootstrap_room,
+            data_parallel_rank=data_parallel_rank,
         )
         generator = self.tokenizer_manager.generate_request(obj, None)
 

python/sglang/srt/managers/data_parallel_controller.py

@@ -248,10 +248,18 @@ class DataParallelController:
 
     def round_robin_scheduler(self, req: Req):
         if self.server_args.disaggregation_mode == "null":
+            if req.data_parallel_rank is not None:
+                logger.debug(f"Direct routing to DP rank {req.data_parallel_rank}")
+                self.workers[req.data_parallel_rank].send_pyobj(req)
+            else:
                 self.workers[self.round_robin_counter].send_pyobj(req)
                 self.round_robin_counter = (self.round_robin_counter + 1) % len(
                     self.workers
                 )
+        else:
+            if req.data_parallel_rank is not None:
+                logger.debug(f"Direct routing to DP rank {req.data_parallel_rank}")
+                self.workers[req.data_parallel_rank].send_pyobj(req)
             else:
                 self.workers[req.bootstrap_room % len(self.workers)].send_pyobj(req)
 

python/sglang/srt/managers/io_struct.py

@@ -106,6 +106,9 @@ class GenerateReqInput:
     bootstrap_port: Optional[Union[List[Optional[int]], int]] = None
     bootstrap_room: Optional[Union[List[int], int]] = None
 
+    # For data parallel rank routing
+    data_parallel_rank: Optional[int] = None
+
     def contains_mm_input(self) -> bool:
         return has_valid_data(self.image_data) or has_valid_data(self.audio_data)
 
@@ -417,6 +420,9 @@ class GenerateReqInput:
             bootstrap_room=(
                 self.bootstrap_room[i] if self.bootstrap_room is not None else None
             ),
+            data_parallel_rank=(
+                self.data_parallel_rank if self.data_parallel_rank is not None else None
+            ),
         )
 
 
@@ -464,6 +470,9 @@ class TokenizedGenerateReqInput:
     bootstrap_port: Optional[int] = None
     bootstrap_room: Optional[int] = None
 
+    # For data parallel rank routing
+    data_parallel_rank: Optional[int] = None
+
 
 @dataclass
 class EmbeddingReqInput:

python/sglang/srt/managers/schedule_batch.py

@@ -451,6 +451,7 @@ class Req:
         bootstrap_host: Optional[str] = None,
         bootstrap_port: Optional[int] = None,
         bootstrap_room: Optional[int] = None,
+        data_parallel_rank: Optional[int] = None,
     ):
         # Input and output info
         self.rid = rid
@@ -605,6 +606,9 @@ class Req:
         self.bootstrap_room: Optional[int] = bootstrap_room
         self.disagg_kv_sender: Optional[BaseKVSender] = None
 
+        # For data parallel rank routing
+        self.data_parallel_rank: Optional[int] = data_parallel_rank
+
         # the start index of the sent kv cache
         # We want to send it chunk by chunk for chunked prefill.
         # After every chunk forward, we do the following:

python/sglang/srt/managers/scheduler.py

@@ -949,6 +949,7 @@ class Scheduler(
                 bootstrap_host=recv_req.bootstrap_host,
                 bootstrap_port=recv_req.bootstrap_port,
                 bootstrap_room=recv_req.bootstrap_room,
+                data_parallel_rank=recv_req.data_parallel_rank,
             )
             req.tokenizer = self.tokenizer
 

python/sglang/srt/managers/tokenizer_manager.py

@@ -570,6 +570,7 @@ class TokenizerManager:
                 session_params=session_params,
                 custom_logit_processor=obj.custom_logit_processor,
                 return_hidden_states=obj.return_hidden_states,
+                data_parallel_rank=obj.data_parallel_rank,
             )
         elif isinstance(obj, EmbeddingReqInput):
             tokenized_obj = TokenizedEmbeddingReqInput(