[BugFix] fix num_lookahead_slots missing in async executor (#4165)

Co-authored-by: Lei Wen <wenlei03@qiyi.com>

vllm/executor/cpu_executor.py

@@ -109,12 +109,14 @@ class CPUExecutorAsync(CPUExecutor, ExecutorAsyncBase):
         blocks_to_swap_in: Dict[int, int],
         blocks_to_swap_out: Dict[int, int],
         blocks_to_copy: Dict[int, List[int]],
+        num_lookahead_slots: int,
     ) -> List[SamplerOutput]:
         output = await make_async(self.driver_worker.execute_model)(
             seq_group_metadata_list=seq_group_metadata_list,
             blocks_to_swap_in=blocks_to_swap_in,
             blocks_to_swap_out=blocks_to_swap_out,
-            blocks_to_copy=blocks_to_copy)
+            blocks_to_copy=blocks_to_copy,
+            num_lookahead_slots=num_lookahead_slots)
         return output
 
     async def check_health_async(self) -> None:

vllm/executor/executor_base.py

@@ -112,6 +112,7 @@ class ExecutorAsyncBase(ExecutorBase):
         blocks_to_swap_in: Dict[int, int],
         blocks_to_swap_out: Dict[int, int],
         blocks_to_copy: Dict[int, List[int]],
+        num_lookahead_slots: int,
     ) -> List[SamplerOutput]:
         """Executes one model step on the given sequences."""
         raise NotImplementedError

vllm/executor/gpu_executor.py

@@ -163,10 +163,12 @@ class GPUExecutorAsync(GPUExecutor, ExecutorAsyncBase):
         blocks_to_swap_in: Dict[int, int],
         blocks_to_swap_out: Dict[int, int],
         blocks_to_copy: Dict[int, List[int]],
+        num_lookahead_slots: int,
     ) -> List[SamplerOutput]:
         output = await make_async(self.driver_worker.execute_model)(
             seq_group_metadata_list=seq_group_metadata_list,
             blocks_to_swap_in=blocks_to_swap_in,
             blocks_to_swap_out=blocks_to_swap_out,
-            blocks_to_copy=blocks_to_copy)
+            blocks_to_copy=blocks_to_copy,
+            num_lookahead_slots=num_lookahead_slots)
         return output

vllm/executor/neuron_executor.py

@@ -84,6 +84,7 @@ class NeuronExecutorAsync(NeuronExecutor, ExecutorAsyncBase):
         blocks_to_swap_in: Dict[int, int],
         blocks_to_swap_out: Dict[int, int],
         blocks_to_copy: Dict[int, List[int]],
+        num_lookahead_slots: int,
     ) -> List[SamplerOutput]:
         output = await make_async(self.driver_worker.execute_model)(
             seq_group_metadata_list=seq_group_metadata_list, )

vllm/executor/ray_gpu_executor.py

@@ -196,6 +196,7 @@ class RayGPUExecutor(DistributedGPUExecutor):
                 "blocks_to_swap_in": blocks_to_swap_in,
                 "blocks_to_swap_out": blocks_to_swap_out,
                 "blocks_to_copy": blocks_to_copy,
+                "num_lookahead_slots": num_lookahead_slots,
             },
             use_ray_compiled_dag=USE_RAY_COMPILED_DAG)