fix ray spmd

Signed-off-by: youkaichao <[email protected]>

vllm/executor/ray_gpu_executor.py

@@ -190,6 +190,9 @@ def sort_by_driver_then_worker_ip(worker):
         # Get the set of GPU IDs used on each node.
         worker_node_and_gpu_ids = []
         for worker in [self.driver_dummy_worker] + self.workers:
+            if worker is None:
+                # driver_dummy_worker can be None when using ray spmd worker.
+                continue
             worker_node_and_gpu_ids.append(
                 ray.get(worker.get_node_and_gpu_ids.remote()) \
             ) # type: ignore

vllm/executor/ray_hpu_executor.py

@@ -165,6 +165,9 @@ def sort_by_driver_then_worker_ip(worker):
 
         worker_node_and_gpu_ids = []
         for worker in [self.driver_dummy_worker] + self.workers:
+            if worker is None:
+                # driver_dummy_worker can be None when using ray spmd worker.
+                continue
             worker_node_and_gpu_ids.append(
                 ray.get(worker.get_node_and_gpu_ids.remote()) \
             ) # type: ignore

vllm/executor/ray_tpu_executor.py

@@ -139,6 +139,9 @@ def sort_by_driver_then_worker_ip(worker):
         # Get the set of TPU IDs used on each node.
         worker_node_and_gpu_ids = []
         for worker in [self.driver_dummy_worker] + self.workers:
+            if worker is None:
+                # driver_dummy_worker can be None when using ray spmd worker.
+                continue
             worker_node_and_gpu_ids.append(
                 ray.get(worker.get_node_and_gpu_ids.remote()) \
             ) # type: ignore

vllm/executor/ray_xpu_executor.py

@@ -18,6 +18,9 @@ def _get_env_vars_to_be_updated(self):
         # Get the set of GPU IDs used on each node.
         worker_node_and_gpu_ids = []
         for worker in [self.driver_dummy_worker] + self.workers:
+            if worker is None:
+                # driver_dummy_worker can be None when using ray spmd worker.
+                continue
             worker_node_and_gpu_ids.append(
                 ray.get(worker.get_node_and_gpu_ids.remote()))  # type: ignore