Fix vllm-project#2540 (vllm-project#2545)

hongxiayang · Jan 22, 2024 · b8b0b59 · b8b0b59
1 parent 5e0cfd0
commit b8b0b59
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/vllm/engine/llm_engine.py b/vllm/engine/llm_engine.py
@@ -209,7 +209,7 @@ def _init_workers_ray(self, placement_group: "PlacementGroup",
             worker.set_cuda_visible_devices.remote(node_gpus[node_id])
 
         distributed_init_method = get_distributed_init_method(
-            driver_ip, get_open_port)
+            driver_ip, get_open_port())
 
         # Lazy import the Worker to avoid importing torch.cuda/xformers
         # before CUDA_VISIBLE_DEVICES is set in the Worker