[Bugfix] Fix ray instance detect issue (vllm-project#9439)

smallcloudai · Oct 28, 2024 · 2adb440 · 2adb440
1 parent feb92fb
commit 2adb440
Showing 1 changed file with 10 additions and 3 deletions.
diff --git a/vllm/executor/ray_utils.py b/vllm/executor/ray_utils.py
@@ -232,9 +232,16 @@ def initialize_ray_cluster(
 
     # Connect to a ray cluster.
     if current_platform.is_rocm() or current_platform.is_xpu():
-        ray.init(address=ray_address,
-                 ignore_reinit_error=True,
-                 num_gpus=parallel_config.world_size)
+        # Try to connect existing ray instance and create a new one if not found
+        try:
+            ray.init("auto")
+        except ConnectionError:
+            logger.warning(
+                "No existing RAY instance detected. "
+                "A new instance will be launched with current node resources.")
+            ray.init(address=ray_address,
+                     ignore_reinit_error=True,
+                     num_gpus=parallel_config.world_size)
     else:
         ray.init(address=ray_address, ignore_reinit_error=True)