use ray for hpu distributed inference

HabanaAI · Sep 23, 2024 · 3bb593a · 3bb593a
1 parent e00ab5a
commit 3bb593a
Showing 1 changed file with 7 additions and 0 deletions.
diff --git a/vllm/config.py b/vllm/config.py
@@ -855,6 +855,13 @@ def __init__(
                 raise ValueError(
                     "TPU backend only supports Ray for distributed inference.")
 
+        if current_platform.is_hpu() and self.world_size > 1:
+            if self.distributed_executor_backend is None:
+                self.distributed_executor_backend = "ray"
+            if self.distributed_executor_backend != "ray":
+                raise ValueError(
+                    "HPU backend only supports Ray for distributed inference.")
+
         if self.distributed_executor_backend is None and self.world_size > 1:
             # We use multiprocessing by default if world_size fits on the
             # current node and we aren't in a ray placement group.