Disable custom all reduce by default (vllm-project#2808)

Author: WoosukKwon

vllm/config.py

@@ -423,16 +423,26 @@ def _verify_args(self) -> None:
         if self.pipeline_parallel_size > 1:
             raise NotImplementedError(
                 "Pipeline parallelism is not supported yet.")
-        if is_hip():
+        if not self.disable_custom_all_reduce and self.world_size > 1:
+            if is_hip():
+                self.disable_custom_all_reduce = True
+                logger.info(
+                    "Disabled the custom all-reduce kernel because it is not "
+                    "supported on AMD GPUs.")
+            elif self.pipeline_parallel_size > 1:
+                self.disable_custom_all_reduce = True
+                logger.info(
+                    "Disabled the custom all-reduce kernel because it is not "
+                    "supported with pipeline parallelism.")
+
+        # FIXME(woosuk): Fix the stability issues and re-enable the custom
+        # all-reduce kernel.
+        if not self.disable_custom_all_reduce and self.world_size > 1:
             self.disable_custom_all_reduce = True
             logger.info(
-                "Disabled the custom all-reduce kernel because it is not "
-                "supported on AMD GPUs.")
-        elif self.pipeline_parallel_size > 1:
-            self.disable_custom_all_reduce = True
-            logger.info(
-                "Disabled the custom all-reduce kernel because it is not "
-                "supported with pipeline parallelism.")
+                "Custom all-reduce kernels are temporarily disabled due to "
+                "stability issues. We will re-enable them once the issues are "
+                "resolved.")
 
 
 class SchedulerConfig: