[AMD] Fix MoRI EP warmup hang by restoring deepep_mode=normal default (#19498)

python/sglang/srt/server_args.py

@@ -2266,6 +2266,9 @@ class ServerArgs:
 
         if self.moe_a2a_backend == "mori":
             self.ep_size = self.tp_size
+            if self.deepep_mode == "auto":
+                self.deepep_mode = "normal"
+                logger.warning("auto set deepep_mode=`normal` for MORI EP")
             logger.warning(
                 f"MoRI MoE is enabled. The expert parallel size is adjusted to be the same as the tensor parallel size[{self.tp_size}]."
             )