change condition

Author: fsx950223

vllm/platforms/rocm.py

@@ -184,14 +184,14 @@ def get_attn_backend_cls(cls, selected_backend, head_size, dtype,
         selected_backend = (_Backend.ROCM_FLASH if selected_backend
                             == _Backend.FLASH_ATTN else selected_backend)
         if envs.VLLM_USE_V1:
-            if envs.VLLM_USE_TRITON_FLASH_ATTN:
-                logger.info("Using Triton Attention backend on V1 engine.")
-                return ("vllm.v1.attention.backends."
-                        "triton_attn.TritonAttentionBackend")
-            else:
+            if envs.VLLM_ROCM_USE_AITER:
                 logger.info("Using Flash Attention backend on V1 engine.")
                 return ("vllm.v1.attention.backends."
                         "flash_attn.FlashAttentionBackend")
+            else:
+                logger.info("Using Triton Attention backend on V1 engine.")
+                return ("vllm.v1.attention.backends."
+                        "triton_attn.TritonAttentionBackend")
         if selected_backend == _Backend.ROCM_FLASH:
             if not cls.has_device_capability(90):
                 # not Instinct series GPUs.