skip AiterInt8ScaledMMLinearKernel

vllm/model_executor/layers/quantization/kernels/scaled_mm/__init__.py

@@ -7,9 +7,9 @@ from typing import TypeVar
 import torch
 
 from vllm.logger import init_logger
-from vllm.model_executor.layers.quantization.kernels.scaled_mm.aiter import (
-    AiterInt8ScaledMMLinearKernel,
-)
+# from vllm.model_executor.layers.quantization.kernels.scaled_mm.aiter import (
+#     AiterInt8ScaledMMLinearKernel,
+# )
 from vllm.model_executor.layers.quantization.kernels.scaled_mm.cpu import (
     CPUInt8ScaledMMLinearKernel,
 )
@@ -51,7 +51,7 @@ _POSSIBLE_INT8_KERNELS: dict[PlatformEnum, list[type[Int8ScaledMMLinearKernel]]]
         CutlassInt8ScaledMMLinearKernel,
         TritonInt8ScaledMMLinearKernel,
     ],
-    PlatformEnum.ROCM: [AiterInt8ScaledMMLinearKernel, TritonInt8ScaledMMLinearKernel],
+    PlatformEnum.ROCM: [TritonInt8ScaledMMLinearKernel], # AiterInt8ScaledMMLinearKernel
 }
 
 # in priority/performance order (when available)