[feature]w8a8量化模型默认不开启aiter

vllm/envs.py

@@ -167,7 +167,7 @@ if TYPE_CHECKING:
     VLLM_MOE_USE_DEEP_GEMM: bool = True
     VLLM_USE_DEEP_GEMM_E8M0: bool = True
     VLLM_USE_DEEP_GEMM_TMA_ALIGNED_SCALES: bool = True
-    VLLM_USE_AITER_MOE_W8A8: bool = True
+    VLLM_USE_AITER_MOE_W8A8: bool = False
     VLLM_DEEP_GEMM_WARMUP: Literal[
         "skip",
         "full",
@@ -1292,7 +1292,7 @@ environment_variables: dict[str, Callable[[], Any]] = {
         int(os.getenv("VLLM_USE_DEEP_GEMM_TMA_ALIGNED_SCALES", "1"))
     ),
     "VLLM_USE_AITER_MOE_W8A8": lambda: bool(
-        int(os.getenv("VLLM_USE_AITER_MOE_W8A8", "1"))
+        int(os.getenv("VLLM_USE_AITER_MOE_W8A8", "0"))
     ),
     # DeepGemm JITs the kernels on-demand. The warmup attempts to make DeepGemm
     # JIT all the required kernels before model execution so there is no