diff --git a/vllm/model_executor/layers/fused_moe/fused_moe.py b/vllm/model_executor/layers/fused_moe/fused_moe.py index 5f9bfd6d9cf7d..d0f5eb498127b 100644 --- a/vllm/model_executor/layers/fused_moe/fused_moe.py +++ b/vllm/model_executor/layers/fused_moe/fused_moe.py @@ -818,6 +818,9 @@ def get_config_file_name( E: int, N: int, dtype: str | None, block_shape: list[int] | None = None ) -> str: device_name = current_platform.get_device_name().replace(" ", "_") + # Set device_name to H200 if a device from the H200 family is detected + if "H200" in device_name: + device_name = "H200" dtype_selector = "" if not dtype else f",dtype={dtype}" block_shape_selector = ( "" if not block_shape or not all(block_shape) else f",block_shape={block_shape}"