diff --git a/vllm/compilation/pass_manager.py b/vllm/compilation/pass_manager.py index 54f00d541521..e07e52be9fdf 100644 --- a/vllm/compilation/pass_manager.py +++ b/vllm/compilation/pass_manager.py @@ -7,11 +7,13 @@ from vllm.config import VllmConfig from vllm.logger import init_logger from vllm.platforms import current_platform -if current_platform.is_cuda(): +if current_platform.is_cuda_alike(): from .fusion import FusionPass - from .collective_fusion import AllReduceFusionPass, AsyncTPPass from .fusion_attn import AttnFusionPass +if current_platform.is_cuda(): + from .collective_fusion import AllReduceFusionPass, AsyncTPPass + from .activation_quant_fusion import ActivationQuantFusionPass from .fix_functionalization import FixFunctionalizationPass from .inductor_pass import CustomGraphPass, InductorPass, get_pass_context