[MoE Refactor] Introduce RoutedExperts alias for FusedMoE and don't store SharedExperts in MK (#40735)

Signed-off-by: Bill Nell <bnell@redhat.com>
Signed-off-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Signed-off-by: Robert Shaw <robertgshaw2@gmail.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: Robert Shaw <robertgshaw2@gmail.com>
This commit is contained in:
bnellnm
2026-05-12 13:37:44 -04:00
committed by GitHub
co-authored by Robert Shaw Robert Shaw
parent 6ff7405b81
commit 4d591db470
53 changed files with 369 additions and 332 deletions
@@ -11,7 +11,7 @@ from vllm.config.quantization import (
)
from vllm.logger import init_logger
from vllm.model_executor.layers.fused_moe import (
FusedMoE,
RoutedExperts,
)
from vllm.model_executor.layers.fused_moe.unquantized_fused_moe_method import (
UnquantizedFusedMoEMethod,
@@ -118,7 +118,7 @@ class OnlineQuantizationConfig(QuantizationConfig):
return Mxfp8OnlineLinearMethod()
else:
return Fp8PerTensorOnlineLinearMethod()
elif isinstance(layer, FusedMoE):
elif isinstance(layer, RoutedExperts):
if should_ignore_layer(
prefix,
ignore=self.ignored_layers,