[MoE Refactor] Combine CompressedTensorsWNA16MarlinMoEMethod with CompressedTensorsWNA16MoEMethod (#44570)

Signed-off-by: Bill Nell <bnell@redhat.com>
Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>
Signed-off-by: Dobrzyniewicz, Agata <agata.dobrzyniewicz@intel.com>
Signed-off-by: Tushar Jain <tushar00jain@users.noreply.github.com>
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>
Co-authored-by: Agata Dobrzyniewicz <160237065+adobrzyn@users.noreply.github.com>
Co-authored-by: Tushar Jain <8455015+tushar00jain@users.noreply.github.com>
Co-authored-by: Tushar Jain <tushar00jain@users.noreply.github.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: Jiangyun Zhu <riverclouds.zhu@qq.com>
This commit is contained in:
bnellnm
2026-07-31 14:13:03 -07:00
committed by GitHub
co-authored by Claude Taneem Ibrahim Agata Dobrzyniewicz Tushar Jain Tushar Jain Wentao Ye Jiangyun Zhu
parent 0c4dc7c488
commit 454ea5b526
14 changed files with 545 additions and 855 deletions
@@ -490,8 +490,9 @@ class AutoGPTQMoEMethod(FusedMoEMethodBase):
moe,
weight_key,
quant_config=self.quant_config,
may_have_zp=True,
may_have_zp=not self.quant_config.is_sym,
may_have_bias=True,
allow_tile_padding=not self.quant_config.desc_act,
)
def create_weights(
@@ -806,16 +807,6 @@ class AutoGPTQMoEMethod(FusedMoEMethodBase):
w2_bias=getattr(layer, "w2_bias", None),
)
def select_gemm_impl(
self,
prepare_finalize,
layer: RoutedExperts,
):
raise ValueError(
f"{self.__class__.__name__} uses the new modular kernel "
"initialization logic. This function should not be called."
)
def apply(
self,
layer: RoutedExperts,