[chore] clean-up weight prepack for INT8 MoE (#50116)

Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
This commit is contained in:
Fadi Arafeh
2026-07-31 10:15:38 +00:00
committed by GitHub
parent 88bc8fb56f
commit 0e9b50074e
4 changed files with 7 additions and 26 deletions
@@ -162,6 +162,7 @@ class CompressedTensorsW8A8Int8MoEMethod(CompressedTensorsMoEMethod):
routing_tables=layer._expert_routing_tables(),
layer=layer,
)
self.moe_kernel.fused_experts.process_weights_after_loading(layer)
def maybe_make_prepare_finalize(
self,
@@ -114,6 +114,7 @@ class Int8OnlineMoEMethod(OnlineMoEMethodBase):
routing_tables=layer._expert_routing_tables(),
layer=layer,
)
self.moe_kernel.fused_experts.process_weights_after_loading(layer)
def get_fused_moe_quant_config(
self, layer: torch.nn.Module