[MoE Refactor] Make SharedExperts class for use with DefaultMoERunner (#35153)

Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
bnellnm
2026-04-01 09:44:08 -04:00
committed by GitHub
parent 5e30e9b9a9
commit 7cf56a59a2
34 changed files with 556 additions and 397 deletions
@@ -483,7 +483,7 @@ class BitsAndBytesMoEMethod(FusedMoEMethodBase):
topk_weights: torch.Tensor,
topk_ids: torch.Tensor,
shared_experts_input: torch.Tensor | None,
) -> torch.Tensor | tuple[torch.Tensor, torch.Tensor]:
) -> torch.Tensor:
from vllm.model_executor.layers.fused_moe import fused_experts
# TODO(bnell): Do these need to be called on the hot path?