[Refactor] Move FusedMoE hidden_size roundup to quant_method (#34285)

Signed-off-by: Bowen Bao <bowenbao@amd.com>
This commit is contained in:
Bowen Bao
2026-03-26 23:38:26 -07:00
committed by GitHub
parent c2b17d71af
commit 0ae89f18fd
12 changed files with 204 additions and 222 deletions
@@ -672,8 +672,6 @@ class Fp8MoEMethod(FusedMoEMethodBase):
params_dtype: torch.dtype,
**extra_weight_attrs,
):
layer.intermediate_size_per_partition = intermediate_size_per_partition
layer.hidden_size = hidden_size
layer.num_experts = num_experts
layer.orig_dtype = params_dtype
layer.weight_block_size = None
@@ -1011,8 +1009,6 @@ class Fp8OnlineMoEMethod(Fp8MoEMethod):
params_dtype: torch.dtype,
**extra_weight_attrs,
):
layer.intermediate_size_per_partition = intermediate_size_per_partition
layer.hidden_size = hidden_size
layer.num_experts = num_experts
layer.orig_dtype = params_dtype
layer.weight_block_size = None