[Refactor] Move FusedMoE hidden_size roundup to quant_method (#34285)
Signed-off-by: Bowen Bao <bowenbao@amd.com>
This commit is contained in:
@@ -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
|
||||
|
||||
Reference in New Issue
Block a user