[Refactor] Remove dead cutlass mxfp8 code (#44681)

Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Shengqi Chen <harry-chen@outlook.com>
This commit is contained in:
Wentao Ye
2026-06-17 21:18:25 -07:00
committed by GitHub
co-authored by Shengqi Chen
parent f428718ffe
commit b4c80ec0fd
10 changed files with 0 additions and 1369 deletions
-70
View File
@@ -1136,76 +1136,6 @@ def cutlass_mxfp4_moe_mm(
)
def mxfp8_experts_quant(
input_tensor: torch.Tensor,
problem_sizes: torch.Tensor,
expert_offsets: torch.Tensor,
blockscale_offsets: torch.Tensor,
quant_output: torch.Tensor,
scale_factor: torch.Tensor,
) -> None:
torch.ops._C.mxfp8_experts_quant(
input_tensor,
problem_sizes,
expert_offsets,
blockscale_offsets,
quant_output,
scale_factor,
)
def cutlass_mxfp8_grouped_mm(
a_tensors: torch.Tensor,
b_tensors: torch.Tensor,
a_scales: torch.Tensor,
b_scales: torch.Tensor,
out_tensors: torch.Tensor,
problem_sizes: torch.Tensor,
expert_offsets: torch.Tensor,
blockscale_offsets: torch.Tensor,
) -> None:
torch.ops._C.cutlass_mxfp8_grouped_mm(
a_tensors,
b_tensors,
a_scales,
b_scales,
out_tensors,
problem_sizes,
expert_offsets,
blockscale_offsets,
)
if hasattr(torch.ops._C, "mxfp8_experts_quant"):
@register_fake("_C::mxfp8_experts_quant")
def _mxfp8_experts_quant_fake(
input_tensor: torch.Tensor,
problem_sizes: torch.Tensor,
expert_offsets: torch.Tensor,
blockscale_offsets: torch.Tensor,
quant_output: torch.Tensor,
scale_factor: torch.Tensor,
) -> None:
return None
if hasattr(torch.ops._C, "cutlass_mxfp8_grouped_mm"):
@register_fake("_C::cutlass_mxfp8_grouped_mm")
def _cutlass_mxfp8_grouped_mm_fake(
a_tensors: torch.Tensor,
b_tensors: torch.Tensor,
a_scales: torch.Tensor,
b_scales: torch.Tensor,
out_tensors: torch.Tensor,
problem_sizes: torch.Tensor,
expert_offsets: torch.Tensor,
blockscale_offsets: torch.Tensor,
) -> None:
return None
# gptq_marlin
def gptq_marlin_repack(
b_q_weight: torch.Tensor,