biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Easy] Eliminate c10::optional usage in vllm/csrc (#17819)

Browse Source
This commit is contained in:
Lu Fang
2025-05-08 03:05:10 -07:00
committed by GitHub
parent a1e19b635d
commit f50dcb7c21
4 changed files with 7 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
csrc/quantization/gptq_allspark/allspark_repack.cu
View File

@@ -100,9 +100,9 @@ void rearrange_kn_weight_as_n32k16_order_ldg16(
void rearrange_kn_weight_as_n32k16_order(
torch::Tensor const& b_qweight, torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& b_zeros, bool has_zp,
std::optional<torch::Tensor> const& b_zeros, bool has_zp,
torch::Tensor& b_qweight_reorder, torch::Tensor& b_scales_reorder,
c10::optional<torch::Tensor> const& b_zeros_reorder, const int64_t K,
std::optional<torch::Tensor> const& b_zeros_reorder, const int64_t K,
const int64_t N, const int64_t N_32align) {
// Verify device and strides
TORCH_CHECK(b_qweight.device().is_cuda(), "b_qweight is not on GPU");