DeepEP normal support deepgemm-contiguous (#5626)

Co-authored-by: Yingyi Huang <yingyihuang2000@outlook.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
Co-authored-by: Xuting Zhou <xutingz@nvidia.com>
Co-authored-by: ZhengHSI <zhenghsi@qq.com>
This commit is contained in:
lukec
2025-05-08 16:20:32 +08:00
committed by GitHub
parent a05bd83a94
commit acc816d8a2
6 changed files with 568 additions and 59 deletions

View File

@@ -308,8 +308,8 @@ def sglang_per_token_group_quant_fp8(
device=x.device,
dtype=torch.float32,
)
sgl_per_token_group_quant_fp8(x, x_q, x_s, group_size, eps, fp8_min, fp8_max)
if x.shape[0] > 0:
sgl_per_token_group_quant_fp8(x, x_q, x_s, group_size, eps, fp8_min, fp8_max)
return x_q, x_s