EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Use deepgemm instead of triton for fused_qkv_a_proj_with_mqa (#6890)

Browse Source
This commit is contained in:
fzyzcjy
2025-06-06 02:37:05 +08:00
committed by GitHub
parent dd1012fcbe
commit 35b65cf0ca
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
python/sglang/srt/layers/quantization/fp8_utils.py
View File

@@ -227,8 +227,8 @@ def deepgemm_w8a8_block_fp8_linear_with_fallback(
output_dtype = input.dtype
dtype_supported = output_dtype == torch.bfloat16
# TODO: add more robust shape check here
shape_supported = weight.shape[0] % 128 == 0 and weight.shape[1] % 128 == 0
# TODO: https://github.com/sgl-project/sglang/pull/6890#issuecomment-2943395737
shape_supported = weight.shape[0] % 64 == 0 and weight.shape[1] % 128 == 0
if not (shape_supported and dtype_supported):
# fall back to triton