minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[XPU] DeepSeek V4: use sgl-kernel-xpu implemetation of flash_mla_sparse_fwd for prefill (#31865)

Browse Source 
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
This commit is contained in:
Xuan Liao
2026-08-05 21:05:53 +08:00
committed by GitHub
co-authored by Ma Mingfei
parent 99709f734d
commit 3b4fac5b99
1 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/layers/attention/deepseek_v4_backend.py
+3
View File
@@ -1780,6 +1780,9 @@ class DeepseekV4AttnBackend(
indices. Chunk-invariant scaffolding lives in indices. Chunk-invariant scaffolding lives in
``self.forward_metadata.sparse_prefill_cache``. ``self.forward_metadata.sparse_prefill_cache``.
""" """
if _is_xpu:
from sgl_kernel import flash_mla_sparse_fwd
else:
from sgl_kernel.flash_mla import flash_mla_sparse_fwd from sgl_kernel.flash_mla import flash_mla_sparse_fwd
# q is (b, 1, h_q, d_qk); flash_mla_sparse_fwd takes (s_q, h_q, d_qk). # q is (b, 1, h_q, d_qk); flash_mla_sparse_fwd takes (s_q, h_q, d_qk).