[MLA Attention Backend] Add TOKENSPEED_MLA backend for DSR1/Kimi K25 prefill + decode on Blackwell (#41778)

Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Roger Wang <hey@rogerw.io>
This commit is contained in:
Yongye Zhu
2026-05-13 23:48:02 -07:00
committed by GitHub
co-authored by Roger Wang
parent fd7d858c8a
commit 0d2732dd91
14 changed files with 639 additions and 88 deletions
@@ -1362,6 +1362,7 @@ def backend_supports_prefill_query_quantization() -> bool:
return backend_cls.get_name() in (
"FLASHINFER",
"TRTLLM_RAGGED",
"TOKENSPEED_MLA",
)