[feat] Add FlashInfer attention backend support for MiniMax H3 - #1832

Open

klhhhhh wants to merge 21 commits into

hao-ai-lab:mainfrom

klhhhhh:flashinfer-h3

Commits on Sep 8, 2026

[feat]support flashinfer attention kernel of prefill
klhhhhh
committed
dd53cdd
[feat]support flashinfer attention kernel of prefill
klhhhhh
committed
fb179b1
add flashinfer enum
klhhhhh
committed
5dcef83
check flashinfer env
klhhhhh
committed
37bfe59
add flashinfer backend in selector
klhhhhh
committed
b95b160
[test]add tests for flashinfer
klhhhhh
committed
ca39615
beanchmark for test
klhhhhh
committed
5868eef
delete benchmark
klhhhhh
committed
86c7c61
fix some bugs
klhhhhh
committed
a52292f
rebase
klhhhhh
committed
0ac9151
change version compatibility for cudnn kernel in flashinfer
klhhhhh
committed
e8db45c
add flashinfer implementation
klhhhhh
committed
05b2379
import cudnn kernels
klhhhhh
committed
ba0812e
add prefill backend for flashinfer
klhhhhh
committed
e8b1669
add cudnn test
klhhhhh
committed
78a7371
pre-commit
klhhhhh
committed
880a773
update docs for flashinfer kernel
klhhhhh
committed
01fcc8c
update test
klhhhhh
committed
ed03e31
add cudnn head size check
klhhhhh
committed
981a04b
update test
klhhhhh
committed
2ac7558
update attention backend for h3
klhhhhh
committed
0e2c9f7