Morph MORPH
SIGN IN SIGN UP
flashinfer-ai / flashinfer UNCLAIMED

FlashInfer: Kernel Library for LLM Serving

0 0 0 Python

ACTIVITY

Commits Reviews
week month year
COMMITS
20
in the last week
CONTRIBUTORS
16
active
STARS
0
total
FORKS
0
total
TOP CONTRIBUTORS
B
Brian K. Ryu
3 commits
S
Sam (Kesen Li)
2 commits
A
ameynaik-hub
2 commits
O
ooo oo
1 commit
B
Ben Glick
1 commit
W
Wei Zhao
1 commit
C
ChristinaZ
1 commit
I
Igor Shovkun
1 commit
A
Andrii Skliar
1 commit
P
Perkz Zheng
1 commit
RECENT COMMITS
O
docs: document replay command in CLI reference (#2919)
ooo oo
3072fe3
B
fix: Update NVSHMEM interface to use NVSHMEM4Py instead of custom bindings (#2960)
Ben Glick
19b825d
W
fix: Add filelock to ensure_symlink (#2979)
Wei Zhao
01729d6
S
fix: snap weight_scale_vec_size to handle block_scale_interleave padding for SM120 (#2898)
Sam (Kesen Li)
c4cb6e0
C
Refactor the routing part (#2803)
ChristinaZ
c26953f
B
fix: use float instead of double in sampling binary search to avoid FP64 bottleneck on SM103 (#2945)
Brian K. Ryu
19329d8
I
feat: Mamba SSU: horizontal MTP kernel (+ DSTATE=96 support) (#2865)
Igor Shovkun
15f11e9
A
feat: SM121 (GB10) tile filtering and autotuner robustness (#2927)
Andrii Skliar
ee3ca01
P
fix: [Fmha] revert blackwell ultra optimization that causes deadlocks. (#2956)
Perkz Zheng
5758837
B
test: skip unsupported mm_mxfp8 configurations on SM12x (#2974)
Brian K. Ryu
582bd7b
J
test: xfail cuDNN FP8 prefill on Blackwell with CUDA <= 12.9 (#2963)
Jonathan Dierksen
fc5441a
Z
fix: avoid re-downloading BMM export headers when flashinfer-cubin is installed (#2903)
Zihao Ye
e0f3729
S
feat: add MXFP8 GEMM support for SM120 (#2902)
Sam (Kesen Li)
fe05393
A
perf: Optimize GDN MTP decode kernel (v15) — eliminate ilp=1 fallback… (#2842)
ameynaik-hub
fc08cd1
Y
ci: remove 1gpu label from H100 runner selector (#2946)
Yong Wu
34b6550
A
feat(gdn): add BF16 state kernel with MTP support beyond T>4 with intermediate caching. (#2679)
ameynaik-hub
7514bf2
Y
feat: implement deterministic topk (#2661)
Yinzuo Jiang
637209a
Y
fix: vectorize get_shuffle_matrix_a_row_indices to eliminate CPU contention (#2935)
youkaichao
5a906be
B
perf: Optimize CuTe-DSL fp4 and fp8 quantization kernels (#2904)
Brian K. Ryu
d476c61
S
read real strides for kv and block scale (#2844)
sychen52
7924dc0