Skip to content

Pull requests: HazyResearch/ThunderKittens

New pull request New
21 Open 81 Closed
21 Open 81 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

m3 + benchmark harness
#195 opened Apr 20, 2026 by Lazarus-931 Loading…
1
fp8 shared vector support (e4m3/e5m2/e8m0)
#193 opened Apr 9, 2026 by Emre-Dinc Loading…
1
fix MMA Pipeline theoretical race
#190 opened Mar 29, 2026 by Topologized Loading…
Possible solution of using DLPack to handle TVM FFI
#181 opened Mar 17, 2026 by haoran35-jpg Loading…
Fix narrowing conversion from char to signed char in packing #116
#179 opened Mar 10, 2026 by florianmattana Loading…
Fixes H100 attn & matmul kernel
#155 opened Oct 28, 2025 by symlons Loading…
added a docker container (only try on GPU wont work on CPU)
#146 opened Aug 24, 2025 by Kushagra481 Loading…
Fix unit test Makefile bugs for A100 and 4090 results in linker failure due to cuda libraries missing
#145 opened Aug 20, 2025 by tomflinda Contributor Loading…
Document and simplify fp16 -> fp8 conversion
#140 opened Aug 3, 2025 by melonedo Loading…
Add Implementation of Native Sparse Attention
#137 opened Jul 22, 2025 by yukavio Loading…
implement group gemm for contiguous case
#136 opened Jul 22, 2025 by XiaobingSuper Contributor Loading…
5
Added LCSF template implementation support for FlashAttention Backward.
#135 opened Jul 22, 2025 by KuangjuX Loading…
2 tasks done
5
Remove redundant register declarations.
#134 opened Jul 22, 2025 by KuangjuX Loading…
Remove unnecessary device and stream syncs
#129 opened Jun 15, 2025 by Edenzzzz Loading…
10
Use -gencode instead of -arch in mla_decode
#126 opened Jun 6, 2025 by lucifer1004 Loading…
1
40% build speedup: get torch and python include paths without subprocesses
#119 opened May 7, 2025 by technillogue Loading…
Gpt2
#103 opened Mar 17, 2025 by oleitersdorf Loading…
1
global_to_shared.cuh row accessing fixes
#95 opened Feb 21, 2025 by dylanllim Collaborator Loading…
batch matrix multiply kernel
#94 opened Feb 19, 2025 by technillogue • Draft
Do not pass l_vec and o to mha_forward in h100_bench
#92 opened Feb 19, 2025 by acforvs Loading…
1
I created a website to write documentation!
#89 opened Feb 14, 2025 by prateekshukla1108 Loading…
ProTip! Mix and match filters to narrow down what you’re looking for.