Skip to content

Pull requests: Dao-AILab/flash-attention

New pull request New
140 Open 532 Closed
140 Open 532 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

Add compress_factor for compressed causal attention
#2418 opened Mar 31, 2026 by jduprat Loading…
4
Fix SM120 forward pass crash: parent __init__ overwrites arch, enabling unsupported TMA path
#2416 opened Mar 31, 2026 by moghon92 Loading…
[Cute,Fwd,Sm90] Support SplitKV
#2415 opened Mar 31, 2026 by imbr92 Loading…
fix noisy logger
#2414 opened Mar 31, 2026 by drisspg Loading…
Feat([FA4][CUTE DSL]) Add head_dim=256 support (forward + backward)
#2412 opened Mar 30, 2026 by wangsiyu Loading…
10
chore(tests): move benchmarks to benchmarks/cute/ and reduce test prints
#2408 opened Mar 29, 2026 by NJX-njx Loading…
3 tasks
1
1
fix(flash_fwd_sm90): zero partial V smem to prevent 0*NaN=NaN in PV GEMM
#2407 opened Mar 29, 2026 by NJX-njx Loading…
3 tasks
1
[CuTe, SM120] Forward kernel with optimized TMA path and full features support
#2406 opened Mar 28, 2026 by sisgrad Loading…
1
fix(cute): SM120 forward/bwd and atomic add compatibility
#2404 opened Mar 28, 2026 by NJX-njx Loading…
1
1
feat: setup_context for FlashAttnFunc (torch.func.grad)
#2405 opened Mar 28, 2026 by NJX-njx Loading…
1
4
build(windows): MSVC conforming preprocessor for CUDA 13+ and ninja warning
#2403 opened Mar 28, 2026 by NJX-njx Loading…
1
1
feat(cute): implement softcap backward pass, correct math formula, and resolve JIT cache bug
#2402 opened Mar 28, 2026 by CaesarG Loading…
1
5
Add FA4 CI: GitHub Actions workflow with Apptainer on B200 runner
#2393 opened Mar 25, 2026 by Johnsonms Loading…
18
[Cute,Sm80,Fwd] Guard kernel body with tile validity check to fix varlen illegal memory access
#2391 opened Mar 25, 2026 by zhuochenKIDD Loading…
1
Add SM80/SM120 block-sparse forward attention support
#2389 opened Mar 25, 2026 by blake-snc Loading…
6
[ROCM] Fix windows issues
#2385 opened Mar 23, 2026 by micmelesse Loading…
30
Fix missing seqlen_info param in softcap scoremod
#2366 opened Mar 17, 2026 by rucnyz Loading…
[ROCM] Migrate CK backend to Aiter
#2364 opened Mar 17, 2026 by micmelesse • Draft
1
[Cute] Fix multi-GPU support: add torch.cuda.device guards for kernel launches
#2356 opened Mar 16, 2026 by NJX-njx Loading…
2 tasks
1
4
[CuTe, Sm100] PackGQA for backward
#2354 opened Mar 15, 2026 by reubenconducts Loading…
5
Fix RuntimeError when out_ is specified in _flash_attn_forward (fix #2073)
#2351 opened Mar 14, 2026 by NJX-njx Loading…
1
2
rocm/ck: switch FlashAttention to the CK vendored via aiter and fix current CK API integration
#2350 opened Mar 14, 2026 by eliasmagn Loading…
4
Add SM120 TMA forward kernel with warp specialization
#2349 opened Mar 14, 2026 by blake-snc Loading…
1
Add SM120 kernel-level paged KV cache support
#2348 opened Mar 13, 2026 by blake-snc Loading…
7
Add SM120 split-KV (FlashDecoding) with FP32 partial outputs
#2336 opened Mar 12, 2026 by blake-snc Loading…
6
ProTip! Add no:assignee to see everything that’s not assigned.