Skip to content

Pull requests: Dao-AILab/flash-attention

New pull request New
179 Open 610 Closed
179 Open 610 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

fix: add weights_only=True to torch.load calls (CVE-2026-31253)
#2585 opened May 23, 2026 by factnn Loading…
1
[Cute,Fwd,Sm100] NVFP4/MXFP8 mixed-precision fwd
#2582 opened May 22, 2026 by Edenzzzz Loading…
5 tasks
[Cute,Fwd,Sm100] Fix FP8 precision loss at tile boundaries
#2578 opened May 21, 2026 by Robin930 Loading…
1
fix(sm100): make arch check robust against CUTLASS DSL Arch enum changes
#2575 opened May 19, 2026 by lingolin128 Loading…
4
Fix FA2 varlen SplitKV early-exit LSE offset
#2573 opened May 19, 2026 by DCwashing Loading…
[CuTe, SM103] Update architecture assertion for SM 10.x and 11.x
#2572 opened May 17, 2026 by ocss884 Loading…
[cute, sm90, flex] support sm90 flex_attention deterministic
#2567 opened May 14, 2026 by geruome Contributor Loading…
Add descriptive kernel name
#2564 opened May 13, 2026 by drisspg Collaborator • Draft
1
Normalize scalar tensors in CuTe JIT cache keys
#2560 opened May 13, 2026 by xyzhang626 Loading…
1
[CuTe,Sm100] Varlen Dynamic Persistent scheduler and metadata
#2559 opened May 13, 2026 by reubenconducts Contributor Loading…
[CuTe,Fwd,Sm100] refactor mla sm100 forward and add page table
#2558 opened May 13, 2026 by jayhshah Collaborator Loading…
3
[CuTe,Bwd,Sm100] allow 2cta with score mod and mask mod in bwd
#2557 opened May 12, 2026 by reubenconducts Contributor Loading…
1
Add vectorized mask_mod support
#2554 opened May 11, 2026 by drisspg Collaborator • Draft
[Cute,Fwd,Sm120] Fix three forward-pass correctness bugs (use_tma_O, smem-store atom, pack_gqa)
#2553 opened May 11, 2026 by jganbar Loading…
Optimize decoding by two synthetic S/P lanes
#2551 opened May 9, 2026 by lightzhan-intellif Loading…
Reserve memory headroom and cap MAX_JOBS in source builds (#2493)
#2548 opened May 8, 2026 by shivam2199 Contributor Loading…
2 of 3 tasks
1
4
[CuTe, Flex] Allow score mod use in varlen backward
#2547 opened May 7, 2026 by reubenconducts Contributor Loading…
[CuTe, sm90, sm100] Feat: enable chunked causal attention for low-latency streaming workloads
#2546 opened May 7, 2026 by gpzlx1 Loading…
Bump AITER submodule to commit 3b2e6f4
#2540 opened May 5, 2026 by sstamenk Loading…
@micmelesse
2
nightly-ci: SM100 hang fixes, CI reliability improvements
#2538 opened May 5, 2026 by Johnsonms Collaborator • Draft
4 tasks
Remove stale guard
#2537 opened May 5, 2026 by drisspg Collaborator Loading…
tests: fix SM100 varlen backward failures on B200
#2534 opened May 3, 2026 by Johnsonms Collaborator • Draft
Add nightly benchmark CI: canonical FA4 benchmark tracking
#2533 opened May 2, 2026 by Johnsonms Collaborator • Draft
[Cute,Sm80] Fix forward GQA correctness — pack_gqa flag without pack_…
#2528 opened Apr 30, 2026 by kbkarthikeyan Loading…
Fix CK varlen num_splits plumbing
#2526 opened Apr 30, 2026 by TideDra Loading…
ProTip! Updated in the last three days: updated:>2026-05-21.