Skip to content

Pull requests: fla-org/flash-linear-attention

New pull request New
34 Open 646 Closed
34 Open 646 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

[Perf] Fuse ds in Ascend chunk_bwd_dqkwg to avoid recomputing do@v.T
#1048 opened Jul 20, 2026 by sunyi0505 Contributor Loading…
[Ops] Add triton-ascend backend for KDA kernels
#1047 opened Jul 20, 2026 by zheliuyu Contributor • Draft
[Perf] TileLang backends for linear-attention chunk backward (simple_gla / gla / delta_rule / gdn)
#1046 opened Jul 19, 2026 by markovchain-builder Contributor Loading…
2
[Feature] support reusable final_state buffer in fused recurrent GDN
#1043 opened Jul 18, 2026 by Bsgg1 Loading…
1
[Fix] Reject non-divisible GQA head counts
#1032 opened Jul 16, 2026 by morluto Contributor Loading…
1
1
[Perf] Optimize the performance of causal conv1D on Ascend NPU
#1024 opened Jul 15, 2026 by ChengQianqian • Draft
2
[Ops] Rewrite NPU chunk_scaled_dot_kkt_fwd
#1023 opened Jul 14, 2026 by OsirisDuan Contributor Loading…
3
[CP] Enable context parallelism for GDP
#1022 opened Jul 13, 2026 by Mellonta Loading…
1
[Model] Add CAT (Compress and Attend Transformer)
#1017 opened Jul 12, 2026 by bicycleman15 Loading…
[GDN] Add batch-invariant mode: bitwise-reproducible prefill/decode across batch sizes
#1003 opened Jul 4, 2026 by yichuan-w Loading…
[GDN2] Add fused BT=16 inference kernels for GDN2 prefill
#990 opened Jun 29, 2026 by Ghd-2077 Loading…
[Attn] Remove erroneous @torch.compile on ParallelAttentionFunction class
#980 opened Jun 25, 2026 by arbi-dev • Draft
1
[Perf] Generalize fused q/k/v short convolution across layers
#977 opened Jun 23, 2026 by zhiyuan1i Collaborator Loading…
3
[Fix] generate() prefills the full prompt instead of only the last token
#962 opened Jun 20, 2026 by Sunt-ing Contributor Loading…
4
[Fix] causal_conv1d_bwd silent dweight corruption with cu_seqlens and B>1
#955 opened Jun 19, 2026 by TimDarcet • Draft
5
[Misc] Add generic concat + conv + split vs N separate grouped convs benchmark
#952 opened Jun 18, 2026 by Costa-SM Contributor Loading…
6
[Model] Add Preconditioned Gated DeltaNet (PGDN) and KDA (PKDA)
#950 opened Jun 17, 2026 by ntumm120 Loading…
8
[GDN] Fix GDN precision on Blackwell
#948 opened Jun 14, 2026 by syeehyn Loading…
1
[GDN] Restrict chunk_delta_h Blackwell autotune stages
#946 opened Jun 12, 2026 by IgorYashch Loading…
[Fix] Zero-init chunk-mode backward gradient buffers to prevent NaN propagation
#892 opened May 12, 2026 by xylian86 Loading…
6
[Fix] Fix shared memory race in tilelang chunk_bwd dg_last accumulation help wanted Extra attention is needed
#890 opened May 11, 2026 by Erix025 Contributor Loading…
6
[SSE] Add SSE integration
#882 opened May 9, 2026 by Pan-Yuqi Contributor Loading…
7
[KDA][AMD]for kda kernel,fix core dump on AMD GPU and tune the config for AMD branch
#869 opened Apr 29, 2026 by binding7012 Loading…
10
[Ops] Fix int32 overflow in pointer arithmetic across all Triton kernels
#818 opened Apr 8, 2026 by tmct Contributor • Draft
43
Add MALA (Magnitude-Aware Linear Attention) to FLA
#809 opened Apr 3, 2026 by drdanielwuwu Loading…
25
ProTip! Type g p on any issue or pull request to go back to the pull request listing page.