Owner
fla-org
2 indexed repositories · View on GitHub
-
flash-linear-attention
🚀 Efficient implementations for emerging model architectures
Python · 6054 stars
-
flame
🔥 A minimal training framework for scaling FLA models
Python · 428 stars
Issues somebody may already be working on are listed last. List everything by date
-
Difficulty 4/5 3-5 days Newbie friendliness 45/100
fla-org/flash-linear-attention#1748 ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 15/100
fla-org/flash-linear-attention#1730 · 5 comments ·
Maintainers usually reply within 1 day
-
ascend-npu bug
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
fla-org/flash-linear-attention#1712 · 3 comments ·
Maintainers usually reply within 1 day
-
[Ops] New operator proposal: decay_max_scan - input-dependent decay running-max scan with O(1) stateOpen
Difficulty 5/5 Over a week Newbie friendliness 35/100
fla-org/flash-linear-attention#1705 · 2 comments ·
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 35/100
fla-org/flash-linear-attention#1691 ·
Maintainers usually reply within 1 day
-
Difficulty 4/5 3-5 days Newbie friendliness 49/100
fla-org/flash-linear-attention#1683 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 35/100
fla-org/flash-linear-attention#1672 ·
Maintainers usually reply within 1 day
-
Difficulty 4/5 3-5 days Newbie friendliness 43/100
fla-org/flash-linear-attention#1654 · 1 comment ·
Maintainers usually reply within 1 day
-
[Bug] MoM synthetic convolution padding corrupts cached short continuationsMay be free again @Cross2pro claimed this 18 days ago, and no pull request is open. Openbug
Difficulty 4/5 3-5 days Newbie friendliness 42/100
fla-org/flash-linear-attention#1611 · 1 comment ·
Maintainers usually reply within 1 day
-
[Bug] chunk_gated_delta_rule: wrong bfloat16 gradients in the early layers of Qwen3.5-4B on A100 (fla 0.5.2)May be free again @Cross2pro claimed this 18 days ago, and no pull request is open. Openbug
Difficulty 4/5 3-5 days Newbie friendliness 45/100
fla-org/flash-linear-attention#1598 · 4 comments ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 40/100
fla-org/flash-linear-attention#1586 ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#1560 ·
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#1556 · 2 comments ·
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 35/100
fla-org/flash-linear-attention#1550 ·
Maintainers usually reply within 1 day
-
ascend-npu enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#1531 · 4 comments ·
Maintainers usually reply within 1 day
-
stale
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#1379 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#1370 · 3 comments ·
Maintainers usually reply within 1 day
-
ascend-npu enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#942 · 13 comments · 17 reactions ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#680 ·
Maintainers usually reply within 1 day
-
[RFC] Context Parallel ImplementationsMay be free again @zhiyuan1i claimed this 290 days ago, and no pull request is open. Openenhancement
fla-org/flash-linear-attention#674 · 3 comments · 6 reactions · 1 assignee ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#659 · 1 comment ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 35/100
fla-org/flash-linear-attention#657 · 2 comments ·
Maintainers usually reply within 1 day
-
[RFC] Add flag to control recomputation in chunk kernelsMay be free again @zhiyuan1i claimed this 364 days ago, and no pull request is open. Openenhancement
fla-org/flash-linear-attention#485 · 1 assignee ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#471 · 2 comments ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 18/100
fla-org/flash-linear-attention#441 · 1 comment ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 20/100
fla-org/flash-linear-attention#439 ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#357 · 2 comments ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 20/100
fla-org/flash-linear-attention#340 ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#328 · 2 comments · 2 reactions ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 35/100
fla-org/flash-linear-attention#318 ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#314 ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#269 · 2 comments · 1 reaction ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 20/100
fla-org/flash-linear-attention#236 · 1 comment · 3 reactions ·
Maintainers usually reply within 1 day
-
[RFC] Unifying `chunk` and `fused_chunk` modeMay be free again A pull request for this issue was closed without being merged. Openenhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#229 ·
Maintainers usually reply within 1 day
-
[Feature Request] Customize initialization / Add a switch for turning off FLA's initializationMay be free again @zhiyuan1i claimed this 614 days ago, and no pull request is open. Openenhancement
fla-org/flash-linear-attention#220 · 4 comments · 1 assignee ·
Maintainers usually reply within 1 day
-
[Feature Request] Allow Q K has different sequence lenMay be free again @yzhangcs claimed this 567 days ago, and no pull request is open. Openenhancement
fla-org/flash-linear-attention#219 · 6 comments · 1 reaction · 1 assignee ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#190 · 1 comment · 1 reaction ·
Maintainers usually reply within 1 day
-
enhancement urgent
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#153 ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#148 · 1 comment ·
Maintainers usually reply within 1 day
-
[RFC] Fuse shortconv and output norm/gate into the kernelMay be free again @yzhangcs claimed this 659 days ago, and no pull request is open. Openenhancement
fla-org/flash-linear-attention#140 · 1 assignee ·
Maintainers usually reply within 1 day
-
enhancement urgent
Difficulty 5/5 Over a week Newbie friendliness 25/100
fla-org/flash-linear-attention#127 ·
Maintainers usually reply within 1 day
-
enhancement urgent
Difficulty 5/5 Over a week Newbie friendliness 30/100
fla-org/flash-linear-attention#117 · 2 comments · 1 reaction ·
Maintainers usually reply within 1 day
-
enhancement stale
Difficulty 5/5 Over a week Newbie friendliness 15/100
fla-org/flash-linear-attention#110 · 3 comments ·
Maintainers usually reply within 1 day
-
precond_kda (new op, #1552) backward kernels hit CUDA illegal memory access on Blackwell under triton 3.3.1, pass under triton 3.7.1Possibly taken @YYuJLiu claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
fla-org/flash-linear-attention#1749 ·
Maintainers usually reply within 1 day
-
[RFC] Preserve normalized decode key-sum state in FP32Possibly taken A pull request linked to this issue is open or already merged. Open
Difficulty 4/5 3-5 days Newbie friendliness 30/100
fla-org/flash-linear-attention#1735 ·
Maintainers usually reply within 1 day
-
[Bug] prepare_sequence_ids and prepare_token_indices skip empty sequences in cu_seqlensPossibly taken @0z5a claimed this 2 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
fla-org/flash-linear-attention#1732 ·
Maintainers usually reply within 1 day
-
[Bug] Missing dy rearrange in causal_conv1d_bwd when x.ndim > 3Possibly taken @shatianming5 claimed this 12 days ago. Open
Difficulty 3/5 1-2 days Newbie friendliness 48/100
fla-org/flash-linear-attention#1660 ·
Maintainers usually reply within 1 day
-
[RFC] Upstream ReWY backward optimizations for Delta-Rule attentionPossibly taken A pull request linked to this issue is open or already merged. Open
Difficulty 5/5 Over a week Newbie friendliness 35/100
fla-org/flash-linear-attention#1653 · 1 comment ·
Maintainers usually reply within 1 day
-
[Bug] GKA FP32 with K > 64 doesn't fit in 99 KB shared memoryPossibly taken @fleeb83 claimed this 12 days ago. Openbug
Difficulty 4/5 3-5 days Newbie friendliness 35/100
fla-org/flash-linear-attention#1638 ·
Maintainers usually reply within 1 day
-
[CI] nvidia-h100-smoke keeps timing out at 120 minutes as the dependent-test fan-out growsPossibly taken @zhiyuan1i claimed this 24 days ago. Open
Difficulty 4/5 3-5 days Newbie friendliness 45/100
fla-org/flash-linear-attention#1522 · 1 comment · 1 assignee ·
Maintainers usually reply within 1 day