Morph MORPH
SIGN IN SIGN UP
dhruvhead / flash-linear-attention UNCLAIMED
0 0 0 Python

COMMITS

main
July 22, 2026
M
[Attn] Support for heterogeneous hybrid-attention configurations (#1051)
Michael R committed
d1ce073
C
[CI] add ascend-a2-benchmark-ci and fix ascend-a2-ci (#1052)
ChunyuWei committed
89af578
July 21, 2026
C
[Ops] Add triton-ascend backend for KDA kernels (#1047)
ChunyuWei committed
33fcbdf
Z
[Cleanup] Remove Ascend chunk_fwd_o kernels superseded by #1049 (#1050)
Zhiyuan Li committed
e537b7c
A
[Perf] Fuse Ascend chunk_fwd_o inter+intra into unified 1D core-grid kernel (#1049)
Ascendyh committed
7e188d7
S
[Perf] Fuse ds in Ascend chunk_bwd_dqkwg to avoid recomputing do@v.T (#1048)
sunyi0505 committed
cac37a9
July 20, 2026
A
[Perf] Add opt-in TileLang RWKV6 intra kernel (#1045)
Annie Guo committed
c70f11c
S
[Perf][Misc] Add LayerNormGated benchmarks and BT-tile Ascend fused_norm_gate (#1044)
sunyi0505 committed
f3acf5d
July 19, 2026
M
[Perf] Cache and bucket Wall attention autotuning (#1041)
morluto committed
cbb0a72
M
[Fix] Support value-split Wall backward with local deltas (#1042)
morluto committed
cff3451
A
[Ops][GDN][Perf] Rewrite recompute_w_u_fwd kernel for Triton Ascend backend (#1021)
Ascendyh committed
ee23e19
July 18, 2026
M
[Refactor] Centralize generation errors for unsupported cache strategies (#1035)
morluto committed
fe8fce9
M
[Fix] Avoid concurrent LSE stores in value-split attention (#1033)
morluto committed
2cec2a3
July 17, 2026
M
[Fix] Store split attention decode outputs at the correct offset (#1031)
morluto committed
ccb0ff9
S
[Perf] Optimize triton-ascend L2Norm with row tiling on Ascend NPU (#1036)
sunyi0505 committed
5f3801f
July 16, 2026
M
[Fix] fall back to Triton when TileLang has no usable nvcc compiler (#1026)
Matt Van Horn committed
b328e7c
H
[Fix] based: only one V-split program stores z in fused_chunk fwd (#1018)
Hao Wu committed
5659576
S
[Fix] Correct causal convolution benchmark dispatch and Ascend CI Python setup (#1027)
sunyi0505 committed
d9bba6e
July 15, 2026
S
[Misc] Add faceted benchmark visualization and multi-device benchmark support (#1019)
sunyi0505 committed
4eb2560
July 12, 2026
C
[Ops] Add GDN kernels for triton-ascend backend (#1011)
ChunyuWei committed
94e2725
July 9, 2026
Z
[Test] Skip flash-attn-dependent tests when flash-attn is not installed (#1014)
Zhiyuan Li committed
7dd7cf5
Z
Add get_max_length to FLA cache layer for transformers 5.x (#1009)
Zhiyuan Li committed
953e435
July 8, 2026
N
[Fix] Correct Mamba-3 decay parameterization (#1012)
Netanel Haber committed
9b04758
July 7, 2026
Y
[Ops] Add Gluon backend for AttnRes (#1010)
Yu Zhang committed
454fe9e
Y
[Fix] Expose cp_context/disable_recompute in chunk_rwkv7 wrapper (#1004)
Yoake_0829 committed
6daaedf
Z
[Fix] avoid SymInt stride reads in activations under torch.compile (#1008)
Zhiyuan Li committed
bce39a4
X
[GDN] Add FlashQLA backend dispatch (#998)
Xi Lin committed
77b509f
July 6, 2026
S
[Fix] [Test] Remove unused GeLU/SquaredReLU NPU kernels and add Ascend UB Manager tests (#1006)
sunyi0505 committed
6bd9069
K
[Fix] drop deprecated tl.extra.cuda.libdevice prefix (#1005)
Kashif Rasul committed
adb8f38
July 3, 2026
C
[Ops] Add triton-ascend l2norm and fused_norm_gate backends for Ascend NPU (#1002)
ChunyuWei committed
6f62104