inclusionAI/cuLA v0.2.0

Init v0.2.0

Repository: inclusionAI/cuLA

Tag: v0.2.0

Published: 2026-08-12T09:48:19Z

Prerelease: no

Release notes:

What's Changed

• Adjust README by @icavan in https://github.com/inclusionAI/cuLA/pull/6
• Add Slack link and modify readme contact by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/7
• Prepare for pypi installation by @icavan in https://github.com/inclusionAI/cuLA/pull/22
• [KDA] Optimize recompute_wu kernel and support disable_recompute=True by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/27
• [KDA] Change to transposed VK state layout for Hopper fused forward by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/33
• Add PR template by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/39
• doc: update README by @icavan in https://github.com/inclusionAI/cuLA/pull/40
• support bf16 beta for KDA (SM90 & SM10X) by @cherhh in https://github.com/inclusionAI/cuLA/pull/34
• [KDA] Support CuTeDSL KDA Decode by @yuan-luo in https://github.com/inclusionAI/cuLA/pull/37
• [KDA] Refactor csrc kernel selection, add tests and sanity check by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/42
• [Doc] Update Kimi CP link by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/45
• docs: Update method for joining WeChat group by @zheyang0825 in https://github.com/inclusionAI/cuLA/pull/46
• doc: update README with bigger logo by @icavan in https://github.com/inclusionAI/cuLA/pull/53
• [KDA] Adapt recompute_wu and delta_h for backward by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/54
• perf(kda_decode): optimize GMEM coalescing, V-tile parallelism by @icavan in https://github.com/inclusionAI/cuLA/pull/52
• fix(linter): fix linter & updates kda decode perf by @icavan in https://github.com/inclusionAI/cuLA/pull/60
• feat: BHVK (K-last) state layout for Lightning Attention prefill & decode by @icavan in https://github.com/inclusionAI/cuLA/pull/56
• perf(la_decode): boost small-batch (B<=32) by 33% via 8-warp CTA by @icavan in https://github.com/inclusionAI/cuLA/pull/59
• [KDA] Optimize recompute_wu with better register allocation by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/61
• fix output_final_state wrapper issue by @yechenzhi in https://github.com/inclusionAI/cuLA/pull/63
• [KDA] fix internal output_final_state wrapper issue in SM90 by @yechenzhi in https://github.com/inclusionAI/cuLA/pull/66
• [Fix] change to umma pipelines for SM100 KDA by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/68
• [Fix] delta_h race risk by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/69
• [KDA] sm90 GVA enhance by @sjmshsh in https://github.com/inclusionAI/cuLA/pull/64
• [Feat] upgrade FLA to v0.5.0 by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/72
• [Fix] add cross-proxy fence for recomp_wu kernel by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/77
• [KDA] sm100 GVA enhance by @sjmshsh in https://github.com/inclusionAI/cuLA/pull/65
• fix: wire h0_indices into Lightning Attention decode for state-pool indexing by @Emre-Dinc in https://github.com/inclusionAI/cuLA/pull/75
• Unify the naming convention by @icavan in https://github.com/inclusionAI/cuLA/pull/79
• Consolidate duplicated benchmark helper logic into benchmarks/utils.py, including shared timing, reporting, and accuracy utilities. by @icavanyu in https://github.com/inclusionAI/cuLA/pull/80
• [KDA] support GVA for SM100 end-to-end by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/73
• [KDA] add backward chunk_wy_dqkg kernel for SM10X by @KevinZeng08 in https://github.com/inclusionAI/cuLA/pull/74
• [KDA] split unit tests into fast/slow modes by @Longxmas in https://github.com/inclusionAI/cuLA/pull/84
• [KDA] Add intra-card CP for chunk_delta_h forward in SM100 by @cherhh in https://github.com/inclusionAI/cuLA/pull/70
• Relax dependency minimum versions by @icavan in https://github.com/inclusionAI/cuLA/pull/87
• [CI] Add GitHub workflow for building and releasing fat wheels by @tongke6 in https://github.com/inclusionAI/cuLA/pull/91
• [ci] add prek lint workflow by @tongke6 in https://github.com/inclusionAI/cuLA/pull/98
• ci: align pytorch version requirements in wheel build by @tongke6 in https://github.com/inclusionAI/cuLA/pull/99
• refactor(kda): reorganize KDA backends into arch-first layout and add lazy imports by @cherhh in https://github.com/inclusionAI/cuLA/pull/100
• feat: intracard cp for sm90 by @Hyaloid in https://github.com/inclusionAI/cuLA/pull/86
• [KDA] KDA MTP decode: recurrent + KVBuffer chunkwise verify + flush by @Longxmas in https://github.com/inclusionAI/cuLA/pull/96
• [LA] Lightning Attention MTP decode + KVBuffer parallel verify / commit by @fkuner in https://github.com/inclusionAI/cuLA/pull/97
• feat: add packed kda decode by @zhouaihui in https://github.com/inclusionAI/cuLA/pull/102
• [Feature] Add GVA support for Lightning by @sjmshsh in https://github.com/inclusionAI/cuLA/pull/85
• [Fix] Wire cudac bindings by @cherhh in https://github.com/inclusionAI/cuLA/pull/105
• feat(kda): SM90 FlashKDA prefill, intracard CP, and backend dispatch by @cherhh in https://github.com/inclusionAI/cuLA/pull/106
• fix(kda): add the missing cross-proxy fence before TMA stores by @cherhh in https://github.com/inclusionAI/cuLA/pull/114
• [Lightning Attention] Add Hopper SM90 CuTe DSL prefill by @Aharrypotter in https://github.com/inclusionAI/cuLA/pull/111
• [KDA] Fuse causal conv1d into MTP verify decode by @Longxmas in https://github.com/inclusionAI/cuLA/pull/107
• Detect tcgen05 load/store API by signature by @icavan in https://github.com/inclusionAI/cuLA/pull/119
• Remove obsolete SM100 fully fused KDA prototype by @icavan in https://github.com/inclusionAI/cuLA/pull/116
• Update SM90 FlashKDA benchmark results by @icavan in https://github.com/inclusionAI/cuLA/pull/120
• feat: isolate cutlass._mlir imports behind compat gateway (#118) by @bikrammajhi in https://github.com/inclusionAI/cuLA/pull/121
• perf(kda): preserve SM90 workspace byte layout with bulk copies by @fkuner in https://github.com/inclusionAI/cuLA/pull/115

New Contributors

• @cherhh made their first contribution in https://github.com/inclusionAI/cuLA/pull/34
• @yuan-luo made their first contribution in https://github.com/inclusionAI/cuLA/pull/37
• @yechenzhi made their first contribution in https://github.com/inclusionAI/cuLA/pull/63
• @sjmshsh made their first contribution in https://github.com/inclusionAI/cuLA/pull/64
• @Emre-Dinc...