Skip to content

Pull requests: lightseekorg/tokenspeed

New pull request New
38 Open 1,603 Closed
38 Open 1,603 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(moe): skip padded stage2 reduction tiles
#1724 opened Sep 22, 2026 by qedawkins Contributor Loading…
ci(amd-kernel): Add GLM 5.3 Flash MoE Kernel Benchmarks
#1723 opened Sep 22, 2026 by Max191 Contributor • 3/3 Loading…
7
ci(amd-kernel): Add GLM 5.3 Flash DSA Kernel Benchmarks
#1721 opened Sep 22, 2026 by Max191 Contributor • 2/3 Loading…
8
perf(kimi3): cut extra passes and undersized launches from gfx1250 prefill
#1720 opened Sep 22, 2026 by Yu-Zhewen Contributor Loading…
10
feat(dcp): support GPU MLA and DSA decode context parallelism
#1718 opened Sep 22, 2026 by OftenDream Collaborator • Draft
feat(amd): add Kimi K3 prefill projection kernel
#1711 opened Sep 22, 2026 by antiagainst Member Loading…
14
refactor(kernel): remove dsv4 pack topk router
#1709 opened Sep 22, 2026 by borontion Contributor • 2/2 Loading…
3
refactor(kernel): move top-k related kernel after moe_topk api
#1706 opened Sep 22, 2026 by borontion Contributor Loading…
14
refactor(kernel): clean up dsv4 mega moe apis
#1701 opened Sep 21, 2026 by borontion Contributor • 1/2 Loading…
33
feat (kernel): Support block-interleaved DCP MLA decode
#1694 opened Sep 21, 2026 by wzhao18 Loading…
3
perf(kernel): Add an optional split-KV floor to MLA decode
#1693 opened Sep 21, 2026 by wzhao18 Loading…
4
perf(kernel): wire PDL into the activation Triton kernels
#1691 opened Sep 21, 2026 by tuanzhangCS Contributor Loading…
7
perf(engram): Fuse input preparation and n-gram hashing
#1687 opened Sep 21, 2026 by chenht2022 Member Loading…
6
fix(sampling): acquire PDL inputs before softmax reads
#1682 opened Sep 21, 2026 by yechank-nvidia Collaborator • Draft
@yechank-nvidia
perf(hyperconnection): fuse batched HC mix up to 256 rows
#1679 opened Sep 21, 2026 by tuanzhangCS Contributor • Draft
7
perf(attention): tune CuTe QSA tiling and asynchronous KV staging
#1677 opened Sep 21, 2026 by tuanzhangCS Contributor Loading…
6
feat(kda): Add optional BF16 KDA state and FlashInfer decode/verify support
#1666 opened Sep 20, 2026 by chenht2022 Member Loading…
6
feat: support tp shard on MoE shared expert
#1652 opened Sep 20, 2026 by byshiue Collaborator Loading…
2
perf(moe): fuse FlashInfer NVFP4 routing-map padding initialization
#1644 opened Sep 19, 2026 by byshiue Collaborator Loading…
5
perf(amd): give gfx1250 MXFP4 decode's spare warp bit to N
#1641 opened Sep 18, 2026 by jerryyin Member Loading…
2
refactor(kernel): unify PyTorch references under numerics/reference
#1626 opened Sep 17, 2026 by antiagainst Member • Draft
1
[Don't Merge][Draft] feat(kimi-k3) DEP + TP o projection on attention
#1610 opened Sep 17, 2026 by byshiue Collaborator Loading…
fix(runtime): aggregate cache flush results across data parallel schedulers
#1608 opened Sep 17, 2026 by yechank-nvidia Collaborator Loading…
@yechank-nvidia
1
perf(sampling): skip unused fallback-index reduction in target-only v…
#1607 opened Sep 17, 2026 by yechank-nvidia Collaborator Loading…
@yechank-nvidia
1
perf(moe): pass unpacked routes to FlashInfer
#1606 opened Sep 17, 2026 by yechank-nvidia Collaborator Loading…
@yechank-nvidia
1
ProTip! What’s not been updated in a month: updated:<2026-08-22.