12月1月2月3月4月5月6月7月8月9月10月11月
周一周三周五
一年内 263947 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 fix/kimi-k3-dp-mla-packing 分支a6ee91b3e6 fix(cache): derive the Kimi-K3 MLA packing from the KDA state size
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 k3-nvfp4-gb300-ci 分支25b14d0162 ci: add GB300 two-node AIME26 gates for K3-NVFP4 and K3-NVFP4+DSpark
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 replace-private-mnnvl 分支31bb248504 perf(comm): route generic MNNVL AR patterns to upstream flashinfer
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 main 分支比较 79 提交 »978ed2cfdc perf(cache): sparsify and budget Kimi-K3 state cache (#1147)53acfa2608 ci: use explicit GB300 runner labels (#1151)94f4d07cf6 perf(kimi3): avoid packed KDA QKV copies (#1129)cb1b5ed9fd perf(kda): forward host cu_seqlens to the CuteDSL prefill wrapper (#1120)1e4a7946ca Support nvidia/Kimi-K3-NVFP4: ModelOpt FP8_PB_WO attention + NVFP4 SiTU MoE (#1128)
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 borontion/amd-dsv4-1 分支e14de7326f dsv4 support
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 feat/kda-replay-commit 分支比较 169 提交 »099a54b977 fix(kda): probe both ops the replay path launchesc39d12d9e9 fix(kda): settle deferred state at release, unblock the flush pathb6e5256edf Merge remote-tracking branch 'origin/main' into feat/kda-replay-commit19320a84e5 test(runtime): actually run the drop-before-zero ordering test978ed2cfdc perf(cache): sparsify and budget Kimi-K3 state cache (#1147)