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 259347 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/rowboatlabs--rowboat 的 main 分支比较 2 提交 »7e1650811a Merge pull request #922 from rowboatlabs/archived-topics534122b3c1 fix(spaces): archived topics are reachable again
- frostbyte_neo 推送了仓库 github-featured/rowboatlabs--rowboat 的 feat/phase-8b 分支3b10ff416d fix(x): pass the relay option through createAuthServer
- frostbyte_neo 推送了仓库 github-featured/rowboatlabs--rowboat 的 arch/server-client-separation 分支80898aa067 feat(x): Phase 8b — token encryption at rest + OAuth loopback relay (#911)
- frostbyte_neo 推送了仓库 github-featured/lum1104--understand-anything 的 main 分支比较 4 提交 »ba450c4342 Merge pull request #190 from Omar-Basel-ALkhasawneh/patch-11f9952d572 Merge pull request #185 from Bortlesboat/docs/windows-update-uninstall3369980e2a fix(windows): make install.ps1 PS5-safe; document update/uninstallea8c8b3618 Enhance multi-agent pipeline documentation
- frostbyte_neo 推送了仓库 github-featured/mempalace--mempalace 的 develop 分支比较 7 提交 »4c1e6d0c11 Merge pull request #2353 from masalu0105-gif/fix/windows-config-utf802ae4f7001 test(config): stop assuming cp1252 rejects UTF-8 CJK bytesabf839ab3f fix(config): read and write project YAML as UTF-898e6dfc206 Merge pull request #1964 from iroiro147/fix/mcp-search-candidate-union-194986f2900c6a Merge pull request #2355 from AmirF194/fix/2325-drawer-mutation-closet-purge
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 ts2/mnnvl-plain-ar 分支比较 12 提交 »87bf9dd3f7 perf(comm): one shared fusion workspace per group serves plain all-reduces4cc9324c12 perf(gemm): pass native K-major scales to the FlashInfer FP8 block-scale GEMM (#1254)b1bb1a6845 fix(kda): select batched replay kernel on NVIDIA (#1253)ea070af475 fix: wire runtime pdl flag to all kernels (#1238)9b196f19dd feat(amd): add gfx950 A8W4 SiTU core (#1138)
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 xuchun/perf-glm 分支比较 14 提交 »7991f9243c fix(glm5): preserve AMD DSA prefill metadata supporta4fb46814e Merge branch 'main' into xuchun/perf-glm4cc9324c12 perf(gemm): pass native K-major scales to the FlashInfer FP8 block-scale GEMM (#1254)b1bb1a6845 fix(kda): select batched replay kernel on NVIDIA (#1253)ea070af475 fix: wire runtime pdl flag to all kernels (#1238)
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 bot/update-tokenspeed-smg-1.10.0.post20260826 分支374e2fee54 Update tokenspeed-smg* pins to 1.10.0.post20260826 / 0.4.16.post20260826 / 0.9.1.post20260826
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 ts2/retire-prepacked-mn 分支52c85a6a4f refactor(gemm): retire the FlashInfer prepacked-MN scale machinery
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 main 分支4cc9324c12 perf(gemm): pass native K-major scales to the FlashInfer FP8 block-scale GEMM (#1254)
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 feat/dp-rank-affinity 分支比较 2 提交 »b8d42ff536 Merge branch 'main' into feat/dp-rank-affinityb1bb1a6845 fix(kda): select batched replay kernel on NVIDIA (#1253)
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 fix/kimi-k3-dp-mla-packing 分支比较 13 提交 »8ea3ed51d8 Merge branch 'main' into fix/kimi-k3-dp-mla-packing4cc9324c12 perf(gemm): pass native K-major scales to the FlashInfer FP8 block-scale GEMM (#1254)b1bb1a6845 fix(kda): select batched replay kernel on NVIDIA (#1253)ea070af475 fix: wire runtime pdl flag to all kernels (#1238)9b196f19dd feat(amd): add gfx950 A8W4 SiTU core (#1138)
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 borontion/amd-dsv4-2 分支比较 3 提交 »71b419dae6 refactor(amd): move MXFP4 decode checks into kernel04d269866e style(amd): remove MXFP4 kernel tagsd0042bdf3c fix(amd): keep DSV4 offsets out of DeepGEMM
- frostbyte_neo 推送了仓库 github-featured/lightseekorg--tokenspeed 的 ci/mi450-simulator 分支870579cd1e fix(amd): harden MI450 simulator coverage
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 gma/xpu-pin-memory 分支比较 3 提交 »e88badfea3 XPU: resolve the SYCL runtime by mapped path instead of a hardcoded soname6e0a597d9d XPU: bind to torch's SYCL runtime instead of loading a second onefa1ca15ded XPU: register native pinned host memory with the SYCL runtime
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 gma/autotp-per-model-meta 分支比较 13 提交 »fff342f671 Port #8299 regression tests to per-model AutoTPMetaf077053ad6 Merge remote-tracking branch 'origin/master' into gma/autotp-per-model-meta715965e027 [Apple Silicon Support Phase 1] Add Metal FusedAdam kernel and CPU Adam build for Apple Silicon (#8300)7154a00f3d Fix AutoTP metadata updates for unsharded modules (#8299)eba5d27393 Avoid redundant copies in MPS P2P staging (#8314)
