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 261573 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/mlx 分支比较 4 提交 »da2b371916 pass grad norm in mlx worker051b2024be Fix studio integration, local dataset files, chat templates without the torch gpu importsaaa2fabd65 Merge remote-tracking branch 'origin/main'797ae4cf40 mlx fixes
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/macos-arm64-venv-autodetect 分支70d4cc4885 fix(install): rebuild x86_64 (Rosetta) venv as arm64 on Apple Silicon
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/mac-llamacpp-binary 分支4fadb05c50 fix: detect Apple Silicon under Rosetta and use arm64 binaries
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/llama-update-banner-downgrade 分支比较 2 提交 »410299b5a6 [pre-commit.ci] auto fixes from pre-commit.com hooksb0dff621b6 Fix llama.cpp update banner offering a downgrade when /releases/latest lags
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/llama-tag-latest 分支比较 3 提交 »90724f6dab fix(studio): suppress fatal error when prebuilt manifest is missingb6ffd94918 docs: add comment explaining latest vs master for llama.cpp tag1902e83a67 fix(studio): revert llama.cpp default tag to latest
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/llama-strix-system-rocm 分支39ee347a39 studio: prefer system ROCm over the bundled runtime on bare-metal Strix Halo
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/llama-gpu-offload-validation 分支比较 10 提交 »2bbd1a679a [pre-commit.ci] auto fixes from pre-commit.com hooksb990701a3e Studio: exempt macOS Metal from the GPU-offload rejectione9f52c2114 [pre-commit.ci] auto fixes from pre-commit.com hooks889e889d9c Studio: gate ROCm hosts out of CPU prebuilt and require offload signal in smoke-test100e27ffe7 Studio: re-validate matching existing installs and reused source builds (#5807)
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/llama-cpp-b8637 分支8ecfb07616 fix(studio): revert llama.cpp default tag to latest
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/lemonade-marker-update-to-fork 分支比较 6 提交 »274cc61ecb Studio: harden marker handling and tighten the data-center routing gateffd7932860 [pre-commit.ci] auto fixes from pre-commit.com hooks10d7ac5eb2 Studio: align data-center lemonade selection with the time-ordered resolver4a02cedf60 [pre-commit.ci] auto fixes from pre-commit.com hooksac5652be6b Studio: hydrate a real upstream source tree for data-center lemonade installs
- frostbyte_neo 推送了仓库 github-featured/unslothai--unsloth 的 fix/lazy-import-llama-cpp 分支737017a511 fix(studio): lazy-import extract_model_size_b in llama_cpp to fix transformers 5.x version switch
