* TP: fix 0-sized tensor slices, AllReduce fallback * fix layer structure <-> GPU count aliasing * add missing std::fill * fix CUDA device set, max ggml ctx size |
||
|---|---|---|
| .. | ||
| cmake | ||
| include | ||
| src | ||
| .gitignore | ||
| CMakeLists.txt | ||
* TP: fix 0-sized tensor slices, AllReduce fallback * fix layer structure <-> GPU count aliasing * add missing std::fill * fix CUDA device set, max ggml ctx size |
||
|---|---|---|
| .. | ||
| cmake | ||
| include | ||
| src | ||
| .gitignore | ||
| CMakeLists.txt | ||