Alessandro Versace
Asked on Nov 04, 2023
The switch to the llama.cpp backend in Tabby's 0.5.0 release is an internal change that improves performance. It replaces the previous CPU/GPU backend on ctranslate2. However, note that the 0.5.0 release is still pre-release and may have some bugs, particularly in CPU support. It is recommended to stay at version 0.4.0 if you are running on CPU.