Note
Most packages dont have binary cache, and building anything CUDA is very painful and slow.
Nix flake for several AI projects focusing on nvidia/CUDA.
- tabbyapi unstable-2024-12-04
- exllamav2 0.2.6
- autogptq 0.7.1
- ava & ava-headless 2024-04-24
- ava-prebuilt 2024-04-21
- tensor_parallel 2.0.0
- text-generation-inference 2.4.1
- comfyui 0.3.10
- dadaptation 3.1
- prodigyopt 1.0
- lycoris-lora 2.0.2
- kohya_ss 22.6.1 (patched to run without having to include web files)
- spandrel, spandrel_extra_arches 0.4.0
- flash-attention 2.6.1
- rouge 1.0.1
nix run github:BatteredBunny/nix-ai-stuff#tabbyapi