By name: gotcha-ollama-acceleration-deprecated
description: USE WHEN configuring CUDA support for Ollama — services.ollama.acceleration = "cuda" is removed in current nixpkgs. Use package = pkgs.ollama-cuda instead. First build is slow (~30 min) — nvcc compiles GGML for many CUDA arches and binary cache often misses.
services.ollama.acceleration = "cuda" is deprecated
Removed in current nixpkgs. Use package = pkgs.ollama-cuda instead. First build of ollama-cuda is slow (~30 min) — nvcc compiles GGML for many CUDA arches and binary cache often misses.