rope/nixos
1
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ollama: revert to CPU inference — M2000 CUDA incompatible with nixpkgs

Browse source 
CUDA ≤12.5 removed from nixpkgs as unmaintained; CUDA 12.6+ requires
driver ≥560 but legacy_535 (Maxwell's last supported branch) caps out
at 12.2. No compatible CUDA path exists for the Quadro M2000.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
ediblerope 2026-05-13 09:25:39 +01:00
parent b86a92293c
commit dc3eebb742
1 changed files with 3 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

11
services/ollama.nix
View file

@ -3,14 +3,9 @@
config = lib.mkIf (config.networking.hostName == "FredOS-Mediaserver") {
services.ollama.enable = true;
# nixpkgs ollama-cuda builds against CUDA 12.8, which requires driver ≥570.
# The Quadro M2000's legacy_535 driver only supports up to CUDA 12.2,
# so we override cudaPackages to build against 12.2 instead.
services.ollama.package = pkgs.ollama.override {
acceleration = "cuda";
cudaPackages = pkgs.cudaPackages_12_2;
cudaArches = [ "sm_52" ]; # Maxwell GM206 compute capability
};
# Quadro M2000 (Maxwell/GM206) uses legacy_535 driver which caps CUDA
# at 12.2. nixpkgs has removed all CUDA versions ≤12.5 as unmaintained,
# and 12.6+ requires driver ≥560. CPU inference is the only option.
services.open-webui.enable = true;
services.open-webui.port = 8888;