vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Nix: fix ARM neon dotproduct on macOS

Browse source
This commit is contained in:
Niklas Korz 2023-03-14 21:14:29 +01:00
parent a296589180
commit 84fb7f516c
1 changed files with 9 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

13
flake.nix
View file

@ -18,21 +18,26 @@
{
packages.default = pkgs.stdenv.mkDerivation {
name = "llama.cpp";
src = ./.;
nativeBuildInputs = with pkgs; [ cmake ];
buildInputs = with pkgs; lib.optionals stdenv.isDarwin [
darwin.apple_sdk.frameworks.Accelerate
];
src = ./.;
cmakeFlags = with pkgs; lib.optionals (system == "aarch64-darwin") [
"-DCMAKE_C_FLAGS=-D__ARM_FEATURE_DOTPROD=1"
];
installPhase = ''
mkdir -p $out/bin
mv main $out/bin/llama-cpp
mv quantize $out/bin/llama-cpp-quantize
mv llama $out/bin/llama
mv quantize $out/bin/quantize
echo "#!${llama-python}/bin/python" > $out/bin/convert-pth-to-ggml
cat convert-pth-to-ggml.py >> $out/bin/convert-pth-to-ggml
cat ${./convert-pth-to-ggml.py} >> $out/bin/convert-pth-to-ggml
chmod +x $out/bin/convert-pth-to-ggml
'';
};
devShells.default = pkgs.mkShell {
packages = with pkgs; [
cmake
llama-python
] ++ lib.optionals stdenv.isDarwin [
darwin.apple_sdk.frameworks.Accelerate