diff options
Diffstat (limited to 'flake.nix')
| -rw-r--r-- | flake.nix | 100 |
1 files changed, 61 insertions, 39 deletions
@@ -6,52 +6,74 @@ outputs = { self, nixpkgs, flake-utils }: flake-utils.lib.eachDefaultSystem (system: let - inherit (pkgs.stdenv) isAarch64 isDarwin; - inherit (pkgs.lib) optionals; - isM1 = isAarch64 && isDarwin; - osSpecific = if isM1 then - with pkgs.darwin.apple_sdk_11_0.frameworks; [ - Accelerate - MetalKit - MetalPerformanceShaders - MetalPerformanceShadersGraph - ] - else if isDarwin then - with pkgs.darwin.apple_sdk.frameworks; [ - Accelerate - CoreGraphics - CoreVideo - ] - else - [ ]; + inherit (pkgs.stdenv) isAarch32 isAarch64 isDarwin; + buildInputs = with pkgs; [ openmpi ]; + osSpecific = with pkgs; buildInputs ++ + ( + if isAarch64 && isDarwin then + with pkgs.darwin.apple_sdk_11_0.frameworks; [ + Accelerate + MetalKit + MetalPerformanceShaders + MetalPerformanceShadersGraph + ] + else if isAarch32 && isDarwin then + with pkgs.darwin.apple_sdk.frameworks; [ + Accelerate + CoreGraphics + CoreVideo + ] + else + with pkgs; [ openblas ] + ); pkgs = import nixpkgs { inherit system; }; + nativeBuildInputs = with pkgs; [ cmake pkgconfig ]; llama-python = - pkgs.python310.withPackages (ps: with ps; [ numpy sentencepiece pip ]); + pkgs.python3.withPackages (ps: with ps; [ numpy sentencepiece ]); + postPatch = '' + substituteInPlace ./ggml-metal.m \ + --replace '[bundle pathForResource:@"ggml-metal" ofType:@"metal"];' "@\"$out/bin/ggml-metal.metal\";" + substituteInPlace ./*.py --replace '/usr/bin/env python' '${llama-python}/bin/python' + ''; + postInstall = '' + mkdir -p $out/bin + mv bin/* $out/bin/ + mv $out/bin/main $out/bin/llama + mv $out/bin/server $out/bin/llama-server + + echo "#!${llama-python}/bin/python" > $out/bin/llama-convert.py + cat ${./convert.py} >> $out/bin/llama-convert.py + chmod +x $out/bin/llama-convert.py + ''; + cmakeFlags = [ "-DLLAMA_BUILD_SERVER=ON" "-DLLAMA_MPI=ON" "-DBUILD_SHARED_LIBS=ON" "-DCMAKE_SKIP_BUILD_RPATH=ON" "-DLLAMA_LTO=ON" "-DLLAMA_SANITIZE_THREAD=OFF -DLAMMA_SANITIZE_ADRRESS=ON" "-DLLAMA_SANITIZE_UNDEFINED=ON" ]; in { packages.default = pkgs.stdenv.mkDerivation { name = "llama.cpp"; src = ./.; - postPatch = if isM1 then '' - substituteInPlace ./ggml-metal.m \ - --replace '[bundle pathForResource:@"ggml-metal" ofType:@"metal"];' "@\"$out/bin/ggml-metal.metal\";" - '' else - ""; - nativeBuildInputs = with pkgs; [ cmake ]; + postPatch = postPatch; + nativeBuildInputs = nativeBuildInputs; buildInputs = osSpecific; - cmakeFlags = [ "-DLLAMA_BUILD_SERVER=ON" "-DLLAMA_LTO=ON" "-DLLAMA_SANITIZE_THREAD=OFF -DLAMMA_SANITIZE_ADRRESS=ON" "-DLLAMA_SANITIZE_UNDEFINED=ON" ] ++ (optionals isM1 [ - "-DCMAKE_C_FLAGS=-D__ARM_FEATURE_DOTPROD=1" - "-DLLAMA_METAL=ON" + cmakeFlags = cmakeFlags + ++ (if isAarch64 && isDarwin then [ + "-DCMAKE_C_FLAGS=-D__ARM_FEATURE_DOTPROD=1" + "-DLLAMA_METAL=ON" + ] else [ + "-DLLAMA_BLAS=ON" + "-DLLAMA_BLAS_VENDOR=OpenBLAS" ]); - installPhase = '' - mkdir -p $out/bin - mv bin/* $out/bin/ - mv $out/bin/main $out/bin/llama - mv $out/bin/server $out/bin/llama-server - - echo "#!${llama-python}/bin/python" > $out/bin/llama-convert.py - cat ${./convert.py} >> $out/bin/llama-convert.py - chmod +x $out/bin/llama-convert.py - ''; + postInstall = postInstall; + meta.mainProgram = "llama"; + }; + packages.opencl = pkgs.stdenv.mkDerivation { + name = "llama.cpp"; + src = ./.; + postPatch = postPatch; + nativeBuildInputs = nativeBuildInputs; + buildInputs = with pkgs; buildInputs ++ [ clblast ]; + cmakeFlags = cmakeFlags ++ [ + "-DLLAMA_CLBLAST=ON" + ]; + postInstall = postInstall; meta.mainProgram = "llama"; }; apps.llama-server = { @@ -68,7 +90,7 @@ }; apps.default = self.apps.${system}.llama; devShells.default = pkgs.mkShell { - packages = with pkgs; [ cmake llama-python ] ++ osSpecific; + packages = nativeBuildInputs ++ osSpecific; }; }); } |