{ stdenv , buildPythonPackage , fetchurl , isPy37 , isPy38 , python , nvidia_x11 , addOpenGLRunpath , future , numpy , patchelf , pyyaml , requests }: let pyVerNoDot = builtins.replaceStrings [ "." ] [ "" ] python.pythonVersion; platform = if stdenv.isDarwin then "darwin" else "linux"; srcs = import ./binary-hashes.nix; unsupported = throw "Unsupported system"; in buildPythonPackage { pname = "pytorch"; # Don't forget to update pytorch to the same version. version = "1.6.0"; format = "wheel"; disabled = !(isPy37 || isPy38); src = fetchurl srcs."${stdenv.system}-${pyVerNoDot}" or unsupported; nativeBuildInputs = [ addOpenGLRunpath patchelf ]; propagatedBuildInputs = [ future numpy pyyaml requests ]; postInstall = '' # ONNX conversion rm -rf $out/bin ''; postFixup = let rpath = stdenv.lib.makeLibraryPath [ stdenv.cc.cc.lib nvidia_x11 ]; in '' find $out/${python.sitePackages}/torch/lib -type f \( -name '*.so' -or -name '*.so.*' \) | while read lib; do echo "setting rpath for $lib..." patchelf --set-rpath "${rpath}:$out/${python.sitePackages}/torch/lib" "$lib" addOpenGLRunpath "$lib" done ''; pythonImportsCheck = [ "torch" ]; meta = with stdenv.lib; { description = "Open source, prototype-to-production deep learning platform"; homepage = "https://pytorch.org/"; license = licenses.unfree; # Includes CUDA and Intel MKL. platforms = platforms.linux; maintainers = with maintainers; [ danieldk ]; }; }