{ stdenv , config , pkgs , lib , fetchFromGitHub , cmake , eigen , onnxruntime , opencv , cudaSupport ? config.cudaSupport , cudaPackages ? { } }@inputs: let effectiveStdenv = if cudaSupport then cudaPackages.backendStdenv else inputs.stdenv; cudaCapabilities = cudaPackages.cudaFlags.cudaCapabilities; # E.g. [ "80" "86" "90" ] cudaArchitectures = (builtins.map cudaPackages.cudaFlags.dropDot cudaCapabilities); cudaArchitecturesString = lib.strings.concatStringsSep ";" cudaArchitectures; in effectiveStdenv.mkDerivation (finalAttrs: { pname = "fastdeploy-ppocr"; version = "0-unstable-2023-10-09"; src = fetchFromGitHub { owner = "MaaAssistantArknights"; repo = "FastDeploy"; # follows but not fully follows https://github.com/MaaAssistantArknights/MaaDeps/blob/master/vcpkg-overlay/ports/maa-fastdeploy/portfile.cmake#L4 rev = "0db6000aaac250824266ac37451f43ce272d80a3"; hash = "sha256-5TItnPDc5WShpZAgBYeqgI9KKkk3qw/M8HPMlq/H4BM="; }; outputs = [ "out" "cmake" ]; nativeBuildInputs = [ cmake eigen ] ++ lib.optionals cudaSupport [ cudaPackages.cuda_nvcc ]; buildInputs = [ onnxruntime opencv ] ++ lib.optionals cudaSupport (with cudaPackages; [ cuda_cccl # cub/cub.cuh libcublas # cublas_v2.h libcurand # curand.h libcusparse # cusparse.h libcufft # cufft.h cudnn # cudnn.h cuda_cudart ]); cmakeFlags = [ (lib.cmakeFeature "CMAKE_BUILD_TYPE" "None") (lib.cmakeBool "BUILD_SHARED_LIBS" true) ] ++ lib.optionals cudaSupport [ (lib.cmakeFeature "CMAKE_CUDA_ARCHITECTURES" cudaArchitecturesString) ]; postInstall = '' mkdir $cmake install -Dm644 ${finalAttrs.src}/cmake/Findonnxruntime.cmake $cmake/ ''; meta = with lib; { description = "MaaAssistantArknights stripped-down version of FastDeploy"; homepage = "https://github.com/MaaAssistantArknights/FastDeploy"; platforms = platforms.linux; license = licenses.asl20; broken = cudaSupport && stdenv.hostPlatform.system != "x86_64-linux"; }; })