一、项目说明1.1 仓库地址GIT URL:https://github.com/huojichuanqi/ds#1.2 项目介绍gnina发音为NEE-na是一个分子对接程序集成了使用卷积神经网络对配体进行评分和优化的支持。它是 smina 的一个分支而 smina 是 AutoDock Vina 的一个分支。二、安装部署2.1 环境准备操作系统Ubuntu22.04CUDA12.0cmake: 2.25 (Ubuntu默认的是2.22.1需要修改CMakeLists.txt)python3.10.12torch: 需要和cuda的版本一致cuda12.4:pip install torch2.5.1 torchvision0.20.1 torchaudio2.5.1 --index-url https://download.pytorch.org/whl/cu124cuda12.6:pip install torch2.7.1 torchvision0.22.1 torchaudio2.7.1 --index-url https://download.pytorch.org/whl/cu126wsl能访问github宿主机配置好代理关闭防火墙2.2 开始编译2.2.1 安装依赖库aptupdate-yapt-getinstallbuild-essentialgitcmakewgetlibboost-all-dev libeigen3-dev libgoogle-glog-dev libprotobuf-dev protobuf-compiler libhdf5-dev libatlas-base-dev python3-dev librdkit-dev python3-numpy python3-pip python3-pytest libjsoncpp-dev libxml2-devaptupgrade cmake## 安装cudasudoapt-getremove nvidia-cuda-toolkitwgethttps://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda_12.4.0_550.54.14_linux.runchmod700cuda_12.4.0_550.54.14_linux.runsudoshcuda_12.4.0_550.54.14_linux.run## 安装cudnnwgethttps://developer.download.nvidia.com/compute/cudnn/9.0.0/local_installers/cudnn-local-repo-ubuntu2204-9.0.0_1.0-1_amd64.debsudodpkg-icudnn-local-repo-ubuntu2204-9.0.0_1.0-1_amd64.debsudocp/var/cudnn-local-repo-ubuntu2204-9.0.0/cudnn-*-keyring.gpg /usr/share/keyrings/sudoapt-getupdatesudoapt-get-yinstallcudnn-cuda-122.2.2 安装 OpenBabel3注意在 3.1.1 及更早版本中存在键级确定错误。gitclone https://github.com/dkoes/openbabel.gitcdopenbabelmkdirbuildcdbuild cmake-DWITH_MAEPARSEROFF-DWITH_COORDGENOFF-DPYTHON_BINDINGSON-DRUN_SWIGON..makemakeinstall2.2.3 安装 gninagitclone https://github.com/gnina/gnina.gitcdgninamkdirbuildcdbuild cmake..-DCMAKE_CUDA_ARCHITECTURES75make-j8sudomakeinstallcmake..\-DOPENBABEL3_INCLUDE_DIR/usr/local/include/openbabel3\-DOPENBABEL3_LIBRARIES/usr/local/lib/libopenbabel.so\-DJSONCPP_INCLUDE_DIR/usr/include/jsoncpp\-DJSONCPP_LIBRARY/usr/lib/x86_64-linux-gnu/libjsoncpp.so\-DBoost_USE_STATIC_LIBSOFFgnina编译的时候对环境有要求如果cmake的版本低于3.25那么修改根目录的CMakeLists.txt的第一行cmake_minimum_required(VERSION 3.25)为实际的cmake的版本号cmake的版本号可以通过cmake --version查看。第二个要修改的也是根目录的CMakeLists.txt在76行将set(CMAKE_CUDA_ARCHITECTURES all-major)改为实际的架构版本实际的查看命令为nvidia-smi --query-gpucompute_cap --formatcsv。#set(CMAKE_CUDA_ARCHITECTURES all-major) set(CMAKE_CUDA_ARCHITECTURES 75)