一、Ubuntu16安装NVIDIA驱动460.39
1、下载地址:https://www.nvidia.cn/drivers/results/170212/
2、安装参考:https://blog.csdn.net/lihe4151021/article/details/90083431
二、安装cuda10.0
1、下载地址:https://developer.nvidia.com/cuda-toolkit-archive
cuda10.0下载地址: https://developer.nvidia.com/cuda-10.0-download-archive?target_os=Linux
2、安装参考:https://blog.csdn.net/lihe4151021/article/details/90237681
三、安装cudnn7.6.5
1、下载地址:https://developer.nvidia.com/rdp/cudnn-archive
2、安装参考:https://blog.csdn.net/dudu815110/article/details/88592558(方法一)
四、安装Pycharm
安装参考:https://blog.csdn.net/shuiyixin/article/details/89530415
五、安装Anaconda3-5.3.1(python3.7)
1、下载可执行文件:https://www.anaconda.com/products/individual
2、安装参考:https://blog.csdn.net/u012243626/article/details/82469174
六、虚拟环境配置
1、创建虚拟环境并进入:
conda create -n *****(虚拟环境名) python=3.7 #创建虚拟环境
source activate *****(虚拟环境名) #进入虚拟环境
2、在虚拟环境中安装pytorch:
(1)使用如下指令查看cuda版本。
nvcc --version
(2)到pytorch官网选择与cuda版本匹配的pytorch:
pytorch官网:https://pytorch.org/get-started/previous-versions/
参考:https://blog.csdn.net/xuzhichao123456/article/details/109218835
我选择的是pytorch1.2.0(对应于cuda=10.0):(pip下载快很多)
pip install torch==1.2.0 torchvision==0.4.0
3、安装cv2、numpy等需要的模块(包):
(1)常用的安装和卸载指令:
sudo apt-get install #卸载: sudo apt-get remove
pip install #卸载: pip uninstall
conda install #卸载: conda uninstall
(2)某些模块(包/库)的安装:
source activate env_name #先激活进入虚拟环境再安装
python -m pip install opencv-python
pip install numpy
pip install scipy
pip install matplotlib
pip install tensorboard
pip install h5py
conda install scikit-image
(3)安装apex用于多GPU训练
cd PATH_TO_INSTALL
git clone https://github.com/NVIDIA/apex
cd apex
git reset --hard 4ef930c1c884fdca5f472ab2ce7cb9b505d26c1a
pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./