Ubuntu 18.04 + GTX 1070 + CUDA10 + cuDNN7 +DeepLabV3+环境配置

 

1.安装Ubuntu 18.04 ,见博客《安装双系统》

2.安装GTX 1070驱动，见博客

http://www.geforce.cn/drivers

Ubuntu18.04 关闭和开启图形用户界面

a. 关闭用户图形界面
sudo systemctl set-default multi-user.target
sudo reboot
b. 开启用户图形界面
sudo systemctl set-default graphical.target
sudo reboot
 

测试安装是否成功
nvidia-smi

3.cuda_10.0.130_410.48_linux.run
sudo sh cuda_10.0.130_410.48_linux.run

在.bashrc配置环境变量
export PATH=/usr/local/cuda-10.0/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
export CUDA_HOME=/usr/local/cuda

测试安装是否成功

nvcc -V

4.安装libcudnn

sudo dpkg -i libcudnn7_7.6.5.32-1+cuda10.0_amd64.deb

5.安装tensorflow_gpu-1.14
sudo pip install tensorflow_gpu-1.14.0-cp27-cp27mu-manylinux1_x86_64.whl

6.下载tensorflow的模型models

下载源码models-master。https://github.com/tensorflow/models

在.bashrc配置环境变量
export PYTHONPATH=/home/lw/models/research:/home/lw/models/research/slim

测试安装是否成功
cd /home/lw/models/research
python deeplab/model_test