First, the host driver installation nvidia
Open a terminal, first remove the old drive:
sudo apt-get purge nvidia*
Disable comes nouveau nvidia drivers
sudo gedit /etc/modprobe.d/blacklist.conf
Whether facie Nouveau has been disabled
lsmod | grep again
If you have no show without explanation disabled, or continue to operate below
sudo vim /etc/modprobe.d/blacklist-nouveau.conf # Create a file (Note: Click the i button, now represents the content insertion)
And add the following:
new blacklist
new options modeset = 0
Note: Exit can be used in any of the two commands:
After press esc, or press shift + zz
After press esc, enter ": wq!" The contents inside double quotes
And then update the look:
sudo update-initramfs –u
Nouveau is an acknowledged has been disabled:
lsmod | grep again
No output anything, explained that it had successfully closed.
Close X-window service:
Ctrl + Alt + F1 to switch to the free Desktop command terminal: Here lightdm is your own display manager, it may be gdm, kdm, in the end is which one can use cat / etc / X11 / default-display-manager to view, and then modify, and turn off the display manager. Here below a few small steps recommended using a mobile phone camera, facing the picture to do, because you may not be familiar
sudo service lightdm stop
In this case officially entered the terminal interface:
Login: User accounts
Password: user password
installation:
cd / home / wlh / tmp # guide to download your driver where to put the address
sudo sh NVIDIA-Linux-x86_64-387.12.run
Installation following steps:
(1)accept
(2)contiuned install
Yes just fine after the default installation
Activate the display: (lightdm just my display manager, you might be in front of said gdm)
sudo service lightdm start
Then press Ctrl + Alt + F7 into the desktop operating
Checks for success
nvidia-smi
Two, docker installation
Download: HTTPS: // download.docker.com/lin UX / Ubuntu / dists / xenial / the pool / stable / AMD64 /
containerd.io_1.2.5-1_amd64.deb
docker-what-cli_18.09.4_3-0_ubuntu xenial_amd64.deb
docker-ce_18.09.4 3-0 ~ ~ ubuntu-xenial_amd64.deb
dpkg -i containerd.io_1.2.5-1_amd64.deb
dpkg -i docker-ce-cli_18.09.4_3-0_ubuntu-xenial_amd64.deb
dpkg -i docker-ce_18.09.4~3-0~ubuntu-xenial_amd64.deb
Try using a docker command is not successfully installed.
Establish docker Group:
$ sudo groupadd docker
The current user to join docker Group:
$ sudo usermod -aG docker $USER
Three, nvidia docker installation
If you have nvidia-docker 1.0 installed: we need to remove it and all existing GPU containers
docker volume ls -q -f driver=nvidia-docker | xargs -r -I{} -n1 docker ps -q -a -f volume={} | xargs -r docker rm -f sudo apt-get purge -y nvidia-docker
Add the package repositories
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | \ sudo apt-key add - distribution=$(. /etc/os-release;echo $ID$VERSION_ID) curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | \ sudo tee /etc/apt/sources.list.d/nvidia-docker.list sudo apt-get update
5 perform the download command to download five packets to the current directory, copy these packets to the server.
apt download libnvidia-container1
apt download libnvidia-container-tools
apt download nvidia-container-runtime-hook
apt download nvidia-container-runtime
apt download nvidia-docker2
Executed on the server, dpkg -i libnvidia NVIDIA to install a five packets.
Test nvidia-smi with the latest official CUDA image
docker run --runtime=nvidia --rm nvidia/cuda:9.0-base nvidia-smi
Fourth, the installation cuda9.0
First Quguan network download cuda9.0, 1.6G download the .run file, the download is complete you can officially installed.
Go to the download directory, add permissions to run the file:
chmod +x ./cuda_9.0.176_384.81_linux.run
Run the installation
sudo ./cuda_9.0.176_384.81_linux.run
Start the installation program, press the space bar has been to the last (Ctrl + c can choose to skip), do not worry, 99% of the time, the input accept to accept the terms
Note: The first time a reminder if you install the driver, select "n", the rest are "y"
After the installation is complete you need to add the environment, this step is very important! ! !
gedit ~/.bashrc
Add the following to the last:
export CUDA_HOME=/usr/local/cuda export PATH=$PATH:$CUDA_HOME/bin export LD_LIBRARY_PATH=/usr/local/cuda-9.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
Test whether the installation is successful
nvcc -V
V. Installation cudnn7
Downloaded directly extract, extract will be a cuda folder, there are two documents include and lib64, the inside of the file copy to the appropriate directory / usr / local / cuda / inside lining. If you unpack at the local, do not move. Just add the file to give permission to read!
sudo chmod a+x /usr/local/cuda/include/cudnn.h sudo chmod a+x /usr/local/cuda/lib64/libcudnn*
And then update the Internet:
cd / usr / local / CUDA / lib64 /
sudo chmod + r libcudnn.so.7.0.5 # View .so own version of
sudo LN -sf libcudnn.so.7.0.5. libcudnn.so.7
sudo LN -sf libcudnn libcudnn.so .so.7
sudo ldconfig
View cudnn version, check whether the installed:
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
Sixth, the load packaged mirror
$ docker load<jq_tensorflow1.8-cuda9.0-cudnn7-devel-ubuntu16.04.tar
Seven, start the container
docker run --runtime=nvidia -it -v /home/dock/Downloads:/usr/Downloads name /bin/bash
Start error Solution:
Systemd drop-in file sudo mkdir -p /etc/systemd/system/docker.service.d sudo tee /etc/systemd/system/docker.service.d/override.conf <<EOF [Service] ExecStart= ExecStart=/usr/bin/dockerd --host=fd:// --add-runtime=nvidia=/usr/bin/nvidia-container-runtime EOF sudo systemctl daemon-reload sudo systemctl restart docker Daemon configuration file sudo tee /etc/docker/daemon.json <<EOF { "runtimes": { "nvidia": { "path": "/usr/bin/nvidia-container-runtime", "runtimeArgs": [] } } } EOF sudo pkill -SIGHUP dockerd 再去开启镜像，done。