Note: I am currently busy with other real-life matters, so I will not be able to continuously keep this repository updated. The scripts and installation guide should still be valid for quite some time. However, if you would like the most recent commits and changes, please pull from the main repo and simply copy the "scripts" directory over. The only change not merged into the main branch from this repository is a small fix for older Linux versions running older FFMpeg binaries.

Universal Installation Method

Install Anaconda3

Download the installer https://www.anaconda.com/distribution/#linux.

Initialize conda for your shell.

export PATH=~/anaconda3/bin:$PATH
conda init bash
# Restart your shell

Install DeepFaceLab

conda create -y -n deepfacelab python=3.6.6 cudatoolkit=9.0 cudnn=7.3.1
conda activate deepfacelab
git clone https://github.com/lbfs/DeepFaceLab_Linux.git
cd DeepFaceLab_Linux
python -m pip install -r requirements-cuda.txt

Native Installation Instructions

Installation for Ubuntu 16.04

An installation script has been created to automatically install all of the required dependencies for Ubuntu 16.04. Clone the repository and run ubuntu16.04-cuda9-installer.sh from the root directory of DeepFaceLab_Linux.

Installation for Ubuntu 18.04

Add NVIDIA package repositories

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
sudo apt-get update

Install NVIDIA Video Driver

sudo apt-get install --no-install-recommends nvidia-driver-418

Reboot your system.

For this method, we will create an Ubuntu 16.04 container on your system. In order to do this, we will need to install and configure LXD.

sudo snap install lxd
sudo lxd init
sudo adduser "$USER" lxd

Reboot or logout so the new group membership can take effect

After you have finished installing and configuring lxd to your needs. We will now need to create the container.

echo "root:$UID:1" | sudo tee -a /etc/subuid /etc/subgid #Only run once and never again!
wget https://blog.simos.info/wp-content/uploads/2018/06/lxdguiprofile.txt #Thanks to Simos Xenitellis for his GUI LXC profile!
lxc profile create gui
cat lxdguiprofile.txt | lxc profile edit gui
lxc launch --profile default --profile gui ubuntu:16.04 deepfacelab
# Wait 30s so the environment can fully setup without issue.
# Logging in before the inital setup is done can cause problems.
# The next command will fix the DeepFaceLab GUI to allow it to show up correctly.
lxc exec deepfacelab -- sh -c "echo 'export QT_X11_NO_MITSHM=1' >> /home/ubuntu/.bashrc"

You can now access your container at any time with the following command

lxc exec deepfacelab -- su ubuntu

While in the container, change to your home directory with cd ~\ and then run the installation instructions for Ubuntu 16.04 and you will have created an identical environment.

WARNING: Make sure you install the same video driver in the container as installed in the host!

diegodorado / DeepFaceLab_Linux