Prerequisite: You need CUDA enabled GPU.
Note: This method installs DFL on Centos 7/8 or Ubuntu 18.04 and provides an option to use Anaconda
To install CentOS, refer here
To install Ubuntu, [coming soon]
The directory structure is important.
This is how the project tree will look like:
DeepFaceLab_NVIDIA_Linux
|-- _internal
| |-- DeepFaceLab # Created via git clone command (Step #2)
| |-- pretrain_CelebA # Manually copied after extracting .EXE (Step #3)
| `-- pretrain_Quick96 # Manually copied after extracting .EXE (Step #3)
|-- install_script # Created via git clone command (Step #0)
|-- scripts # Created via git clone command (Step #0)
`-- workspace # Created through one of the scripts in /scripts
Clone this repository
git clone https://github.com/jwchoi09/DeepFaceLab_NVIDIA_Linux.git
cd DeepFaceLab_NVIDIA_Linux
Install NVIDIA driver and Dependencies
Navigate to DeepFaceLab_NVIDIA_Linux/install_script
directory and run:
- CentOS:
bash 1_CentOS_install_dependencies.sh
- Ubuntu:
bash 1_Ubuntu_install_dependencies.sh
on Anaconda: (recommended)
Install AnacondaURL #4: https://www.anaconda.com/distribution/#linux
URL #5: https://docs.anaconda.com/anaconda/install/linux/Note: Download Python 3.7 version (at the time of writing: 05/22/2020)
Through the terminal, locate to Anaconda download destination and run the following:bash Anaconda3-2020.02-Linux-x86_64.sh
The installer prompts "In order to continue the installation process, please review the license agreement."
Click Enter to view license terms.
Scroll to the bottom of the license terms and enter "Yes" to agree.
Accept the default install location.
Type "yes" to the prompt, "Do you wish the installer to initialize Anaconda3 by running conda init?"
Close and open your terminal window for the installation to take effect, or you can enter the command:source ~/.bashrc
Create the DeepFaceLab environment with the following libraries in the terminal:
conda create -n deepfacelab -c main python=3.6.8 cudnn=7.6.5 cudatoolkit=10.0.130
Activate the environment
conda activate deepfacelab
on system:
Install python 3.6.8, cudnn 7.6.5, cudatoolkit 10.0.130CentOS:
Ubuntu:
Install Python Dependencies
python3 -m pip install -r ./DeepFaceLab/requirements-cuda.txt
This step involves deleting the already existing DeepFaceLab directory from the _internal
directory, and
cloning DeepFaceLab from the main repository to the _internal
directory
sudo rm -rf DeepFaceLab
git clone https://github.com/iperov/DeepFaceLab.git
Download the latest NVIDIA build(.EXE) from the main repo for Windows
To download, refer to the Google Drive link provided in here: https://github.com/iperov/DeepFaceLab/
Extract the build.
7z x DeepFaceLab_NVIDIA_build_XX_XX_XXX.exe
Once the build has been extracted, there will be _internal
directory.
From it, copy both pretrain_CelebA
and pretrain_Quick96
to the directory DeepFaceLab_NVIDIA_Linux/_internal/
(Further guide coming soon!)