stylegan3-manager

StyleGAN3 Manager ...

Setup

Linux setup assumes Debian/Ubuntu

Mac OS setup requires Homebrew to be installed (https://brew.sh/)

Setup - Install Imagemagick

You have to setup imagemagick on your local machine, here are the instructions:

Setup - Install Imagemagick - Linux

apt-get install

Setup - Install Imagemagick - Mac OS

brew install imagemagick

Setup - Install Imagemagick - check

Check if imagemagick and mogrify image conversion command line utilities are installed correctly by running the which mogrify command (it should print the mogrify command location)

Setup - Local image preparation

NOTE: You don't need to do this step if you are doing the image preparation on the server

Please skip this step if you are a novice

Otherwise if you want full control install nvidia-docker on your machine

Setup - Install Docker / Docker Desktop

Follow the instructions here to setup docker ... TODO... https://docs.docker.com/engine/install/

Setup - Install Nvidia Drivers

Follow the instructions here to install nvidia drivers ... TODO... https://docs.nvidia.com/datacenter/tesla/tesla-installation-notes/index.html#abstract

Setup - Install Nvidia-Docker

Setup - Verify installation

Check both which nvidia-docker and nvidia-smi commands, they shouldn't error.

Congratulations! if you reached here you have a local setup including Nvidia Docker!

Prepare

The first step you need to do is to prepare the dataset, the dataset can be any set of images in the following formats: JPG, PNG, GIF or TIFF

The images will be converted to squared 1024 x 1024 pixel images (PNG RGB) and packaged up in a ZIP file (PyTorch Dataset zip file).

You will use the zip file to train the StyleGAN3 model on your GPU VM and you'll be able to monitor the training process.

Prepare - Prepare Images

The first step is to create a directory named source_images_start and to copy all your dataset images into that directory.

Prepare - Run command

The command you want to execute is:

rake prepare

This command should finish with a success message.

You will notice a file source_images_dataset.zip

Great, you have now a dataset ready and you can start training!

Run Training

To run the training (create and) cd into the directory of your machine learning project.

This directory needs:

• the file source_images_dataset.zip to be present in the directory

• a subdirectory output (which we will use to download and check the files on our local machine)

When you are ready execute:

rake

This will start training.