🐳 Docker

[Ujstor]

I've noticed that many people have a hard time running this code outside of Linux. Also, there are individuals who won't monetize others' hard work. These are the reasons for this pull request. I am open to suggestions

Install Docker and Nvidia-container-runtime. Place pretrained_models as expected. Docker will ignore models in the building stage.

Building and Running the Docker Image

Run the following command to build and start the Docker image/container::

docker-compose up -d

Open a terminal and create an SSH tunnel to the Docker container:

ssh -L 7860:127.0.0.1:7860 -p 2222 models@localhost

Password: root

Execute ./entrypoint.sh to run the Gradio demo. The Gradio demo will be available at http://localhost:7860.

The SSH tunnel needs to remain active at all times for the Gradio demo to work. Additionally, you need to be inside the Docker container to execute commands.

Before running commands like:

python3 -m magicanimate.pipelines.animation --config configs/prompts/animation.yaml

or any other command that is not a script that automatically activates the environment, make sure to have the conda environment active:

conda activate manimate

A volume is mounted at the root directory and all changes, files, etc. will be saved in the root directory and accessible from the Docker container.