Docker.installation.md

#Docker installation notes

It would be much easier to install the application in a docker container. Here the basic steps to run the application in Docker.

Build the base image, and the running image

docker pull nvcr.io/nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04
docker pull nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04

docker buildx build -t ubuntu22.04-cu118-conda:torch2.0.1-py39 -f Dockerfile.cu118.torch2.0.1.py39   .

docker buildx build -t genfaceplus:0219 -f Dockerfile.genface .

Preparing the model checkpoints

Notes:

• https://github.com/yerfor/GeneFacePlusPlus/blob/main/docs/prepare_env/install_guide-zh.md
• https://github.com/yerfor/GeneFacePlusPlus/blob/main/README-zh.md
• Put trainval_dataset.npy to data/binary/videos/May/trainval_dataset.npy
• Unzip audio2motion_vae.zip, motion2video_nerf.zip to checkpoints directory.
• Copy 8 files from BFM2009 to deep_3drecon/BFM/

# Assume that all model files has been downloaded into ~/GeneFace++Models
mkdir -p data/binary/videos/May/
cp ~/GeneFace++Models/trainval_dataset.npy data/binary/videos/May/
(cd checkpoints && unzip ~/GeneFace++Models/audio2motion_vae.zip)
(cd checkpoints && unzip ~/GeneFace++Models/motion2video_nerf.zip)
cp ~/GeneFace++Models/BFM2009/* deep_3drecon/BFM/

Start a docker container

docker run -it --name geneface -p 7869:7860 --gpus all -v ~/.cache:/root/.cache -v ~/workspace/GeneFacePlusPlus:/data/geneface/  genfaceplus:0219 /bin/bash

Activate the inference environment including download other necessary models

Run the following commands in the docker container.

source ~/.bashrc
conda activate pytorch

cd /data/geneface/
export PYTHONPATH=./

export HF_ENDPOINT=https://hf-mirror.com
python inference/genefacepp_infer.py --a2m_ckpt=checkpoints/audio2motion_vae --head_ckpt= --torso_ckpt=checkpoints/motion2video_nerf/may_torso --drv_aud=data/raw/val_wavs/MacronSpeech.wav --out_name=may_demo.mp4

Finally, start the Gradio demo app

Run the following command in the docker container.

export PYTHONPATH=./

python inference/app_genefacepp.py --server 0.0.0.0 --a2m_ckpt=checkpoints/audio2motion_vae --head_ckpt= --torso_ckpt=checkpoints/motion2video_nerf/may_torso

And after that, you can access the server at http://127.0.0.1:7869

And you can quit the docker container at anytime.

Restart the container, and start the service

After the container stopped, you can restart the container to run gradio service at anytime.

docker start geneface
docker exec -it geneface /bin/bash

source ~/.bashrc
conda activate pytorch

cd /data/geneface/
export PYTHONPATH=./

python inference/app_genefacepp.py --server 0.0.0.0 --a2m_ckpt=checkpoints/audio2motion_vae --head_ckpt= --torso_ckpt=checkpoints/motion2video_nerf/may_torso