installation.md

Installation

We provide some tips for XRMoGen installation in this file.

• Installation
  • Requirements
  • Prepare environment

    • a. Create a conda virtual environment and activate it.

    • b. Install PyTorch and torchvision

    • c. Install MMHuman3D

    • d. Install Other Needed Python Packages

  • Another option: Docker Image

Prepare Environment

a. Create a conda virtual environment and activate it.

conda create -n xrmogen python=3.8 -y
conda activate xrmogen
conda install ffmpeg -y

b. Install PyTorch and torchvision

1. Check the version of pytorch-cuda，and select a suitable on at here
2. Use the conda install command for corresponding version to install PyTorch and Torchvision.

conda install pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=10.1 -c pytorch

c. Install MMHuman3D

mkdir ./workspace && cd ./workspace
git clone https://github.com/open-mmlab/mmhuman3d.git
cd mmhuman3d && pip install -e .
cd ../..

d. Install Other Needed Python Packages

pip install -r requirements.txt
pip install -e .

Another option: Docker Image

XRMogen provides a Dockerfile to build the docker image directly

docker build -f ./Dockerfile --rm -t xrmogen .

Attention Users need to make sure that nvidia-container-toolkit is properly installed.

Run the following command to start the docker image

docker run --gpus all -it xrmogen /workspace

Open a teiminal in your host computer, copy project into docker container

docker cp ProjectPath/xrmogen [DOCKER_ID]:/workspace

where [DOCKER_ID] is the docker id that can be obtained by

docker ps -a