Installation¶

Setting up the environment¶

There are three options for installing the dependencies using either docker, singularity or conda depending on your preference.

Using docker¶

If it is not already done, install docker. Follow this tutorial to install it on Ubuntu.

Steps to build your docker image:

Build the dev-image in dev.Dockerfile

sudo docker build --target dev-image -t instadeep/qdax:$USER -f dev.dockerfile .

Save time: use the Makefile

The command make build automatically builds the container.

Setup your image, mount your directory and start developing

To develop code in your IDE while running it inside your container, mount this folder into your container. For example, you might first get the path to your QDax root folder inside an environment variable:

cd ..
export QDAX_PATH=$(pwd)

and then mount QDax into your container:

cd docker
sudo docker run --rm -it -v $QDAX_PATH:/app instadeep/qdax:$USER /bin/bash

Save time: use the Makefile

The command make dev_container automatically runs the image, mounts the qdax directory, setups the GPU and forwards Neptune's credentials. Pretty handy, right?

How to use GPUs with Docker?

Docker also allows you to use GPUs. First, be sure you have already installed the Nvidia drivers on your computer and that you have installed nvidia-docker. Then, to use all your GPUs inside docker run:

sudo docker run --rm -it --gpus all -v $QDAX_PATH:/app instadeep/qdax:$USER /bin/bash

If you want to use only a subset of the available GPUs inside docker (for instance only GPUs with ID 0 and 1), run:

sudo docker run --rm -it --gpus '"device=0,1"' -v $QDAX_PATH:/app instadeep/qdax:$USER /bin/bash

Using conda¶

If it is not already done, install conda from here

Install some necessary packages on your machine

sudo apt-get update
sudo apt-get install -y --no-install-recommends \
                    ffmpeg \
                    libgl1-mesa-dev \
                    libgl1-mesa-glx \
                    libglfw3 \
                    libosmesa6-dev \
                    patchelf \

Clone qdax

git clone git@github.com:adaptive-intelligent-robotics/QDax.git

Create a conda environment with all required libraries
```
cd QDax
conda env create -f environment.yaml
```
Activate the environment and manually install the package qdax
```
conda activate qdaxpy39
pip install -e .
```

Install GPU support

This tutorial only covers the installation on a CPU hardware. If you wish to run the code on GPU, consider using the docker tutorial. You can also install cuda on your machine and follow the instructions to setup jax on GPU from here.