Skip to content

Latest commit

 

History

History
85 lines (60 loc) · 2.17 KB

GETTING_STARTED.md

File metadata and controls

85 lines (60 loc) · 2.17 KB

Getting Started

Prerequisites

  • Linux or macOS
  • Python 3.6+
  • docker
  • CUDA 10.0+
  • (Optional) AWS

Installation

Dockerized environment is encouraged for all DGP contributors and users. Alternatively, you can use python virtual environments. Please see virtual environment setup for instructions.

To setup DGP docker image:

  • You can pull the latest master docker via:

    dgp$ docker pull ghcr.io/tri-ml/dgp:master && docker image tag ghcr.io/tri-ml/dgp:master dgp:latest

    ---or---

  • Build the docker from scratch via:

    dgp$ make docker-build

Inspect if docker image dgp:latest has been pulled or built successfully:

dgp$ docker inspect --type=image dgp:latest

If you get a response, then you already have DGP docker image on the machine!

To check if DGP docker image is built successfully, run the unit tests via:

dgp$ make docker-run-tests

Develop within docker

In order to start development, the quickest way to get started would be use the interactive docker mode via:

dgp$ make docker-start-interactive

The DGP base directory is mounted within the docker container, and gives you a sandbox to develop quickly without needing to set up a local virtual environment.

Within the interactive docker container (after make docker-start-interactive), you can now build the proto definitions (make build-proto) and run the tests (make test) to make sure everything is functioning properly.

dgp$ make build-proto
dgp$ make test

To set up linters wherever you plan on using git from (either on your host or inside the container), run

# If in Docker, run the following; otherwise skip this block.
dgp$ apt update
dgp$ apt install -y git
dgp$ git config --global --add safe.directory /home/dgp

# Set up the linters.
dgp$ make setup-linters

You may need to first pip install pre-commit if you use git in an environment that does not have DGP's requirements-dev.txt installed. Note that if you run linting outside of Docker, ensure that you use the same Python version DGP's Docker environment uses for a consistent linting experience.