Skip to content

Getting Started

Requirements

  • Python 3.12+
  • Linux (tested on Ubuntu 20.04+), macOS, or Windows (via WSL2)
  • A container runtime: Docker, Podman, or Apptainer (e.g. for rootless runs on HPC)
  • Sufficient hardware resources (recommended ~3 CPUs, ~6 GB RAM per environment)

Installation

pip install deepracer

uv add deepracer

git clone https://github.com/uzairakbar/deepracer.git
cd deepracer
uv venv --python 3.12
uv sync --extra dev
source .venv/bin/activate

First launch downloads the simulator image

On first use, the simulator image is downloaded automatically. It is several GBs, so the first launch may take a few minutes. Later launches reuse the cached image.

Rollout

import gymnasium as gym
import deepracer                        # register the deepracer-v0 env in gym

env = gym.make("deepracer-v0")          # starts a simulator service on demand

observation, info = env.reset()         # reset env, start the episode rollout

print("Start DeepRacer rollout")
total_reward = 0
episode_over = True

while not episode_over:
    observation, reward, terminated, truncated, info = env.step(
        env.action_space.sample()       # take a single step via random action
    )

    total_reward += reward
    episode_over = terminated or truncated

print(f"Episode finished! Total reward: {total_reward}")
env.close()                             # stop, & remove the simulator service

Instantiation via gym.make("deepracer-v0") works out of the box with the packaged defaults. For more details on environment configuration, environment management, or interfacing via Gymnasium, please see the User Guide section.