This repository is archived as all needed code is in the Simapp/Robomaker repository
The DeepRacer Core repository is a utility that is pulling together the different components required for DeepRacer local training. It is not meant for direct usage. If you are looking for the end-user interface to run training locally please go to Deepracer-for-Cloud
The primary components of DeepRacer are four docker containers:
- Robomaker Container: Responsible for the robotics environment. Based on ROS + Gazebo as well as the AWS provided "Bundle". Uses components of AWS Robomaker
- Sagemaker Container: Responsible for training the neural network. Uses components of AWS Sagemaker
- Reinforcement Learning (RL) Coach: Responsible for preparing and starting the Sagemaker environment.
- Log-Analysis: Providing a containerized Jupyter Notebook for analyzing the logfiles generated. Uses Deepracer Utils.
The different sub-modules will include build.sh scripts.
Requirements:
- Scripts made assuming a Ubuntu 20.04 installation.
- New version of Docker (~version 24), including BuildX (
docker-buildx-plugin)
The built Docker Containers can be found on Docker Hub: