Skip to content

WIP Updating the Repo from the AWS Bundle

Chris Rhodes edited this page Jul 6, 2019 · 2 revisions

Before we begin

Make sure your local copy the repo is clean and up to date, git is the easiest way to see what has changed and where. We are going to copy over the files and manually restore the updates by using git diff to see what has changed.

Rough Steps

  1. Download the bundle
  2. Extract bundle and then extract the sub-bundle. We don't need the metadata.tar
  3. All the files we need are in opt/install
  4. Copy all needed files out the sub directories.
  • This is difficult because it might all change and it really requires just knowing what goes where, or just trying. Command Dump
  1. Update all the references to deepracer_simulation_envionrment to deepracer_simulation, or just rename the folder and fix the build. One day I will do that but this command is easier find simulation/ -type f -exec sed -i 's/deepracer_simulation_environment/deepracer_simulation/g' {} \;
  2. Commit
  3. Go back through all the files restoring any changes. I suggest you use a diff tool like vimdiff to do this
  4. Commit
  5. Rebuild sagemaker and robomaker images
  • Robomaker you can just run docker build -t crr0004/deepracer_robomaker:console -f Robomaker-kinetic-debug.docker ./
  • Sagemaker requires you to rebuild sagemaker_containers which is just in the sagemaker_container folder, run python setup.py sdist and then copy the tar.gz out of the dist folder to the sagemaker-rl-tensorflow folder.
  • Then you can run docker build -t crr0004/sagemaker-rl-tensorflow:console --build-arg sagemaker_container=./sagemaker_containers-2.4.4.post2.tar.gz -f Sagemaker-rl.docker ./
aws s3 --profile drrobomaker cp s3://deepracer-managed-resources-us-east-1/deepracer-simapp.tar.gz ./deepracer-simapp.tar.gz
find simulation/ -type f -exec sed -i 's/deepracer_simulation_environment/deepracer_simulation/g' {} \;
You can’t perform that action at this time.