See the mission description at HACKtheMACHINE: Unmanned Track 3.

It's recommended to use VSCode. A dev environment is already configured to run in VSCode terminal.

Add a .env file to the project with your AWS credentials:

AWS_ACCESS_KEY_ID=…
AWS_SECRET_ACCESS_KEY=…
CDK_DEFAULT_ENV_NAME=user-…

If you prefer another editor then checkout .devcontainer/Dockerfile for instructions on setting up a dev environment.

Install Node.js dependencies:

yarn install

Install Python dependencies:

poetry install
poetry run python model/ray/patch.py

Run the following command with your {ACCOUNT_ID}:

aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 763104351884.dkr.ecr.us-east-1.amazonaws.com aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin {ACCOUNT_ID}.dkr.ecr.us-east-1.amazonaws.com

The cdk.json file tells the CDK Toolkit how to execute your app.

If this is the first time a CDK project has been deployed to this particular account, run:

yarn cdk bootstrap

At this point you can now synthesize the CloudFormation template for this code.

yarn cdk synth

To add additional dependencies, for example other CDK libraries, just add them to your package.json file and rerun the yarn install command.

• yarn cdk ls list all stacks in the app

• yarn cdk synth emits the synthesized CloudFormation template

• yarn cdk deploy --outputs-file cdk.outputs.json deploy this stack to your default AWS account/region

• yarn cdk diff compare deployed stack with current state

• yarn cdk docs open CDK documentation

• yarn cdk destroy destroy stack

• yarn sagemaker:algorithm:create create Sagemaker training algorithm

• yarn sagemaker:algorithm:ydescribe describe Sagemaker training algorithm

• yarn sagemaker:algorithm:update update Sagemaker training algorithm

• yarn cdk deploy --outputs-file cdk.outputs.json && yarn sagemaker:algorithm:update deploy and update

Add a .wandb_api_key to the root of the project. Register for an API key at Weights & Biases.

Commands are stored in the scripts section of ./package.json:

• poetry run python model/server/Forge.py generate Generate model environments
• poetry run python model/server/Forge.py train --RENDER=True Train model
• ./model/client/run.sh Visualize model training

Training for the original Neural MMO competition config:

poetry run python model/server/Forge.py generate --config=CompetitionRound1 --TERRAIN_RENDER=True
poetry run python model/server/Forge.py train --config=CompetitionRound1 --RENDER=True

Training for HACKtheMACHINE:

poetry run python model/server/Forge.py train --config=EastPacificOcean --RENDER=True

Running the UI client:

./model/client/run.sh

The website’s source code and README is located in the ./site directory. It's build is output to the ./docs directory and hosted using GitHub Pages.