Dino Player Reinforcement Learning Model

This model is built using TensorFlow.js Node.js implementation and it's using the DDQN algorithm to play the Dino Chrome game. Browser automation is done using Puppeteer.

Steps

  1. Clone the repository: git clone https://github.com/gajojr/tfjs-dino-player.git

  2. Change to the project directory: cd tfjs-dino-player

  3. Install the dependencies: npm install

  4. Clone the Dino Chrome game repository: git clone https://github.com/wayou/t-rex-runner.git

  5. Install http-server globally: npm i -g http-server

  6. Change to the t-rex-runner directory: cd t-rex-runner

  7. Start the http-server in the t-rex-runner directory: http-server

  8. Open another command line in the tfjs-dino-player directory.

  9. Run the Dino bot with the desired mode:

  • To train the model, run:
    node dino-bot.js --train
    
  • To play the game, run:
    node dino-bot.js --play