Skip to content

edomerli/procgen-PPO

Repository files navigation

procgen-PPO

Project of the course Autonomous and Adaptive Systems from Unibo, 2024.

Requirements

To install the required packages:

pip list --format=freeze > requirements.txt

Training

To train the agent on the cloud (usually with Colab or Kaggle), move the file notebook.ipynb to the chosen platform and run it with the desired configuration.

To train the agent using "local" resources, modify the configuration dictionary in main.py with the desired game, difficulty and hyperparameters. Then run:

python main.py 

I didn't provide the game and difficulty settings as command line parameters as training has been done using cloud GPU resources.

Demo

To run a demo of the trained models, use the command:

python demo.py --game <game> --difficulty <difficulty>

Supported (game, difficulty) configurations are:

  • (coinrun, hard)
  • (miner, easy)
  • (bossfight, easy)
  • (climber, easy)

About

Project of the course Autonomous and Adaptive Systems from Unibo, 2024

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors