Skip to content

mati3230/MetaMarathon

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 

Repository files navigation

MetaMarathon

Task

Clone the repository and switch to one of the branches named easy, medium and hard. Fill in the gaps which are marked with TODO. Execute script dqn/src/main.py. All scripts can include TODOs. After finishing development, you can download pretrained weights from the Nextcloud-folder and see the agent play. You switch between training and playing process by changing the train-Flag. The slides can be downloaded as pdf from the Nextcloud (filename: "Build your own Hero.pdf").

Aim

Develop an AI which is able to play Atari games. Reinforement Learning with Deep Q Networks is used to train an autonomous agent.

alt text

Requirements

Setup

Install VirtualBox

Download Image

Import Image

  • Click New-Button
  • Type an appropriate name for the image
  • Choose Type Linux and Version Ubuntu (64 Bit)
  • If you are not able to choose 64 bit:
  • Choose Use an existing virtual hard drive file
  • Click on the folder button and browse for the .vdi file which you downloaded in the previous step
  • Click Create

alt text

Start the Image

  • Double-Click on image created in previous step
  • If error occur: make sure Hyper-V is disabled
  • Login as modalg with password 12345678

Start Eclipse

  • Click on Ubuntu-Button and search for eclipse
  • Execute eclipse
  • Click ok when workspace dialog appears

alt text

Run Test Project

  • Open test folder in modalg181-Project
  • Double click on any of the files
  • Execute file with click on run button (run as Python Run)

alt text

Clone this repository and setup eclipse

  • Open explorer
  • Navigate to folder Projekte
  • Right click in explorer and select Open in terminal option
  • Type git clone https://github.com/mati3230/MetaMarathon.git in terminal
  • If closed, start eclipse
  • Right click in PyDev Package Explorer
  • Select New/Project... option

alt text

  • Select PyDev/PyDev Project and click Next

alt text

  • Type project name MetaMarathon
  • Select grammar version 3.0-3.5
  • Select interpreter /home/modalg/miniconda3/envs/modalg/bin/python3.5
  • Press Finish

alt text

Work with the repository

Clone the repository

Update to latest state

  • Open explorer
  • Navigate to folder Projekte
  • Right click in explorer and select Open in terminal option
  • Type git pull origin master in terminal

Switch to branch with certain level of difficulty

  • Open explorer
  • Navigate to folder Projekte
  • Right click in explorer and select Open in terminal option
  • There are the the branches easy, medium and hard. Type git checkout easy in terminal to switch to easy-branch. Type git checkout medium in terminal to switch to medium-branch
  • If closed, start eclipse
  • Left click on the project MetaMarathon you created before in step Clone this repository and setup eclipse and press F5 to refresh the project. Now you should see a dqn folder.

Work with Python and the Frameworks

Python

OpenAI

Tensorflow

Pretrained Model

  • You can download pretrained weights of the model here
  • Place the files to /dqn/src/experiments/Breakout-v0/checkpoints

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages