https://github.com/wkwan/mkii-subzero-ppo2agent

https://github.com/wkwan/mkii-subzero-ppo2agent

Last synced: 7 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/wkwan/mkii-subzero-ppo2agent
• Owner: wkwan
• Created: 2020-08-07T02:30:32.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-08-07T13:39:04.000Z (over 4 years ago)
• Last Synced: 2023-10-20T23:49:06.297Z (over 1 year ago)
• Language: Python
• Size: 106 KB
• Stars: 41
• Watchers: 4
• Forks: 8
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Mortal Kombat II - Sub-Zero PPO2 Agent

## Video: https://youtu.be/-oUVr_B_cQo

The final Sub-Zero model I demo'ed in the video: https://drive.google.com/file/d/1_jyXBvkc7t1KzemiBqWyAsFkDGeqzsSd/view?usp=sharing  

The second-last model I demo'ed in the video, it's the same one but before I trained it specifically on the Baraka fight (works better on the first 4 fights): https://drive.google.com/file/d/1DZtZyaWInQlageeaBpAjupA9nj6S3kyA/view?usp=sharing

## Setup:

1. Install pip packages and any dependencies.

2. Find a Mortal Kombat II Sega Genesis ROM and install it with ```python3 -m retro.import /path/to/your/ROMs/directory/```

3. Copy the save states and the scenario.json (reward function definition) in gym-retro-files to /your-pip-install-folder/retro/data/stable/MortalKombatII-Genesis

Now you can play around with the train.py and play.py scripts

Refer to the Gym Retro docs for more info: https://retro.readthedocs.io/en/latest/index.html