https://github.com/alexiush/ml-agents-configuration

Automated generation and management of config files for Unity ML-Agents
https://github.com/alexiush/ml-agents-configuration

configuration-management ml-agents unity unity-package

Last synced: 10 days ago
JSON representation

Automated generation and management of config files for Unity ML-Agents

• Host: GitHub
• URL: https://github.com/alexiush/ml-agents-configuration
• Owner: Alexiush
• Created: 2025-01-15T14:07:09.000Z (27 days ago)
• Default Branch: master
• Last Pushed: 2025-01-15T14:57:29.000Z (27 days ago)
• Last Synced: 2025-02-01T19:14:43.076Z (10 days ago)
• Topics: configuration-management, ml-agents, unity, unity-package
• Language: C#
• Homepage:
• Size: 2.9 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# ML-Agents Configuration

This tool generates and manages configurations for ML-Agents training scenarios. It provides models for entities like `Behavior`, `NetworkSettings`, etc and composes a `Configuration` with them. 

To preserve the settings components that contain settings for behaviors and configuration can be added to scene. Settings here work explicitly - file is created without defaults omitted.

## Usage

### Installation

The package can be installed through the UPM by the link `https://github.com/Alexiush/ML-Agents-Configuration.git`. 

It also requires [SerializeReferenceExtensions](https://github.com/mackysoft/Unity-SerializeReferenceExtensions) and [VYaml](https://github.com/hadashiA/VYaml) packages.

### Settings

Default path for generated config can be changed in the project settings under `ML Agents Config` tab.

### Config generation

To generate config open `Window/ML Agents Configuration/Configuration Generator` where all the found behaviors will be present as well as environment-level settings. 

Notice, that settings are not preserved unless there are `BehaviorConfig` and `EnvironmentConfig` components in the scene.

## Samples

**Roller** - RollerBall agent environment ready for generation.

## TODO:

* Behavioral Cloning and Self-Play support

* Optimization

## Credits

[SerializeReferenceExtensions](https://github.com/mackysoft/Unity-SerializeReferenceExtensions)

[VYaml](https://github.com/hadashiA/VYaml)