- Selective weight reset mechanism based on performance
- Population-based training pipeline for MARL
- Performance monitoring and threshold evaluation
- Configurable hyperparameters for resets and evaluations
- Seamless integration with PyTorch
- Support for cooperative and competitive environments