Custom Implementation Guide
This section will guide you to implement your own CLArena components, including:
| Component | Description |
|---|---|
| CL Dataset | Continual learning dataset |
| MTL Dataset | Multi-task learning dataset |
| STL Dataset | Single-task learning dataset |
| CL Algorithm | Continual learning algorithm |
| CUL Algorithm | Continual unlearning algorithm |
| MTL Algorithm | Multi-task learning algorithm |
| STL Algorithm | Single-task learning algorithm |
| Backbone Network | The backbone neural network architecture |
| Metrics | The evaluation metrics used to assess model performance |
| Callbacks | The callbacks (additional operation hooks) used during training and evaluation |
Implement CL Modules Outside the Package
You can implement your custom components anywhere outside the package source code, and specify them in your configs for experiments and evaluations.
To properly locate your implement module for the config, you need to add the path of your module to the environment variable PYTHONPATH. This should be done in the beginning of every terminal sessions. Either do it once by:
export PYTHONPATH=<path-to-your-module>or set it every time the command is executed:
PYTHONPATH=<path-to-your-module> clarena ...Contributing to CLArena
If you’re interested in contributing to this package with your custom modules, feel free to submit a pull request!