PyMLRF is a general framework to aid with research projects developing and applying machine learning algorithms. The package will implement:
- Classes for handling file systems
- Training and validation loops for Pytorch as well as for more specific frameworks including d3rlpy
- Integration with weights and biases
## Motivation
- When experimenting with ML models there are often a number of artifacts that get reused between models;
- It useful to have configs of these in memory and it is difficult to manage the locations of all such artifacts
- Two base classes have been defined
FileHandler
andDirectoryHandler
which tie an object in memory to a disk location
- Model tracking is performed using the
Tracker
andExperiment
class. - The
Tracker
is an