This repository contains many small ML experiments that mess around with model architecture. The main focus lies on text processing and all (or most) training experiments are done on text sequence-to-sequence tasks.
The code currently only correctly works under linux, due to limiations coming
from jax. To run it, simply run run_ml_py.sh, which will open up a selection
menu.
The code is structured around the concept of an experiment, which is defined
as an independent program performing some sort of test. The experiments are
defined in their own files and registered in experiments/__init__.py. The rest
of the code is a minimal framework to make creating new experiments easier.