[P] Hydralette: Simple but powerful configs based on dataclasses

Hi r/ML,

i want to share a little side project of mine: [hydralette](https://github.com/ValeKnappich/hydralette). I mainly built this for my own work but thought why not get some feedback and potentially make someone else's work a little easier as well.

I think we all agree that having a flexible configuration is crucial to successful ML experimentation. There are a million python config libraries out there, some dedicated to configs like `hydra` and others that support configs as a convenience feature like …

experimentation feature hydra libraries machinelearning python simple support think transformers