TransformerRegression

Transformer model with SOAP optimizer work better than anything else here.

FirstVersion

We use mamba as the neural architecture backbone and non discrete target.
Not working great.