learning_dynamics

TODO:

Nonlinear encoder and decoder (try combination)
Train curriculum learning with more steps (1to3 3to5)
Update forward function for curriculum learning
LQR!!!!!!!!!!
Better naming for checkpoints and losses