Training code

[VitorGuizilini-TRI]

Hi, congrats on the great work! Are you still planning to release training code?
Or at least the relevant diffusion methods required to generate the training loss, that should be enough.