ok it is working

it is done

use base class from audiolm-pytorch for type validation

override parameters, so if mulan embed quantizer is passed to audiolm…

… trainers, mulan is not trained

working inference example

complete the residual vector quantization of the joint embedding musi…

…c-text space of mulan, with fetching of learned conditioning embeddings, setup so all three transformers in audiolm can have its own

add decoupled contrastive learning for mulan as an option

complete mulan, lay out a gameplan for remaining week