text-autoencoder
My attempts at applying Soundstream design on learned tokenization of text and then applying hierarchical attention to text generation