Kindof yeah - predictivity is a question though for larger layers - when trying to scale this up. But yeah, this is a "95% predictor in latent space is a 7x imp

Kindof yeah - predictivity is a question though for larger layers - when trying to scale this up. But yeah, this is a "95% predictor in latent space is a 7x improvement in speed if done right" approach.