2 ms·From Sparse to Soft Mixtures of Experts. Outperforms Dense/Sparse models2 points by famouswaffles 3y ago