Fetching the paper…

Dense Training, Sparse Inference: Rethinking Training of Mixture-of-Experts Language Models · Around