Shortcut-connected Expert Parallelism for Accelerating Mixture-of-Experts

Open in new window