Communicate the Change: Efficient Expert Parallelism via Activation Compression Paper • 2611.22215 • Published 3 days ago • 1
McGill-NLP/Nemotron-3.5-Lightning-30B-A3B-SFT-EchoMoE-teacher Text Generation • 32B • Updated 1 day ago • 647 • 1
McGill-NLP/EchoMoE-Nemotron-3.5-Lightning-30B-A3B-s4-r768-c1408 Text Generation • 28B • Updated 1 day ago • 630