Search papers, labs, and topics across Lattice.
1
0
3
15
Achieving 4.7x to 8.2x higher throughput for trillion-parameter MoE models could redefine the limits of large-scale model training.
