Search papers, labs, and topics across Lattice.
1
0
2
DMuon slashes training time for large models, achieving up to 163x faster optimizer steps while maintaining the benefits of matrix-aware updates.
