Search papers, labs, and topics across Lattice.
1
0
2
6
Sparsifying FFNs can lead to nearly double the decoding speed without sacrificing model quality, thanks to a novel channel-selection strategy.
