Search papers, labs, and topics across Lattice.
Tsinghua University
1
0
2
Adding depthwise convolutions to Transformers can boost accuracy on downstream tasks while barely increasing model size.
