Search papers, labs, and topics across Lattice.
Huawei Technologies
1
0
2
6
Adding depthwise convolutions to Transformers can boost accuracy on downstream tasks while barely increasing model size.
