RWKV-LM: RNN with transformer-level performance, without using attention (github.com) 4 points by vletal 4y ago ↗ HN
0 comments
[ 3.5 ms ] story [ 12.3 ms ] threadNo comments yet.