alt
Hacker News
bellowsgulch
•
today at 7:44 PM
•
0 replies
•
view on HN
Self-attention is an O(n^2) operation. You don't want to train on individual characters, or bytes.