No Train No Gain:Revisiting Efficient Training Algrthm for Transformer-BasedLMarxiv.org11 ポイント·投稿者 froster·3 年前·1 コメント
Ask HN: What true thing do you believe that few people agree with you on?2 ポイント·投稿者 froster·4 年前·6 コメント
Ask HN: What is going on the Internet that everyone should be aware of?2 ポイント·投稿者 froster·4 年前·2 コメント
froster·3 年前·議論Recent paper highlights the difficulty of creating a new optimizer as drop-in replacement. Sophia and Lion were recently proposed as superior alternatives to Adam, but appeared worse in an independent eval
froster·4 年前·議論No, I'm not a marketer, I'm a tech student. I just want to know your perspective on this. That is why I'm asking.