3 ms·As usual LSTMs are shit at generating text. But attention models (like BERT) are a whole different game.by visarga 8y agoAs usual LSTMs are shit at generating text. But attention models (like BERT) are a whole different game.