issue about MDN #1
Description
In line 88 of the file rnn_train.py, why should we subtract v_max first, and then add v_max.squeeze()?
v = (v - v_max).exp().sum(dim=1).log() + v_max.squeeze()
Is there a corresponding formula？
{{ message }}
In line 88 of the file rnn_train.py, why should we subtract v_max first, and then add v_max.squeeze()?
v = (v - v_max).exp().sum(dim=1).log() + v_max.squeeze()
Is there a corresponding formula？