as far as I know, dropout is only applied if the number of layers is greater than 1 right?
How can I apply the same dropout effect if I’m using just one layer?
Is just as simple like this:
self.dropout = nn.Dropout(0.5) ... out, h = self.rnn(X, (h,c)) out = self.dropout(out) out = self.f_c(out)
Does this mean the same dropout mask is applied to each time step? So in every item in the hidden state, the same units are set to zero ?
Is this a common way?
Thanks for helping.