Question about the inference function

[vikigenius]

In your inference function you have

if self.bidirectional or self.num_layers > 1:
    # unflatten hidden state
     hidden = hidden.view(self.hidden_factor, batch_size, self.hidden_size)

hidden = hidden.unsqueeze(0)

Shouldn't the unsequeeze happen only in an else condition if bidirectional is false and number of layers is one? Otherwise if the model uses a bidirectional GRU or has num_layers > 1 you will have a 4d tensor: (1, hidden_factor, batch_size, hidden_size)

[timbmg]

I think you are right. Thanks for pointing out! But now I am thinking about it, that bidirectional option does not make too much sense right now anyway. We'd need another linear layer to combine the two directions. Will add this soon.

[MBAnslow]

Would it be possible to set the bidirectional option to false as the default value in argparse? This took me quite some time to figure out and really, default options should work.

[timbmg]

parser.add_argument('-bi', '--bidirectional', action='store_true')

If --bidirectional is not provided, it is disabled.

[MBAnslow]

Ok, it is just an artifact of how I'm running it then. Thanks for this repo btw, it is very interesting :).