using lstm with cuda gives error

[htoyryla]

Command

python train.py runoja.txt --n_epochs 5000 --model lstm --cuda

gives

Traceback (most recent call last):
  File "train_.py", line 95, in <module>
    loss = train(*random_training_set(args.chunk_len, args.batch_size))
  File "train_.py", line 55, in train
    hidden = hidden.cuda()
AttributeError: 'tuple' object has no attribute 'cuda'

Here https://github.com/spro/char-rnn.pytorch/blob/master/model.py#L38-L39 it looks like the hidden layer is indeed a tuple. Not sure about the correct way to make this work. Anyhow, I tried this

    if args.cuda:
        if args.model == "gru":
            hidden = hidden.cuda()
        else:
            hidden = (hidden[0].cuda(), hidden[1].cuda())

and it appears to work. Except that the same problem seems to exist also in generate.py, too.
This is what I used there:

        if isinstance(hidden, tuple):
            hidden = (hidden[0].cuda(), hidden[1].cuda())
        else:
            hidden = hidden.cuda()

[berkersonmez]

Yes these changes seem to fix it.

[olivatooo]

See my fork, I'm trying to keep updated

https://github.com/olivatooo/char-rnn.pytorch