global variables are being used instead of instance variables in `DecoderWithAttention` in "Show, Attend and Tell.ipynb"

[naveenmarthala]

I could identify some places, where the global variables are being used inside DecoderWithAttention, like this line in forward method:

        # flatten image
        encoder_out = encoder_out.view(batch_size, -1, encoder_dim)

I presume encoder_dim should have been self.encoder_dim!
and here

        # create tensors to hold word prediction scores and alphas
        predictions = torch.zeros(batch_size, max(decode_lens), vocab_size).to(device)

vocab_size should instead be self.vocab_size.

Please correct me If I got this wrong.