generate with pre-trained models gives gibberish data

[lpanchma]

When I try to run generate on the pre-trained models included in the repo, I get gibberish output

python3 /sentiment-discovery/generate.py --gen_length 100 --load_model /SDModels/imdb_clf_16.pt
Creating mlstm
/sentiment-discovery/generate.py:112: UserWarning: invalid index of a 0-dim tensor. This will be an error in PyTorch 0.5. Use tensor.item() to convert a 0-dim tensor to a Python number
neuron = neuron[0]

ªS¦.��ßý;Ù·XÜ¿}íø{ÁU=�0é�)ò$EE�Ù¢N'j$¼ûe�ä²0Ã(C#¤±��ß�ÿ[×�fn�xXÔ/s��©

I have torch

import torch
print (torch.version)
0.4.1

[raulpuric]

As I've mentioned on other issues in this repo, we do not provide anymore support for generate.py unfortunately.