seq-learning-basic issues #2
Description
Hello,
The attention sin wave has a perfect prediction in the pdf file on Windows but the LSTM has slightly narrowed amplitudes and the CNN has a vertical result which is shrinked in size about 20x. Python 3.6 and latest PyTorch installed on a new machine. I'm not a Python programmer but it seems there is a bug somewhere ..