generate more data

[Chris-Lea]

Two points in my view could make the project more perfect. Welcome discussion and correction if I was wrong.

1. When scrolling the window, the last one timestep is forgotten. It could add 1 the for loop in line26 and line45 to add 1 sample, like this "range(self.len_test - seq_len+1)".
2. Rather than scrolling the window in train and test dataset respectively, I think it could be better scrolling the window first and split the train and test dataset. This could add 50 (one seq_len) samples.