Implementation change question

[AlexandruGhiurutan]

What's the difference between applying the loss function after each cell of the RNN vs applying on the entire sequence? In your example you feed only 1 character at a time and the results are good. I tried feeding the entire sequence of 300 characters at a time and the results were very bad. In my mind the 2 approaches are basically the same, but I can't figure out the difference. Any hint would help. Thanks a lot and sorry for posting this as an issue, but i didn't found the answer anywhere else.