Hey so I was just following the same code that Jeremy has provided in his notebook in Lesson-6 (Part 1) in course.18 fastai but when I run my model I found out that my Multi Output RNN loss is even worse that the basic pytorch RNN loss. Whereas Jeremy got 3x better on the same code? Can anyone tell me why is this happening?
Code link[https://www.kaggle.com/abhishekydav/recurrent-nn-from-scratch]