Pretty basic questions, but I couldn’t understand the docs completely. fit_one_cycle exposes the neural net the data and from the fit_one, I’m assuming it shows the data once. So, What are the 2 parameters you feed in? for example on the third lesson learn.fit_one_cycle(10, slice(lr)). what’s the 10?
10 is the number of epochs (how many times you will go over your training data). The slice is your learning rate. The one cycle policy is a paper by Leslie Smith here:
Detailing how it can speed up training time (number of epochs) with a cyclical learning rate schedule rather than flat
Also see this post: