I’m interested in doing this as well since I’ve got an idea for a webapp that I want to build but unfortunately, I’m struggling with doing the following in Fast.ai (which can be done in pure Pytorch):
-
Online Learning: receiving a single labeled example from the user and training the neural net on just that one example (batch).
-
Regression: unfortunately, I’m having a lot of problems performing regression on images using fast.ai.
Also, I’m very interested in implementing ULMfit in Pytorch since it’s only available on Fast.ai right now.