Lesson 1 Personal Data Set: Achieving > 90% F1 Score


I have used the tutorial as specified to download about 3k images but needed to gather more images for my project in order to generalize to the problem I’m dealing with.

My question is a general one: I have now about 4.6k images which the optimal learning rate was suggested to be ~ 1e-3. Using this learning rate, batch=16, I’ve seen a consistent increase in f1 and slow (but very spiky) loss decrease.

I’ve now reached the point where i’m still seeing small improvements at 200 epochs which seems quit high. Is this reasonable?

Obviously I could increase batch size, decrease learning rate. But the answer to many ML questions is: “It depends”. Curious to hear other thoughts on other routes to go.


Great work!
You might be able to increase the learning rate as well! You should keep going on- watch the second lesson as well. There’s a method in fastai (lr_find) to set the learning rate based on some analysis. You’ll immediately see faster training-you may be able to reach close to the accuracy after 200 epochs in just a few epochs! After that, you can tweak your model as seems necessary.
Cheers, stay safe!

By the way, I hope you’re doing the 2019 course and not the 2018 course.

I am thanks. I did use the method lr_find actually thanks for bringing it up. That is actually what is leading to my challenge right now. Still having a lot of epochs…