Lr_find() volatile implications

Hi there! First of all, thanks for the excellent lessons and community.

For my own education, what would cause a volatile learning rate finder result? And on top, what should be done about it?

Can you show us how you set up the model? In particular what model you use, what batch size and what data you are using. This might give us some ideas what’s going on.

Thanks for reaching out–found a solution.

The motivation for this came from:

The prompt is very similar to the rossmann lesson. However, while training data is given at a day-level granularity, the result expects a month-level granularity. I wasn’t correctly taking this goal into account. This oversight meant my model wasn’t learning much at all.