Lr_find() getting interrupted

Hi, I am new to and ml in general, but every time I run learn.lr_find() it says “interrupted”
I’m using google colab and I do have hardware accelerator set to GPU. It doesn’t always happen after 17 cycles, it has happened as late as 28, and as early as 6.

Any help would be greatly appreciated!

1 Like

Yes, it will end when the train_loss starts increasing too fast (so fast it’s not improving). That is expected :slight_smile:


@muellerzr Ok, but I ran it on the same model (just unfreezing it again) and it stopped this time after 18 cycles, also the graph changed from
Is that really normal?

Yes, as you train that plot will dynamically change due to how different your weights will become