Large loss value and ipynb widget

fastai
1

Is there a way to dynamically expand the widget for large loss values? I am getting the following behavior:

58%20PM

I am using v1.0.22 FYI. Is there an easy way to approach fixing this issue on my end (e.g., make a callback)? Or is this a problem with the widget? FYI, MSE is just the nn.MSELoss() (which is also, redundently, the loss function). I can provide more details if desired.

I tried to look this up in the forums but didn’t see this problem addressed but I may have missed something, so sorry if this has been posted before. Thanks for the help

2

Eh, I didn’t expect losses that large. It’s all done in fast progress so I’ll change the code there (when I have time). Maybe raise an issue on the fastprogress repo so that I know it’s there?

3

Raised an issue. Thanks for the response!