Y_range for vision learner

Greetings, I am conducting regression with a vision learner. My response data vary between 0 and 1. Often my models predict below zero. While I can clamp negative prediction values to zero, this seems an inelegant solution, and I worry that not being able to specify a range for response negatively impacts model training. I see that tabular regression learners do have an argument for y_range. Are there plans to implement y_range for vision learners as well?

1 Like

I found a solution that works for me, described here: Image Regression using fastai