U-Net: What is VAL_SUBSPLITS used for when training the model?

Hi everyone,

I was just finishing the second assignment from week 3 and wondered what VAL_SUBSPLITS is used for. Is that the number of subsplits of the training data used for cross-validation?

Since I couldn’t quickly find anything on Google, it would be great if someone could give me a answer so that I can stop wondering.

Thank you,
Grit

It will help greatly if you provide a screen capture image that shows where you see this reference.

Thanks for the hint.

This is from Image_segmentation_Unet_v2.ipynb:

I checked the code a little bit, and compared it to some other tutorials.

In this assignment, it isn’t used. I deleted that line and the notebook runs fine without it.

I think it’s only there by accident, maybe it was intended to be used, or it was copied from some other source but not fully implemented.

Typically it would be used if the .fit method was being used to perform validation automatically, as is shown in this example:

Great, thank you very much for the detailed response!