I have been very excited since the announcement of keras core and all that it brings to the table. However I stumbled across this question that has been bugging me for some time now.
In keras core, you can write one training loop that works with both pytorch dataloaders or tensorflow datasets, however in pytorch, images are presented in the order CxHxW whiles for tf datasets it is HxWxC. I am wondering if there is going to be any code changes to deal with this descanpcy or if all this is taken care of under the hood by keras with no action from the user?