Hi,

this post is really useful. I’m trying out TF-2.0 now. I’m not sure I understand how to split the sample I pass to the fit function into a validation and training. In native keras, I’d do model.fit(X,y,validation_split=0.2). Here x is a tensor, so y is not needed. But TF complains if I pass a validation_split parameter. I suppose this feature should be made clearer for end users.