I have already used keras on some applications of the neural network. Yet I am curious how does keras organize its training procedures. Are there any links or tutorials?
Keras is just a front-end for TensorFlow. It makes using TensorFlow easier.
I previous experience is that keras organizes as a tf.data.Dataset object. But I want to bypass this step and input my features directly. Should I convert it to the same object or is an np object okay for training?
Have you attended any of the DLAI courses which use TensorFLow?
Yes I have finished the full DLS.
This link might be helpful.