See more
is pa… with the Dataset class. For other formats (csv, numpy etc.) this page has a good write-up, however I suggest using tfrecords as they offer better performance and are easier to integrate with a Tensorflow development pipeline.