New code example on http://keras.io : an explainer on how to create TFRecord files. TFRecord is an efficient binary data format for machine learning, that helps you manage your datasets at scale.https://keras.io/examples/keras_recipes/creating_tfrecords/ …
-
What do you think of the myriad of APIs for dataset IO; http://tf.data .experimental.{load/save} and the dataset.{cache, checkpoint}? Is building a TFRecordwriter the long term way to do it? Thanks for your work!
