Hi everyone. I was surprised to see that searching for “TFlearn” (a relatively new keras-style high-level tensorflow API) in the forum resulted in a null search. Given that it was optimized for tensorflow by google, wouldn’t it be better to use this rather than keras with tensorflow backend (for part 2)? I have not tried it out, but if somebody in the forum has, please let us know about your experience!
TensorFlow since version 1.1 actually contains the Keras API and will support it natively, so afaik there should be no reason to use TFLearn anymore. See https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/keras and https://www.tensorflow.org/api_docs/python/tf/contrib/keras.
You can also read the explanation of François Chollet, Keras author and Google employee, on the issue in Jeremy’s Reddit of this course.
Thanks for the clarification!