Tensorflow, a coding library in python developed by Google, is a revolution for machine learning (actually it is deep learning) community.
In a few years ago, no many people / professional were really paying good attention to deep learning, due to its lack of a standardized coding style. Google saw the potential at that time and created Tensorflow.
In my opinion, this could probably be the greatest invention from Google. Because of it, we had Alpha Go, we had self-driving car, we had pay the attention that deep learning deserves, and this is awesome.
However, such popularity made so many developers dying to contribute to the development of Tensorflow, and leaving their name in the mighty history. They usually would wrap up and modify Tensorflow library, creating a new "Deep learning coding library", claiming that this library is way more convenient than Tensorflow.
It did get convenient!!! And Google got really cocky by the support and love from the community and accept the libraries as extra supporting API.
This is where the nightmare began. Libraries are Keras, TFlearn, TF contribute, tf-slim provide exactly the same kind of function, but with different guideline and coding style. If you are a user of, say, an Iphone, and you find out that there are four buttons doing exactly the same thing, it would be really annoying right?
This is a complete failure in Tensorflow and users (not developers) are complaining that this is so fxxking annoying.
Here is one example: http://blog.nateharada.com/tensorflow-i-love-you-but
Apple, on the other hand, was always deemed as a late mover. In the last WWDC in June, Apple release the supporting library for machine learning, and it did the right thing. As you can see, only Keras is supported as a deep learning library for tensorflow.
This is what developers / users want. And I didnt expect that Google need to be embarrassed by Apple in its most dominant field.
