Those seem like fair criticisms, although I'm just a beginner with Tensorflow right now. I don't doubt there's room for improvement!
We were building a character based RNN for generating text, with some special tweaks to it. This wasn't a good fit for Keras based on what we were reading, and there was a great straightforward Tensorflow example already out there which we could work off of so it made more sense to us at that point in time.
We were building a character based RNN for generating text, with some special tweaks to it. This wasn't a good fit for Keras based on what we were reading, and there was a great straightforward Tensorflow example already out there which we could work off of so it made more sense to us at that point in time.