Tensorflow 1.0 in general is not famed for its ease of use, nor for being well documented. Tensorflow 2.0 has set out to fix all those issues, and it does make things easier with advent of eager execution and unified RNNs & unified Optimizers etc. There are a lot to take in version 2.0, but without proper context and comparison to the 1.0 version, some features are beyond magical. I will provide proper explanation and context through this blog.
* Graph
* Variable
* Placeholder
* Session
* unified RNNs
* unified Optimizers
* Eager Execution
* Autograph
Tensorflow 2.0 is awesome.