Frames White

It isn't exactly building a computational graph of Julia functions. The julia functions (like for all operations available in TensorFlow.jl) return Tensors. So they are building the Computational Graph, which...

> Can we use normal Julia function, such as `apply_mask(V, mask) = V.*mask`, as an node of the computation graph? Generally, no. And as I said before I'm not sure...

Oh does this mean: [this](https://www.tensorflow.org/extend/adding_an_op) I think, maybe, just maybe, this could be done somehow via CXX.jl? @stevengj mentioned this in another issue far away.

use `select(mask, B, A)`. We should probably add this as an second overload to `where` since this keep catches people out. I think it comes down to our `where` does...

`select` is what the operation is called in the libtensorflow. Python tensorflow does basically have a conditional, that results in `where` something hitting the libtensorflow `where` and other times it...

This has broken some tests. https://travis-ci.org/malmaud/TensorFlow.jl/jobs/353723466#L1184 I've not looked too closely as to why, Might the the tests are too fragile, might not.

It bothers me that we have a function that takes a `name` argument and returns a Node. and even if we provide the name argument directly, the returned node does...

Hmm, Can you pop out some python graphs and compare?

This looks good to me. @malmaud can you give this a once over?

In short (and @malmaud will prob come along with the long), because of issues in how the dynamic RNNs stuff is implemented up-stream in the backend, and in the python...