Adam Pocock

That won't give you an appropriate ROC curve as it's not on the same data and doesn't represent how changing the classification threshold would change the false positive & true...

You'll need to use a model which supports generating probabilities, and then you can use the methods on LabelEvaluation to compute the AUC, or LabelEvaluationUtil to compute the ROC curve...

That information isn't stored in the ROC class, the number of correctly classified points is stored in your `LabelEvaluation`.

What model are you using, how big are the inputs and how big are the outputs?

Ok, so you have 6 inputs which are roughly 20-30 ints long? Or have you embedded them externally? And the output is what dimension? Presumably you're closing all the input...

A `SavedModelBundle` exposes a call operator for the main function (and other functions) in the saved model - https://github.com/tensorflow/java/blob/master/tensorflow-core/tensorflow-core-api/src/main/java/org/tensorflow/SavedModelBundle.java#L446, that might be worth exploring. The eager session might be causing...

I believe people eventually got TFDF models to work with TF-Java, there were a bunch of issues with how TFDF was exporting its symbols which were incompatible with TF-Java's build,...

What are the properties of the RandomUniform op? Is it stateful (i.e. the initial seed is used to generate a value which stays in the op and is updated each...

I've created an issue for the spec clarification, see #6408.

The PR probably doesn't want to be merged in it's current state, as we don't have enough insight into how the process is run internally to generate the docs. What...