Example of use with Google Cloud Services

[rohit507]

I would like to use this library with Google's Machine Learning service but frankly, I have no idea where to start. Is there an example or at least some pointers on how use any of google's cloud services with this library?

Thanks.

[blackgnezdo]

I didn't use it with Google Cloud. When I ran in a distributed setting the interface to poke is runSessionWithOptions in particular sessionTarget.

Chances are you'll need to upgrade the native TF dependency first.

[rohit507]

I've been looking to upgrading that dependency, but I've hit a snag with operator generation.

Ops have attributes with the "func" type, but that type doesn't seem to have enough information to allow me to turn it into the corresponding haskell type.

In ops.pbtxt I see things like "FilterDataset" which has an attr "predicate" of type "func". There's usually a human readable description of what the type of that function should be, but nothing that I can programmatically extract. Which means, that during opgen I have no clue what the type should be for those attributes. Is there some sort of wrapper structure that might make sense? Possibly just making those new type parameters that would be caught when the graph is being built?

[blackgnezdo]

We do not have support for "func" type and have been disabling them in the blacklist. It is fine to extend this list to allow for the upgrade to proceed and if/when you are feeling ambitious consider the API for supporting such ops.