JVM, .NET Language Support about tensorflow HOT 19 CLOSED

Although we may not pursue this ourselves, the internal formats used for communication (GraphDef, etc.) are all platform independent protobuf. Thus, it is possible to implement all or part of the tensorflow API in Java while preserving communication compatibility with the existing code. Since a Java version would likely be slower, one useful bit would be a pure inference layer that evaluates graphs but isn't necessarily able to build them; this would allow graphs built in Python and trained in Python / C++ on GPUs to be run from Java servers.

from tensorflow.

Since raw performance is one of the goals of this library I think that pure implementations don't make too much sense.
A prettier API on top of the generated interfaces would be nice though.

from tensorflow.

Is there a canonical / repeatable test suite, so language bindings can have a target and level of confidence? Is there a build server? Could there be a build server?

from tensorflow.

There's a testsuite with fairly good converge, but currently it's mostly Python with a few C++ tests. There's also a lot of functionality for building graphs that's currently Python only, in particular the automatic differentiation functionality, though that doesn't matter for evaluation of graphs in Java. There are plans to move this functionality into the underlying C++ in future, at which point Java SWIG bindings would be more useful for creating graphs.

If someone takes up the Java SWIG challenge, we'd be happy to accept it upstream pending review, etc., at which point it would be part of our continuous testing. The details of accepting contributions is in flux at the moment, but that will stabilize.

from tensorflow.

Also perhaps https://github.com/bytedeco/javacpp can be used to generate bindings.

from tensorflow.

@girving What aspect of automatic differentiation is not available from other languages? Is that just for defining new ops?

from tensorflow.

If you search for RegisterGradient in the python directory, you'll see all the places we define the gradients of various ops. The gradients are defined by Python code, so they aren't available from pure C++ as yet (we hope to change this at some point).

from tensorflow.

I don't fully grasp the architecture/order of operations, but could one populate the gradients using python, while still calling into the C++ code via a bridge?

Alternatively: Run the python code once to generate the gradients for every op, serialize all those gradients to a file, and then other runtimes can read them off and register them?

from tensorflow.

Both of those might work, but I don't know how much of a savings they are compared to doing it the right way and porting the code to C++.

from tensorflow.

I fully agree, just sounding out a temporary solution so I'm not blocked (I imagine it might be a while before all that code is ported)

from tensorflow.

If all the functionality is not available in the C++ core then porting will be very difficult and some cases not possible (with respect to a fully functional version). Also I opened this: #476

from tensorflow.

+1 for Java

from tensorflow.

+1 for java

from tensorflow.

+1 for Java.

from tensorflow.

+1 for Java

from tensorflow.

+1 for Java/Scala

from tensorflow.

+1

from tensorflow.

@girving I want to contribute to tensorflow java api,
I used to implement the PHP frontend of Apache Spark alone.
Where to start?

from tensorflow.

Closing this out as a duplicate of #5 and #18.

from tensorflow.