The tensorflow-ocaml project provides some OCaml bindings for TensorFlow.

These bindings are in a very early stage of their development and are not ready for real-world usage. They are likely to change a lot in the near future. Expect some segfaults if you use them for now.

Installation

Install the dependencies

opam install ocamlbuild ctypes ctypes-foreign core

You may also have to install the python-dev package.

Clone the repo

git clone https://github.com/LaurentMazare/tensorflow-ocaml.git

Install TensorFlow.

Copy the TensorFlow shared library to lib/libtensorflow.so in the cloned repo. You may have to tweak the following command depending on where TensorFlow was installed.

cp ~/.local/lib/python2.7/site-packages/tensorflow/python/_pywrap_tensorflow.so lib/libtensorflow.so

Try executing the examples from the root of the cloned repo

make run

Enjoy!

Examples

• examples/load.ml contains a simple example where the TensorFlow graph is loaded from a file (this graph has been generated by examples/load.py).
• examples/simple.ml and examples/var.ml show how to generate a TensorFlow graph from OCaml and execute it.

Generating the TensorFlow Graph from OCaml

The TensorFlow graph operators are defined in src/graph/ops.ml. This file has been automatically generated from src/gen_ops/ops.pbtxt which comes from the TensorFlow distribution. Only the basic operators are likely to work for now.

Dependencies

• ocaml-ctypes is used for the C bindings.
• Core is only necessary when generating the graph from OCaml, the wrapper itself does not need it.
• The code in the piqi directory comes from the Piqi project. There is no need to install piqi though.