Clojure and TensorFlow

A very light layer over Java interop for working with TensorFlow.

(ns example.core
  (:require [clojure-tensorflow.ops :as tf]
            [clojure-tensorflow.core :as run]))

;; Define some operations
(def input (tf/constant [[0. 1.] [0. 0.] [1. 1.] [1. 0.]]))
(def weights (tf/variable (repeatedly 2 #(vector (dec (* 2 (rand)))))))

(def shallow-neural-net (tf/sigmoid (tf/matmul input weights)))

;; Run operations
(run/session-run
 [(tf/global-variables-initializer)
  shallow-neural-net])

Usage

1. Add this library your dependencies in project.clj

2. Add a version of TensorFlow to your dependencies

The easiest option is to add [org.tensorflow/tensorflow "1.1.0-rc1"] to your dependencies. However, TensorFlow is fastest when you use a version compiled for your hardware and OS, and this is especially true if you have a GPU.

Read the official Installing TensorFlow for Java guide for more information.

NOTE: TensorFlow requires at least Java 8 to run. This will already be the default on most machines, but if it isn't for you, it's possible to force lein to use it by adding the :java-cmd "/path/to/java" key to your project.clj.

License

Copyright © 2017 Kieran Browne

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.