https://github.com/jsflo/kotlin-tensorflow-ext

Kotlin extension functions to wrap the Tensorflow Java API
https://github.com/jsflo/kotlin-tensorflow-ext

Last synced: about 1 month ago
JSON representation

Kotlin extension functions to wrap the Tensorflow Java API

• Host: GitHub
• URL: https://github.com/jsflo/kotlin-tensorflow-ext
• Owner: JsFlo
• Created: 2017-09-10T16:14:31.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-09-11T23:16:37.000Z (over 7 years ago)
• Last Synced: 2024-04-20T17:21:01.372Z (8 months ago)
• Language: Kotlin
• Homepage:
• Size: 11.7 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# kotlin-tensorflow-ext

I've been using the Python API of [Tensorflow](https://www.tensorflow.org/)

recently for creating some simple models. I also have been trying to learn a new

programming language called Kotlin since [Google announced that Kotlin is an officially

                                          supported language for Android Developement](https://www.youtube.com/watch?v=d8ALcQiuPWs).

I noticed that the Java API of Tensorflow is in its early stages and had an ["experimental API"](https://www.tensorflow.org/api_docs/java/reference/org/tensorflow/package-summary)

which felt like the perfect opportunity to try to use and learn

Kotlin's extension functions to provide a better API and to learn the current limitations, compared to the python api, of the Java API.

## Constants

```kotlin

Graph().use { graph ->

    val value = "Hello from  ${TensorFlow.version()}"

    val myConst = graph.addConstant("aConstant", value)

    Session(graph).use { sess ->

        sess.runner()

        .fetch(myConst)

        .runFirstTensor({ output ->

        Assert.assertEquals(value, output.bytesValue().toUTF8String())

        })

    }

}

```

## Placeholders

```kotlin

Graph().use { graph ->

    // Creates a graph for y = a (placeholder) + b (placeholder)

    val a = graph.addPlaceholder("a", DataType.FLOAT)

    val b = graph.addPlaceholder("b", DataType.FLOAT)

    val y = graph.Operation("y", OperationType.ADD, a, b)

    Session(graph).use { sess ->

        val ta = Tensor.create(10f)

        val tb = Tensor.create(10f)

        sess.runner()

        .feed(a, ta)

        .feed(b, tb)

        .fetch(y)

        .runFirstTensor {

            println("${ta.floatValue()} + ${tb.floatValue()} = ${it.floatValue()}")

            Assert.assertEquals(ta.floatValue() + tb.floatValue(), it.floatValue())

        }

    ta.close()

    tb.close()

    }

}

```

## Operations

```kotlin

val a = graph.addPlaceholder("a", DataType.FLOAT)

val b = graph.addPlaceholder("b", DataType.FLOAT)

val y = graph.Operation("y", OperationType.ADD, a, b)

val y = graph.Operation("y", OperationType.SUB, a, b)

val y = graph.Operation("y", OperationType.MUL, a, b)

val y = graph.Operation("y", OperationType.DIV, a, b)

val y = graph.Operation("multiArg", OperationType.Add, a, b, c, d, e, f)

```