{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exporting models from CNTK to ONNX"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this tutorial, we will demonstrate how to export a CNTK model to the ONNX format."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Installation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To export to ONNX, simply make sure you have CNTK 2.3.1 or higher installed. <br>\n",
    "Follow CNTK installation instructions __[here](https://docs.microsoft.com/en-us/cognitive-toolkit/Setup-CNTK-on-your-machine)__."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## API Usage"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To save a CNTK model to the ONNX format, specify the ONNX format in the format parameter of the save function."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Using Python API **  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```python\n",
    "import cntk as C\n",
    "\n",
    "x = C.input_variable(<input shape>)\n",
    "z = create_model(x) #your create model function\n",
    "z.save(<path of where to save your ONNX model>, format=C.ModelFormat.ONNX)\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Exporting in C# **"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```csharp\n",
    "var x = CNTKLib.InputVariable(<specify input variable parameters>);\n",
    "Function z = CreateModel(x); //your create model function\n",
    "z.Save(<path of where to save your ONNX model>, ModelFormat.ONNX);\n",
    "```\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Trying it out with ResNet-20"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's go through an example of exporting a pretrained CNTK model to ONNX."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 1: Prepare a CNTK model to export"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For this tutorial, we will be using a pretrained ResNet-20 model (trained on the CIFAR-10 dataset) from the collection of pretrained CNTK models found [here](https://github.com/Microsoft/CNTK/blob/master/PretrainedModels/Image.md). Download the model to your working directory. (Note that not all of the models found here are exportable to the ONNX format yet.)  \n",
    "Download link: https://www.cntk.ai/Models/CNTK_Pretrained/ResNet20_CIFAR10_CNTK.model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 2: Load the model into CNTK"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import cntk as C"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model_path = \"ResNet20_CIFAR10_CNTK.model\"\n",
    "z = C.Function.load(model_path, device=C.device.cpu())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 3: Export the model to ONNX"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next, export the CNTK model by saving it out to the ONNX format."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "z.save(\"model.onnx\", format=C.ModelFormat.ONNX)"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}