notebooks/graph/graph_00.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import json\n",
    "from keras.models import Model\n",
    "from keras.layers import Input, Permute, Concatenate\n",
    "from keras import backend as K\n",
    "from collections import OrderedDict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def format_decimal(arr, places=6):\n",
    "    return [round(x * 10**places) / 10**places for x in arr]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "DATA = OrderedDict()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### graph 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_seed = 10000\n",
    "data_in_shape = (3, 3, 5)\n",
    "\n",
    "input_layer_0 = Input(shape=data_in_shape)\n",
    "branch_0 = Permute((1, 3, 2))(input_layer_0)\n",
    "\n",
    "input_layer_1 = Input(shape=data_in_shape)\n",
    "branch_1 = Permute((2, 3, 1))(input_layer_1)\n",
    "\n",
    "output_layer = Concatenate()([branch_0, branch_1])\n",
    "model = Model(inputs=[input_layer_0, input_layer_1], outputs=output_layer)\n",
    "\n",
    "data_in = []\n",
    "for i in range(2):\n",
    "    np.random.seed(random_seed + i)\n",
    "    data_in.append(np.expand_dims(2 * np.random.random(data_in_shape) - 1, axis=0))\n",
    "\n",
    "# set weights to random (use seed for reproducibility)\n",
    "weights = []\n",
    "for i, w in enumerate(model.get_weights()):\n",
    "    np.random.seed(random_seed + i)\n",
    "    weights.append(2 * np.random.random(w.shape) - 1)\n",
    "model.set_weights(weights)\n",
    "\n",
    "result = model.predict(data_in)\n",
    "data_out_shape = result[0].shape\n",
    "data_in_formatted = [format_decimal(data_in[i].ravel().tolist()) for i in range(2)]\n",
    "data_out_formatted = format_decimal(result[0].ravel().tolist())\n",
    "\n",
    "DATA['graph_00'] = {\n",
    "    'inputs': [{'data': data_in_formatted[i], 'shape': data_in_shape} for i in range(2)],\n",
    "    'weights': [{'data': format_decimal(w.ravel().tolist()), 'shape': w.shape} for w in weights],\n",
    "    'expected': {'data': data_out_formatted, 'shape': data_out_shape}\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### export for Keras.js tests"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "filename = '../../test/data/graph/00.json'\n",
    "if not os.path.exists(os.path.dirname(filename)):\n",
    "    os.makedirs(os.path.dirname(filename))\n",
    "with open(filename, 'w') as f:\n",
    "    json.dump(DATA, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\"graph_00\": {\"inputs\": [{\"data\": [0.09094, -0.656269, 0.705864, -0.639553, 0.1774, -0.75845, 0.109398, -0.835948, 0.98541, -0.494561, 0.06592, 0.784278, -0.515813, -0.164759, 0.735891, -0.42734, -0.724601, -0.898413, 0.856582, 0.5599, 0.388949, 0.906877, 0.863885, -0.585201, -0.553286, 0.995037, -0.015873, -0.54845, 0.975432, 0.513181, 0.165166, 0.132054, -0.256313, -0.094073, 0.634537, -0.795987, 0.834313, -0.912258, 0.149097, -0.639002, 0.271403, 0.788287, 0.31112, -0.49155, 0.193158], \"shape\": [3, 3, 5]}, {\"data\": [0.614539, -0.766986, 0.408571, 0.047078, -0.6846, 0.784464, 0.388013, 0.800908, 0.168812, -0.389806, -0.317231, -0.455732, 0.005666, -0.753868, 0.656084, 0.433935, -0.129529, 0.017374, -0.074927, -0.853786, 0.046388, -0.110576, 0.585628, 0.932795, 0.616368, 0.041046, -0.725788, -0.466342, -0.866621, -0.401956, -0.83338, -0.006886, -0.895257, -0.871749, -0.763091, 0.195421, 0.652286, -0.133466, 0.539261, -0.252054, -0.385686, -0.272843, -0.743031, -0.914618, 0.082555], \"shape\": [3, 3, 5]}], \"weights\": [], \"expected\": {\"data\": [0.09094, -0.75845, 0.06592, 0.614539, 0.433935, -0.83338, -0.656269, 0.109398, 0.784278, -0.766986, -0.129529, -0.006886, 0.705864, -0.835948, -0.515813, 0.408571, 0.017374, -0.895257, -0.639553, 0.98541, -0.164759, 0.047078, -0.074927, -0.871749, 0.1774, -0.494561, 0.735891, -0.6846, -0.853786, -0.763091, -0.42734, 0.388949, 0.995037, 0.784464, 0.046388, 0.195421, -0.724601, 0.906877, -0.015873, 0.388013, -0.110576, 0.652286, -0.898413, 0.863885, -0.54845, 0.800908, 0.585628, -0.133466, 0.856582, -0.585201, 0.975432, 0.168812, 0.932795, 0.539261, 0.5599, -0.553286, 0.513182, -0.389806, 0.616368, -0.252054, 0.165166, -0.795987, 0.271403, -0.317231, 0.041046, -0.385686, 0.132054, 0.834313, 0.788287, -0.455732, -0.725788, -0.272843, -0.256313, -0.912258, 0.31112, 0.005666, -0.466342, -0.743031, -0.094073, 0.149097, -0.49155, -0.753868, -0.866621, -0.914618, 0.634537, -0.639002, 0.193158, 0.656084, -0.401956, 0.082555], \"shape\": [3, 5, 6]}}}\n"
     ]
    }
   ],
   "source": [
    "print(json.dumps(DATA))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}