mnist-variational-autoencoder.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from read_mnist import read_images, read_labels\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def load_dataset(name):\n",
    "    def featurize(image):\n",
    "        return image.astype(float) / 255.0\n",
    "    images = np.array([featurize(image) for image in read_images(name + '-images-idx3-ubyte')])\n",
    "    labels = np.array([label for label in read_labels(name + '-labels-idx1-ubyte')])\n",
    "    return images, labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import os\n",
    "test_in, test_out = load_dataset('../data/t10k')\n",
    "train_in, train_out = load_dataset('../data/train')\n",
    "\n",
    "def random_batch(inputs, outputs, count=64):\n",
    "    indices = np.random.randint(0, len(inputs)-1, count)\n",
    "    return inputs.take(indices, axis=0), outputs.take(indices, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAAA7CAYAAACqsfm3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f74f2ec8790>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "def imshow(im):\n",
    "    plt.figure()\n",
    "    plt.imshow(im, cmap='gray', vmin=0, vmax=1)\n",
    "# imshow(test_in[0])\n",
    "# print test_in[0].shape\n",
    "\n",
    "def imshow_multi(images):\n",
    "    f = plt.figure()\n",
    "    for n, im in enumerate(images):\n",
    "        f.add_subplot(1, len(images), n+1)  # this line outputs images on top of each other\n",
    "        # f.add_subplot(1, 2, n)  # this line outputs images side-by-side\n",
    "        fig = plt.imshow(im, cmap='gray', vmin=0, vmax=1)\n",
    "        fig.axes.get_xaxis().set_visible(False) # this is the worst api in the world\n",
    "        fig.axes.get_yaxis().set_visible(False)\n",
    "    plt.show()\n",
    "\n",
    "imshow_multi(test_in[:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "dropout_keep_prob = tf.placeholder(tf.float32, name='dropout_keep_prob')        \n",
    "\n",
    "def weight_var(shape, stddev=0.1, weight_decay=0, name=None):\n",
    "    initial = tf.truncated_normal(shape, stddev=stddev)\n",
    "    v = tf.Variable(initial, name=name)\n",
    "    if weight_decay > 0:\n",
    "        l2 = tf.nn.l2_loss(v) * weight_decay\n",
    "        tf.add_to_collection('losses', l2)\n",
    "    return v\n",
    "\n",
    "def leaky_relu(x, leak=0.2, name=\"lrelu\"):\n",
    "    with tf.variable_scope(name):\n",
    "        f1 = 0.5 * (1 + leak)\n",
    "        f2 = 0.5 * (1 - leak)\n",
    "        return f1 * x + f2 * abs(x)\n",
    "\n",
    "def relu(x):\n",
    "    return leaky_relu(x)\n",
    "\n",
    "def create_fc(input, out_size):\n",
    "    # input_dropped = tf.nn.dropout(input, dropout_keep_prob)\n",
    "    in_size = input.get_shape()[-1].value\n",
    "    w = weight_var([in_size, out_size], weight_decay=0.004)\n",
    "    b = weight_var([out_size], weight_decay=0.004)\n",
    "    x = tf.matmul(input, w)\n",
    "    return relu(x + b)\n",
    "\n",
    "def create_conv(input, out_channels, patch_size=5, stride=1, batch_norm=False, dropout=False):\n",
    "    in_channels = input.get_shape()[-1].value\n",
    "    w = weight_var([patch_size, patch_size, in_channels, out_channels])\n",
    "    b = weight_var([out_channels], stddev=0)\n",
    "    conv = tf.nn.conv2d(input, w, strides=[1,stride,stride,1], padding='SAME')\n",
    "    if batch_norm: conv = create_batch_norm(conv)\n",
    "    activation = relu(conv + b)\n",
    "    if dropout: activation = create_dropout(activation)\n",
    "    return activation\n",
    "\n",
    "def create_deconv(input, input_shape, out_channels, patch_size=5, stride=1, use_relu=True):\n",
    "    # for best results, patch_size should be a multiple of stride (or vice-versa?)\n",
    "    input_w, input_h, input_channels = input_shape[-3:]\n",
    "    \n",
    "    w = weight_var([patch_size, patch_size, out_channels, input_channels])\n",
    "    b = weight_var([out_channels], stddev=0)\n",
    "    \n",
    "    batch_size = tf.shape(input)[0]\n",
    "    output_shape = tf.pack([batch_size, input_w*stride, input_h*stride, out_channels])\n",
    "    \n",
    "    deconv = tf.nn.conv2d_transpose(input, w, output_shape, strides=[1,stride,stride,1], padding='SAME')\n",
    "    \n",
    "    activation = relu(deconv + b) if use_relu else deconv + b\n",
    "    return activation\n",
    "\n",
    "def create_max_pool(inputs, ksize=2, stride=2):\n",
    "    return tf.nn.max_pool(inputs, ksize=[1, ksize, ksize, 1], strides=[1, stride, stride, 1], padding='SAME')\n",
    "\n",
    "def create_batch_norm(inputs):\n",
    "    return batch_norm(inputs, is_training=TRAIN)\n",
    "\n",
    "def create_dropout(inputs):\n",
    "    return tf.nn.dropout(inputs, dropout_keep_prob)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "encoded_size = 10\n",
    "# batch_size = 64\n",
    "\n",
    "def encoder(images):\n",
    "    images = tf.reshape(images, [-1, 28, 28, 1])\n",
    "    images = create_conv(images, 32, patch_size=5, stride=2)\n",
    "    images = create_conv(images, 32, patch_size=5, stride=2) # now 7x7x32\n",
    "    images = tf.reshape(images, [-1, 7*7*32])\n",
    "    fc1 = create_fc(images, 1500)\n",
    "    # fc1 = create_dropout(fc1)\n",
    "    z_mean = create_fc(fc1, encoded_size)\n",
    "    z_stddev = create_fc(fc1, encoded_size)\n",
    "    return z_mean, z_stddev\n",
    "\n",
    "def decoder(vectors):\n",
    "    fc1 = create_fc(vectors, 1500)\n",
    "    fc2 = create_fc(fc1, 7*7*32)\n",
    "    images = tf.reshape(fc2, [-1, 7, 7, 32])\n",
    "    \n",
    "    dc1 = create_deconv(images, [7,7,32], 16, patch_size=2, stride=2) # now 14x14x16\n",
    "    dc2 = create_deconv(dc1, [14,14,16], 1, patch_size=2, stride=2, use_relu=False) # now 28x28x1\n",
    "    dc2 = tf.sigmoid(dc2)\n",
    "    return tf.reshape(dc2, [-1, 28, 28])\n",
    "\n",
    "def sample_z(z_mean, z_stddev):\n",
    "    samples = tf.random_normal(tf.shape(z_stddev), 0, 1, dtype=tf.float32)\n",
    "    return z_mean + samples * z_stddev\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "images = tf.placeholder(tf.float32, [None, 28, 28], name='images')\n",
    "z_mean, z_stddev = encoder(images)\n",
    "z_vals = sample_z(z_mean, z_stddev)\n",
    "\n",
    "z_input = tf.placeholder(tf.float32, [None, encoded_size], name='z_input')\n",
    "use_z_input = tf.placeholder(tf.int32, shape=[], name=\"use_z_input_condition\")\n",
    "decoder_input = tf.cond(use_z_input > 0, lambda: z_input, lambda: z_vals)\n",
    "\n",
    "reconstructions = decoder(decoder_input)\n",
    "use_log_loss = False # this seems to cause nan loss ...?\n",
    "if use_log_loss:\n",
    "    images_flat = tf.reshape(images, [-1, 28*28])\n",
    "    reconstructions_flat = tf.reshape(reconstructions, [-1, 28 * 28])\n",
    "    image_similarity_loss = -tf.reduce_sum(images_flat * tf.log(1e-8 + reconstructions_flat) + (1-images_flat) * tf.log(1e-8 + 1 - reconstructions_flat) ,1)\n",
    "    image_similarity_loss = tf.reduce_mean(image_similarity_loss)\n",
    "else:\n",
    "    image_similarity_loss = tf.reduce_mean(tf.nn.l2_loss(tf.reshape(images, [-1, 28*28]) - tf.reshape(reconstructions, [-1, 28*28])))\n",
    "kl_divergence = tf.reduce_mean(0.5 * tf.reduce_sum(tf.square(z_mean) + tf.square(z_stddev) - tf.log(tf.square(z_stddev)) - 1, 1))\n",
    "loss = image_similarity_loss + kl_divergence\n",
    "\n",
    "random_vecs = tf.random_normal([16, encoded_size], 0, 1, dtype=tf.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "learn_rate = tf.placeholder(tf.float32, name='learning_rate')\n",
    "optimizer = tf.train.AdamOptimizer(learn_rate)\n",
    "global_step = tf.contrib.framework.get_or_create_global_step()\n",
    "train_step = optimizer.minimize(loss, global_step=global_step)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Restored from checkpoint models/mva2/model.ckpt-1500\n"
     ]
    }
   ],
   "source": [
    "save_path = 'models/mva2'\n",
    "\n",
    "session = tf.Session()\n",
    "init_op = tf.group(tf.global_variables_initializer(), tf.local_variables_initializer())\n",
    "session.run(init_op)\n",
    "\n",
    "saver = None\n",
    "if save_path:\n",
    "    if not os.path.exists(save_path):\n",
    "        os.mkdir(save_path)\n",
    "    saver = tf.train.Saver()\n",
    "    ckpt = tf.train.get_checkpoint_state(save_path)\n",
    "    if ckpt and ckpt.model_checkpoint_path:\n",
    "        saver.restore(session, ckpt.model_checkpoint_path)\n",
    "        print 'Restored from checkpoint', ckpt.model_checkpoint_path\n",
    "    else:\n",
    "        print 'Did not restore from checkpoint'\n",
    "else:\n",
    "    print 'Will not save progress'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Step: 100, learn rate: 0.001, loss: 3626.59521484\n",
      "Step: 200, learn rate: 0.001, loss: 2974.48901367\n",
      "Step: 300, learn rate: 0.001, loss: 2626.23779297\n",
      "Step: 400, learn rate: 0.001, loss: 2305.30664062\n",
      "Step: 500, learn rate: 0.001, loss: 2409.29003906\n",
      "Saved\n",
      "Step: 600, learn rate: 0.001, loss: 2419.12280273\n",
      "Step: 700, learn rate: 0.001, loss: 2088.93310547\n",
      "Step: 800, learn rate: 0.001, loss: 2421.25170898\n",
      "Step: 900, learn rate: 0.001, loss: 2088.87207031\n",
      "Step: 1000, learn rate: 0.001, loss: 2287.10229492\n",
      "Saved\n",
      "Step: 1100, learn rate: 0.001, loss: 1976.65808105\n",
      "Step: 1200, learn rate: 0.001, loss: 1949.72607422\n",
      "Step: 1300, learn rate: 0.001, loss: 1749.30273438\n",
      "Step: 1400, learn rate: 0.001, loss: 1770.34277344\n",
      "Step: 1500, learn rate: 0.001, loss: 1817.59130859\n",
      "Saved\n",
      "Step: 1600, learn rate: 0.001, loss: 1675.75683594\n",
      "Step: 1700, learn rate: 0.001, loss: 1800.18798828\n",
      "Step: 1800, learn rate: 0.001, loss: 1890.37573242\n",
      "Step: 1900, learn rate: 0.001, loss: 2101.77905273\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-11-709b7f114507>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     13\u001b[0m         \u001b[0mz_input\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mzeros\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mencoded_size\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     14\u001b[0m         learn_rate: learn_rate_}\n\u001b[0;32m---> 15\u001b[0;31m     \u001b[0mloss_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mglobal_step_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0m_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msession\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mloss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mglobal_step\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrain_step\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     16\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     17\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mglobal_step_\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;36m100\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    764\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    765\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 766\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    767\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    768\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    962\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    963\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m--> 964\u001b[0;31m                              feed_dict_string, options, run_metadata)\n\u001b[0m\u001b[1;32m    965\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    966\u001b[0m       \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_run\u001b[0;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1012\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1013\u001b[0m       return self._do_call(_run_fn, self._session, feed_dict, fetch_list,\n\u001b[0;32m-> 1014\u001b[0;31m                            target_list, options, run_metadata)\n\u001b[0m\u001b[1;32m   1015\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1016\u001b[0m       return self._do_call(_prun_fn, self._session, handle, feed_dict,\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m   1019\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1020\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1021\u001b[0;31m       \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1022\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1023\u001b[0m       \u001b[0mmessage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_run_fn\u001b[0;34m(session, feed_dict, fetch_list, target_list, options, run_metadata)\u001b[0m\n\u001b[1;32m   1001\u001b[0m         return tf_session.TF_Run(session, options,\n\u001b[1;32m   1002\u001b[0m                                  \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1003\u001b[0;31m                                  status, run_metadata)\n\u001b[0m\u001b[1;32m   1004\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1005\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_prun_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msession\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "for i in xrange(1000000):\n",
    "    global_step_ = session.run(global_step)\n",
    "    ti, tl = random_batch(train_in, train_out, count=256)\n",
    "    \n",
    "    learn_rate_ = 0.001\n",
    "    if global_step_ > 5000:\n",
    "        learn_rate_ = 0.0001\n",
    "    \n",
    "    feed_dict = {\n",
    "        dropout_keep_prob: 0.5, \n",
    "        images: ti,\n",
    "        use_z_input: 0,\n",
    "        z_input: np.zeros(shape=(0, encoded_size)),\n",
    "        learn_rate: learn_rate_}\n",
    "    loss_, global_step_, _ = session.run([loss, global_step, train_step], feed_dict=feed_dict)\n",
    "    \n",
    "    if global_step_ % 100 == 0:\n",
    "        print 'Step: {}, learn rate: {}, loss: {}'.format(global_step_, learn_rate_, loss_)\n",
    "        if saver and global_step_ % 500 == 0:\n",
    "            saver.save(session, save_path + '/model.ckpt', global_step=global_step_)\n",
    "            print 'Saved'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAAA7CAYAAACqsfm3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEhIi4P8BS2IAfrJrjKoAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fab1182b350>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAAA7CAYAAACqsfm3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4A0yDx48ePDgMQT4/4++2/QmVFBEAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fab1012b7d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "def reconstruction_demo():\n",
    "    ti, tl = random_batch(train_in, train_out, count=10)\n",
    "    # imshow(ti[0])\n",
    "    feed_dict = {\n",
    "        dropout_keep_prob: 1,\n",
    "        use_z_input: 0,\n",
    "        z_input: np.zeros((0, encoded_size)),\n",
    "        images: ti\n",
    "    }\n",
    "    reconstructed = session.run(reconstructions, feed_dict=feed_dict)\n",
    "    imshow_multi(ti)\n",
    "    imshow_multi(reconstructed)\n",
    "\n",
    "reconstruction_demo()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAAA7CAYAAACqsfm3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ng3YSAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f369f2c8cd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "def generation_demo():\n",
    "    rand = np.random.normal(size=(16, encoded_size))\n",
    "    feed_dict = {\n",
    "        use_z_input: 1,\n",
    "        z_input: rand,\n",
    "        images: np.zeros((0, 28, 28))\n",
    "    }\n",
    "    images_ = session.run(reconstructions, feed_dict=feed_dict)\n",
    "    imshow_multi(images_[:10])\n",
    "\n",
    "generation_demo()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 5.50972652  0.46766824  2.48728442  1.71777141  0.44927651  3.16284323\n",
      "  3.76665998 -2.93620014 -1.27996302  1.05184591]\n",
      "[ 4.16117859  2.51626611  4.85881281  2.96656823 -0.03133375  0.79681724\n",
      "  3.29458308  1.32304287  4.39395094  5.74068451]\n",
      "[ 2.07452369  0.549528    4.79504347  7.09625387  2.87136245  3.77796531\n",
      "  5.73144388 -0.49003765  3.79606915 -0.75874317]\n",
      "[ 9.74973297  1.23781765  2.74592471  7.47834969  4.77851725  3.36044383\n",
      "  1.10687459 -2.58537769 -0.53342634  5.57272959]\n",
      "[-0.52251011  2.63840866  7.58703995  3.07208514  5.62446499 -0.28510398\n",
      "  6.35791731 -1.08272314  0.04777831  0.88749641]\n",
      "[ 0.86131221  3.78881216  0.24704948 -0.06277622  0.11256009  0.24238397\n",
      " -0.08645609  0.54086167  6.01901293  1.69932652]\n",
      "[ 3.4191215   3.39315414  3.5098989   4.23779202  4.20887232  5.07336426\n",
      "  0.07902446 -2.34544611  4.84207439  5.17160988]\n",
      "[ 1.3043797   2.66852355  0.79584235  5.3798008   3.23818445  0.36497191\n",
      "  4.7644043  -0.50749695  5.21184111  1.37258863]\n",
      "[ 9.90791416 -0.40507725  2.39001346  3.02796149  2.04013062  0.84306675\n",
      "  2.61146498  9.17674828  1.79642236  0.90905923]\n",
      "[ 4.58278084  2.60902834  5.01338387  3.10160184  1.82819808  0.65865284\n",
      "  4.24889517  2.16205263  2.678334    7.7526679 ]\n",
      "[ 7.66235781  7.84045458  3.67812395  0.25048068  3.04333854  3.45766401\n",
      "  1.04885769 -2.05838346  1.45828569  4.27376842]\n",
      "[-0.94764042 -0.15719603  1.15040064  7.07688761  1.16448414  2.30655313\n",
      "  9.20170689 -1.12042546  3.08666444  0.36673397]\n",
      "[ 4.28190708  2.8435092   4.575109    2.11466742  2.84034777 -0.11681342\n",
      "  8.20570278 -4.47571802  1.83373988  0.26384899]\n",
      "[ 0.85904747  2.79819918  2.5873785   8.34438992 -0.01080403  3.61529827\n",
      "  0.40323979 -5.14826918 -0.7746976  -0.2329625 ]\n",
      "[ 2.48578191  0.68330026  3.00113773  3.83534312  5.16338825  1.29386091\n",
      "  6.26083994 -8.4579792  -1.87400091  4.34584665]\n",
      "[-0.2304903   0.70833886  2.66412234  0.89207119  3.0407958   2.45954609\n",
      "  5.50883532 -0.53016502  3.13062477  3.83981729]\n"
     ]
    }
   ],
   "source": [
    "ti, tl = random_batch(train_in, train_out, count=16)\n",
    "for vec in session.run(z_vals, feed_dict={images: ti}):\n",
    "    print vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 1.21247959  1.53361011  1.61471784  3.96641254  2.69023848  4.05767488\n",
      "   5.34224939 -7.95813751 -0.0837506   5.23074532]\n",
      " [ 1.49431205  0.98597068  1.86497402  4.47098923  2.58632708  3.64798379\n",
      "   5.1012249  -6.89040804  0.6463843   4.6715498 ]\n",
      " [ 1.77614427  0.43833125  2.11523008  4.97556543  2.48241544  3.23829246\n",
      "   4.86020088 -5.82267809  1.3765192   4.11235428]\n",
      " [ 2.05797672 -0.10930818  2.36548615  5.48014164  2.37850404  2.82860136\n",
      "   4.61917639 -4.75494862  2.10665393  3.55315876]\n",
      " [ 2.33980894 -0.65694773  2.61574221  5.98471832  2.2745924   2.41891003\n",
      "   4.37815189 -3.68721867  2.83678889  2.99396324]\n",
      " [ 2.6216414  -1.2045871   2.86599851  6.48929453  2.170681    2.00921893\n",
      "   4.1371274  -2.61948895  3.5669241   2.43476748]\n",
      " [ 2.90347362 -1.75222659  3.11625433  6.99387074  2.0667696   1.5995276\n",
      "   3.89610291 -1.55175912  4.29705858  1.87557185]\n",
      " [ 3.18530607 -2.29986596  3.36651039  7.49844694  1.96285796  1.18983638\n",
      "   3.65507841 -0.48402938  5.02719355  1.31637621]]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeUAAABFCAYAAAB0W/44AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\naKLJjMj/A+MW2G4qG/QaAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f74b16cfc10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def find_digit(n):\n",
    "    for im, label in zip(test_in, test_out):\n",
    "        if label == n:\n",
    "            return im\n",
    "\n",
    "def blend_demo(digit1, digit2):\n",
    "    im1 = find_digit(digit1)\n",
    "    im2 = find_digit(digit2)\n",
    "    feed_dict = {\n",
    "        use_z_input: 0,\n",
    "        z_input: np.zeros((0, encoded_size)),\n",
    "        images: np.array([im1, im2])\n",
    "    }\n",
    "    v1, v2 = session.run(z_vals, feed_dict=feed_dict)\n",
    "    steps = 7\n",
    "    def blend(p):\n",
    "        return v1 * (1-p) + v2 * p\n",
    "    feed_dict = {\n",
    "        use_z_input: 1,\n",
    "        z_input: np.array([blend(s * 1.0 / steps) for s in range(steps+1)]),\n",
    "        images: np.zeros((0, 28, 28))\n",
    "    }\n",
    "    print np.array([blend(s * 1.0 / steps) for s in range(steps+1)])\n",
    "    # print 'ok'\n",
    "    reconstructed = session.run(reconstructions, feed_dict=feed_dict)\n",
    "    imshow_multi(reconstructed)\n",
    "\n",
    "blend_demo(0, 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12+"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}