draw_without_attention.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DRAW\n",
    "\n",
    "### References:\n",
    "- [tf implementation](https://github.com/ericjang/draw)\n",
    "- [paper](https://arxiv.org/pdf/1502.04623.pdf)\n",
    "- [paper review](https://github.com/tensorflow/magenta/blob/master/magenta/reviews/draw.md)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from read_mnist import read_images, read_labels\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from nbutil import imshow_multi, to_pil\n",
    "import os\n",
    "\n",
    "BATCH_SIZE = 32"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "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\n",
    "\n",
    "test_in, test_out = load_dataset('../data/t10k')\n",
    "train_in, train_out = load_dataset('../data/train')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# train_image = tf.convert_to_tensor(train_in)\n",
    "# train_label = tf.convert_to_tensor(train_out)\n",
    "# image_q, label_q = tf.train.slice_input_producer([train_image, train_label], num_epochs=None, shuffle=True)\n",
    "# image_batch, label_batch = tf.train.shuffle_batch([image_q, label_q], batch_size=BATCH_SIZE, capacity=512, min_after_dequeue=256)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUkAAAFKCAYAAACQHq0GAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUpIkKZEgKUmSlEiQlCRJSiRISpIkJfp/olYqqoaJEWcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f3f1b4adad0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "imshow_multi([train_in[0]])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def weight_var(shape, stddev=0.1, name=None):\n",
    "    initial = tf.truncated_normal(shape, stddev=stddev)\n",
    "    v = tf.Variable(initial, name=name)\n",
    "    return v\n",
    "\n",
    "def create_fc(x, out_size):\n",
    "    w = tf.get_variable(\"w\", [x.get_shape()[1], out_size], initializer=tf.contrib.layers.xavier_initializer()) \n",
    "    b = tf.get_variable(\"b\", [out_size], initializer=tf.constant_initializer(0.0))\n",
    "    return tf.nn.elu(tf.matmul(x, w) + b)\n",
    "\n",
    "\n",
    "def create_linear(x, out_size):\n",
    "    w = tf.get_variable(\"w\", [x.get_shape()[1], out_size], initializer=tf.contrib.layers.xavier_initializer()) \n",
    "    b = tf.get_variable(\"b\", [out_size], initializer=tf.constant_initializer(0.0))\n",
    "    return tf.matmul(x, w) + b\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Z_SIZE = 2\n",
    "LSTM_SIZE = 256\n",
    "STEPS = 5\n",
    "IMG_SIZE = 28\n",
    "BATCH_SIZE = 100\n",
    "EPS = 1e-8\n",
    "\n",
    "image = tf.placeholder(tf.float32, [None, IMG_SIZE, IMG_SIZE], name='image')\n",
    "\n",
    "# when generating images, still need to fill the `image` placeholder with zeros\n",
    "generate_images = tf.placeholder_with_default(tf.constant(0.0), [], name='generate_images')\n",
    "\n",
    "def read_without_attention(target_image, current_canvas, prev_decoder_state):\n",
    "    # when we add attention, the decoder gets to \"ask\" where to attend to in the image\n",
    "    flat_img_shape = [-1, IMG_SIZE * IMG_SIZE]\n",
    "    return tf.concat(1, [tf.reshape(target_image, flat_img_shape), tf.reshape(current_canvas, flat_img_shape)])\n",
    "\n",
    "def write_without_attention(canvas, h_dec, reuse):\n",
    "    with tf.variable_scope('write', reuse=reuse):\n",
    "        write = create_linear(h_dec, IMG_SIZE * IMG_SIZE)\n",
    "        write = tf.reshape(write, [-1, IMG_SIZE, IMG_SIZE])\n",
    "    return canvas + write\n",
    "\n",
    "def encoder(read_data, enc_state, reuse_vars):\n",
    "    with tf.variable_scope('encoder', reuse=reuse_vars):\n",
    "        h_enc, enc_state = encoder_lstm(read_data, enc_state)\n",
    "        return h_enc, enc_state\n",
    "\n",
    "def sample_z(z_mean, z_stddev):\n",
    "    rand = tf.random_normal(tf.shape(z_stddev), 0, 1, dtype=tf.float32)\n",
    "    return z_mean + rand * z_stddev\n",
    "\n",
    "def decoder(z, dec_state, reuse_vars):\n",
    "    # returns a new draw operation\n",
    "    with tf.variable_scope('decoder', reuse=reuse_vars):\n",
    "        h_dec, dec_state = decoder_lstm(z, dec_state)\n",
    "        return h_dec, dec_state\n",
    "\n",
    "encoder_lstm = tf.nn.rnn_cell.LSTMCell(LSTM_SIZE, state_is_tuple=True)\n",
    "decoder_lstm = tf.nn.rnn_cell.LSTMCell(LSTM_SIZE, state_is_tuple=True)\n",
    "\n",
    "# the paper says we should learn the initial encoder/decoder states, but eric jang's implementation doesn't\n",
    "enc_state = encoder_lstm.zero_state(BATCH_SIZE, tf.float32)\n",
    "dec_state = decoder_lstm.zero_state(BATCH_SIZE, tf.float32)\n",
    "\n",
    "z_means = []\n",
    "z_log_sigmas = []\n",
    "canvases = [tf.zeros_like(image)] # canvas = c[t], in the paper\n",
    "\n",
    "for step in xrange(STEPS):\n",
    "    reuse = step > 0\n",
    "    \n",
    "    error_image = image - canvases[-1] # error_image = x_hat, in the paper; jang's implementation uses the sigmoid of c instead\n",
    "    read_data = read_without_attention(image, canvases[-1], dec_state) # r_t in paper\n",
    "    h_enc, enc_state = encoder(read_data, enc_state, reuse)\n",
    "    \n",
    "    with tf.variable_scope('z_mean', reuse=reuse):\n",
    "        z_means.append(create_linear(h_enc, Z_SIZE))\n",
    "    with tf.variable_scope('z_sigma', reuse=reuse):\n",
    "        z_log_sigmas.append(create_linear(h_enc, Z_SIZE))\n",
    "    \n",
    "    real_z = sample_z(z_means[-1], tf.exp(z_log_sigmas[-1]))\n",
    "    generated_z = sample_z(tf.random_normal(tf.shape(real_z), 0, 1, dtype=tf.float32), tf.random_normal(tf.shape(real_z), 0, 1, dtype=tf.float32))\n",
    "    z = real_z * (1 - generate_images) + generated_z * generate_images\n",
    "    \n",
    "    h_dec, dec_state = decoder(z, dec_state, step > 0)\n",
    "    canvases.append(write_without_attention(canvases[-1], h_dec, reuse))\n",
    "\n",
    "# from http://blog.fastforwardlabs.com/post/149329060653/under-the-hood-of-the-variational-autoencoder-in\n",
    "# def kullback_leibler(mu, log_sigma):\n",
    "#     \"\"\"(Gaussian) Kullback-Leibler divergence KL(q||p), per training example\"\"\"\n",
    "#     with tf.name_scope(\"KL_divergence\"):\n",
    "#         # = -0.5 * (1 + log(sigma**2) - mu**2 - sigma**2)\n",
    "#         return -0.5 * tf.reduce_sum(1 + 2 * log_sigma - mu**2 -\n",
    "#                                     tf.exp(2 * log_sigma), 1)\n",
    "\n",
    "def compute_kl_loss(z_means, z_log_sigmas):\n",
    "    # from https://github.com/ericjang/draw/blob/master/draw.py\n",
    "    kl_terms=[0]*STEPS\n",
    "    for t in range(STEPS):\n",
    "        mu2=tf.square(z_means[t])\n",
    "        sigma2=tf.square(tf.exp(z_log_sigmas[t]))\n",
    "        logsigma=z_log_sigmas[t]\n",
    "        kl_terms[t]=0.5*tf.reduce_sum(mu2+sigma2-2*logsigma,1)-STEPS*.5 # each kl term is (1xminibatch)\n",
    "    KL=tf.add_n(kl_terms) # this is 1xminibatch, corresponding to summing kl_terms from 1:T\n",
    "    Lz=tf.reduce_mean(KL) # average over minibatches\n",
    "    return Lz\n",
    "    \n",
    "# image loss:\n",
    "def binary_crossentropy(t,o):\n",
    "    # from https://github.com/ericjang/draw/blob/master/draw.py\n",
    "    return -(t*tf.log(o+EPS) + (1.0-t)*tf.log(1.0-o+EPS))\n",
    "\n",
    "recons = tf.nn.sigmoid(canvases[-1])\n",
    "binarized_image = tf.round(image)\n",
    "image_loss_per_image = tf.reduce_sum(binary_crossentropy(binarized_image, recons))\n",
    "image_loss = tf.reduce_mean(image_loss_per_image)\n",
    "\n",
    "# KL divergence:\n",
    "z_means_tf = tf.pack(z_means)\n",
    "z_log_sigmas_tf = tf.pack(z_log_sigmas)\n",
    "# kl_loss = tf.reduce_mean(kullback_leibler(z_means_tf, z_log_sigmas_tf))\n",
    "kl_loss = compute_kl_loss(z_means_tf, z_log_sigmas_tf)\n",
    "\n",
    "# loss = image_loss + kl_loss * 0.0001\n",
    "loss = image_loss + kl_loss\n",
    "\n",
    "session = tf.InteractiveSession()\n",
    "\n",
    "canvases_tf = tf.pack(canvases)\n",
    "\n",
    "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)\n",
    "\n",
    "session.run(tf.group(tf.global_variables_initializer(), tf.local_variables_initializer()))\n",
    "tf.train.start_queue_runners()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# for v in tf.all_variables():\n",
    "#   print \"%s : %s\" % (v.name,v.get_shape())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def random_batch(inputs):\n",
    "    indices = np.random.randint(0, len(inputs)-1, BATCH_SIZE)\n",
    "    return inputs.take(indices, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Step: 1, loss: 54390.6523438, 54397.828125+-7.17723560333\n",
      "Step: 401, loss: 15539.1564502, 11356.2119141+44.2380485535\n",
      "Step: 801, loss: 10737.9556616, 9954.27539062+46.7574691772\n",
      "Step: 1201, loss: 9460.81701172, 8516.40039062+41.4191894531\n",
      "Step: 1601, loss: 8899.96064819, 8663.578125+50.0029563904\n",
      "Step: 2001, loss: 8495.65726807, 8797.62011719+47.783203125\n",
      "Step: 2401, loss: 8265.23181763, 8289.66015625+45.9290771484\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-9-0930ede5265b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      4\u001b[0m     \u001b[0minputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mrandom_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_in\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m     \u001b[0mfeed\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0mimage\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlearn_rate\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;36m0.01\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m     \u001b[0m_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstep_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mloss_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mil_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkl_\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[0mtrain_step\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mglobal_step\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mimage_loss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkl_loss\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeed\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      7\u001b[0m     \u001b[0mlosses\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mloss_\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mstep_\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;36m400\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\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": [
    "losses = []\n",
    "\n",
    "while True:\n",
    "    inputs = random_batch(train_in)\n",
    "    feed = {image: np.array(inputs), learn_rate: 0.01}\n",
    "    _, step_, loss_, il_, kl_ = session.run([train_step, global_step, loss, image_loss, kl_loss], feed_dict=feed)\n",
    "    losses.append(loss_)\n",
    "    if step_ % 400 == 1:\n",
    "        print \"Step: {}, loss: {}, {}+{}\".format(step_, sum(losses) / len(losses), il_, kl_)\n",
    "        losses = []\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original and final images:\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAADnCAYAAADPTSXjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAA4Aj4/1vMjZCswIwJAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f3ebc3e8e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Drawing steps:\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeUAAABWCAYAAADxGY4NAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMDAYExhSNjAwMDAwGBP8H/3xUwRl4/eBAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f3f1b48ab90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "# Visualize the reconstruction of sample digits:\n",
    "\n",
    "if True:\n",
    "    inputs = random_batch(train_in)\n",
    "    image_, recons_, canvases_ = session.run([binarized_image, recons, canvases_tf], feed_dict={image: np.array(inputs)})\n",
    "    # imshow_multi([image_[0], recons_[0]])\n",
    "    print \"Original and final images:\"\n",
    "    imshow_multi([image_[0], recons_[0]])\n",
    "    print \"Drawing steps:\"\n",
    "    imshow_multi(canvases_[:,0])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeUAAABkCAYAAAC4qyirAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nwcHBwSEk+P+G5Y4OHKRTXQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f3f1e30f110>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "# Random generation:\n",
    "if True:\n",
    "    images_zero = np.zeros_like(np.array(random_batch(train_in)))\n",
    "    images_ = session.run(recons, feed_dict={image: images_zero, generate_images: 1})\n",
    "    imshow_multi(images_[:5])"
   ]
  },
  {
   "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
}