Ex11.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exercise 11: Autoencoder Networks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torchvision"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use built-in Pytorch function to download MNIST data in usable format \n",
    "# the images are automatically normalised to a mean of zero and sigma of 1 \n",
    "\n",
    "batch_size = 100\n",
    "\n",
    "train_loader = torch.utils.data.DataLoader(\n",
    "  torchvision.datasets.MNIST('.', train=True, download=True,\n",
    "                             transform=torchvision.transforms.Compose([\n",
    "                               torchvision.transforms.ToTensor(),\n",
    "                               torchvision.transforms.Normalize(\n",
    "                                 (0.1307,), (0.3081,)) # initial mean and width\n",
    "                             ])),\n",
    "  batch_size=batch_size, shuffle=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([100, 1, 28, 28])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADuCAYAAAA+7jsiAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAABrtJREFUeJzt3U2IzX0fx/FzhDwUKWwsLCR7ZqVpJF1WiqLkISl2Jkk2kmTKTknDlIWkKOV5Y2VhSNlYjpSmXE3ZzI2y0TRx7tW9u8733PN0zXyO12v78T/nX653/6t+zjnNVqvVALIsmu8bAKZOuBBIuBBIuBBIuBBIuBBIuBBIuBBIuBBo8VT+cLPZ9M+sYI61Wq1mpz/jiQuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBhAuBFs/3DXSDZcuWlfvu3bvL/a+//mq7bdiwobx279695d5sNsu91WqV+1zqdG/Pnj1ruz158qS89t69e9O6pxSeuBBIuBBIuBBIuBBIuBBIuBBIuBCoOZVzvGazOX+HfnNo/fr15b5jx45yP3fuXLlv27Ztyvc0WzqdlY6NjZX7u3fvpv3emzZtKvetW7eWe/Xf5uTkZHltf39/ud++fbvc51Or1ar/0hqeuBBJuBBIuBBIuBBIuBBIuBBIuBDojznHXbt2bdvtxYsX5bWdzhtn+pnXr1+/tt1u3rxZXvvgwYNy//79e7lPTEyU+48fP8q9snz58nJfvXp1uVefx+3p6SmvHR4eLvddu3aV+3xyjgtdSrgQSLgQSLgQSLgQSLgQSLgQ6I/5XuUrV6603Tqd03by6dOnch8aGir3ly9ftt0+fPgwrXtaCH7+/Fnuvb295d7prPZP5okLgYQLgYQLgYQLgYQLgYQLgYQLgbrmHPfOnTvlfvjw4Tl77+3bt5f7t2/f5uy9F7JDhw6V++Dg4L90J93HExcCCRcCCRcCCRcCCRcCCRcCdc1x0MjISLlXX6Ha6etVO+nr6yv36mtGF7rTp0+33a5duzaj1160qH5u/P79u+02Pj5eXtvpK3fTeeJCIOFCIOFCIOFCIOFCIOFCIOFCoK45x7169Wq5b968ue124sSJGb339evXy/3t27fl3ulMcj5VX7E6lZ9o/SfVOW2j0Wi8evWq7Xbq1Kny2o8fP07nlmJ44kIg4UIg4UIg4UIg4UIg4UIg4UKg5lTO4prN5swO7ubRypUr225nzpwpr718+XK5d/o87/v378v90qVL0762kzVr1pT7hQsXyn3fvn1ttxUrVkzrnv5nbGys3Hfu3Nl2+/z584zeeyFrtVodPyDuiQuBhAuBhAuBhAuBhAuBhAuBhAuB/phz3Mq6devKvdNnPy9evFjuM/3c6kx0OmOey3v78uVLuVfntI1GozE6OjqbtxPDOS50KeFCIOFCIOFCIOFCIOFCIOFCIOe4s6C3t7fcDx48WO7Hjx9vu33//r289vnz5+X+5s2bcj979my59/T0lHul0/n3rVu3pv3a3cw5LnQp4UIg4UIg4UIg4UIg4UIgx0FdrtPXrw4MDEz7tZ8+fVruBw4cmPZr/8kcB0GXEi4EEi4EEi4EEi4EEi4EEi4Eco7b5X79+lXunf7+q5/53LNnT3nt+Ph4ufPPnONClxIuBBIuBBIuBBIuBBIuBBIuBFo83zdAbenSpeU+ODg4p+//+PHjtptz2vnjiQuBhAuBhAuBhAuBhAuBhAuBhAuBfB53gTty5Ei53717t9ybzfqjnY8ePSr3Y8eOtd0mJibKa5ken8eFLiVcCCRcCCRcCCRcCCRcCORjfQtAX19f2+3GjRtz+t5DQ0Pl7shnYfLEhUDChUDChUDChUDChUDChUDChUDOcReA/fv3t91WrVo1o9c+evRouQ8PD8/o9ZkfnrgQSLgQSLgQSLgQSLgQSLgQSLgQyNez/gtOnjxZ7tVPZS5ZsqS8dnR0tNy3bNlS7iw8vp4VupRwIZBwIZBwIZBwIZBwIZBwIZDP486CjRs3lvv58+fLvTqrnZycLK8dGBgod7qTJy4EEi4EEi4EEi4EEi4EEi4EEi4Eco47C86dO1func55Kw8fPiz3+/fvT/u1yeWJC4GEC4GEC4GEC4GEC4GEC4EcB82CkZGRGV3/+vXrtlt/f/+MXpvu5IkLgYQLgYQLgYQLgYQLgYQLgYQLgfzMJiwwfmYTupRwIZBwIZBwIZBwIZBwIZBwIdBUP4/7n0aj8fdc3AjQaDQajf/ru3yn9A8wgIXB/ypDIOFCIOFCIOFCIOFCIOFCIOFCIOFCIOFCoP8CSwpDiYBEIlIAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Helper function to draw MNIST image\n",
    "# expect a 2D tensor as input\n",
    "def draw_image(im):\n",
    "    plt.imshow(im, cmap='gray', interpolation='none')\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "examples = enumerate(train_loader)\n",
    "for batch_idx, (example_data, example_targets) in examples:\n",
    "\n",
    "    # First print the shape of the example_data tensor\n",
    "    print(example_data.shape) # 100 examples, 1 color, 28x28 pixel images\n",
    "    \n",
    "    # Show the first 6 images\n",
    "    for i in range(6):\n",
    "        draw_image(example_data[i][0])\n",
    "        \n",
    "\n",
    "    break # Abort the loop over the input data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# A few useful functions\n",
    "\n",
    "# Produce random numbers as input for generator\n",
    "# torch.randn produces random numbers following \n",
    "# a Gaussian with mean=0 and sigma=1\n",
    "def noise(size):\n",
    "    return torch.autograd.Variable(torch.randn(size, 100))\n",
    "\n",
    "# Translate images to flat vectors (784 = 28*28)\n",
    "def images_to_vectors(images):\n",
    "    return images.view(images.size(0), 784)\n",
    "\n",
    "# Translate flat vectors to images in the format (color, vertical, horizontal)\n",
    "def vectors_to_images(vectors):\n",
    "    return vectors.view(vectors.size(0), 1, 28, 28)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# The code below implements a (technically) fully functional variational autoencoder\n",
    "# However it will need to be a bit more complex to give good performance\n",
    "\n",
    "class VAE(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(VAE, self).__init__()\n",
    "\n",
    "        # Layers for the encoder\n",
    "        self.fc1 = nn.Linear(784, 20)\n",
    "        self.fc2 = nn.Linear(784, 20)\n",
    "        \n",
    "        # Layers for the decoder\n",
    "        self.fc3 = nn.Linear(20,784)\n",
    "\n",
    "        \n",
    "    def encode(self,x):\n",
    "        # Encoder part of the network\n",
    "        # We have one layer to calculate mu and one to calculate sigma\n",
    "        # Ideally we would have a few shared layers for mu and sigma \n",
    "        # and the final layers separate. Of course the intermediate\n",
    "        # layers should also include activiation functions\n",
    "        mu    =  self.fc1(x)\n",
    "        sigma =  self.fc2(x)\n",
    "            \n",
    "        return mu,sigma\n",
    "    \n",
    "    def decode(self, z):\n",
    "        # Decoder part of the network\n",
    "        # This could also be more complex. Make sure to keep the sigmid\n",
    "        # as final actication\n",
    "        return F.sigmoid(self.fc3(z))\n",
    "        \n",
    "    def sample(self, mu, sigma):\n",
    "\n",
    "        # randn_like produces random numbers using a Gauss(0,1) distribution\n",
    "        # as a tensor with the same shape as its argument\n",
    "        eps = torch.randn_like(sigma)\n",
    "        z = mu + eps*sigma\n",
    "        \n",
    "        return z\n",
    "    \n",
    "        \n",
    "    def forward(self, x):\n",
    "        \n",
    "        # this is the forward pass\n",
    "        \n",
    "        # run the encoder\n",
    "        mu,sigma = self.encode(x)\n",
    "        \n",
    "        # sample from Gauss(mu,sigma)\n",
    "        z = self.sample(mu,sigma)\n",
    "                \n",
    "        # and put it into the decoder\n",
    "        out = self.decode(z)\n",
    "\n",
    "        return out, mu, sigma\n",
    "\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')          \n",
    "model = VAE().to(device)\n",
    "\n",
    "# We need two optimisers, one for the generator, one for the discriminator network\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reconstruction loss function\n",
    "def loss_function(recon_x, x, mu, sigma):\n",
    "\n",
    "    # Binary cross entropy\n",
    "    BCE = F.binary_cross_entropy(recon_x, x.view(-1, 784), reduction='sum')\n",
    "    \n",
    "    # Here we are missing the Kullback-Leiber divergence\n",
    "    # You should calculate it following the formula derived in the lecture\n",
    "    # For a tensor T you can call T.pow(2) to square it\n",
    "    # and T.log() to calculate the logarithm\n",
    "    # torch.sum(T) sums over the different axes of a tensor and return a scalar\n",
    "    KLD = 0\n",
    "    \n",
    "    \n",
    "    return BCE + KLD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 0, Batch: 0, Loss: 56618.83203125\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 0, Batch: 500, Loss: -526814025.6027832\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1, Batch: 0, Loss: -1468174.25\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1, Batch: 500, Loss: -746432739.375\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 2, Batch: 0, Loss: -1530067.375\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 2, Batch: 500, Loss: -764647302.125\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 3, Batch: 0, Loss: -1520712.25\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 3, Batch: 500, Loss: -757261826.875\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 4, Batch: 0, Loss: -1468817.5\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 4, Batch: 500, Loss: -741664109.25\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 5, Batch: 0, Loss: -1447511.125\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 5, Batch: 500, Loss: -725033929.125\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 6, Batch: 0, Loss: -1451153.25\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 6, Batch: 500, Loss: -710987843.875\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 7, Batch: 0, Loss: -1415893.5\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 7, Batch: 500, Loss: -699516214.25\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADuCAYAAAA+7jsiAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAEHJJREFUeJzt3UtsFeTWxvG1W1p6L7dCK9AKQlAopGKoQi3eyqUaMdEB0QFO1IkxThw51rkDTdSB0WhiIooQIxoVISlopAhYAohcBARCL7TUtrS0tD2jfXIm77PYkJPj+r7/b/p07bZ7d7nNXqz3zUxOThqAWPL+1z8AgNzRuEBANC4QEI0LBETjAgHRuEBANC4QEI0LBETjAgFNyeWLi4uLJysrK5N5Xp7+78Do6GgyU4/r1d5MXlBQIHNleHhY5lOm6KfR+96q3vuXbdeuXbvlxzYzm5iYkHlJScktf+/y8nKZ9/b2yrysrCyZXb9+XdZ6vNeksLBQ5oODg7f82GNjY8lsYGDAhoeHM/IBLMfGraystC1btiTzqVOnyvqLFy8ms40bN95yrZnZuXPnZF5dXZ3MvP/gHD58WOZz5sy5rXz27NnJzPsDPXjwoMxnzZolc+/x77333mR26NAhWfvggw/KfOvWrTJvampKZqdPn5a1njvuuEPm8+fPl/nevXuTmfd6X7p0KZl98cUXsjaL/1UGAqJxgYBoXCAgGhcIiMYFAsrpU+WJiQn5MXhDQ4OsV58Me5/cepYuXSpz9Qno8uXLZa33qXN3d7fM1XNmpkcuFy5ckLXeqKqqqkrm3qeru3fvTmbeJ/neiK+0tFTm6hPzadOmydqVK1fK3PtUuq+vT+bq76m1tVXWnj17Npnd7JiLd1wgIBoXCIjGBQKicYGAaFwgIBoXCIjGBQLK5HIgel1d3eTrr7+ezOfOnSvr1YrZ+++/L2u9TZPOzk6Zq1ltcXGxrP3jjz9kvmfPHpk/9thjMldzXm8+PTAwIPN3331X5m+++abMDxw4kMzUPNLM/9nq6+tlrja6ioqKZO2nn34qc+816enpkbmaj3///feydsGCBcns888/t66uLnetj3dcICAaFwiIxgUConGBgGhcICAaFwgop7W+q1ev2o4dO5L5mjVrZP2iRYuSmXfA1rfffivzF154QebHjh1LZt6hZ96KmLf219HRIfO///47mT377LO39b29tcAbN27IXJ0Cef78eVlbW1t7y49tZtbe3p7MampqZK23MuiNQYeGhmQ+ffr0ZLZ582ZZq8aL+fn5sjaLd1wgIBoXCIjGBQKicYGAaFwgIBoXCIjGBQLKaY5bXl5uzc3Nydxbb1PHbc6cOVPWPvHEEzL3ZqXr1q1LZt4aViajt6xaWlpk7t1qt2HDhmTmzWm9W+Wef/55mXvHlKqf3XtevAusHnroIZmrlcJNmzbJWu+YU+/vRR2Za6ZXVPft2ydr1azWm6tn8Y4LBETjAgHRuEBANC4QEI0LBETjAgHRuEBAOc1xS0tLrbGxMZn//PPPsn7ZsmXJ7Ndff5W16qhOM5N7wmZmZ86cSWZqH9bM7LvvvpO5ek7M9AzZzGzevHnJTM0LzcwKCgpk7l3x6R1Nq+aZ4+Pjstabzat9WzP9u+3cuVPWeteL9vf3y9y7elXt1HpXl86aNSuZeb9XFu+4QEA0LhAQjQsEROMCAdG4QEA0LhAQjQsElNMct7+/X840vbOR1XzLO+fWe+zW1laZ//jjj8lMzdXMzOrq6mS+du1amXvnMquZo7ePe+XKFZl710X+8MMPMld7q+fOnZO1o6OjMvdmzGqO682vvefNm92rc5PN9BzXu3705ZdfTmZTp06VtVm84wIB0bhAQDQuEBCNCwRE4wIB0bhAQDmNg8bHx62vry+Ze9cuqnGQtyLmfXy/f/9+mauP971xz+rVq2V+1113ydxbb1OjC29koq4PNTP74IMPZK7WHc3MLl68mMy8o0TV34qZ2djYmMxHRkaSmXd8aldXl8zVKqWZHveY6fHlU089JWvV+qv3emfxjgsEROMCAdG4QEA0LhAQjQsEROMCAdG4QEA5zXHz8vKsqKgomW/evFnWqzWviYkJWXvo0CGZd3d3y1ytBQ4MDMhab97oHXHqPb6a3bW1tcnad955R+ZHjx6VuXf8a01NTTLzrvj0ZpLe86rm297aXkVFhcwXLlwoc++IVbVOOWPGDFmrVi293+vfX3dTXwXgH4XGBQKicYGAaFwgIBoXCIjGBQKicYGAcprjmpllMplktmfPHlmr5lcnTpyQtd7OrDfHHRoakrly6tQpmXszw8WLF8t8165dyeztt9+WtZcuXZK5er3M/F3he+65J5l5R+p6P9vt8OadTU1NMm9oaJC5N+fdvn17MnvvvfdkrdoFVjvI/4l3XCAgGhcIiMYFAqJxgYBoXCAgGhcIiMYFAsppjpvJZOQ+rndGrzo7+dFHH5W127Zt0z+cQ+0Cnzx5UtZ6Vx96e6UHDhyQuZrjdnZ2ylpvlurNmNevXy/z5557Lpl55yr/9ddfMvdmlvn5+cnMm+t782nvefGu8VTnNjc2NspadQ2n93pm8Y4LBETjAgHRuEBANC4QEI0LBETjAgHRuEBAOc1xCwoKrKqqKpm/9NJLsl6dReudTfzVV1/J3DvDV80EKysrZe3ly5dl7v3s3q6wenzv3GNvHund1fraa6/JfP78+cns8OHDslbdSWxmVl5eLnO1S+zN1r386tWrMvfOVd64cWMyUz1ipu969l7vLN5xgYBoXCAgGhcIiMYFAqJxgYBoXCCgnMdBc+fOTebesZTNzc3J7HbGOWb+aEFd4+kd9VlSUiJzj7f+ph6/rKxM1tbX18v81VdflXltba3M1XOjVjzNTP6tmJn19vbKXI2D7r//flnrjcnUEalmevXOzGzHjh3J7PHHH5e1alTlHaebxTsuEBCNCwRE4wIB0bhAQDQuEBCNCwRE4wIB5TTHHRwctLa2tmTe2toq648fP57MvDWqwsJCmXvHWqr5mDen9db+vO+tVuPMzGpqapLZtGnTZO2GDRtkfuedd8r8do5I7ejokLXqSFwzszlz5shc/e5erTfX92br3s/+9NNPJ7Nr167JWnVt6/Xr12VtFu+4QEA0LhAQjQsEROMCAdG4QEA0LhAQjQsElNMct6SkxO67775k/vvvv8t6tVs6NDQka6urq2XuzSPVHLi0tFTWevu63t6pt7eqZtjevuyKFStkro4CNfOPKVXXfHr7tN6OtXd8q3rNvdfkdve3vZ/tjTfeSGbe38PN7twqvOMCAdG4QEA0LhAQjQsEROMCAdG4QEA0LhBQTnPcyclJucfo7Z3u3r07mfX19cnamTNnytzbqVVz3P7+/tt6bO8MX28WOzY2lswKCgpkrXfVZU9Pj8y9Wev4+Hgy83ZavdfM23NWs1TvOV+4cKHMT548KXNvP1zNar1aNXs/dOiQrM3iHRcIiMYFAqJxgYBoXCAgGhcIiMYFAqJxgYBymuP29/fbN998k8yXL18u69Uur7oz1MzsxIkTMl+1apXMT58+ncy8M5u9eaM3z/Rm1Gq3tL29/ba+t7dvq+a0ZnoW6+36ejuzXr3aqfX2bb397rq6Opl7z4s619n73mvWrElm3u52Fu+4QEA0LhAQjQsEROMCAdG4QEA0LhBQTuOg/Px8ecSqN/ZQowtvdHD33XfLfNmyZTKvqqpKZt7PffToUZl7VyMeOXJE5mpk413h6a393c4RqGZ6RW14eFjWXrp0SebeFaDqukrvefG+tzdG89YG1fPy22+/ydrPPvssmXlH3mbxjgsEROMCAdG4QEA0LhAQjQsEROMCAdG4QEA5zXHLysps7dq1ydybdy5ZsiSZXbx40f3eSnFxsczVnPfPP/+Utfv375e5NzO8cOGCzNU81Js3enNcb9XSO1J32rRpyezjjz+Wtd6q5ujoqMwHBgaS2ZUrV2StdyTuggULZK6OzDXTa6Leta1dXV3JzJtPZ/GOCwRE4wIB0bhAQDQuEBCNCwRE4wIB0bhAQDnNcW/cuCFnUN4s9fjx48nM2830ZqEtLS0yV/Mxb0bszeW8fV41jzQzGxkZSWZqJ9XMbMaMGTL39pwvX74sc3VMqfd7e8evetdwqiNYvb817/hWbw7sXZVZU1OTzLyrMp988slk5v2bgSzecYGAaFwgIBoXCIjGBQKicYGAaFwgIBoXCCinOe7w8LB1dHQkc7Wra2Z2/vz5ZKb2G838mZ/aGzXT80zvms2VK1fK/PDhwzL3ZrE9PT3JbGJiQtZ6O6/enFbtSJvpGbS3T+vtCvf398tcnQntzZBLSkpkrmbnZmbl5eUy37dvXzJ7+OGHZa26qtabfWfxjgsEROMCAdG4QEA0LhAQjQsEROMCAeU0DiouLrYVK1Yk8127dsl6NdpoaGiQtZlMRubempZaw/LW9urq6mS+atUqmZ84cULm6ghW77hOb2XQuwLUOxZXXfvovSZqJdDMH8Op0Yi31uetgS5cuFDm3ghPraEePHhQ1jY1NSWztrY2WZvFOy4QEI0LBETjAgHRuEBANC4QEI0LBETjAgHlNMcdHx+Xq1jeLFat1nlrfd4RqkeOHJG5mmd6K1yVlZUy9+bA3sxQHXnrrfV5c96ioiKZe/NKdQWodzSst3rnrSSqvxdvRjxr1iyZv/LKKzLfunWrzNXVqvPmzZO16t8ceNeqZvGOCwRE4wIB0bhAQDQuEBCNCwRE4wIB0bhAQDnNcSsqKmz9+vXJXB076fGuNVy6dKnMd+zYIXM1c9yyZYus9XZ9p0zRT6N3jKmaI3tzWu8YUm9v1aufP39+Mquurpa1x44dk7l3Baj63SsqKmTt6tWrZb5t2zaZe6/piy++mMy+/vprWavm9sxxgf/DaFwgIBoXCIjGBQKicYGAaFwgIBoXCCinOW5vb6998sknyXzTpk2yXl1N6F0HuW7dOpl7VzqqOfFbb70la9Us08zf5/XOH1az1vz8fFnrzWG960e9XeLZs2cnM+9KyAULFsjcm1mqq1UXL14sa73Zuzof3Mysvb1d5rt3705mtbW1srazszOZebPtf3/dTX0VgH8UGhcIiMYFAqJxgYBoXCAgGhcIiMYFAsp4+57/qaqqavKZZ55J5t4ZwIsWLUpm3kxP7TCamT3wwAMy//DDD5OZN9NTZwub+fffenfYtra2JjPvbGLvbGNvjuvNQ3/66adkpma8Zv5MUt29a2Z25syZW35sb/buvWaNjY0y379/fzLzfi815/3yyy+tu7tbD/6Nd1wgJBoXCIjGBQKicYGAaFwgIBoXCCintb68vDy5RuaNTc6ePZvMvLHG+fPnZb5kyRKZq2Nlf/nlF1lbV1cn85aWFpl7Ixf1vHgrgd5anjeK2r59u8y9Y1AV7zWrr6+X+dDQUDLzrg/1/h7mzJkjc++438LCwmQ2ffp0WaueU2+NM4t3XCAgGhcIiMYFAqJxgYBoXCAgGhcIiMYFAsppjuvxrsI8ePBgMlNHcZr562neeuKpU6eSmXeMaFlZmczV6puZ2eDgoMzVHNe7otObf4+MjMi8ublZ5jt37kxmPT09stY7Mtdb5VTPi7dS+NFHH8ncu370kUcekXlNTU0y27t3r6xVV3h6c/ss3nGBgGhcICAaFwiIxgUConGBgGhcICAaFwgop+NZM5lMt5md++/9OMD/e3WTk5NV3hfl1LgA/hn4X2UgIBoXCIjGBQKicYGAaFwgIBoXCIjGBQKicYGAaFwgoH8BAE7cDXXNcUMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 8, Batch: 0, Loss: -1352049.375\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 8, Batch: 500, Loss: -692027172.75\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 9, Batch: 0, Loss: -1405118.75\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 9, Batch: 500, Loss: -684016699.125\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# The code below handles the training of the GAN\n",
    "\n",
    "# Total number of epochs to train\n",
    "n_epochs = 10\n",
    "\n",
    "# Loop over epochs\n",
    "for epoch in range(n_epochs):\n",
    "    \n",
    "    # Loop over training examples\n",
    "    examples = enumerate(train_loader)\n",
    "    \n",
    "    train_loss = 0\n",
    "    \n",
    "    for i_batch, (data, _) in examples:\n",
    "\n",
    "        # Prepare input data\n",
    "        data = torch.autograd.Variable(images_to_vectors(data))\n",
    "\n",
    "        # reset readients\n",
    "        optimizer.zero_grad()\n",
    "                \n",
    "        # apply model to data and get the decoder output as well as the\n",
    "        # latent space variables\n",
    "        recon_batch, mu, sigma = model(data)\n",
    "        \n",
    "        # calculate the loss function and update gradients\n",
    "        loss = loss_function(recon_batch, data, mu, sigma)\n",
    "        loss.backward() \n",
    "        optimizer.step()\n",
    "        \n",
    "        train_loss += loss.item()\n",
    "        \n",
    "        # Show status every few batches\n",
    "        if (i_batch) % 500 == 0: \n",
    "        \n",
    "            # throw random numbers and let the autoencoder build images\n",
    "            sample = torch.randn(20).to(device)\n",
    "            sample = model.decode(sample).cpu()\n",
    "    \n",
    "            print(\"Epoch: {0}, Batch: {1}, Loss: {2}\".format(epoch,i_batch,train_loss))\n",
    "            draw_image(sample.view(1, 28, 28).detach().numpy()[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADuCAYAAAA+7jsiAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAD+tJREFUeJzt3UtvFWTXxvG7SGlLC7TQQgvUgpZjLcWKGk9AIIGEYIwmJiYy0LGfQEf6ERw4UhOmDnSg8ZCoQROtQKCClpZDOZRToQVaSktbCu0z6psneXNfV3ebJ++zfP+/6cXq3t17L7dhse67aGpqKgGIZd7/9RMAUDgaFwiIxgUConGBgGhcICAaFwiIxgUConGBgGhcIKD5hfzh8vLyqcrKymz+6NEjWV9aWprN5s3T/w0ZGxuTeXFxscwnJydnlc1ERUWFzAcHB2WunvvIyMicHvuxxx6TufuXcw8fPsxmJSUlc3ps956q92ViYkLWOuqzmFJKRUVFMi8vL89m9+/fl7XqNR0YGEgjIyP6wVOBjVtZWZnee++9bH7nzh1Z39TUlM3Kyspk7enTp2VeW1src/ViuuZwH8Dnn39e5t9++63M6+rqstkff/wha19++WWZq//QpuSbR72na9askbVLliyR+blz52SuntuVK1dkrWu8zZs3y9x9EWzbti2bnThxQtb29fVls08++UTWTuN/lYGAaFwgIBoXCIjGBQKicYGACvpb5eLi4lRTU5PN169fL+uHh4ez2cKFC2Wt+xvM7u5umavnXVVVJWs7Ojpk7p7b2rVrZa7+BnT//v2ytqGhQeaHDh2S+cqVK2Xe09OTzVpaWmTt7du3ZT5/vv74qRHhvn37ZG1bW5vML126JPMVK1bIXP1tu5uuqM+ie02m8Y0LBETjAgHRuEBANC4QEI0LBETjAgHRuEBABc1xR0dHU1dXVzZ380w197t586asdWtYbjtIrVK5eeOuXbtk3tnZKfPly5fLXK1DznXDxs3H3WZUdXV1NnPzyrNnz8pcbUWllNLGjRuzmdtqam5ulvmxY8dm/dgppXTmzJls5rbJvvzyy2w2MDAga6fxjQsEROMCAdG4QEA0LhAQjQsEROMCARU0Dlq0aJE8nOzUqVOy/tatW9nMreW5w8G2b98+68e+fv26rHVjj8bGRpm70cLBgwez2Ycffihr7969K/Pff/9d5m5UpQ74Gx8fl7VuFHX06FGZq8/EG2+8IWvb29tlvnjxYpm7scyCBQtmlaWU0p49e7KZWyGdxjcuEBCNCwRE4wIB0bhAQDQuEBCNCwRE4wIBFbnb2v5dQ0PD1Pvvv5/N1epcSvpmOXe51bp162SuLlJKSc9SFy1aJGvdrXT19fUyd0duqou53JzW3TTo6t1KozrG1D22mp2nlNLOnTtl7ubnivtcu3VG97upGbWbIavazz//PPX29trb+vjGBQKicYGAaFwgIBoXCIjGBQKicYGAaFwgoIL2cScmJtKNGzeyuTtuU80z3az03r17MldXMqakZ7Vujut+tjtadmhoSObqd3dzVrcTq67JTCml8+fPy1w9d3WsbEr+mFL1WUpJ7wK7fVr3WXTveVlZmczV775ly5ZZP7bbYZ7GNy4QEI0LBETjAgHRuEBANC4QEI0LBETjAgEVPMe9du1aNm9tbZX16vxiNyvdtGmTzGtqamReXFyczcrLy2WtO3fZ7QK76yZ7e3uz2eXLl2Wt2pdNyV9H6fZ1i4ryq6Fu59VdEep+t8HBwWzmdqDd7+3m326XWJ1HvWzZsln/bLfTPo1vXCAgGhcIiMYFAqJxgYBoXCAgGhcIqKBxUEVFhbxm012VqcYqe/fuLeSp/C8rVqyQeXV1dTab6xGohw4dkvmZM2dkrkZsDx48kLXuuY2Ojs6pXj2+GhWl5I9XdY/d2dmZze7fvy9r3fGrVVVVMnfHAatrX91qnhpFzfS4ZL5xgYBoXCAgGhcIiMYFAqJxgYBoXCAgGhcIqKA5blFRkVyPc7MzNd9Ss8yZ5G6lUB0V6mZnamaXkp9XuvU1tYLmVuPczNAdoerW39Qc+D89Yx4eHs5m7vOgalPyc1z1OU9Jv+cDAwOytra2Npu592sa37hAQDQuEBCNCwRE4wIB0bhAQDQuEBCNCwRU0Bx3ZGQkHTlyJJtv27ZN1qvrBVeuXClrd+/eLfOJiQmZq7mbu4rSHa967tw5mZeWlspc7Qq3tLTI2p07d8rcva5dXV0yP3r0aDZrb2+Xtf39/TJ375k6ntXt4y5YsEDm7gpQt1uuZtSvvPKKrJ3LjvM0vnGBgGhcICAaFwiIxgUConGBgGhcICAaFwiooDluSUlJamxszObqHNyU9IzKzfReffVVmVdUVMhc7W9+9dVXstbNK91Msa6uTuYHDhzIZq+99pqsdVc6ul3jVatWyVzNoN3+9dDQkMzdLrB67m4P2e0Ku2s03Zx3Lju1GzduzGZlZWWydhrfuEBANC4QEI0LBETjAgHRuEBANC4QEI0LBFTQHHdyclLO3tT+ZEopPfHEE9nMzSNv3LghczdDvnnz5qxr3b6uO6N3z549Mt+/f382W716tax15/+qu1hnUq/OJ3Z7yO5sY3fuspr7P3z4cE4/e6759u3bs5l7TdWur5ttT+MbFwiIxgUConGBgGhcICAaFwiIxgUCKnitb+3atdn87t27sl6tM7lrE0+cOCHzvr4+mR8/fjybuWsy3YqXek1SSumpp56SuRqFzZun/9vqxgfuPfn5559l/vXXX2eze/fuyVr3urmjSEtKSrJZQ0ODrJ0/X3+0y8vLZa7Ghynptb7r16/LWnUNJ9dsAv9gNC4QEI0LBETjAgHRuEBANC4QEI0LBFTQHHdsbExey+iuwjx58mQ2u3z5sqx1R4GeOnVK5uo4Trci5mZ+lZWVMncri+poWjeHdVeAuvy7776TuXpubvXNHZnrjiJdunRpNluyZImsdXNct+7oPhPq3w24fxeg1v64ZhP4B6NxgYBoXCAgGhcIiMYFAqJxgYBoXCCggua4xcXFcg/x8OHDsv6FF17IZhcuXJC1blbqrrpUM0c3j1y0aJHM1WuSkr/yUc0E3bG0v/zyi8zV7Dwlvwet9kO3bt0qa93Vqe51U7Na9367Oay7ArS5uVnmHR0d2Wz9+vWyVn3W3fOexjcuEBCNCwRE4wIB0bhAQDQuEBCNCwRE4wIBFTTHTUmflVtTUyNrv/jii2ymZrwp6fN9U/Ln0ar9TDc7c7udcz0/WJ0Z7c6T/u2332Tuzgd2v7uapbrrQ6urq2Xu9nEVd3axO4/6119/nVO92rF2/yahqakpm830NeEbFwiIxgUConGBgGhcICAaFwiIxgUConGBgAqa486fPz9VVVVlc3dWraqtr6+XtW5W6nZm1d6p+9luptfb2ytzd+azOjvZ3V/r9nXdedTqDtqUUlq9enU2u3r1qqzdtWuXzN2+rtpjdrNxd3dva2urzNva2mReV1eXzTZs2CBru7u7s5l7TabxjQsEROMCAdG4QEA0LhAQjQsEROMCARU0DhofH089PT3Z/KWXXpL16q/B1XGXKflxjxpbpKTHQe541sHBQZm7cdCRI0dm/fPdz3bPbcGCBTJ34yC11ufGf4sXL5a5G8Op98xds+leN/VZTElfhZmSfl2Hh4dlrfosu9dkGt+4QEA0LhAQjQsEROMCAdG4QEA0LhAQjQsEVNAct6ioSM6ZPvvsM1n/4osvZrOFCxfK2j///FPmo6OjMlfzyjt37sjagYEBmbs5sDtKVF0ZqVb+ZvLY7mjZdevWyVwdsfrOO+/I2tLSUpmPjY3JXK0s9vf3y1r3urnXxX0eL168mM2effZZWauO3HXHDE/jGxcIiMYFAqJxgYBoXCAgGhcIiMYFAqJxgYAKmuM+evRIHnu5atUqWX/79u1s9vfff8vaxx9/XOZdXV0yV8eUqjlqSvqY0JT8zNAdJap+vpvrVVRUyHzLli0y37p1q8zVdZIqS8nvxLorPtXr5q6jdHN9d6xteXm5zNXnqb29XdaqfV3muMA/GI0LBETjAgHRuEBANC4QEI0LBETjAgEVNMednJyUMyg311OzN3eV5YULF2Tu9ivVObnu7GE3b3RzXHcmtPr5bi90x44dMn/99ddlrnakU9JnI09NTclatyt8/vx5mavPhJvDuufm5sDuTOihoaFs5s74Vp839zmexjcuEBCNCwRE4wIB0bhAQDQuEBCNCwRU0DhoampKrh2ptb2UklwJfPrpp2XtN998o5+csXTp0mzm1u4mJibm9NhuxUxdGdna2iprDxw4IPPdu3fL3K2vqbHKlStXZK07llatxqWkjzF1Y5NLly7JfKbXWeaoMdrhw4dlbVNTUzZzY9H/+XMz+lMA/qvQuEBANC4QEI0LBETjAgHRuEBANC4QUEFz3Hnz5snrKpcvXy7r1dzOXaO5b98+mX/88ccyV1dlurU9t1qnVh1T8rPSqqqqbObmsO5KR7dS6OaGan3Nze2PHz8uczcHVtefuqtR3Xu6efNmmbvrR7u7u7NZTU2NrFU5a33APxiNCwRE4wIB0bhAQDQuEBCNCwRE4wIBFTTHLSkpkfOtnp4eWa+OrRwfH5e1V69elfnKlStlruZ6fX19stbt47pZqPvdFDcDdj97bGxM5u4YUzXHdfu2HR0dMnefF7XH7H4vN3v//vvvZe6uJ1XXvrrjfo8dO5bN3JWv0/jGBQKicYGAaFwgIBoXCIjGBQKicYGAaFwgoILmuBMTE3Keqs4HTknP1urq6mSt2yt1V11+8MEH2eytt96StW4n1p35rM6iTknPSo8ePSpr1XnRKaXU3NwsczcH/uuvv7KZ+73VzmpKfhZbWVmZzUpLS2Wtm8O6+bibzau99AcPHsjaucyAp/GNCwRE4wIB0bhAQDQuEBCNCwRE4wIB0bhAQAXNcVPS8y13z6y6s3SuO63V1dUyv3btWja7ceOGrH3yySdl7nZa3ZnPxcXF2czdret2Yt2s1J3jq+a47izsN998U+YHDx6U+aeffprNfvjhB1lbUVEhc/eeqvckpZROnjw568d2M+SZ4BsXCIjGBQKicYGAaFwgIBoXCIjGBQIqcqOMf1dXVzf17rvvZnN1XWRKczumVI2SUkrpmWeekfmtW7ey2aZNm2StW9NSa3kp+VGVG/ko7rpJN+pqamqadf3ixYtl7enTp2VeW1sr8x9//DGbvf3227LWHamrPg8p+RXVycnJbNbZ2Slrn3vuuWz20UcfpYsXL+q5auIbFwiJxgUConGBgGhcICAaFwiIxgUConGBgApa63v48GHq7+/P5o2NjbJeXSHojjCd6xqWuh707NmzstZd8bls2TKZu9W79evXZ7OffvpJ1ra0tMh8zZo1Mj937pzM6+vrs5lalUwppb1798q8ra1N5jt27Mhmw8PDstatM27YsEHmbv49MjKSzdx7op6764NpfOMCAdG4QEA0LhAQjQsEROMCAdG4QEA0LhBQQfu4RUVF/Smlnv/c0wH+32uYmpqqcX+ooMYF8N+B/1UGAqJxgYBoXCAgGhcIiMYFAqJxgYBoXCAgGhcIiMYFAvoX0Dh6hmlmnbUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Produce 10 examples\n",
    "sample = torch.randn(10,20).to(device)\n",
    "sample = model.decode(sample).cpu()\n",
    "for i in range(10):\n",
    "    draw_image(sample.view(-1,1, 28, 28)[i].detach().numpy()[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Homework\n",
    "\n",
    "\n",
    "The code above should be fully functional but not do a very good job. \n",
    "\n",
    "  *  The loss function is missing the KL-Divergence term that ensures the latent space distribution to be Gaussian with the correct parameters. \n",
    "  * The network architecture is seriously underpowered.\n",
    "  \n",
    " Fix these two issues and train a working MNIST VAE. Some more details are in the code above.\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}