08_Autoencoder/3_Variational_Autoencoder.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Variational Autoencoder\n",
    "- MNIST dataset\n",
    "- Variational Inference\n",
    "- [Auto-Encoding Variational Bayes](https://arxiv.org/abs/1312.6114)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Settings\n",
    "### 1) Import required libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "import torch.nn.init as init\n",
    "import torchvision.datasets as dset\n",
    "import torchvision.transforms as transforms\n",
    "from torch.utils.data import DataLoader\n",
    "from torch.autograd import Variable\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2) Set hyperparameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 128\n",
    "learning_rate = 0.0005\n",
    "num_epoch = 10\n",
    "hidden_size = 50"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Data\n",
    "\n",
    "### 1) Download Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "mnist_train = dset.MNIST(\"./\", train=True, transform=transforms.ToTensor(), target_transform=None, download=True)\n",
    "mnist_test = dset.MNIST(\"./\", train=False, transform=transforms.ToTensor(), target_transform=None, download=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2) Set DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_loader = torch.utils.data.DataLoader(mnist_train,batch_size=batch_size, shuffle=True,num_workers=2,drop_last=True)\n",
    "test_loader = torch.utils.data.DataLoader(mnist_test,batch_size=batch_size, shuffle=False,num_workers=2,drop_last=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Model & Optimizer\n",
    "### 1) Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Encoder(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(Encoder,self).__init__()\n",
    "        self.fc1_1 = nn.Linear(784, hidden_size)\n",
    "        self.fc1_2 = nn.Linear(784, hidden_size)\n",
    "        self.relu = nn.ReLU()\n",
    "                        \n",
    "    def encode(self,x):\n",
    "        x = x.view(batch_size,-1)\n",
    "        mu = self.relu(self.fc1_1(x))\n",
    "        log_var = self.relu(self.fc1_2(x))\n",
    "                \n",
    "        return mu,log_var\n",
    "    \n",
    "    def reparametrize(self, mu, logvar):\n",
    "        std = logvar.mul(0.5).exp_()\n",
    "        \n",
    "        eps = torch.FloatTensor(std.size()).normal_()\n",
    "        eps = Variable(eps).cuda()\n",
    "        \n",
    "        return eps.mul(std).add_(mu)\n",
    "    \n",
    "    def forward(self,x):\n",
    "        mu, logvar = self.encode(x)\n",
    "        reparam = self.reparametrize(mu,logvar)\n",
    "        \n",
    "        return mu,logvar,reparam\n",
    "        \n",
    "encoder = Encoder().cuda()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Decoder(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(Decoder,self).__init__()\n",
    "        self.fc1 = nn.Linear(hidden_size, 784)\n",
    "        self.sigmoid = nn.Sigmoid()\n",
    "    \n",
    "    def forward(self,x):\n",
    "        out = self.fc1(x)\n",
    "        out = self.sigmoid(out)\n",
    "        out = out.view(batch_size,28,28,1)\n",
    "        \n",
    "        return out\n",
    "        \n",
    "decoder = Decoder().cuda()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2) Loss func & Optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "reconstruction_function = nn.BCELoss(size_average=False)\n",
    "\n",
    "def loss_function(recon_x, x, mu, logvar):\n",
    "    BCE = reconstruction_function(recon_x, x)\n",
    "\n",
    "    # see Appendix B from VAE paper:\n",
    "    # Kingma and Welling. Auto-Encoding Variational Bayes. ICLR, 2014\n",
    "    # https://arxiv.org/abs/1312.6114\n",
    "    # 0.5 * sum(1 + log(sigma^2) - mu^2 - sigma^2)\n",
    "    KLD_element = mu.pow(2).add_(logvar.exp()).mul_(-1).add_(1).add_(logvar)\n",
    "    KLD = torch.sum(KLD_element).mul_(-0.5)\n",
    "\n",
    "    return BCE + KLD\n",
    "\n",
    "parameters = list(encoder.parameters())+ list(decoder.parameters())\n",
    "optimizer = torch.optim.Adam(parameters, lr=learning_rate)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. Train "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "--------model restored--------\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gunho/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/functional.py:767: UserWarning: Using a target size (torch.Size([128, 1, 28, 28])) that is different to the input size (torch.Size([128, 28, 28, 1])) is deprecated. Please ensure they have the same size.\n",
      "  \"Please ensure they have the same size.\".format(target.size(), input.size()))\n",
      "/home/gunho/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/serialization.py:147: UserWarning: Couldn't retrieve source code for container of type Encoder. It won't be checked for correctness upon loading.\n",
      "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
      "/home/gunho/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/serialization.py:147: UserWarning: Couldn't retrieve source code for container of type Decoder. It won't be checked for correctness upon loading.\n",
      "  \"type \" + obj.__name__ + \". It won't be checked \"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " 22765.7402\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21904.7910\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22009.3418\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21833.4297\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21094.2734\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21506.8809\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21382.5098\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21497.7910\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21222.3516\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21390.6484\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21167.9219\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21796.5176\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21589.0586\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 20944.2637\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22058.2363\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 20930.8125\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22475.0312\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21683.2598\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21684.8398\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22011.9980\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22223.5293\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 20886.3828\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21517.1484\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 20274.2344\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21377.1855\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21657.9785\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21323.4922\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21849.2539\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21553.0820\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21307.3496\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21851.3477\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21537.6543\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21536.2637\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 20159.7305\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21645.5625\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22105.4141\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21099.5996\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21168.9609\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 20988.2207\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21825.2090\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22199.8438\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21319.4297\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22613.1582\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21830.9629\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21780.5645\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21420.4844\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21426.3262\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 22582.5820\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21294.1094\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n",
      "Variable containing:\n",
      " 21640.4004\n",
      "[torch.cuda.FloatTensor of size 1 (GPU 0)]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    encoder, decoder = torch.load('./model/variational_autoencoder.pkl')\n",
    "    print(\"\\n--------model restored--------\\n\")\n",
    "except:\n",
    "    print(\"\\n--------model not restored--------\\n\")\n",
    "    pass\n",
    "\n",
    "for i in range(num_epoch):\n",
    "    for j,[image,label] in enumerate(train_loader):\n",
    "        optimizer.zero_grad()\n",
    "        \n",
    "        image = Variable(image).cuda()\n",
    "        mu,log_var,reparam = encoder(image)\n",
    "        output = decoder(reparam)\n",
    "        \n",
    "        loss = loss_function(output, image, mu, log_var)\n",
    "        \n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        if j % 100 == 0:\n",
    "            torch.save([encoder,decoder],'./model/variational_autoencoder.pkl')\n",
    "            print(loss)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. Check with Train Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([128, 28, 28])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1a1fb364a8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f19c47f8c18>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAD9pJREFUeJzt3V+MXOV5x/Hfwxr8Z72AbWAx2MTEspAjS5jKQpVAVaqU\niKJIJjcoXFSuhOJcBCmRclFEL8olqvJHuYrkCBRTpSSVEgQXURuKimglMDaIYmM3Bowje1nWgMG7\ni//gtZ9e7HG0gZ3nHebMzJnd5/uRLO/OM2fm3bF/e87Mc97zmrsLQD6XNT0AAM0g/EBShB9IivAD\nSRF+ICnCDyRF+IGkCD+QFOEHklrSzyczM04nBHrM3a2d+9Xa85vZ3Wb2BzN7y8weqvNYAPrLOj23\n38yGJB2WdJek45L2Srrf3Q8G27DnB3qsH3v+2yW95e5H3P1TSb+StL3G4wHoozrhv1HSsTnfH69u\n+zNmttPM9pnZvhrPBaDLev6Bn7vvkrRL4rAfGCR19vxjktbP+X5ddRuABaBO+PdK2mRmN5vZFZK+\nJemZ7gwLQK91fNjv7jNm9qCk/5A0JOlxd3+jayMD0FMdt/o6ejLe8wM915eTfAAsXIQfSIrwA0kR\nfiApwg8kRfiBpAg/kBThB5Ii/EBShB9IivADSRF+ICnCDyTV10t3A1+EWTw5rZ8zUhcj9vxAUoQf\nSIrwA0kRfiApwg8kRfiBpAg/kBR9/gWg1O+O6pddFv9+r/PY3ahHSn38ixcvhvULFy50/NgZziFg\nzw8kRfiBpAg/kBThB5Ii/EBShB9IivADSdXq85vZUUlTki5ImnH3bd0Y1GJTtxe+ZEn8zxTVh4aG\nwm1L5wHUPU+g9PyRmZmZsH727NmwHvXqo3MAsujGST5/7e4fdOFxAPQRh/1AUnXD75J+b2avmNnO\nbgwIQH/UPey/093HzOw6Sc+a2f+5+wtz71D9UuAXAzBgrFsTGMzsEUnT7v7D4D6Lf7bEPPjArzN1\nP/A7f/58y1rpA7+FPLHH3duaTdXxYb+ZDZvZyKWvJX1d0oFOHw9Af9U57B+V9FT1m3+JpH9193/v\nyqgA9FzXDvvberJFethf99B32bJlYf2qq64K6yMjIy1rK1euDLddunRpWC+95Sj97NGc++np6XDb\nDz/8MKyfPHkyrJ85c6ZljcN+Wn1AWoQfSIrwA0kRfiApwg8kRfiBpLh0d5uiM91Krbzh4eGwvm7d\nurC+ZcuWsD46OtqydsUVV4TbRu0wqXwWXalVGP3sk5OT4bb79+8P66dPnw7r586da1krXfZ7Ibf6\n2sWeH0iK8ANJEX4gKcIPJEX4gaQIP5AU4QeSos9fKU1Njfr8y5cvD7e96aabwvodd9wR1jdv3hzW\no+mpx48fD7cdGxsL66U+/6ZNm8J6NPZoKrIkrVq1KqxPTU2F9TpTekv/HxbDeQDs+YGkCD+QFOEH\nkiL8QFKEH0iK8ANJEX4gKfr8lTqr5qxevTrc9tZbb61VL13a+8CB1mul7NmzJ9z23XffDeul6wFs\n2LAhrK9fv76jmlTu8x87diysj4+Pt6xFc/2l8nz/xYA9P5AU4QeSIvxAUoQfSIrwA0kRfiApwg8k\nVezzm9njkr4h6YS7b6luWy3p15I2SDoq6T53/6h3w6wvmo/fTj3qtZd63bfddltYL50ncOjQobD+\n/PPPt6y988474bbnz58P66XlwUvnAaxZs6Zl7corrwy3veaaa8L6DTfcENajsZXO62A+/6xfSLr7\nM7c9JOk5d98k6bnqewALSDH87v6CpJOfuXm7pN3V17sl3dvlcQHosU7f84+6+6VzJ9+T1Hq9KAAD\nqfa5/e7uZtbyDZCZ7ZS0s+7zAOiuTvf8E2a2VpKqv0+0uqO773L3be6+rcPnAtADnYb/GUk7qq93\nSHq6O8MB0C/F8JvZk5JelHSLmR03swckPSrpLjN7U9LfVN8DWECK7/nd/f4Wpa91eSw9VerLlvr8\nK1asaFkrXVe/dN3+6PrykvTiiy+G9SNHjrSsffLJJ+G2pZ+7pLR9dN3/06dPh9uW6qV/0+gaDKU+\nfgac4QckRfiBpAg/kBThB5Ii/EBShB9Iikt3tymafnrLLbeE25ampu7duzesHz16NKxHrcLSJahL\nrbpSS6zUjpuYmGhZi1pxkjQ9PR3WS2O7/PLLW9aGhobCbUtTnRcD9vxAUoQfSIrwA0kRfiApwg8k\nRfiBpAg/kBR9/kqpZ3zddde1rN18883htsuXLw/rpX72p59+GtajXn2pn1269PbIyEjHzy3FU4pL\n265cuTKsly4rvnTp0pa1ulN6F8OlvdnzA0kRfiApwg8kRfiBpAg/kBThB5Ii/EBS9Pkr0dxvKb78\n9rXXXhtuW+qll5aqLi1FHc3ZL83nLy0PXrpWwcaNG8N6tHz5unXrwm3PnTsX1teuXRvW6/T5F0Mf\nv4Q9P5AU4QeSIvxAUoQfSIrwA0kRfiApwg8kVezzm9njkr4h6YS7b6lue0TStyW9X93tYXf/Xa8G\n2Q11luCW4p5yqSc8MzMT1q+++uqwvnXr1rAe9ctL5xhcf/31HT+2VL6WQXR+ROnnLl3HoLQeQvRv\nzhLd7e35fyHp7nlu/4m7b63+DHTwAXxeMfzu/oKkk30YC4A+qvOe/0Eze93MHjezVV0bEYC+6DT8\nP5O0UdJWSeOSftTqjma208z2mdm+Dp8LQA90FH53n3D3C+5+UdLPJd0e3HeXu29z922dDhJA93UU\nfjOb+9H3NyUd6M5wAPRLO62+JyV9VdI1ZnZc0j9J+qqZbZXkko5K+k4PxwigB4rhd/f757n5sR6M\npVHR3G8pXof+4MGD4balud9vv/12WJ+cnAzry5Yta1mL1htop15ac6B0Xf/o2vqlcxBKStcqOHPm\nTMta6dyLxTBfv4Qz/ICkCD+QFOEHkiL8QFKEH0iK8ANJpbl0d2lK74ULF8L6/v37W9YOHz4cbvv+\n+++H9Y8++iisl9pSUbutNCV3dHQ0rK9Zsyasl6bV1pk6W/q5x8fHw/qpU6c6fuxSq28xtALZ8wNJ\nEX4gKcIPJEX4gaQIP5AU4QeSIvxAUmn6/CVnz54N61EvP5o62s5jl84xKPXKp6amOn7u0jkGpT7+\n5s2bw3p0+e3z58+H237wwQdh/eWXXw7rUZ+/9Jovhj5+CXt+ICnCDyRF+IGkCD+QFOEHkiL8QFKE\nH0gqTZ+/1LctLQcd9ctL25YuMV0aW51rEZw7dy7cdnp6OqwPDw+H9dI5Dh9//HHLWqnP/9JLL4X1\nUp8/+jfL0McvYc8PJEX4gaQIP5AU4QeSIvxAUoQfSIrwA0kV+/xmtl7SE5JGJbmkXe7+UzNbLenX\nkjZIOirpPnePJ4c3qNTXLfWco+3r9vFL8/VLff6oXhpb6ecuncNQunZ+dB2E0mM/9dRTYX1sbCys\nR+c/0Odvb88/I+kH7v4VSX8p6btm9hVJD0l6zt03SXqu+h7AAlEMv7uPu/ur1ddTkg5JulHSdkm7\nq7vtlnRvrwYJoPu+0Ht+M9sg6TZJeySNuvulY773NPu2AMAC0fa5/Wa2UtJvJH3f3Sfnvk91dzez\ned9EmdlOSTvrDhRAd7W15zezyzUb/F+6+2+rmyfMbG1VXyvpxHzbuvsud9/m7tu6MWAA3VEMv83u\n4h+TdMjdfzyn9IykHdXXOyQ93f3hAeiVdg7775D0d5L2m9lr1W0PS3pU0r+Z2QOS/ijpvt4MsTtK\nLa/SpZybVGoFRm2r0lLUdaf8HjlyJKwvWdL6v1hpOvCBAwfCeqlVGKnzmi4WxfC7+/9IavVKfa27\nwwHQL5zhByRF+IGkCD+QFOEHkiL8QFKEH0iKS3dX6i6T3attpXrTckvblupRn14qT+kdGhpqWStN\nJy7VS1Odo9c9Qx+/hD0/kBThB5Ii/EBShB9IivADSRF+ICnCDySVps9fUur71ukL97rPX+ey4iXR\nMteSNDk5GdanpqZa1kZGRsJtS/WlS5eG9ehaBb38914o2PMDSRF+ICnCDyRF+IGkCD+QFOEHkiL8\nQFL0+fugbs+47hLfdR67VK9zHsHw8HBYX7FiRVgvzeePZOjjl7DnB5Ii/EBShB9IivADSRF+ICnC\nDyRF+IGkin1+M1sv6QlJo5Jc0i53/6mZPSLp25Ler+76sLv/rlcDRWfqrldQdz7/xMREy9qpU6fC\nbU+cOBHWo2sFSPHPRp+/vZN8ZiT9wN1fNbMRSa+Y2bNV7Sfu/sPeDQ9ArxTD7+7jksarr6fM7JCk\nG3s9MAC99YXe85vZBkm3SdpT3fSgmb1uZo+b2aoW2+w0s31mtq/WSAF0VdvhN7OVkn4j6fvuPinp\nZ5I2Stqq2SODH823nbvvcvdt7r6tC+MF0CVthd/MLtds8H/p7r+VJHefcPcL7n5R0s8l3d67YQLo\ntmL4bXbK2GOSDrn7j+fcvnbO3b4p6UD3hwegV6yN6aJ3SvpvSfslXZq/+bCk+zV7yO+Sjkr6TvXh\nYPRY9FcWmDrLYJe2L207MzMT1utelnyxcve25ngXw99NhH/hIfwLT7vh5ww/ICnCDyRF+IGkCD+Q\nFOEHkiL8QFK0+oBFhlYfgBDhB5Ii/EBShB9IivADSRF+ICnCDyTV7yW6P5D0xznfX1PdNogGdWyD\nOi6JsXWqm2P7Urt37OtJPp97crN9g3ptv0Ed26COS2JsnWpqbBz2A0kRfiCppsO/q+Hnjwzq2AZ1\nXBJj61QjY2v0PT+A5jS95wfQkEbCb2Z3m9kfzOwtM3uoiTG0YmZHzWy/mb3W9BJj1TJoJ8zswJzb\nVpvZs2b2ZvX3vMukNTS2R8xsrHrtXjOzexoa23oz+y8zO2hmb5jZ96rbG33tgnE18rr1/bDfzIYk\nHZZ0l6TjkvZKut/dD/Z1IC2Y2VFJ29y98Z6wmf2VpGlJT7j7luq2f5Z00t0frX5xrnL3fxiQsT0i\nabrplZurBWXWzl1ZWtK9kv5eDb52wbjuUwOvWxN7/tslveXuR9z9U0m/krS9gXEMPHd/QdLJz9y8\nXdLu6uvdmv3P03ctxjYQ3H3c3V+tvp6SdGll6UZfu2BcjWgi/DdKOjbn++MarCW/XdLvzewVM9vZ\n9GDmMTpnZaT3JI02OZh5FFdu7qfPrCw9MK9dJytedxsf+H3ene7+F5L+VtJ3q8PbgeSz79kGqV3T\n1srN/TLPytJ/0uRr1+mK193WRPjHJK2f8/266raB4O5j1d8nJD2lwVt9eOLSIqnV3ycaHs+fDNLK\nzfOtLK0BeO0GacXrJsK/V9ImM7vZzK6Q9C1JzzQwjs8xs+HqgxiZ2bCkr2vwVh9+RtKO6usdkp5u\ncCx/ZlBWbm61srQafu0GbsVrd+/7H0n3aPYT/7cl/WMTY2gxri9L+t/qzxtNj03Sk5o9DDyv2c9G\nHpC0RtJzkt6U9J+SVg/Q2P5Fs6s5v67ZoK1taGx3avaQ/nVJr1V/7mn6tQvG1cjrxhl+QFJ84Ack\nRfiBpAg/kBThB5Ii/EBShB9IivADSRF+IKn/B5fwtO0HgOkdAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f19c400fbe0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nivADhSL8QKEIP1Aowg8U6v8A05YiVz2pzH4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f19846ef1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f198457acc0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "out_img = torch.squeeze(output.cpu().data)\n",
    "print(out_img.size())\n",
    "\n",
    "for i in range(5):\n",
    "    #plt.imshow(torch.squeeze(image[i]).numpy(),cmap='gray')\n",
    "    #plt.show()\n",
    "    plt.imshow(out_img[i].numpy(),cmap='gray')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "for i in range(1):\n",
    "    for j,[image,label] in enumerate(test_loader):\n",
    "       \n",
    "        image = Variable(image,volatile=True).cuda()\n",
    "        output,mean,var = encoder(image)\n",
    "        output = decoder(output)\n",
    "        \n",
    "    if j % 10 == 0:\n",
    "        print(loss)        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([128, 28, 28])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKcIPJPX/2a1N1RqaDBsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1a1fac6518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1984580e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n8ANJEX4gKcIPJEX4gaT+DxJ6aQvrEx/MAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1984522588>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nlnuQJ+mQAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1a1facda58>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f19843cfe80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIvxAUIQfCIrwA0ERfiCo/wMi6ZzxpchqMwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1984740d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/DYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f19c404b550>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCD+QFOEHkvp/Adb20NBNhqkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1984752358>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRfiBoAg/EBThB4L6Hyu4ct6owBXqAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f19842b5ac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f19842edfd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "out_img = torch.squeeze(output.cpu().data)\n",
    "print(out_img.size())\n",
    "\n",
    "for i in range(5):\n",
    "    plt.imshow(torch.squeeze(image[i].cpu()).data.numpy(),cmap='gray')\n",
    "    plt.show()\n",
    "    plt.imshow(out_img[i].numpy(),cmap='gray')\n",
    "    plt.show()"
   ]
  }
 ],
 "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}