VAE with fastai.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.vision import *\n",
    "from vae import *\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist = untar_data(URLs.MNIST_TINY)\n",
    "tfms = get_transforms(do_flip=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = (ImageImageList.from_folder(mnist/'train')\n",
    "        .split_by_rand_pct(0.1, seed=42)\n",
    "        .label_from_func(lambda x: x)\n",
    "        .transform(tfms)\n",
    "        .databunch(num_workers=0, bs=16)\n",
    "        .normalize(do_y=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ImageDataBunch;\n",
       "\n",
       "Train: LabelList (639 items)\n",
       "x: ImageImageList\n",
       "Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28)\n",
       "y: ImageList\n",
       "Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28)\n",
       "Path: C:\\Users\\C64062\\.fastai\\data\\mnist_tiny\\train;\n",
       "\n",
       "Valid: LabelList (70 items)\n",
       "x: ImageImageList\n",
       "Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28)\n",
       "y: ImageList\n",
       "Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28),Image (3, 28, 28)\n",
       "Path: C:\\Users\\C64062\\.fastai\\data\\mnist_tiny\\train;\n",
       "\n",
       "Test: None"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "image_size = data.one_batch()[0].shape[-1]\n",
    "vae = VAE(size_in = image_size, size_emb=2)\n",
    "my_learner = Learner(data,\n",
    "                     vae,\n",
    "                     opt_func=torch.optim.Adam,\n",
    "                     loss_func=vae.vae_loss_function)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Find the learning rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "my_learner.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "my_learner.recorder.plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fit for 5 epochs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:27 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>-187683.203125</td>\n",
       "      <td>-493635.093750</td>\n",
       "      <td>00:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>-513658.718750</td>\n",
       "      <td>-689428.625000</td>\n",
       "      <td>00:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>-637746.500000</td>\n",
       "      <td>-705161.812500</td>\n",
       "      <td>00:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>-686498.500000</td>\n",
       "      <td>-697282.312500</td>\n",
       "      <td>00:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>-708810.250000</td>\n",
       "      <td>-685905.937500</td>\n",
       "      <td>00:05</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "my_learner.fit(5, 1e-2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show 4 samples and their reconstructions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAArcAAAP4CAYAAADH5NR6AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3Xu0nXV5L/rnl6wEQeQqWIVqxKWIWHFTBREriRtUkiXDy7bs1rGLgofDEe/F2jNOkaTUqnV4q6XdW21pz+Z4aatnH10BLGqi7CogiEWlRFeBiqCACIiA5PaeP+bMzkqcv5m15lprXp71+YzBSPI+833fZ82smd83v8z5UJqmCQAAyGDJoBsAAID5ItwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAswS6WU15ZSmvZ/K9rH/rb961vn6R47rr92Pq4HsFgIt5BQKWXjfAatOfaydkdQm8U5S0spPy2lfLRSXzEt/O347+FSyvdLKe8vpTx2/r6CGfu3iLg6Iq6f6QmdQvI0V7f/+9G8dTjzvt5WStlWSjm0Q21jh+d+9//+tt8970kpZWxaf3806H6AhTM26AYAOnhhRBwcEf9jBo+9PVoB8HER8dSIOC8iTi6l/GbTNNs7nVBKWd40zeb5ajYiommaCyPiwnm83vPm61o9eHlEfL1pmrs61G6MiEe1f35IRBzR/vm/RsTP2z//t7ncvJSyLCK2Nv4vQ0AP7NzCIjFtx21jKeXcUsqtpZQHSimTpZRfqzzuTaWUH7Z3RdeXUg7v9Lhpx3bZpW3XLphW37Fz9to9tPvyiLg3Ir46gy/tE03TPK9pmidHxD+2jz07Ip612z3f337rwP0R8Zl27TGllA+WUm4ppWwupfy4lPJfSykHTOu5lFIuKKXc1X6+/ntE7N/h+f2VtyW0zz2nlHJdKeWhUsov2j9/fnt38+Jpl7hl+tsQOr0toZTyxFLK/11K+UkpZUsp5fZSysem77BO76OU8upSyk2llAdLKV8rpRy5pyezlHJwRJwYlb9YNE3zhvbz/bzYNcz/r+PtoB+llE+VUn7Qft42t3v6cCnlMdPud0m736lSylmllFsi4pGIeHQpZa9SykWllPtKKT8rpXyolPK+9uO37tb3a0op17S/1gdLKV8upZzQrp0cEVumPfzCHffc0/MBjB47t7D4PD8ijo+I2yJi34hYExEfiIjX7Pa450XEcRFxa0QcHhGrI+Kz7XNn6saIGI+Iw9q/vrr94917OO+0iFjfNM3WPTxuNt4cEVsj4uaIeKSUsjwiNkbEsRGxOVo7j0+JiP89Ip5bSnle0zRbIuL/iIi17Wv8OCJWRcQrZnjPP4+IN7Z/fm9E3BERR0fE06K1u3lz7Nz5/Ha0Ql3HtyG0A+w3IuIJ7cd9P1o71f9bRKwspRzbNM0vpp1yWET8P+377B0RvxURfxOt4NrNaRGxNGa2a74nr4iIX7R7eEy0vta3RGuX/Xd2e+yvR8THIuIHsfP74z0R8Yb2z2+NiP8SO3eN/5dSyjsj4r3tX061H/OiiNhYSvmtiLg/Wt97O753fxStHf/b5vLFAcPJzi0sPksj4oSmaZ4WEf9v+9h/rDzu2KZpnhERb28fO66UsmqmN2qa5g0R8Ylpv96xs7e+dk4p5T9ExIqYebh6fSnlqvaO339qH/uXiLhht8f9PCKObJrmN6IV5P9ztILt1mh9ncdEK3huax//7fZ572z/eE27rxUR8c09NVVa76E9t/3Lz0fEE5qmeWZEPD4ivtrhbQyvaD83n4jOzo1WsG0i4reapjk6WkE0ohVyX7fb48ci4lVN0xwVER9uH3t+KWXvPbT+8oj4btM0c3prQdvzmqZ5bNM0z26a5ikR8b728Ve2/3Ix3fKIOLtpmqdH6zkqsTPYfi5awfiIiLhz+kmllH1j578OXNg0zVMj4kkR8ZX2Ndc1TfPNiHjBtNP+W/u5fvU8fI3AkBFuYfH5TtM0327//Mb2j7/ywaGIuKFpmpvaP//MtOPPXLDOWl4eEb+MiMtn+PjDorUj92vR2rX7QESc3OH9tp9tmuZHERFN02yLnbt4YxHx3fZbKW6NVqiPiHheKWW/iHhi+9f/o2maze3d5M/NoK/nRiugRUR8sGmaX7bvfW/TNLfM8Gvb/XoREVPtsBZN01werR3hiIjn7Pb4+5um+UL75zdOO97p9zoiIkop+0TEKTE/u7YREaeUUr7bfltLEzv/orA8Wu+pnu6BaO0sR/v37qkRsVe79pmm5ecRsftfjH4jWjvTERHnt++zLVo7txGtf4EAFhFvS4DF575pP9/xz/6l0wOn6VTf8WGfpdOO/cp7UXvw8oj4UtM0D87w8euaplk7g8f9ZLdf7/iatkTEtzo8/s4Ox3Y/dxBm+iGrTr/PEd17f0m0guKcw20p5YyI+LP2L++I1lsBHhs734axdLdT7uryAbJuX/P0r2f6h9p2mM+3tgAjwM4tUPOsaR9A+k/Tjn+3/eOOT9KvKK3RXY+KiFM7XOehHT8ppTy62w3b/5T/rJi/ncNurmn/OBYRb532IakXRMS6iLikvVO4432Zp5VSlpdSxqIVwPfkm7EzlL21lLJXREQpZf+yc+zXQ9Me3/W5iZ1vhXhqKeW57Wu9NCIObB+/dgY97cnLI+K2pmmum4dr7dgxvS8iVjRNc3y03iowU9+P1nuLIyJeFRHR3klfs9vjbojWTn9ExBXResvNjt/L18XO90tvi9Z7qyP2/FwDI0y4BWoeiYhvlVK+Fzvfs3lttD6EFRHx5faPh0dr5/O70fqn5N3dNO3n32u/P/aIDo+LaH0AaXtEfKFSn0+fitaHuEpEfL2U8r1Syr9G68NHl0brvbURO3cfnxcRt7T/e/6eLt40za0RcVH7ly+PiDtKKTdE60NpK9vHpz83X2o/N7UPfF3UPrdExJWllO9G6728Ea23Y1xcOW9GSilLoxUc/7+5XGeaHe95PiBakyBujohXzvTk9ofj/rL9y9Pb598crbef7P64He9dfnNE3F5Kub6Ucle03o7xn9uPayJiU/txb2tPVpi30W3A8BBugZprI+Jt0ZqosDla74F95bR/Or44WtMAfhqt96V+JSI+0uE6kxHx8Yi4J1of9Dk+Ivap3LPbfNV51TTNI9EKmR+M1nttnxqtua3fi4g/iZ071BdFxB9H6+s8ICKuioj/a4a3eXO0PhR1fbS+5iOi9U/nP2j3cEO0gtmd0Qptx8fOndjd+70rWgH7v0drN/TIaE0V+EREnLjbpIRezGa28Ex8LFrfD/dEa1LCV6K1Iz4b/2e0Au790XpePhURf9euPbzjQU3T/GlE/F60JiLsH61pFPdGxN9G+328bW+K1u/rkmi9h7nTX8aAEVfMyAamK63ZtCdF6xP9K/t434OjFfLe2TTNB/p1X1pKKR+J1qitQ+d5BFvPSmv+8kPtt4fs+MDbdRHx9Ii4smmaFw6yP2A4+UAZMCwOjtaO6acH3cgi9b2IeNOwBNu2F0TExaWUa6M1TeG4aM3I3RrT/ucgANMJt8BQaJrm+7Hzwz/0WdM0Hxt0Dx38W7Te0vGsaL3d4GfRek/we5qmubrbicDi5W0JAACk4QNlAACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaYz182allKaf94Nh0DRNGXQPvRgfH/d6ZdGZmpoayddrhDWWxanTGmvnFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANMYG3cCoOOigg6q1008/vVp72cteVq3ddttt1drDDz/c8fjnPve56jnf+973qrV77rmnWgP2bOmSpdXaqatP7WMndT/4t5s7Ht/0rzf2uROYHWss88nOLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkUZqm6d/NSunfzebZDTfcUK0dffTRfeyk7pZbbqnWzjjjjGrtn//5nxeiHdqapimD7qEX4+PjI/t67WZiYmLQLQyVycnJQbcwVKampkby9RphjV1o1tjh1GmNtXMLAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGmMDbqBUfHQQw/N+zU/85nPVGunn376rK/35Cc/uVr7yle+Uq29/e1vr9YuuuiiWfcBg2bc18x1e66MCaNfrLHMJzu3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkUZqm6d/NSunfzebZfvvtV61dfvnl1drxxx9frT3lKU+p1u68886Ox9/xjndUz/n93//9au0xj3lMtbZt27Zq7Zhjjul4/MYbb6yew66apimD7qEX4+PjI/t6ffGLX1ytLV++vI+djLbFOC1hampqJF+vEdbY3VljF4dOa6ydWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIY2zQDYyKn//859XaaaedVq2dd9551Vq30Se33nprx+N//Md/XD1n06ZN1dqnPvWpam3JkvrfcU499dSOx40pYZj90z/9U7XWbRTY5s2bF6KdWVsxUR9l9szo3yizJxz67Grtjru+3bc+yM8auytr7NzYuQUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANErTNP27WSn9uxm72L59e7XW7Xvgr//6rzseP/vss+fc02LRNE0ZdA+9GB8f93odkDVrJqq10sfvpsnJyf7dbEhMTU2N5Os1who7SNbYwem0xtq5BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0xgbdAPPnuOOOm/drfvzjH5/3awIRJ598crXWz3FfwMxYY0eHnVsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIQbgEASMMosCF1zDHHdDz+xje+sXrO0UcfPe99PPzww/N+Tchm2bLOf5S++JSXVM8pS4Zj3tfkpYPuAPrPGpubnVsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIwLWFIrV69uuPxM888s699nHTSSR2PP/LII9VzfvCDHyxUOzAwhx56aLV23HHH9bGTebZ9ctAdQN9ZY3OzcwsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaZSmafp3s1L6d7MRt2zZso7H3/SmN1XP+fVf//Vq7S1veUu11sv3wB133FGtnXzyydXapk2bZn2vUdc0TRl0D70YHx/3ep1mYmJi0C0siJu7jBW6cRG+Xqempkby9RphjZ0Na2wendZYO7cAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZRYIvEfvvtV62df/751dqb3/zmjseXL19ePef666+v1k444YRq7ZFHHqnWRplRYDm86IUvrdb22W+sj530z+Tk5KBb6DujwOiFNXZwjAIDACA14RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSMAqOrc889t+PxD3/4w9Vzliyp/53pb/7mb6q1N7zhDR2Pb9mypXrOKDAKLL+JiYlBt9B3WceEGQVGP1lj584oMAAAUhNuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIwygwevJXf/VX1drZZ5/d0zUf//jHdzx+11139XS9YWEU2OK1bGxZtXbwIQdXaw/cd1+1tnLlqmqtLF06s8YW0KiPCDMKjGFgjZ05o8AAAEhNuAUAIA3hFgCANIRbAADSEG4BAEjDtAR6cthhh1VrP/zhD3u65hFHHNHx+L//+7/3dL1hYVoC/fL0p/9mx+Pj450/Jb0QTEsYHGtsHtbYmTMtAQCA1IRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSGBt0A4ymn/70p9XaN77xjWrthBNOqNZe//rXdzx+/vnnz7wxWMRuuum6jsfHxyf61kMp9Ula/Rw9CaPMGjs3dm4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA2jwGboQx/6ULU2NTVVrV100UUL0c7APfrRj67WjjzyyD52AgwT477ohTV2V9bYubFzCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpGAU2Q69//eurteXLl1dr3cbi/OVf/uWceuqHvffeu+Pxiy++uHrOQQcd1NO9rr322p7Og92dunp1tbZ1y5Zq7f5776/Wrrn2mln3UUrpUq3/2dDrNK2JiYneToQBs8buyho7N3ZuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIw7SEGfrMZz5Trb3uda+r1j760Y9Wa895znOqtQ996EMza2yG9t9//2rtjDPOqNZe+cpXdjx+wAEH9NTHZZddVq198Ytf7OmasLulS+p/b1+6117V2qG/dmi1ZhLBzHTbMdnety4YNdbYXVlj58bOLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkUZqm6d/NSunfzebZsmXLqrV3v/vd1dpZZ51VrfU66qMXpZRqrZfvgc2bN1drf/d3f1etveMd76jWHnjggVn3MQqapqk/+UNsfHx8ZF+vB0/UR3qdEMf1sZO8Lr/88o7Ht27d2udO5tfU1NRIvl4jrLG7s8buajGtsXZuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACANo8AW2KpVq6q1Cy+8sFo74YQT5rWPXseUXHLJJR2Pf/7zn6+e89nPfnbmjS0CRoENl26vhTVr1vSxk+HQ7fW/fv36PnYyHIwCGy3WWIwCAwAgNeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0jAKDBWYUWH777LNPx+PLli2vntNs316tHfWMo6q1b37zm9Xa9i7XZGaMAoPRYhQYAACpCbcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQxNugGAEbdQw89VKv0dL2rr76692YAFjk7twAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAapWmaQfcAAADzws4tAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpjPXzZqWUpp/3g2HQNE0ZdA+9GB8f93pl0ZmamhrJ12uENZbFqdMaa+cWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0xgbdAMAomJiY6Nu9mqap1kop1dq2bds6Hv/Jj++snnP9t78188YARoCdWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIY6RHgd18883V2ooVK/rWR7fRPN1G+gxLH5///OertX/8x3/sePyee+6pnnPZZZdVa9AP/RzbtRC6vZa7Wbp0acfjhx3+hOo5N3znX6q12mgxFgdr7Pz0YY3tPzu3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJBG6fMYjXm92fbt26u1URgPkrWPe++9t1p75zvfWa1dcsklHY8/8sgjc+5pkJqm6W2u04CNj4/375umR6M+8mvYrV+/vlrr558p/TQ1NTWSr9cIa+xi6cMau6tOa6ydWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hjpaQl/+Id/WK299rWvnc9bxYYNG6q1H//4x9XaihUrqrVbb721WnvJS17S8fjBBx9cPafbJzmf/OQnV2tLly6t1nrR6ydKn/Oc53Q8fv3118+5p0EyLWHhmJawsK688spq7f777+9jJ/1jWsJO1thdWWOHk2kJAACkJtwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJDGSI8CY+aOPfbYau3AAw+s1t7xjnd0PH7KKadUz+l1TMlVV13V8fiJJ55YPWcUGAW2cJ58xBEdjx+w/wHVcx7zmEdXa9/4dufvwYiIX1tWv+bmZqzj8Z/ed1f1nG3bt1drL53Yv1obixdUa/00OTk56BYWhFFg9MIaOzhGgQEAkJpwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGp3n15DOt771rZ7Ou++++zoe7zampFcPPPDAvF+T3G65+ea+3eu2uHter9dtnM+wjPvq56hIGGXW2OFi5xYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0jAKbEitWLGi4/Gjjjqqes4dd9xRrT3rWc+q1l71qldVa6eddlq1VtNtxFFt7ElExJlnnjnre8GoeuFJJw26hT360hVXDLoFWBDW2Nzs3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGkaBDdDf//3fV2unnnpqx+P77LNP9ZzNmzdXa8uXL595Y9M0TTPrc6ampqq1lStXVmvdxqzAMDt1defX69IlS/vcyeyt71bcsqVfbcC8s8buajGtsXZuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIw7SEATrmmGOqtW6f2KzZa6+9qrVePpEZEfHwww93PP5nf/Zn1XM++MEPVmu/+MUveuoDBm1iYmLQLSyINV1qkz3+uQHDwBq7eNm5BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0jAKjq+3bt3c8fv3111fPMYqEUZV13FevVq9eXa1deumlfewEcrLGLgw7twAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQRmmapn83K6V/NxsB5557brV2/PHHdzy+atWq6jmHHXZYtdbP3+eLL764Wnv/+99frW3atGkh2hm4pmnKoHvoxfj4eMrXayn13441a9b0sZOcJicnB93CnExNTY3k6zXCGrs7a+yuFtMaa+cWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIwCiyRE088sVq78MILq7WVK1fOax/dRi11+3773d/93Y7HP/3pT8+5p0EyCmx0vGjVi6q1fR69Tx87GV1bt26t1i6//PI+dtIbo8CoscYOJ6PAAABITbgFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACANo8AWiaVLl1ZrRx11VLX2sY99rOPxI488snrOgQceWK11+37btm1bx+MXXHBB9Zz3vOc91dqwMAosv8c//vCOxzdv+2X1nL2XLavWfnb3PfWbdRkD9Mi2LR2Pn/rS1fXr9dHk5OSgW9gjo8DohTV2cIwCAwAgNeEWAIA0hFsAANIQbgEASEO4BQAgDdMS6MkTnvCEau1d73pXtXbmmWdWa7VPm9Y+4RkRcfLJJ1drX/va16q1fjItgUGbmJgYdAsRYVrCQrPG5mGNnTnTEgAASE24BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgjbFBN5DdE5/4xGrtj/7oj6q1m2++uePx9773vXPuaT7ccccd1do555xTrd1+++3V2tq1azser40viYg4/fTTq7VhGVMCtHUbsmWIFT2wxu7KGtti5xYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0ihN07/5K6WURTfs5SMf+Ui19sY3vrFa+8UvftHx+FFHHVU9p9vokGHRbazIJz/5yXm9V7fxJv3UNE23AUhDa3x8fOhfr6tXr+54fMmS3v7ePjk5OZd2htbExMSgW4iI0Xh+p6amRvL1GmGN3Z01dleLaY21cwsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaYwNuoHsXvOa11Rrpcx+4syWLVvm0k5fnHXWWdXaX/zFX1RrvTwf//AP/zDrc8hj65atHY8v32t5T9frNjLr33/4o2rtu9/5l47Hlyypf09v27Z95o1Nv2ap70msXtN5NBpkZY3dlTW2xc4tAABpCLcAAKQh3AIAkIZwCwBAGsItAABpmJawwJqm6am27777djx+xRVXVM/56le/Wq1dd9111dqmTZuqteXLO3/q/Lzzzque87KXvaxa6+X5ePDBB6vn/MEf/EG1Rn5btnb+ZHOv0xK6edITD++pBiwca+yurLEtdm4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAII3SbWzEvN+slP7dbEiH5WG1AAAVfElEQVS8733vq9be8pa3VGvLli2b9b1KKdVan3+fe+rjnnvu6Xj8937v96rnXH755TNvbECapqk/IUNsfHx8ZF+vExMTg25h0ZqcnBx0C3MyNTU1kq/XCGvs7qyxu1pMa6ydWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIY2zQDWT3zne+s6fz3vrWt3Y8PjY22r9lV1xxRbVW+5pvuummhWqHpLqNo3r2s59drR1++OEL0c7IGvWxXuRnjd2VNbbFzi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApFGapunfzUrp381G3AknnNDx+DHHHFM959WvfnW1tu+++1ZrmzZtmnljbbfccku19o1vfKNa+9KXvlStbd26ddZ9jIKmacqge+jF+Pi41+sMLV26tOPxUuq/9du3b6/WnrTiydXaHbf/qFrbvHlzx+P9/HN+1E1NTY3k6zXCGjsb1tg8Oq2xdm4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA2jwGCBGQUGo8MoMBgtRoEBAJCacAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaZSmaQbdAwAAzAs7twAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApDHWz5uVUpp+3g+GQdM0ZdA99GJ8fNzrlUVnampqJF+vEdZYFqdOa6ydWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hgbdAPMn5UrV1ZrF1xwQU/nzbdVq1ZVaxs3buxbH9DJ0qX1v+9v397M+noHH/KEau345xxTrZUlo7vvMDk5OegWYEFYY0fH6P4JCgAAuxFuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIozTN7Mfb9HyzUvp3sz5au3ZttdZtPAi7qo0p6TbaZBQ0TVMG3UMvxsfHR/b12u17Zp999qnWShnJ36qh0m1NWb9+fR876c3U1NTIfhNYY+lmMa2xdm4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAII2xQTcwbFauXNnx+IYNG/rbSA/WrVtXrXUbpTLfeh3bUnvuu40W6jbCpDb2hDwe97hDOx437mtwmmb7oFtgiFlj584au2d2bgEASEO4BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDaPA5kG3cRjdRocM8xiNueh1JEq3ESa9nJP1+WWne++7r1IZ8XFfXUbzbN22rVobGxv8H+l3/vTuQbdAMtbYXVlj98zOLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkMfi5MUOmNtpi1apVsz6HX3XSSSfN6/W6jYEhv82PbO54/NJL11fPWbNmzbz3sX379kofl877vbqZmJjo6/06ue6aawfdAkPMGruwrLEtdm4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEjDtIQZ8mnNmVu5cmVPtRqfomW2mqap1jZ9/wfV2pFPe2q11m3yQW1aQr9995rrOh5/5nG/Oe/36vYcw2z5s3zmrLF7ZucWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIwCoyedBs3smHDhp6uWRs5MqqjSBhOP/j+pp5qo2AhRn7VrF+/vm/3gsXGGjs3dm4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA2jwOiqNo6k11Ek69atq9bWrl3b0zVhMTl19eq+3atpmr7dCxYja+zCsHMLAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkYBUZXF1xwwazP2bhxY7W2mEaRwEJYuqR/exLr16/v271gMbLGLgw7twAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQRmmapn83K6V/N2PGNmzYUK2tXLly1tcrpcyhm3yaphnJJ2R8fNzrdUAmJib6dq/rr/l2tXb7XT/qWx/DYmpqaiRfrxHW2GFljV1YndZYO7cAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQxNugG6I9un8js5dOaq1at6r0ZIJ5/4vGDbiEiFudEBJhv1tjhYucWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIwCiyRbuNGNmzY0NM1161b1/H4xo0be7oeLCallGrtoAMP6Vsfk5OTfbsXZGWNHR12bgEASEO4BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDaPAEpnvUSQREWvXru2xG2DNmjV9vV+zfXtf7weLiTV2dNi5BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0jAIbMb2OIulm48aN835NWExWr1496BYiIuLSSy8bdAsw0qyxOdi5BQAgDeEWAIA0hFsAANIQbgEASEO4BQAgDdMShlTtE5srV67s6XqrVq2q1nySE/bs6KOeXq0tWdK/fYI7frqpWmui6VsfMMqssbnZuQUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIwCG1K9jCNZt25dtWYUCczNk454yqBbiIiI66+eGnQLMPKssbnZuQUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIwCG6C1a9fO+pxu40Z6uR6w00tOeXG1tqSUPnZS1zTNoFuAkWCNXbzs3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGqWfY2VKKWbYTNPLc79u3bqe7mWEyeA0TTMcM6RmaXx8POXr9dDHHlKtHfe84+f1Xnff/dNqbenS+t7CVVddVa1t3759Tj3R3dTU1Ei+XiOssbuzxi4OndZYO7cAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkMbYoBvIbr7Hg1xwwQXzer0II0xYXJ57/HHzfs3ayKHHPvbgnq43Nlb/o3nz5s09XRMyssbSiZ1bAADSEG4BAEhDuAUAIA3hFgCANIRbAADSKLVP+S7IzUrp381GwIYNG6q1r371q7O+3saNG3uqsbCapimD7qEX4+PjXq8sOlNTUyP5eo2wxu7OGrs4dFpj7dwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBpGgcECMwoMRodRYDBajAIDACA14RYAgDSEWwAA0hBuAQBIQ7gFACAN4RYAgDT6OgoMAAAWkp1bAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSGOvnzUopTT/vB8OgaZoy6B56MT4+7vXKojM1NTWSr9cIayyLU6c11s4tAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGsItAABpjA26AYBhsXRp/e/7TdPlxErtoIMPrZ5y3HP+Q7W2ZGxpl5sNt8nJyUG3ACxydm4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA2jwBJZtmxZtbbffvtVa29/+9tnfa/HPe5x1dqZZ5456+tFRNx0000dj59//vnVcz772c/2dC9ye8ELXlCt7b///tVaKWUh2llUVq9ZXa1duv7SPnYC88saOzrs3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGkaBDam999674/FuIztOOumkau2EE06o1pqmmXljM9Dr9Y488siOx0899dTqOaM6poT5URvrZdzXAM3znyewEKyxO2VcY+3cAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAaRoEtsMMPP7xaO/bYY6u18847r+PxE088cc497W7btm3V2i233DLr69VGrEREHHbYYbO+HtQ88sgjg26h77qNARqGMWc/ueuuQbfAImKN3ZU1tsXOLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkYRTYDHUbvfGud72rWjvrrLOqtYMPPnhOPc3GL3/5y2rtt3/7t6u1ycnJjse7PR+/8zu/U619/OMfr9Zq7rzzzlmfw+JQ+76+7LLLquesXr16/hupjOdav/7S+ilRH+nVq4mJiXm/5mx969rrBt0CI8gauytr7NzYuQUAIA3hFgCANIRbAADSEG4BAEhDuAUAIA3TEnbzqEc9quPxP//zP6+ec+aZZ857H7fddlvH43fccUf1nI0bN1ZrGzZsqNauuOKKGfe1wymnnFKt9fJpzYiIK6+8suPxP/mTP+npeixe27dvr9Z+9rOfVWsHHXRQtfblL3+5Wnv44Ydn1tgCu/Wm73c8vuLpT5v3ezWVCRHQjTV2Zqyxc2PnFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANIRbAADSMApshu6///5q7bLLLqvWLrnkkmrtRz/60axrt99+e/WcLVu2VGu9Oumkkzoe/8QnPtHT9T75yU9Wa+9+97s7Hh+WMUvk8PWvf33QLSyYhRj5VbN+/fq+3Yv8rLG7ssbOjZ1bAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEijNE3Tv5uV0r+bMWOPetSjqrXauJ+VK1dWz+k2iuScc86p1h588MFqbZQ1TVMG3UMvxsfHvV6H0Eteemq1tmxsad/6mJyc7Nu9+mlqamokX68R1thhZY1dWJ3WWDu3AACkIdwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJDG2KAboD/22Wefau073/lOtbZixYqOx2+88cbqOWeffXa19vDDD1drwJ4Z9wXDxxo7XOzcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAaRoEtEscff3y1VhtFEhFx//33dzx+zjnnVM8xigTmZmJiom/32rpla9/uBVlZY4eLnVsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIwLSGRE088sVq7/PLLq7UHH3ywWlu9enXH41ddddXMGwN+xX777TvoFiIi4vIv1v9sAHayxo4OO7cAAKQh3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZRYCNm5cqV1doXv/jFnq75ohe9qFq79tpre7om0N0LX7iyb/f6+oYr+3YvGGXW2Bzs3AIAkIZwCwBAGsItAABpCLcAAKQh3AIAkIZwCwBAGkaBDam999674/FPf/rT1XPGxuq/nZdcckm1ZhQJLIyJiYlBtxARET978P5BtwBDxRqbm51bAADSEG4BAEhDuAUAIA3hFgCANIRbAADSEG4BAEjDKLABeu5zn1utfeADH+h4/JBDDqmes379+mrtjDPOmHljwKy89KUvHXQLERFx5df+56BbgKFhjV287NwCAJCGcAsAQBrCLQAAaQi3AACkIdwCAJCGaQkD9La3va1aO/HEEzsev/vuu6vnXHDBBXPuCehsxZOeWK2NjfXvj9Km2Vat3f/z+/rWBww7a+ziZecWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIwCmyBvfKVr6zWVq9eXa3deeedHY+fdtpp1XOuv/76mTcGzMrRz/yNQbcQERFX/+u/DLoFGBrWWDqxcwsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaRgFNg8OOuigam3dunXV2oMPPlitveIVr+h4/Nprr515Y8CsvOSUF1drpZQ+dlL3s1t/MugWoK+sscyWnVsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIQbgEASMMosBk68MADq7Vzzz23WnvGM55RrV144YXV2jXXXDOzxoBZOeSxh1Rry/ZaPq/3+slP7qzWli9fVq1dffXV1VrTNHPqCYaRNZb5ZOcWAIA0hFsAANIQbgEASEO4BQAgDeEWAIA0hFsAANIwCmyGVqxYUa2tXbu2WvvoRz9arf3pn/7pHDoCejH+tPFqrduYrVLq1/z5zx/oePyggw+qnrN0SX1voVsfRoGRkTWW+WTnFgCANIRbAADSEG4BAEhDuAUAIA3hFgCANEo/P3lbShn6j/kecMABHY9/4QtfqJ7T7VOeJ510UrV28803z7gvRlfTNF0+Zz+8xsfHh/71CvNtampqJF+vEdbY3VljF4dOa6ydWwAA0hBuAQBIQ7gFACAN4RYAgDSEWwAA0hBuAQBIY2zQDQyb9773vR2PP+UpT6mec8opp1RrRpEAQIs1ln6wcwsAQBrCLQAAaQi3AACkIdwCAJCGcAsAQBrCLQAAaZSmaQbdAwAAzAs7twAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAgCQhnALAEAawi0AAGkItwAApCHcAvD/t1sHMgAAAACD/K3v8RVFABtyCwDAhtwCALAhtwAAbMgtAAAbcgsAwEbG0sbdUQIB5gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 864x1152 with 12 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "my_learner.show_results(rows=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2436df6e9b0>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sample = torch.randn(64, 2)\n",
    "epoch = 5\n",
    "sample = vae.decode(sample).cpu()\n",
    "plt.imshow(sample[0].permute(1, 2, 0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "save_image(sample.view(64, 3, 28, 28), 'results/sample_' + str(epoch) + '.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plot samples in 2D space"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x154162ceef0>"
      ]
     },
     "execution_count": 152,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "emb = vae.encode(data.one_batch()[0])\n",
    "sns.scatterplot(x=emb[0][:, 0], y=emb[0][:, 1])"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "fastai (3.6.7)",
   "language": "python",
   "name": "fastai"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}