Real-NVP_normalizing_flow.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Lab exercise: Real NVP"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import scipy.stats\n",
    "import matplotlib.pyplot as plt\n",
    "import itertools\n",
    "import random\n",
    "import math\n",
    "import time\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.autograd as autograd\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# samples1-2 shape must be (n samples, 2)\n",
    "def plot_samples(samples1, samples2=None):\n",
    "    fig, ax = plt.subplots()\n",
    "\n",
    "    ax.scatter(samples1[:,0], samples1[:,1], marker=\"x\", color=\"blue\")\n",
    "    if samples2 is not None:\n",
    "        ax.scatter(samples2[:,0], samples2[:,1], marker=\"x\", color=\"red\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import sklearn.datasets\n",
    "\n",
    "target_samples, target_classes = sklearn.datasets.make_moons(1000, noise=0.1)\n",
    "target_samples = torch.from_numpy(target_samples).float()\n",
    "\n",
    "plot_samples(target_samples)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "class RealNVPLayer(nn.Module):\n",
    "    def __init__(self, size, reverse=False):\n",
    "        super().__init__()\n",
    "        \n",
    "        self.mask = torch.zeros(size, requires_grad=False)\n",
    "        mid = int(size / 2)\n",
    "        if reverse:\n",
    "            self.mask[mid:] = 1.\n",
    "        else:\n",
    "            self.mask[:mid] = 1.\n",
    "        \n",
    "        ## the two operations\n",
    "        self.scale = nn.Sequential(\n",
    "            nn.Linear(size, 10),\n",
    "            nn.Tanh(),\n",
    "            nn.Linear(10, size),\n",
    "        )\n",
    "        self.transpose = nn.Sequential(\n",
    "            nn.Linear(size, 10),\n",
    "            nn.Tanh(),\n",
    "            nn.Linear(10, size),\n",
    "        )\n",
    "        \n",
    "    # project from the latent space to the observed space,\n",
    "    # i.e. x = g(z)\n",
    "    def forward(self, z):\n",
    "        # you will need this!\n",
    "        n_mask = 1. - self.mask\n",
    "        z_masked = z * self.mask\n",
    "\n",
    "        # s = self.scale(z_masked)*n_mask\n",
    "        # t = self.transpose(z_masked)*n_mask\n",
    "        # x = z_masked + n_mask* (z * torch.exp(s) + t)\n",
    "        s = self.scale(z_masked)\n",
    "        t = self.transpose(z_masked)\n",
    "        x = z_masked + ( n_mask * (z * torch.exp(s) + t) )\n",
    "\n",
    "        return x\n",
    "    \n",
    "    # project from the observed space to the latent space,\n",
    "    # this function also return the log det jacobian of this inv function\n",
    "    def inv(self, x):\n",
    "        # you will need this!\n",
    "        n_mask = 1. - self.mask\n",
    "        x_masked = x * self.mask\n",
    "        \n",
    "        # log_det_jacobian = x.new_zeros(x.shape[0])\n",
    "        # print(log_det_jacobian)\n",
    "        s = self.scale(x_masked)\n",
    "        t = self.transpose(x_masked)\n",
    "        z = ( n_mask * (x-t)*torch.exp(-s)) + x_masked\n",
    "        log_det_jacobian = -s.sum(dim=1)\n",
    "        # print(log_det_jacobian)\n",
    "        # log_det_jacobian = x.new_zeros(x.shape[0])\n",
    "        # log_det_jacobian -= s.sum(dim=1)\n",
    "        # print(log_det_jacobian)\n",
    "\n",
    "        return z, log_det_jacobian\n",
    "\n",
    "layer = RealNVPLayer(2, reverse=False)\n",
    "\n",
    "with torch.no_grad():\n",
    "    x = torch.rand(1, 2)\n",
    "    z, _ = layer.inv(x)\n",
    "    xx = layer(z)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In the 3 vectors below, the first element must be equal\n",
      "This two vectors should be equal:\n",
      "tensor([[0.0813, 0.6206]])\n",
      "tensor([[0.0813, 0.6206]])\n",
      "This vector should be different to the two above\n",
      "tensor([[0.0813, 0.5971]])\n",
      "\n",
      "In the 3 vectors below, the second element must be equal\n",
      "This two vectors should be equal:\n",
      "tensor([[0.3981, 0.3910]])\n",
      "tensor([[0.3981, 0.3910]])\n",
      "This vector should be different to the two above\n",
      "tensor([[0.5981, 0.3910]])\n"
     ]
    }
   ],
   "source": [
    "# Test!\n",
    "layer = RealNVPLayer(2, reverse=False)\n",
    "\n",
    "with torch.no_grad():\n",
    "    x = torch.rand(1, 2)\n",
    "    z, _ = layer.inv(x)\n",
    "    xx = layer(z)\n",
    "\n",
    "    print(\"In the 3 vectors below, the first element must be equal\")\n",
    "    print(\"This two vectors should be equal:\")\n",
    "    print(x)\n",
    "    print(xx)\n",
    "    print(\"This vector should be different to the two above\")\n",
    "    print(z)\n",
    "print()\n",
    "\n",
    "layer = RealNVPLayer(2, reverse=True)\n",
    "\n",
    "with torch.no_grad():\n",
    "    x = torch.rand(1, 2)\n",
    "    z, _ = layer.inv(x)\n",
    "    xx = layer(z)\n",
    "\n",
    "    print(\"In the 3 vectors below, the second element must be equal\")\n",
    "    print(\"This two vectors should be equal:\")\n",
    "    print(x)\n",
    "    print(xx)\n",
    "    print(\"This vector should be different to the two above\")\n",
    "    print(z)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "class RealNVP(nn.Module):\n",
    "    def __init__(self, size, n_layers):\n",
    "        super().__init__()\n",
    "        \n",
    "        self.prior = torch.distributions.normal.Normal(torch.zeros(2), torch.ones(2))\n",
    "        \n",
    "        self.layers = nn.ModuleList(\n",
    "                            RealNVPLayer(size, i % 2 == 0)\n",
    "                            for i in range(n_layers)\n",
    "        )\n",
    "    \n",
    "    def forward(self, z):\n",
    "        x = z\n",
    "        for i in range(len(self.layers)):\n",
    "            x = self.layers[i](x)\n",
    "        return x\n",
    "    \n",
    "    def inv(self, x):\n",
    "        log_det_jacobian = 0.\n",
    "        z = x\n",
    "        for i in reversed(range(len(self.layers))):\n",
    "            z, j = self.layers[i].inv(z)\n",
    "            # remember here, we just have to sum all log det jacobians!\n",
    "            log_det_jacobian = log_det_jacobian + j\n",
    "        return z, log_det_jacobian\n",
    "\n",
    "    \n",
    "    def sample(self, n_samples):\n",
    "        z = self.prior.sample((n_samples,))\n",
    "        x = self(z)\n",
    "        return x\n",
    "\n",
    "    def log_prior(self, z):\n",
    "        x, det = self.inv(z)\n",
    "        ret = self.prior.log_prob(x).sum(1) + det\n",
    "        \n",
    "        return ret"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loss 9.982192993164062\n",
      "loss 8.60582447052002\n",
      "loss 7.847801208496094\n",
      "loss 7.424689292907715\n",
      "loss 7.180319309234619\n",
      "loss 6.98622465133667\n",
      "loss 6.78571081161499\n",
      "loss 6.567846298217773\n",
      "loss 6.3333611488342285\n",
      "loss 6.084957599639893\n",
      "loss 5.825464725494385\n",
      "loss 5.558241844177246\n",
      "loss 5.289393424987793\n",
      "loss 5.033022880554199\n",
      "loss 4.815653324127197\n",
      "loss 4.65077543258667\n",
      "loss 4.499691486358643\n",
      "loss 4.32443380355835\n",
      "loss 4.121725559234619\n",
      "loss 3.9086854457855225\n",
      "loss 3.703927755355835\n",
      "loss 3.515209436416626\n",
      "loss 3.3407680988311768\n",
      "loss 3.1805546283721924\n",
      "loss 3.038893461227417\n",
      "loss 2.9191460609436035\n",
      "loss 2.816019296646118\n",
      "loss 2.7210958003997803\n",
      "loss 2.637404680252075\n",
      "loss 2.5736019611358643\n",
      "loss 2.5265893936157227\n",
      "loss 2.4866750240325928\n",
      "loss 2.448934316635132\n",
      "loss 2.4081568717956543\n",
      "loss 2.3612639904022217\n",
      "loss 2.3126723766326904\n",
      "loss 2.274400472640991\n",
      "loss 2.2511115074157715\n",
      "loss 2.2299551963806152\n",
      "loss 2.203063726425171\n",
      "loss 2.1768596172332764\n",
      "loss 2.157109022140503\n",
      "loss 2.140624523162842\n",
      "loss 2.1214346885681152\n",
      "loss 2.098905563354492\n",
      "loss 2.0781726837158203\n",
      "loss 2.0626299381256104\n",
      "loss 2.048015594482422\n",
      "loss 2.0293920040130615\n",
      "loss 2.0096757411956787\n",
      "loss 1.9941173791885376\n",
      "loss 1.9826406240463257\n",
      "loss 1.9714555740356445\n",
      "loss 1.9583485126495361\n",
      "loss 1.9444565773010254\n",
      "loss 1.9320117235183716\n",
      "loss 1.921371340751648\n",
      "loss 1.910897135734558\n",
      "loss 1.8997249603271484\n",
      "loss 1.8891535997390747\n",
      "loss 1.880518913269043\n",
      "loss 1.873205304145813\n",
      "loss 1.8656784296035767\n",
      "loss 1.857444405555725\n",
      "loss 1.8489741086959839\n",
      "loss 1.840224266052246\n",
      "loss 1.8307310342788696\n",
      "loss 1.820995569229126\n",
      "loss 1.8119218349456787\n",
      "loss 1.8034391403198242\n",
      "loss 1.7950547933578491\n",
      "loss 1.7867881059646606\n",
      "loss 1.778714656829834\n",
      "loss 1.770326852798462\n",
      "loss 1.7611130475997925\n",
      "loss 1.7514020204544067\n",
      "loss 1.741861343383789\n",
      "loss 1.7325224876403809\n",
      "loss 1.7230936288833618\n",
      "loss 1.7137023210525513\n",
      "loss 1.7044708728790283\n",
      "loss 1.6949992179870605\n",
      "loss 1.685046911239624\n",
      "loss 1.6749062538146973\n",
      "loss 1.6647390127182007\n",
      "loss 1.6544326543807983\n",
      "loss 1.6440678834915161\n",
      "loss 1.633771538734436\n",
      "loss 1.6233893632888794\n",
      "loss 1.6128525733947754\n",
      "loss 1.6022734642028809\n",
      "loss 1.5915806293487549\n",
      "loss 1.5807232856750488\n",
      "loss 1.569859266281128\n",
      "loss 1.5589604377746582\n",
      "loss 1.5479155778884888\n",
      "loss 1.5368599891662598\n",
      "loss 1.5257948637008667\n",
      "loss 1.5146161317825317\n",
      "loss 1.503422737121582\n",
      "loss 1.492227554321289\n",
      "loss 1.4810073375701904\n",
      "loss 1.4698750972747803\n",
      "loss 1.4588209390640259\n",
      "loss 1.4478363990783691\n",
      "loss 1.4369478225708008\n",
      "loss 1.4261503219604492\n",
      "loss 1.4154983758926392\n",
      "loss 1.4050203561782837\n",
      "loss 1.3947588205337524\n",
      "loss 1.384676456451416\n",
      "loss 1.3747620582580566\n",
      "loss 1.36503005027771\n",
      "loss 1.3554352521896362\n",
      "loss 1.345996618270874\n",
      "loss 1.3366541862487793\n",
      "loss 1.3274247646331787\n",
      "loss 1.3183343410491943\n",
      "loss 1.3094103336334229\n",
      "loss 1.3006885051727295\n",
      "loss 1.2922173738479614\n",
      "loss 1.2840797901153564\n",
      "loss 1.2763702869415283\n",
      "loss 1.2691463232040405\n",
      "loss 1.2624609470367432\n",
      "loss 1.2563261985778809\n",
      "loss 1.2507253885269165\n",
      "loss 1.2456092834472656\n",
      "loss 1.240912675857544\n",
      "loss 1.236572027206421\n",
      "loss 1.2325458526611328\n",
      "loss 1.2287815809249878\n",
      "loss 1.2252475023269653\n",
      "loss 1.2219167947769165\n",
      "loss 1.2187697887420654\n",
      "loss 1.2157843112945557\n",
      "loss 1.2129548788070679\n",
      "loss 1.210269808769226\n",
      "loss 1.2077221870422363\n",
      "loss 1.2052949666976929\n",
      "loss 1.2029732465744019\n",
      "loss 1.2007465362548828\n",
      "loss 1.1986087560653687\n",
      "loss 1.1965620517730713\n",
      "loss 1.1946089267730713\n",
      "loss 1.1927467584609985\n",
      "loss 1.1909538507461548\n",
      "loss 1.189197063446045\n",
      "loss 1.1874479055404663\n",
      "loss 1.1856908798217773\n",
      "loss 1.1839303970336914\n",
      "loss 1.1821777820587158\n",
      "loss 1.1804487705230713\n",
      "loss 1.178757667541504\n",
      "loss 1.1771154403686523\n",
      "loss 1.175528645515442\n",
      "loss 1.1739975214004517\n",
      "loss 1.1725175380706787\n",
      "loss 1.1710842847824097\n",
      "loss 1.169690728187561\n",
      "loss 1.1683297157287598\n",
      "loss 1.1669950485229492\n",
      "loss 1.1656856536865234\n",
      "loss 1.1644206047058105\n",
      "loss 1.1632587909698486\n",
      "loss 1.162505865097046\n",
      "loss 1.162582278251648\n",
      "loss 1.1668806076049805\n",
      "loss 1.1584229469299316\n",
      "loss 1.1573989391326904\n",
      "loss 1.1624140739440918\n",
      "loss 1.1556079387664795\n",
      "loss 1.1529340744018555\n",
      "loss 1.1523128747940063\n",
      "loss 1.1536272764205933\n",
      "loss 1.15632164478302\n",
      "loss 1.147647738456726\n",
      "loss 1.153793454170227\n",
      "loss 1.149688720703125\n",
      "loss 1.1450105905532837\n",
      "loss 1.144106388092041\n",
      "loss 1.1465874910354614\n",
      "loss 1.144079327583313\n",
      "loss 1.1416257619857788\n",
      "loss 1.1387418508529663\n",
      "loss 1.1379400491714478\n",
      "loss 1.1383757591247559\n",
      "loss 1.1377955675125122\n",
      "loss 1.137735366821289\n",
      "loss 1.134531021118164\n",
      "loss 1.1323261260986328\n",
      "loss 1.1308969259262085\n",
      "loss 1.1301465034484863\n",
      "loss 1.1301953792572021\n",
      "loss 1.129753589630127\n",
      "loss 1.1303304433822632\n",
      "loss 1.1275581121444702\n",
      "loss 1.1258798837661743\n",
      "loss 1.1236552000045776\n",
      "loss 1.1221888065338135\n",
      "loss 1.1212618350982666\n",
      "loss 1.120635747909546\n",
      "loss 1.120758056640625\n",
      "loss 1.1210275888442993\n",
      "loss 1.1223448514938354\n",
      "loss 1.1164737939834595\n",
      "loss 1.1160560846328735\n",
      "loss 1.1203001737594604\n",
      "loss 1.1140203475952148\n",
      "loss 1.112449049949646\n",
      "loss 1.1140025854110718\n",
      "loss 1.1182478666305542\n",
      "loss 1.1095207929611206\n",
      "loss 1.1131895780563354\n",
      "loss 1.1144174337387085\n",
      "loss 1.107280969619751\n",
      "loss 1.110655665397644\n",
      "loss 1.1117397546768188\n",
      "loss 1.1048415899276733\n",
      "loss 1.1068224906921387\n",
      "loss 1.1157060861587524\n",
      "loss 1.103929042816162\n",
      "loss 1.1318402290344238\n",
      "loss 1.130195140838623\n",
      "loss 1.1014926433563232\n",
      "loss 1.1203969717025757\n",
      "loss 1.1006916761398315\n",
      "loss 1.1279138326644897\n",
      "loss 1.1254444122314453\n",
      "loss 1.098921537399292\n",
      "loss 1.1073315143585205\n",
      "loss 1.1002212762832642\n",
      "loss 1.0976076126098633\n",
      "loss 1.0994179248809814\n",
      "loss 1.1019052267074585\n",
      "loss 1.0947715044021606\n",
      "loss 1.1097192764282227\n",
      "loss 1.0968708992004395\n",
      "loss 1.1240366697311401\n",
      "loss 1.132638931274414\n",
      "loss 1.097824215888977\n",
      "loss 1.1618417501449585\n",
      "loss 1.2257939577102661\n",
      "loss 1.1710636615753174\n",
      "loss 1.1223857402801514\n",
      "loss 1.1278756856918335\n",
      "loss 1.1510673761367798\n",
      "loss 1.1427756547927856\n",
      "loss 1.1053684949874878\n",
      "loss 1.11670982837677\n",
      "loss 1.1273229122161865\n",
      "loss 1.09279465675354\n",
      "loss 1.152034878730774\n",
      "loss 1.2125344276428223\n",
      "loss 1.1644831895828247\n",
      "loss 1.1260762214660645\n",
      "loss 1.1223260164260864\n",
      "loss 1.1349356174468994\n",
      "loss 1.1382001638412476\n",
      "loss 1.114533543586731\n",
      "loss 1.0981067419052124\n",
      "loss 1.1047805547714233\n",
      "loss 1.090149164199829\n",
      "loss 1.123228907585144\n",
      "loss 1.1433591842651367\n",
      "loss 1.0961123704910278\n",
      "loss 1.1327507495880127\n",
      "loss 1.1669310331344604\n",
      "loss 1.129019021987915\n",
      "loss 1.1041620969772339\n",
      "loss 1.11449134349823\n",
      "loss 1.1186254024505615\n",
      "loss 1.1109094619750977\n",
      "loss 1.0954618453979492\n",
      "loss 1.104225754737854\n",
      "loss 1.1012998819351196\n",
      "loss 1.0918159484863281\n",
      "loss 1.0960339307785034\n",
      "loss 1.0877636671066284\n",
      "loss 1.089713454246521\n",
      "loss 1.088317632675171\n",
      "loss 1.0878151655197144\n",
      "loss 1.0876153707504272\n",
      "loss 1.0830577611923218\n",
      "loss 1.0920398235321045\n",
      "loss 1.0832798480987549\n",
      "loss 1.0872899293899536\n",
      "loss 1.086268424987793\n",
      "loss 1.0818848609924316\n",
      "loss 1.0922014713287354\n",
      "loss 1.08656644821167\n",
      "loss 1.0865355730056763\n",
      "loss 1.0878493785858154\n",
      "loss 1.0877538919448853\n",
      "loss 1.0815595388412476\n",
      "loss 1.0925908088684082\n",
      "loss 1.0924928188323975\n",
      "loss 1.0788326263427734\n",
      "loss 1.0788058042526245\n",
      "loss 1.0879850387573242\n",
      "loss 1.0809400081634521\n",
      "loss 1.0875126123428345\n",
      "loss 1.090301275253296\n",
      "loss 1.074101448059082\n",
      "loss 1.0789762735366821\n",
      "loss 1.0758321285247803\n",
      "loss 1.0722192525863647\n",
      "loss 1.0753586292266846\n",
      "loss 1.0764352083206177\n",
      "loss 1.0723655223846436\n",
      "loss 1.0760297775268555\n",
      "loss 1.0718083381652832\n",
      "loss 1.0705695152282715\n",
      "loss 1.073174238204956\n",
      "loss 1.0715223550796509\n",
      "loss 1.0695008039474487\n",
      "loss 1.0690034627914429\n",
      "loss 1.068286418914795\n",
      "loss 1.0688953399658203\n",
      "loss 1.0687918663024902\n",
      "loss 1.069304347038269\n",
      "loss 1.0690339803695679\n",
      "loss 1.0683351755142212\n",
      "loss 1.066861629486084\n",
      "loss 1.0660146474838257\n",
      "loss 1.0651289224624634\n",
      "loss 1.0648529529571533\n",
      "loss 1.064474105834961\n",
      "loss 1.0644463300704956\n",
      "loss 1.0645986795425415\n",
      "loss 1.065191626548767\n",
      "loss 1.066392183303833\n",
      "loss 1.063102126121521\n",
      "loss 1.0622050762176514\n",
      "loss 1.0628979206085205\n",
      "loss 1.0659701824188232\n",
      "loss 1.0615959167480469\n",
      "loss 1.0608806610107422\n",
      "loss 1.0628399848937988\n",
      "loss 1.0649895668029785\n",
      "loss 1.0598809719085693\n",
      "loss 1.0631117820739746\n",
      "loss 1.0631660223007202\n",
      "loss 1.058671236038208\n",
      "loss 1.0608001947402954\n",
      "loss 1.0666112899780273\n",
      "loss 1.0591813325881958\n",
      "loss 1.082940936088562\n",
      "loss 1.081251859664917\n",
      "loss 1.0587149858474731\n",
      "loss 1.0695258378982544\n",
      "loss 1.0575437545776367\n",
      "loss 1.0760343074798584\n",
      "loss 1.0680594444274902\n",
      "loss 1.065190315246582\n",
      "loss 1.0646418333053589\n",
      "loss 1.0623383522033691\n",
      "loss 1.0597202777862549\n",
      "loss 1.0718390941619873\n",
      "loss 1.0716379880905151\n",
      "loss 1.0569610595703125\n",
      "loss 1.056421160697937\n",
      "loss 1.066961646080017\n",
      "loss 1.0600625276565552\n",
      "loss 1.0714646577835083\n",
      "loss 1.0743727684020996\n",
      "loss 1.054841160774231\n",
      "loss 1.066699504852295\n",
      "loss 1.055620551109314\n",
      "loss 1.0797183513641357\n",
      "loss 1.0858073234558105\n",
      "loss 1.0541768074035645\n",
      "loss 1.0910223722457886\n",
      "loss 1.1038886308670044\n",
      "loss 1.0602388381958008\n",
      "loss 1.0914499759674072\n",
      "loss 1.1307003498077393\n",
      "loss 1.0926562547683716\n",
      "loss 1.0624135732650757\n",
      "loss 1.0868154764175415\n",
      "loss 1.0884956121444702\n",
      "loss 1.065011739730835\n",
      "loss 1.0753726959228516\n",
      "loss 1.085209608078003\n",
      "loss 1.082130789756775\n",
      "loss 1.0677766799926758\n",
      "loss 1.0596436262130737\n",
      "loss 1.0689282417297363\n",
      "loss 1.0576422214508057\n",
      "loss 1.0755265951156616\n",
      "loss 1.0866254568099976\n",
      "loss 1.0697004795074463\n",
      "loss 1.0664701461791992\n",
      "loss 1.0717836618423462\n",
      "loss 1.0796290636062622\n",
      "loss 1.0704883337020874\n",
      "loss 1.05495285987854\n",
      "loss 1.059410572052002\n",
      "loss 1.0582733154296875\n",
      "loss 1.0519086122512817\n",
      "loss 1.0639911890029907\n",
      "loss 1.0622594356536865\n",
      "loss 1.0566543340682983\n",
      "loss 1.0565954446792603\n",
      "loss 1.0627529621124268\n",
      "loss 1.055700659751892\n",
      "loss 1.0562641620635986\n",
      "loss 1.057246446609497\n",
      "loss 1.0556082725524902\n",
      "loss 1.0523203611373901\n",
      "loss 1.059198260307312\n",
      "loss 1.0576437711715698\n",
      "loss 1.0528711080551147\n",
      "loss 1.0513451099395752\n",
      "loss 1.058089256286621\n",
      "loss 1.0564254522323608\n",
      "loss 1.0513688325881958\n",
      "loss 1.0500444173812866\n",
      "loss 1.0562363862991333\n",
      "loss 1.0523113012313843\n",
      "loss 1.0542261600494385\n",
      "loss 1.053604245185852\n",
      "loss 1.0515871047973633\n",
      "loss 1.0505971908569336\n",
      "loss 1.0528814792633057\n",
      "loss 1.0517126321792603\n",
      "loss 1.0495718717575073\n",
      "loss 1.0473828315734863\n",
      "loss 1.056572675704956\n",
      "loss 1.0547643899917603\n",
      "loss 1.0472129583358765\n",
      "loss 1.0460292100906372\n",
      "loss 1.0538203716278076\n",
      "loss 1.0500991344451904\n",
      "loss 1.0511990785598755\n",
      "loss 1.0506246089935303\n",
      "loss 1.0473082065582275\n",
      "loss 1.0456184148788452\n",
      "loss 1.0544013977050781\n",
      "loss 1.0532063245773315\n",
      "loss 1.0449154376983643\n",
      "loss 1.043308973312378\n",
      "loss 1.0521475076675415\n",
      "loss 1.0468500852584839\n",
      "loss 1.054449200630188\n",
      "loss 1.0567432641983032\n",
      "loss 1.0420405864715576\n",
      "loss 1.0495402812957764\n",
      "loss 1.0413906574249268\n",
      "loss 1.0490987300872803\n",
      "loss 1.0412249565124512\n",
      "loss 1.0555658340454102\n",
      "loss 1.049597144126892\n",
      "loss 1.0476722717285156\n",
      "loss 1.0492006540298462\n",
      "loss 1.0423908233642578\n",
      "loss 1.042283058166504\n",
      "loss 1.0469213724136353\n",
      "loss 1.042471170425415\n",
      "loss 1.0536079406738281\n",
      "loss 1.053889274597168\n",
      "loss 1.0400059223175049\n",
      "loss 1.042900800704956\n",
      "loss 1.0421346426010132\n",
      "loss 1.038349986076355\n",
      "loss 1.0436204671859741\n",
      "loss 1.038075566291809\n",
      "loss 1.0419456958770752\n",
      "loss 1.0394772291183472\n",
      "loss 1.037128210067749\n",
      "loss 1.0381988286972046\n",
      "loss 1.037680745124817\n",
      "loss 1.0376079082489014\n",
      "loss 1.036316156387329\n",
      "loss 1.0365443229675293\n",
      "loss 1.0361295938491821\n",
      "loss 1.0363850593566895\n",
      "loss 1.0358166694641113\n",
      "loss 1.0354841947555542\n",
      "loss 1.034980058670044\n",
      "loss 1.035084843635559\n",
      "loss 1.035064458847046\n",
      "loss 1.0351327657699585\n",
      "loss 1.0349252223968506\n",
      "loss 1.0342824459075928\n",
      "loss 1.0337456464767456\n",
      "loss 1.0335049629211426\n",
      "loss 1.0338510274887085\n",
      "loss 1.0340266227722168\n",
      "loss 1.0347117185592651\n",
      "loss 1.0343120098114014\n",
      "loss 1.033347249031067\n",
      "loss 1.0321539640426636\n",
      "loss 1.0322154760360718\n",
      "loss 1.0330637693405151\n",
      "loss 1.0345087051391602\n",
      "loss 1.0324628353118896\n",
      "loss 1.031381607055664\n",
      "loss 1.0316637754440308\n",
      "loss 1.0330958366394043\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x29a4f1751e0>]"
      ]
     },
     "execution_count": 87,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "trained_distrib = RealNVP(2, 50)\n",
    "optimizer = torch.optim.Adam(trained_distrib.parameters(), lr=1e-3)\n",
    "\n",
    "batch_size = 1000\n",
    "losses = list()\n",
    "n_epoch = 500\n",
    "for _ in range(n_epoch):\n",
    "    for i in range(0, target_samples.shape[0], batch_size):\n",
    "        batch = target_samples[i:i+batch_size]\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        loss = -trained_distrib.log_prior(batch).mean()\n",
    "        print(\"loss {}\".format(loss.item()))\n",
    "        losses.append(loss.item())\n",
    "\n",
    "        loss.backward()\n",
    "        torch.nn.utils.clip_grad_norm_(trained_distrib.parameters(), 5)\n",
    "        optimizer.step()\n",
    "    \n",
    "plt.plot(np.arange(len(losses)), losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# sample from the model\n",
    "with torch.no_grad():\n",
    "    samples = trained_distrib.sample(1000)\n",
    "    plot_samples(target_samples, samples)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# print the latent space corresponding to each half moon in a different color\n",
    "with torch.no_grad():\n",
    "    source_sample1, _ = trained_distrib.inv(target_samples[target_classes == 0])\n",
    "    source_sample2, _ = trained_distrib.inv(target_samples[target_classes == 1])\n",
    "    plot_samples(source_sample1, source_sample2)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.4 64-bit",
   "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": "3.10.4"
  },
  "vscode": {
   "interpreter": {
    "hash": "33184612fe23ea789693bd8266f7c5cf5860109e6e433b1efdda378c5f430ce8"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}