wgan.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "from torch.nn import functional as F\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import torchsummary\n",
    "import math\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Number of workers for dataloader\n",
    "workers = 4\n",
    "\n",
    "# Batch size during training\n",
    "batch_size = 128\n",
    "\n",
    "# Spatial size of training images. All images will be resized to this\n",
    "#   size using a transformer.\n",
    "image_size = 64\n",
    "\n",
    "# Number of channels in the training images. For color images this is 3\n",
    "nc = 3\n",
    "\n",
    "# Size of z latent vector (i.e. size of generator input)\n",
    "nz = 100\n",
    "\n",
    "# Size of feature maps in generator\n",
    "ngf = 64\n",
    "\n",
    "# Size of feature maps in discriminator\n",
    "ndf = 64\n",
    "\n",
    "# Number of training epochs\n",
    "num_epochs = 5\n",
    "\n",
    "# Learning rate for optimizers\n",
    "lr = 0.0002\n",
    "\n",
    "# Beta1 hyperparam for Adam optimizers\n",
    "beta1 = 0.5\n",
    "\n",
    "# Number of GPUs available. Use 0 for CPU mode.\n",
    "ngpu = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----------------------------------------------------------------\n",
      "        Layer (type)               Output Shape         Param #\n",
      "================================================================\n",
      "            Conv2d-1           [-1, 64, 32, 32]           3,072\n",
      "         LeakyReLU-2           [-1, 64, 32, 32]               0\n",
      "            Conv2d-3          [-1, 128, 16, 16]         131,072\n",
      "         LeakyReLU-4          [-1, 128, 16, 16]               0\n",
      "            Conv2d-5            [-1, 256, 8, 8]         524,288\n",
      "         LeakyReLU-6            [-1, 256, 8, 8]               0\n",
      "            Conv2d-7            [-1, 512, 4, 4]       2,097,152\n",
      "         LeakyReLU-8            [-1, 512, 4, 4]               0\n",
      "            Conv2d-9              [-1, 1, 1, 1]           8,192\n",
      "================================================================\n",
      "Total params: 2,763,776\n",
      "Trainable params: 2,763,776\n",
      "Non-trainable params: 0\n",
      "----------------------------------------------------------------\n",
      "Input size (MB): 0.05\n",
      "Forward/backward pass size (MB): 1.88\n",
      "Params size (MB): 10.54\n",
      "Estimated Total Size (MB): 12.46\n",
      "----------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "class Discriminator(nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        self.layers = nn.Sequential(\n",
    "            nn.Conv2d(3, 64, 4, 2, 1, bias=False),\n",
    "            nn.LeakyReLU(0.2, inplace=True),\n",
    "            \n",
    "            nn.Conv2d(64, 128, 4, 2, 1, bias=False),\n",
    "            # nn.BatchNorm2d(128),\n",
    "            nn.LeakyReLU(0.2, inplace=True),\n",
    "            \n",
    "            nn.Conv2d(128, 256, 4, 2, 1, bias=False),\n",
    "            # nn.BatchNorm2d(256),\n",
    "            nn.LeakyReLU(0.2, inplace=True),\n",
    "            \n",
    "            nn.Conv2d(256, 512, 4, 2, 1, bias=False),\n",
    "            # nn.BatchNorm2d(512),\n",
    "            nn.LeakyReLU(0.2, inplace=True),\n",
    "            \n",
    "            nn.Conv2d(512, 1, 4, 1, 0, bias=False),\n",
    "            # nn.Sigmoid()\n",
    "        )\n",
    "        \n",
    "        self.reset_weights()\n",
    "    \n",
    "    def forward(self, input):\n",
    "        return self.layers(input)\n",
    "    \n",
    "    def reset_weights(self):\n",
    "        for layer in self.layers:\n",
    "            if layer.__class__.__name__.find(\"Conv\") != -1:\n",
    "                nn.init.normal_(layer.weight.data, 0.0, 0.02)\n",
    "            elif layer.__class__.__name__.find(\"BatchNorm\") != -1:\n",
    "                nn.init.normal_(layer.weight.data, 1.0, 0.02)\n",
    "                nn.init.constant_(layer.bias.data, 0)\n",
    "    \n",
    "torchsummary.summary(Discriminator(), [3, 64, 64])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----------------------------------------------------------------\n",
      "        Layer (type)               Output Shape         Param #\n",
      "================================================================\n",
      "   ConvTranspose2d-1            [-1, 512, 4, 4]         819,200\n",
      "       BatchNorm2d-2            [-1, 512, 4, 4]           1,024\n",
      "              ReLU-3            [-1, 512, 4, 4]               0\n",
      "   ConvTranspose2d-4            [-1, 256, 8, 8]       2,097,152\n",
      "       BatchNorm2d-5            [-1, 256, 8, 8]             512\n",
      "              ReLU-6            [-1, 256, 8, 8]               0\n",
      "   ConvTranspose2d-7          [-1, 128, 16, 16]         524,288\n",
      "       BatchNorm2d-8          [-1, 128, 16, 16]             256\n",
      "              ReLU-9          [-1, 128, 16, 16]               0\n",
      "  ConvTranspose2d-10           [-1, 64, 32, 32]         131,072\n",
      "      BatchNorm2d-11           [-1, 64, 32, 32]             128\n",
      "             ReLU-12           [-1, 64, 32, 32]               0\n",
      "  ConvTranspose2d-13            [-1, 3, 64, 64]           3,072\n",
      "             Tanh-14            [-1, 3, 64, 64]               0\n",
      "================================================================\n",
      "Total params: 3,576,704\n",
      "Trainable params: 3,576,704\n",
      "Non-trainable params: 0\n",
      "----------------------------------------------------------------\n",
      "Input size (MB): 0.00\n",
      "Forward/backward pass size (MB): 3.00\n",
      "Params size (MB): 13.64\n",
      "Estimated Total Size (MB): 16.64\n",
      "----------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "class Generator(nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        \n",
    "        self.layers = nn.Sequential(            \n",
    "            nn.ConvTranspose2d(100, 512, 4, 1, 0, bias=False),\n",
    "            nn.BatchNorm2d(512),\n",
    "            nn.ReLU(inplace=True),\n",
    "            \n",
    "            nn.ConvTranspose2d(512, 256, 4, 2, 1, 0, bias=False),\n",
    "            nn.BatchNorm2d(256),\n",
    "            nn.ReLU(inplace=True),\n",
    "            \n",
    "            nn.ConvTranspose2d(256, 128, 4, 2, 1, 0, bias=False),\n",
    "            nn.BatchNorm2d(128),\n",
    "            nn.ReLU(inplace=True),\n",
    "            \n",
    "            nn.ConvTranspose2d(128, 64, 4, 2, 1, 0, bias=False),\n",
    "            nn.BatchNorm2d(64),\n",
    "            nn.ReLU(inplace=True),\n",
    "            \n",
    "            nn.ConvTranspose2d(64, 3, 4, 2, 1, 0, bias=False),\n",
    "            nn.Tanh(),\n",
    "        )\n",
    "        \n",
    "        self.reset_weights()\n",
    "    \n",
    "    def forward(self, input):\n",
    "        return self.layers(input)\n",
    "    \n",
    "    def reset_weights(self):\n",
    "        for layer in self.layers:\n",
    "            if layer.__class__.__name__.find(\"Conv\") != -1:\n",
    "                nn.init.normal_(layer.weight.data, 0.0, 0.02)\n",
    "            elif layer.__class__.__name__.find(\"BatchNorm\") != -1:\n",
    "                nn.init.normal_(layer.weight.data, 1.0, 0.02)\n",
    "                nn.init.constant_(layer.bias.data, 0)\n",
    "                \n",
    "torchsummary.summary(Generator(), input_size=[100, 1, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "import fastai"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [],
   "source": [
    "def set_trainable(mod, is_trainable):\n",
    "    for param in mod.parameters():\n",
    "        param.require_grad = is_trainable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2, imutils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [],
   "source": [
    "import fastai.vision.data as data\n",
    "from fastai import Path\n",
    "images_ds = data.ImageClassificationDataset.from_single_folder(Path(\"/Users/alvaro/Downloads/pokemon/\"), classes=[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [],
   "source": [
    "def transform_image(image_tensor):\n",
    "    image = image_tensor.data.numpy().swapaxes(0, 2)\n",
    "    image = imutils.resize(image, 64, 64)\n",
    "    image = (image - 0.5) / 0.5\n",
    "    return torch.Tensor(image.swapaxes(0, 2))\n",
    "\n",
    "def collate_fn(images):\n",
    "    return torch.stack(list(transform_image(i[0].data) for i in images))\n",
    "\n",
    "batch_size = 16\n",
    "dl = torch.utils.data.DataLoader(images_ds, batch_size=batch_size, collate_fn=collate_fn)\n",
    "num_batches = math.ceil(len(images_ds)/batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x10e7e91d0>"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvIxREBQAAIABJREFUeJztnXl0XNdxp3/VjcZCgiCIhSBIkAQp\n7hIlUiu1S5YlS46X2JYVL+PReJQoJ5Pk2HG8ZplxEidjz8yJ7ZmTUSyPnegktrzGkSxvkmnJliWZ\nErVRpLhT3BeQBBdsvd/5A81XC9AgKAINUq++c3hYr+/t926/7otXdatuFYUQ4DhOvEhM9AAcx6k8\nPvEdJ4b4xHecGOIT33FiiE98x4khPvEdJ4b4xHecGHJWE5+IbieizUS0jYg+PVaDchxnfKHXG8BD\nREkAWwDcCmAvgOcAvD+E8OrYDc9xnPGg6izeeyWAbSGEHQBARN8C8E4AZSc+EXmYoOOMMyEEOl2f\ns1H1ZwHYI473ll5zHOcc52ye+KOCiO4FcO94X8dxnNFzNhN/H4DZ4rij9JoihHA/gPsBV/Ud51zh\nbFT95wAsJKJ5RFQN4H0AHh6bYTmOM5687id+CCFPRH8E4GcAkgC+HkLYMGYjcxxn3Hjd7rzXdTFX\n9R1n3BnNqv64L+45zvlGx0Wd6vjQ9v2RnBvIVng044OH7DpODPGJ7zgxxFV9xzG8+4/fp4437Nga\nyau/8P2y71twxdJIXvqWlarth5/75hiNbmzwJ77jxBCf+I4TQ3ziO04McRvfcQDc8uG3RvL0RTNU\nW38byzVTJ6m2uz72wUhe9uYVkfyjB380xiMcW/yJ7zgxxCe+48QQV/Wd+EAcyXrTB29VTTd/6LZI\nDkNmRTGSrvkPN6uWy+9YFcm7jx2I5D0vv3YWAx1//InvODHEJ77jxBBX9Z3zntr6ukieu3y+arvg\n0sWRfOENl0RyQ8tU1S9UsRlQDAXV1rWVVfhLrrlYtfXkelkeGIjkVEJPrbd86nciefrCmarte5/4\naiQPHOtFJfAnvuPEEJ/4jhNDfOI7TgxxG98ZwoKViyJ52ZUXRnLHwtmq39TmxkhO1NaqtupJNZFc\nSLHNnM9r+znTl47kniMnVFsiyXb3lGa2yRtaG1S/wdouJUbIKKVakjpJDQkbf8faraptXgdnjZ8+\nt1215fL5SC5m2e23+MZLVL9ps5ojeeuT61Vbpex6iT/xHSeG+MR3nBjiqn5MqJ82RR1f9+4bIvmG\nd9yo2trmaHX2FAStHudyrDzn80XV1pfr437IRXJNUavidcIV1zhj2rDXHYLV5oV6T0aFDwVuy2Y4\nX97hvQdVv0KWVfa5i+eqtkkNvDGnYEyJIr8NR17riuTGmU16jOIRu+7hZzDR+BPfcWKIT3zHiSE+\n8R0nhriN/wajurY6km/7z3dE8nV33qT6NTeyDZqU7jCDNGmzOW3f9mc4RDVdSKu2UMV9i/J95lGT\nEE1Fa7sLc71Y4DWE9Il+1a2mnl2JqaqUPoVw02WO8xhnX9hZ7lJImHIU8h5Yb2EQ4yqmRc59s6Zy\nYu/hSO6fAPed5bRPfCL6OhF1EdF68VoTET1GRFtL/49yVcZxnHOB0aj6/wzgdvPapwGsDiEsBLC6\ndOw4znnCaVX9EMKviKjTvPxOADeV5AcAPAHgU2M4LmeUNM1uVcd3/+3vRfKM+eyWSyWqVb8+oZZW\nmZ+BVGfzwl81kNfqfDEIF575JcmoOxLdDmzRldRPdh2L5JqaOtWWFCdtmsWmSav5zFkRPZc3O+uk\n769h+lTxqtbZratSIh2V0j0IAMUit2b62W05xZw/3afv3UTzehf32kIIp/YqHgTQNlJnx3HOLc56\ncS+EEEaqgktE9wK492yv4zjO2PF6J/4hImoPIRwgonYAXeU6hhDuB3A/4GWyx4rGjpZIvvNvPqza\nmjumR3Iux2poLmhVU6SfG7KqXxDqa1Hq6aTV4aS0Hkzb/i17Inn3hp2RvPSK5arf0gu57BQZBbTc\nj6Ugw+UAFEOGzxF0BKFahS8K80NbPmr4QwMDpevBjEWs6ifEzzthTQc6tzznr3c0DwO4uyTfDeCh\nsRmO4ziVYDTuvAcBPANgMRHtJaJ7AHwewK1EtBXAm0vHjuOcJ4xmVf/9ZZpuGeOxOI5TITxy7zyA\nTCjZDb/PYRXJWh2pVpQ71eQ5zK41qesVSBuu0uRPBu5oHV65HLvOtm3Yrtoy3bw7b9W7ro/kxpRO\nohFy8sBY18Lwzosdfn0FHblXLLsaoEmI84ec/jTyHEPOJl145iZId55cXgjmszS2m916E8y5teLg\nOE5F8InvODHkDavqz71wsTq+6LorI7mvj9XQHS/o/Ge7X90yvgN7HSx7y2XquKGV87dlsjnVViiy\n+p1Kiq93SMSZcG0lyqvKCWFm5IvaJDhwiDeeDBw6qdouuoErxxbF3pVeDKh+ibx0gWm3YhbstisI\n2Zo+cozW3RbE+YN8m9kRJDVzaxZRdWL4jgByBR6XsIpg9z3J3P/NnTre7ejOQ6g0/sR3nBjiE99x\nYohPfMeJIW8oG3/SVHYV/enXv6jakiJBRX+Bjc6sCf88cYR3i63990dV2zMPPhzJAyd6zm6wZ8DC\nG3S9tqRILkFBjz8n/GM1IilF0dj40rVFRZt5go+7du2P5F1bDqhueZFPYnqrTtB5dDcb2y8+8bNI\nftu9b1H9MtJmFt8LoCJsld1t3YokHHD2cw7JqhF1NLvzxPlT1eWfh5mMqQuQ4/stdyvaCF2x9IKF\n1+uwZbfxHcepCD7xHSeGvKFU/Ts/9V8iuWDUtUyBVTKt3muVr5DmXWyN05tV27s++ruR3NLCbdOm\n6ZLLqXrOw94yc4ZqS6TYzyNdQQDQdYDdYz3C5ZgRMgD0dR+P5GKNHv+OQ6w2TkoJV1lRq6gtYhdf\n1069uTI/wCbCwovZzFj5wd9S/RLCDdiT1bv/ToqkHT3Cx7Zn0y7Vb8b8OZFM5XNhaDdd0rgmc+Wj\n7tR+UGneGBOgKjVChKKoGdA3oD9nRuTqr66TW/6MX1Gcv3OVdjW/8G+/juRshRJ2+BPfcWKIT3zH\niSHntao/bYbOvXbh7bwZpDevV4hlDpCk0CkH9mo1N3GI1ej3vP+9qi0pIuHEwvqQv559eU4MUTCJ\nIWRutxrUqLbqTk4T3SdWuIM9h/gsOaNRpuWmEaHmpo9rL8SuJ9ZG8rXv0Cp8fUN9JKfE50wXdJTg\nEaH2po13RKrLTbNnRvL2X7yq+rV1cgXekXJVyE0vxZxtFSv+Vk+Xq/cJuXJvovPEYcGs+Pen+fvs\ny2hVPJ3l76l1Hv8ej+zQK/Wt89i0SkzWG6uWvZUjM1/67lOoBP7Ed5wY4hPfcWKIT3zHiSHntY1/\nxV1vU8fZArusimYX1eQku9Faanmn1JRly1S/4lL+W5izJZHFcU5Et9kcF9JEtCan2khmfE9F0bs2\nwXZg3kTnZYSRa91X8i85CcN1dpuOrFvwgXeJ9xiXZmC7tTfHck9OG9e5Mkk/AO3MUp64qsm6o3zj\nCCW0VK4Qm8dSJsq0N1XcEBmRF8yOxKJwd/YYl1qPWMvoSeskIAnx6aR3Nl+w91SMI6mft5fesSqS\nNz/6UiQPnNBu3LHEn/iOE0N84jtODDnvVH1K8N+qy377VtWW7mXVqMFUTW2fwtF1k1PajSbJDinZ\nWg4RLWbeUiuuffzYMdW2ed2GSD506IhqO5ljNTKXY72x/8Rx1W/Zyksief6lF5UZFZAU/rEeU/5q\noCDdmyhLWuSNz9iNLVIeYu6UKTGb0GWySOW6My62oDqWvZh09QUzjpSImCuKZB6ZtDZb+kVJsT7h\nvgOA3jQnDykMGLelDIgU5khhQJ9f5viz+4im1PE9uf0j747kH/z1v2C88Ce+48QQn/iOE0N84jtO\nDDnvbPyWuR2RfPLoCdWWGmBbbOkVK1VbnbC7pe1bMAa6DMXNG1tMllbOidLM659ao/rt6+LkFVnj\nNkpMZnvuZI22F0OKv47qlHDn5XWCyo5F83gcRetyZDknSkZPMi4kGbZsXZ/Slh9pzWOk3XRFZeLz\nQeu8uarfznWc3LTz4gXmAkIUySvlbrzBa4kagWYX4oD4ngrC35Y1OyN7xK5Mae8DQCIt3JZ2QUfm\n0hcvt14wU3Xr3ss7L5s6dKh5QXyemYt4t+LCVReqflt/swFjxWhKaM0moseJ6FUi2kBEHym93kRE\njxHR1tL/08ZsVI7jjCujUfXzAP40hLAMwCoAf0hEywB8GsDqEMJCAKtLx47jnAeMpnbeAQAHSnIP\nEW0EMAvAOwHcVOr2AIAnAHxqXEYpyAtXy6F1eqfXbSKSj6zLR8kiQYVRV6vEC8WC3vq2/oVXInnj\nFla7GmbphB2JJt7dVrSuoQy77KzWWCMiywYOdUfylddfp/olJ3P0mzVVpGuuTqj31mUn78GA+ZzS\nxBnJZZcQLwzNgyfKRwv/VVVK/+QOHODPOWOZvleyXDeqWM6m9XgLImLTqvDpPLvV8iLisWjz6mf4\nnEljSgTrI5Rt8kDejyqdWD+XEfn3h9bhFjKfY9WdN6puW9eI3/uQk5wZZ7S4R0SdAFYCWAOgrfRH\nAQAOAmgr8zbHcc4xRr24R0T1AL4P4KMhhJPyiRpCCEQ07J8gIroXwL1nO1DHccaOUT3xiSiFwUn/\njRDCv5VePkRE7aX2dgBdw703hHB/COHyEMLlYzFgx3HOntM+8Wnw0f41ABtDCH8vmh4GcDeAz5f+\nf2hcRmhY+Ra2d294j87RnhwhhYt080j3lS1nnBUusIf+5bv6/C1su9dN5xDgk33a3ZbOsq2aN0kX\npTuo2iQETQu7/tIrr4rkKSZhp7Tr7SeWdr1cv7DqWFrY3dZtKd9XJQ17uytOyKZUHPIoM0ZjW2fz\n/M7uPl1/ryDGGETi0IKx4zPCfi5m9f1OiEMdXmvrDLA8dJNgeXs6iPuTEr5g+46aeg4Tz5iEnYVa\nTs4qxzG1TTvKlt7Aodobf/kSzobRqPrXAvgQgFeI6NTV/gyDE/47RHQPgF0A7jqrkTiOUzFGs6r/\nawxdtD3FLWM7HMdxKsF5Ebk3fyVHML3zo/8pkoNRqGSkWi307jyZ9NImwJRkhWtrzqJFqm3dqy/z\n+Vsb+bpG9VS+M7MVKyfMgLpMtWq76u0isUgNt1mXnaTaROSVU+/TZhx5cU7r6qspU65qqArMrTaZ\nh0wQSsLMstZYPs8vnDzeq9qkN09W/K6yLtiiNN10m/yq5fmGLEWXt2j0mExrlfg8YuPokPO3dLDL\nd/+mvbqtuWnYa9ldfAtEJN/Zqvoeq+84McQnvuPEkHNS1bdRTx/83J8M2+/sYpcGyRT0po6eHB+3\nXbRQtbX3cEKM13Zui+RJ07WqJleS0yd1jrZFS5ZG8pylelNKXqiscnXXRhfKDTZWxS6KyC9pIljj\nRq7Wp8wF5NOgKM4/NDqPsd+FjKajPMvVRk9vbm6J5P7DetPVpCZe7VZVs6w6P8KKvBxkQv7aR/jx\nBP2TUPfU3gTZtZAV98pGOYqbWjS/OdmmIkfNlzb7ws5Ibp6rPT1Hdx3EmeBPfMeJIT7xHSeG+MR3\nnBhyTtr4b/rQu9TxnLmcnCAp7K2egqljJurN5U1ChlSC1w02r9sYyS+tXaf63fyB3xZH2hBcumpF\nJM8UyTBefvxp1W/GPK4Ht+BNOhmmdNFYd81o0SZ5eeeTvgN2F195C13lxJc79YZE7ok1BONHywu7\nHsJFmjTrCc0zuabcz7+xWrVd8SHenSavZe+bKttn7GIq8zmt2y+RHLbbYJvwzQ1JCCrdheJ9dk1F\n5vu3uz5zoiZhKsn1ExOmHLhcGlhwrU7S4Ta+4zinxSe+48SQc0bVT9WxiqPVbaB784FIzuzkDTF9\nc3XEXNsFrGLLjTIAUBCRcKt/9LNIblk8X/WT0YDVRi2VJaknTeNNOte95w7VTyaoGCnqLmVU56J4\nn3TtJY2OnT7JJa+nTG1QbTJHnsydZw2C7gNcxjlzXLvR5i7niEWViAMWfmVI9KKoCzBS0ohJ03j8\n81dcqtr2vbg7kmev5O/WXko+veyV9F4c4Qa1917eKzMr1PCNnSE9fXIcBXO35G9i0pRJqu3gNo7k\nm7eUf49DSqwJu+KCK5eotjXf1GbS6fAnvuPEEJ/4jhNDfOI7Tgw5Z2z8y9/BO3wz3TrM9ebFnHBj\n69EnI/mxZ3+t+rUvYLff7tf0DqjZizojubGFw0Srzd++pLDNssaek64haXfXmD+fss0muRgYwYcn\n7U55jv5Dh1W/STW8HpJM6PDmrEguKW3TnY89r/rtfOw3kXxbj84j/8p0Pv/cj74vkhtNaLK0Yu2O\nx2JW2vjidWO4yl2CkxumqLZFi9gVuvHFZyO5bZFO7yhPmTQZQeTySD4v3XIGZcebc9i+8m3S3Sm+\nwFoTmizdmFMadanwHlEfIitcewmT3iRb5O9pikjoCgCTmwfXSgbMDsdy+BPfcWKIT3zHiSHnjKq/\n8q03R/Js6J1HTZNFEo1vfDuS516jXXFSKes5ostT5xZ0RvKC5axCTmnRec2kpyhh3GgyYUWtOMj2\nadNk3ZPPRfLia02OUVES2aRvV19GvSjvfKK7R/VbegXv6uvO6Hx/WZHHP7/xaCS/97r3qH7rH10b\nycs72lXbJf38eb7y8S9F8qr7PqP6VYtyYJmcLgtdFHnwsuKD5k1Im9ytN3BMuxVra/n8F15+TSS/\n/NQvVb/25VxWrTAkx8rwO+aGRO6JtqLJo6/SDlqzrszOOhuhqKgyJ5HRnMJXWSCTP1BEEBZNbODM\nRYP3YPe6HeWvK881ql6O47yh8InvODFkwlT9NrHKDgCpDK9g3vXu21TbU9/5TiTPufjiSN56TKfy\nzwk9r8pEtOWEbjf3Ek6GYTeeSNUtYdZzhfaNE11HIvmX3/+h6lfXytdObdKq15wVvLnC5o6bJFTA\nrWvXR/KKlStUP1kdNm+yRvSv4yjHe3/rg5G8+mv3qX4rV1zG5zNqen7L9kj+j628GeknP/yV6nfB\ne9kTUxjQ5xjo5zHKDSpWBa4SJlONMB0Au2LOP9WFK1apfi//kqPWFl6/uOw5JNaMk/1MHhgkZQ7C\nIWXE5NFI6//yfPp5O2PhzOE7Gk+JTLoSjBnQOm/QPN6/ec+oxuBPfMeJIT7xHSeG+MR3nBgyYTb+\ngkt1goppJ9lWbZis881vf/aFSF7exFFbCy9dpvrtFzb+ZOOmU7vMxJ+7GptoUuVV1G39YhfbmtWP\nR/LUeTqS7PBujrTrXFKr2lLCXqwdweXTe4JdeHWT9Dl6cpyA5LUXNqm233vzeyN5wWyOtPvRttdU\nv9A+N5LrZ2l33tKPco3T33zkLyJ5znP6HCfvYFdi/0kd/SftermFrdpuSRTYKEQStnBORAKmanTk\n2/Lrb43kFx/7iWpb+CZez5HrC7Z0mjTek6bQgDTJTVCfcqolVGIS3U9ernqy/j6HLBxEbzLHI/gj\na+oHz5kYyY0oOO0Tn4hqiehZInqZiDYQ0V+VXp9HRGuIaBsRfZuIqk93Lsdxzg1Go+pnALwphHAJ\ngBUAbieiVQC+AOCLIYQFAI4BuGf8huk4zlgymtp5AcCpyP9U6V8A8CYAHyi9/gCAzwK4z76/HLNM\nTvk3X311JKfTOolGbie7KKYuZ3dedZ12/8gE5Q2N2p0nFSCZYMMmuVCnM00/++b3I7l5MUeLZfI2\nhxorhO0L5qo2XSC3fB2nBcLlmCnqzBMDOb4/s9M6qcOyBTyuosw72K2j4tDO125brpM6NC5kF17z\nCjbJLhcmFwB8/dE1kTxzuc4BJ29dqka8PuR2i8Qnk/VnOdHN1XOraritt7tb9dv/GrsfybgEX36Y\nN/csvoHHWNekNwSNlMwDI5QKk8hNQDZCUb7LugsTtobZ0MuellM5DkfIeaKvOZpORJQsVcrtAvAY\ngO0AjocQTv0i9wKYNfphOo4zkYxq4ocQCiGEFQA6AFwJYMlp3hJBRPcS0VoiWnv63o7jVIIzcueF\nEI4DeBzA1QAaiaLsZB0A9pV5z/0hhMtDCJcP1+44TuU5rY1PRK0AciGE40RUB+BWDC7sPQ7gTgDf\nAnA3gIdOezUiVJWSXjbN1DvwrhE2/uYtW1Tb5J4+liexDZc1oaZJkSUxldIfTdrrWtZ/+2R5Z2su\nzV3EtfQOinDhuuapql9KZIOor9PODp1/vrwR19TKyUL6ctpVtu5xtq0/f/fHVFtCrHPseG1nJDdm\nze65I7xzb5JZD5HUixoBR557UbftEglCLtJGbbX42DKBiXWjyaPW+XNU296NHO48YyGvCW19Qa81\n1Au3ZWODdq02zuHjHWvZHTltpk5k0bG8c/hBGWyF9WxeuvCEnNWOv3Qfu2AbW/S1R0sYwV3Ye2xw\nGa5QsA7H4RmNH78dwANElMSghvCdEMIjRPQqgG8R0ecAvAjga6O6ouM4E85oVvXXAVg5zOs7MGjv\nO45znlHRyL2aulrMvmgwZ3tdTY1qk+pgb6/OG1ZflRq2XyqVUv2k0m49JFKlTwo5DCl2JMZk1NLL\nbr0+kvdv3RnJzz7xpOpXIxO/mXPIMVpFPyf0t7TYdWfz2TUUWI9uE/kDLXv3cF76xqI2aQr7ueRS\nziTAkKQP8S5EO97WExy5V0hhBMrnupPHSZMwT9ZakD1nLdI78A52s7u3ps5ExYlb17qM3ZRp4SoE\ngFcffTmSl920VLUlJ/NvNWOzp4hx9Rzg5C95Y4Y2zeZSYTQkum505p9U77NmHH0lVb9o/Yhl8Fh9\nx4khPvEdJ4ZUVNVPVqfQOGtwlbVuhJXktja9MvtKNauAiQGOWstA65dSyakmrTZWi0QO8q9dplg+\nl3KwudeE3LGwc1gZAI4cZPU4azQvaYLkzdKsLH9VleBx2RXcZZ0LMRqqUmwSHM7q3HyzhZl05Hld\nMbjjds5/2LNNJBIxWmhb4Dt5kMrfK4ld1R+pRFdT+wzRxq3T58xW/Q7u2cn9akzlXxWSx+eY3Nyo\n+jXO4mttemKjaqudyiepa9QRfwWh0k+dMW1YeXBgIvrPPG6lUZAaQe0viNTsNk15956u0nhMfbEy\n+BPfcWKIT3zHiSE+8R0nhlTUxk9UJVHfMhhllTKJMtbv4B1WF87T+fJ7hfujb92rkTxAel8QtbO7\npiahP1ptko/TxfJ2kErwWLaXdsBkjb1V1cSRZLkhkWqyvJYt0TX8tW0a9jrxtQ2xmcVJVlxySSQ/\nY8qGdzbwSU+aiLwt/4s3WdYKm7G2pVn1ay3wusEhlEdGtOXtkorMZ2/qjSWEe48SwiVY1N/MrM4L\nIvnAQZ0sRCa90NWuTcmvDEfWzVyxSLUdeHlrJPcePqDaLryDQ1wSyk2ssTvyJOV2Bo5UljwU9fiP\nbD9gO4+IP/EdJ4b4xHecGFJRVT9VU4PppcQUNdVa9/n37Zww4ZGf6bxpmTp2PYUsb1g5eUKXyZpa\nxSpatc3fJnRKmZfexjnJ6Dkb/Scr6cp+WaN2yWtVGTeXzM1v3XSykK5ssvn9W6eyK4pGSCRSX8+5\n6XrMZqHGZjZHamp1FGV60zbRZiLhBIfzmbJtefFhpIfJmiYJtYFHn0OaRfJm2TJWzbM4L33XAb1J\nVN1jUYJqyBcvNrdkjT0yY1lnJOeP64i8Ay+xkTPjIo7Oq6ozUaUjZPqQ36C8B9YUlPd03zpt0pwp\n/sR3nBjiE99xYohPfMeJIRXPqx9KobQFY+AmmziEt+Ha6artYDvbqg+uZ7fftAs7Vb+UCNMtmHBb\n6b2R9vnA0LrK4hz6WNr80k0XjA0+SXRMmZ1YyuQ0Lp60DMkcwRw93MVJQEZy58m25puuVv1q9vPu\nNOt6erlrfyR3i+ygm5r1OkFmGYfONpubJTenkbStzZpEQYzR1jGQty4pCg3aZY2cuPbMefNU22vb\nN0RybYNIgDEk771wOdq1F5HTP2fKkl987Y2R/PzPH43kedfopCIF4c8rDlmXEb8lVTJb98qLcbz0\n0NM4G/yJ7zgxxCe+48SQiqr6+UwWR7fsAgA8d1wn21h+x3WRXJPUw5ohcrGlTJ56SU7oxNUmqYP+\nC8cdrXop1S6bc1+qntXifdXG7zeChw0kzj+Sq09GA5LRS3+4jXPObXtQq54LFrNLMyvqExycryMl\nv9zAbtFkk84Z2DT30kguiM/ZmNW5/4ppVj1PHutRbTWi5oFUX4P5LPIODMlZnxhevS+OkGtiaquO\nLqzbzeq9Orv9jsQFRkj9j4IxDYvCt3jJjbyrce2jP1L9ltwmS8aNzjyz0YVrv8tlyk/sP4qzwZ/4\njhNDfOI7TgyhIVVDx5FkVTLU1g+qXnd97lOqbcfzr0Ty/FWXqLY5y1h9bZzCK/x2I45U0YppHWHV\nPJXf15XmFe3C6/z8erHeqq+jq31kI/f6yngY7Nnk+6znQanO4s+6jf5TXgOjUmYyrNLnennzSrKg\nNzet/8UzkTxzqa6xUlPHJa/kJp3ESCvapqlcag/bb6SIvF6RT/DQ7l2RnB7oU/2KYhxk7mlSnHRy\nnU7EMXcF/1blsNI9+vyb1rGa3nnZBapNXm7gRH8kb/31BtVvw0+ei+T8gDa71PlsBplh8Ce+48QQ\nn/iOE0N84jtODKmojU8ihGvOxbqs8ts//geRfHjnXtXWc+J4JM9czJFZM0zJpcmiRHLupM4VP7uD\nk3b05NgFNtKn3yNy5wM6z3tbBycEtfbzSOeUNnna2PSj/SakfT4kRTukW4rlbF7b54UMr4EU+7W9\nmBU2/msiSUc2rV2HdQ28w6/tAh0xJ912IyY0Gcn1WWa3oj2hXDawP2cqk+RyyL2WrmBTI0CuS5jb\niFxWrFEkyn/mnS9yQtP+gk6a0TCD7+OBjVwj4Pnv6noNuf7yuyElY2rjl0plv0hEj5SO5xHRGiLa\nRkTfJqLq053DcZxzgzNR9T8CQOYd/gKAL4YQFgA4BuCesRyY4zjjx6hUfSLqAPAAgL8F8DEAbwdw\nGMCMEEKeiK4G8NkQwltOc56yF2tfyHn2bvuDD6u26bNbI7lrF5sBx9drlalmBrvs2i/RefumzeSN\nPymhy1WldIRfUmVMKK8xJUY2jQJpAAAWY0lEQVRQL6GSfuimAZHkwebVh1LhpZ5rEjKI9wVbMknu\nMMnxtbp36QQV2W5W21/5td7w0XQBb76Zv4JNsmpTnkptIrFfrdA2ZU4Ua8Y1zWYTzHr6lFtU3HCb\ng1DVQrD3dHSeVWWaVJnE9wVRripv/ady85D8KmxUZhW/b/UDD6q2w7u51Nm+V3aObsAjMJaq/pcA\nfBJsCTUDOB5COGXx7AUwa7g3Oo5z7nHaiU9EbwPQFUJ4/vVcgIjuJaK1RLT29bzfcZyxZzSbdK4F\n8A4ieiuAWgANAL4MoJGIqkpP/Q4A+4Z7cwjhfgD3AyOr+o7jVI4zcucR0U0APh5CeBsRfRfA90MI\n3yKifwSwLoTwf0/z/lFdbJKpq3fb770/khddxXnMD+/Qf2v2/5JdJovbtI3/9KY1fI6VF0dyf73q\nhilt7FqZ1taq2nJZdoElkqwspfu0m0vWL8uZWmZ5YeNn+tKqrf8wJw9tqOaQ16mTdZhofx+Hddb2\nm+SVaTY0q/tZ7jUhqgtvviGSn962RrUtuP6ySJbWojUchfdqSOlnWcH82H5OSNnddVz161jGJa+r\njPFOInlFdY0ciHGfCtek/T3LaGQZOmyXAuSxXSdQefxtAtZJ7MyiGn6OFs13G8TW0cO79DrHg3/2\nd6Lj2T8bxztk91MAPkZE2zBo83/tLM7lOE4FOaP9+CGEJwA8UZJ3ALhy7IfkOM54M2GRe6+XJddd\nHslvufeDqi0l8sNv+cWzqm3mEVbJls5m9TJhk7SLLA8Hj+rCUEmhIEn3T3WVDvWqFuWpkyapSONk\nacZojSwndr8VIHP/v74Iv+6T3ZG8saA/S+N1yyK5pqZOtclbosoT2F1rop/Je6LMon2bOE/i3JUX\nqX6Far4/tSa/f+MUHpe8H+m0iTTsZ7U6mF2ZWRHtlhT2R85k88gLdd6q+lV1PK5JLdo2TNXwd93T\nyyZYoVsnmimKeuk2GckvvvqvkbzxibPLpQf47jzHccrgE99xYsh5p+pLJpvV/6vf+7ZIvuTW61Xb\nQA+rXiee5/JDFwzofHNXLOIV/+pqraYnZfXWcqVtgVHr4vbeyxV/qSqnM1q17ROVXY/26FXyNbte\n5vFewfkJmxboDU0JoQ1aNb06NcpwN9Ht4PZdqqlmMnslWhZwJOCkhsmqX6qaVWWbg7BK5twTr+eK\nOu/0gLg//UdPqrZD23jTS3NHeyTbCLzuLjaLqqdOUm0dS3gDUnKEdOndxznvYKFbe1FkhKXdWNV7\nnDeUfe0PPo2zxVV9x3GGxSe+48QQn/iOE0MqXkJrLOk7ru25n3/1m5E8qUXnkZ8yjW35GbdwgsT9\nwrYDgH9ay2WQliRaVNsV85dHcoOJplOM1kS2O7jEmkK12EF4pE/nUF+ze30kH27Rf7tbfvsqPodw\nQ9n1BBEUp6Lshg6SRWvHH9zO9vMFq5artjaxpjBF2Ps5Y34WVCJOfWk1ZNGWMiXQc1V838hE/x07\ndDiSm2bNiOTjXfqeUi3fhDnLdNSnjP47uE3fg53reKf6opu5TJmthSDXUWz59YZmLnveeRmvMe18\nfh3GC3/iO04M8YnvODHkvFb1hyBU51V33qGa9mzcGsmb1nAeucY2XZl3zh3XRPLJrN5g88B63plc\nvY3dNXPr9WaevkNsPtSSvsUyf3vVFO02Olrgc/bXiw0qnXqMTReviOR2mxhCRqAJXdm66KqEvpnP\n6Wi3Xes3R3L3Qd441NypXYKLb2azYub8maqtTrjpZBXgvK3uW0YGdISbzGtovaU5mQjPRhem+P7L\nCso9vTqy7uJVrKbnM/p+vPqr30Ry/0ltXqbE+Q9s4AjF6e3tqp+06mz5hLxInrL8lmsj2VV9x3HG\nFJ/4jhNDfOI7Tgx5Y9n4wn7Mm91X7Yu5XtnMJSx379UJO/ds3Mbn6NXJFOTurtab2SWYNu6lVJJD\nZYsmcWNerBvYPJnCrEe9aEsYP5eqj0fWPcZvPLaHkzgWzY6zvj4R5tqjE4nMvoh3LyYns42f7tOl\nsGfOYztW2vQAkBPXk+sax/fpXYL1Iuy6xqx5hHKyWSeQ4c1Fs+aREOPavWlHJC+7Re8ol67hjU/p\nnZ1Tmth1m6jSYeIFkdA0fYLXDQqtehzyHtgoeVmOvWlWZVJX+hPfcWKIT3zHiSFvLFVf0H2gSx1P\nm8EuN6ket87WbpemWVwaq+eoLsM1cJR3wu3bzK4b64aS5akSJklH7zFWKfNpbUoM9HMih6nT2YVX\nzOvdeVU1nN/+pIlerJvKEYrThfstmdCq+JQprG9Obi8fanjiEN/HmRfMUG2Tajgy0Oawk6x56LFI\nTtXp+1HXwKrzRVdfptpUnQHxsk2iURDmkx3G4d37I3nqXP6ue03k3tYXXorkaTOa9fmFOh+MKZHP\ncqKPfL9U503ePiEP+b2Ivs0zW1AJ/InvODHEJ77jxJA3rKrfc/CwOm6ZKaPfWLXKmjTIW55+LpJP\nHNNJLpbfyFFVC68Qm1JMYoj+k6yaHzuu1fS6NpF7zeT7kw6A6hG+mZyI9GrQl4bMdaI04qF1vsR7\ndItSU8WYLrxOq+KFEdJVHxWpwvMiX97UaXpVPH2cV8KHVLpVB+I7syVrRQITYwUgLVKRX30t52vc\n/swLqt+0mWwKBhtaJwYm8yICQH+Wv99EQdysUW7UAvSGKRLXmmrSu584pH/TZ4M/8R0nhvjEd5wY\n4hPfcWLIG9bG79q+Wx3Pv5TzuW9//pVIPrprh+rXvpBdYNmEts+PHWIXUEOzSPRBOnIvJPi2Zk/s\nV22ZvIjSatc2XHWV3IHGtl46bfLqCxufjDEp8ywmpBPJlp2SuULNn/+8sHGbO9i1ZdcCjh3mXYg7\nN+r72NrZEcktM9kN2N+t3Wj9R9gdaRNxJNUF+bNkTPJRGZWYt1FxInKvXkQGLr3pKtXv2Yd+xmOf\nZUqnCXdqX7pftfUX2XWbKo7yOWo+Z1F8GTKnPyXG77k8qolPRDsB9AAoAMiHEC4noiYA3wbQCWAn\ngLtCCMfKncNxnHOHM/mTcnMIYUUI4dTS6KcBrA4hLASwunTsOM55wNmo+u8EcFNJfgCDNfU+dZbj\nGTO2r3lJHbfNF1FsvaxezlrUqfrlRJQWmdJSx7rZvTdpGudJk8kvACBVy7e1ZW6HapOJHE7s0+bI\ntBnscqydwvnni+bvc0Hqs2anT165osrnzpentN6rA5s4acnyG9lt+cJPnlT9aoQaPblZu+l2vcrn\nuPI2rnFwbI82fZo72Aywqr781JkCfy95UyZL3oKCiYacMZ83vVQL04HMpqJZ8zsj+egB/b3kxLgG\ncqbGQZ6vV5uX06l8sQXrtiwoi4zf12fcyWPJaJ/4AcCjRPQ8Ed1beq0thHBqa9tBAG3Dv9VxnHON\n0T7xrwsh7COi6QAeI6JNsjGEEMpVySn9obh3uDbHcSaGUT3xQwj7Sv93AfgBBstjHyKidgAo/d9V\n5r33hxAuF2sDjuNMMKd94hPRZACJEEJPSb4NwF8DeBjA3QA+X/r/ofEc6Jmy+em16njhSk4ucfG1\nl0Zyd49OupgRiScLJg413aN3wkVYZUeYd9WT9G40SnJSh6lNOjf/gS3sEtu+ltco5l66QvWrqWf7\nv6raJvoQMbwiKYd1xcmwX5uk4+Rh/ht+WCSvmLVgnuonXU8HD+uEJm1t7AZMCeO9da5OyjlSmUHl\n0hShsYWcDptNiVuw8ZnnVdvFb70hkrt7OZmpDTFuXdgZyccOHVFth/Zz/YBirUmeKn4v1CvuN8oz\n9DML+1+8M5fODOk5VoxG1W8D8IPSokMVgG+GEH5KRM8B+A4R3QNgF4C7xm2UjuOMKaed+CGEHQAu\nGeb1owBuGY9BOY4zvrxhI/cKWe3y6T3J+eJk2vSc2d0mSRvXULFfunKEqyxhcsUnZASecfVVcfIK\nO8aZS7h0U+1Udo898Ec6RGLOSnaxda64SLW1zmZVukHkds/l9ThkmeiEiRCrn8aJPtrmcolrqx9n\nshzFdnDra6pt5YffH8nlcuedjqxw4eUGWO01VbKQETssd63brNoWrboikosorzrLHYnt8xaptpoa\njvjbvG69auvo4Psz58olfL6yVxqKjL7s6x4/F57EY/UdJ4b4xHecGOIT33FiyBvWxre8vJrrny2+\nnndmyTz0AJQf5tAresfZjR94TyRXiffZ3PbSFLauMhmSmarVYaOZNLupnv3BT+1HiNj94ivDygBw\n+dtv42uJWm7VU3RIbS7DdnHfUV0qfNHV7PqUdQBCUq9JZNLsCp3aqMuSJ0SMsMwpb3cTjmQND4hd\neCFbvj7eiz/9VSS3X6jXPHI54SojEY5thiGdrkUTOzxtJoddXzVLh2CjzGez8WxBbYfUZ5CXO7pn\nDyqBP/EdJ4b4xHecGBIbVf/Izn2RvO2ljZHc2qFzxadFdNeqd92u2lKitDTJHWHGJZhXx8YMUBqf\nVgd3vsLj2vTLZ/B62PwbTiL5zr/8ZCTngt6el0jytbc++ZRqq5vCEYXHTxwb9j2AdmM2NDeptnIK\n/EgRbf1m51uuj0t7BZFF89gBnXQykxVmRVZHYkImshTfmR2fPLYJO8Pw21AAAFVCT5deUZlcAwAK\nwn1qd+dJl++ul14te62xxJ/4jhNDfOI7TgyJjaovOb6L1f7b7n6XapNVdk8c0ZtySKzky6A7qxpK\nFd7ms5Oru0Wj8/38Hx4YeeCjoOcwbzD56Zfui+Qbf//Dql9tA0ejFXO6Wm4xzx9OlX4yJs3BnXsj\n+Zo7367aSMnlc/j3ixs50KPz2QWZFEW88edfeVD12/bcujJXBqY1s9ky/9KLI9k+8dS2HyrviUma\n8VeJGSRNvoJJ/ievVzBmgLTCtq99GZXAn/iOE0N84jtODPGJ7zgxJJY2/vOPPB7Jq973W6qtYQaX\nKSZjg2czbMxLuz4MCUYbPh+8PX76Wzp3Se/Rsc1OfmQH75jb/eNfq7ZiK+/AazEuzeOi7mDtJE76\n0dOlE1TMXrogkhubGqGRdfVY7s1ql126V6wvZHSCDRkRue4XHHmpbXqLvt8vPMIluhdczjb+kLLe\nVH4dQrrpbKp7mag0L+z6pDmJjM4rmlLbe9dvieSx/g2Uw5/4jhNDfOI7TgyJpaqfFPpa19M6cQPN\n58i9o/3avTR9CeecC0nW8U52afVsSgtHsZkcHTi4g3O2P/vdH53BqM+O5375aNm2e77yt+p45rKF\nkdy/j3PpzRGJQgYpH4cnS2j3ZjgBhlLtAZDcBGTu1YEd7C788Ze/XvZaI3FwK29U2vQUl0Bfcv0V\npmf5cmNqw411xRlX7imsq1aaFtmMTgiy+qv/MvxJxhF/4jtODPGJ7zgxxCe+48SQWNr4HR1zI7l/\nX49qu/t3OUv4uo27VNtPtrCdPH8au8CmD2hX1h5hxx7Zs1e1PfR3/4cPrFE7QVBtrTqeVs81A5uF\nXT+kJLeQM2aL4ok+Xh/Ji2SYZAr1Sdv34A59r77z5/8jkm0u/dfDo/f9cyRfv0Db+JszHMadqNW1\nENoyHN58uFr/XppmcXLTA+s5cUu1qSU4dTq7iZ/81++ptp7DunR4JfAnvuPEEJ/4jhNDYqnq79jO\nkVKFvHat/Ne//PNI/s1TT6i2Qg3/nXz7rRzx113fqvp99b7/GcnF/NmrqOPB4iWcA/7CdItq6z12\nIpKP7ToYyW35etVvVx2r8/XT9T04uH67aON8fLX1+hzbn+VSYT/+4ldVW25A1zU4a4RptWXDBtX0\n+C9+EslTmnX+wK/8I6vmf/M3H1dtG8ARhX/3ib+P5H/9pt5pufoHP47kTU/+BhPNqJ74RNRIRN8j\nok1EtJGIriaiJiJ6jIi2lv6fdvozOY5zLjBaVf/LAH4aQliCwXJaGwF8GsDqEMJCAKtLx47jnAdQ\nOM3KMhFNBfASgPlBdCaizQBuCiEcKJXJfiKEsLjceUrvOTeWsR3MnstRiJ2mZNSWPZsi+S8+wX/P\nb3zz76h+//u+/x7Jh4JOWvLfPvxnkfyJz/5JJHed1PnyNjz25JkMe0KQiThGmi81wjuSSY+xmXIG\nhDBk29gQRvPEnwfgMIB/IqIXiej/lcplt4UQTsVzHsRgVV3Hcc4DRjPxqwBcCuC+EMJKAH0wan1J\nExj2TyER3UtEa4lo7XDtjuNUntFM/L0A9oYQ1pSOv4fBPwSHSio+Sv93DffmEML9IYTLQwiXj8WA\nHcc5e05r4wMAET0J4HdDCJuJ6LMATmVnOBpC+DwRfRpAUwjhk2VPArfxz0fmzb8gkpdfeo1qW/2T\nH0RyX5/OZ3/VNTdF8pqnnxiPoTllGI2NP1o//h8D+AYRVQPYAeDDGNQWvkNE9wDYBeCuEd7vOM45\nxKgmfgjhJQDDqeq3jO1wHMepBKNS9cfsYq7qO864M1buPMdx3mD4xHecGOIT33FiiE98x4khPvEd\nJ4b4xHecGFLpRBxHMBjs01KSJ5JzYQyAj8Pi49Cc6Tjmnr5Lhf340UWJ1k507P65MAYfh49josbh\nqr7jxBCf+I4TQyZq4t8/QdeVnAtjAHwcFh+HZlzGMSE2vuM4E4ur+o4TQyo68YnodiLaTETbSsk7\nKnXdrxNRFxGtF69VPD04Ec0moseJ6FUi2kBEH5mIsRBRLRE9S0Qvl8bxV6XX5xHRmtL38+1S/oVx\nh4iSpXyOj0zUOIhoJxG9QkQvnUoTN0G/kYqksq/YxCeiJIB/AHAHgGUA3k9Eyyp0+X8GcLt5bSLS\ng+cB/GkIYRmAVQD+sHQPKj2WDIA3hRAuAbACwO1EtArAFwB8MYSwAMAxAPeM8zhO8REMpmw/xUSN\n4+YQwgrhPpuI30hlUtmHECryD8DVAH4mjj8D4DMVvH4ngPXieDOA9pLcDmBzpcYixvAQgFsnciwA\nJgF4AcBVGAwUqRru+xrH63eUfsxvAvAIAJqgcewE0GJeq+j3AmAqgNdQWnsbz3FUUtWfBWCPON5b\nem2imND04ETUCWAlgDUTMZaSev0SBpOkPgZgO4DjIYRTNb8q9f18CcAnAZwqo9s8QeMIAB4loueJ\n6N7Sa5X+XiqWyt4X9zByevDxgIjqAXwfwEdD0JUoKjWWEEIhhLACg0/cKwEsOc1bxhwiehuArhDC\n85W+9jBcF0K4FIOm6B8S0Q2ysULfy1mlsj8TKjnx9wGYLY47Sq9NFKNKDz7WEFEKg5P+GyGEf5vI\nsQBACOE4gMcxqFI3EtGp/RuV+H6uBfAOItoJ4FsYVPe/PAHjQAhhX+n/LgA/wOAfw0p/L2eVyv5M\nqOTEfw7AwtKKbTWA9wF4uILXtzwM4O6SfDcG7e1xhQZrMX0NwMYQwt+LpoqOhYhaiaixJNdhcJ1h\nIwb/ANxZqXGEED4TQugIIXRi8PfwixDCBys9DiKaTERTTskAbgOwHhX+XkIIBwHsIaJTpehuAfDq\nuIxjvBdNzCLFWwFswaA9+ecVvO6DAA4AyGHwr+o9GLQlVwPYCuDnGKwLMN7juA6Dato6DNYjfKl0\nTyo6FgAXA3ixNI71AP5r6fX5AJ4FsA3AdwHUVPA7ugnAIxMxjtL1Xi7923DqtzlBv5EVANaWvpt/\nBzBtPMbhkXuOE0N8cc9xYohPfMeJIT7xHSeG+MR3nBjiE99xYohPfMeJIT7xHSeG+MR3nBjy/wEd\nainul2FA3AAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "batch = next(iter(dl))\n",
    "plt.imshow(batch[0].permute(1, 2, 0)*0.5 + 0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [],
   "source": [
    "discriminator = Discriminator()\n",
    "generator = Generator()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [],
   "source": [
    "dis_optimizer = torch.optim.Adam(discriminator.parameters(), lr=0.0001, betas=(0, 0.9))\n",
    "gen_optimizer = torch.optim.Adam(generator.parameters(), lr=0.0001, betas=(0, 0.9))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███       | 16/52 [02:07<04:46,  7.95s/it]"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-141-04716ccea07c>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     23\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     24\u001b[0m         \u001b[0mloss\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mdiscriminator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimages\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmean\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mdiscriminator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfake_images\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmean\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;36m10\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mgrad_penalty\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 25\u001b[0;31m         \u001b[0mloss\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     26\u001b[0m         \u001b[0;31m# with torch.no_grad():\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     27\u001b[0m         \u001b[0;31m#     for param in discriminator.parameters():\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/.asdf/installs/python/3.6.3/lib/python3.6/site-packages/torch/tensor.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, gradient, retain_graph, create_graph)\u001b[0m\n\u001b[1;32m     91\u001b[0m                 \u001b[0mproducts\u001b[0m\u001b[0;34m.\u001b[0m \u001b[0mDefaults\u001b[0m \u001b[0mto\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     92\u001b[0m         \"\"\"\n\u001b[0;32m---> 93\u001b[0;31m         \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     94\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     95\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mregister_hook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/.asdf/installs/python/3.6.3/lib/python3.6/site-packages/torch/autograd/__init__.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables)\u001b[0m\n\u001b[1;32m     88\u001b[0m     Variable._execution_engine.run_backward(\n\u001b[1;32m     89\u001b[0m         \u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_tensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 90\u001b[0;31m         allow_unreachable=True)  # allow_unreachable flag\n\u001b[0m\u001b[1;32m     91\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     92\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "for images in tqdm(iter(dl), total=num_batches):\n",
    "# for i in range(20):\n",
    "    # images = torch.zeros(batch_size, 3, 64, 64)\n",
    "    noise = torch.randn(len(images), 100, 1, 1)\n",
    "\n",
    "    set_trainable(discriminator, True)\n",
    "    set_trainable(generator, False)\n",
    "\n",
    "    for _ in range(5):\n",
    "        gen_optimizer.zero_grad()\n",
    "        dis_optimizer.zero_grad()\n",
    "        \n",
    "        images = torch.Tensor(images)\n",
    "        alpha = torch.randn(batch_size, 1)\n",
    "        fake_images = generator(noise)\n",
    "        interpolates = alpha.view(len(images), 1, 1, 1)*images + (1 - alpha.view(len(images), 1, 1, 1))*fake_images\n",
    "        \n",
    "        discriminator(interpolates).mean().backward(retain_graph=True)\n",
    "        grad_penalty = ((torch.cat(list(p.grad.pow(2).view(-1) for p in discriminator.parameters())).mean()).sqrt() - 1).pow(2)\n",
    "        \n",
    "        generator.zero_grad()\n",
    "        discriminator.zero_grad()\n",
    "        \n",
    "        loss = (discriminator(images).mean() - discriminator(fake_images).mean()) + 10 * grad_penalty\n",
    "        loss.backward()\n",
    "        # with torch.no_grad():\n",
    "        #     for param in discriminator.parameters():\n",
    "        #         param.grad.mul_(1)\n",
    "        \n",
    "        dis_optimizer.step()\n",
    "\n",
    "        # for params in discriminator.parameters():\n",
    "        #     params.data.clamp_(-0.01, 0.01)\n",
    "\n",
    "    set_trainable(discriminator, False)\n",
    "    set_trainable(generator, True)\n",
    "\n",
    "    gen_optimizer.zero_grad()\n",
    "    dis_optimizer.zero_grad()\n",
    "\n",
    "    loss = - discriminator(generator(noise)).mean(0)\n",
    "    loss.backward()\n",
    "    gen_optimizer.step()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/alvaro/.asdf/installs/python/3.6.3/lib/python3.6/site-packages/matplotlib/figure.py:457: UserWarning: matplotlib is currently using a non-GUI backend, so cannot show the figure\n",
      "  \"matplotlib is currently using a non-GUI backend, \"\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAARsAAAEYCAYAAABsuVKPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ndtWyTFG4AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "with torch.no_grad():\n",
    "    fake_img = generator(noise).detach()\n",
    "\n",
    "fig, plots = plt.subplots(1, 1)\n",
    "for i, plot in enumerate(np.array(plots).reshape(-1)):\n",
    "    plot.imshow(fake_img[i].data.numpy().transpose(1, 2, 0)*0.5 + 0.5)\n",
    "    \n",
    "#fig.set_figwidth(16)\n",
    "#fig.set_figheight(5)\n",
    "fig.tight_layout()\n",
    "fig.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/alvaro/.asdf/installs/python/3.6.3/lib/python3.6/site-packages/torch/serialization.py:241: UserWarning: Couldn't retrieve source code for container of type Generator. It won't be checked for correctness upon loading.\n",
      "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
      "/Users/alvaro/.asdf/installs/python/3.6.3/lib/python3.6/site-packages/torch/serialization.py:241: UserWarning: Couldn't retrieve source code for container of type Discriminator. It won't be checked for correctness upon loading.\n",
      "  \"type \" + obj.__name__ + \". It won't be checked \"\n"
     ]
    }
   ],
   "source": [
    "torch.save(generator, \"generator.pkl\")\n",
    "torch.save(discriminator, \"discriminator.pkl\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}