train.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "bd1298d1-7db6-492c-9732-5ea46c607898",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": ""
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "import time\n",
    "from pandas.core.common import flatten\n",
    "from IPython.display import display_markdown\n",
    "import torch\n",
    "import torch.onnx\n",
    "import torchvision\n",
    "\n",
    "from torch import optim, nn\n",
    "from torchvision import datasets, transforms\n",
    "from torch.optim.lr_scheduler import StepLR\n",
    "from torch.utils.tensorboard import SummaryWriter\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "from tqdm.notebook import tqdm_notebook\n",
    "from functools import reduce\n",
    "\n",
    "from datetime import timedelta\n",
    "\n",
    "from networks import CNN\n",
    "\n",
    "# Setup\n",
    "use_cuda = torch.cuda.is_available()\n",
    "device = torch.device(\"cuda\" if use_cuda else \"cpu\")\n",
    "loss_fn = torch.nn.CrossEntropyLoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "e54f14c0-9aa0-424c-a9ab-152486e84000",
   "metadata": {
    "editable": false,
    "slideshow": {
     "slide_type": ""
    },
    "tags": [
     "parameters"
    ]
   },
   "outputs": [],
   "source": [
    "# Parameters\n",
    "input_size = 28 * 28\n",
    "hidden_layer_size = 512\n",
    "num_classes = 10\n",
    "num_epochs = 15\n",
    "batch_size = 64\n",
    "test_batch_size = 1000\n",
    "training_size = 60000\n",
    "test_size = 10000\n",
    "mean_val = 0.1307\n",
    "standard_deviation_val = 0.3081\n",
    "learning_rate = 1\n",
    "gamma = 0.7\n",
    "step_size=1\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "97c99c08-030f-415e-a941-3fef56cd688c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dataloaders\n",
    "\n",
    "toTensor = transforms.ToTensor()\n",
    "normalize = transforms.Normalize((mean_val,), (standard_deviation_val,))\n",
    "randomAffine = transforms.RandomAffine(degrees=30, translate=(0.5, 0.5), scale=(0.25, 1), shear=(-30, 30, -30, 30))\n",
    "\n",
    "\n",
    "# load datasets\n",
    "mnist_training = datasets.MNIST(\n",
    "    root='/tmp/mnist', \n",
    "    train=True, \n",
    "    download=True, \n",
    "    transform=transforms.Compose([\n",
    "        randomAffine,\n",
    "        toTensor,\n",
    "        normalize\n",
    "    ])\n",
    ")\n",
    "mnist_val = datasets.MNIST(\n",
    "    root='/tmp/mnist',\n",
    "    train=False,\n",
    "    download=True,\n",
    "    transform=transforms.Compose([\n",
    "        toTensor,\n",
    "        normalize\n",
    "    ])\n",
    ")\n",
    "\n",
    "loader_opts = { 'num_workers': 1, 'pin_memory': True } if use_cuda else {}\n",
    "\n",
    "train_loader = torch.utils.data.DataLoader(mnist_training, batch_size=batch_size, shuffle=True, **loader_opts)\n",
    "test_loader = torch.utils.data.DataLoader(mnist_val, batch_size=test_batch_size, **loader_opts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "146fb526-f881-4c7a-8d48-4a7a280c22ec",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "## Training Data"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA9EAAAFKCAYAAAAaDtzaAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy81sbWrAAAACXBIWXMAAA9hAAAPYQGoP6dpAABJ8ElEQVR4nO3dfXwU1aH/8e9CwoohiWAgISCQBjRVfEAELaKg1ihKVazKT3tbn4s8KcWWC0ULooVUbctVQAQpUKtFvVLFFoUoGPSiFW2piAVFCU2rIUIlCUIISc7vD5pxJ09Mkt3sztnP+/U6r9eZObOzJ/PNbHJ2ngLGGCMAAAAAAHBU7aLdAQAAAAAA/IJBNAAAAAAAHjGIBgAAAADAIwbRAAAAAAB4xCAaAAAAAACPGEQDAAAAAOARg2gAAAAAADxiEA0AAAAAgEcMogEAAAAA8MjqQfSyZcsUCAT07rvvhmV9gUBAEyZMCMu6Qtc5c+bMFr22sLBQgUCgwbJixYqw9jOW2J6rJB0+fFj33Xef+vTpo2AwqJycHD366KPh62CMiodsQ7366qvOPrtnz56wrDNWxUO299xzj0aOHKkePXooEAjopptuClvfYlU85PrRRx/pu9/9rjp37qxjjz1WZ599tlatWhW+DsYo27N97733NH78eJ166qlKTk5Wenq6vv3tb2vdunVh7WMssj1bKT4/jyX7sy0qKtKoUaP0jW98Q0lJSUpNTdWAAQM0b948VVVVhbWfR2P1IDpeTJw4UW+99ZarXHzxxdHuFlph3LhxmjNnjsaPH681a9Zo1KhRuuuuuzR79uxodw1hsn//ft1+++3KzMyMdlcQJr/+9a+1d+9eXXHFFerQoUO0u4MwKCws1Le+9S1t375dCxcu1HPPPaeuXbvqqquu0vPPPx/t7qEVfv/73+udd97RLbfcohdffFFPPPGEgsGgLrroIv32t7+NdvfQSnwe2+mrr75SSkqK7r33Xq1atUorVqzQ0KFDNXHiRN1xxx1t2peENn03RESvXr10zjnnRLsbCJOtW7dqyZIl+vnPf66f/OQnkqThw4dr7969euCBB3THHXeoS5cuUe4lWmvq1Knq3LmzLr/8cj3wwAPR7g7CoLy8XO3aHflu+sknn4xybxAOeXl5OnDggNasWaMePXpIki699FKdeuqp+tGPfqRRo0Y5mcNfpkyZoocfftg177LLLtOZZ56pWbNm6Qc/+EGUeoZw4PPYTjk5OVq+fLlr3ogRI1RSUqLly5dr/vz5CgaDbdKXuP/kr6io0N13360zzjhDqamp6tKli771rW/pxRdfbPQ1jz/+uE488UQFg0GdfPLJDZ46XVxcrDFjxqhnz57q0KGDsrKydN9997X5qQbxys+5vvDCCzLG6Oabb3bNv/nmm3Xw4EG98sorYXsvP/JztrXeeOMNLVq0SE888YTat28f9vX7ld+zZTDVMD/n+n//9386/fTTnQG0JLVv314jRoxQUVGR3nnnnbC9lx/5Odtu3brVm9e+fXsNHDhQRUVFYXsfv/JzthKfx03xe7YN6dq1q9q1a9em/1PF/ZHoQ4cO6d///rd+/OMfq0ePHqqsrNSrr76qq6++WkuXLq33TeSqVau0fv16zZo1S0lJSVqwYIGuv/56JSQk6JprrpF05Jdo8ODBateunX72s58pOztbb731lh544AEVFhZq6dKlTfapT58+ko6cRuZFXl6efvrTnyohIUFnnnmmpkyZoiuuuKLZ28Imfs71gw8+UNeuXZWRkeGaf9pppznt8czP2UrSwYMHdeutt2rSpEk688wz4+LaSq/8ni0a5udcKysrGzzzp/ZIx/vvvx/XZ4L5OduGVFVV6Y033tApp5zS7NfaxrZs8TUbsjXGqLq6WuXl5Vq7dq2WLVumu+++WwkJbTi0NRZbunSpkWQ2bdrk+TVVVVXm8OHD5tZbbzUDBgxwtUkyHTt2NMXFxa7lc3JyTN++fZ15Y8aMMZ06dTK7du1yvf7hhx82kszWrVtd65wxY4ZruezsbJOdnX3Uvn722Wfm9ttvN88++6x54403zFNPPWXOOeccI8ksXrzY88/sN7bnevHFF5uTTjqpwbYOHTqYH/7wh0ddh1/Znq0xxtx9993mG9/4hjlw4IAxxpgZM2YYSeaLL77w9Hq/iodsQyUlJZkbb7yx2a/zG9tzveqqq8xxxx1nysvLXfPPO+88I8nMnj37qOvwK9uzbcj06dONJPPCCy+06PV+EW/ZxsvnsTHxk+2cOXOMJCPJBAIBM336dM+vDRfOdZD03HPP6dxzz1WnTp2UkJCgxMRELVmyRH//+9/rLXvRRRcpPT3dmW7fvr1Gjx6tHTt26J///Kck6Y9//KMuuOACZWZmqqqqyikjRoyQJBUUFDTZnx07dmjHjh1H7Xf37t21aNEiXXvttRo6dKhuuOEGbdiwQQMGDNDUqVPj/tRxv+YqHblzYUva4oVfs33nnXc0d+5cPf744+rYsWNzfuS44dds0TS/5jphwgSVlpbqBz/4gT799FPt3r1b9957rzZu3CiJU0Yl/2Zb1xNPPKGf//znuvvuu3XllVc2+/U2siVb1Of3bG+66SZt2rRJa9as0ZQpU/TQQw9p4sSJnl8fDnH/6b9y5Updd9116tGjh373u9/prbfe0qZNm3TLLbeooqKi3vJ1T7ENnbd3715J0u7du/XSSy8pMTHRVWpPD4rko2wSExM1evRo7d27Vx9//HHE3ifW+TnX448/3nnPUF999VWjpxbGEz9ne8stt+jqq6/WWWedpX379mnfvn1On8vKylReXh6W9/ErP2eLxvk514suukhLly7Vhg0blJ2drYyMDK1cuVL333+/JLmulY5Hfs421NKlSzVmzBj98Ic/1EMPPRT29fuRLdmiPhuyzcjI0FlnnaXc3Fzl5eVp1qxZmjdvnv7617+G9X2aEvfXRP/ud79TVlaWnnnmGdcRvkOHDjW4fHFxcaPzjj/+eElSWlqaTjvtNP385z9vcB2RfqSNMUZSfH9D7udcTz31VK1YsULFxcWuD64tW7ZIkvr37x+W9/ErP2e7detWbd26Vc8991y9tuzsbJ1++unavHlzWN7Lj/ycLRrn91xvvPFGfe9739PHH3+sxMRE9e3bV3PmzFEgENB5550XtvfxI79nKx0ZQN9222268cYbtXDhQs72+g8bskXDbMx28ODBkqSPPvpIAwYMiOh71Yr7QXQgEFCHDh1cv0TFxcWN3qHutdde0+7du53TGqqrq/XMM88oOztbPXv2lCSNHDlSq1evVnZ2tjp37hz5HyLE4cOH9cwzzygtLU19+/Zt0/eOJX7O9corr9Q999yj5cuX67//+7+d+cuWLVPHjh116aWXRuy9/cDP2a5fv77evGXLlmn58uV64YUX4v6olp+zReNsyDUhIUHf/OY3JUmlpaVatGiRrrzySvXu3Tvi7x3L/J7tsmXLdNttt+m//uu/9MQTTzCADuH3bNE4G7Ot/f+qLcc+cTGIXrduXYN3e7vssss0cuRIrVy5UuPGjdM111yjoqIi3X///erevXuDp0OnpaXpwgsv1L333uvcoW7btm2uW73PmjVL+fn5GjJkiO68806ddNJJqqioUGFhoVavXq2FCxc6v3QNqf0FONq1AZMnT9bhw4d17rnnKiMjQ0VFRXr00Ue1efNmLV261PpH59ia6ymnnKJbb71VM2bMUPv27TVo0CCtXbtWixYt0gMPPBAXp3Pbmu3w4cPrzXv99dclSeeee67S0tKafL0NbM1WOnLN1xdffCHpyD8Zu3bt0v/+7/9KkoYNG6auXbsedR1+ZWuuJSUl+uUvf6lzzz1XycnJ2rZtmx588EG1a9dO8+fP97h1/M3WbJ977jndeuutOuOMMzRmzJh6jysbMGBAmz1vNlpszVaK789jyd5sZ8yYod27d+v8889Xjx49tG/fPr3yyitavHixrr32Wg0cONDjFgqDNr+VWRuqvUNdY2Xnzp3GGGPy8vJMnz59TDAYNN/85jfN4sWLnTvmhpJkxo8fbxYsWGCys7NNYmKiycnJMU899VS99/7iiy/MnXfeabKyskxiYqLp0qWLGThwoJk+fbrZv3+/a51171DXu3dv07t376P+fEuWLDGDBw82Xbp0MQkJCaZz587mkksuMWvWrGn2tvIT23M1xpjKykozY8YM06tXL9OhQwdz4oknmkceeaRZ28mP4iHbuuLt7tw2Zzts2LBGf77169c3Z3P5hu257t271+Tm5pquXbuaxMRE06tXLzNx4kTr91dj7M/2xhtv9PTz2cj2bI2Jz89jY+zPdtWqVebb3/62SU9PNwkJCaZTp05m8ODB5pFHHjGHDx9u9vZqjYAx/7mAFgAAAAAANCl+7zwFAAAAAEAzMYgGAAAAAMAjBtEAAAAAAHjEIBoAAAAAAI8YRAMAAAAA4FHEnhO9YMECPfTQQ/r88891yimnaO7cuTrvvPOO+rqamhp99tlnSk5O5qH3UWaMUXl5uTIzM9Wu3ZHvW1qaq0S2sYRs7UW29gpntuQaO9hn7UW29iJbOzWUa1MLh92KFStMYmKiWbx4sfnwww/NXXfdZZKSksyuXbuO+tqioqImn29GaftSVFTU6lzJNjYL2dpbyNbeEo5syTX2CvusvYVs7S1ka2epzbUpERlEDx482Nxxxx2ueTk5OWbq1Kn1lq2oqDClpaVO+cc//hH1DUdxl3379jU7V7L1RyFbewvZ2ltaki25xn5hn7W3kK29hWztLLW5NiXs10RXVlbqvffeU25urmt+bm6uNm7cWG/5OXPmKDU11Sm9evUKd5fQSoFAoNm5SmTrB2RrL7K1V0uyJdfYxz5rL7K1F9naycsp9WEfRO/Zs0fV1dVKT093zU9PT1dxcXG95adNm6bS0lKnFBUVhbtLCIPm5iqRrV+Qrb3I1l78rbUT+6y9yNZeZBufInZjsbojeGNMg6P6YDCoYDAYqW4gzLzmKpGt35CtvcjWXvyttRP7rL3I1l5kG1/CfiQ6LS1N7du3r/fNS0lJSb1vaOAf5GovsrUX2dqLbO1ErvYiW3uRbXwK+yC6Q4cOGjhwoPLz813z8/PzNWTIkHC/HdoIudqLbO1FtvYiWzuRq73I1l5kG6eOeuuxFqi9zfuSJUvMhx9+aCZNmmSSkpJMYWHhUV9bWloa9TuyUdyltLS01bmSbWwWsrW3kK29JRzZkmvsFfZZewvZ2lvI1s5Sm2tTIjKINsaY+fPnm969e5sOHTqYM8880xQUFHh6Hb9EsVdCf5FamivZxmYhW3sL2dpbwpEtucZeYZ+1t5CtvYVs7SxeBtEBY4xRDCkrK1Nqamq0u4EQpaWlSklJafV6yDb2kK29yNZe4ciWXGMP+6y9yNZeZGsnL7mG/ZpoAAAAAABsxSAaAAAAAACPGEQDAAAAAOARg2gAAAAAADxiEA0AAAAAgEcMogEAAAAA8IhBNAAAAAAAHjGIBgAAAADAIwbRAAAAAAB4xCAaAAAAAACPGEQDAAAAAOBRQrQ7AAAAEAmPPvqoU9+1a5er7eGHH27r7gAALMGRaAAAAAAAPGIQDQAAAACARwyiAQAAAADwiGuiAQCAFX70ox+5pseOHevU//Wvf7nauCYaANBSHIkGAAAAAMAjBtEAAAAAAHjE6dwAAMCXqqurXdNbt251Ta9du9apP/74423SJwCA/TgSDQAAAACARwyiAQAAAADwiEE0AAAAAAAecU00AADwjUcffdSpv/766662K6+80jW9b98+p75p0yZX24svvhj2vgFAW3vppZdc06effrpT79mzZ4vWuXz5ctf0zTff3KL12Iwj0QAAAAAAeNTsQfSGDRv0ne98R5mZmQoEAnrhhRdc7cYYzZw5U5mZmerYsaOGDx9e726Z8B9ytRfZ2ots7USu9iJbe5Gtvcg2PjX7dO6vvvpKp59+um6++WZ997vfrdf+4IMP6le/+pWWLVumE088UQ888IAuvvhibd++XcnJyWHpNNoeudqLbO1FtnaaO3duXOV69tlnu6bHjh3r1C+++GJX2/79+13Thw4dcuo5OTkR6F14sc/ai2ztFe1sR40a5Zru3r27U7/jjjsafd2QIUNc08OGDXPqN954o6uN07nra/aR6BEjRuiBBx7Q1VdfXa/NGKO5c+dq+vTpuvrqq9W/f38tX75cBw4c0NNPP93g+g4dOqSysjJXQWxpSa4S2foB2dqLbO312GOPkauF2GftRbb2Itv4FdZronfu3Kni4mLl5uY684LBoIYNG6aNGzc2+Jo5c+YoNTXVKSeccEI4u4QwKCwsbHauEtn6Adnai2zttXv3bnK1EPusvcjWXmQbv8I6iC4uLpYkpaenu+anp6c7bXVNmzZNpaWlTikqKgpnlxAGJSUlkpqXq0S2fkC29iJbu5Grfdhn7UW29iLb+BWRR1wFAgHXtDGm3rxawWBQwWAwEt1AmDUnV4ls/YRs7UW2doqnXPv06dNo23vvvdfka0N/5oYuQ4tF8ZRtvCFbe0Uz26qqKtd06IB8+vTpjb7uuuuuc02HXhMd+ihBNCysR6IzMjIkqd43LyUlJfW+oYF/dOvWTRK52ohs7UW2diNX+7DP2ots7UW28Susg+isrCxlZGQoPz/fmVdZWamCgoJ6d4CDf/Tp04dcLUW29iJbe6Wnp5Orhdhn7UW29iLb+NXs07n379+vHTt2ONM7d+7U5s2b1aVLF/Xq1UuTJk3S7Nmz1a9fP/Xr10+zZ8/WscceqxtuuCGsHUfbCQQC5GopsrUX2dpr7NixcZXrrl27Wvza0NMpv/jii3B0J2LYZ+1FtvaKhWzrXvLy5JNPNrrs8ccf79Sbeuzf1KlTW90v2zV7EP3uu+/qggsucKYnT54s6cjzxJYtW6YpU6bo4MGDGjdunL788kudffbZWrt2Lc/A8zlytRfZ2ots7TRp0iQZY8jVQuyz9iJbe5FtfAoYY0y0OxGqrKxMqamp0e4GQpSWliolJaXV6yHb2EO29iJbe4UjWz/les4557im33zzTafepUsXV1vdZ61WV1c79dq76Nbq3r17uLoYFuyz9iJbe8VCtpE4Ep2UlOSaPnjwYIv65ldecg3rNdEAAAAAANgsIo+4AgAACIemjpYcPnzYNd23b1/X9GeffebU6x55njhxolPncS4A/KpTp06u6aFDh3p6Xd2Tkf/0pz859Xg78twSHIkGAAAAAMAjBtEAAAAAAHjE6dwAAMSoq666yqkfPnzYdbpdvFiyZIlr+qWXXnLqdU85XL58uWt60KBBTv3zzz93tf3617926itWrHC1xfrjsACg1gcffOCaDn20X1Pmzp3rmr7zzjvD1aW4wJFoAAAAAAA8YhANAAAAAIBHDKIBAAAAAPCIa6IBAIhRf/jDH5x6WVmZUlNTo9ib2NOtWzfX9Nlnn+2aLi4uduqXXnqpq2316tVOve710qGPw+L6aACxJjEx0aknJLiHc14fT/XWW2+5prkmunk4Eg0AAAAAgEcMogEAAAAA8IhBNAAAAAAAHvnimuhevXo59Q4dOrjahgwZ4tSHDh3qajvuuOOc+ne/+90Wvfc///lP1/S3vvUtp/7www+72q6//nqnXlFR4WrLy8tzTd93330t6g/QUjU1NU69rKzMtX/YYOrUqU79hRdecLVt27atjXsDhEfofhtajyd1r0keOXKkU697LfOOHTsaXc+aNWs8v2d2drZT79q1q6tt165dTv2rr77yvE4ACJdDhw459fnz57va9u/f3+jrQv9XMsa42rxeS40jOBINAAAAAIBHDKIBAAAAAPAoYOoey4+y2kd4nHrqqWrfvr0kad26dU57WzzeI/SUuVtuucXV1tQpEqGnlX355Zeutu3bt4epd22vtLRUKSkprV4Pj2eJvKeffto1/f/+3/9z6i+//LJTP3DggK699lqrsg39KAsEAlHsSWywKdt4Vl1d7dTLysrUuXPnsGTrp1zrPsbqb3/7m1Ov+2iXuqdeN2Xs2LFOfcGCBa620O3elLrv3xrss/YiW3tFK9vf/OY3Tv2mm25qdLlPPvnENf3KK6849YkTJ3rvYJzxkitHogEAAAAA8IhBNAAAAAAAHjGIBgAAAADAo5h9xNU///lP57rGvXv3OvPDdS1I6KMuLrjgAldbZWWlU3/yySfD8n5ApIRew1/3d/mGG25o6+4ACKOqqqoG6/Ek9G+yJF111VVO/c9//nOL1/vYY481WJfU5LVwZWVlLX5PAAiH0Hs21b1/E9oGR6IBAAAAAPCIQTQAAAAAAB7F7OncoY+I+slPfuLUR44c6Vrur3/9q1N/5JFHGl1f3VOzvvrqK6d+yimnuNruuuuu5nUWaEOhp29L0t133+3UCwoK2ro7MeO//uu/ot0FICzOOOMMp/6vf/3LqZeXl0ehN9G3b98+13RrTuH2ilO2AQBN4Ug0AAAAAAAeNWsQPWfOHA0aNEjJycnq1q2brrrqKm3fvt21jDFGM2fOVGZmpjp27Kjhw4dr69atYe00ooNs7USu9iJbf9q9e7e2b9+u999/X4MGDdKYMWP06aefupYhW3uRrZ3I1V5kG5+aNYguKCjQ+PHj9fbbbys/P19VVVXKzc11nRr94IMP6le/+pXmzZunTZs2KSMjQxdffHHcnoZmE7K1E7nai2z9af/+/UpLS1O/fv3029/+VlVVVfrBD36gAwcOOMuQrb3I1k7kai+yjVOmFUpKSowkU1BQYIwxpqamxmRkZJi8vDxnmYqKCpOammoWLlzoaZ2lpaVGUovKQw895CrXX3+9U1q6TopMaWlp1LON93Ldddc55ZFHHnGV1qx33759rc61Jdk++eSTTmnO61JSUpzy0ksvucrhw4edEu28YqFEK1tK68vw4cOdYszXf2tXr14dtmzJNfYKf2vtLXwe21vI1s5SWlp61MxadU10aWmpJKlLly6SpJ07d6q4uFi5ubnOMsFgUMOGDdPGjRsbXMehQ4dUVlbmKog9ZGunwsLCZucqka0fkK09av/Wdu7cWVLLsiVXf+BvrZ34PLYX2cavFg+ijTGaPHmyhg4dqv79+0uSiouLJUnp6emuZdPT0522uubMmaPU1FSnnHDCCS3tEiKIbO1UUlIiqXm5SmTrB2Rrh9C/tSeffLKklmVLrv7A31o78XlsL7KNXy1+xNWECRP0/vvv680336zXFggEXNPGmHrzak2bNk2TJ092psvKypr1i5SUlOTUBw4c6GoLXe/evXtdbWvXrvX8HvhaW2aLr61YscKpz58/39UWuk0XLFjganv99ded+sGDB/WnP/1Jb775plJSUpSamuq0NSdXqfnZnnbaaY22NeV//ud/XNPvvPOOU3/ppZdcbX//+99b9B62i3S2CL/Q/bZdu6+/6w7dZyU+j21Gtnbi8zg2VFdXu6bbt2/f6nWSbfxp0SB64sSJWrVqlTZs2KCePXs68zMyMiQd+Sa1e/fuzvySkpJ639DUCgaDCgaDLekG2hDZ+ttrr72mPXv2OPts7WlD3bp1k9S8XCWy9QOytVdLsiVXf+BvrZ34PLYX2cavZp3ObYzRhAkTtHLlSq1bt05ZWVmu9qysLGVkZCg/P9+ZV1lZqYKCAg0ZMiQ8PUZUkK0/GWP02muvaceOHQ3us3369CFXS5GtvcjWXvyttRP7rL3INn4160j0+PHj9fTTT+vFF19UcnKyc65/amqqOnbsqEAgoEmTJmn27Nnq16+f+vXrp9mzZ+vYY4/VDTfcEJEfAG2DbP3ptdde07Zt23TllVe69tnaU4zI1V5kay+ytRfZ2olc7UW28atZg+jHHntMkjR8+HDX/KVLl+qmm26SJE2ZMkUHDx7UuHHj9OWXX+rss8/W2rVrlZycHJYO1xX6jOrbb7/d1faXv/zFqS9evNjVtn79eqf+7rvvutpCrzk1xoSlnzZo62zxtfvvv9+p/+xnP3O1jR8/3qnPmzfP1fa3v/1NkvTss8/q2WefbXDdbZFr6HU/krRq1apGlw3d55q6nqjuz/OrX/2qhb2zF/usvcjWXmRrJ3KNrk8++cSpX3bZZWFdN9nGp4CJsVFiWVlZvZuneJWdne2aDh1E79u3z9XGINq70tJSpaSktHo9rck23t13331Ove4gOvR3tO4g+s4772xyvW2V7bJly1zTf/rTn5z6c88952oLxyD67bffbrK/8YD91l7hyJZcYw/7rL3INvpCB9Hjxo1zta1Zs6bF6yVbO3nJtVXPiQYAAAAAIJ60+BFXsSj0WyZJzinm0pFTzkN9//vfb7AuuR+b9Ytf/CKMPQRaZsaMGQ3W/eLgwYOu6dB9ru6R6KaOPjclJyfHqXMkGgAA1Ordu7dT79evn6utNUeiEb84Eg0AAAAAgEcMogEAAAAA8IhBNAAAAAAAHll1TXRdf/jDH5z6xx9/7GoLvZPvRRdd5GoLvW6iR48errZ//etf4ewiEBfGjh0b9nVed911rulFixY59bp3AwcAIFKCwaBTP3ToUBR7glp1n65TXV3t1Os+yQRoCY5EAwAAAADgEYNoAAAAAAA8svp07lAffPCBa/rJJ5906hdeeKGrbcyYMU697m3wL7744gj0DkBr/fCHP4x2FwAAlurTp49T37lzp6st9NTh888/36lXVVXxyMUoCT19W5ISEuJmyIM2wpFoAAAAAAA8YhANAAAAAIBHDKIBAAAAAPAobi4QmDVrlmt60KBBTj0QCDT6urS0NNd0u3Zff+9QU1MTpt4BAAAgVn366adOfceOHa62yZMnO/UNGzY49bKyMh133HER7xuOGDFihFNv6n97IBw4Eg0AAAAAgEcMogEAAAAA8Miq07lPOukk1/S6deucekZGhuf1hN4W//PPP3e1cQo3AABAfLn//vud+owZM1xt06dPd+rFxcVOvby8PPIdg+Pll1926ieeeGIUe4J4wJFoAAAAAAA8YhANAAAAAIBHMXc6tzGmxa8NPQ1bcp9Gc+yxx7ZoPQcOHGhxf2zRmkwisR6ED9nai2ztFY5MyDX2sM/GtoqKikbbDh065NRD//fcv3+/JLKNhra6/JJs7eQlj5gbRLfm+pG6jxzIyclpbXegI5mkpqaGZT2ILWRrL7K1VziyJdfYwz4b2+bMmdNo28MPP9xgvRbZtr2dO3e2yfuQrZ285BowMfbVR01NjT777DMZY9SrVy8VFRUpJSUl2t2KKWVlZTrhhBMivm2MMSovL1dmZqbr+dgtRbZHR7Z2aqtcpchku337dp188snk2gC/Zss+e3R8HtuLbO3k189jib+1RxOL+2zMHYlu166devbsqbKyMklSSkoKv0iNaIttE45v12qRrXdka6e22i7hzrZHjx6SyLUpfsuWfdY7Po/tRbZ28tvnscTfWq9iaZ/lxmIAAAAAAHjEIBoAAAAAAI9idhAdDAY1Y8YMBYPBaHcl5vh92/i9/5Hk923j9/5Hit+3i9/7H0l+3zZ+738k+X3b+L3/keT3beP3/keK37eL3/sfSbG4bWLuxmIAAAAAAMSqmD0SDQAAAABArGEQDQAAAACARwyiAQAAAADwiEE0AAAAAAAeMYgGAAAAAMCjmB1EL1iwQFlZWTrmmGM0cOBAvfHGG9HuUpuaM2eOBg0apOTkZHXr1k1XXXWVtm/f7lrGGKOZM2cqMzNTHTt21PDhw7V169Yo9dibeM9VIltb2ZqrRLZkay9bs433XCWytZWtuUpk67tsTQxasWKFSUxMNIsXLzYffvihueuuu0xSUpLZtWtXtLvWZi655BKzdOlS88EHH5jNmzebyy+/3PTq1cvs37/fWSYvL88kJyeb559/3mzZssWMHj3adO/e3ZSVlUWx540j1yPI1k425moM2RpDtjazMVtyPYJs7WRjrsaQrTH+yzYmB9GDBw82d9xxh2teTk6OmTp1apR6FH0lJSVGkikoKDDGGFNTU2MyMjJMXl6es0xFRYVJTU01CxcujFY3m0SuDSNbO9mQqzFk2xCytZcN2ZJrw8jWTjbkagzZNiTWs42507krKyv13nvvKTc31zU/NzdXGzdujFKvoq+0tFSS1KVLF0nSzp07VVxc7NpOwWBQw4YNi8ntRK6NI1s7+T1XiWwbQ7b28nu25No4srWT33OVyLYxsZ5tzA2i9+zZo+rqaqWnp7vmp6enq7i4OEq9ii5jjCZPnqyhQ4eqf//+kuRsC79sJ3JtGNnayYZcJbJtCNnay4ZsybVhZGsnG3KVyLYhfsg2oc3f0aNAIOCaNsbUmxcvJkyYoPfff19vvvlmvTa/bSe/9TfSyNZONuUq+bPPkUK29rIpW7/1N9LI1k425Sr5s8+R4odsY+5IdFpamtq3b1/vG4WSkpJ63zzEg4kTJ2rVqlVav369evbs6czPyMiQJN9sJ3Ktj2ztZEuuEtnWRbb2siVbcq2PbO1kS64S2dbll2xjbhDdoUMHDRw4UPn5+a75+fn5GjJkSJR61faMMZowYYJWrlypdevWKSsry9WelZWljIwM13aqrKxUQUFBTG4ncv0a2drJtlwlsq1FtvayLVty/RrZ2sm2XCWyreW7bCN/77Lmq73N+5IlS8yHH35oJk2aZJKSkkxhYWG0u9Zmxo4da1JTU83rr79uPv/8c6ccOHDAWSYvL8+kpqaalStXmi1btpjrr78+pm/hT65HkK2dbMzVGLI1hmxtZmO25HoE2drJxlyNIVtj/JdtTA6ijTFm/vz5pnfv3qZDhw7mzDPPdG5vHi8kNViWLl3qLFNTU2NmzJhhMjIyTDAYNOeff77ZsmVL9DrtQbznagzZ2srWXI0hW7K1l63ZxnuuxpCtrWzN1Riy9Vu2gf90GgAAAAAAHEXMXRMNAAAAAECsYhANAAAAAIBHDKIBAAAAAPCIQTQAAAAAAB4xiAYAAAAAwCMG0QAAAAAAeMQgGgAAAAAAjxhEAwAAAADgEYNoAAAAAAA8YhANAAAAAIBHDKIBAAAAAPCIQTQAAAAAAB4xiAYAAAAAwCMG0QAAAAAAeMQgGgAAAAAAjxhEAwAAAADgEYNoAAAAAAA8YhANAAAAAIBHDKIBAAAAAPCIQTQAAAAAAB4xiAYAAAAAwCMG0QAAAAAAeMQgGgAAAAAAjxhEAwAAAADgEYNoAAAAAAA8YhANAAAAAIBHDKIBAAAAAPDI6kH0smXLFAgE9O6774ZlfYFAQBMmTAjLukLXOXPmzBa//p577tHIkSPVo0cPBQIB3XTTTWHrW6yKh1xDvfrqqwoEAgoEAtqzZ09Y1hmrbM+2qKhIo0aN0je+8Q0lJSUpNTVVAwYM0Lx581RVVRXWfsYa27MtLCx09tO6ZcWKFWHtZyyxPVf2WXuzrfXBBx/o2muvVdeuXRUMBtWnTx+NGzcuPB2MUfGQbTz+fyzZn+3MmTMb/Vvb1n9vE9rsnRARv/71r3Xaaafpiiuu0G9+85todwdhtn//ft1+++3KzMzUZ599Fu3uoJW++uorpaSk6N5771WvXr1UWVmp1atXa+LEidq8ebOeeOKJaHcRrTRx4kTdcMMNrnn9+vWLUm/QWuyzdlu/fr0uv/xynXfeeVq4cKHS0tL0j3/8Q3/961+j3TW0Ev8f2+m2227TpZdeWm/+7bffrk8++aTBtkhhEO1z5eXlatfuyAkFTz75ZJR7g3CbOnWqOnfurMsvv1wPPPBAtLuDVsrJydHy5ctd80aMGKGSkhItX75c8+fPVzAYjFLvEA69evXSOeecE+1uIEzYZ+114MABfe9739OFF16ol156SYFAwGn7/ve/H8WeIRz4/9hOPXv2VM+ePV3zCgsLtXXrVn3ve9/Tcccd12Z9sfp0bi8qKip0991364wzzlBqaqq6dOmib33rW3rxxRcbfc3jjz+uE088UcFgUCeffHKDpw4UFxdrzJgx6tmzpzp06KCsrCzdd999YT/9q/YDAm5+z1WS3njjDS1atEhPPPGE2rdvH/b1+5UN2dbVtWtXtWvXLu5ztjFb2Jkr++wRfs72ueee0+eff66f/OQnrgE0jvBzthL/HzfF79nW9Zvf/EbGGN12220RfZ+64v5I9KFDh/Tvf/9bP/7xj9WjRw9VVlbq1Vdf1dVXX62lS5fqBz/4gWv5VatWaf369Zo1a5aSkpK0YMECXX/99UpISNA111wj6cgv0eDBg9WuXTv97Gc/U3Z2tt566y098MADKiws1NKlS5vsU58+fSQd+WYFLeP3XA8ePKhbb71VkyZN0plnnqlVq1a1aDvYyO/ZSpIxRtXV1SovL9fatWu1bNky3X333UpIiO+PZBuyzcvL009/+lMlJCTozDPP1JQpU3TFFVc0e1vYxIZc2Wcb5udsN2zYIEmqrq7W0KFD9c477ygpKUmXXnqpfvnLXyozM7NlG8USfs4WTbMp25qaGi1btkx9+/bVsGHDmvXaVjMWW7p0qZFkNm3a5Pk1VVVV5vDhw+bWW281AwYMcLVJMh07djTFxcWu5XNyckzfvn2deWPGjDGdOnUyu3btcr3+4YcfNpLM1q1bXeucMWOGa7ns7GyTnZ3tuc+1kpKSzI033tjs1/lNPOR69913m2984xvmwIEDxhhjZsyYYSSZL774wtPr/SoesjXGmDlz5hhJRpIJBAJm+vTpnl/rV7Zn+9lnn5nbb7/dPPvss+aNN94wTz31lDnnnHOMJLN48WLPP7Pf2J5rLfZZb/yU7SWXXGIkmeOOO85MmTLFrFu3zixcuNAcf/zxpm/fvuarr77y/HP7je3Z1hUv/x8bE3/Zvvzyy0aSmTNnTrNf21qc66Ajp/Sce+656tSpkxISEpSYmKglS5bo73//e71lL7roIqWnpzvT7du31+jRo7Vjxw7985//lCT98Y9/1AUXXKDMzExVVVU5ZcSIEZKkgoKCJvuzY8cO7dixI4w/YXzya67vvPOO5s6dq8cff1wdO3Zszo8cN/yaba2bbrpJmzZt0po1azRlyhQ99NBDmjhxoufX28yv2Xbv3l2LFi3Stddeq6FDh+qGG27Qhg0bNGDAAE2dOjXuTx33a6612Gcb59dsa2pqJEmjR4/WL37xC11wwQUaM2aMlixZoh07dujpp5/2vA1s5ddscXS2ZLtkyRIlJCRE5e7rcT+IXrlypa677jr16NFDv/vd7/TWW29p06ZNuuWWW1RRUVFv+YyMjEbn7d27V5K0e/duvfTSS0pMTHSVU045RZKsf0xRLPBzrrfccouuvvpqnXXWWdq3b5/27dvn9LmsrEzl5eVheR+/8nO2oe9/1llnKTc3V3l5eZo1a5bmzZsX93eEtSHbUImJiRo9erT27t2rjz/+OGLvE+tsyJV9tmF+zvb444+XJF1yySWu+ZdccokCgYD+8pe/hOV9/MrP2aJptmS7Z88erVq1SpdffnmDfYy0+L6YR9Lvfvc7ZWVl6ZlnnnHdWOLQoUMNLl9cXNzovNoP5LS0NJ122mn6+c9/3uA64v06m7bg51y3bt2qrVu36rnnnqvXlp2drdNPP12bN28Oy3v5kZ+zbczgwYMlSR999JEGDBgQ0feKZTZma4yRFN83ubExV/bZI/yc7WmnndbkM2XjeZ+V/J0tmmZLtk8++aQqKyvb/IZiteJ+EB0IBNShQwfXL1FxcXGjd6h77bXXtHv3bue0hurqaj3zzDPKzs52brk+cuRIrV69WtnZ2ercuXPkfwjU4+dc169fX2/esmXLtHz5cr3wwgvq0aNHxN7bD/ycbWNqM+/bt2+bv3cssS3bw4cP65lnnlFaWlpcZ2tbrhL7bC0/Zztq1ChNnz5dL7/8skaNGuXMf/nll2WMiftH1fk5WzTNlmyXLFmizMxM55TxthYXg+h169Y1eLe3yy67TCNHjtTKlSs1btw4XXPNNSoqKtL999+v7t27N3j6XVpami688ELde++9zh3qtm3b5vo2c9asWcrPz9eQIUN055136qSTTlJFRYUKCwu1evVqLVy4sN4zzkLV/lH2cm1AQUGBvvjiC0lHfql37dql//3f/5UkDRs2TF27dj3qOvzK1lyHDx9eb97rr78uSTr33HOVlpbW5OttYGu2M2bM0O7du3X++eerR48e2rdvn1555RUtXrxY1157rQYOHOhxC/mXrdlOnjxZhw8f1rnnnquMjAwVFRXp0Ucf1ebNm7V06VLrH4Vka67ss/Zmm5OTo/Hjx2vBggVKTk7WiBEj9NFHH+mee+7RgAEDdN1113ncQv5la7ZSfP9/LNmdrST9+c9/1tatW/XTn/40en9f2/xWZm2o9g51jZWdO3caY4zJy8szffr0McFg0Hzzm980ixcvdu6GHEqSGT9+vFmwYIHJzs42iYmJJicnxzz11FP13vuLL74wd955p8nKyjKJiYmmS5cuZuDAgWb69Olm//79rnXWvUNd7969Te/evT39jMOGDWv051u/fn1zNpdvxEOudcXb3bltzXbVqlXm29/+tklPTzcJCQmmU6dOZvDgweaRRx4xhw8fbvb28hPbs12yZIkZPHiw6dKli0lISDCdO3c2l1xyiVmzZk2zt5Wf2J4r+6y92Rpz5C7DeXl5pm/fviYxMdF0797djB071nz55ZfN2VS+Ew/ZxuP/x8bER7bGGHP77bebQCBgPvnkE8+vCbeAMf+5YAsAAAAAADQpvu+aAAAAAABAMzCIBgAAAADAIwbRAAAAAAB4xCAaAAAAAACPGEQDAAAAAOBRxJ4TvWDBAj300EP6/PPPdcopp2ju3Lk677zzjvq6mpoaffbZZ0pOTnY9BBxtzxij8vJyZWZmql27I9+3tDRXiWxjCdnai2ztFc5syTV2sM/ai2ztRbZ2aijXphYOuxUrVpjExESzePFi8+GHH5q77rrLJCUlmV27dh31tUVFRU0+34zS9qWoqKjVuZJtbBaytbeQrb0lHNmSa+wV9ll7C9naW8jWzlKba1MiMogePHiwueOOO1zzcnJyzNSpU+stW1FRYUpLS53yj3/8I+objuIu+/bta3auZOuPQrb2FrK1t7QkW3KN/cI+a28hW3sL2dpZanNtStivia6srNR7772n3Nxc1/zc3Fxt3Lix3vJz5sxRamqqU3r16hXuLqGVAoFAs3OVyNYPyNZeZGuvlmRLrrGPfdZeZGsvsrWTl1Pqwz6I3rNnj6qrq5Wenu6an56eruLi4nrLT5s2TaWlpU4pKioKd5cQBs3NVSJbvyBbe5Gtvfhbayf2WXuRrb3INj5F7MZidUfwxpgGR/XBYFDBYDBS3UCYec1VIlu/IVt7ka29+FtrJ/ZZe5Gtvcg2voT9SHRaWprat29f75uXkpKSet/QwD/I1V5kay+ytRfZ2olc7UW29iLb+BT2QXSHDh00cOBA5efnu+bn5+dryJAh4X47tBFytRfZ2ots7UW2diJXe5Gtvcg2Th311mMtUHub9yVLlpgPP/zQTJo0ySQlJZnCwsKjvra0tLTJu6UlJia6SlPLUsJTSktLW52rl2wpZEuxO9ucnByn3Hfffa4S7e3lpxKObNlnY6/E4j5LIVsK2cZjqc21KRG5Jnr06NHau3evZs2apc8//1z9+/fX6tWr1bt370i8HdoIudqLbO1FtvYiWzuRq73I1l5kG38CxhgT7U6EKisrU2pqaqPtiYmJrunDhw9Huktxr7S0VCkpKa1ez9GyRdsjW3vFYrY5OTlO/frrr3e1zZgxIyzvEQ/CkS37bOyJxX0W4UG29iJbO3nJNezXRAMAAAAAYKuIPeIqUup+S1NSUuLU27XjOwEAaEtPPPGEa/rUU0916u+9956rLfToc91veO+55x6nvnfvXldbt27dWt1PAIhndU88bezRSwC8YdQJAAAAAIBHDKIBAAAAAPDIF6dzp6WlOfXQ07frOuOMM1zTmzdvjlCPACB+HX/88U69pqbG1Zadne3U615iE3o5Tnl5uatt0qRJTv33v/99OLoJAPiPXr16RbsLgFU4Eg0AAAAAgEcMogEAAAAA8IhBNAAAAAAAHvnimuimcIt+AIi8adOm6ZhjjpEkff/733fm9+7d27XcvHnznPpdd93VNp0DADSpqKjINT1s2DCnXlBQ0NbdAXyPI9EAAAAAAHjEIBoAAAAAAI98fzq3Mcap155qCAAIrylTpiglJUWS9OyzzzrzTz75ZNdyFRUVbdovAEDzvf76606dSyOB5uNINAAAAAAAHjGIBgAAAADAIwbRAAAAAAB45Itroquqqjwt9/bbb0e4JwAQn9q1a6d27Y587/r73//emc810ADgP0uWLIl2FwBf40g0AAAAAAAeMYgGAAAAAMAjX5zO/dvf/rbRtv/5n/9pw54AQHzq06eP8xiUO++805l/8OBB13L5+flt2i8AQPPddttt0e4C4GsciQYAAAAAwCMG0QAAAAAAeMQgGgAAAAAAj3xxTTQAILq+/PJLpz5v3jynvnv3btdyp5xyilPftm1b5DsGoFUeffRRpz5x4sQo9gQA/IMj0QAAAAAAeNTsQfSGDRv0ne98R5mZmQoEAnrhhRdc7cYYzZw5U5mZmerYsaOGDx+urVu3hqu/iBJytRfZ2ots7USu9iJbe5Gtvcg2TplmWr16tZk+fbp5/vnnjSTzhz/8wdWel5dnkpOTzfPPP2+2bNliRo8ebbp3727Kyso8rb+0tNRIcpVVq1Y5paamxlV++9vfOqXu6yjhKaWlpa3OtbFsKWRL8Xe21dXVrrJmzRqnpKSkuEq0t4ktZebMmeyzFpZI7rNvv/22U/70pz+5SrR/7ngo/K21t5CtnaW0tPSomTV7EO16sdyD6JqaGpORkWHy8vKceRUVFSY1NdUsXLiwwXVUVFSY0tJSpxQVFdX7QRhER7fs27ev2bl6zZZCthR/Z8sguu1Leno6+6yFJZL7LINoe7OlkC0l/MXLIDqs10Tv3LlTxcXFys3NdeYFg0ENGzZMGzdubPA1c+bMUWpqqlNOOOGEcHYJYVBYWNjsXCWy9QOytRfZ2mv37t3kaiH2WXuRrb3INn6FdRBdXFwsSUpPT3fNT09Pd9rqmjZtmkpLS51SVFQUzi4hDEpKSiQ1L1eJbP2AbO1FtnYjV/uwz9qLbO1FtvErIo+4CgQCrmljTL15tYLBoILBYIvf64orrmjxa9E8zclVan22aDtka69wZbtjxw4lJyfXW2fddU2bNs2pl5WVtajPODr2WXuFK9tLLrlEiYmJkqSzzz670fWj7bDf2ots409Yj0RnZGRIUr1vXkpKSup9QwP/6NatmyRytRHZ2ots7Uau9mGftRfZ2ots41dYB9FZWVnKyMhQfn6+M6+yslIFBQUaMmRION8KbahPnz7kaimytRfZ2is9PZ1cLcQ+ay+ytRfZxq9mn869f/9+7dixw5neuXOnNm/erC5duqhXr16aNGmSZs+erX79+qlfv36aPXu2jj32WN1www3Nep8rrrjCOQ1p5MiRzvy6pwced9xxzf0R0EyBQCBsuSK2kK29wp1t3759PS130kknNXvdaJ6xY8eyz1oo3PvsjTfeqGOPPVaS9Pbbbzvz58+f71ru2WefdeoFBQUt7D2awt9ae5Ft/Gr2IPrdd9/VBRdc4ExPnjxZ0pEP62XLlmnKlCk6ePCgxo0bpy+//FJnn3221q5d61xLB38iV3uRrb3I1k6TJk2SMYZcLcQ+ay+ytRfZxqeAMcZEuxOhysrKlJqa6joS/dxzz7naQ3EkOvJKS0uVkpLS6vXUZovYQbb2ila2oUeit2/f3ur3R33hyJZ9NvaEe599+umnnSPRoddm/uUvf3Etz5HoyONvrb3I1k5ecg3rNdEAAAAAANgsIo+4CoeqqqoGbw3foUMH13Tv3r2d+q5duyLeLwBA4zj6DMSGyy+/3DmSEnrSYUKC+1+/8ePHt2m/AMAGHIkGAAAAAMAjBtEAAAAAAHgUs6dzr1692qlv27bNqefk5LiW++///m+nPm7cuMh3DAAAIMZxkyIAiByORAMAAAAA4BGDaAAAAAAAPGIQDQAAAACARzF7TXSotWvXOvUePXq42iZPntzW3QEAAAAAxCmORAMAAAAA4BGDaAAAAAAAPGIQDQAAAACARzF7TXReXp6OOeYYSdJNN93kzO/cubNruZqamrbsFoAYMWXKFNf0D3/4Q6e+aNGiRl83d+5c13RlZWVY+wUAAIDYMm3aNNf0/fffX2+ZsrIydenSxdP6OBINAAAAAIBHDKIBAAAAAPAoZk/nHjVqlJKTkyVJKSkpzvwrr7zStdwf/vCHNu0XgLZz8sknu6a3bt3q6XW/+MUvGm3btGmTa3r9+vXN7xgAAACibtiwYU69f//+rrZHHnnEqRtjwvq+HIkGAAAAAMAjBtEAAAAAAHjEIBoAAAAAAI9i9pro//u//9Oxxx4rSfrmN7/pzP/LX/4SrS4BaANjxoxx6rNmzQrLOvft2+fUv/Od77jauCYaAAAgdnTv3t01/ec//7nRZTMzMxtt69atm1NPS0tztd11111OPfQxqV5xJBoAAAAAAI8YRAMAAAAA4FHMns59yy23RLsLANpAly5dXNOhp3B36tSp0df94x//cE336tWr0WUHDhzo1D/99NPmdhEAAAAR9O1vf9upv/LKK40ud+jQIdf0pEmTnPq8efMafd3evXtd0y05hTsUR6IBAAAAAPCoWYPoOXPmaNCgQUpOTla3bt101VVXafv27a5ljDGaOXOmMjMz1bFjRw0fPlxbt24Na6cRHWRrJ3K1F9nai2ztRbZ2Ild7kW18atYguqCgQOPHj9fbb7+t/Px8VVVVKTc3V1999ZWzzIMPPqhf/epXmjdvnjZt2qSMjAxdfPHFKi8vD3vn0bbI1k7kai+ytRfZ2ots7USu9iLbOGVaoaSkxEgyBQUFxhhjampqTEZGhsnLy3OWqaioMKmpqWbhwoWe1llaWmokUWKolJaWkq2lZd++fa3OtbXZ3nHHHa6yefNmp9S1ZcsWpzRHTk6OU6K9zeMpW0rsZmtLrmeddZar1G2fNm2aU/r06eMq0e573cLfWnsLn8f2FrINX6mqqmq0fPrpp04599xzXaWpdXbu3Nkp+/btc5W//e1vTsnIyDAZGRmmW7duRjryeXw0rbomurS0VNLXNwbauXOniouLlZub6ywTDAY1bNgwbdy4scF1HDp0SGVlZa6C2EO2diosLGx2rhLZ+gHZ2qsl2ZKrP/C31k58HtuLbONXiwfRxhhNnjxZQ4cOVf/+/SVJxcXFkqT09HTXsunp6U5bXXPmzFFqaqpTTjjhhJZ2CRFEtnYqKSmR1LxcJbL1A7K1V0uyJVd/4G+tnfg8thfZxq8WP+JqwoQJev/99/Xmm2/WawsEAq5pY0y9ebWmTZumyZMnO9NlZWX8IsUwsrVTc3KVwpvtwoULXdN79uxx6tOnT3e1nXHGGZ7W+fzzz7umt23b1qK+2SCa2SKy4uXz+Iknnmi07dZbb3VN19TUuKbbtfv6WMH777/vaissLGx95yIkXrKNN3we24tsm6+qqqrRtrpH8c8///wWvccXX3zR6PuF/v/Z1BcejWnRIHrixIlatWqVNmzYoJ49ezrzMzIynI50797dmV9SUlLvG5pawWBQwWCwJd1AGyJbO3Xr1k1S83KVyNYPyNZeLcmWXP2Bv7V24vPYXmQbv5p1OrcxRhMmTNDKlSu1bt06ZWVludqzsrKUkZGh/Px8Z15lZaUKCgo0ZMiQ8PQYUUG2durTpw+5Wops7UW29uJvrZ3YZ+1FtvGrWUeix48fr6efflovvviikpOTnUPfqamp6tixowKBgCZNmqTZs2erX79+6tevn2bPnq1jjz1WN9xwQ0R+ALQNsrUTudqLbO1FtvYiWzuRq73INn41axD92GOPSZKGDx/umr906VLddNNNkqQpU6bo4MGDGjdunL788kudffbZWrt2rZKTk8PSYUQP2dop1nINvc9C+/btW72OeBZr2SJ8bMu2Odc9h/rrX//qmh4wYECjy/7xj390TQ8aNMipv/vuu0frYpuxLVscQa72Ilvvvvvd7zr1v/3tb662gQMHtnr9TV1nfeedd7qmFy1a1Kr3ChhjTKvWEGZlZWVKTU2NdjcQorS0VCkpKa1eD9nGnljMtvaaQElau3atq+3UU0/1tI4f/ehHrum5c+e2ul9+E4vZIjzCkW2s5doWg+i6Ym0QzT5rL7K1F9k2T+gg+qc//amrLdKD6HHjxrmmmxpEe8m1Vc+JBgAAAAAgnrT4EVdApB08eNCp33jjja62Z599tq27gwj53ve+55p+8sknnXpTj4eoKzs726l/+umnre8YgLAK/UyXpNdff92pX3rppY2+ru5j8E466SSnfuWVV7raysrKGl1P3aPUmzdvbnRZAED4hT6CtO7jSFsq9KTquo85DD2Fu6kznlqCI9EAAAAAAHjEIBoAAAAAAI8YRAMAAAAA4BHXRKPNffnll84d75p6hNExxxzj1LkG2t/+/ve/u6b79u3r1BMSWvYxNG/ePNc010ED0Vf3HgdnnXWWUw/9TJekYDDo1NevX+9qu/DCCz293/XXX++5b9u2bfO8LAAgNr388suu6dDroJ9++mlX21NPPdXgcuHAkWgAAAAAADxiEA0AAAAAgEeczo02N3LkyAZP4X388cdd0815vBFiz6hRo5x6Tk5O2NffvXv3sK8TwNF16tTJNV1eXu7pdXv37nVNez1lu64OHTo49d/85jeutsOHD7umJ06c6NSrq6sbXWf//v1d0zfffLNT/8UvfuFq27Nnj1MP9+mBAID6VqxY4dQvvvhiV9uPfvQjp/7II4+0WZ84Eg0AAAAAgEcMogEAAAAA8IhBNAAAAAAAHnFNNNrcSy+95Dzi6t1333Xm9+7d27XcmDFj2rRfaL2RI0cqMTFRknTCCSc48ysqKlzL1X3UjVfPP/+8U7/mmmtatA4ArTNjxgzPy4Y+zu7AgQNhef/Q+2cc7bNk4cKFDdbrqnttc7t2Xx9jGDhwoKtt9erVTv3BBx9surMAgGarey+d0P/5PvnkE1dbW14HHYoj0QAAAAAAeMQgGgAAAAAAjzidG20u9BFX69atc+afdtpp0eoSwuSPf/yjUx8xYoRTv+WWWxp9Td3Hnc2bN8+p1572X+uMM85w6nVP/9+1a1ez+gqgZa6++mrX9GWXXebUP/roI1db3dPuWqKqqso13b59+1avs67Q07frGjZsmGv6448/Dvv7A0C8Cz2F+4MPPnC1hT6u8LHHHmuzPjWFI9EAAAAAAHjEIBoAAAAAAI9i7nRuY0y0u4A6wpVJ7XpCT80rKytz6tXV1WF5H3gX7mxDHT58uMF6XXXvihv6O1FXeXl5o6+DWySzRXSFI5PWrCN0P5Tc+3ck9su6nwmROJ27OQ4ePBiR9bLP2ots7UW24RM6Dqj7uV9ZWdmmffGSR8wNouv+cUb0lZeXKzU1NSzrkaR33nnHmdelS5dWrxctF+5sQ7366qstWtczzzzT2u5Akc0W0RWObFuTa+i9CdpCvPydYJ+1F9nai2zDJ/R+E9H+3PeSa8DE2FcfNTU1+uyzz2SMUa9evVRUVFTv5kLxrqysTCeccELEt40xRuXl5crMzGzypiteke3Rka2d2ipXKTLZbt++XSeffDK5NsCv2bLPHh2fx/YiWzv59fNY4m/t0cTiPhtzR6LbtWunnj17OofxU1JS+EVqRFtsm3B8u1aLbL0jWzu11XYJd7Y9evSQRK5N8Vu27LPe8XlsL7K1k98+jyX+1noVS/ssNxYDAAAAAMAjBtEAAAAAAHgUs4PoYDCoGTNmKBgMRrsrMcfv28bv/Y8kv28bv/c/Uvy+Xfze/0jy+7bxe/8jye/bxu/9jyS/bxu/9z9S/L5d/N7/SIrFbRNzNxYDAAAAACBWxeyRaAAAAAAAYg2DaAAAAAAAPGIQDQAAAACARwyiAQAAAADwiEE0AAAAAAAexewgesGCBcrKytIxxxyjgQMH6o033oh2l9rUnDlzNGjQICUnJ6tbt2666qqrtH37dtcyxhjNnDlTmZmZ6tixo4YPH66tW7dGqcfexHuuEtnaytZcJbIlW3vZmm285yqRra1szVUiW99la2LQihUrTGJiolm8eLH58MMPzV133WWSkpLMrl27ot21NnPJJZeYpUuXmg8++MBs3rzZXH755aZXr15m//79zjJ5eXkmOTnZPP/882bLli1m9OjRpnv37qasrCyKPW8cuR5BtnayMVdjyNYYsrWZjdmS6xFkaycbczWGbI3xX7YxOYgePHiwueOOO1zzcnJyzNSpU6PUo+grKSkxkkxBQYExxpiamhqTkZFh8vLynGUqKipMamqqWbhwYbS62SRybRjZ2smGXI0h24aQrb1syJZcG0a2drIhV2PItiGxnm3Mnc5dWVmp9957T7m5ua75ubm52rhxY5R6FX2lpaWSpC5dukiSdu7cqeLiYtd2CgaDGjZsWExuJ3JtHNnaye+5SmTbGLK1l9+zJdfGka2d/J6rRLaNifVsY24QvWfPHlVXVys9Pd01Pz09XcXFxVHqVXQZYzR58mQNHTpU/fv3lyRnW/hlO5Frw8jWTjbkKpFtQ8jWXjZkS64NI1s72ZCrRLYN8UO2CW3+jh4FAgHXtDGm3rx4MWHCBL3//vt6880367X5bTv5rb+RRrZ2silXyZ99jhSytZdN2fqtv5FGtnayKVfJn32OFD9kG3NHotPS0tS+fft63yiUlJTU++YhHkycOFGrVq3S+vXr1bNnT2d+RkaGJPlmO5FrfWRrJ1tylci2LrK1ly3Zkmt9ZGsnW3KVyLYuv2Qbc4PoDh06aODAgcrPz3fNz8/P15AhQ6LUq7ZnjNGECRO0cuVKrVu3TllZWa72rKwsZWRkuLZTZWWlCgoKYnI7kevXyNZOtuUqkW0tsrWXbdmS69fI1k625SqRbS3fZRv5e5c1X+1t3pcsWWI+/PBDM2nSJJOUlGQKCwuj3bU2M3bsWJOammpef/118/nnnzvlwIEDzjJ5eXkmNTXVrFy50mzZssVcf/31MX0Lf3I9gmztZGOuxpCtMWRrMxuzJdcjyNZONuZqDNka479sY3IQbYwx8+fPN7179zYdOnQwZ555pnN783ghqcGydOlSZ5mamhozY8YMk5GRYYLBoDn//PPNli1botdpD+I9V2PI1la25moM2ZKtvWzNNt5zNYZsbWVrrsaQrd+yDfyn0wAAAAAA4Chi7ppoAAAAAABiFYNoAAAAAAA8YhANAAAAAIBHDKIBAAAAAPCIQTQAAAAAAB4xiAYAAAAAwCMG0QAAAAAAeMQgGgAAAAAAjxhEAwAAAADgEYNoAAAAAAA8YhANAAAAAIBH/x9OQ5A3XR85BgAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 1200x400 with 16 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "## Test Data"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1200x400 with 16 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot some digits.\n",
    "\n",
    "def display_sample_data(data, data_label, rows=2, cols=8):\n",
    "    fig, axes = plt.subplots(nrows=rows, ncols=cols, figsize=(1.5*cols, 2*rows))\n",
    "    for i, ax in enumerate(axes.flatten()):\n",
    "        image, label = data[i]          # returns PIL image with its labels\n",
    "        ax.set_title(f\"Label: {label}\")\n",
    "        ax.imshow(image.squeeze(0), cmap='gray')  # we get a 1x28x28 tensor -> remove first dimension\n",
    "    display_markdown(f\"## {data_label}\", raw=True)\n",
    "    plt.show()\n",
    "\n",
    "display_sample_data(mnist_training, 'Training Data')\n",
    "display_sample_data(mnist_val, 'Test Data')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2dc04abb-02a5-48d9-b605-d64c6d8a8353",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Lib\n",
    "def train(\n",
    "    model: nn.Module, \n",
    "    train_loader: torch.utils.data.DataLoader, \n",
    "    device: torch.device, \n",
    "    optimizer: torch.optim.Optimizer, \n",
    "    epoch: int,\n",
    "    bar: tqdm_notebook\n",
    "):\n",
    "    model.train()\n",
    "    losses = []\n",
    "    for batch_idx, (data, target) in enumerate(train_loader):\n",
    "        imgs, labels = data.to(device), target.to(device)\n",
    "        opt.zero_grad()\n",
    "        predictions = model(imgs)\n",
    "        # Compute the loss\n",
    "        loss = loss_fn(predictions, labels)\n",
    "        loss.backward()\n",
    "        losses.append(float(loss))\n",
    "        opt.step()\n",
    "        bar.update(batch_size)\n",
    "    return losses\n",
    "\n",
    "def test(\n",
    "    model: nn.Module,\n",
    "    test_loader: torch.utils.data.DataLoader,\n",
    "    device: torch.device,\n",
    "    bar: tqdm\n",
    "):\n",
    "    model.eval()\n",
    "    test_loss = 0\n",
    "    correct = 0\n",
    "    with torch.no_grad():\n",
    "        for data, target in test_loader:\n",
    "            data, target = data.to(device), target.to(device)\n",
    "            img_grid = torchvision.utils.make_grid(data)\n",
    "            output = model(data)\n",
    "            test_loss += loss_fn(output, target).sum().item()\n",
    "            pred = output.argmax(dim=1, keepdim=True)\n",
    "            correct += pred.eq(target.view_as(pred)).sum().item()\n",
    "            bar.update(test_batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f9a72193-d851-415e-88bd-82aef2796d2f",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": ""
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5ca4d911af2d403baf51fa7c431a87e1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "            00:00/?"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9a6b7a5e72fa4b149dca720f0879a40b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Train:   0%|          | 0/60000 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ab552c4c03394654ad62e1e913308bf2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Test :   0%|          | 0/10000 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Finished\n"
     ]
    }
   ],
   "source": [
    "torch.backends.cudnn.benchmark = True\n",
    "\n",
    "model = CNN().to(device)\n",
    "\n",
    "opt = torch.optim.Adadelta(params=model.parameters(), lr=learning_rate)\n",
    "\n",
    "losses = []\n",
    "test_losses = []\n",
    "\n",
    "# progress bars\n",
    "epoch_bar    = tqdm_notebook(position=1, total=num_epochs, bar_format=\"{desc} {bar} {elapsed}/{remaining}\")\n",
    "training_bar = tqdm_notebook(position=2, total=training_size, desc=\"Train\", mininterval=0.05)\n",
    "testing_bar  = tqdm_notebook(position=3, total=test_size,     desc=\"Test \")\n",
    "\n",
    "\n",
    "scheduler = StepLR(opt, step_size=step_size, gamma=gamma)\n",
    "for epoch in range(num_epochs):\n",
    "    # update and reset progress bars\n",
    "    epoch_bar.set_description_str(f'Epoch: {epoch}')\n",
    "    training_bar.reset()\n",
    "    testing_bar.reset()\n",
    "    \n",
    "    epoch_losses = train(model, train_loader, device, opt, epoch, bar=training_bar)\n",
    "    test(model, test_loader, device, bar=testing_bar)\n",
    "    losses.append(epoch_losses)\n",
    "    \n",
    "    epoch_bar.update(1)\n",
    "    scheduler.step()\n",
    "print('Training Finished')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "172b47d5-56ff-484b-b19b-c7c1c97603cb",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": ""
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "# Accuracy: 97.31%"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot Learning Curve\n",
    "plt.plot(list(flatten(losses)))\n",
    "\n",
    "# Determine the accuracy of our clasifier\n",
    "loader = torch.utils.data.DataLoader(mnist_val, batch_size=test_size, **loader_opts)\n",
    "data, target = next(iter(loader))\n",
    "images, labels = data.to(device), target.to(device)\n",
    "\n",
    "predictions = model(images).to(device)\n",
    "\n",
    "# For each row determine the column index with the maximum score. This is the\n",
    "# predicted class.\n",
    "predicted_classes = torch.argmax(predictions, dim=1)\n",
    "\n",
    "# Accuracy = number of correctly classified images divided by the total number\n",
    "# of classified images.\n",
    "accuracy = sum(predicted_classes.cpu().numpy() == labels.cpu().numpy()) / test_size\n",
    "display_markdown('# Accuracy: {:.2f}%'.format(100. * accuracy), raw=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "c3c0f19c-57ea-43a9-a9b6-ce6ea82b8066",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": ""
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model exported to demo/modes/\n"
     ]
    }
   ],
   "source": [
    "# export to onnx format\n",
    "def onnxExport(net: nn.Module, state_dict):\n",
    "    pytorch_model = net(inference=True)\n",
    "    pytorch_model.load_state_dict(state_dict)\n",
    "    pytorch_model.eval()\n",
    "    dummy_input = torch.zeros([280 * 280 * 4])\n",
    "    torch.onnx.export(pytorch_model, dummy_input, './demo/models/mnist_inference_model.onnx', verbose=True, opset_version=9)\n",
    "\n",
    "\n",
    "model_state = model.state_dict()\n",
    "onnxExport(CNN, model_state)\n",
    "print('model exported to demo/models/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b87cd038-d304-4b16-8e7d-dc74ea82e60d",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}