dlFall2022MiniProjCode.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "274f57d8-a974-4286-b08e-d0f40d49faaa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: torchvision in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (0.14.0)\n",
      "Requirement already satisfied: typing-extensions in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torchvision) (4.4.0)\n",
      "Requirement already satisfied: torch==1.13.0 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torchvision) (1.13.0)\n",
      "Requirement already satisfied: requests in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torchvision) (2.28.1)\n",
      "Requirement already satisfied: numpy in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torchvision) (1.23.4)\n",
      "Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torchvision) (9.3.0)\n",
      "Requirement already satisfied: nvidia-cublas-cu11==11.10.3.66 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torch==1.13.0->torchvision) (11.10.3.66)\n",
      "Requirement already satisfied: nvidia-cudnn-cu11==8.5.0.96 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torch==1.13.0->torchvision) (8.5.0.96)\n",
      "Requirement already satisfied: nvidia-cuda-runtime-cu11==11.7.99 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torch==1.13.0->torchvision) (11.7.99)\n",
      "Requirement already satisfied: nvidia-cuda-nvrtc-cu11==11.7.99 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from torch==1.13.0->torchvision) (11.7.99)\n",
      "Requirement already satisfied: setuptools in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from nvidia-cublas-cu11==11.10.3.66->torch==1.13.0->torchvision) (65.5.1)\n",
      "Requirement already satisfied: wheel in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from nvidia-cublas-cu11==11.10.3.66->torch==1.13.0->torchvision) (0.38.4)\n",
      "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from requests->torchvision) (1.26.12)\n",
      "Requirement already satisfied: idna<4,>=2.5 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from requests->torchvision) (3.4)\n",
      "Requirement already satisfied: charset-normalizer<3,>=2 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from requests->torchvision) (2.1.1)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from requests->torchvision) (2022.9.24)\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    }
   ],
   "source": [
    "pip install torchvision "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "1f9f6246-1b8b-414c-9f25-9d0f299c102f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: matplotlib in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (3.6.2)\n",
      "Requirement already satisfied: cycler>=0.10 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (0.11.0)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (1.4.4)\n",
      "Requirement already satisfied: pyparsing>=2.2.1 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (3.0.9)\n",
      "Requirement already satisfied: python-dateutil>=2.7 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (2.8.2)\n",
      "Requirement already satisfied: packaging>=20.0 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (21.3)\n",
      "Requirement already satisfied: numpy>=1.19 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (1.23.4)\n",
      "Requirement already satisfied: contourpy>=1.0.1 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (1.0.6)\n",
      "Requirement already satisfied: fonttools>=4.22.0 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (4.38.0)\n",
      "Requirement already satisfied: pillow>=6.2.0 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from matplotlib) (9.3.0)\n",
      "Requirement already satisfied: six>=1.5 in /home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages (from python-dateutil>=2.7->matplotlib) (1.16.0)\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    }
   ],
   "source": [
    "pip install matplotlib"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e271ed87-ea9e-4eeb-8639-4ce78e242e67",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torchvision\n",
    "from torchvision import datasets, models, transforms\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "31fab4a6-901f-47d5-99d8-0902c2154728",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files already downloaded and verified\n",
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "source": [
    "#data transforms\n",
    "dset_transform = transforms.Compose([\n",
    "    transforms.RandomCrop(32, padding=4), #data augmentation\n",
    "    transforms.RandomHorizontalFlip(),    #data augmentation\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize([0.4914, 0.4822, 0.4465],\n",
    "                         [0.247, 0.243, 0.261])])\n",
    "\n",
    "batch_size = 64\n",
    "#resnet = 'resnet[3,3,3]'\n",
    "trainset = torchvision.datasets.CIFAR10(root='./data', train=True,\n",
    "                                        download=True, transform=dset_transform)\n",
    "\n",
    "testset = torchvision.datasets.CIFAR10(root='./data', train=False,\n",
    "                                       download=True, transform=dset_transform)\n",
    "\n",
    "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size,\n",
    "                                          shuffle=True)\n",
    "\n",
    "testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size,\n",
    "                                         shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "146ba0c3-9b09-4d33-be06-ea05a3d011be",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([3, 32, 32])"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature, label = trainset[3]\n",
    "feature.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "6c087283-1fec-4e84-8533-d8d6943a6d69",
   "metadata": {},
   "outputs": [],
   "source": [
    "class BasicBlock(nn.Module):  # two conv as a res block\n",
    "\n",
    "    def __init__(self, in_planes, planes, stride=1):\n",
    "        super(BasicBlock, self).__init__()\n",
    "        self.conv1 = nn.Conv2d(\n",
    "            in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)  # conv1 3x3 filter\n",
    "        self.bn1 = nn.BatchNorm2d(planes)  # batchnorm\n",
    "        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3,  # conv2\n",
    "                               stride=1, padding=1, bias=False)\n",
    "        self.bn2 = nn.BatchNorm2d(planes)  # batchnorm\n",
    "\n",
    "        self.shortcut = nn.Sequential()  # skip connection\n",
    "        if stride != 1 or in_planes != planes:\n",
    "            self.shortcut = nn.Sequential(\n",
    "                nn.Conv2d(in_planes, planes,\n",
    "                          kernel_size=1, stride=stride, bias=False),\n",
    "                nn.BatchNorm2d(planes)\n",
    "            )\n",
    "\n",
    "    def forward(self, x):\n",
    "        out = F.relu(self.bn1(self.conv1(x)))\n",
    "        out = self.bn2(self.conv2(out))\n",
    "        out += self.shortcut(x)\n",
    "        out = F.relu(out)\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "8b133a89-b2cb-46f3-be19-5d6b67f0071c",
   "metadata": {},
   "outputs": [],
   "source": [
    "class ResNet(nn.Module):\n",
    "    def __init__(self, block, num_blocks, num_classes=10):\n",
    "        super(ResNet, self).__init__()\n",
    "        self.in_planes = 68 \n",
    "        self.conv1 = nn.Conv2d(3, 68, kernel_size=2,\n",
    "                               stride=1, padding=1, bias=False)\n",
    "        self.bn1 = nn.BatchNorm2d(68)\n",
    "        self.layer1 = self._make_layer(block, 68, num_blocks[0], stride=1)\n",
    "        self.layer2 = self._make_layer(block, 136, num_blocks[1], stride=2)\n",
    "        self.layer3 = self._make_layer(block, 272, num_blocks[2], stride=2)\n",
    "        #self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)\n",
    "        self.linear = nn.Linear(272, num_classes)\n",
    "        self.dropout = nn.Dropout(0)\n",
    "\n",
    "\n",
    "    def _make_layer(self, block, planes, num_blocks, stride):\n",
    "        strides = [stride] + [1] * (num_blocks - 1)\n",
    "        layers = []\n",
    "        for stride in strides:\n",
    "            layers.append(block(self.in_planes, planes, stride))\n",
    "            self.in_planes = planes\n",
    "        return nn.Sequential(*layers)\n",
    "\n",
    "    def forward(self, x):\n",
    "        out = F.relu(self.bn1(self.conv1(x)))\n",
    "        out = self.dropout(out)\n",
    "        out = self.layer1(out)\n",
    "        out = self.dropout(out)\n",
    "        out = self.layer2(out)\n",
    "        out = self.dropout(out)\n",
    "        out = self.layer3(out)\n",
    "        out = self.dropout(out)\n",
    "        #out = self.layer4(out)\n",
    "        #out = self.dropout(out)\n",
    "        out = F.avg_pool2d(out,8)\n",
    "        out = out.view(out.size(0), -1)\n",
    "        #print(out.shape)\n",
    "        #out = out.view(-1, 512*2*2)\n",
    "        out = self.dropout(out)\n",
    "        out = self.linear(out)\n",
    "        return out\n",
    "\n",
    "\n",
    "def project1_model():\n",
    "    return ResNet(BasicBlock, [3,3,3])  # modify"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "dade2bbe-1d6c-47ab-84e1-860168d3cfc8",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.optim as optim\n",
    "device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
    "model = ResNet(BasicBlock, [3,3,3]).to(device)\n",
    "Loss = nn.CrossEntropyLoss()\n",
    "optimizer = optim.Adam(model.parameters(), lr = 0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "6f73e849-0729-4639-a634-d3cd18f50e4d",
   "metadata": {},
   "outputs": [],
   "source": [
    "def count_parameters(model):\n",
    "    return sum(p.numel() for p in model.parameters() if p.requires_grad)\n",
    "    # torch.numel() returns number of elements in a tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "b523e3a1-d640-404b-99a8-f4373de4e724",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4883906\n"
     ]
    }
   ],
   "source": [
    "print(count_parameters(model))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b79a2363-7cfd-4838-aaaf-4b3219d8fd49",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 0, Train loss 1.4546997434342914, Test loss 1.1277937232309085\n",
      "Accuracy  0.5915\n",
      "Epoch 1, Train loss 0.9697476545410693, Test loss 0.8863740271064127\n",
      "Accuracy  0.6847\n",
      "Epoch 2, Train loss 0.7760512032700927, Test loss 0.7271787848821871\n",
      "Accuracy  0.7502\n",
      "Epoch 3, Train loss 0.6384995113438963, Test loss 0.5981914877511894\n",
      "Accuracy  0.7928\n",
      "Epoch 4, Train loss 0.54498966522229, Test loss 0.5569150391847465\n",
      "Accuracy  0.8087\n",
      "Epoch 5, Train loss 0.47835423820235234, Test loss 0.5231712340928947\n",
      "Accuracy  0.8253\n",
      "Epoch 6, Train loss 0.4292399855449681, Test loss 0.4844585899144981\n",
      "Accuracy  0.8367\n",
      "Epoch 7, Train loss 0.39107344328137617, Test loss 0.4197577597324256\n",
      "Accuracy  0.8553\n",
      "Epoch 8, Train loss 0.3545526782874866, Test loss 0.423999510088544\n",
      "Accuracy  0.855\n",
      "Epoch 9, Train loss 0.3203332851000149, Test loss 0.39340270343859485\n",
      "Accuracy  0.8673\n",
      "Epoch 10, Train loss 0.29412450142147595, Test loss 0.3751935107502968\n",
      "Accuracy  0.874\n",
      "Epoch 11, Train loss 0.27227082606071557, Test loss 0.37342208339150545\n",
      "Accuracy  0.8734\n",
      "Epoch 12, Train loss 0.24639611554991864, Test loss 0.35815647842398113\n",
      "Accuracy  0.882\n",
      "Epoch 13, Train loss 0.22166134284623443, Test loss 0.3485242694046847\n",
      "Accuracy  0.886\n",
      "Epoch 14, Train loss 0.20960415378594033, Test loss 0.3600012243363508\n",
      "Accuracy  0.8868\n",
      "Epoch 15, Train loss 0.19652933862222277, Test loss 0.34167638174287834\n",
      "Accuracy  0.8874\n",
      "Epoch 16, Train loss 0.17972581791203193, Test loss 0.3312068563547863\n",
      "Accuracy  0.8933\n",
      "Epoch 17, Train loss 0.16496020144261325, Test loss 0.3519020028364886\n",
      "Accuracy  0.8891\n",
      "Epoch 18, Train loss 0.15615162206932787, Test loss 0.3155314420723611\n",
      "Accuracy  0.8997\n",
      "Epoch 19, Train loss 0.13906862964386793, Test loss 0.34770531579852104\n",
      "Accuracy  0.8945\n",
      "Epoch 20, Train loss 0.1342416286439923, Test loss 0.33369440457243826\n",
      "Accuracy  0.898\n",
      "Epoch 21, Train loss 0.12356541342818943, Test loss 0.30632697579682255\n",
      "Accuracy  0.9024\n",
      "Epoch 22, Train loss 0.113573396628451, Test loss 0.34119462516087634\n",
      "Accuracy  0.9018\n",
      "Epoch 23, Train loss 0.10792322351914042, Test loss 0.35157773653223257\n",
      "Accuracy  0.9012\n",
      "Epoch 24, Train loss 0.10031789149362547, Test loss 0.3669305046557621\n",
      "Accuracy  0.902\n",
      "Epoch 25, Train loss 0.09484890083808575, Test loss 0.32728640569053635\n",
      "Accuracy  0.9062\n",
      "Epoch 26, Train loss 0.08659379779810177, Test loss 0.36077782516456713\n",
      "Accuracy  0.8976\n",
      "Epoch 27, Train loss 0.08563320280607704, Test loss 0.32960031561221287\n",
      "Accuracy  0.9103\n",
      "Epoch 28, Train loss 0.08039849836979528, Test loss 0.34549110128906124\n",
      "Accuracy  0.9052\n",
      "Epoch 29, Train loss 0.07332575821574977, Test loss 0.346287755449866\n",
      "Accuracy  0.9066\n",
      "Epoch 30, Train loss 0.07113846684770321, Test loss 0.34077132490884726\n",
      "Accuracy  0.9104\n",
      "Epoch 31, Train loss 0.06924847725336976, Test loss 0.3341687246919817\n",
      "Accuracy  0.9096\n",
      "Epoch 32, Train loss 0.06155188848196631, Test loss 0.3531129117224627\n",
      "Accuracy  0.9096\n",
      "Epoch 33, Train loss 0.06222852338474635, Test loss 0.3438497094354432\n",
      "Accuracy  0.9091\n",
      "Epoch 34, Train loss 0.057882394889653056, Test loss 0.35848547670112296\n",
      "Accuracy  0.9107\n",
      "Epoch 35, Train loss 0.05868622429205624, Test loss 0.34823987379100674\n",
      "Accuracy  0.9074\n",
      "Epoch 36, Train loss 0.05020958090604752, Test loss 0.37410910439434325\n",
      "Accuracy  0.9089\n",
      "Epoch 37, Train loss 0.05538849828555189, Test loss 0.35606340373492545\n",
      "Accuracy  0.908\n",
      "Epoch 38, Train loss 0.052268875683622096, Test loss 0.37230607919442427\n",
      "Accuracy  0.906\n",
      "Epoch 39, Train loss 0.04750866046421887, Test loss 0.35894250838904623\n",
      "Accuracy  0.9124\n",
      "Epoch 40, Train loss 0.044815144265446305, Test loss 0.3579916575342227\n",
      "Accuracy  0.9139\n",
      "Epoch 41, Train loss 0.04637625378512187, Test loss 0.35083986409483064\n",
      "Accuracy  0.9125\n",
      "Epoch 42, Train loss 0.042133417490826885, Test loss 0.3563885664104656\n",
      "Accuracy  0.9146\n",
      "Epoch 43, Train loss 0.0422901416275873, Test loss 0.382194654149994\n",
      "Accuracy  0.9119\n",
      "Epoch 44, Train loss 0.040180608222577865, Test loss 0.368549346330629\n",
      "Accuracy  0.9187\n",
      "Epoch 45, Train loss 0.041121426578986285, Test loss 0.3583471791427227\n",
      "Accuracy  0.9127\n",
      "Epoch 46, Train loss 0.03876501450817103, Test loss 0.3879126602202464\n",
      "Accuracy  0.9115\n",
      "Epoch 47, Train loss 0.03829795777317746, Test loss 0.374297310331255\n",
      "Accuracy  0.914\n",
      "Epoch 48, Train loss 0.03829768998112947, Test loss 0.4080303444224558\n",
      "Accuracy  0.9083\n",
      "Epoch 49, Train loss 0.03632283960806999, Test loss 0.36441394473147243\n",
      "Accuracy  0.9161\n",
      "Epoch 50, Train loss 0.03499814256978974, Test loss 0.36564434653112465\n",
      "Accuracy  0.9167\n",
      "Epoch 51, Train loss 0.0334193563650233, Test loss 0.35558005234902834\n",
      "Accuracy  0.9185\n",
      "Epoch 52, Train loss 0.03414948797255368, Test loss 0.3847894696340819\n",
      "Accuracy  0.9132\n",
      "Epoch 53, Train loss 0.03271004744037829, Test loss 0.34750950084939886\n",
      "Accuracy  0.9165\n",
      "Epoch 54, Train loss 0.0317959298272236, Test loss 0.3828223616857627\n",
      "Accuracy  0.9128\n",
      "Epoch 55, Train loss 0.03100702885831313, Test loss 0.42054854674193604\n",
      "Accuracy  0.9057\n",
      "Epoch 56, Train loss 0.03412327086651975, Test loss 0.362624128627929\n",
      "Accuracy  0.9164\n",
      "Epoch 57, Train loss 0.02900974402207729, Test loss 0.3568342022455422\n",
      "Accuracy  0.9188\n",
      "Epoch 58, Train loss 0.026552385495751715, Test loss 0.35899930882748143\n",
      "Accuracy  0.9175\n",
      "Epoch 59, Train loss 0.0300673645565032, Test loss 0.3979747644892544\n",
      "Accuracy  0.9143\n",
      "Epoch 60, Train loss 0.028513841084590184, Test loss 0.3593863974307563\n",
      "Accuracy  0.9184\n",
      "Epoch 61, Train loss 0.029006071874591678, Test loss 0.3714478063830145\n",
      "Accuracy  0.9168\n",
      "Epoch 62, Train loss 0.025257144392656503, Test loss 0.38372319533376936\n",
      "Accuracy  0.913\n",
      "Epoch 63, Train loss 0.026136633668467343, Test loss 0.3773359424037159\n",
      "Accuracy  0.9184\n",
      "Epoch 64, Train loss 0.02496292698695777, Test loss 0.41882708716461325\n",
      "Accuracy  0.915\n",
      "Epoch 65, Train loss 0.025590173879657965, Test loss 0.40722733673159106\n",
      "Accuracy  0.9117\n",
      "Epoch 66, Train loss 0.026369149664799613, Test loss 0.3885484633457129\n",
      "Accuracy  0.9114\n",
      "Epoch 67, Train loss 0.024464335840488988, Test loss 0.39814286512933716\n",
      "Accuracy  0.916\n",
      "Epoch 68, Train loss 0.025104162282757924, Test loss 0.4081534178583485\n",
      "Accuracy  0.9169\n",
      "Epoch 69, Train loss 0.02170472623111726, Test loss 0.40113102163004266\n",
      "Accuracy  0.917\n",
      "Epoch 70, Train loss 0.022059043625175693, Test loss 0.38864828087982667\n",
      "Accuracy  0.9183\n",
      "Epoch 71, Train loss 0.022624730346239824, Test loss 0.3687281900435496\n",
      "Accuracy  0.9228\n",
      "Epoch 72, Train loss 0.022205407016637365, Test loss 0.403979852071898\n",
      "Accuracy  0.9161\n",
      "Epoch 73, Train loss 0.023683504864644638, Test loss 0.36258229478054743\n",
      "Accuracy  0.9198\n",
      "Epoch 74, Train loss 0.020342432223894213, Test loss 0.37310982795705083\n",
      "Accuracy  0.9173\n",
      "Epoch 75, Train loss 0.022334952593029714, Test loss 0.3758790003978143\n",
      "Accuracy  0.92\n",
      "Epoch 76, Train loss 0.020930496541152716, Test loss 0.3886154273133369\n",
      "Accuracy  0.9199\n",
      "Epoch 77, Train loss 0.019544624414684547, Test loss 0.4034997390784845\n",
      "Accuracy  0.9168\n",
      "Epoch 78, Train loss 0.021779937078467915, Test loss 0.3689366091208853\n",
      "Accuracy  0.9196\n",
      "Epoch 79, Train loss 0.019709250396404886, Test loss 0.37988219174087806\n",
      "Accuracy  0.9182\n",
      "Epoch 80, Train loss 0.020328824076477525, Test loss 0.3792325081956235\n",
      "Accuracy  0.9186\n",
      "Epoch 81, Train loss 0.020345134655902947, Test loss 0.39895089974354025\n",
      "Accuracy  0.9177\n",
      "Epoch 82, Train loss 0.02227895940606565, Test loss 0.37300251585662747\n",
      "Accuracy  0.9195\n",
      "Epoch 83, Train loss 0.018883626226203383, Test loss 0.3838340991123277\n",
      "Accuracy  0.9188\n",
      "Epoch 84, Train loss 0.019467306676307854, Test loss 0.3762446246139563\n",
      "Accuracy  0.9214\n",
      "Epoch 85, Train loss 0.018437309792004627, Test loss 0.3845292416869835\n",
      "Accuracy  0.9179\n",
      "Epoch 86, Train loss 0.016305900678974087, Test loss 0.38897298364217875\n",
      "Accuracy  0.9186\n",
      "Epoch 87, Train loss 0.01851665517656178, Test loss 0.3845994704563147\n",
      "Accuracy  0.917\n",
      "Epoch 88, Train loss 0.017462003732082936, Test loss 0.3891040481104972\n",
      "Accuracy  0.9206\n",
      "Epoch 89, Train loss 0.016939077326139765, Test loss 0.40202041145912404\n",
      "Accuracy  0.9181\n",
      "Epoch 90, Train loss 0.018227668657048193, Test loss 0.3768398527316986\n",
      "Accuracy  0.9217\n",
      "Epoch 91, Train loss 0.01664710865680506, Test loss 0.39032569079167523\n",
      "Accuracy  0.9194\n",
      "Epoch 92, Train loss 0.017037402525565157, Test loss 0.42033092089139734\n",
      "Accuracy  0.92\n",
      "Epoch 93, Train loss 0.016578283186953674, Test loss 0.41746011459428795\n",
      "Accuracy  0.9204\n",
      "Epoch 94, Train loss 0.016226660895511804, Test loss 0.43293898473167497\n",
      "Accuracy  0.9194\n",
      "Epoch 95, Train loss 0.01758724292485474, Test loss 0.3722385814662572\n",
      "Accuracy  0.9236\n",
      "Epoch 96, Train loss 0.014720729331849976, Test loss 0.3953746450207772\n",
      "Accuracy  0.9229\n",
      "Epoch 97, Train loss 0.017826666026823538, Test loss 0.3997957315058655\n",
      "Accuracy  0.9195\n",
      "Epoch 98, Train loss 0.016284743882765945, Test loss 0.38613011518339063\n",
      "Accuracy  0.921\n",
      "Epoch 99, Train loss 0.012886590714759715, Test loss 0.38363075931408225\n",
      "Accuracy  0.9249\n",
      "Epoch 100, Train loss 0.0165839371053069, Test loss 0.41512198486383173\n",
      "Accuracy  0.9193\n",
      "Epoch 101, Train loss 0.01557143493971202, Test loss 0.362912401367145\n",
      "Accuracy  0.9248\n",
      "Epoch 102, Train loss 0.014685667394571372, Test loss 0.4068906919639202\n",
      "Accuracy  0.9204\n",
      "Epoch 103, Train loss 0.014618198925047851, Test loss 0.3933460212306707\n",
      "Accuracy  0.9265\n",
      "Epoch 104, Train loss 0.012880037361437202, Test loss 0.4187648746772272\n",
      "Accuracy  0.9239\n",
      "Epoch 105, Train loss 0.016982728596180503, Test loss 0.3906150194964591\n",
      "Accuracy  0.9221\n",
      "Epoch 106, Train loss 0.013014358471852163, Test loss 0.42532962699462273\n",
      "Accuracy  0.9177\n",
      "Epoch 107, Train loss 0.017314754553101473, Test loss 0.3859203820276982\n",
      "Accuracy  0.9235\n",
      "Epoch 108, Train loss 0.014620746828468397, Test loss 0.41097403443448105\n",
      "Accuracy  0.919\n",
      "Epoch 109, Train loss 0.013606406515575839, Test loss 0.39138931054977855\n",
      "Accuracy  0.9215\n",
      "Epoch 110, Train loss 0.014128545098404353, Test loss 0.4147982294819537\n",
      "Accuracy  0.9214\n",
      "Epoch 111, Train loss 0.012370657766183294, Test loss 0.4037833616944255\n",
      "Accuracy  0.9219\n",
      "Epoch 112, Train loss 0.015312756470929811, Test loss 0.38156844930616535\n",
      "Accuracy  0.9251\n",
      "Epoch 113, Train loss 0.012579237168341226, Test loss 0.42145462949896695\n",
      "Accuracy  0.9235\n",
      "Epoch 114, Train loss 0.015657629770660784, Test loss 0.39862030127328957\n",
      "Accuracy  0.9231\n",
      "Epoch 115, Train loss 0.013259846521961658, Test loss 0.38726592119812825\n",
      "Accuracy  0.9255\n",
      "Epoch 116, Train loss 0.012202655236025674, Test loss 0.4132138678839632\n",
      "Accuracy  0.9231\n",
      "Epoch 117, Train loss 0.01262994104243383, Test loss 0.39821696078559016\n",
      "Accuracy  0.926\n",
      "Epoch 118, Train loss 0.012369528153861808, Test loss 0.4087851764396972\n",
      "Accuracy  0.9232\n",
      "Epoch 119, Train loss 0.01492939594280779, Test loss 0.36296547370351806\n",
      "Accuracy  0.9264\n",
      "Epoch 120, Train loss 0.012492973777918678, Test loss 0.4268034242663034\n",
      "Accuracy  0.9238\n",
      "Epoch 121, Train loss 0.014255425313545447, Test loss 0.383100010276695\n",
      "Accuracy  0.9246\n",
      "Epoch 122, Train loss 0.00995939327749232, Test loss 0.4331503340286339\n",
      "Accuracy  0.9205\n",
      "Epoch 123, Train loss 0.014450254755511406, Test loss 0.4035408735856604\n",
      "Accuracy  0.924\n",
      "Epoch 124, Train loss 0.01234464936055178, Test loss 0.41201409984641013\n",
      "Accuracy  0.9209\n",
      "Epoch 125, Train loss 0.014302635659430478, Test loss 0.40173006523400545\n",
      "Accuracy  0.9243\n",
      "Epoch 126, Train loss 0.012042824860662018, Test loss 0.42573320679959786\n",
      "Accuracy  0.9234\n",
      "Epoch 127, Train loss 0.010586018887362578, Test loss 0.42642525990440205\n",
      "Accuracy  0.9193\n",
      "Epoch 128, Train loss 0.011108020126444814, Test loss 0.4098840804094342\n",
      "Accuracy  0.9245\n",
      "Epoch 129, Train loss 0.012532281784556674, Test loss 0.4086751769101069\n",
      "Accuracy  0.9254\n",
      "Epoch 130, Train loss 0.010435966919958958, Test loss 0.40061399231480943\n",
      "Accuracy  0.9233\n",
      "Epoch 131, Train loss 0.012731117007171304, Test loss 0.40565394147471256\n",
      "Accuracy  0.9269\n",
      "Epoch 132, Train loss 0.010227312539546287, Test loss 0.384584048454454\n",
      "Accuracy  0.9283\n",
      "Epoch 133, Train loss 0.011468306725257435, Test loss 0.3917959121287249\n",
      "Accuracy  0.9269\n",
      "Epoch 134, Train loss 0.013904599356653453, Test loss 0.43018382749739725\n",
      "Accuracy  0.9218\n",
      "Epoch 135, Train loss 0.008656549417674702, Test loss 0.43161333523761886\n",
      "Accuracy  0.9217\n",
      "Epoch 136, Train loss 0.01250969386084827, Test loss 0.3963224684736531\n",
      "Accuracy  0.9256\n",
      "Epoch 137, Train loss 0.011458023308243071, Test loss 0.42261344799475303\n",
      "Accuracy  0.9195\n",
      "Epoch 138, Train loss 0.010996644936175566, Test loss 0.4214121656860136\n",
      "Accuracy  0.9194\n",
      "Epoch 139, Train loss 0.011087461958666032, Test loss 0.4213348068534189\n",
      "Accuracy  0.9258\n"
     ]
    }
   ],
   "source": [
    "epochs = 140\n",
    "train_save_loss = []\n",
    "test_save_loss = []\n",
    "train_save_acc = []\n",
    "test_save_acc = []\n",
    "for epoch in range(epochs):\n",
    "    train_loss = 0.0\n",
    "    train_corrects = 0.0\n",
    "    test_loss = 0.0\n",
    "    test_corrects = 0.0\n",
    "\n",
    "    for i, data in enumerate(trainloader):\n",
    "        inputs, labels = data\n",
    "        inputs = inputs.to(device)\n",
    "        labels = labels.to(device)\n",
    "        optimizer.zero_grad()\n",
    "        outputs = model(inputs)\n",
    "        fit = Loss(outputs, labels)\n",
    "        fit.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        train_loss += fit.item()\n",
    "        train_corrects += (torch.eq(torch.max(outputs, 1)[1], labels).sum()).data.cpu().numpy()\n",
    "    for i, data in enumerate(testloader):\n",
    "        with torch.no_grad():\n",
    "            inputs, labels = data\n",
    "            inputs = inputs.to(device)\n",
    "            labels = labels.to(device)\n",
    "            outputs = model(inputs)\n",
    "            fit = Loss(outputs, labels)\n",
    "            test_loss += fit.item()\n",
    "            test_corrects += (torch.eq(torch.max(outputs, 1)[1], labels).sum()).data.cpu().numpy()\n",
    "\n",
    "    train_loss = train_loss / len(trainloader)\n",
    "    test_loss = test_loss / len(testloader)\n",
    "    train_save_loss.append(train_loss)\n",
    "    test_save_loss.append(test_loss)\n",
    "    train_corrects = train_corrects / len(trainloader.dataset)\n",
    "    train_save_acc.append(train_corrects)\n",
    "    test_corrects = test_corrects / len(testloader.dataset)\n",
    "    test_save_acc.append(test_corrects)\n",
    "    print('Epoch %s, Train loss %s, Test loss %s' % (epoch, train_loss, test_loss))\n",
    "    print(\"Accuracy \", test_corrects)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3688c47c-9475-4b40-bdcd-32ba68ab0e68",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "7e9408fa-05a0-4d20-8d41-af6bd4d6a190",
   "metadata": {},
   "outputs": [],
   "source": [
    "#A = test_save_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "745b8e4a-3878-4388-a0c6-2ff9d879528d",
   "metadata": {},
   "outputs": [],
   "source": [
    "#NA = test_save_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "7f4ffdfa-1209-4df8-83c4-304f035d9f4a",
   "metadata": {},
   "outputs": [],
   "source": [
    "#p3 = test_save_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "34d35eff-582a-4122-b743-85ced0fc33eb",
   "metadata": {},
   "outputs": [],
   "source": [
    "#p4 = test_save_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "f0854109-3d27-405c-95b5-2f0b0fe6dc7c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjcAAAGwCAYAAABVdURTAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAACMH0lEQVR4nOzdd3xUVdrA8d/01EkvJARC7x1EwIKCICoKoiKyUmyrYkGsrIriqqwd18aKr7iuDSuiiIgUsSA9FOmQEEoqKZM+7b5/nGRCSIAEJhkSni+f+SS5c+fe596Euc+c85xzdZqmaQghhBBCNBF6XwcghBBCCOFNktwIIYQQokmR5EYIIYQQTYokN0IIIYRoUiS5EUIIIUSTIsmNEEIIIZoUSW6EEEII0aQYfR1AQ3O73Rw5coTg4GB0Op2vwxFCCCFELWiaRkFBAXFxcej1J2+bOeeSmyNHjpCQkODrMIQQQghxGg4ePEjz5s1Pus45l9wEBwcD6uRYrVYfRyOEEEKI2rDZbCQkJHiu4ydzziU3FV1RVqtVkhshhBCikalNSYkUFAshhBCiSZHkRgghhBBNiiQ3QgghhGhSJLkRQgghRJMiyY0QQgghmhRJboQQQgjRpEhyI4QQQogmRZIbIYQQQjQpktwIIYQQokmR5EYIIYQQTYokN0IIIYRoUiS5EUIIIUSTcs7dOLO+OFwOjpYeRdM0mgU183U4QgghxDlLWm68ZNvRbVz25WXc9tNtvg5FCCGEOKdJcuMlZr0ZALvb7uNIhBBCiHObJDdeYjKYALC7JLkRQgghfEmSGy+xGCyAJDdCCCGEr0ly4yWebilJboQQQgifkuTGSzzdUm47mqb5OBohhBDi3CXJjZdUdEsBONwOH0YihBBCnNskufESs8Hs+V66poQQQgjfkeTGS0x6k+f7MleZDyMRQgghzm2S3HiJXqf3JDjSLSWEEEL4jk+Tm1WrVjFy5Eji4uLQ6XQsWLCg1q/9/fffMRqN9OzZs97iq6uKrinplhJCCCF8x6fJTVFRET169OCtt96q0+vy8vKYMGECQ4YMqafITk9FUbF0SwkhhBC+49MbZ44YMYIRI0bU+XV33nknN910EwaDoU6tPfWtoltKbsEghBBC+E6jq7mZN28e+/fv56mnnqrV+mVlZdhstiqP+lLRLeVwSc2NEEII4SuNKrnZs2cPjz32GB999BFGY+0anWbNmkVISIjnkZCQUG/xSbeUEEII4XuNJrlxuVzcdNNNzJw5k/bt29f6ddOnTyc/P9/zOHjwYL3F6OmWkoJiIYQQwmd8WnNTFwUFBaxfv55NmzZxzz33AOB2u9E0DaPRyE8//cSll15a7XUWiwWLxVJteX2Q0VJCCCGE7zWa5MZqtbJ169Yqy95++22WL1/Ol19+SatWrXwUWSXPncGloFgIIYTwGZ8mN4WFhezdu9fzc3JyMklJSYSHh9OiRQumT5/O4cOH+fDDD9Hr9XTt2rXK66Ojo/Hz86u23Fc8N8+UlhshhBDCZ3ya3Kxfv55LLrnE8/O0adMAmDhxIh988AFpaWmkpqb6Krw6s+iloFgIIYTwNZ8mN4MHD0bTtBM+/8EHH5z09U8//TRPP/20d4M6A56h4HL7BSGEEMJnGs1oqcZACoqFEEII35PkxosqkhvplhJCCCF8R5IbLzLrpeVGCCGE8DVJbrxIam6EEEII35PkxoukW0oIIYTwPUluvEi6pYQQQgjfk+TGi2S0lBBCCOF7ktx4kSe5kdsvCCGEED4jyY0XScuNEEII4XuS3HiR58aZktwIIYQQPiPJjRd5CoqlW0oIIYTwGUluvEjuCi6EEEL4niQ3XiTdUkIIIYTvSXLjRTLPjRBCCOF7ktx4kcxQLIQQQvieJDdeJPPcCCGEEL4nyY0XVXRLOVxy40whhBDCVyS58SLplhJCCCF8T5IbL5JuKSGEEML3JLnxoorkxul24tbcPo5GCCGEODdJcuNFFfPcgAwHF0IIIXxFkhsvqigoBumaEkIIIXxFkhsvMuqNnu+l5UYIIYTwDUluvEin08ktGIQQQggfk+TGy+QWDEIIIYRvSXLjZRV3Bpe5boQQQgjfkOTGyyq6pRxumaVYCCGE8AVJbrzMM5GfdEsJIYQQPiHJjZfJLRiEEEII35Lkxss8N8+UbikhhBDCJyS58TLplhJCCCF8S5IbL5NuKSGEEMK3JLnxMpnnRgghhPAtSW68TLqlhBBCCN+S5MbLPMmN3DhTCCGE8AmfJjerVq1i5MiRxMXFodPpWLBgwUnX//rrr7nsssuIiorCarUyYMAAlixZ0jDB1pJ0SwkhhBC+5dPkpqioiB49evDWW2/Vav1Vq1Zx2WWX8cMPP7BhwwYuueQSRo4cyaZNm+o50tqTbikhhBDCt4y+3PmIESMYMWJErdefPXt2lZ+ff/55vv32W7777jt69erl5ehOj+eu4NItJYQQQviET5ObM+V2uykoKCA8PPyE65SVlVFWVjks22az1WtMFTfOlJYbIYQQwjcadUHxyy+/TGFhITfccMMJ15k1axYhISGeR0JCQr3GVNFyI/PcCCGEEL7RaJObTz75hJkzZ/L5558THR19wvWmT59Ofn6+53Hw4MF6jUtuvyCEEEL4VqPslvrss8+47bbb+OKLLxg6dOhJ17VYLFgslgaKTGYoFkIIIXyt0bXcfPrpp0yePJlPP/2UK6+80tfhVCOjpYQQQgjf8mnLTWFhIXv37vX8nJycTFJSEuHh4bRo0YLp06dz+PBhPvzwQ0B1RU2cOJHXX3+d/v37k56eDoC/vz8hISE+OYbjebqlXNItJYQQQviCT1tu1q9fT69evTzDuKdNm0avXr2YMWMGAGlpaaSmpnrWf/fdd3E6nUyZMoVmzZp5Hvfff79P4q+JdEsJIYQQvuXTlpvBgwejadoJn//ggw+q/Lxy5cr6DcgL5PYLQgghhG81upqbs510SwkhhBC+JcmNl8k8N0IIIYRvSXLjZZ4ZiqVbSgghhPAJSW68TIaCCyGEEL4lyY2XeW6cKcmNEEII4ROS3HhZRUGxdEsJIYQQviHJjZfJXcGFEEII35LkxsukW0oIIYTwrUZ548yzkqMU8g9hLssDwKW5cLqdGPVyioUQQoiGJFdebzm8Hj64EnNEG7CqRXaXXZIbIYQQooFJt5S3WFRGYy4r9CxyuGWWYiGEEKKhSXLjLX4quTGW2tDr1GmVuhshhBCi4Uly4y3lLTc4S7Do5c7gQgghhK9IcuMtlmDPtya93IJBCCGE8BVJbrzFYAJTAADm8iJi6ZYSQgghGp4kN95U3jVl0UlyI4QQQviKJDfeVF5UbJKCYiGEEMJnJLnxpvK6G7NOB0hyI4QQQviCJDfeVNEtVX5apaBYCCGEaHiS3HhTRbeUpn6UlhshhBCi4Uly400VsxRrKruReW6EEEKIhifJjTf5hQBg0dyA3H5BCCGE8AVJbrypoqDYrZIbabkRQgghGp4kN95U0S3ldgFScyOEEEL4giQ33uRXNbmRbikhhBCi4Uly400VLTculdRIt5QQQgjR8CS58aaKlhun6o6SbikhhBCi4Uly400VBcUuSW6EEEIIX5Hkxpssaii42aG6oyS5EUIIIRqeJDfe5OmWKk9u5PYLQgghRIOT5MabZIZiIYQQwuckufEmkx8YzFjKkxuHS4aCCyGEEA1NkhtvswR7Wm6kW0oIIYRoeJLceJvFikm6pYQQQgifkeTG2/ys0i0lhBBC+JAkN95msVZ2S8lQcCGEEKLB+TS5WbVqFSNHjiQuLg6dTseCBQtO+ZqVK1fSu3dvLBYLbdu25YMPPqj3OOvELwSTym0oc0u3lBBCCNHQfJrcFBUV0aNHD956661arZ+cnMyVV17JJZdcQlJSElOnTuW2225jyZIl9RxpHViCpVtKCCGE8CGjL3c+YsQIRowYUev158yZQ6tWrXjllVcA6NSpE7/99huvvfYaw4cPr68w60a6pYQQQgifalQ1N6tXr2bo0KFVlg0fPpzVq1ef8DVlZWXYbLYqj3rlZ8WMjJYSQgghfKVRJTfp6enExMRUWRYTE4PNZqOkpKTG18yaNYuQkBDPIyEhoX6DPLblRua5EUIIIRpco0puTsf06dPJz8/3PA4ePFi/O/STbikhhBDCl3xac1NXsbGxZGRkVFmWkZGB1WrF39+/xtdYLBYsFktDhFe+w8qCYrvLjqZp6HS6htu/EEIIcY5rVC03AwYMYNmyZVWWLV26lAEDBvgoohpYQjwzFGtoODWnjwMSQgghzi0+TW4KCwtJSkoiKSkJUEO9k5KSSE1NBVSX0oQJEzzr33nnnezfv59HHnmEnTt38vbbb/P555/zwAMP+CL8mvlZMWuVP0rXlBBCCNGwfJrcrF+/nl69etGrVy8Apk2bRq9evZgxYwYAaWlpnkQHoFWrVixatIilS5fSo0cPXnnlFd57772zZxg4VCkoBkluhBBCiIbm05qbwYMHox2TCByvptmHBw8ezKZNm+oxqjPkZ8UAGDUNp04nw8GFEEKIBtaoam4aBUswAFa3GwCbvZ7n1RFCCCFEFZLceJs5CHR6rC6V3OSX5fs4ICGEEOLcIsmNt+l0YAkm1C3JjRBCCOELktzUB0sIIeXJTV5Znm9jEUIIIc4xktzUBz8rIS4XIC03QgghREOT5KY+WII9LTf5dkluhBBCiIYkyU19sFgJlYJiIYQQwickuakPftbKlhtJboQQQogG1ahunNloWKxSUCyEEKfgdmv8mXyUojIXnZoFEx/q3yA3Gna7NRxuNxaj4YTraJpGZkEZfkYDIQGmKsuP5JeSlldC2+ggQgPMnueyC8tISs3Dz2Sgc5yV8EBzTZuuN3anG1upA5dbIyrIgl6vzqXbrbEj3caa/TkczC0ms6CMTFspmgbRVgvRwX60jAjg8q6xNAup+SbUALlFdowGHcF+phOuA7DhQA5mg4FuzUO8enx1Uefk5qmnnuKWW26hZcuW9RFP0yAFxUKcFXKL7CzbmUmZ08XIHnFYT/Gm7Eu5RXZyiu0kRgRi0J/ZBd7udLM9zUaA2UBiRCBmY+0b6YvtTn7dk83KXZlkFdgJshgIsBiJDDTTJzGcPi3DCLKoS4fbrZFuK2X1vqOs2pPF73uz8TcbuL5PAmP7JRARaGblriw+W5fKupRceiSEMqxzDBe2i2TV7izm/Z7C/uwiz76tfkY6xAaTEB5Ai/AAEiMCGdQ2kqhgi2edv47k89WGw+h1MKxLLH1ahnnO19HCMg7kFKNpoNeBTqfD7nRT6nBR4nCxK72ADQdy2ZiaS0Gpk4hAM81C/YgJ9sPPZMBi1KPT6ThwtIhdGQUUlKobH8dYLbSPCcZk0LPlUB7ZhZW31UkI96dddDD7swpJOVpc5VzGWv2ItlooLHVSUKa2dX7rCC7tGMX5rSNIzipi9f6jbDiQS5DFSK8WYfRuEYrRoGdN8lHWJueQnF1EkMWI1c9EWKCJPi3DubRjNK0iA8ktsrNw8xG+3nSY3ekFlDhcnn37mfQkRgQSbfVjy6E88oodp/zdP/P9dga2ieDqHnFEB/uBTv2ONxzI5ZfdWfx1xIbZoOfqnnFMHpRIl7jK5MXt1vh5Rwb/WbWfDQdyGdQ2go9vO7/Wf3feptNOdv+DGvTs2ZNt27Zx8cUXc+uttzJmzBgsFsupX3iWsNlshISEkJ+fj9VqrZ+d/PoK23+dxdj4ZkT5R7H8huX1sx/RJGw/YmP5zgx6JoQxqG3EaX9ydbs1ih0uiu1OsgrKOJRbwsGcYpxujcu7xJIYGVhlfbvTjcmgO+X+8ksc7M4oYFd6AfuzijAZdUQFWYgKttAqMpBOzayYDOriWeZ0sWZ/DlsP5xMWoC4c8aH+tI4MxGiovMDanW5W7MrkcG4JPRJC6RYfUusLsKZplDrUJ9SCUiduTcOtabjcGrlFDrIKS0nPL+O3vVn8uT8Hl1u9xQX7GZk4IJGx/RLYmV7A73uz2XQwj6ggMx1jrXSIDaawzMlfR/LZdthGVkEZOh3odTr8TQa6xFvp3SKMLnFWDueVsPVQPn8dUTOQR1stxFj9CLIYKXO6sTvduDWNYIuRkAAT/iYDmQVlHMot5lBuCW5NI9BsJMhiJL/EwfY0G2n5pQCE+Js4r1U4fcsv2kVl6ncaZDESF+pPs1A/cosc/LEvm9X7jnIot4TWUYF0jA0mPsyfLYfyWZ+S67nQGfQ6EiMCCA80U+pQF/pSp8vzvcPlJshiIjTARKDFyI40G3an+4Tn36DX0TYqiMIyJxm2Upzumi8hBr2OsABTlUSgJsEWI/Fh/uzNLKxxWzod9G4RxqC2kfy2J4uNqXlVno8MstA+Jog9mYVkFXj3djcGvc7z93Mso15HVLDF8zs7VrvoIBwud7VEx9sSwv1Jzy/F4aoen14Hx4cdaDbQr1U4HWKDiQn2I8bqB0BmQSkZtjI2puayNjmnTjF0bmYlyM+IXgdp+aUcKD9ms0HP6F7xPDu6q+e9wRvqcv2uc3IDsGnTJubNm8enn36K0+nkxhtv5JZbbqFfv36nHXRDaZDkZu1cjvz0KMMT4jHrzaz/2/oGaWoVvqFpGgdzSjiSX0KGrZTsQjtWPyPRVj+igizklzjYn11IclYRGuoNoUu8lWK7i7dX7OPnHRmebbWLDmLiwEQubh9FRJCZALORMqeLfZlF7Ey3UWR3MbxLjPpUBThdbj5dm8rbK/fV+EZ7rPMSwxnRLZbUnGLWpeSw/YiNiCALg9pEcEG7KC5uH1XlE7Ld6eZfi3fywR/J1d4oj+VvMtAjIYQgi4k/9mVTbHdVWyfYYmRAmwguaBfJodwSvtpwiKNFlRc9P5OeXglh9GsVTv9W4fRqEUqAubJh+WBOMYu3pfHD1nS2H7Fhd5344nu8Ts2sOFxu9mYW1vo1vmIx6ik7SWJRF2EBJpwuzdNiUBcJ4f4M6RhDu5ggistcFNmdpOYUszY5h0O5JVXW1eugS1wIF7WP5MJ2UaTll/DpmoOsTVEXyvBAM2N6xzO0UwwbUnNZ8lcGmw/mkRgRwORBrRjTp3l5Uuhib2YhezMLPYn59jQbWw5Vbf026nUM7xqLxajn5+0Z2EqrHl9ciB9Ggx63pqFp6pxaTAb8THriQ/3p2zKMvonhxIWq5CAtv4SsgjLKnG7KnC4cLo3mYf50iA2mVWQgdqebPZmF7EovwOFy0zU+hM7NrPiZDOSXOPjrcD57MgtpERFA74QwTxdWYZmTnWk28oodBPsZCfYzUWR3snJXJst3ZrEjzUaM1cKA1hGc1yqCojInG1NVq5LTpdE3MYzzWkXQJc5KicOFrcRBen4pq/ZksWZ/jicR7NzMynV9mnNJx2jCA8wE+RnVe1JuCcnZhaTll9Ix1kr35iGnTDQO5hSzYNNhftmdhd2lEnS3G9rFBDG4QxQXtI3iYG4x835P4YetadUSv2A/Izef35JJAxOJLk+evKnek5sKDoeD7777jnnz5rFkyRI6duzIrbfeyqRJkwgJ8V1f28k0SHKz5XOKvrmD8xMTAFg7fi3+xhP3Y4pT23Y4n682HuLCduoN9Pj/pHanm02pufy2N5vDeSUEW9SbSYDFgP6YxNJs0ONX/kZnK3FwJL+UI3klGPQ6ercIo19iOO1jgsgvcZCWX0p2YRl+JgNWPxMhASaigiyeFoYyp4tvk44wd9V+9pzBhVOng4FtIkhKzaPouMQgwGzA7nRX+URr1OsY1iWGC9pG8X+/7WdfVlG17YUFmEkI86d5eAAFpU5+25N10gQFwGzUc9N5Lbh7cBtcmsaUjzd6PiXHhfjRPjaYdtFBuNyQVaj67Hek2apdXKKDLfRvrd6sj+SVcCi3hMIaLrCRQRa6Nw8h6WAeOUVVP90b9DoCzQZMBj16va7GT+QV6xgNek8XRFiAiahgC5FBFjo3szK8vMXK7db4aXs6b67Yy7bDNlpFBjKobQT9EsPJK3awM93GrvQCAsxGusRb6RIXQkKY+j+rAXnFdpJS89iYmsfOdBvxof50jQ+hW7y6YGQUlJJpK6PY7sRiNGA26tGhLnD5JQ4Ky5xEBVtICAugeZg/JoOewjInRWVO/EwGOjWz0rFZMAEmA9uO2Phz/1G2HspXx2gxEmA2YCv/mzySX4LFaOD81uEMbBNJ2+gg9mcVsjO9gNSjxXRsFszANpG0jwkCIN1Wyp6MQgrLnPiZ9PgZDfiZDeqrSY/JoKeg1ElesR1bqYPWUUG0iw464Qeyw3kl7Eq3ERpgJtbqR1SwpcaL5t7MAtLzy+jXKqxabUuJ3YXFqPfUhJxMWn4JS7dnsCY5hw4xwdx4XoInuXe43Py5/yhpeaW0iwmifUwwgZbGUUpaYnfhZ9Kf1gffglIH6w/kEmv1o1OzerqOnUJafgmbD+Z7Wk6Nej0XtIv0dFnWhwZLbux2O9988w3vv/8+y5cvZ+DAgRw5coSMjAzmzp3L2LFjT3fT9aZBkptdP6J9OpbeiS1w6mDpdUuJDYytn301EaUOF3/sy+bXPdnEh/rzt/Nb4mdSb4grd2Vy10cbPc3sEYFmrureDH+zkUxbKem2UjYfrJ4YnC6dDk70v0Kvg/gwfxIjAtmdUUCGTV10zQY98WH+xFjVhdVW6iTTVkpWQRlBfkZaRwbSKjIIt6ax/YiN7Wk2ShwurukZx5RL2tImKghbqYMv1x9i/rqDJB8tqtI1YPUz0jHWit3lJulgXpWYwgPNTB3ajiu6NSPQbKzxDTM9v5SvNh5i9b6jtI4KpG9iOL0SQjmUW8Lve7NZsSvT08ViMeo9n0qD/Yy8ekNPLuscU+P5cLs19mUVsv5ALrYSB4PaRtK5mbXKRcvl1vjrSD6/7lHdKIEWA2N6q0+aJoMeTVPbWJucy9ryOoMjx7VC6XWqVmFEt2Zc1C6SiCALgWZDnS8MmqZR4nBVaRUSQjQO9Z7cbNiwwdMtZbFYmDBhArfddhtt27YF4I033uDZZ58lIyPjFFtqeA2S3Bz4A+aNYHDLFhzVw5cjv6RDeIf62ZcPOFxu8kscRAbVXGt1tLCMn3dk8POOTHKL7PibDfibDASYDfib1adQs1FPUZmTwlInWYVlrEvJodRReTFPjAjgn6O6crTQzkNfbMbp1ugSZ/V0+9QkItDMoLaRdGwWTHGZy/PJuOIPXNPA7iqvOXC4KmsYQvwosbtYm5LDxgO5niQpMshMZJAFu8uNrcSJrcRRrTskxmph8qBW3NS/RZ2KVU81WkPTNIrsLo4WlmEy6GkW4ue5kO9Is/HxmgOsT8nl4g5RTLmk7RkXymqaxh/7jvLq0t1sOJALQNd4K2/f1IcWEQFntO3TkWkrpbDMicOl4XC5iQv1b/CRJ0KIs0u9JjfdunVj586dDBs2jNtvv52RI0diMFR9g87OziY6Ohq32zt9x97UIMlN+laYcwGjEpqzz6jnvWHv0b9Z//rZVwPblV7A3R9vYF9WEee1Cue63s0Z0ima3RmF/Ln/KH/sy2bDgdxTdoHUpFmIHxe1i2LFrkwyj+uGuLpHHC9f3wOdDn7bk83SHRlYjHpirH5EB6uRDMe3GJwOp8tNdqGdsEBTtcRD0zSyCspIzi4i5WgRQRYTl3WOqdNIlLOdpmn8tjebfZmF3HheC0/rmRBC+Fpdrt91bpu94YYbuOWWW4iPjz/hOpGRkWdlYtNgLOqkh7icYDQ3uuHgmqaxJjmHRVvSiAq2MKRTNJ2bWflq42GeWLDV08KyNjnnhNX13eJDGNa5vCDR7qLY7qKk/Guxw4nd6SbQbPQU2vVIUEV6Op2OglIHr/y0mw9Xp+DWYPKgRJ68srMncbmkYzSXdIyul2M3GvTEhtRcCKfT6Yi2+hFt9aN/64h62b+v6XQ6LmwXxYXtonwdihBCnLY6JzdPPvlkfcTRtPip5MbqdILFfFZO5OdyaxTZVbeNergoKnOyL6uQj/5MZVdGgWfdV5fuJiLQ7BndcmG7SB4b0ZGVu7L4auMh9mcVERWsqv7Pbx3BxR2iiA89/QLqYD8TT1/dhZv6tyAtv5SL2kXKaDMhhBC1VufkZsyYMZx33nk8+uijVZa/+OKLrFu3ji+++MJrwTVa5S03oeWtVza7zZfRAFBU5uT3vdms3J3FL7uyOJxXctL1/U0Gru4RR06xnd/2ZHO0yI5eB1OHtmfKJW0x6HV0iQvh7sFtyC12EBZg8noC0j4mmPYxwV7dphBCiKavzsnNqlWrePrpp6stHzFiBK+88oo3Ymr89AawhJwV95dyuNy892sybyzfU+P8I8byoaaBZgOBFiOhASYu79qM6/o0J8RfFamWOlysSc4hPMBcbTptnU4nhZ5CCCHOKnVObgoLCzGbq1/MTCYTNpvvWyjOGkFRhDozgYa7v1Spw8WBo8VYjHoCzAb2ZRXx1MJt7M5Qc7AkhPtzaYdoBneIplvzEIIsRs904yfjZzJwcXupwRBCCNE41Dm56datG/Pnz2fGjBlVln/22Wd07tzZa4E1eoFRWHPTgfpvuSksc/LRnwd479dksgurT3YWHmjmH1d0YkzveKldEUII0eSdVkHxtddey759+7j00ksBWLZsGZ9++qnU2xwrMIrQ7Pq9eaamafxn1X7eXrHXM0tsoFkN3S12uDDodFzXpzmPXt6RMOk6EkIIcY6oc3IzcuRIFixYwPPPP8+XX36Jv78/3bt35+eff+biiy+ujxgbp8Coeq25sTvdPPLlZhYkHQGgdVQgdw9uyzU94zyzvrrcWpWbFQohhBDngtOag/zKK6/kyiuv9HYsTUtQtGe0lLdrbmylDu783wb+2HcUo17HzGu6cGO/FhiOmcBOp9NhNEgXlBBCiHOP3GClvgRGElI+VX++PR9N00673mVtcg5PL/yrvCVGx9FCO+m2UgLNBt75Wx8ukmJfIYQQwqPOyY3L5eK1117j888/JzU1Fbu96n1+cnJqnrH2nBMYhbW85cbpdlLiLCHAVPd79GQWlHL3xxurFQpHB1uYN7kfXeLOzruvCyGEEL5S5+Rm5syZvPfeezz44IM88cQTPP7446SkpLBgwYJqI6jOaYHR+GsaZg3sOtU1VdfkxuXWuP/TJLILy+gYG8yTV3XG4XLj1jT6tAgnJODMbpYohBBCNEV1Tm4+/vhj5s6dy5VXXsnTTz/NuHHjaNOmDd27d+fPP//kvvvuq484G5/AKHRAiNtNlkFPflk+cUFxddrEv5ftYfX+owSYDbw1vjdtooLqJ1YhhBCiCanzUJr09HS6desGQFBQEPn5aiTQVVddxaJFi7wbXWMWpOpgQlxqiHZdi4pX7c7i38v3APD86G6S2AghhBC1VOfkpnnz5qSlpQHQpk0bfvrpJwDWrVuHxWLxbnSNmcUKBnOVouLa+nLDIW7773o0Dcadl8CoXie+A7sQQgghqqpzcjN69GiWLVsGwL333suTTz5Ju3btmDBhArfccovXA2y0dDoIrBwOnl966uTG6XLzz++389AXm7G73FzWOYanRnap70iFEEKIJqXONTf/+te/PN+PHTuWli1b8scff9CuXTtGjhzp1eAavcBIQhypwKlbbuxON3//33pW7MoC4L4h7Zg6pB16vcxVI4QQQtRFnZIbh8PB3//+d5588klatWoFwPnnn8/5559fL8E1ekHRhGSnACevudE0jce/2cqKXVn4mfS8ekNPrujWrGFiFEIIIZqYOnVLmUwmvvrqq/qKpekJjCLEfer7S729ch9fbDiEXgfvjO8jiY0QQghxBupcczNq1CgWLFjgtQDeeustEhMT8fPzo3///qxdu/ak68+ePZsOHTrg7+9PQkICDzzwAKWlpV6Lx6sCIwktLyi2ldlqXGXh5iO8tGQXADOv7sIlHaMbLDwhhBCiKapzzU27du145pln+P333+nTpw+BgYFVnq/LPDfz589n2rRpzJkzh/79+zN79myGDx/Orl27iI6ufpH/5JNPeOyxx3j//fcZOHAgu3fvZtKkSeh0Ol599dW6Hkr9C4z23Dyzpm6pDQdyeOiLzQDcekErbh6Q2IDBCSGEEE2TTtM0rS4vqKi1qXFjOh379++v9bb69+9Pv379ePPNNwFwu90kJCRw77338thjj1Vb/5577mHHjh2e0VoADz74IGvWrOG3336r1T5tNhshISHk5+djtVprHetp2TyfdYvv5ZZmMbQKacXCUQs9Tx04WsTot/8gp8jOZZ1jmPO3PlVufCmEEEKISnW5fte55SY5Ofm0AzuW3W5nw4YNTJ8+3bNMr9czdOhQVq9eXeNrBg4cyEcffcTatWs577zz2L9/Pz/88AM333zzCfdTVlZGWVnlfZlstpq7h+pFUFTlPDfH1NzkFduZ/ME6corsdIsP4fUbe0piI4QQQniJz+4Knp2djcvlIiYmpsrymJgYdu7cWeNrbrrpJrKzs7ngggvQNA2n08mdd97JP/7xjxPuZ9asWcycOdOrsddaYJSnWyq/TN0Z3OHS+Pv/NrA/q4i4ED/+b2JfAsxyc3YhhBDCW+p8VT3VRH3vv//+aQdzKitXruT555/n7bffpn///uzdu5f777+ff/7znzz55JM1vmb69OlMmzbN87PNZiMhIaHeYqzimJobl+ai0FHIp39msiY5hyCLkfcn9yPa6tcwsQghhBDniDonN7m5uVV+djgcbNu2jby8PC699NJabycyMhKDwUBGRkaV5RkZGcTGxtb4mieffJKbb76Z2267DYBu3bpRVFTEHXfcweOPP45eX33wl8Vi8d1tIQIi8NPAz+2mVK9unvndZnXrikcv70DH2Hqu+RFCCCHOQXVObr755ptqy9xuN3fddRdt2rSp9XbMZjN9+vRh2bJljBo1yrOdZcuWcc8999T4muLi4moJjMFgANREeGcdgxECwgkpT252Z2Ww9XA+eh0yl40QQghRT+o8z02NG9HrmTZtGq+99lqdXjdt2jTmzp3Lf//7X3bs2MFdd91FUVERkydPBmDChAlVCo5HjhzJO++8w2effUZycjJLly7lySefZOTIkZ4k56wTWFlUvHy3KsY+v3UEEUFyk1EhhBCiPnitknXfvn04nc46vWbs2LFkZWUxY8YM0tPT6dmzJz/++KOnyDg1NbVKS80TTzyBTqfjiSee4PDhw0RFRTFy5Eiee+45bx2G9wVGEVqaA8DqAweB9oyQVhshhBCi3tR5nptji3NBdQelpaWxaNEiJk6c6Jmz5mzVoPPcAHwxmUcyf2FxUCClGVfizL2QNf8YQnSwFBILIYQQtVWv89xs2rSpys96vZ6oqCheeeWVU46kOicFRhHtUveX0htt9EsMl8RGCCGEqEd1Tm5WrFhRH3E0XUFRRDlVcqMz2riia80jwYQQQgjhHXUuKE5OTmbPnj3Vlu/Zs4eUlBRvxNS0HNNyozPauLyr1NsIIYQQ9anOyc2kSZP4448/qi1fs2YNkyZN8kZMTUtgNFHlyY2/fxGxIdIlJYQQQtSnOic3mzZtYtCgQdWWn3/++SQlJXkjpqYlMIro8m4pDLazcz4eIYQQogmpc3Kj0+koKCiotjw/Px9XeQuFqOQOiCSy/Lw4tFKKHEU+jkgIIYRo2uqc3Fx00UXMmjWrSiLjcrmYNWsWF1xwgVeDawqSSwMJ0DSCyyfyyyzJ9HFEQgghRNNW59FSL7zwAhdddBEdOnTgwgsvBODXX3/FZrOxfPlyrwfY2G1KtxOrWYhyuSgw6MkqzqJ1SGtfhyWEEEI0WXVuuencuTNbtmzhhhtuIDMzk4KCAiZMmMDOnTvp2rVrfcTYqG0+mMdRzeopKs4slpYbIYQQoj6d1u0X4uLieP75570dS5OUdDCPo4QQ7VR3U88qyfJxREIIIUTTVueWm3nz5vHFF19UW/7FF1/w3//+1ytBNRWlDhc70mxkaqFEudR9t7KKJbkRQggh6lOdk5tZs2YRGRlZbXl0dLS05hznryM2nG6NPGPlRH7SLSWEEELUrzp3S6WmptKqVatqy1u2bElqaqpXgmoqkg7mAWAIjSesTCU30i0lhBBC1K86t9xER0ezZcuWass3b95MRESEV4JqKjaXJzfW6BbSciOEEEI0kDonN+PGjeO+++5jxYoVuFwuXC4Xy5cv5/777+fGG2+sjxgbrYqWm9iE1p7RUlnFWTJLsRBCCFGP6twt9c9//pOUlBSGDBmC0ahe7na7mTBhAs8995zXA2yscorspOYUA9CqdXssP6vkxu62Y7PbCLGE+DI8IYQQosmqc3JjNpuZP38+zz77LElJSfj7+9OtWzdatmxZH/E1WhVdUq2jAgmOagFAqMtFnsFAZnGmJDdCCCFEPalzt1SFdu3acf3113PVVVcRFhbGO++8Q9++fb0ZW6NW0SXVMyEUTP7gH1ala0oIIYQQ9eO0JvGrsGLFCt5//32+/vprQkJCGD16tLfiavSqJDcA1niinRnsMcv9pYQQQoj6VOfk5vDhw3zwwQfMmzePvLw8cnNz+eSTT7jhhhvQ6XT1EWOjo2kaWw7lAccmN3FE5R4BpOVGCCGEqE+17pb66quvuOKKK+jQoQNJSUm88sorHDlyBL1eT7du3SSxOcbRIju5xQ50OmgfE6wWBjeT+0sJIYQQDaDWLTdjx47l0UcfZf78+QQHB9dnTI3e/qwiAOJD/fEzGdRCazxRTpnITwghhKhvtW65ufXWW3nrrbe4/PLLmTNnDrm5ufUZV6OWnF0IQKvIwMqF1jjPRH7SLSWEEELUn1onN//5z39IS0vjjjvu4NNPP6VZs2Zcc801aJqG2+2uzxgbnf3ZquWmTVRQ5UJrXGW3lBQUCyGEEPWmTkPB/f39mThxIr/88gtbt26lS5cuxMTEMGjQIG666Sa+/vrr+oqzUanolqrWclPeLZVdnI1bk4RQCCGEqA9nNM/N888/z8GDB/noo48oLi5m3Lhx3oyt0UrOrjm5iShvuXFqTnJLpVtPCCGEqA+nndx4NqDXM3LkSBYsWMDBgwe9EVOj5nJrHDiqkpvWUcckNxYrJnMQ4S4pKhZCCCHq0xknN8eKjo725uYapUO5xThcGmajnrgQ/8ondDoIbubpmpLh4EIIIUT98GpyIyqLiVtFBKLXHzf3zzFFxTJiSgghhKgfktx4WXJWDV1SFazxnuHgMmJKCCGEqB+S3HjZ/prmuKlgjfNM5JdRlNGQYQkhhBDnjDonN61bt+bo0aPVlufl5dG6dWuvBNWYVYyUan3sHDcVrM1o4XQAkFqQ2pBhCSGEEOeMOic3KSkpuMq7Vo5VVlbG4cOHvRJUY5Zc0xw3FazxtHQ4ATiQf6AhwxJCCCHOGbW+t9TChQs93y9ZsoSQkBDPzy6Xi2XLlpGYmOjV4BqbYruTI/mlALQ+QbdUS4dqucksyaTYUUyAKaAhQxRCCCGavFonN6NGjQJAp9MxceLEKs+ZTCYSExN55ZVXvBpcY5OSXQxAWICJsEBz9RWs8YS4NcJdLnIMBlJsKXSO6NzAUQohhBBNW62Tm4r7R7Vq1Yp169YRGRlZb0E1VjXOTHysgAgwmGnpcJBjMHDAdkCSGyGEEMLL6lxzk5ycXC2xycvLO+0A3nrrLRITE/Hz86N///6sXbv2pOvn5eUxZcoUmjVrhsVioX379vzwww+nvX9v2p9VMVKqhmJi8EzkV1F3k2JLaaDIhBBCiHNHnZObF154gfnz53t+vv766wkPDyc+Pp7NmzfXaVvz589n2rRpPPXUU2zcuJEePXowfPhwMjNrngPGbrdz2WWXkZKSwpdffsmuXbuYO3cu8fHxdT2MelE5UuoELTdQtajYJkXFQgghhLfVObmZM2cOCQkJACxdupSff/6ZH3/8kREjRvDwww/XaVuvvvoqt99+O5MnT6Zz587MmTOHgIAA3n///RrXf//998nJyWHBggUMGjSIxMRELr74Ynr06FHXw6gXFbMT11hMXMHajFblRcUyYkoIIYTwvjonN+np6Z7k5vvvv+eGG25g2LBhPPLII6xbt67W27Hb7WzYsIGhQ4dWBqPXM3ToUFavXl3jaxYuXMiAAQOYMmUKMTExdO3aleeff77GoekVysrKsNlsVR71QdO0ym6pk7bcxFXpltI0rV7iEUIIIc5VdU5uwsLCPHf//vHHHz3JiaZpJ00yjpednY3L5SImJqbK8piYGNLT02t8zf79+/nyyy9xuVz88MMPPPnkk7zyyis8++yzJ9zPrFmzCAkJ8TwqEjNvyymyYyt1otNBYsRJkpvY7iQ4Heg0KHQUcrS0+oSIQgghhDh9dU5urr32Wm666SYuu+wyjh49yogRIwDYtGkTbdu29XqAx3K73URHR/Puu+/Sp08fxo4dy+OPP86cOXNO+Jrp06eTn5/veVQkZt5WUW8TF+KPn8lw4hXbX45FbyHOKXU3QgghRH2o9VDwCq+99hqJiYkcPHiQF198kaAgNTIoLS2Nu+++u9bbiYyMxGAwkJFR9R5LGRkZxMbG1viaZs2aYTKZMBgqk4dOnTqRnp6O3W7HbK4+t4zFYsFisdQ6rtMVHezH/UPaYTaeIl/0s0K7y0jMXcNhk5EDtgP0ielT7/EJIYQQ54o6Jzcmk4mHHnqo2vIHHnigTtsxm8306dOHZcuWeSYIdLvdLFu2jHvuuafG1wwaNIhPPvkEt9uNXq+SiN27d9OsWbMaE5uG1CIigAcua1+7lbuMpuWK3/kdSMlPqc+whBBCiHPOad0V/H//+x8XXHABcXFxHDigulVmz57Nt99+W6ftTJs2jblz5/Lf//6XHTt2cNddd1FUVMTkyZMBmDBhAtOnT/esf9ddd5GTk8P999/P7t27WbRoEc8//zxTpkw5ncPwnfaX01LNiUhKZt2GzwshhBDi5Oqc3LzzzjtMmzaNESNGkJeX5ykiDg0NZfbs2XXa1tixY3n55ZeZMWMGPXv2JCkpiR9//NFTZJyamkpaWppn/YSEBJYsWcK6devo3r079913H/fffz+PPfZYXQ/DtyxBJMb0BOBA7h7fxiKEEEI0MTqtjmORO3fuzPPPP8+oUaMIDg5m8+bNtG7dmm3btjF48GCys7PrK1avsNlshISEkJ+fj9Vq9VkcRzZ+wPCtr2DUYP3NmzAY6txDKIQQQpwz6nL9Pq3bL/Tq1avacovFQlFRUV03d86K7XIdZk3DqYMj+5f6OhwhhBCiyahzctOqVSuSkpKqLf/xxx/p1KmTN2I6J+gtQbTQBwBwYMfXPo5GCCGEaDpqndw888wzFBcXM23aNKZMmcL8+fPRNI21a9fy3HPPMX36dB555JH6jLXJSQxtDUDKwd/A5fBxNEIIIUTTUOtCj5kzZ3LnnXdy22234e/vzxNPPEFxcTE33XQTcXFxvP7669x44431GWuT07JZP8j9ixR3Cez6ATpf4+uQhBBCiEav1snNsXXH48ePZ/z48RQXF1NYWEh0dHS9BNfUJYa1AeCAyQjr3pPkRgghhPCCOtXc6HS6Kj8HBARIYnMGWoW0AmCP2YyWvAqydvk4IiGEEKLxq9P44/bt21dLcI6Xk5NzRgGdSzqEdcCgM3DUABkGA7Hr34cRL/g6LCGEEKJRq1NyM3PmTEJCQuorlnOOn9GPdmHt2Jmzk20WM7FJn8CQGWA+yV3FhRBCCHFSdUpubrzxRumG8rKukV3ZmbOTrSHRDE07AFu/gD6TfB2WEEII0WjVuubmVN1R4vR0i+wGwLawZmrB2vegbpNGCyGEEOIYtU5u6niXBlFLXSK6APCXIw+30Q8ytkKa3ExTCCGEOF21Tm7cbrd0SdWDNqFt8Df6U+QsJqXdYLVwy+c+jUkIIYRozOp8+wXhXUa9kU7h6rYVW+O7qoVbvwCX04dRCSGEEI2XJDdnga6RKqnZZgD8w6EoE5JX+jQmIYQQorGS5OYs4CkqProduo5RC6VrSgghhDgtktycBbpEqqLinbk7sXcrT252fAdlhT6MSgghhGicJLk5CzQPak6oJRSn28lu/yAIbw2OYti5yNehCSGEEI2OJDdnAZ1O56m72Xp0G3Qfq57YMt+HUQkhhBCNkyQ3ZwlPUXH2Nuh+g1q4fwUUZPgwKiGEEKLxkeTmLOEpKs7eprql4vuC5oY9P/k4MiGEEKJxkeTmLFExU3FyfjL5ZfmQOEg9cXi9D6MSQgghGh9Jbs4SEf4RtApphYbGmrQ1quUG4NAG3wYmhBBCNDKS3JxFLoi/AIDfDv8GzcuTm8y/wF7kw6iEEEKIxkWSm7NIRXLz++Hf0YKbQXCcqrs5kuTbwIQQQohGRJKbs0ifmD74G/3JLMlkd+7uytabQ+t8G5gQQgjRiEhycxaxGCz0i+0HHNc1JUXFQgghRK1JcnOWqVJ3I0XFQgghRJ1JcnOWqUhukjKTKIxqBzoDFBwB2xEfRyaEEEI0DpLcnGUSghNItCbi1Jysyd4K0Z3VE4eka0oIIYSoDUluzkKD4tUEfr8e/lWKioUQQog6kuTmLHRs3Y0W30ctPCx1N0KIelSSC4segt1yy5d6l38Yvph09rbIl+ZDxnbYuww2z1fxNjJGXwcgqusb0xeLwUJGcQZ7Q2JoB3BkE7icYJBfmRDCyxwl8Ok4SF0Ne3+G9sN8HVHTUHQUnCUQ0rzq8tVvwl/fQOYOuGs16GvZzuAsg7IC9XCWQmR70Bu8G/PWL+Hr29UcaxViu8PfV4FO59191SNpuTkL+Rn96N+sPwA/2XaDxQqOYkhLUp+qfnoC9v/i2yCFEE2Dywlf3qoSG4DcZCjM9G1Mxzu6D94ZBOv+z9eR1J6mwbwR8Nb51QeE7F2mvmbthN2La7e9Hd/D8/HwUhv4d094+3yYewmU2rwaNqvfUomNXwhEdwGDGdK3VO89KM5Rx3F4I+QdVDPpO+3gcqi/Kbe75u03EEluzlJXtroSgO/2f487rqda+N5Q+OR6+OMN1aRpL/ZZfKIJ2/QRzP+b+nTYUEpy1SdGudVI/dv6JXw0BpY8Dtu+gu/vh12LwGCBwGi1TuqfvoktNwX2rai+/PfZkLENVs5SF05vWvMuzGoB+5Z7d7u5KZC9C+wFsOO7yuV5B9XyCr++qhKhk7EXwQ8Pg9uhfjYFqKQjbTN8dSu4Xd6JOXsPHNmoRuneswHu/gO6XKue2zCvcj23Gz66Vj3mXgKzu8LzcfBsFPwzEv4ZAe/7tvVPkpuz1CUtLiHQFMjhwsNsatahfKkG1ngIiICSHEj62KcxiibIXgyLH1Nvxtu/bbj9/vSkepP+v+HqzV/Uj9J8+P4B1fW0+k348haVzOr0cN3/Qaer1HoH15z5vg5vhHXvnfrCXcFeDPOuhP+Nqlr3U2qDrV+p74uyYP/K2m3P5YDNn6lWnxNxu1XiVJYPC+8/+QdGl7NuHyiPrac5NrnZV95qE9URjH5qktaU306+rd9eU1OChLaEf6TB42kw+Uf1+j3lrfnesPkz9bXtUAiKUt/3mai+bvu6spXor69VqYTBoq5JBnMNG/NtF9ZZkdy89dZbJCYm4ufnR//+/Vm7dm2tXvfZZ5+h0+kYNWpU/QboA/5Gfy5reRkA3wX4w7XvqT7PB/6Cix9TK61+y3sZuxAAOxepT5oAB/5omH06y2D7QvV9xlaYeykcrN17gKij9fOgzAZhraDfbRDXG/zDYeTr0GkkJJyv1jvTlhtHKXwyFhY9CHuW1u41f74NtkPq+5WzKpOirV+A45gWva1fVH3d3mXw++vVW3RWPAff/B3mXAh/Lah5n6mrwVZeLJufCqteqnk9twveHw6vdqp5cIezrPqyY0e4Hvhd1d+ASiwBuo6BXn9T3//2ankMh+DrO+DTmyqT/NwU+P3f6vvhz4E5QH3fvA+MnqO+//NtWPmC+r0d3Xd6rfpuN2z5XH3fY2zl8hYDVG2Poxi2famSxuXPqucuehimbYcnMmH6YXgsFR5NgUeSYfwX1XbRkHye3MyfP59p06bx1FNPsXHjRnr06MHw4cPJzDx5n29KSgoPPfQQF154YQNF2vCubnM1AD+lLqO080ho1kMVdPUaD/5hqm/82E8E4tyRugbeHlBzE/6Z2Pxp5fcHfvfutk9k3wr1yTkoBmK6QVEmfHDlyVuOnGWwdAa81hV+flr1/4uTc5bBn++o7y96GK58Be5YAY8mQ+8JankLVetH2mZVZHy6tn2pfo9Q2VJRwV6susSO/dstyobfZlf+fGSjSgI0rbI7pMto9XXn95UXb1ua6kJdOgN+faXy9dl74I831feOIvhiolrn+ARo25fqa2R79fWPNyBrF9Vs/ky1sJTmwUfXQeZOtdzlhGXPqC6ZX1+t+pqK5EanVzUsu35QiUFFvWSbITDwXtUFtG85LH4U3uwHW+arbsI5F6g6m5+eAFcZtLoIOl5VdR9dRsOl5a02K59XCdgbveGFRFj6VNWuZUeJOqcF6dWPD1Sil5+qajw7XFG5XKeD3uWtNxv+C5v+p649gVFw/l2V61iCVJ2OfxgEhIN/aM37aSA+T25effVVbr/9diZPnkznzp2ZM2cOAQEBvP/++yd8jcvlYvz48cycOZPWrVs3YLQNq09MH5oFNqPAUcDKQysrnzAHQr/b1fd//Lv2zb6i6Vj2DGRur/qGfqZsabC/4oKjU58YG2Jm7L++Vl+7XAu3/KjewF12+ObOyovIsTK2q9ad31+H/IOqyX52N3VOSvJOvq9NH8Fr3dTF+3QcXAvvXXZ2fqgoyob176v6pZps/gwK01U3Qrfra14ntKVKMt0O1a10OjQNVr9d+fPxCfimj1SX2MfXqa4OgF9eUC2GzXrAgHvUspX/UjGkb1XdH1e+quKzF6pEAVQLgqO4chsH16r9L35EHUPboSqBAPX38vmEyvdLp12NWAIY8QK0H6Fes+jBqu+pjhLVCgTq4l2SA/8brfb14TXq/6DbCRs+qHydo0QV4QL0vEl93fGd6qoqs6kEIK4nhCWqFhyANXPUsbQYoFrUSvNg/nj1Op0BLn+h5tFKFz4EQ2ZA835qe6ZAlQz9Phve6KNa+L+dAi+3V7VW/zcMygqrb6fig03nq8HkX/W5HuPKa3ySVNIEcNEjKqE5S/k0ubHb7WzYsIGhQ4d6lun1eoYOHcrq1atP+LpnnnmG6Ohobr311lPuo6ysDJvNVuXRWOh1eq5qrTL17/Yd92Z63u3qP/zhDQ3XfSDODpk74EB5H/2xzd1nast89Qkz4Xxo1r18+7X82zq0Htb8p+4jJBylsLP8QtVltHqzvOFDaD1YvdF/MbGyyNjtVi0P7w5WxaUBEXDZMxDbTV3wfn1FFd3nHqh5Xy6nuhjmp1a9+NZW2mZ1cTi0VtUI1fZYM7bDh6O838p2rNwD8H+XqXqar26r/rzbpT4IAZx/NxhrqpFAXTwTyltvDh7XNZV3UJ3/eVfC881VIW5N9q+EzL9U0atOr4pnj02Sd5R3Qbqdqs7ql5dUUgZw2T9h0P1g9FctJQvLE50uo1RrQEVStvVLlfRU1B0mnA+aSx375k9VS4jBDCNehGHPwnXzVH3KrkWQ9El5nCtUIhgYDYkXqQTH6A8pv1bGA7B2ruq6sjaHu/9UtTIFR9T5PvAbmIPUvvIOqNFPAGlb1PEFRsP5Uyr3V9Ea2ebSyiHcFz6oEpKASBj1DkxeDLcsqUzKAPrdCjGdT/w7u/BBuO1nuH8z/OMwjJsP4a2hMAOW/EMllGXl1768A7BsZtVtOEoqY+sxrvo+AiMqW43KbBDaAvpMqjmes4RPk5vs7GxcLhcxMTFVlsfExJCeXnPT2W+//cb//d//MXfu3FrtY9asWYSEhHgeCQkJZxx3Q7qqjfqD+v3w7xwtOeYiFhQNPcv/CCvetMS54djhsBXN3WdK0yo/ufUcBy3VLNme4cEn43aproHFj8D2BXXb775l6hO7NV598gT1pn/tXNWCkLUTfnhEXRw/uhZ+fEx9Km03TM0PMuh++PuvMPZjtY2je9RFJ21L9X3t+QkK0tT3u35QiVVtZe+B/11beYHITYbkWkzHYC9WrQX7V8DCe+u2z9rK3KG6I3L2q5/3/lx95M/ORXB0r2p5qCgQPZEWFXU3xxQV//SEGhHz42Pqgm4vgMUPq26c4/1Znjj2uhma9VTfVxQBFx2tTJg7XqX+flc8qxKBdsOg9cXqva3vLeXHtl197TNZfe1+Q/kxLoXvpwGaavEb/zmEtFAX7gXlXSUD74OINur7rtfCJf8oP5bHVSvX1vIuqS6j1fxhYS3h4ofVskXTVCJcnFPZOnrJP8AaBzd/oy7uADFd4Y5fVDIOlf8XK7qkmveD6E4Q0Va1Rq57Ty1vM6TyfEV3hKlb4IFtqpVHp1PJ57Bn4eYFqsZyyFPVz/OJ6HTQ4XK4e41KFuN6qW7HSYvgb+UtZWvfrVrEvGux+tsOSYAWA2ve7rHJzCWPnzhBPkv4vFuqLgoKCrj55puZO3cukZGRtXrN9OnTyc/P9zwOHmxcIzFah7Sma0RXXJqLH5KPu4gNuBfQwe4fKwsyRdNSlF21j7ysoHJEQ6uL1FdvdJGkJalEwmCBzqOgZfkbXG1abpJXVSYNW+bXbb8V3QKdR1WdyCwoGsa8pz75J32kahH2r1CfrK98BW76HILLPxTpdGqUz61L1b3YCjNg3hXV54La+N/K78tstR/6m5eqWl6Ks1W3Sfcb1fJjh8aeyE+Pq4QLVBdaxcWtLnKSVUK7fWHVofL2YtWtM2+EOv9RnSpj++nJysEGzjL49WX1fb/bwRJ88v1VFBUfWqtap1J+r0xiWgyE4c+rxAFU0nNs12jWLpVEooPz74Q2l6jlFcnN7sWqhSWmG9zwP1XUDOr3PPSY1oRB96mWFlAtJRUJV1QHNaGc26niM5hh6FMqabv2XbUdUBfpCx+selzn3632W5Krup52LlLLj+2iGzS1ssVk1UvwzkDVPRTVCXqUn1trHNy2DMb8n2otiWwLHUao53aVz1njSW76lv99jlQ/VwzlbntMcgMQGFm9KwjU+btk+ul1/xjN6jzesRKufgMSL1D7raif+fYe9X6S9Kn6ewGVPJ5oQsHEC1WrTvexJ+7WPIv4NLmJjIzEYDCQkZFRZXlGRgaxsbHV1t+3bx8pKSmMHDkSo9GI0Wjkww8/ZOHChRiNRvbtqz7kz2KxYLVaqzwam5Ft1H+MBXsXoB3bFxzZFgaUN3l+O+XkQx5F41NWqIoKX+9ZeSHe8rn61BzRTjW5g7ro12Yir9J8+HNOzXUpSeWtNh2vVIWALQaonzO3n7pYt2KEBahWg6LsE6+be0CNCAHVFF5xMeh6bfV1W11UOTLQXqhaAe78VV0Qa6o9CIlXTfqJF6pz9NlNKjEANX38nvLhxe2Gq68VtT4nk7Ub3h+hRvFEtleffAeVX9h3LoKCjBO/dtfiyu6NnuWjYn59Wf0eKuQfqrmAtTgHfnlR/f7/3VO1JHx+M7zYGj65UbWUvdgavpysLtbN+8HkH+DyWepCn7FNJcGOEnUe0jarro/+d576mJt1V0lkSS5k7YAfHlLL+94CtyxW7znD/gmDy1tClj2jak+WPlU5JLnjlapbpKJFY/9K1Tq443v1c6eR6iJ6xcsw+l3VjXJst0twrEpGQCVSx/6+j72w9v+7qjMBaDkAhj6tjvOq1ypHFVUwmNSoMHSqhdFRpGp4Ku7fB6rVcNizMPo/KtGvSNqHPl11JuCgaOh2XWVC0v5y9fXQevU3UTEMvKI1siK5AdXaE1z9+tZghj2ruthyk+HVzrDgTvX3bY2Hvicp9dDr1eisa9/1/qzI9cCnyY3ZbKZPnz4sW1ZZTe92u1m2bBkDBgyotn7Hjh3ZunUrSUlJnsfVV1/NJZdcQlJSUqPrcqqtK1tfiUlvYnfubnbk7Kj65NCn1YWozAafTzyzEQ6i4Wz8EOYOOfmQ543/VW+uzhJ1Qdv9U2WXVL9b1SfaiHaqubviwn0yPzwMPz4KS5+sutxprxw1UlH8GBgJkeXzK51sWLC9uLKGwj9MfaLedoKkYcvnqsBxdjfVVfPHGyppCUmAinuoHe+ih2DwdFVbc+tSiGx38mP0D4W/faVaGOyFaiiwy6lqDjQ3tLxAjRQClXyc7P/LofWqu8d2SJ3nmxeo8xLTRdWluJ1q5EiF7D3qGHd8D7uXqA8coApkR76ufl8luaqwFdQ9e97oC2+dp/7vHt2nWlvW/Z8a8bLiOVVXojOobsKwRDXl/u7FqrXOWaK6Ry54QMUWEK4eFce3/J/w8fUq4TQFwLhPK+cuORmDqfL3seBuleD6h8Olx/3dDH5UFbKCSl5+n135d1iRmCT0V4lSYfkFvyJJr5hPR6dTw45rut3DkBlw/xY1OvRY3a5TxxMYrYppjzXofnj8CLS7rOZja95H1Sseu62aEuUeN6pELqarag1rP7zm7VWwxqnuHzRVWGw7pFqR4nqp5+N6q+QBqrfaNDQ/K1xd/jdYUdw8dCbcs059QGgifH6jomnTpjFx4kT69u3Leeedx+zZsykqKmLyZNXHOmHCBOLj45k1axZ+fn507dq1yutDQ0MBqi1vSkIsIQxpMYQfU35kwd4FdI445hOOwaSK5f5zoZoj5LupcPEj6oJxlveJNgqa5v37qeQfUomGs1SNuhj/RWU3UAVnWeVQ1rBW6lPWp2PVBdror5qHK5q7f3tVXey6XXfifeYkV9YYbPsGLv+XGnUHqsiy+CgExULrSypf03KgKgY98Dt0vKL6NkFdaO2F6iLb/y5YMh22fAb976hcR9NUi0XF3BigihcrChi7jDrxOdYbYPBjJz6umhgtcO1/1HT9B9eobpONH6rn+kxUn9RDElQ30d6fq36qdrshZ5865h//oT7dx/eBm75QRZUV+kxW2974X5VcrH8ffpxe2e1QIaabukgbjOrrZzepYubCDJVwec7HAjXEObSl2j+oqe8H3K1G8QRGqPOYuV3Vdbhdqisktnv1c3feHaqmIi9VJcfm4PK/seofGE+oRX9VW5OWpH6+bKZKnI534YOqVubgWsj4S8UX263y79loUd/vW6a66Fxl6u85+gTFscfS6VQdzPGscXDXH6rV5HSGG1/6pEpsC9Iru/FqEt8H7qrDdAgdrlAT260u/38b3aWyO0mnU+/Lq9+u7BbypbZDYdQc1dXae4Jq7WtifF5zM3bsWF5++WVmzJhBz549SUpK4scff/QUGaemppKWlubjKH1vdFs1x8Oi/Ysocx03YZS1mer/1enVheWN3vBstOrO2PVjwwfbVOxboe7l8suL1Z/b/q0aOXI6kyiumKUSG71RJQYfjVF1K8faMl+NyAhuBnf+pooeK25k1/36yjf1igvznqUnb4X449+q1gHKp4P/vvK5DeW1KL3+VvXGrBVFxSeru6nokup2g+ou0BnUCL7svWq5ywHf3VeZ2Ay8Vx1Pj3Hq+PVG1YfvbaEt4IryCdlWPq8+SfuFQqer1YWmyyj1XEUrU/4hNXHav1rAm33hu/tVYtPmUpiwsGpiA+r1fqEqgfjgStV143aompyE/uoTf1xvuO59dYEHdfFL6K9aXCoSm4sfVQXRbS9TLUE5+8ASAiNeUpN29vpb5b51OtVqdNHDKuGrmPfqeEaLatEFddGasKBuiQ1U1t2A6lqp6FarSWw31ZJ41atqKP8VL1WNq6LupmLW404jz/wDQ3ir0+/a8bPC7cvV32FU+zOL41gVdTcVRefHdneBKsi9Z21lkbOv9Ryn/j82wcQGQKdp59YkKTabjZCQEPLz8xtV/Y3L7eLyry8nvSidly56ictbXV59paRPVdNw7gH1Bgqq4O5vX1UWn55t3G5IWaU+gdb0ydBXXE5VTFhxD5ixH1c2pW/5Ar4uL4TsMlqN7DGYarfdjO0wZ5BKVCZ+r+Zo2bdMFU+OKZ/+3u1SBbQ5+1T/+MB7VTyLpqni8UmLKrtnNE1NZGc7BDd+qlpYHKXqAldxAbGlwevdVfdVmyFqf60uhokLVYvOv3uq9e7fXFm/AOqC/1oXlbA8llq9qLHoKLzSXl2Up6xVxZ4fX6+6Ji56RNWmfD5R7U+nVzVCx3YJFKSruqLItrX9rdSNpql7sFWM4Op/F4z4l/r+8AY1V44pQNURfDdVzV8C6ncR01V1H1z40IlbQH+cXjkyqKIgduC9J79wp66BD65Qw4evnVu1Oyb5V1Uk22tC7bqPTiXlN9VKcjpdDSW58HIHlbDdvkLNyXK60req2qEKty6FhPNOf3tnK01TXa755YNWrnm7epeaOCN1uX77vFtK1I5Bb+DqNlfz7pZ3+WbvNzUnNz3HqYemqXuwLJqmuis+vQkmfX9mb1D1oWKyrXVzVdP7TZ/5OqJKmz+penO7BXepIZ2FGfDt3ZXL//pGJRPXfwAmv1Nvd9lMldh0uhpaXag+FX8+AfYsURN2nXeHag7P2adaBiqGXxqMcHUNQ/4ruqbWvKMmvUMrH9LZAka9pZLa1W+qxKbFAFVo+Xp31VKUd7CyZqT1JVUTG4CQ5qoFJC9V1d20G1r1+b++VolNsx4qsQHVCrPnJzWsfNdi1VVqClAtGBWfbCsEx8IpBu6cEZ1OHe+hdeou130nVz4X11t1AeUdUOcf1HFcNVsl2oZavDX2vVV1dxn94Pp5tfsA0aK/GqJbUR9zrFYXqoe3JF5w6nVOxD8Mbi5v1TrT943oLmo226Is1fUZ3/fUr2mMdDr1N762fP6fimJi4RM+75YStTeqzSgAVh9ZTVrhSbrqdDpVzX/te5UjRz4aU9lVUEHT1NDUBVN8M339qpdVYgPqglh43C03ljyuZtOs79iydquLfUUjpr0YVjyvvh86UzXRl9lUvcRnN6lEodNINRzZ6KfqTj4ao/rTkz5Rhb813d065XfV8qIzVBZimvxg7EeVs7KufVcVwYIaCXKqYbtQOXqkLL+ySTw/Ff57tTqH68uHLF/4kKphSLwQ0FTXyKbySdBONPdJRdfU/PGquPTQBnW+dnxfORLo2G6lDuWtEvkHVWITGK1amo5PbBpKQLjq3rl7dWUCBuVdU6Mrf+53u2pRiO9du8QGVIvTvRtg6ta6tYxGtj27WilPJPGCM0uQKuj1laOmOl554qHGTUHFbQv8w9TcNsJnpFuqkbllyS2sS1/HPT3v4e89/n7qF5TaVE1A+hbVt3rlq6rw1FGqaiEq5iVpN0wNx2yoN5718+D7qep7v1A1l8SIlyoLUY/uU7VDAJc8UTm51pkoyVPn4NhugwN/qMnZnCWqbuSqV9WMpMtmqtaPe9er5Oo/F6pPnqA+9U9apIaaJv+qbhDoOC6ZCYyCC6ap1gJNUy08q15ShcF9b1EtCsfbs1S1vhRnq9aOB/6q/UUwbYuq3wmMUveGWfFc1XldYruri7xOp7ovF9ypuixddjUz6rQdNXe/ZPylkq30rTXvV6dXrz22/uHbKSpxiuygCllrKgo9GxRkqKHLnUaqKedF/clLVTNYX/CAGnHWVGmaqsWLaq+KdoVX1eX6LclNI/Pdvu/4x2//ID4onh+u/QG9rhbJSGGmugAfKb9XTJfR6s3m8AbViqA3qlEMQ2ZUn/gK1Midb/6uLpojXz/zYsBNH6nZWjW3Ko4MiFAznzbvpybFAtXiUDHqIDBafTquTbfPiSR9qrqT4nrDqLfVp/jDG+C/11TeBRtUjUJxjmoFGf2fyom7kn9VrTPBseoTfsUEcqAu/Bs/VCOOSnJVy0bF3Y2DYlShb0WLSkCkGoFxomLIgnR1A8HEQVVH8ZyOvxaoBLY0v2rNkL1I3WfGXn5/mYH3qXlLTkTTVNfO2rmqfkVvUjU/ke3ULLMVxbkVinPUHDCdRvr85nlCiKZDkpuTaOzJTYmzhCFfDKHAXsDbQ97mwua17KN3OdSQ2F9erBw14x8G1/9X1R0svFd9Cp+wsHq//7J/Vs5wettyNVdEhd0/qbsyD3ny1F0P9iI1BLrifjC9J6pkqTADXu2kkp37klRC8Gon1ZpT0bpwzVtq5AiogtvdS9SNHctsapbNiDbqBnQ1Vf5n7lT3I/IUWVvUENv189Q+Ei9U86l8e09lMWBMN9XScWxLVkGG2v6pkiyXQx3jLy9VJjlhiWrIZa+bVZdhQynKVolsfO+qyxdMUTP/AtyzofZFvS6nGp7t7eHxQghxCpLcnERjT24AXlj7Ah/t+IjBzQfzxpAa7u1yMoc3qmGueiNc939qFlFNU/UUmz9RrSR//0XNJQGqdeO9yyoToj6Tymf5RL3u7fPVtP16o7rhYccra95v1i6Yf7Mq0tXp1cRsFz5YOdPlh9eoicAufUIVHS68p/zmbJNVF1F0ZzW3Baj4j+1yqWD0V61SfW+BhPJiPkeJGhWTuV0lMUaLmtukQvN+6l4xlmDV6vL9A2oI+I2fqNaTM+EsU0mYX4ja99lUa3B4g6pnanvZ2VXILYQQJyDJzUk0heRmf/5+rllwDXqdnh+v/ZFmQc3OfKP2YnhviEoCAsvv65PQH/5zkUpIYrqqKd3NwfDQblVvsm+5moSuwokSHLcb3uyjbuwX3Ext+/hCxU0fq26jyA6qZSRtsyrm7TNJDUe2F6oh7UeS1Myr6KDzNar1yVQeS9Yxsze3GAgXTlPdIxvmqWO663dVk5L0iZpQLKKtqgnxD6saS31M3Hc2sh0pP3813NNGCCHOMpLcnERTSG4Abl1yK2vT13JH9zu4t9e93tlobgp8Oq78Trw6NQnVoXWqm+iu1fDepWqdilqUj29QQ5j73qq6d7Z9peoxbvyk6vwde5epOzpbQtTokprm8CjNh5faqdofUF1H03aoCcwWP6aGOoe0UKOAoGrxMVTWhWz4ALZ+obqyPHSqdaZiMjFQ3Ss6/dnVmiKEEOKE6nL9lnf2Rur6Dmr479d7vsZx/JTvpyssUd3ttvcEQKu8s+1Vs1WSUTFL6cb/qdFMe5YAOnUjvdHvQpdr1aRf392nRmNVqLh7co8bTzw5mV9I1fu3dB1TOTPr+XeqRKQisRl4X9XEBlRLS8J5qlj4/i1qaLWp/PYCF06rmtiAGu4riY0QQjRJ8u7eSA1JGEKEXwTZJdmsSF3hvQ2bA+DqN1SyEhSjkoSK+wr1HAfo1D1nfpyulrUbpop5DUY106u1ubqfTUVCU5BeeefnE82lUuHYu/32u63y+7DEyjlJuo5R3VUnY20Gw5+DB7bB5B+r3/BPCCFEkybJTSNlMpi4tt21AHy++3Pv76DHWHhwl0oSKoQ0r7yj7Z4l6uv5d1Y+b7SoUUcAv76q6ng2faRmsW1+nrovzsm0H66Spd4Tqo7IAtV6NO4z1SVW2xaXgHB1T51zoX5GCCGEhyQ3jdh17a9Dr9OzJm0N27K3eX8HNSUFFcOxAaI6Vr2LdMXzoS2hKFPNtlsxqunYqe9PxGhRBb5X1zACzM+qhprX9h5OQgghzlmS3DRicUFxXNlKjUx6fePrDbPTDldUji7qf2f1BMhggosfUd8vf1bNsWIJgc6jGiY+IYQQ5zxJbhq5u3vejVFv5M+0P1l9ZHX979BoUUO5L360aivOsbrfCOFtVHExqEJic0D9xyaEEEIgyU2j1zy4OWM7qBsXvr7xdRpkZH/boXDJP07cRWQwwuDHKn+uuLO1EEII0QAkuWkCbu92OwHGAP46+hdLDyz1dThK1zHqTssXPwoxnX0djRBCiHOIJDdNQIR/BBO7qGHWb2x6A6fb6eOIULdVuPJl1cIjhBBCNCBJbpqICZ0nEGYJI8WWwic7PvF1OEIIIYTPSHLTRASZg7iv930AvJn0JocLD/s4IiGEEMI3JLlpQq5tdy19YvpQ4izhn6v/2TDFxUIIIcRZRpKbJkSv0/PUgKcw6838fuR3FiUv8nVIQgghRIOT5KaJaRXSir/3+DsAL659kdzSXB9HJIQQQjQsSW6aoMldJtMurB25ZbkNN3OxEEIIcZaQ5KYJMhlMPNH/CQAW7F3A/vz9Po5ICCGEaDiS3DRRvWN6MzhhMC7NxRsba7gRpRBCCNFESXLThE3tPRW9Ts/PqT+zOWuzr8MRQgghGoTR1wGI+tMmtA2j2o7i6z1f8+r6V/ng8g/QHX8XbyGEaIJcLhcOh8PXYYg6MpvN6PVn3u4iyU0Td1ePu1i0fxEbMzey6tAqLk642NchCSFEvdE0jfT0dPLy8nwdijgNer2eVq1aYTabz2g7ktw0cbGBsYzvNJ73t73PaxteY1D8IIx6+bULIZqmisQmOjqagIAAaa1uRNxuN0eOHCEtLY0WLVqc0e9OrnLngFu73crXe75mX/4+vtr9FWM7jvV1SEII4XUul8uT2ERERPg6HHEaoqKiOHLkCE6nE5PJdNrbkYLic4DVbOWuHncB8FbSWxTYC3wckRBCeF9FjU1AQICPIxGnq6I7yuVyndF2JLk5R1zf4XpahbQityyXuVvm+jocIYSoN9IV1Xh563cnyc05wqQ38VDfhwD4aMdHHCw46OOIhBBCiPohyc055ML4Czm/2fk43A5eXvey3DVcCCFEkyTJzTlEp9PxcL+H0ev0LD+4nMd+fQy7y+7rsIQQ4pw2ePBgpk6d6tVtTpo0iVGjRnl1m42JJDfnmPZh7Zk5cCZGnZEfkn/g9p9uJ680z9dhCSGEEF5zViQ3b731FomJifj5+dG/f3/Wrl17wnXnzp3LhRdeSFhYGGFhYQwdOvSk64vqRrUdxdtD3ybIFMTGzI3cvPhmDhce9nVYQgjhVZqmUWx3+uRR227/SZMm8csvv/D666+j0+nQ6XSkpKQAsG3bNkaMGEFQUBAxMTHcfPPNZGdne1775Zdf0q1bN/z9/YmIiGDo0KEUFRXx9NNP89///pdvv/3Ws82VK1fWuP8ff/yRCy64gNDQUCIiIrjqqqvYt29flXUOHTrEuHHjCA8PJzAwkL59+7JmzRrP89999x39+vXDz8+PyMhIRo8eXbdfVD3w+Tw38+fPZ9q0acyZM4f+/fsze/Zshg8fzq5du4iOjq62/sqVKxk3bhwDBw7Ez8+PF154gWHDhvHXX38RHx/vgyNonAbEDeB/I/7H3cvuJsWWwoTFE5h72Vxah7b2dWhCCOEVJQ4XnWcs8cm+tz8znADzqS+xr7/+Ort376Zr164888wzgJrrJS8vj0svvZTbbruN1157jZKSEh599FFuuOEGli9fTlpaGuPGjePFF19k9OjRFBQU8Ouvv6JpGg899BA7duzAZrMxb948AMLDw2vcf1FREdOmTaN79+4UFhYyY8YMRo8eTVJSEnq9nsLCQi6++GLi4+NZuHAhsbGxbNy4EbfbDcCiRYsYPXo0jz/+OB9++CF2u50ffvjBS2fx9Ok0H1eV9u/fn379+vHmm28CaobChIQE7r33Xh577LFTvt7lchEWFsabb77JhAkTqj1fVlZGWVmZ52ebzUZCQgL5+flYrVbvHUgjlVmcyd+X/p29eXsJtYQyZ+gcukR28XVYQghRZ6WlpSQnJ9OqVSv8/PwotjvP+uQGVM1Nz549mT17tmfZs88+y6+//sqSJZXxHzp0iISEBHbt2kVhYSF9+vQhJSWFli1bVtvmpEmTyMvLY8GCBXWKOzs7m6ioKLZu3UrXrl159913eeihh0hJSakxQRo4cCCtW7fmo48+qtN+TuT43+GxbDYbISEhtbp++7Tlxm63s2HDBqZPn+5ZptfrGTp0KKtXr67VNoqLi3E4HCfMSmfNmsXMmTO9Em9TFB0Qzbzh87jr57vYdnQbt/50K3OGzqFndE9fhyaEEGfE32Rg+zPDfbbvM7F582ZWrFhBUFBQtef27dvHsGHDGDJkCN26dWP48OEMGzaM6667jrCwsDrtZ8+ePcyYMYM1a9aQnZ3taZFJTU2la9euJCUl0atXrxNeY5OSkrj99tvrfoD1zKc1N9nZ2bhcLmJiYqosj4mJIT09vVbbePTRR4mLi2Po0KE1Pj99+nTy8/M9j4MHZX6X44X6hfLe8PfoF9uPIkcR96+4nyOFR3wdlhBCnBGdTkeA2eiTx5lORldYWMjIkSNJSkqq8tizZw8XXXQRBoOBpUuXsnjxYjp37swbb7xBhw4dSE5OrtN+Ro4cSU5ODnPnzmXNmjWeWhq7XY2k9ff3P+nrT/W8r5wVBcWn61//+hefffYZ33zzTbXmqwoWiwWr1VrlIaoLNAXy5qVv0jG8IzmlOdy7/F6KHcW+DksIIZo8s9lc7XYDvXv35q+//iIxMZG2bdtWeQQGBgIqeRs0aBAzZ85k06ZNmM1mvvnmmxNu83hHjx5l165dPPHEEwwZMoROnTqRm5tbZZ3u3buTlJRETk5Ojdvo3r07y5YtO91Drzc+TW4iIyMxGAxkZGRUWZ6RkUFsbOxJX/vyyy/zr3/9i59++onu3bvXZ5jnjABTAP++5N9E+EWwO3c3j/36GG7N7euwhBCiSUtMTGTNmjWkpKR4uoamTJlCTk4O48aNY926dezbt48lS5YwefJkXC4Xa9as4fnnn2f9+vWkpqby9ddfk5WVRadOnTzb3LJlC7t27SI7O9tz361jhYWFERERwbvvvsvevXtZvnw506ZNq7LOuHHjiI2NZdSoUfz+++/s37+fr776ylM68tRTT/Hpp5/y1FNPsWPHDrZu3coLL7xQ/yftVDQfO++887R77rnH87PL5dLi4+O1WbNmnfA1L7zwgma1WrXVq1fXeX/5+fkaoOXn559WvOeCTRmbtF4f9tK6ftBVe2XdK74ORwghaqWkpETbvn27VlJS4utQ6mTXrl3a+eefr/n7+2uAlpycrGmapu3evVsbPXq0Fhoaqvn7+2sdO3bUpk6dqrndbm379u3a8OHDtaioKM1isWjt27fX3njjDc82MzMztcsuu0wLCgrSAG3FihU17nvp0qVap06dNIvFonXv3l1buXKlBmjffPONZ52UlBRtzJgxmtVq1QICArS+fftqa9as8Tz/1VdfaT179tTMZrMWGRmpXXvttad9Lk72O6zL9dvno6Xmz5/PxIkT+c9//sN5553H7Nmz+fzzz9m5cycxMTFMmDCB+Ph4Zs2aBcALL7zAjBkz+OSTTxg0aJBnO0FBQTUWXh2vLtXW57Lv9n3HP377BwD3976f27rd5uOIhBDi5E420kY0Dk1itBTA2LFjycrKYsaMGaSnp9OzZ09+/PFHT5Fxamoqen1l79k777yD3W7nuuuuq7Kdp556iqeffrohQ2/SRrYZydGSo7yy4RVe3/g6gaZAxnUc5+uwhBBCiFPyectNQ5OWm7p5c9Ob/GfLfwB4ZuAzjG7n+5knhRCiJtJy0/h5q+WmUY+WEvVvSs8p/K3T3wCY8ccMHlz5IFnFWT6OSgghhDgxSW7ESVXcSfz2brdj0Bn46cBPXLPgGr7c/WWt750ihBBCNCRJbsQp6XV67ut9H59d9RldIrpQ4Chg5uqZfLzjY1+HJoQQQlQjyY2otY7hHfn4io+5s8edALy0/iX+OPKHj6MSQgghqpLkRtSJQW/g7h53c02ba3Brbh7+5WEO2A74OiwhhBDCQ5IbUWc6nY4ZA2bQPao7NruN+5bfx8ECuWeXEEKIs4MkN+K0mA1mZg+eTXRANPvz93PF11cwftF4Ptr+Efll+b4OTwghxDlMkhtx2qICopgzdA79Y/ujQ8eW7C28sO4Fxiwcw+aszb4OTwghxCkMHjyYqVOn+joMr5PkRpyRdmHteG/4eyy7fhmPnfcYCcEJZBRnMGnxJD7a/pEMFxdCiFOojwRj0qRJjBo1yqvbbEwkuRFeERUQxfhO4/n8qs8Z1nIYTs3JC+te4MFfHpRuKiGEEA1KkhvhVUHmIF6++GUeO+8xjHojSw8s5brvrmN9+npfhyaEONdoGtiLfPOoZav1pEmT+OWXX3j99dfR6XTodDpSUlIA2LZtGyNGjCAoKIiYmBhuvvlmsrOzPa/98ssv6datG/7+/kRERDB06FCKiop4+umn+e9//8u3337r2ebKlStrFU9ubi4TJkwgLCyMgIAARowYwZ49ezzPHzhwgJEjRxIWFkZgYCBdunThhx9+8Lx2/PjxREVF4e/vT7t27Zg3b17tflde5vMbZ4qmR6fTMb7TeHpG9eSRVY+QWpDKLUtuYXLXyUzsMpFwv3BfhyiEOBc4iuH5ON/s+x9HwBx4ytVef/11du/eTdeuXXnmmWcAiIqKIi8vj0svvZTbbruN1157jZKSEh599FFuuOEGli9fTlpaGuPGjePFF19k9OjRFBQU8Ouvv6JpGg899BA7duzAZrN5kovw8Nq9706aNIk9e/awcOFCrFYrjz76KFdccQXbt2/HZDIxZcoU7HY7q1atIjAwkO3btxMUFATAk08+yfbt21m8eDGRkZHs3buXkpKS0zyBZ0aSG1FvukR24YuRX/Cvtf/im73f8P629/nf9v8xtOVQbmh/A31i+qDT6XwdphBC+ExISAhms5mAgABiY2M9y99880169erF888/71n2/vvvk5CQwO7duyksLMTpdHLttdfSsmVLALp16+ZZ19/fn7KysirbPJWKpOb3339n4MCBAHz88cckJCSwYMECrr/+elJTUxkzZoxnX61bt/a8PjU1lV69etG3b18AEhMT635CvESSG1GvAkwBPDPoGS5ufjHvbX2PbUe3sTh5MYuTF9M3pi/T+kyjW1S3U29ICCHqyhSgWlB8te8zsHnzZlasWOFpFTnWvn37GDZsGEOGDKFbt24MHz6cYcOGcd111xEWFnba+9yxYwdGo5H+/ft7lkVERNChQwd27NgBwH333cddd93FTz/9xNChQxkzZgzdu3cH4K677mLMmDFs3LiRYcOGMWrUKE+S1NCk5kY0iCEth/DpVZ/y2VWfMabdGMx6M+sz1nPTDzfx0C8PsTt3t69DFEI0NTqd6hryxeMMW6ULCwsZOXIkSUlJVR579uzhoosuwmAwsHTpUhYvXkznzp1544036NChA8nJyV46eTW77bbb2L9/PzfffDNbt26lb9++vPHGGwCMGDGCAwcO8MADD3DkyBGGDBnCQw89VK/xnIgkN6JBdYnowtMDn+b70d9zdZur0aFjScoSxiwcw/hF4/lq91cUO4p9HaYQQjQYs9mMy+Wqsqx379789ddfJCYm0rZt2yqPwEBVy6PT6Rg0aBAzZ85k06ZNmM1mvvnmmxNu81Q6deqE0+lkzZo1nmVHjx5l165ddO7c2bMsISGBO++8k6+//poHH3yQuXPnep6Liopi4sSJfPTRR8yePZt33323zufDGyS5ET7RLKgZz13wHF+M/ILLWl6GUWdkS/YWnl79NCO+HsE3e77Brbl9HaYQQtS7xMRE1qxZQ0pKCtnZ2bjdbqZMmUJOTg7jxo1j3bp17Nu3jyVLljB58mRcLhdr1qzh+eefZ/369aSmpvL111+TlZVFp06dPNvcsmULu3btIjs7G4fDcco42rVrxzXXXMPtt9/Ob7/9xubNm/nb3/5GfHw811xzDQBTp05lyZIlJCcns3HjRlasWOHZ54wZM/j222/Zu3cvf/31F99//73nuYYmyY3wqQ7hHXh18KssvX4p0/pMIyE4gZzSHGb8MYPxi8bzx+E/OGA7QF5pHi533T6FCCFEY/DQQw9hMBjo3LkzUVFRpKamEhcXx++//47L5WLYsGF069aNqVOnEhoail6vx2q1smrVKq644grat2/PE088wSuvvMKIESMAuP322+nQoQN9+/YlKiqK33//vVaxzJs3jz59+nDVVVcxYMAANE3jhx9+wGQyAeByuZgyZQqdOnXi8ssvp3379rz99tuAai2aPn063bt393SdffbZZ/Vz0k5Bp51jU8jabDZCQkLIz8/HarX6OhxxHIfLwSc7P+Gdze9Q5Ciq8pxZb+bihIsZ1XYUA+MGYtRLPbwQolJpaSnJycm0atUKPz8/X4cjTsPJfod1uX5Ly404q5gMJiZ2mch3o75jdNvRNAtsRoBRjTqwu+0sPbCUKcumMPSLoczZPIdCe6GPIxZCCHG2kY++4qwUFRDFM4Oe8fzscDnYl7+PhfsWsmj/Io6WHuWtpLf4aMdHTOoyiTHtxhDmd/pDIIUQQjQd0i0lGh2H28HSlKXM2TKH5PzKYY8hlhBaWlvSMrglLawtSLQm0ia0DW1D28pkgUKcA6RbqvHzVreUtNyIRsekN3FF6ysYnjicxSmL+b+t/8fevL3kl+WzJWsLW7K2VFm/f7P+PNjnQTpF+KZqXwghRMOS5EY0Wga9gataX8VVra+i2FHMwYKDHLAdILUglZT8FFILUtmWvY01aWsY+/1Yrmx9Je3D2lPsLKbYUUx8UDyXtriU2MDaT08uhBDi7CfJjWgSAkwBdAjvQIfwDlWWHyo4xL83/ZvFyYv5fv/31V43a+0sOoV3omtkV4qdxRTaCzHqjdzc+Wb6xPRpqPCFEEJ4kdTciHPCtuxtfL7rc5xuJwGmAPwMfmzN3sqmzE1o1Pxf4Oo2VzOtzzSyS7JZuG8hPx34CZPeRI+oHvSI6kHfmL60CW0j9TxCnCWk5qbxk5obIeqga2RXukZ2rbb8aMlRVh1axZGiIwSZgggyBbE1eytf7/mahfsW8kPyDzjdziqvOVhw0NMK1DqkNZe3upxhLYfROqS1JDpCCHEWkJYbIWqwJWsLz/75LDtydmDUGxncfDAj24zEYrCwOWszSZlJrM9Yj8NdOaV5kCmIjuEdaRvaliJHEenF6WQWZxIXGMfghMEMThhMXFCcD49KiKZNWm4aP2+13EhyI8QJuNwudubuJD4wnlC/0GrPF9gLWHFwBT8m/8iatDXY3fZTbjPcLxxN03BpLkx6EwnBCSSGJNI8qDl2t538snwK7AWE+4V7hrHHBsZi1Bsx6AzVvhr0hno4ciEaJ0lulMTERKZOncrUqVN9uo3TId1SQtQzg95Al4guJ3w+2BzM1W2u5uo2V+NwO9ift5+dOTvZn78fq9lKbGAsEf4R7Dy6k5WHVrIpcxM5pTlVtnG09ChJWUmnFZ8OHTGBMbQIbkELawuaBTYjOiCaaP9o9TUwmmBTsKerTNM07G47FoOlTvvRNE2624SoR4MHD6Znz57Mnj3bK9tbt26d587h5ypJboTwApPeVONoLYDzm53PpK6TyC/LJ70oHYPOgF6vp8xZxgHbAZJtyRwpPIKfwY8QSwjB5mAyizPZm7eXvXl7ySnNqVb3A6ChkV6UTnpROmvT19YYl5/BD6vZSrGzmCJHERoaAcYAogOiiQqIItQSSrA5mGBTMMHmYILMQVjNVkqcJWzO2syWrC0cLDhI39i+XN3maoa2GEqAKaDaflxuF0XOoirJlBDCezRNw+VyYTSe+rIdFRXVABGd3SS5EaKBhFhCCLGEVFlWl4kF3Zobl9uFU3PidDspc5VxqOAQBwsOklqQSkZRBpnFmWQUZ5BVkkV+WT6lrlJKS0qrbKfYWUyKLYUUW0qt970mbQ1r0tbwrPFZekT1oG1oW9qHtcdmt7EufR0bMjZQ6CjEz+BHTGAMkf6RnhjtLjutQ1ozIG4AA5oNIDYolkJ7IQX2ApyakzBLGCGWEPQ6das7p9tJkaOIYodKyIqcRQQYA2ge3Bx/o3+tYxZC0zRKnCU+2be/0b9Wif6kSZP45Zdf+OWXX3j99dcBSE5OJiUlhUsuuYQffviBJ554gq1bt/LTTz+RkJDAtGnT+PPPPykqKqJTp07MmjWLoUOHerZ5fJeSTqdj7ty5LFq0iCVLlhAfH88rr7zC1VdfXevjSU1N5d5772XZsmXo9Xouv/xy3njjDWJiYgDYvHkzU6dOZf369eh0Otq1a8d//vMf+vbty4EDB7jnnnv47bffsNvtJCYm8tJLL3HFFVfU4YzWjSQ3QjQSep0evUGPCRMAwQQT6R9Jz+ieNa5f6iwlqySLQnshAaYAAk2BmA1mckpyyCrJIrM4k/yyfAodKtE49qHT6egS0YWe0T1pFtiM5anL+W7/dxywHeDPtD/5M+3PmvfpKuWA7QAHbAeqLN+fv5+fU38+4bEZdAYCTYGUOktPWrsU5R9FmF8YJc4SSpwluDU3XSK60CemD71jehNgDKDMVVb5cKqvLs2FxWDxdMlVHH9OaQ4trS3pE9OHDmEd0Ov05JblkmpLpcRZQpR/FFEBUVjN1moXKrfmJr0oHYvBQrhf+CkvZG7NTWZxJocKDhHhH0GiNVFauepZibOE/p/098m+19y0psZWzuO9/vrr7N69m65du/LMM+p+elFRUaSkpADw2GOP8fLLL9O6dWvCwsI4ePAgV1xxBc899xwWi4UPP/yQkSNHsmvXLlq0aHHC/cycOZMXX3yRl156iTfeeIPx48dz4MABwsPDTxmj2+3mmmuuISgoiF9++QWn08mUKVMYO3YsK1euBGD8+PH06tWLd955B4PBQFJSEiaTeq+aMmUKdrudVatWERgYyPbt2wkKCjrlfs+EJDdCNFF+Rj8SghOqLbearSSGJNZpW+3C2nFH9zvYmbOTHTk72JO7hz15e/Az+NE3pi/9mvWjlbUVR0uOkl6cztHSo5j0JvwMfujQsSV7C3+m/cnmrM2eLrZAUyB69BQ4CnBpLmx2W5V9mvVmAkwBBBgDVNLlKCCrJIuskqwq6/16+Fd+Pfxr3U5ODY6N53hmvZmogCiiA6IJsYSQXpROSn4KpS7VKhZsCqaltSWhfqGUOEsodhRjd9k9cyg53U7Si9KrJG7hfuH0ielDojXRk6y5NBdWs5VQSyihfqE0D2pOq5BWxATEUOYq78bMT6bEWeJZx2q24mf0w8/gh1FvJL8sn+ySbHJKcwgwBRAbGEtsQCz+Rn+cmhOHy4FTc2LWmzEbzJ4WswqaplHoKCSrJIsAYwAxATFVkrD8snyS85MJ8wsjOiC6XlvTXG5XjUXzRY4iQP3OGruQkBDMZjMBAQHExlafLf2ZZ57hsssu8/wcHh5Ojx49PD//85//5JtvvmHhwoXcc889J9zPpEmTGDduHADPP/88//73v1m7di2XX375KWNctmwZW7duJTk5mYQE9Z7y4Ycf0qVLF9atW0e/fv1ITU3l4YcfpmPHjgC0a9fO8/rU1FTGjBlDt27dAGjduvUp93mmJLkRQtSKTqejU0Snk3alBZgCSLBWT6gGxg/kzh53UuIsodRZSrA5GKNevf04XA5yy3IptBfiZ/Qj0BRIgDEAk8Hkeb2maeSX5XOw4CAF9gL8Tf4EGAOwu+wkZSWxIWMDW7O24sbtaaE59qHT6bC77Nhddty4ifSPJDogmlBLKLtydrEpcxOFjkLP/mIDYwk0BpJVkoXNbsPutnO48DCHCw9XOS6j3ojL7aLAUcC2o9tOeQ6NOiMxgTFkFWeRU5rD0gNLT/kaULVTZa6yE044WRs6dDW+3qg3YjFYMOvNmPQmbHabJ2kDlYR1jexKqCWUbdnb2J+/v8rrQywhdAzrSJ+YPvSJ6YPdbWdN2hr+TPuTQwWHCDIHEWIOwWqxEmJWXbNWs9Xz+9fr9ASbg4kLiiMuKI4SRwm/Hv6VVYdWsS9vH50iOnFB/AWcF3see3L3sPzgcjZmbMStuWlhbUHH8I60CW1DmCWMcGM4Uc4oCu2FlOpKcbvdLLt+GSa9CZPOhAtXlVY/i8GCv9Efk97kWV7mKsOoN2I2mPEz+GHSm1SdnE6PTqdD07Qq51GHSvzsbjtlrjJKnaXodXocLgd2vV1973aoZFfTCDAFYDaYPa93uBw43U5KnaUUO4oxG8zo0FHmKgOgTdc2FNoLsRgtGHVGioqKePrpp1m0aBFpaWk4nU5KSkpITU096e+/e/fuuDU3OnQEBgZitVrJzMysso5bc2N32atMcQGwY8cOEhISaN68uWeAQefOnQkNDWXHjh3069ePadOmcdttt/G///2PoUOHcv3119OmTRsA7rvvPu666y5++uknhg4dypgxY+jevftJ4z1TZ0Vy89Zbb/HSSy+Rnp5Ojx49eOONNzjvvPNOuP4XX3zBk08+SUpKCu3ateOFF16o1747IYR3+Bv9q33SNxlManRXQPQJX6fT6Qj1C61xSH63qG7c3PnmM4rL5XaxL38fOnQkBCfgZ6wcglrmKiOrOIvskmwyizPJLc0lOiCa1qGtiQ+Kx6W5OGg7SIothQJ7AQGmAPyN/qrVqrzFQ6/TExMQ4xnWb3fZ2Za9jQ0ZG8guyfacF71OT4G9gLyyPHJKczhgO8ChgkOeZMNqttI6pDVB5iDyy/LJLc3FZrd5uuEqznGEXwTh/uEUO4pJL0qn0FF4wsTI6VY1XEUUVVkeZAqixFlCTmkOqw6tqvJcdEA0BfYCSpwl5JflsyZ9DWvS19S4/UJHIelF6af3iwG2H93O9qPbeXfLu9WeO74LtJm5GY+2fRRjkRG9XV9t/eOVOkvJL8uvtrwi2bBRtTWxIrmpjQJ79RbAChVJfLGzmBKHSqiKHEUk5yd71jlSeERthwLPMep1ep556Bn+WPkHTz73JK3atMJsMXPnhDvJLshmT+4enG4nDreDrOIsDtgOYNKrDwlZZVnsOLoDvU6PxWgBHeSW5HK44LCnRa/ibwjA4XaQW5qrfs8O1aq4K3cXbrcbg96AQWfArbnJKs4i1ZbKrQ/eyiVXX8KyJcv44acfeOqpp/jss88YPXo0t912G8OHD2fRokX89NNPzJo1i1deeYV77723VufydPg8uZk/fz7Tpk1jzpw59O/fn9mzZzN8+HB27dpFdHT1N7s//viDcePGMWvWLK666io++eQTRo0axcaNG+natfoMtEIIcSoGvYH2Ye1rfM5isNA8uDnNg5vX+LwRI23D2tI2rG2t92c2mOkd05veMb1Pua7D7eBIoZpB+2S1PW7NjdPtrNIqUKHAXkCZqwyT3oTFYMGgM2B3q5asMleZ58Jmd9sJNgUTGRCJv9GfUmcpu3J3sS17G/ll+XSO6Ez3qO6e+ZoKHAUcKTzC5szNbMjYwIbMDZj0Js6LPY/zm51Ph/AOFDuKybfnYyuzYbPbPHM5OTUnmqbh1tzkl+VzuOgwRwqP4NbcDIgbwEXxF9ElsgtJmUn8dvg3NmVuIiE4gUsSLuGShEsIMAV4ukkPFRxSSYpTjVz0M/phMpnQ6XQ43U7VYqe50el0+Bn8VCJZPmKxxFmC0+1UrTgmf/wNqvuuohXG6Xbi1twAJ01sKpIGP4Mfbs3tSTg1TcOoN3paIkscqvWy1FnZOuZn8UOv6THpTZ5WE4NOdcdV1MpVHMOGNRu4+sarGTBsAADFhcUcPHCQ3gN6Y3dVdnu63C4K7ZWtkRXdwW7NTYmjRHU/2gvJK8urdhwVrWplzjJSbamEtwznyKEjHDp4iGbxzXC6nezatQtbvo3mbZt7ErnIFpGMvX0sY28fy2N3Psa8efMYPXo0AAkJCdx5553ceeedTJ8+nblz5zbt5ObVV1/l9ttvZ/LkyQDMmTOHRYsW8f777/PYY49VW//111/n8ssv5+GHHwZUf+PSpUt58803mTNnToPGLoQQ9c2kN9HS2vKU6+l1+hoTG1BzMgUTXHW7BtMpa1b8jH6ee6kdT6fTYTVbsYZb6RjekbEdx54yxtPR0tqSa9peU+Nzg+IHMSh+kOfnigngEoITqk0A53K70Ol01WqMQF3wa1p+7PMuzaW6ZNCh0+k83VFa+T+jzlgt8azowjp2206309Pq5WfwI9gcTKe2nUhKSsJsMxMaGEpYWBhZVlVb1sLagtDQUE+XUcf2Hfll8S+Mvno0Op2OF/75ApqmEWQOIjEkEaPOiElvItwvnGZBzXC4VLIUHRBNh/AOuNwuSl2l6HQ6As2BRAdEq+SrPPE16tVxGPVGAswBmPQmLhlyCZ26dOKpe57itVdfo8xRxtP3Pc0FF17A8AuHU1xczIzpM7h69NW0TGzJkcNH+GvTX1x/3fUATJ06lREjRtC+fXtyc3NZsWIFnTrVfqTo6fBpcmO329mwYQPTp0/3LNPr9QwdOpTVq1fX+JrVq1czbdq0KsuGDx/OggULaly/rKyMsrLKpjabzVbjekIIIZquk83mfbLEpuL5U61Tk2OToApGvZEwvzDCCPMse+ihh5g4cSKdO3empKSE5OTk4zeFXqfHz+jHv2f/m1tuuYWrhlxFZGQkjz76KKVFpVgMlirJqtloJtyvciSUn1EVnBv1RixGCzp0hFpCiQqoeU4cHSp5bR+uWjQXfbeIe++9l0svubTKUPBQSygBugCK8ou45/Z7yMjIIDIykmuvvZaZM2cC4HK5mDJlCocOHcJqtXL55Zfz2muv1fl81oVPk5vs7GxcLpdnnHyFmJgYdu7cWeNr0tPTa1w/Pb3mPt1Zs2Z5TrAQQghxtmnfvn21D/SJiYk1doMlJiayfPnyKsumTJlS5eeKYeQVatpOXl7eSWM6fhstWrTg22+/rXFds9nMp59+esJtvfHGGyfdV32oeyrayEyfPp38/HzP4+DBg74OSQghhBD1yKctN5GRkRgMBjIyMqosz8jIqHG8P0BsbGyd1rdYLFgsdbuXjhBCCCEaL5+23JjNZvr06cOyZcs8y9xuN8uWLWPAgAE1vmbAgAFV1gdYunTpCdcXQgghxLnF56Olpk2bxsSJE+nbty/nnXces2fPpqioyDN6asKECcTHxzNr1iwA7r//fi6++GJeeeUVrrzySj777DPWr1/Pu+9WnwNBCCGEEOcenyc3Y8eOJSsrixkzZpCenk7Pnj358ccfPUXDqamp6PWVDUwDBw7kk08+4YknnuAf//gH7dq1Y8GCBTLHjRBCCODk89GIs5u3fnc67Rz7K7DZbISEhJCfn4/VavV1OEIIIbzE5XKxe/duoqOjiYiI8HU44jTk5+dz5MgR2rZt67nxZoW6XL993nIjhBBCeIPBYCA0NNRzz6SAgAC583oj4na7ycrKIiAgAKPxzNITSW6EEEI0GRUjZ4+/KaRoHPR6PS1atDjjpFSSGyGEEE2GTqejWbNmREdH43A4Tv0CcVYxm81V6mxPlyQ3QgghmhyDwYDBcOJbLoimrcnPUCyEEEKIc4skN0IIIYRoUiS5EUIIIUSTcs7V3FRM62Oz2XwciRBCCCFqq+K6XZvp+c655KagoACAhIQEH0cihBBCiLoqKCggJCTkpOucczMUu91ujhw5QnBwsNcnd7LZbCQkJHDw4MFzdvZjOQdyDs714wc5ByDnAOQcgHfPgaZpFBQUEBcXd8rh4udcy41er6d58+b1ug+r1XrO/iFXkHMg5+BcP36QcwByDkDOAXjvHJyqxaaCFBQLIYQQokmR5EYIIYQQTYokN15ksVh46qmnsFgsvg7FZ+QcyDk4148f5ByAnAOQcwC+OwfnXEGxEEIIIZo2abkRQgghRJMiyY0QQgghmhRJboQQQgjRpEhyI4QQQogmRZIbL3nrrbdITEzEz8+P/v37s3btWl+HVG9mzZpFv379CA4OJjo6mlGjRrFr164q65SWljJlyhQiIiIICgpizJgxZGRk+Cji+vevf/0LnU7H1KlTPcvOhXNw+PBh/va3vxEREYG/vz/dunVj/fr1nuc1TWPGjBk0a9YMf39/hg4dyp49e3wYsXe5XC6efPJJWrVqhb+/P23atOGf//xnlXvfNLVzsGrVKkaOHElcXBw6nY4FCxZUeb42x5uTk8P48eOxWq2EhoZy6623UlhY2IBHcfpOdvwOh4NHH32Ubt26ERgYSFxcHBMmTODIkSNVttGYjx9O/TdwrDvvvBOdTsfs2bOrLK/vcyDJjRfMnz+fadOm8dRTT7Fx40Z69OjB8OHDyczM9HVo9eKXX35hypQp/PnnnyxduhSHw8GwYcMoKiryrPPAAw/w3Xff8cUXX/DLL79w5MgRrr32Wh9GXX/WrVvHf/7zH7p3715leVM/B7m5uQwaNAiTycTixYvZvn07r7zyCmFhYZ51XnzxRf79738zZ84c1qxZQ2BgIMOHD6e0tNSHkXvPCy+8wDvvvMObb77Jjh07eOGFF3jxxRd54403POs0tXNQVFREjx49eOutt2p8vjbHO378eP766y+WLl3K999/z6pVq7jjjjsa6hDOyMmOv7i4mI0bN/Lkk0+yceNGvv76a3bt2sXVV19dZb3GfPxw6r+BCt988w1//vkncXFx1Z6r93OgiTN23nnnaVOmTPH87HK5tLi4OG3WrFk+jKrhZGZmaoD2yy+/aJqmaXl5eZrJZNK++OILzzo7duzQAG316tW+CrNeFBQUaO3atdOWLl2qXXzxxdr999+vadq5cQ4effRR7YILLjjh8263W4uNjdVeeuklz7K8vDzNYrFon376aUOEWO+uvPJK7ZZbbqmy7Nprr9XGjx+vaVrTPweA9s0333h+rs3xbt++XQO0devWedZZvHixptPptMOHDzdY7N5w/PHXZO3atRqgHThwQNO0pnX8mnbic3Do0CEtPj5e27Ztm9ayZUvttdde8zzXEOdAWm7OkN1uZ8OGDQwdOtSzTK/XM3ToUFavXu3DyBpOfn4+AOHh4QBs2LABh8NR5Zx07NiRFi1aNLlzMmXKFK688soqxwrnxjlYuHAhffv25frrryc6OppevXoxd+5cz/PJycmkp6dXOQchISH079+/yZyDgQMHsmzZMnbv3g3A5s2b+e233xgxYgRwbpyDY9XmeFevXk1oaCh9+/b1rDN06FD0ej1r1qxp8JjrW35+PjqdjtDQUODcOH63283NN9/Mww8/TJcuXao93xDn4Jy7caa3ZWdn43K5iImJqbI8JiaGnTt3+iiqhuN2u5k6dSqDBg2ia9euAKSnp2M2mz3/mSvExMSQnp7ugyjrx2effcbGjRtZt25dtefOhXOwf/9+3nnnHaZNm8Y//vEP1q1bx3333YfZbGbixIme46zp/0ZTOQePPfYYNpuNjh07YjAYcLlcPPfcc4wfPx7gnDgHx6rN8aanpxMdHV3leaPRSHh4eJM7J6WlpTz66KOMGzfOc9PIc+H4X3jhBYxGI/fdd1+NzzfEOZDkRpyRKVOmsG3bNn777Tdfh9KgDh48yP3338/SpUvx8/PzdTg+4Xa76du3L88//zwAvXr1Ytu2bcyZM4eJEyf6OLqG8fnnn/Pxxx/zySef0KVLF5KSkpg6dSpxcXHnzDkQNXM4HNxwww1omsY777zj63AazIYNG3j99dfZuHEjOp3OZ3FIt9QZioyMxGAwVBsFk5GRQWxsrI+iahj33HMP33//PStWrKB58+ae5bGxsdjtdvLy8qqs35TOyYYNG8jMzKR3794YjUaMRiO//PIL//73vzEajcTExDT5c9CsWTM6d+5cZVmnTp1ITU0F8BxnU/6/8fDDD/PYY49x44030q1bN26++WYeeOABZs2aBZwb5+BYtTne2NjYaoMtnE4nOTk5TeacVCQ2Bw4cYOnSpZ5WG2j6x//rr7+SmZlJixYtPO+NBw4c4MEHHyQxMRFomHMgyc0ZMpvN9OnTh2XLlnmWud1uli1bxoABA3wYWf3RNI177rmHb775huXLl9OqVasqz/fp0weTyVTlnOzatYvU1NQmc06GDBnC1q1bSUpK8jz69u3L+PHjPd839XMwaNCgalMA7N69m5YtWwLQqlUrYmNjq5wDm83GmjVrmsw5KC4uRq+v+jZqMBhwu93AuXEOjlWb4x0wYAB5eXls2LDBs87y5ctxu93079+/wWP2torEZs+ePfz8889ERERUeb6pH//NN9/Mli1bqrw3xsXF8fDDD7NkyRKggc6BV8qSz3GfffaZZrFYtA8++EDbvn27dscdd2ihoaFaenq6r0OrF3fddZcWEhKirVy5UktLS/M8iouLPevceeedWosWLbTly5dr69ev1wYMGKANGDDAh1HXv2NHS2la0z8Ha9eu1YxGo/bcc89pe/bs0T7++GMtICBA++ijjzzr/Otf/9JCQ0O1b7/9VtuyZYt2zTXXaK1atdJKSkp8GLn3TJw4UYuPj9e+//57LTk5Wfv666+1yMhI7ZFHHvGs09TOQUFBgbZp0yZt06ZNGqC9+uqr2qZNmzyjgWpzvJdffrnWq1cvbc2aNdpvv/2mtWvXThs3bpyvDqlOTnb8drtdu/rqq7XmzZtrSUlJVd4fy8rKPNtozMevaaf+Gzje8aOlNK3+z4EkN17yxhtvaC1atNDMZrN23nnnaX/++aevQ6o3QI2PefPmedYpKSnR7r77bi0sLEwLCAjQRo8eraWlpfku6AZwfHJzLpyD7777TuvatatmsVi0jh07au+++26V591ut/bkk09qMTExmsVi0YYMGaLt2rXLR9F6n81m0+6//36tRYsWmp+fn9a6dWvt8ccfr3Iha2rnYMWKFTX+/584caKmabU73qNHj2rjxo3TgoKCNKvVqk2ePFkrKCjwwdHU3cmOPzk5+YTvjytWrPBsozEfv6ad+m/geDUlN/V9DnSadsxUmkIIIYQQjZzU3AghhBCiSZHkRgghhBBNiiQ3QgghhGhSJLkRQgghRJMiyY0QQgghmhRJboQQQgjRpEhyI4QQQogmRZIbIYQQQjQpktwIIc5JOp2OBQsW+DoMIUQ9kORGCNHgJk2ahE6nq/a4/PLLfR2aEKIJMPo6ACHEuenyyy9n3rx5VZZZLBYfRSOEaEqk5UYI4RMWi4XY2Ngqj7CwMEB1Gb3zzjuMGDECf39/WrduzZdfflnl9Vu3buXSSy/F39+fiIgI7rjjDgoLC6us8/7779OlSxcsFgvNmjXjnnvuqfJ8dnY2o0ePJiAggHbt2rFw4ULPc7m5uYwfP56oqCj8/f1p165dtWRMCHF2kuRGCHFWevLJJxkzZgybN29m/Pjx3HjjjezYsQOAoqIihg8fTlhYGOvWreOLL77g559/rpK8vPPOO0yZMoU77riDrVu3snDhQtq2bVtlHzNnzuSGG25gy5YtXHHFFYwfP56cnBzP/rdv387ixYvZsWMH77zzDpGRkQ13AoQQp89r9xcXQohamjhxomYwGLTAwMAqj+eee07TNE0DtDvvvLPKa/r376/dddddmqZp2rvvvquFhYVphYWFnucXLVqk6fV6LT09XdM0TYuLi9Mef/zxE8YAaE888YTn58LCQg3QFi9erGmapo0cOVKbPHmydw5YCNGgpOZGCOETl1xyCe+8806VZeHh4Z7vBwwYUOW5AQMGkJSUBMCOHTvo0aMHgYGBnucHDRqE2+1m165d6HQ6jhw5wpAhQ04aQ/fu3T3fBwYGYrVayczMBOCuu+5izJgxbNy4kWHDhjFq1CgGDhx4WscqhGhYktwIIXwiMDCwWjeRt/j7+9dqPZPJVOVnnU7H/7dv96yJRHEUxo9BBIV0vjCdnUwsk3SxsrITtBOZVgySJo1NMh9A1FqwUxQsbEQUSSmInV1K7QRLETKNpliQDdsE4q7u8Py6uRcuZ7rDnf/s93tJUiqV0mq10nA41GQyUTKZ1OPjoyqVysnzAjgtZm4AXKTZbPbHs2makiTTNLVYLLTb7Y770+lUV1dXisViur6+VjQa1dvb248yhEIhWZalVquler2uRqPxo/MA/Bvc3AA4C8dxtF6vv6x5vd7j0G6v19Pd3Z0eHh7Ubrc1n8/VbDYlSblcTq+vr7IsS7Zta7PZqFQqKZ/PKxKJSJJs21ahUFA4HFYqldJ2u9V0OlWpVPpWvpeXF93e3ioej8txHA0Gg2O5AnDZKDcAzmI0GskwjC9rsVhM7+/vkn79ydTtdlUsFmUYhjqdjm5ubiRJgUBA4/FYT09Pur+/VyAQUCaTUbVaPZ5lWZY+Pj5Uq9X0/PysYDCobDb77Xw+n0/lclnL5VJ+v1+JRELdbvcEbw7gb/McDofDuUMAwO88Ho/6/b7S6fS5owD4DzFzAwAAXIVyAwAAXIWZGwAXh6/lAH6CmxsAAOAqlBsAAOAqlBsAAOAqlBsAAOAqlBsAAOAqlBsAAOAqlBsAAOAqlBsAAOAqn5RULXTHWN8LAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Commented out IPython magic to ensure Python compatibility.\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# %matplotlib inline\n",
    "\n",
    "fig, plt = plt.subplots()\n",
    "plt.plot(test_save_acc, label='test acc')\n",
    "#plt.plot(NA, label='without data augmentation')\n",
    "#plt.plot(p3, label='p=3')\n",
    "#plt.plot(p4, label='p=4')\n",
    "plt.plot(test_save_loss, label='test loss')\n",
    "plt.plot(train_save_loss, label='train loss')\n",
    "plt.legend(loc = 'upper right')\n",
    "#plt.title('Loss and Accuracy for '+resnet)\n",
    "#plt.savefig(resnet+'.png')\n",
    "plt.set_xlabel(\"Epochs\")\n",
    "plt.set_ylabel(\"Test Accuracy\")\n",
    "torch.save(model.state_dict(), 'project1_model.pt')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d20f9dbb-cb2e-4e19-9e42-a2172b5d4398",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.8.2 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": "3.8.2"
  },
  "vscode": {
   "interpreter": {
    "hash": "31f2aee4e71d21fbe5cf8b01ff0e069b9275f58929596ceb00d14d90e3e16cd6"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}