siamese_fish_facial.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "siamese_fish_facial.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "06a7fab708fa40d4ab7c0771a46d7041": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_view_name": "HBoxView",
            "_dom_classes": [],
            "_model_name": "HBoxModel",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "box_style": "",
            "layout": "IPY_MODEL_5d01deed5bc24d4c984e5df8f074fa7d",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_1d484deccf3b4c7fafb838cb6062cbcf",
              "IPY_MODEL_241c4892df344527ab2f1191c6dbf359"
            ]
          }
        },
        "5d01deed5bc24d4c984e5df8f074fa7d": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "1d484deccf3b4c7fafb838cb6062cbcf": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_ce503bca810c4c3497d2907595a42a79",
            "_dom_classes": [],
            "description": "100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 102502400,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 102502400,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_3cf37e1540724473861d15d0e9243048"
          }
        },
        "241c4892df344527ab2f1191c6dbf359": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_90707766238743989629e329a53da2b2",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 97.8M/97.8M [00:01&lt;00:00, 65.1MB/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_faed169eb4944c6ab02f1b77d8d3795a"
          }
        },
        "ce503bca810c4c3497d2907595a42a79": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "ProgressStyleModel",
            "description_width": "initial",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "bar_color": null,
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "3cf37e1540724473861d15d0e9243048": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "90707766238743989629e329a53da2b2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "DescriptionStyleModel",
            "description_width": "",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "faed169eb4944c6ab02f1b77d8d3795a": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        }
      }
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cjY-OGIW71JA",
        "outputId": "0a6ec8c9-685c-44c7-f266-363a54e8ac29"
      },
      "source": [
        "%matplotlib inline\n",
        "! realpath IKT450\n",
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')\n",
        "!ls /content/gdrive/MyDrive\n",
        "! git clone https://github.com/JoarVarpe/IKT450.git\n",
        "!nvidia-smi\n",
        "\n",
        "from collections import Counter\n",
        "import torchvision\n",
        "import torchvision.datasets as dset\n",
        "import torchvision.transforms as transforms\n",
        "from torch.utils.data import DataLoader,Dataset\n",
        "import matplotlib.pyplot as plt\n",
        "import torchvision.utils\n",
        "import numpy as np\n",
        "import random\n",
        "import glob\n",
        "from PIL import Image\n",
        "import torch\n",
        "from torch.autograd import Variable\n",
        "import PIL.ImageOps    \n",
        "import torch.nn as nn\n",
        "from torch import optim\n",
        "import torch.nn.functional as F\n",
        "import torchvision\n",
        "\n",
        "random.seed(7)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/IKT450\n",
            "Mounted at /content/gdrive\n",
            " 20201124_025526.jpg\t\t\t     res_50_100+0.pt    res_50_1.pt\n",
            " Candito+Linear+Program.pdf\t\t     res_50_100+10.pt   res_50_20.pt\n",
            " classifier0.pt\t\t\t\t     res_50_100+15.pt   res_50_25.pt\n",
            " classifier12.pt\t\t\t     res_50_100+1.pt    res_50_2.pt\n",
            " classifier13.pt\t\t\t     res_50_100+20.pt   res_50_30.pt\n",
            " classifier14.pt\t\t\t     res_50_100+25.pt   res_50_35.pt\n",
            " classifier16.pt\t\t\t     res_50_100+30.pt   res_50_38.pt\n",
            " classifier1.pt\t\t\t\t     res_50_100+35.pt   res_50_39.pt\n",
            " classifier2.pt\t\t\t\t     res_50_100+40.pt   res_50_3.pt\n",
            " classifier3.pt\t\t\t\t     res_50_100+45.pt   res_50_40.pt\n",
            " classifier4.pt\t\t\t\t     res_50_100+50.pt   res_50_45.pt\n",
            " classifier5.pt\t\t\t\t     res_50_100+55.pt   res_50_48.pt\n",
            " classifier6.pt\t\t\t\t     res_50_100+5.pt    res_50_49.pt\n",
            " classifier7.pt\t\t\t\t     res_50_100+60.pt   res_50_50.pt\n",
            " classifier_body0.pt\t\t\t     res_50_100+65.pt   res_50_55.pt\n",
            " classifier_body11.pt\t\t\t     res_50_100+70.pt   res_50_5.pt\n",
            " classifier_body1.pt\t\t\t     res_50_100+75.pt   res_50_60.pt\n",
            " classifier_body2.pt\t\t\t     res_50_100+7.pt    res_50_65.pt\n",
            " classifier_body3.pt\t\t\t     res_50_100+80.pt   res_50_70.pt\n",
            " classifier_body5.pt\t\t\t     res_50_100+85.pt   res_50_75.pt\n",
            "'Colab Notebooks'\t\t\t     res_50_100+90.pt   res_50_80.pt\n",
            "'Husleieavtale Joar Varpe 2-1 signert.pdf'   res_50_100+95.pt   res_50_84.pt\n",
            "'Mal nr 1 Joar.gdoc'\t\t\t     res_50_10.pt       res_50_85.pt\n",
            " node_wow.gslides\t\t\t     res_50_15.pt       res_50_88.pt\n",
            " R300129866-123417.pdf\t\t\t     res_50_17.pt       res_50_90.pt\n",
            " res_50_0.pt\t\t\t\t     res_50_19.pt       res_50_95.pt\n",
            "Cloning into 'IKT450'...\n",
            "remote: Enumerating objects: 5253, done.\u001b[K\n",
            "remote: Total 5253 (delta 0), reused 0 (delta 0), pack-reused 5253\u001b[K\n",
            "Receiving objects: 100% (5253/5253), 356.60 MiB | 34.94 MiB/s, done.\n",
            "Checking out files: 100% (4227/4227), done.\n",
            "NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Qd0GkM4ihaC6"
      },
      "source": [
        "def imshow(img,text=None,should_save=False):\n",
        "    npimg = img.numpy()\n",
        "    plt.axis(\"off\")\n",
        "    if text:\n",
        "        plt.text(75, 8, text, style='italic',fontweight='bold',\n",
        "            bbox={'facecolor':'white', 'alpha':0.8, 'pad':10})\n",
        "    plt.imshow(np.transpose(npimg, (1, 2, 0)))\n",
        "    plt.show()    \n",
        "\n",
        "def show_plot(iteration,loss):\n",
        "    plt.plot(iteration,loss)\n",
        "    plt.show()"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TSBM0-xbhdc_"
      },
      "source": [
        "class Config():\n",
        "    training_dir = \"/content/IKT450/0_8/train/\"\n",
        "    testing_dir = \"/content/IKT450/0_8/test/\"\n",
        "    training_dir_body = \"/content/IKT450/0_8_body/train/\"\n",
        "    testing_dir_body = \"/content/IKT450/0_8_body/test/\"\n",
        "    train_batch_size = 64\n",
        "    train_number_epochs = 100"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ly92j4eUhgez"
      },
      "source": [
        "class SiameseNetworkDataset(Dataset):\n",
        "    \n",
        "    def __init__(self,imageFolderDataset, percent_same=None, transform=None):\n",
        "        self.imageFolderDataset = imageFolderDataset    \n",
        "        self.transform = transform\n",
        "        self.percent_same = percent_same\n",
        "        print(self.percent_same)\n",
        "        \n",
        "        \n",
        "    def __getitem__(self,index):\n",
        "        img0_tuple = self.imageFolderDataset.imgs[index]\n",
        "        #we need to make sure approx 50% of images are in the same class\n",
        "        if self.percent_same != None:\n",
        "          should_get_same_class = self.percent_same\n",
        "        else:\n",
        "          \n",
        "          should_get_same_class = random.randint(0,1) \n",
        "\n",
        "        if should_get_same_class:\n",
        "            while True:\n",
        "                #keep looping till the same class image is found\n",
        "                img1_tuple = random.choice(self.imageFolderDataset.imgs) \n",
        "                if img0_tuple[1]==img1_tuple[1]:\n",
        "                    break\n",
        "        else:\n",
        "            while True:\n",
        "                #keep looping till a different class image is found\n",
        "                \n",
        "                img1_tuple = random.choice(self.imageFolderDataset.imgs) \n",
        "                if img0_tuple[1] !=img1_tuple[1]:\n",
        "                    break\n",
        "\n",
        "        img0 = Image.open(img0_tuple[0])\n",
        "        img1 = Image.open(img1_tuple[0])\n",
        "        img0 = img0.convert(\"L\")\n",
        "        img1 = img1.convert(\"L\")\n",
        "        \n",
        "\n",
        "        if self.transform is not None:\n",
        "            img0 = self.transform(img0)\n",
        "            img1 = self.transform(img1)\n",
        "        \n",
        "        return img0, img1 , torch.from_numpy(np.array([int(img1_tuple[1]!=img0_tuple[1])],dtype=np.float32))\n",
        "    \n",
        "    def __len__(self):\n",
        "        return len(self.imageFolderDataset.imgs)"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1tMccP37hmYB"
      },
      "source": [
        "folder_dataset = dset.ImageFolder(root=Config.training_dir)\n",
        "folder_dataset_body = dset.ImageFolder(root=Config.training_dir_body)"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xMsFWgbVhq1s",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "09495ec2-4b66-47ba-c65b-d9d3f0563af8"
      },
      "source": [
        "siamese_dataset = SiameseNetworkDataset(imageFolderDataset=folder_dataset,\n",
        "                                        transform=transforms.Compose([transforms.Resize((100,100)),\n",
        "                                                                      transforms.ToTensor()\n",
        "                                                                      ])\n",
        "                                       )\n",
        "\n",
        "siamese_dataset_body= SiameseNetworkDataset(imageFolderDataset=folder_dataset_body,\n",
        "                                        transform=transforms.Compose([transforms.Resize((100,100)),\n",
        "                                                                      transforms.ToTensor()\n",
        "                                                                      ])\n",
        "                                       )"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "None\n",
            "None\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GlO3MVfOqjPb"
      },
      "source": [
        "class SiameseNetwork(nn.Module):\n",
        "    #from https://hackernoon.com/facial-similarity-with-siamese-networks-in-pytorch-9642aa9db2f7\n",
        "    #Output size after convolution\n",
        "    #w = initial img size, k = kernel size, s = stride, p = padding\n",
        "    #((w - k + 2P)/s) + 1\n",
        "    \n",
        "    #input here is 100 px x 100 px \n",
        "    \n",
        "    def __init__(self):\n",
        "        super(SiameseNetwork, self).__init__()\n",
        "        self.res_net = torch.hub.load('pytorch/vision:v0.6.0', 'resnet50', pretrained=True)\n",
        "        self.fc1 = nn.Linear(1000, 128)\n",
        "        self.cnn1 = nn.Sequential(\n",
        "            nn.ReflectionPad2d(1),\n",
        "            nn.Conv2d(1, 4, kernel_size=3),\n",
        "            nn.ReLU(inplace=True),\n",
        "            nn.BatchNorm2d(4),\n",
        "            \n",
        "            nn.ReflectionPad2d(1),\n",
        "            nn.Conv2d(4, 8, kernel_size=3),\n",
        "            nn.ReLU(inplace=True),\n",
        "            nn.BatchNorm2d(8),\n",
        "\n",
        "\n",
        "            nn.ReflectionPad2d(1),\n",
        "            nn.Conv2d(8, 8, kernel_size=3),\n",
        "            nn.ReLU(inplace=True),\n",
        "            nn.BatchNorm2d(8),\n",
        "\n",
        "\n",
        "        )\n",
        "\n",
        "        self.fc1 = nn.Sequential(\n",
        "            nn.Linear(8*100*100, 500),\n",
        "            nn.ReLU(inplace=True),\n",
        "\n",
        "            nn.Linear(500, 500),\n",
        "            nn.ReLU(inplace=True),\n",
        "\n",
        "            nn.Linear(500, 5))\n",
        "\n",
        "    def forward_once(self, x):\n",
        "        output = self.res_net(x)\n",
        "        output = output.view(output.size()[0], -1)\n",
        "        output = self.fc1(output)\n",
        "        return output\n",
        "\n",
        "    def forward(self, input1, input2):\n",
        "        output1 = self.forward_once(input1)\n",
        "        output2 = self.forward_once(input2)\n",
        "        return output1, output2"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iGKSs9hcqnKI"
      },
      "source": [
        "class ContrastiveLoss(torch.nn.Module):\n",
        "    \"\"\"\n",
        "    Contrastive loss function.\n",
        "    Based on: http://yann.lecun.com/exdb/publis/pdf/hadsell-chopra-lecun-06.pdf\n",
        "    \"\"\"\n",
        "\n",
        "    def __init__(self, margin=2.0):\n",
        "        super(ContrastiveLoss, self).__init__()\n",
        "        self.margin = margin\n",
        "\n",
        "    def forward(self, output1, output2, label):\n",
        "        euclidean_distance = F.pairwise_distance(output1, output2, keepdim = True)\n",
        "        # d = euclidean_distance.detach().numpy()\n",
        "        # if (d < 0).sum() > 0:\n",
        "          # print(\"Got euclidean distance: \", d)\n",
        "        loss_contrastive = torch.mean((1-label) * torch.pow(euclidean_distance, 2) +\n",
        "                                      (label) * torch.pow(torch.clamp(self.margin - euclidean_distance, min=0.0), 2))\n",
        "\n",
        "\n",
        "        return loss_contrastive"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xwfNDV6sqp1w",
        "outputId": "378b3245-8762-4255-c3e5-5159cc0646f4"
      },
      "source": [
        "preprocess = transforms.Compose([\n",
        "    transforms.Resize(256),\n",
        "    transforms.CenterCrop(224),\n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),\n",
        "])\n",
        "\n",
        "preprocess = transforms.Compose([transforms.Resize((100,100)),\n",
        "                                                                      transforms.ToTensor()\n",
        "                                                                      ])\n",
        "\n",
        "folder_dataset_body_test = dset.ImageFolder(root=Config.testing_dir_body)\n",
        "siamese_dataset_body_test = SiameseNetworkDataset(imageFolderDataset=folder_dataset_body_test,\n",
        "                                        transform=preprocess)\n",
        "\n",
        "train_dataloader_body = DataLoader(siamese_dataset_body,\n",
        "                        shuffle=True,\n",
        "                        num_workers=8,\n",
        "                        batch_size=Config.train_batch_size)\n",
        "\n",
        "test_dataloader_body = DataLoader(siamese_dataset_body_test,\n",
        "                        shuffle=False,\n",
        "                        num_workers=8,\n",
        "                        batch_size=Config.train_batch_size)\n",
        "\n",
        "train_count_body = len(glob.glob(Config.training_dir_body + \"/**/*.jpg\"))\n",
        "test_count_body = len(glob.glob(Config.testing_dir_body + \"/**/*.jpg\"))\n",
        "print(train_count_body, test_count_body)\n",
        "\n",
        "folder_dataset_test = dset.ImageFolder(root=Config.testing_dir)\n",
        "siamese_dataset_test = SiameseNetworkDataset(imageFolderDataset=folder_dataset_test,\n",
        "                                        transform=preprocess)\n",
        "\n",
        "train_dataloader = DataLoader(siamese_dataset,\n",
        "                        shuffle=True,\n",
        "                        num_workers=8,\n",
        "                        batch_size=Config.train_batch_size)\n",
        "\n",
        "test_dataloader = DataLoader(siamese_dataset_test,\n",
        "                        shuffle=False,\n",
        "                        num_workers=8,\n",
        "                        batch_size=Config.train_batch_size)\n",
        "\n",
        "train_count = len(glob.glob(Config.training_dir + \"/**/*.jpg\"))\n",
        "test_count = len(glob.glob(Config.testing_dir + \"/**/*.jpg\"))\n",
        "print(train_count, test_count)\n",
        "\n"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "None\n",
            "1687 426\n",
            "None\n",
            "1690 423\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457,
          "referenced_widgets": [
            "06a7fab708fa40d4ab7c0771a46d7041",
            "5d01deed5bc24d4c984e5df8f074fa7d",
            "1d484deccf3b4c7fafb838cb6062cbcf",
            "241c4892df344527ab2f1191c6dbf359",
            "ce503bca810c4c3497d2907595a42a79",
            "3cf37e1540724473861d15d0e9243048",
            "90707766238743989629e329a53da2b2",
            "faed169eb4944c6ab02f1b77d8d3795a"
          ]
        },
        "id": "aGY2iHMOqy0r",
        "outputId": "6af62576-3ab7-4b21-cee9-911e52c388b0"
      },
      "source": [
        "counter = []\n",
        "loss_history = [] \n",
        "iteration_number= 0\n",
        "net = SiameseNetwork().cuda()\n",
        "criterion = ContrastiveLoss()\n",
        "optimizer = optim.Adam(net.parameters(),lr = 0.0005 )\n",
        "best_loss = 1e10\n",
        "for epoch in range(0,Config.train_number_epochs):\n",
        "    \n",
        "    total_train = 0\n",
        "    correct_train = 0\n",
        "    loss = 1.0\n",
        "    net.train()\n",
        "    for i, data in enumerate(train_dataloader,0):\n",
        "        img0, img1 , label = data\n",
        "        img0, img1 , label = img0.cuda(), img1.cuda() , label.cuda()\n",
        "        optimizer.zero_grad()\n",
        "        output1,output2 = net(img0,img1)\n",
        "        loss_contrastive = criterion(output1,output2,label)\n",
        "        loss_contrastive.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "        loss = loss_contrastive.item()\n",
        "        \n",
        "        #accuracy\n",
        "\n",
        "    net.eval()\n",
        "    total_loss = 0\n",
        "    for i, data in enumerate(test_dataloader,0):\n",
        "        with torch.no_grad():\n",
        "          img0, img1 , label = data\n",
        "          img0, img1 , label = img0.cuda(), img1.cuda() , label.cuda()\n",
        "          optimizer.zero_grad()\n",
        "          output1,output2 = net(img0,img1)\n",
        "          loss_contrastive = criterion(output1,output2,label)\n",
        "          total_loss += loss_contrastive.item()\n",
        "        #print\n",
        "\n",
        "    print(\"Epoch number {}\\n Current loss {}\\n\".format(epoch,loss_contrastive.item()))\n",
        "    print(\"Current Accuracy: Not Implemented, find thresh first\")\n",
        "    iteration_number +=1\n",
        "    counter.append(iteration_number)\n",
        "    loss_history.append(loss_contrastive.item())\n",
        "\n",
        "    # if total_loss < best_loss:\n",
        "    #   print(\"saving model :)))))\")\n",
        "    #   best_loss = total_loss\n",
        "    #   model_save_name = 'classifier{}.pt'.format(epoch)\n",
        "    #   path = F\"/content/gdrive/MyDrive/{model_save_name}\" \n",
        "    #   torch.save(net.state_dict(), path)\n",
        "    #   print(\"Save is done Ok Ok Ok.\")\n",
        "    #   print(\"Total_loss: \", total_loss)\n",
        "    # else:\n",
        "    #   print(\"Loss {} is not better than best loss {}\".format(total_loss, best_loss))\n",
        "\n",
        "\n",
        "show_plot(counter,loss_history)"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading: \"https://github.com/pytorch/vision/archive/v0.6.0.zip\" to /root/.cache/torch/hub/v0.6.0.zip\n",
            "Downloading: \"https://download.pytorch.org/models/resnet50-19c8e357.pth\" to /root/.cache/torch/hub/checkpoints/resnet50-19c8e357.pth\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "06a7fab708fa40d4ab7c0771a46d7041",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, max=102502400.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "RuntimeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-14-c10fe0f3d4c5>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0mloss_history\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0miteration_number\u001b[0m\u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mnet\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mSiameseNetwork\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0mcriterion\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mContrastiveLoss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0moptimizer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0moptim\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mAdam\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnet\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparameters\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mlr\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0.0005\u001b[0m \u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36mcuda\u001b[0;34m(self, device)\u001b[0m\n\u001b[1;32m    461\u001b[0m             \u001b[0mModule\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    462\u001b[0m         \"\"\"\n\u001b[0;32m--> 463\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;32mlambda\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    464\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    465\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mcpu\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mT\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mT\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_apply\u001b[0;34m(self, fn)\u001b[0m\n\u001b[1;32m    357\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_apply\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    358\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mchildren\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 359\u001b[0;31m             \u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfn\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    360\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    361\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0mcompute_should_use_set_data\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtensor_applied\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_apply\u001b[0;34m(self, fn)\u001b[0m\n\u001b[1;32m    357\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_apply\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    358\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mchildren\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 359\u001b[0;31m             \u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfn\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    360\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    361\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0mcompute_should_use_set_data\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtensor_applied\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_apply\u001b[0;34m(self, fn)\u001b[0m\n\u001b[1;32m    379\u001b[0m                 \u001b[0;31m# `with torch.no_grad():`\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    380\u001b[0m                 \u001b[0;32mwith\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mno_grad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 381\u001b[0;31m                     \u001b[0mparam_applied\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparam\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    382\u001b[0m                 \u001b[0mshould_use_set_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcompute_should_use_set_data\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparam\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparam_applied\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    383\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mshould_use_set_data\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m<lambda>\u001b[0;34m(t)\u001b[0m\n\u001b[1;32m    461\u001b[0m             \u001b[0mModule\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    462\u001b[0m         \"\"\"\n\u001b[0;32m--> 463\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;32mlambda\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    464\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    465\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mcpu\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mT\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mT\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/cuda/__init__.py\u001b[0m in \u001b[0;36m_lazy_init\u001b[0;34m()\u001b[0m\n\u001b[1;32m    170\u001b[0m         \u001b[0;31m# This function throws if there's a driver initialization error, no GPUs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    171\u001b[0m         \u001b[0;31m# are found or any other error occurs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 172\u001b[0;31m         \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_C\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_cuda_init\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    173\u001b[0m         \u001b[0;31m# Some of the queued calls may reentrantly call _lazy_init();\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    174\u001b[0m         \u001b[0;31m# we need to just return without initializing in that case.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mRuntimeError\u001b[0m: No CUDA GPUs are available"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V94gE0kEqs0N"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2Sj5gJSnqwYr",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "a7efec3a-7ebc-42e0-f8df-8fe870252bf4"
      },
      "source": [
        "counter = []\n",
        "loss_history = [] \n",
        "iteration_number= 0\n",
        "net = SiameseNetwork().cuda()\n",
        "criterion = ContrastiveLoss()\n",
        "optimizer = optim.Adam(net.parameters(),lr = 0.0005 )\n",
        "best_loss = 1e10\n",
        "for epoch in range(0,Config.train_number_epochs):\n",
        "    \n",
        "    total_train = 0\n",
        "    correct_train = 0\n",
        "    loss = 1.0\n",
        "    net.train()\n",
        "    for i, data in enumerate(train_dataloader_body,0):\n",
        "        img0, img1 , label = data\n",
        "        img0, img1 , label = img0.cuda(), img1.cuda() , label.cuda()\n",
        "        optimizer.zero_grad()\n",
        "        output1,output2 = net(img0,img1)\n",
        "        loss_contrastive = criterion(output1,output2,label)\n",
        "        loss_contrastive.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "        loss = loss_contrastive.item()\n",
        "        \n",
        "        #accuracy\n",
        "\n",
        "    net.eval()\n",
        "    total_loss = 0\n",
        "    for i, data in enumerate(test_dataloader_body,0):\n",
        "        with torch.no_grad():\n",
        "          img0, img1 , label = data\n",
        "          img0, img1 , label = img0.cuda(), img1.cuda() , label.cuda()\n",
        "          optimizer.zero_grad()\n",
        "          output1,output2 = net(img0,img1)\n",
        "          loss_contrastive = criterion(output1,output2,label)\n",
        "          total_loss += loss_contrastive.item()\n",
        "        #print\n",
        "\n",
        "    print(\"Epoch number {}\\n Current loss {}\\n\".format(epoch,loss_contrastive.item()))\n",
        "    print(\"Current Accuracy: Not Implemented, find thresh first\")\n",
        "    iteration_number +=1\n",
        "    counter.append(iteration_number)\n",
        "    loss_history.append(loss_contrastive.item())\n",
        "\n",
        "    if total_loss < best_loss:\n",
        "      print(\"saving model :)))))\")\n",
        "      best_loss = total_loss\n",
        "      model_save_name = 'classifier_body{}.pt'.format(epoch)\n",
        "      path = F\"/content/gdrive/MyDrive/{model_save_name}\" \n",
        "      torch.save(net.state_dict(), path)\n",
        "      print(\"Save is done Ok Ok Ok.\")\n",
        "      print(\"Total_loss: \", total_loss)\n",
        "    else:\n",
        "      print(\"Loss {} is not better than best loss {}\".format(total_loss, best_loss))\n",
        "\n",
        "\n",
        "show_plot(counter,loss_history)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch number 0\n",
            " Current loss 1.211300253868103\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "saving model :)))))\n",
            "Save is done Ok Ok Ok.\n",
            "Total_loss:  7.358695030212402\n",
            "Epoch number 1\n",
            " Current loss 1.1253553628921509\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "saving model :)))))\n",
            "Save is done Ok Ok Ok.\n",
            "Total_loss:  6.767774760723114\n",
            "Epoch number 2\n",
            " Current loss 0.6040231585502625\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "saving model :)))))\n",
            "Save is done Ok Ok Ok.\n",
            "Total_loss:  5.469441056251526\n",
            "Epoch number 3\n",
            " Current loss 0.8224618434906006\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "saving model :)))))\n",
            "Save is done Ok Ok Ok.\n",
            "Total_loss:  5.329757750034332\n",
            "Epoch number 4\n",
            " Current loss 0.7855862379074097\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.5618703961372375 is not better than best loss 5.329757750034332\n",
            "Epoch number 5\n",
            " Current loss 0.7384808659553528\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "saving model :)))))\n",
            "Save is done Ok Ok Ok.\n",
            "Total_loss:  4.986589133739471\n",
            "Epoch number 6\n",
            " Current loss 0.752373218536377\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.12444007396698 is not better than best loss 4.986589133739471\n",
            "Epoch number 7\n",
            " Current loss 1.2211894989013672\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.859697580337524 is not better than best loss 4.986589133739471\n",
            "Epoch number 8\n",
            " Current loss 0.6374872326850891\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.090625107288361 is not better than best loss 4.986589133739471\n",
            "Epoch number 9\n",
            " Current loss 0.6431881785392761\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.240486562252045 is not better than best loss 4.986589133739471\n",
            "Epoch number 10\n",
            " Current loss 1.6502400636672974\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.293642103672028 is not better than best loss 4.986589133739471\n",
            "Epoch number 11\n",
            " Current loss 0.6435938477516174\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "saving model :)))))\n",
            "Save is done Ok Ok Ok.\n",
            "Total_loss:  4.720241725444794\n",
            "Epoch number 12\n",
            " Current loss 0.6146579384803772\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.000130236148834 is not better than best loss 4.720241725444794\n",
            "Epoch number 13\n",
            " Current loss 0.7599909901618958\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.148191154003143 is not better than best loss 4.720241725444794\n",
            "Epoch number 14\n",
            " Current loss 1.267630696296692\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.044652223587036 is not better than best loss 4.720241725444794\n",
            "Epoch number 15\n",
            " Current loss 0.8083236217498779\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.532421171665192 is not better than best loss 4.720241725444794\n",
            "Epoch number 16\n",
            " Current loss 0.6729697585105896\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.511082887649536 is not better than best loss 4.720241725444794\n",
            "Epoch number 17\n",
            " Current loss 0.7968661785125732\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.991964340209961 is not better than best loss 4.720241725444794\n",
            "Epoch number 18\n",
            " Current loss 0.7468221783638\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.304822385311127 is not better than best loss 4.720241725444794\n",
            "Epoch number 19\n",
            " Current loss 0.951432466506958\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.074324727058411 is not better than best loss 4.720241725444794\n",
            "Epoch number 20\n",
            " Current loss 0.5359262824058533\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.34843373298645 is not better than best loss 4.720241725444794\n",
            "Epoch number 21\n",
            " Current loss 1.2389072179794312\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 4.918285608291626 is not better than best loss 4.720241725444794\n",
            "Epoch number 22\n",
            " Current loss 0.7801105976104736\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.023428857326508 is not better than best loss 4.720241725444794\n",
            "Epoch number 23\n",
            " Current loss 0.9329195618629456\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.964715898036957 is not better than best loss 4.720241725444794\n",
            "Epoch number 24\n",
            " Current loss 0.9109846949577332\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.618227899074554 is not better than best loss 4.720241725444794\n",
            "Epoch number 25\n",
            " Current loss 0.9747149348258972\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.369483053684235 is not better than best loss 4.720241725444794\n",
            "Epoch number 26\n",
            " Current loss 0.7009304761886597\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.166028797626495 is not better than best loss 4.720241725444794\n",
            "Epoch number 27\n",
            " Current loss 0.743273138999939\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.297712504863739 is not better than best loss 4.720241725444794\n",
            "Epoch number 28\n",
            " Current loss 1.1721919775009155\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.302236497402191 is not better than best loss 4.720241725444794\n",
            "Epoch number 29\n",
            " Current loss 0.6894263625144958\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.904631733894348 is not better than best loss 4.720241725444794\n",
            "Epoch number 30\n",
            " Current loss 0.4911867380142212\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.343018710613251 is not better than best loss 4.720241725444794\n",
            "Epoch number 31\n",
            " Current loss 0.497427374124527\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.1986179649829865 is not better than best loss 4.720241725444794\n",
            "Epoch number 32\n",
            " Current loss 0.8977314233779907\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.320931613445282 is not better than best loss 4.720241725444794\n",
            "Epoch number 33\n",
            " Current loss 1.2205564975738525\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.831489026546478 is not better than best loss 4.720241725444794\n",
            "Epoch number 34\n",
            " Current loss 0.6598514318466187\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.209815204143524 is not better than best loss 4.720241725444794\n",
            "Epoch number 35\n",
            " Current loss 0.4672660827636719\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.050000965595245 is not better than best loss 4.720241725444794\n",
            "Epoch number 36\n",
            " Current loss 0.6376231908798218\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.608995318412781 is not better than best loss 4.720241725444794\n",
            "Epoch number 37\n",
            " Current loss 0.8293817043304443\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.5626057386398315 is not better than best loss 4.720241725444794\n",
            "Epoch number 38\n",
            " Current loss 0.5539815425872803\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.610308885574341 is not better than best loss 4.720241725444794\n",
            "Epoch number 39\n",
            " Current loss 1.1922061443328857\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.260360598564148 is not better than best loss 4.720241725444794\n",
            "Epoch number 40\n",
            " Current loss 0.8390476703643799\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.956066608428955 is not better than best loss 4.720241725444794\n",
            "Epoch number 41\n",
            " Current loss 0.8979753255844116\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.884509861469269 is not better than best loss 4.720241725444794\n",
            "Epoch number 42\n",
            " Current loss 1.3548572063446045\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.599437475204468 is not better than best loss 4.720241725444794\n",
            "Epoch number 43\n",
            " Current loss 1.226853370666504\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.05245578289032 is not better than best loss 4.720241725444794\n",
            "Epoch number 44\n",
            " Current loss 0.7236697673797607\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.084579944610596 is not better than best loss 4.720241725444794\n",
            "Epoch number 45\n",
            " Current loss 0.9220654964447021\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.823273777961731 is not better than best loss 4.720241725444794\n",
            "Epoch number 46\n",
            " Current loss 1.7686527967453003\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.220443546772003 is not better than best loss 4.720241725444794\n",
            "Epoch number 47\n",
            " Current loss 0.7097635269165039\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.573551952838898 is not better than best loss 4.720241725444794\n",
            "Epoch number 48\n",
            " Current loss 0.8755154609680176\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.22073745727539 is not better than best loss 4.720241725444794\n",
            "Epoch number 49\n",
            " Current loss 1.269531488418579\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.374281525611877 is not better than best loss 4.720241725444794\n",
            "Epoch number 50\n",
            " Current loss 1.3871113061904907\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 9.079195082187653 is not better than best loss 4.720241725444794\n",
            "Epoch number 51\n",
            " Current loss 0.5537486672401428\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.5370959639549255 is not better than best loss 4.720241725444794\n",
            "Epoch number 52\n",
            " Current loss 0.5482946038246155\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.0567678809165955 is not better than best loss 4.720241725444794\n",
            "Epoch number 53\n",
            " Current loss 1.8809404373168945\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.325153470039368 is not better than best loss 4.720241725444794\n",
            "Epoch number 54\n",
            " Current loss 0.5338197946548462\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.095134496688843 is not better than best loss 4.720241725444794\n",
            "Epoch number 55\n",
            " Current loss 1.3399733304977417\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.921728849411011 is not better than best loss 4.720241725444794\n",
            "Epoch number 56\n",
            " Current loss 1.1103545427322388\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.046229362487793 is not better than best loss 4.720241725444794\n",
            "Epoch number 57\n",
            " Current loss 0.6649426817893982\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.763745427131653 is not better than best loss 4.720241725444794\n",
            "Epoch number 58\n",
            " Current loss 1.029518723487854\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.459087699651718 is not better than best loss 4.720241725444794\n",
            "Epoch number 59\n",
            " Current loss 0.7658537030220032\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.984558999538422 is not better than best loss 4.720241725444794\n",
            "Epoch number 60\n",
            " Current loss 1.177369236946106\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.000925183296204 is not better than best loss 4.720241725444794\n",
            "Epoch number 61\n",
            " Current loss 1.333686351776123\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.930673718452454 is not better than best loss 4.720241725444794\n",
            "Epoch number 62\n",
            " Current loss 0.7023392915725708\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.748497486114502 is not better than best loss 4.720241725444794\n",
            "Epoch number 63\n",
            " Current loss 2.003831386566162\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 9.08639407157898 is not better than best loss 4.720241725444794\n",
            "Epoch number 64\n",
            " Current loss 2.036019802093506\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.15178644657135 is not better than best loss 4.720241725444794\n",
            "Epoch number 65\n",
            " Current loss 0.5591673851013184\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.833233118057251 is not better than best loss 4.720241725444794\n",
            "Epoch number 66\n",
            " Current loss 0.9649456739425659\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.647343099117279 is not better than best loss 4.720241725444794\n",
            "Epoch number 67\n",
            " Current loss 1.3028396368026733\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 9.121308088302612 is not better than best loss 4.720241725444794\n",
            "Epoch number 68\n",
            " Current loss 1.1261366605758667\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.853416264057159 is not better than best loss 4.720241725444794\n",
            "Epoch number 69\n",
            " Current loss 0.8913645148277283\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.4926228523254395 is not better than best loss 4.720241725444794\n",
            "Epoch number 70\n",
            " Current loss 1.5878180265426636\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.611929655075073 is not better than best loss 4.720241725444794\n",
            "Epoch number 71\n",
            " Current loss 0.9466652870178223\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.384243845939636 is not better than best loss 4.720241725444794\n",
            "Epoch number 72\n",
            " Current loss 1.4684392213821411\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.570538878440857 is not better than best loss 4.720241725444794\n",
            "Epoch number 73\n",
            " Current loss 0.6086542010307312\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.517834544181824 is not better than best loss 4.720241725444794\n",
            "Epoch number 74\n",
            " Current loss 1.5654529333114624\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 10.17104059457779 is not better than best loss 4.720241725444794\n",
            "Epoch number 75\n",
            " Current loss 1.5402376651763916\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.731033682823181 is not better than best loss 4.720241725444794\n",
            "Epoch number 76\n",
            " Current loss 0.7412971258163452\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.30634480714798 is not better than best loss 4.720241725444794\n",
            "Epoch number 77\n",
            " Current loss 0.8358034491539001\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.056041896343231 is not better than best loss 4.720241725444794\n",
            "Epoch number 78\n",
            " Current loss 1.2321654558181763\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 9.364318311214447 is not better than best loss 4.720241725444794\n",
            "Epoch number 79\n",
            " Current loss 1.7786792516708374\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 9.016408860683441 is not better than best loss 4.720241725444794\n",
            "Epoch number 80\n",
            " Current loss 0.8848573565483093\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.237154424190521 is not better than best loss 4.720241725444794\n",
            "Epoch number 81\n",
            " Current loss 1.416346549987793\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.513695478439331 is not better than best loss 4.720241725444794\n",
            "Epoch number 82\n",
            " Current loss 0.7943236827850342\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.341342389583588 is not better than best loss 4.720241725444794\n",
            "Epoch number 83\n",
            " Current loss 0.7126728892326355\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.513447701931 is not better than best loss 4.720241725444794\n",
            "Epoch number 84\n",
            " Current loss 1.5318530797958374\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.097413539886475 is not better than best loss 4.720241725444794\n",
            "Epoch number 85\n",
            " Current loss 1.0485339164733887\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.46859586238861 is not better than best loss 4.720241725444794\n",
            "Epoch number 86\n",
            " Current loss 1.4833197593688965\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 10.93652868270874 is not better than best loss 4.720241725444794\n",
            "Epoch number 87\n",
            " Current loss 1.3192272186279297\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.796658158302307 is not better than best loss 4.720241725444794\n",
            "Epoch number 88\n",
            " Current loss 0.5204808712005615\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.814725875854492 is not better than best loss 4.720241725444794\n",
            "Epoch number 89\n",
            " Current loss 0.4520387351512909\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.9588033854961395 is not better than best loss 4.720241725444794\n",
            "Epoch number 90\n",
            " Current loss 1.399552822113037\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.483920335769653 is not better than best loss 4.720241725444794\n",
            "Epoch number 91\n",
            " Current loss 1.0110269784927368\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.033943057060242 is not better than best loss 4.720241725444794\n",
            "Epoch number 92\n",
            " Current loss 0.9731783270835876\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.571174561977386 is not better than best loss 4.720241725444794\n",
            "Epoch number 93\n",
            " Current loss 0.9879310131072998\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 8.335495829582214 is not better than best loss 4.720241725444794\n",
            "Epoch number 94\n",
            " Current loss 1.00172758102417\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 5.839535236358643 is not better than best loss 4.720241725444794\n",
            "Epoch number 95\n",
            " Current loss 1.087787389755249\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.597834825515747 is not better than best loss 4.720241725444794\n",
            "Epoch number 96\n",
            " Current loss 1.342297077178955\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 10.22938060760498 is not better than best loss 4.720241725444794\n",
            "Epoch number 97\n",
            " Current loss 1.1267918348312378\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 7.672565281391144 is not better than best loss 4.720241725444794\n",
            "Epoch number 98\n",
            " Current loss 1.1037108898162842\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 6.806054532527924 is not better than best loss 4.720241725444794\n",
            "Epoch number 99\n",
            " Current loss 1.3228999376296997\n",
            "\n",
            "Current Accuracy: Not Implemented, find thresh first\n",
            "Loss 9.291492998600006 is not better than best loss 4.720241725444794\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOy9eZzk2FUm+l1JoVhyrarM2qv3vd3eKLwzbmOMbXiMDcM8sBnzDJhmftisfjMwbwbMsDx2eCw2xgN+PfgNxh4WY4Px7rbxOl2Ne1+rq7u6stasqlwiY9Vy3x/Ska6keyVFhCIzI1vf79e/rsyMUChCoXO/+53vnMM45yhRokSJEpMPbatPoESJEiVKFIMyoJcoUaLEDkEZ0EuUKFFih6AM6CVKlCixQ1AG9BIlSpTYITC26oUXFhb4VVddtVUvX6JEiRITiXvuueci53xR9rctC+hXXXUVjh07tlUvX6JEiRITCcbYSdXfSsmlRIkSJXYIyoBeokSJEjsEZUAvUaJEiR2CMqCXKFGixA5BGdBLlChRYoegDOglSpQosUNQBvQSJUqU2CEoA3qJEhOO4xc28MmHzm31aZTYBtiywqISJUqMjksbPbzlz7+OjuXgtbfu3+rTKbHFKBl6iRITCsfl+OkP3Yuza11YtrvVp1NiGyAzoDPGjjDGPs8Ye5gx9hBj7Kckj2GMsT9kjB1njN3PGHvheE63RIkShD/63BP45ycu4sjuOmy3nDxWIh9DtwG8k3N+C4CXAHg7Y+yW2GNeD+B6/787APxJoWdZokSJCL70xEX8wWefwPe88BC+67kHy4BeAkCOgM45P8s5/xf/300AjwA4FHvYGwD8BffwNQDzjLEDhZ9tiRIlAAB/fc8p7Jky8WtvvA2GrsFxOcr5wCUG0tAZY1cBeAGAr8f+dAjAKeHnJSSDPhhjdzDGjjHGji0vLw92piVKlAjQs13snjJRN3VUNAbA09RLPLuRO6AzxqYB/A2An+acrw/zYpzz93HOj3LOjy4uStv5lihRIgcsh8PQvNtX172AXsouJXIFdMZYBV4w/x+c87+VPOQ0gCPCz4f935UoUWIMsF0XFT+QG1oZ0Et4yONyYQD+HMAjnPPfUzzsowB+0He7vATAGuf8bIHnWaJECQG2w6H7gVz3mbrjlAH92Y48hUUvB/AWAA8wxu71f/d/AbgCADjn7wXwcQDfAeA4gDaAHyr+VEuUKEGwXReG7gXySiC5lF70ZzsyAzrn/EsAWMZjOIC3F3VSJUqUSIftcFQrvoZeJkVL+CgrRUuUmEBYbpgUJQ3dKgP6sx5lQC9RYgJhO2FStNTQSxDKgF6ixATCFmyLpYZeglAG9BIlJhCW6wb+c720LZbwUQb0EiUmEI7LgwrRwIdeSi7PepQBvUSJCYTt8MC2SNJL6XIpUQb0EiUmEJaYFC019BI+yoBeosSQ4Jyj3be35LVtiW2x1NBLlAG9RIkh8fnHLuDor34G611r01/bclwY8aRoqaE/61EG9BIlhsTp1S7afQdr7c0P6J5t0QvkFb3U0Et4KAN6iRJDwnY8zXorpA6xl0toWyw19Gc7yoBeosSQIImDAvumvnZpWywhQRnQS5QYEsTMN5uhe+PmIGHoZUB/tqMM6CVKDIlActlkZmz5r0tJ0VJDL0EoA3qJEkOCuhtam6xdExOvaKWGXiKKMqCXKDEkiKFvNjOm19VLDb1EDHlG0L2fMXaBMfag4u9zjLGPMcbuY4w9xBgrpxWVeFaAArm1yUlRyw/cwUzRUnIp4SMPQ78TwOtS/v52AA9zzp8H4HYAv8sYM0c/tRIltjeswOWyyQzdJQ29rBQtEUVmQOecfxHA5bSHAJjxh0lP+4/dmnroEiU2ERRYN19y8V7P0OLtc0sN/dmOIjT0PwZwM4AzAB4A8FOcc+k3izF2B2PsGGPs2PLycgEvXaLE1sHeIsklSIrGGXqpoT/rUURAfy2AewEcBPB8AH/MGJuVPZBz/j7O+VHO+dHFxcUCXrpEia3DVlWK2jHb4qRq6Bs9G/ecXNnq09hRKCKg/xCAv+UejgN4CsBNBRy3RInC0LfdwgNeUCm6yYHUCiSXydbQP3z3KXzfn34Vnb6z1aeyY1BEQH8GwKsBgDG2D8CNAE4UcNwSJQrDG979Zbz788cLPSb50De79D9IisY19C1oQTAKVtt92C5H1yoDelEwsh7AGPsgPPfKAmNsCcC7AFQAgHP+XgC/AuBOxtgDABiAn+OcXxzbGZcoMQSWLrdxeqVT6DEdd6sqRX2Grsd86BPG0Fs+M+9P2EK0nZEZ0Dnnb8r4+xkA317YGZUoMQb0HLfw5KW1RZKLE0uKMsaga2ziNPQ2BXS7DOhFoawULbHjwTlH33YLZ4JhUnSTJRcnKrkAnuwyaQydpj31yoBeGMqAXmLHgwJ50Qw9tC1usuTikuQS3r6GxiZOQy8ZevEoA3qJHQ8KGEUH3q3qh06vR6X/gB/QJ5Shlxp6cSgDeokdj549Loa+NT50Wph0QXIxdG3iNPRWz2PovdLlUhjKgF5ix4MYetFb+63u5VIRJBdPQ58sptspXS6FowzoJXY8xsXQnWBi0WZLLtFeLvTvSSv9b5HkUmrohaEM6CV2PMaloVtbVfofsy0Cnid90iSXTpkULRxlQC+x49GzvcAxLpfLViVFjUhSVJu4pGirTIoWjjKgl9jx6I8rKeqMh/lnIbAtapOroTsuR9fyzrf0oReHMqCX2PHojcu26AfWrRpBN8kaekdwtpSSS3EoA3qJHY/xMfQtToqKksuEaejkQQfKgF4kyoA+AC63+njifHOrT6PEgBi3D33zJReZbXGyNPR2T2DopYZeGMqAPgDe8/njeOv/e/dWn0aJAUFJ0XH50DebGTsq2+IEaejtfim5jANlQB8Aax0L611rq0+jxIAYl23R2aIRdJQU1ePNuSZIQy8ll/GgDOgDwBpDC9YS48e4JJfAh77ZlaKOi4rO4M1l91CZOA29lFzGgcyAzhh7P2PsAmPswZTH3M4Yu5cx9hBj7AvFnuL2geXwTddLS4wOYoC2y+EWGPQCH/oWFBaJ7ByYQA29ZOhjQR6GfieA16n+yBibB/AeAP+ac34rgH9bzKltP/T8uZSTxIRKRH3OVkE6M+d8y0r/LcdFRYveupOmobeEpGjpQy8OmQGdc/5FAJdTHvJmeEOin/Eff6Ggc9t2sMbUV7vEeCEywKJ2WCIb3nzJhUcsi8Dk+dDbvg/dNLSSoReIIjT0GwDsYozdxRi7hzH2gwUcc1uCAnmp+U0WyOUCAFZBwUMMnpvuQ3fdyHALYAJ96D1PctnVqJT3U4EoIqAbAL4JwHcCeC2AX2CM3SB7IGPsDsbYMcbYseXl5QJeenMRMPSSUUwUogy9mGsnSjdbwdArY9DQOee488tPYa09ficXJUXn6hX07bIfelEoIqAvAfgk57zFOb8I4IsAnid7IOf8fZzzo5zzo4uLiwW89OYi6KtdMoqJgqjRFnXtHCGIW1uQFE0w9AI09FOXO/iljz2MTz50bqTj5EG7b6Ne0VGr6KXkUiCKCOh/D+AVjDGDMdYA8GIAjxRw3G2Hvn8TW/bkbG1LjEdDjzL0zU+KyjR0Z8T31rY8GUR0oIwL7b6DqaoOU9fKpGiBMLIewBj7IIDbASwwxpYAvAtABQA45+/lnD/CGPsEgPsBuAD+jHOutDhOMkoNfTIhXq+iJBdRZtn85lw8UiUKeBr6qJILdT/sWPk/oz/67BOoVXT86L+6ZqDXavcd1E29TIoWjMyAzjl/U47H/DaA3y7kjLYxSpfLZEJMihYVPMSAvtnfB9t1I61zAWqfO2pA9z6nzgAzPv/u3tM4sdxCx3Lwk6++Pvfz2n0bU6YB09Cw0Rv/juDZgrJSdACMazZlifFiHElRUa/eiiHRlYTkoo0s/VBAH2Roc992Yeoafu/Tj+M9dx3P/bx230HD9CSX8n4qDpkMvUSIkqFPJnpj9KHXKtqWDImWJUVHlX5CyWWwgP6G5x9E33HxW594DPN1E29+8RWZz/MCulFKLgWjZOgDoHS5TCZ6tguSnAuzLfrHqVf0LemHHtfQ9QI0dJKmOv38Ad1yXNRNHb/7b5+Hm/bP4CPfOJ3rea2e7TF0Y/ik6H//ytO40OwO9dydijKgDwBidyWjmCz0bBdTVW8zWtRiTKy8VtG3gKHzSC90gGyLm6+h920XFV2DoWvYP1eL5CvSQJJL1dCGuibLzR7e9dGH8I/3nx34uTsZZUAfAP1Acilti5OEvu1ixg/ohVWKukJA34IRdPHmXIamwXE5OB/+XIiZdwcI6JbDYRpeGKkOwLbbfQeNqjG0hk7nuNEtE6oiSg09J8SmXKWGPlno2U7A0AvT0P3vgMfQN9uHLkuKej87brLPS1507cE0dM45+o4b7BaqRv4ioXbfRqOig7Hhdry0cGxsgmd+klAy9JywxuBlLrE56NsupmsU0Itm6NoWVIpKbIt+EB9lt0Cst5vTh0471uqADN11OTqWz9CHlFxI2mmVlscIyoCeE+KXrqxsmyz0bRfTRWvoFNANfWsKixQMfbSA7jP0nElR2u2YPkP3EpzZz+3aDjiHb1vUh2pJTaxebMNbogzouWGNwctcYnPQEwJ6UQlMklnqpj6ydj3wa0uTot7Po5T/hww9X5CkoEryT9XQ0cvB7qkx15TvchGPlReB5CJh6J966Byeutga6Hg7BWVAzwlRey27LU4W+oLLpTjbYii5iD9vBmzHlZb+A6O18g1sizkDOn2WpqEDAKqVfJJL22fVdd+HDgwf0GWSy89++D78yQBFTjsJZUDPiaiGXrpcJgk92wkYetGVorWKF8w2U3axJIlPvUjJRRLQ/9sXT+DhM+uR3yUZuqeHZ435oyZgIkPvOYNJJ31FQLcdFxs9u2ToJdIxjhasJcYP23HhcmCq6gXewtrnCrZFoLjRdnngMfSkDx0oJika19Bdl+PXPv4I/v6+aNFQP2DooctF/L0KrYCh66jqwzJ037YYC+h07BPLZUAvkQKR2ZWFRZMDWogDyaWg1se0S6v7AX0zi4vkSdHiNPSeHWXZxNjj33v6WXS5AMjU0WnBmKqOILlY8qToetcbznGp1d+UQR3bDWVAz4nStjiZoEBRr+jQNVZg+1ySXLTIz5sBy3WTSdECNHTRrijuSFu+1zuuj4eSix/Q/c8iy+lCx2uISdEBPz96fJyhiz+fuLgx0DF3AsqAnhNlQJ9MUBAyDQ0VvbiATt7zgKFvooYu7eVShOQiBGJRR6ckZpxFWwrJJSsxSgy9YRqB5TGPO0YEdYRs9e2Iw6gpVI4+G3X0MqDnRERDLyWXiUEoC+io6MMVscjgCJWiwOZJLpxzxQg62imMnhQFogGdGLVKcgkrRYmhZ2jo/TApSqx+0OtCr8F5aIMEgI1eKLM8G3X0zIDOGHs/Y+wCYyx1ChFj7JsZYzZj7HuLO73tA9HZ0i9dLhMD2v6bhgZT1wqvFK1uclKUkrEJ26JQ+j8sepYD5h9WTIxSwIxLKfGkaOBYyZBc2kJS1FQkRbP074gkJMgsxNArOisZugJ3Anhd2gMYYzqA3wTwqQLOaVuiLCyaTPSExF1F18aWFN0s2yItJAnbYiEauoPZWiX4N4F0aRVDNwdk6G1RcpEkRe85uYIX/uqnsbTSVh5DfPyGJKDfuH8GTy6XGnoCnPMvAric8bCfAPA3AC4UcVLbEaXLZTIhssiKUZyG7rhhshXYvEWeXqcyDtui7WJXIxnQAw099h5VtsUsPbzdt1E1NOgaE1h9+JyllTYcl+P8urrXubgLEJ0uFNCfe3geT19qZXridxpG1tAZY4cAfDeAPxn9dLYv6MtrFOiUKDF+UHAhhl6Uhk4MnVjpZmno9Doq2+JoGrqD+YYJQK6hxwN1kBQd0OXS7ofdL6sSlwsF6E5ffa16Coa+0bOgawy3HJhF13JxNmVRIPzupx7DD/zZ1zIfNwkoIin6/wD4Oc555p3CGLuDMXaMMXZseXm5gJfePPQFP3MZ0CcHYkfAQTT0J843cddj6g2n7Xo9yY0CuhwOAtLqE0lRfTQNnXOOruUEDD2ioZPkEmfolBSN+9BzJEVpZ2PqeuRYgMfgxf/L0E/R0GdqBq5ZnAIAPJUjMXriYguPnm1mPm4SUERAPwrgrxhjTwP4XgDvYYy9UfZAzvn7OOdHOedHFxcXC3jpzQMxsumqUSZFJwhkbyOXS962Db/1ycfwn/72AeXfyTpIDo/N8qFTwK4obIvDJmcth8PlwC4pQ1cUFsW6LQ5iW6TKXZmGHjD0lJ4yMp884A28mKkZuHZxGkA+L3rfdrHetTa1wdq4MPKAC8751fRvxtidAP6Bc/6RUY+73dC3KZGjB/8usf0R0dBz+tA557j31GrqY6njYRHa9SAIJRe5hj5spSgFzzmZhp5RWJRIimY092r1HdRNL/SEAV14Pb/XS1rXx57tYLZmYL1rRySX9a6N6WoFe2eqmDL1XNbFvu3Ccji6ljcfdZKRGdAZYx8EcDuABcbYEoB3AagAAOf8vWM9u20EYnae5DL5K/mzBaT7mrqvoedIaJ9b72K52UMj5eamMXCbLrkIuRwRgYY+5HlQEA4Yej+ZdFS6XIy4hp7F0G1MmTGGLiye7UBDVwf0vu1iz3QV6107Irls9CzM1AwwxnD14hRO5LAu0me61rF2fkDnnL8p78E4528d6Wy2MegLN101gn4RJbYGrsvx1/cs4Y0vOBQEBBUCDb2iwTS0XBNu7ju1CiA9MFmuNwYuTEZujuSisi2OqqFTUVGgoVtJTTv+eagqRbMWzVbPwcF573VkPnSSUDopbpme7WKu7h1jI+Zy2T9bAwBcvTCNe0+tpJ6L+NrrXQv752qZj9/OKCtFc4K+vHUz/9zEEuPBfUur+I9/cz/+6cHsie/EPImh59ld3be0BsALjqpATR0PKZBu1q4tZOixEXTaaD50Kvuf8xl6V6KhJwqLbBcaC187b1K0Yzlo+JILtd6NJEXzaOiWi3pFx5Spxxi6p6EDwNULU1ha6WQO7KBFf70z+UStDOg50be9oQLU87nE1mHd9xo/FOvPLUPI0PXcGjoxdPH5cdh+T/Kgy6HAjLuWg9/6xKOpLo1hQRp6fEh0ZUTbIgW9ekVHraLFfOjywiLLcSM7pKAvS1Zzrp4dJEUZ87zoPSfJ0LM0dNPQMFU1Ei4Xmh977eIUOAeeuawuUBLf16g770sbPbz+D/4Zxy9sXUFTGdBzwvKnmxdZPl5iOGz4Af3B02uZj01o6BnXznU5HlhaCzRq1W6MXC6yLof/cnIF77nrSXz5+KXsN5OB3/v047jzy0+Fr6uwLeoFSS61ioZaRZe7XBw34gTp2dGuj5rGYOrZU4s6fQf1Sqj2VmO5DaokTdPQe7aLqqFhumpEfehdGzN+tevVC551MSsxGjL00RbgR8428cjZdRx7OqsOc3woA3pOWA73+oEYxZWPlxgO1IDpoTPrmVazvuOCMY/R5lmMT1xsodmzcdvhOQBq+cB2XRi6FjBjUXKh55zLUdSShQ/+r2fw2UdDP3zA0BW9XIZNiooMvV7RY71cvEDHeWwUo+MGMguhamiplaKcc7T6IUMHPA1e5itvZyRFqxU9wtC7loO+E86PDQJ6hnWRXnttRMllecO73mdWO6mP+/1PP44vPj6eOpwyoOdE32foRVYblhgOVN691rFwOuPm6dsuTF0DYyxXLxeSW1501W4A6jJ2YuhBDxXhOxEE9LX0c8vCWsfCcrMXCa4UsHVl+9whNXQ/oNcooEtK/4GoBEWfrQhvrmg6s3Y5Im6SeECnQJ4uuXiLyVRVD1w4xNRnfcllplbBwnQVT2c4XayCNPQL6z0AwJk19UJuOy7+6HNP4NjJ7GTtMCgDek54X14KCmVA30qIW+wsHZ1ufAC5erncv7SKKVPHLQdnAQB9xaxL8qFXJMyYAtq5tV7GO0kHNZcSg2uQFI0F0pE1dDsquciacwFRCarvuEGVKKFq6KmSCwXrKTOUXMxYXqoduFzSA7rpSy5N//xooScNHQDm6kZiqlEcRWnoF5re9T6bspCfb/bgcuDAmNw0ZUDPCcv/8lYMVjL0LUarZ8PUNWgMeChDR/dufI8NGlq25HLv0hqec2gu6HPeVTB0K/ChJ22LoeQyGkN/8kIyoKuSolka+r2nVvHy3/gcLrf60r93hYrauqlH3ne77wSefJF9W46EoRvpGjrJIxGGPpSG7vgMPZRcKLcyU60I56NnJmmDgD6ihh4E9FU1Qz/r7yjLgL7FoC9vtUyKbjk2ejbmGxVcuzidg6E7AUM3jXTbYt928ciZdTz/yLy0aZQI2/F86LJJQaHkMpqG/qSfzOtGJBe5bdHIKP1/4PQaTq928MhZ+efVU0gupHlTwVGEodvJUXimoaVWitJx4wydPjPX5WFAz5Rcohp602fYIkP3JKD0+5W+EyNr6E3vep9e7ShzOyTHHJyvj/RaKpQBPSf6Ng80dJdv7gzJElE0uzamqwaec2guM6D3Rcklw7b46Ll19B0Xzz08H7Z1VTB0x+UwNC1gypGAbpHkMlpAPy5h6JaKoWeU/q/6zPzpS3I9Oepy0QJ23LVccA7smvJYb2Ryl28UEJGXoTcUGrr4XlUaOufck0BjLheSXmbEgJ4jSRu4XAqSXHq2ixXFgA7Kq5QMfYvRDySXpKuhxOZio+d5jW89OItz611c3FBr1aS1At6oNNvlyh7ZlBB93pG5sOpRsQBYrhvxocskl1bfCVjjMDgh0dBVtsUsl8uqzz5PXpJ7ssWkqKihkydcztAdieSSXngXzhMNA3pVCOhioy0VQxc7aE6ZBnq2C9txAw19EMlFvL6jBvTl9V4QqFVOlzOrXcxUjcBaWTTKgJ4Tlu2i6jN0IN8MxHbfxh985olSoikYGz5Dp8RlGkuPMnR/MVbIEvctrWHPlIlD8/XMRlOBD11LVoqKAW1Ylt63XZy83IauMXQtN1iEguZcMZcLY578o9LQV9oeQz+pYui2A93vHilKLuRwoYAen62bYOgZLpeWMK2IYBp6UFhEr1eraErbojiFiuyPrZ6DDT8gJxh6WgsH4bqNoqF3+g6aPRvPOzwPADiruO5n1zpjbS9QBvSc8JKi4YSVPEH6S09cxO9/5nHcv7Sa+dgS+bHR8wL6rQc9r/hDZ9SJUaooBMJKRtXu6txaF1fsaYAxllnGbjmeD13TGDQWTUaKzxnWi37yUguOy3H93unIMVW9XABPdlEtVqvtLIbuoua/57oZBnRizLunKKBH5Z9BJRdysDSq8qQovd6eqWokdyBCHFpCnvONvi11uVQr6TsG+pvGRtPQl3255XlHvICuYuhn17o4MCb9HCgDem6QD92U9J5Qgb6c9EUrUQxIcpmrV3Bkdz0HQ/eCB+nOKtvpetcKGj7J+nSLcPzmXIAnf4iBVAx6KqaWBbIsPueQt2hRgCVpJ54UBbxiI6WGHjD0tjRh17WcwNlTFyQXCsDzftOuZFI0urDktS0mJRcn8veFaTOH5KIHk49aPa+Nbq2iRRK1WQsMvZ/dUyaaXWvokXUX/ITozQdmYOoaziisi2dWuzhYMvStB2X0g217Doa+ESt4KFEMiKEDwK0H5lKti32h30glY3e13rGCIclZwxpsl0P3g2pFYxH/d892MeOf3/mhA7onjdzqy0oU3FRJUcBj6EoN3WfoHcsJ2KSIruUGAd3T0D2ZpxWTXMSA7vVyibab9QKoWnJpSyWX0IdOSdM901XYLpdeq2BoSUVg6D076IUuwsw4Hzr+wnQVLo9q+IOAEqL7ZmvYP1eTWhf7touLGz0cmCsZ+paDGhENEtCpqVGelq2D4Of/5n780kcfKvSYkwLOeaChA8BzDs3i6UttZfKxZyU1dFX+Y71rY7Yem3WpCAaW4wZFRXHtume5mK1XsHvKzDXTUobjFzZwcK6GPdNVAGEyUZUUpd+laehX7G4AAJ6WyC5d2wn6mVNg79luwNB3TSU19J6MoVfSXSVtmctFkFzofe6ZSk5OEl8X8CWXWpShz9aiHcGzXC50rAX/c14fcjd9wb/OizNVHJirSSUXGno9LocLUAb03LAcDlPXpFPKVaAEUNGSy4Nn1p61unzPdmG7PLiRSUd/RDETUmToaRo65zzC0LOus+N3WwTgt+WNVlBWDQ37ZmsjMPQNXLt3Opi92Y0x9HhSFCCGnjxf1+VY61h4vq/vyqyLPctBzSDJxXvvHcsRGLpEcpH2ckmXXFp9J2iURjAjLhc/oPsBVqaji4M1yM/e6tlodq2Ifp7nfPpBQPcWkDWF3TALF5o9GBrD7oaJg/N1qdRGQf7A/BYGdMbY+xljFxhjDyr+/gOMsfsZYw8wxr7CGHte8ac5frznruP48N2nlH8nvTArsSYiqGArmKF3LXfbyTjv++KT+HzKUOWiQO+bJI3r/KShql+HjKHLdlcdy4HtcszWSXJJ19Athwcs2dBjkovlJWIPzNWG0tA553jywgauXQwDOjHVYKaohKHHpR9Cs2vD5Z58Y2gMz0gYesdyUKuESVH6XTtuW3Rikou0UlQtcTS7VrALIogBnV6PAmw6Q9cFycUJ5onGz6fvuEptXJRcgOGti8vNHhamq9A0hoPzNZxb7yZ2S/Rd2GrJ5U4Ar0v5+1MAXsk5vw3ArwB4XwHnpcQDS2v42Q/dW3iv6b++ZwkfTxmYQO1zB5Jc+tGS5KLQFZjTdsF7v3ACH/nG6bG/zkbMyUA3tEr7jGjoKQltsqwRQzf81gKp3RZ9lmxoWqJSNGDoQ0gu59a7aPUdj6GbPlvuR5OiEoIOXZfbFsmyuGe6isO76lKGHtfQ6TWJMQeSixBglZWitquslFzv2sFnLD4n1NCJoXuvJ7Mu0oIRtS3aQcGZCJKRVDIb/Z52BMM26LrQ7GHvrHeMA3N1OC5P5CrCgL6FDJ1z/kUAyga/nPOvcM6pddjXABwu6NykWOtY+NtvnMbXTozea1pEq2ent+v0A0NgW8whuVBSdNhEiwpdyx2pYKVouC7Haruf+vkVBWLolPwSXfQYWdkAACAASURBVA4y9CwndLmkJEWJmYnsMa0oxfOhCww95nKpGjoOzNVwqdXP7CUSx5MXvIB77eJUGFxJcvHdNYwlI3p8YSFQUdGuRgVX7JmSWhfjLhf6Xatng7Gwg2GCoUtsi5yrC5zWOxZm6rGA7k+S8sr+bWgMmK+nMHRLkFyEpKg3rSh67CC5rdDRaXFfnBlRQ2/2sNc/xkFfUol3Aj271sFszQjOeRwoWkP/EQD/VPAxI/jmq3ehYer4/KPF9hPe6NqpzYBoe0ksrzdAUrRoDb1nOdjo2Zm9wDcL610LLk9vppQXjsuVDaSA8LMkZuYlqlnAJOPIq6ETMxPZY7ytqwjbDROChsTlUq1owWxLaquaF2RZvG4xqaHT6DsZVBo6MfT5homr9jTw9KVW4rvTlUgutBOcMo3EvFDX5bAcnmDoWe6g9a6VSFyKg6Lp9YJzkGnogm2xamgwNIZWz8Z610oy9CAXkm6BJIlnWIa+3OwGiwJJKvGui2dWu2Pr4UIoLKAzxl4FL6D/XMpj7mCMHWOMHVteHi4gVw0dL7t2Dz7/2IXCAprrcrT6TqqMk7At5mLoY9LQbccLoBmzEjcLFICL2Il87L4zeMVvfk7JuEMNPQy88TFkBAo6FMjT5LKQoYtl42oPs+2ESdF4F0eqTqWKwEGLi45f2MBMzcDiTDXUs/thUlRWVOSdh1xDXw0CegVX7G6g2bUDGyPBKyyKSS6+ht4wdX9XEAZq8t3LKkUBdYWtmHgOniME9HbfRt3UAxeMXEMPJRfGGKb8fi4ql4v3nHSGToVTwxQX2Y6LS60+Fme8601BO25dPLvWGavcAhQU0BljzwXwZwDewDlXaiGc8/dxzo9yzo8uLi4O/XqvvHEvllY6gVd3VARTxhUsz3E5XI6o5JIjKUoSRJEauuMHKWD7+NuJARbB0M+sddDuO8r+LBS4RTfDlCnveR3OE41p6LKAHmjo0U59MobOOY/40I2Ydk2dACmgD5oYfXLZS4gyxhJJUW9nIL9t4+dBoOC9q2Hiqj3eFJ+4jt6zHdTMqORCGvpU1fBmfwr2wn6QmExKLt7xsq2hBLGIi14v/r4j52pFF5PpqoEL6z1wjqTLpZK+Y6D3Ua/omKkaQyVFL7X64ByB5DJbMzBl6onionFXiQIFBHTG2BUA/hbAWzjnj49+Stm4/QZvMbirIFdFK2PKOLEvmikq/i71uP3iGXpk8ECBC8VGz8YvfOTBoc51peXdBEVo6LTFVjGlZqChCwG9qksZujhPFEDq7krG0FXzMUkfJh+6VykaHXAhMvRBrYsnL7Vx1R7PMx7X0KmHjAy6Fj0PwkrbAmPAXL2Cqxa848YHJysZes8O2LLY5pYC4cCSi4Sh0/Xp2y46/o5ATMzGITbnArzrT37/pIaeLrnQfWwaGmbrlaH6uZCkRgGdMYYD8/WIF71rObjc6uPA7BYzdMbYBwF8FcCNjLElxtiPMMb+PWPs3/sP+UUAewC8hzF2L2Ps2BjPFwBwZHcD1+2dxhcKmstHMyqzmgFVdBYk1nKV/o9BclFNkhkV95xcwQe+dhL/MsRorMs+Qy8ioFPgiksChGCIgcDEGqYhlXt6DlUUhgEJSNfQo42d5B5mYsGBbVFj0W6Llqfbz1QNNEx9YIa+3rUw79sEPVkhXOhsVx3QDY3BkWjoq+0+ZmsV6BrD4V0NMAY8fTEe0BUaet8Oknji50GfoSwpCsgDaM920LPdyKIpHqNvSzT0FIZO13WqagSLZlxDz6onEBem2XplKIZOZf+koQNIWFapSdu4GXpmupVz/qaMv78NwNsKO6OcuP2GRfzFV0+i1bNHzhqTG6Vnu3BcnpjXKK7iadv2ONpjKP3vCl/MIhk6uWaGqWpdDQL66OcTBHQFQ9/oWdA1FtnqTys09EAWyKWhe31AqkIpu6psPNyxCUlRiW2RMYb9c4NZF12XR7Rgkl3EXi6yKtHgPKQauhX0YqlVdByYrUW6LtqOV6yVdLm4aPedoGpTbHOrZOiBhp78jCmhrUqKUmXqrikzOIcs2yLgSy5NYugKDV3lchEZes0YSkOnsv+9Avs+OFePFLuR/DLOPi7ABFeKvuqmveg7Lr765Oj2RTEwymSXIKAPILnQpBc6flEJ3HExdLrZmkMc83Ir7BMy6vvs9L3Pda0td7pQ2b9o22uYuuLGH0RDlyfrpJKLEx3UXNG1CEOnifQAsH+2ljpjMvH++jY4j0oHYkC33JSkaIoPnRg/AFy5ZwonBclFnCdKrwdQpaiNhk+YRL+4GAhFpEkugZNIYlsEBA3dNBJSk4i+7YKxsFp2yjRAbzsZ0JOj8+LHovfhSS6DB3Tymy9Ohwz94HwdFzd6wetSgnTba+hbhaNX+fbFAnR0MTDKWCZNiq/oYlI0PaB3LW+6+WzNgO3yXK0C8mBcAZ0WtVEYOufqGZx50c2SXHpOYlstTq0REdysOV0uMilAJq3ZMclFbIrFOY+Mvds/W8P5AWyLAYsVEoe1ih4sdLbjBgOh41Bp6KttKyjdB4Ar9zQiDF0cbgGErLbTd9DuO5gyw9+TeyX8bKOLi5kiuawHDF0huTihZq9rXqtqVaUo7YAARHboag09g6HrGmZrlaEsxheaXexqVCKLG5X3k9RCi/r+rdbQtys8++IC7npseWRWKAaDbj954Wnye0VozpWlodMx9/kXsKjgKwbMYhn68JKL6BsfVXbJI7nEWVijmpehpyRFOzLLm1xDJ693JWDoodRh+46oIKD7kkvetqyhli8wdFMXfOhqhl5RaeidPubrYkCfwsWNfvD9CQK6z2Y1X9KiwiLqjJiPoasljpChqyUXcrkA3s5L5kMnFxFhuir+O3ps2nXk0dDnhmToF9Z72DsTDdQHfS/6mVUK6F7QF4djjwMTG9AB4FU3LeL06uj2RTGItS0Z0/MTQHo4oaafYVukwBYE9IL0btHfW2TBErGnjSFaCohsetTEaDZDT5Z3T2Uy9Fg/dFlSVMLQxT7dIuKSi1eh6b2W2GcE8AK67XJcbOVj6U1J0jcpuaQUFsk09JYVkVzIQUMsvWtFFz7AW0TaxNCpiEuwLYYyZLx9rnp0X+AkUvnQ7dCHHn/fIsShJUCcoSsklxQHm655055m6waaPVvZsVKFC81eJCEKAId3eQH9kw+dA+fcsyyOsYcLYaID+g37ZgAkK7IGRVRySdHQ/W2eGeuuJwNZIQtn6EKAKbItL53fUAy93Q96i4wa0OkGViWnNrrJJPiUaaBvu4lrEiTP4gxdwmIH0dBFGyvg9VChQEqBwxQkFyD/KLqmJOjVK3pg33OEHjJxyDR0y3HR7NlBcy0gWfgSl1zoNde7FmyXBwzdGy8XT4rGB1ykSC7k9U8snGEC1HJ4IPF4AT35+ZPkQqDvA2MIui8mz0fN0EmSo8980LYay0LZP+HKPQ18/zcfwZ1feRr/9WMP48xqJ2gJME5MdEAnf+yojarEYCvzvcZv4LSScEIrYOjVxGuMgu0ouay2+0HgGllyCXzo8qRo059WJIJu6Hbse6DU0G0ZQ08WvKgKi0Lboi+5CBq62KsbQFgtmjOgr0vmYtaEkXBWhg893kOFFsZ5QUOPV7BS8I0H9Esb3jWgABspLFJJLikuF9l7E49BBWq0gNSEhUxEPKDTjm3aNKBp8QUm2rIgDnEuKi00g3jROfeacC3ORgM6Ywy//j234W2vuBp3fuVpPHquOdZZooSJDujTGY2Z8qKVwdDjFq2KznIw9PFILuMqLGoGkstgx+ScY6Vt4ZC/xRy1WjRLcmn17KB1LoECzkZsMaHgSjcsba3j1y7eC52gKiwKe5JTpWjocunHdHtixnntcKHkIjJ0LdLLRVUpWpH0chHL/gl7pkxoLBy4QCShJgTJWkXHJT83EnG5jFBYtN6xYGhh9SuBFly65iTxiLkDEV5L5PAYtKDHF3pAWGCUSdGwHw3lUAbxoq91LPQdN6GhA15Q/8/feTPe+ZobACAYMDJOjK/t1yaAVvJRWWGWbTHORuIDDWQIJZfxMPT5RmUoi6EKFEgG7cey3vU0x0PzddyNFWWTrLzITIpKWqSGDD167rLydNliHO+FTqhW5N0Wg6lBWuhDtxIM3Qs44oi0PMjS0O0U26IumSm6IpT9Ewxdw+JMNdg1yCSXWkXDqRVPyiQZQ+w+mV36L3O5eHmKeKfIgKG3ogy9XtGl97bYcA0IP+M48wfCxSLNtkjnTPNkB/GiBx70mORCYIzhJ159PV5+/QJu9CXicWKiGXrQC3nEICIWcnRktkWqiotILumJk1YsKVpU8KWbb2G6OhYNfdCkKN2ExNALk1zaVsK95PhN1OJMTBU06SY2IwFdSyTs4r3QCSQxxM8jYOjBkOhQu44XvQTtXXPuptY7FkxDi8ofph5tzqWwLRo6S9gWifWKDB3wtP1zcYYee83LAUMPK20TSdGBXC5JJ5F4jBUJQ5dq6JYjl1wkBYaaxlDRWaptMSm55A/oSyuen39fhh3xhVfsGmvbXMJEB/SaoYOxJDMbFBs9O6jySkuKBhq6JCjEQcGWjltU8KWk6MK0uS00dNI9KdE2uuTiuQ76jpvYLdEiGb9xKZcSv3b9GFsGIE1oy3qhe8/T4Ep6e8enBondFuMyj2loqBpa7mslGwBRi1WKygZEA8nZpkB4fUSGDiAyfCNk6ILLpaIHx5qS2RYVkouha9A1eQCVOYnouEAoD4kMXSq5CIVbQLgAxD3ohKqhKytFLWEuahDQB5Bc/vmJi6gaGm47NJf7OePERAd0TWNoVPSRGXqrZwdVXukaujBDMiMpSschvbI4Dd31j1stVENfH7KwiALGoXli6MNfC8fl6Dsu9vnb17iOHkwrUkguSYaeZJHetYsGPVkvdPF58YQa6eWGxIce9BkRFpGZWiX34ISmpF+4GNhsSWsKgqFFK1YBuYYOeIlRklw6Uskl/HcjUliUztAByj3IXC7JPAU9HhCTooLLRXE/iqPvplM09OC8U/qhBwydNPQBkqJfeGwZL7lmz9j95Xkx0QEd8BI2o27zmz0bu6dNMJbe3S3Q0I18SdGKzlCr6MpKxmFAVYgzNaMwGadnO0HQGvQ8qez/cAGSCwUtcgPEA7qsdS4gaOiKpGhEQ5dcO1mnRfF5cbZpxVwuuuBDpyI08TVnavmv/7pkLma9osNyOCzHTW+fK2Hoq20vERlfBPfN1rDuD3WJFxbRaxKmhaRoL7YTkZ2LaG8U0ZQ4iYBkUjRg6Kbahy565un6y+QcIL2vvbg4TFcNaCw/Qz95qYUTF1u4/cbhW4EXjYkP6FOmPrJtkZwTDUUhg9jLhf6fR3KhL+ZMrVJcQPdnP6oaUg2DjaAk2zvmIJW3xAAXZ2owNDYSQ+/EAno8OSVrnQuEW+64/h+3LQJZGrq8l3acoTtBUjR0PQW2RUmRznTVwEbOINGUyBJi58HU9rkSDX3Fb8wVT0QG/vj1bqKiVnxNINTQq0JOgfII8aQo/U46t7UrZ+ikc1MinK5nTVlYJPehyzR0733JK34Bj6zRosQYw2y9kjspetdjXrfX22/cm+vxm4GJD+gNc3SGTsUqddPIaVvM40MPe45MVfVCbYu1iobpmneug1a1yUDOiv1zNbgD9mO53Op7VXY1I6guHBa0O9o/67H9uBdd1joXCDXeeC6l57Mv0ZucrqHLm0bFt+uJpKjmzdF0hJ490dL0/Ay9KWHoYqMqb2KRyraoSRh6tDEXIejVvt5F13LAWDQ4i5JL4HKhBc5xlRo6oG6ZsN6xpRo64H3WMg2973dAFSHzoc/WjED2S56PpqwUFX3ogCe75U2K3vXYBVy1p4GrF6ZyPX4zMPEB3RtuMHwQofFz037varnLxf/yComuPJILaYFFSi5df+jxoHa4NIQB3bshmr38SaGVtoVdDROMMf/zG/5a0Nb/gEJyiQ+IJtQrXnI8vmOJ36wAWU7lGnqibFzhYbYTPnRqKeAmXC6AJxHlbdPQ7FqR8Xr0/gCvz5A4yzQOSoqKO6zVthXp40IgOy0FdLHZFZAsMgKiXRH7jhP4+uOQadZ920tyq2QR009AA4KGbvpNwqzkscQFU9cYPvvO2/EDL7lSeuwsyUW8VrN1I1e+o2s5+OqJS9uKnQM7IKCPytBF54SqDWvctigLCsnjOkLBQ3Ge8a7legy9yIDuB/ADgSMnf1BeafWDTn5TikETeRGXXOJedGLoU9VoAkqVHBe7HhJkPnRZL3QgGsBEBD50oR86EGXo4kIyI1nQXZfjh++8OzGkxWOxMQ1dmK9pp9kWhfMgxFvnEvYJLQm6lpso9qGfG6Ye7HDEJLE4qzWOakVLuEqail0QQXQFEeuvm2Qlzr6uizNVZW6haujKHbUVK9TK26DraycuoWu5eOU20s+BfBOL3s8Yu8AYe1Dxd8YY+0PG2HHG2P2MsRcWf5pqjMp+KXhNVQ2lZhd3uZh5JJeeHdqpCtS7u7bjaei1YqpkgajkMugxV9p97PIHINRHZOj03N1Tpr8FlzP0OIMF5IOiVQw9fu1U7otwHmX0PRFDrwiVovR7WcHNjIShN7s2PvfoBXzpiTCgW75VM26/q0cklxSG7v9etFmudaKtc8NzqmDK1HHOZ+i1RED3zr8h9EYRk8R9W30eMsllXSGXEeg6TZnJnYFoXfTaE7tS7V4FL0mb7XIBPMklj4Z+12PLqBoaXnrNntznsRnI86ncCeB1KX9/PYDr/f/uAPAno59WfqhYdV5sCM4JlWTQ928i2pJWcksuY9LQBcmliI6L8YA+yAK50g4Z+qjXQrTPzTUqSQ29J2fo3u8MCUNPBnSZXKb0R+sKyUXB0C3XlWvovstFlEIoaNBwBECdIxDna6bZFmmBsWMMnRbcOPb5rX27tpsM6CaNd4tOcAJIcnFhGsnrAMglF5U1NDi2nlxAZIOibZeDc7ldUoW8Lhc6vzwB/QuPL+Ol1+5JfG5bjcxPhXP+RQCXUx7yBgB/wT18DcA8Y+xAUSeYBRkzGwShLqurJZfYRa/oLNvl0reFCrbiXC5dy0W1YMmFHBjDMXQLuwOGbqCtSD7lATGxekXHfL0iZej1ii5NCsoGRcf1UUClocsrGFUaetjLJawUBXzJxU8wiux1ulqB4/JIsnnVX6yWN8KArmovG0ouXpuFtPa5AILy/67loGu5QUl7HPtna77kkpQwaoHkEn4u4SAKV/rZEmQBVJV4Do8dSjyEQEMX7knZgpkFMyWgW44b5MYA4KYDM7jQ7OHTD59XHu+J8008dbEVDKvfTihCQz8E4JTw85L/u00BBeFhh1y0hERb3TSUtkXxoudpn9vuOWFS1GdoeYccpIG2xyS5FMH8A4Y+YKtfzjlWWqFGO2XqI1XtUsCrmzrmJPaxZjfZaZHQMCWSiyOTXOQ+dFmgEft0iwgKiyin4jNjLynqJhKMdM5islnG0GV9XICQqdLfK6rCokBy8c5vVdLHRQRNU5JJLvSzKIGEvcXTJRfTSGroqvYK4nOAsBGYeA7iPUluFdFimQWxB00cvRhZ+4EXX4mb9s/gv3zkgcT3r2s5ePfnj+O73/MVVA0N33bLvtznsFnY1KQoY+wOxtgxxtix5eXl7CfkwFTVG/GWZ2izDE0h0dZQNgPikcRJ3va5QdMgKnwZgb0SenboQwcK0tB7XlKQqgnzJkU3ejZsl2N3I9TQi5Bc6hUd8w05Q1d5jaeryYSslzyLBiq5D12hoSsKi+xEYZEfSB3Sd6OvSddfXHwpWFzcCGWlsL2sXEOn72oWQ6fzC8v+5UGUJJdO34mU/YuvKQZYcVScJVksCbIAqmqvEDxHV2vokYAuqS3IQlWywBDiOw3T0PBb3/tcLDd7+PWPPwLAIy7/cP8ZfOvv3IXf/uRjeNm1e/BPP/UtOLxr/N0TB0UR3WJOAzgi/HzY/10CnPP3AXgfABw9erSQqclBH49e8ubNg5aQaFMFpLjOluVysR0XXcsVNPTwhlYFpLzwNHQtSAzmdc9wzhPFJYRm18J0tTLwIrHSijZ+aigq+/KCttb1io65uhmZmk7npfr8GqYu7Ycev/HlPnR5BaNq2g0FzErMtujNjnUSgU4mj1FAv9zqB06LcABETEP3pYeAoStH0EU1dAroc4qAvn/Wm6Z0dq2L6/dNR/4WaOhCgBVzCh5DH0ByydLQiaELEk9D4nKJtyfOgzQNPe5yAYDnHp7Hj/6ra/CnXziB2w7P4R/vP4uvPHkJtx6cxe993/Pxkm2WCBVRBEP/KIAf9N0uLwGwxjk/W8Bxc4GKHoa1y4mJNlX/5TgbkbE8EcTEKaEUyCMD+LtVoO1xUB2ZQ3L5y68/g5f9xucSfT4Iza6nIdNnmVdyoYBBGvqoFtIgKWpqPkNPFhalMXRZL5f4jW/oLNLLRdULHYhqxiLoc9SDXi4USF2pA4MkFDGBLe4+aJCEbFoRIDJ07+/qARdRDX0tQ3IhL/qZtU6k7F98zYjLRcgpyOQs8XEJJ1HXgq6xiEYuInC5VCUMfUQNXdkG2fEGucvex8982w24emEK//nvHsRDZ9bxK298Dj76jlds62AO5GDojLEPArgdwAJjbAnAuwBUAIBz/l4AHwfwHQCOA2gD+KFxnawMVJY87FY/4nIRemaIq3bcKmb6OqyK9baCRSIquRThSCEfuqFrqFW0zEWCc44/+9IJnF3r4lKrL23zSdWJmn/D5WXol4PGTxTQdXQtr7JP5cRIQ9dyoDGPCc7XK2j1nYj1sNmzg54xcTQkVb4yhh7vZa/qhQ6oW8EGDF0pucQYuiSgi17n5WYP++dqSg29FpNcdFUvFz102wBhO9p4Yy4CfRc4R0JyodcUBzBHCotSGXrStkikQbVLrEoYek1SWBS0RB5UcrGT92tQXyIJ6LWKjne/+YX4+ANn8UMvvwp7puX9zrcbMgM65/xNGX/nAN5e2BkNiIChD6klb/hNtKqGHmwz230Hc/V4QI/qbNxvqyrb/pIG3QisX0bk93nR6Tuo6CzQTDnngQ8dIPdM+jHvObmCE/4Q7eVmTxHQrSDoTEm0aBVWEww91DyHkZY6fcev+mRBEFrrWMEA3o2ehZmqfEjAdFVHq29Hbtp4m1UgubtKS9apGbq3YAU2VlFysWQaesU//6TkAgDLG10Ac4HOHP/sKrqGis6C5yuTor7kQoVFwfi5uiIpKoxEU9kWRQ1dTBJbjqvs762yLaocLoAouaT70IeVXGT3a1r7AgC45eAsbjk4m/t1tgN2QKXoaHNFvQIgwz+WqjItGtCD2ZQKCSN0zkSbBg0quXznH/0z/vjzx4Of+47rsynq/5xdVPWhu0MD0sUN+eT5jZ4dBB1Pusj3WVKnRUq6UWXfsLJLx3KCQDIXjG4LZRfZgGjCVNUA59GdmlRDj/nQ05J1gWYcY+hWbFAzBVLbL/2PB5vQkRRe/9W2FejT5HRpdm1MmXJbZq2ih5JLVlLUZ57NrtdpMc6+CYvT1WC4d8LlIin0EcfL9Z3kZys+znJ4pGJV1uddBB1LfL1aUZKLYixez0kOQJl0TPw7CdjvsBq6oMuGgxKix5Jp6IB82LB4LmG3xcEll3bfxonlFk5eage/6wa9tkO9Ma2L30bPxj8+cDaoZhMdFSLEhlAyP7cKK60+NBay24bkBhwEHStMbJN3WtSaW73ktCJCQ/I9kCUoPduiEGhSknU0rIFa4hLiHQ+DwiJfcokHurAbZJShX7vXS0SGAd1SDmmoV/Sg2lKVFI2X/qtGvonvb8GXEuKL0Fy9ghddvRsvuGJX8LtIYVGa5FJJ2j09hq7etclsi7QzkUkug1aKAsnkdtjSY3B5cLtixwT0YVmhaIWT+V4BJPpWqLbiBHJbJBl6/nM8s+oNH2gKAZu+kKHkEmXof3/vafzMh+4Nttr/cN8ZtPsO3vGt1wFQM3QvoIf9WAZJiu5qmEGvj2Ak4JC7pZ7lBgx9PjbfsWc76DtuSlI0+doyPbuiex0JXSHoAeqCF5nlLV7cQ/92XK/0Py7zVA0dpqFFHElrHQt7Z6qYrRlBQJf1cSHUTT20LSp6uegxDV3WuTEOkl3iSVFD1/DhH3spXn7dQvC7MKB7XR9VzFbWpXJd0nQs8hzJjgBIttANWiIPWCnqnU/0Og5zrO2OiX8nUyNKLmJADzRgSXItmhRND+gBQ6/GNfT8Af30qjegV5yeEp/9GNfQ/+exJfzdN07j+/70qzi31sWHjp3CdXun8bJr96Bh6rjYTAZ0x+XeZ1ALF5/cDL3djyTcgmZK1giSi//e6LjE0FVl8YSGJJeiCuhAGPRUvdAJMstbPEkuJiNVfUZmY/1c1nxNeWGmGlSLNnvpDD2UXNJtiwFDV7h3RFBOJU8Je7KXSzpD79nRXEUuhm5GHxNvxyEbWpIFMyug69urfH8UTHxAb4zI0KMautwxE0+KVgz/BlZ4W4OGX8IsxjhDy8Jpf+K6OD2F5omSJupp6OHfn7jQxK0HZ7G00sF3/fGX8I1nVvF9R4+AMYaF6aqUodPiMyskRXMz9FZY9g+oP7+8oKQoECbyqOPiRiwvEUfcQ9+1PIdMXHM3g/xHPoYuKyKjpCjBiLhckmX0dH7xwqL5uonF6WpEQ1ctWJ6GTgw93bYYaujZDJ2siyqdXUSyl4taQweiuQfVcIvgOaShx/r01GMMPRwgMoyGnryvAbWENYmY+IBOAWAkhl4LR14BkmHDKg09IykqfjlnYjd0Fk6vetq5yOrio8LEpl/rXQvn13v4zucewId+7CX+eTJ89wu9LgwL06ZUQ49b5QbpjRNvzUrXYuiAbjmomWHClzFgrU0e7agVNI74YnLG3+HEbY5089JirOqFTpBVPXpJUUFyEZOiEpcLELZ/oMdt9GzM1StYnKkKkos66NUrevB8VVJU7CkDZAdRIGz3kIehGxoDY/l6VpZcuQAAIABJREFUuQBhALUcF+2+k9PlkrRsRhn6cLZFIJnclrU6nnQUUSm6pdA1hrqiZD8PNno2ps2YyyUmGVixjL6qCx8hnhQFojd0HhBDFzV0ueTiHfPJCxsAgOsWp3HrwTn840+8AufWu0HSa2G6imcuhwlWQjOwypHLRR9IQ3/e4fng51HzGV3LCRijpjHM1SsBQw8retMZOp370goF9Gh5NvXkocVY1QudIE66Jzgx+1vEtigpZqLzCxdf7/9zdSMS0NMYdd3Ug0Ctsi0SQxc19DSZAxAll+ygxhgLJKj09rnR+4MWY5WsBcgLi4DkXFHZuLwsqFwupYa+TTFVTQ43yAvRORFq6MkLH5Vcshl6vaJHtuVTkuZRaSANXWzqFTB0QXKx/G3+cQrovnNi72wNzxWC7YIQOETEtempqoGu5SqrSkXQvErCyJKLoKEDntOCNPQHTq8BiHqnRcSlN/r8DiUYejT/kaUzy5KidmwMXNhDxUXfdqTscboaDjlZFUryF2eqaPUdtPt2JDkdh/i5pI2gA8JK0fWOWpMnqJKiKtAsgNReLpVoAA2cRGkMnYZaVJKNybqygD6My0UhuQzC9rc7dsQ7aZjGUF3+XD8hOBVzuSRti9GMflyHjcObVhS9QQYZQwaELheXh4y/K3G5AF5QPr68AVPXcMVuecOghekqLrf7iUAdl1wCLTojKJNGLd6kdUVSOS86fScynHjeZ+iuy/GBr53E0St34ZrFaelzp4O2Bd5rL620YWgM+2aiFX4yDT1LCogzdMuJ+tArwjFVDF3Md4gFP4v+DmpppYO+4yoZdS0S0DM0dJfDdly0+k6m5PLcw/N45Q2LeN6R+dTHEUzDC7BekU4+yUXVFjjynEooI4pIaOhDNucCkpJLydC3KRrmcAydeq6Q5U3lcrEGLCxq95IFMLIxZCrYjhuRSyjodm2SXGh7GsoMx89v4OqFKSV7W5w2wXlYrk+Id/jL68iRbaMbRWjoIkNvmFjrWPjiE8s4eamNt7xUPjMSEFpACJLLgfla4vOIXztVL3SClKG7PBJUg+Zcim6LQHRq0ZrAWKkK9sTyhv84BUM3he+fagSd0D43mO6UkRSdq1fw33/4RTioGLAcR1VI7quTolHJJWw6pg7or75pL37q1dfjSEwiq8VcLtQKQuWtl5+Pdz3iCzP9XAb0bYZhh1yQ3ED6MRUyxNvc9h03cLYA0Wy/9Lg9J5HcGURDP7feheNy3HzAK3MPAjr1gjaiDL3pM3SSW2SgxeFiMxrQZUlRIDugy9whhu65eYYt8upZ0ck58/UK1tp9fOCrJ7EwXcXrn6Oem1LxX3ujHwb0w/PJ3QrpvnTtxIVTBllS1Ha5NClKO7s0lwvnPAjoc0JAf9Jvz6BaXETJRdUnRywsyhNEh0HV0ILvRlqlKBAy4mZG61zAkwh/5jU3BDUNhIYkKTqI3ELnTM8VEbpcdkQYBLBDAvqwDF020qwe+wJxzhNlzkFQUDH0vp0okBhk0aGE6E37KaD7xTUxyYWC8OVWH6cut4PKQxkW/MARty7Gmdy0pKpRBlWFpWqMXxZsx7PCiYFrvlHBmdUuPvfYBbzpRUcymdSU0EL39EpH2shLzH84LsfJSy2ljAPIp93YcR+6H4RoZyIN6DUjSJoGkktDDOgeQ09zuQTvQVkpSm4bLuy8ivU9mIYWkABlYdEQkosKsqTooG2yw0pRhculDOjbC9PV4TR02ba0YRqRgOTQDEOJyyUtKSqTXPJq6JTQu2m/1xgoZOhRyYUY+gOn1+By5GPosYDe9Nua1gXnjPce0oNyILnEWJc3JGTwgE5yUj3G0PuOC40xvPnFV2QegxbNnu3gfLObSIgCUQ399EoHlsNxzcKU8pjVPD50P8DSzkTmkRY7blJb27l6BXumvH4q1EBNaZ/MkRTVBdviKEE0DVVDC3a2aoY+uOSiQlxDT7NLqs853eUy6PG2M3bEO5G1Ts2DwC8uDqY19YjkQskzcQRdpg9dlhStGuj57oAsEEO/0WfodHPGk6K0aHzjmVUAwPWpAd3ziycDulcpS5qkrO+IDMrpOkMydHpOzYxq6ADwmpv34cBctsY7ZXqdIs+udsF50rIIRK/diYseK756UR3QpQzdledU6PtUlblcamG+Y61joWHqqPi9YnZPVbM1dDGgZ0gulusqW/GOCtPQwq6POStF17sWNJYs68+DWsVryUxOL1UlbtY5e88tJZeJwJTfOnVQ0JdebPjkSS7hsfqSi56lobd6dmSREF8jj+xyZq2DPVMm9vrb8fUgKepA11hwLnSz3ntqFRoDrk5hmtNVA1VDSxQXbcS8z3mnFqnazg7SfleEOCCaQHLED75MnQyNvraOVs8RPOgSyUWQy5666LHitM9N1tvb09DDoEpsnXY1ch+630K3a2O1Y0UGNy/OVINrnNbLJXwPiqRoREMPdwFFQgzo2ZWivuTi2ycHSWQS6H3TNehZyYZrWShdLhMGz7YYXX3zDGSOt7n1jqUnWrAC0Y5soZdZYVuUSC5TwpY7C0srHRzaVQ/YWqihu0FbU/G8L270cGR3I7XaLyj/j3nR12Pe57zdK1VtZ+s5JZf/eewU3vGX/xL83JEE9Nfeug8fuuMleNm1C4nny0CLydKKV0AlD+g+Q7e9gD5TM7BnSt4vHCDJJdmsLdKcKwjo6qRo2HHTwlosoNPuyXtcHh96toY+LoZeNfRshh6TXB4521TaabMQnys6DEM3NAaNlYVFARhjr2OMPcYYO84Y+3nJ369gjH2eMfYNxtj9jLHvKP5U1Zgy9aAcGfBsYS/4lU/jg//rmdTnyfqDxOeKWhJrkykEhTg452hLJJcZwWKYhdOrHRyar3uTiTQm2Bajtr6GqYNIz3UpiT2C2AiK0OxakerLvJ0h1zter+16bBHJmxT97CMX8IkHzwULbzBPVLDnVQ0dLx5g5BcVb51e7UDXWFDWLsIMkqIcT11s4ZqFqVTmKGvOZcd86IwxGBoLNXRZ6T8t6L7kEmfoAFJlCfG6q2yLMg191Bm2cZh+t0pArT2LgzDafRvfOLWCl1033Oi2oLbBD+j9IZKiXoWrpIWDf2+rJKxJRGZAZ4zpAN4N4PUAbgHwJsbYLbGH/RcAH+acvwDA9wN4T9EnmgaqEqSg8NTFFtY6Fn75Yw/jmUvJcndC6HKJMvSulQzosuZcMpdLz3Zhu1xqWxRfUwXOOc74AZ0x5vuXSUOPshPGWFBQc92+7IC+KOnnEi83rxqerpvHhz4jGSmWd67oqZU2bDe08HVi+YFh0BAkl/2zSQ86ENPQl1upcgsQHV9GiLfPBeB/ZmqXC33GG10b64qAniZLiJKLntEPnTR01bCMURAhNoqAzhgLcg93P70Cy+F4ec5dVhzxuaKyASJ5UK0kF+aeM7infbsjzyfzIgDHOecnOOd9AH8F4A2xx3AANKtpDsCZ4k4xG0ELXT+QnPVdIj3bwX/46/uU8kurZ8PQWOQGjCdYZWOq0hg6PTfOtOgGvqQYMkG41Oqja7mBQ2OmVon40OMBjxaKXAxd0nFxoxcN6IwxTJl6ZiMxVYVlXLJS4ZTfV4bOR6ahD4ppQXJRzR6loNfsWjiz1sHVC+mfm8joCZbrJvqpVHQt9KErerkA3ue9GmuZQNWiafLIIElRx+GZI9+GhXivpCUTaQzdV45fREVnOHrVLuVj05AM6INLLnQ+8ZxX33alCexJRp53cwjAKeHnJf93In4JwL/zh0h/HMBPyA7EGLuDMXaMMXZseXl5iNOVI97H4+yaVzb/H193E77+1GV84Gsnpc+jToviCl2LacCyajLd7zonc6zEB0QTrvUD7uPnm5Hf92wHb/nzr+NLT1wEEDpcDvmVe7N1I2JbjFviKFCkWRYJC9NVXG71IwucbEpOnjF0qh4oeQL6WscKkoDL8YA+hBOCQLbFpZWO1OEChNfx+IUNcJ7ucAHkrVfjtkXA07WJocv6a8ddLiqGrkLUh66QXITS/2bGyLdhkYehA2Ey+ctPXsQLrtiV2LHmhUxyGUbzliW3LceNuNd2Aop6N28CcCfn/DCA7wDwAcZY4tic8/dxzo9yzo8uLi4W9NIhG6YgdHatg6qh4cf+1TV45Q2L+I1/ehQnL7USz9uQuFE8DThkp+GYqqjU4Q0bTjJ/2iXIkqJX7mng0XPrkd8/craJf37iIn71Hx+G6/Kg7WvA0KuVwLHQs51EVzwKFGlFRYSFaROOy7Hil/9z7t348bFueYqg1hWd/OoxH78Mp4SujyQBdWKtgYfBlOnNsjy33lUydAqGj/kLa5oHHZA7muJJUcBjxySnyRh61dBh6houbfTRsRxpQE9rQUC5BcbUlaKMMegaCzT0ohOiQDSIp/URrxoaLqx38dCZ9aHlFgCB04tsncMUFtH5xDV02czZSUeed3MawBHh58P+70T8CIAPAwDn/KsAagCGv4oDglZ/Ki46s9bFQV+D/o1/cxs0BvzmJx5NPC9u2fOO5fnQSTNVeVWrenILB6gZOgDcvH8Wj5yNMnTqIvjouSY+/cj5sEvgPEkuRlRyMZIMff9sLRcbC6tFvSBKen/8M8hjPWwqRoo1/AR1mt+eXCgAAtcNdbgclaED8D3oqoDuBaHHznnX4aocGjoQdUg4brJ1rKFpqS4XwLuW1Od+TugjvzcHQyepTZUQDc+DBRr6eCQXcWh0uuTylScvgXPg5UMmRAFv53lovo7PPHIBwHCl/4CvoUtsizvJ4QLkC+h3A7ieMXY1Y8yEl/T8aOwxzwB4NQAwxm6GF9CL01QyEO8QeHa1gwN+W9ADc3W87VuuwccfOIcHltYiz2v1k/bCuqmD8/AGDjX0mGZqaArJRa6hA8BNB2bw9KVWJGn40Ok1zNUruGpPA3/42SewtNLBlKkHDM7T0MOkaJyhv/lFV+AnXn2d8rMREa8WVRUHxWeVyqAaKZanhe6py53g35dafkAvICkq7rZkVaJAuDCvtL2ZnlkuEBlD94ZExxi6zmAH7g/5e5iuGYFHPsLQp73vaipD9z8XlWUxOA+NeRr6JjD0tNFtpqGh3XfQMPVIG+dBwRjDt928F186vhx0+BwmCJu6bJQgT91lTCIyPxnOuQ3gHQA+CeAReG6Whxhjv8wY+9f+w94J4EcZY/cB+CCAt3LRFjBmBJ32BA1d7Jv9tm+5GvONCn7nU49FnrfRtRM3dHxyvaojmzc9XpYUTWHoB2bBecgOAY+h33ZoDm9/1XV46Mw6PnbfGRzaVQ90/QRDjwW81992AD/w4nyFN/GAHniVE/KQnqs5l1xDjzqOZDi10sZMzRvuQM3CikiKip95vGsfQdxpZTlcAHnZuOW6icAqJipVDHK6agQ7MDGgz9a9oq+5RoqG7i+UKrmFoGvewpJnnugwiCRFjRTJxb+OL7p698gs+NU370PXcvHl4xdHSIombYs924U5gsS3HZHrk+Gcf5xzfgPn/FrO+a/5v/tFzvlH/X8/zDl/Oef8eZzz53POPzXOk45jKhgQ7MB2XFxo9nBQKBWfqVXw47dfiy88voyvn7gU/F4cEE0I5Bs/wFgSlwv9LLMt0kAGGfO75YBnBHrUD+g928Hj55t4zqE5vPEFh3Bkdx2XWv1AbgE81rbR94ZcxH3og4LcFOKEHCDprvA0dHVAThspFjJ09YKwtNLBkV2NiOum06cq2OEZEy3sGlMPwtA1FgTFazISooC8bNyJVYoC0f4qqgA2XTWC74cY0BljeO9bvgk//PKrledBUltWmbqha7B9yWUcDL0aYejpkguAkfRzwouv2Y0pU8dnHrkwvIYusS16TfeeZQx9EiAy9OWNHhyX48B89Ib+wZdehX2zVfzOpx7DqcttvP0v/wVPLrdweHd0a069RCgxGiRFYzepbHgw4LknqoYm7S99aL6O6aqBR856idHHzjVhORy3HZpDRdfw9ts96USUC2brFXDuFaTIJJdBMFs3YOph+X84rWgwySU+5UiEai6riFOXPVuhN+c0lFzqFX0kTzAtogfm6qmBjxaNfAw9KrlwzpVJ0fhz4hA/r/nYYviqG/fiSEo1pebba7OKYKgQzXb5WDT0SFI0Q0MHMHRBUfRYOl554yI+88h5OC4f0uWS1NCtZ6mGvu1BMkmr5wSTfg7GmjnVKjre8a3X4+6nV/Cq37kLn33kPH7y1dfjp199g/RY7UBy8f4fDxCmLtfQHzvfxHV7p6VbY01juGn/TBDQKSF626E5AMD3vPAwvvWmvXjVjXuD54gl413LGYqdEBhj2CMEUbJQ7opt9cnlolLN0jr50W5JFdA55x5D393A4nQ1WFxkctKgoN3BoYxhDXQtszzoQDIpqprrSRKMrjFlMY+4axumx0rdb+iVBkNjWGmPp3UuIO86KkOtomNXo4Kb988qHzMIXn3TvmBnWZTkEh/+vhMw8UOiAW+bWTW8wo6za55GGWfoAPB9R4/gEw+exb7ZGv7Da2+UdvCLTy2ybDlDr+iadATdE+c38LJr1azkpgMz+PtvnAHnHA/6CdEj/i7BNDS8/63fHHl82M/FTgyAGAYkc6y1LfzR557Ai6/enfCwT1fD3t2y10trh1rPkFwu+ra9I7vq0DWG5Y0eOOceQzdHu7koYKocLgQKRINILsTQKfEZr9akJGlasBF3QsO2ks1Kiuo6w0rLWyTHoqFX8gX0H7/9Wqy0jyQGVgyLV920FxrzRjIWWVg0jkVvK7Fj3g1Z7c76DF0WrE1Dw/9420tSjxMEJCuaFE24XHSW+IKsdSycW+/i+n0zyuPffGAW/9/XnsHSSgcPnF7Dcw7NpsoM9IVb61joO6NJLoDnRV/e6OH3P/M41joW3vVdtyZeX+y4KA3oAUNXu1xUSdFTvmXxyO5G0H/Hk5OckRKi3mvnC+jUtlaVOBURLyyyA4aeTJJ7j1dfH/L7z9SMzOSmDJ4klf4YQ9Nw2Q/o42HovttGY6nB+gVXDFcZqsLuKRMvvGIXjp1ckfabz4JMQ7ecZ6cPfSLQML0+HmfWPNtfmgUs/ThRl4ZqMrhsePATvoRxQ0pflZv9xOj9S2t47JyXEE0DsTqSSYpg6CeWW/jA107iTS+6ArccTG6JwzF08qDcVNgdgWzbIhUVHdndEMbi9dDpjy657Jky8eYXX4HXpYyqAzx3xpFd9VzbbTMmudCQ7ThTpgCdJonRQjlsS9taRU/YJePwJBefoY9RQ98KqeLbbtnnvfYQQVhWKfps9aFPBKjT3jnfsjhscq0e19AVLTYrEg398fNeNdsNKQz9xn0zYAz4+3tPBwnRNBDLurDuB/QRv4ALM1WvG6Sp453ffqP0MVlj6ELJRcbQo20Y4hB7lYc2yn5iQPQw0DSG//u7b5MuUiJqhp46dk5EXEMnmU01fDotQNC1HDag180ckovGgu/uOG2LWzEU4nW37oepMBxkQVYp2tuBlaI7RnJpVL0eIs2ePdQFJ9QTLhe5bVGWFH38fBMNU09Nyk1VDVy5u4HPPepVvmUFdLoplwti6GRd/NnX3IDdij7gWT3RZQOiCXkY+p4pEw3TwB5hilLHchPOj3HhV9/4HOxK6YEuIh7QbVfectXQckguIzL02ZoBPYOoiAF/2F1qGraSoV+1MIV7f/E1Q0lzVcNrC+G4YR8eq0yKbl+Q1e7sagc33jh8n5hGvBkQMTJJd724hv74+Sau3zeTmQi6af8snr7UxmzNyGz8T6yOMvyjBvTvuO0AOpaDf/cSdTESBXQ1Q7fAGILWvSLiO5w4Tq20cdh/z4tCoVO376Au6V8+DgzSY50klCApqvg+EGNPa+1KAX0+pYAoDb/wv90SaPgqiJLMWCWXLWK2wzb5EpPbRNp2ostlx7ybhqljrW1heaOXa/6kCjLJxdSTPZNNI+lyefz8Bm7I0SSLdPTnHJrLlIZqFa+pUxjQR7tk++dqePurrkvtk00McrUtb/W77lfYyhYuTWOoVbTIYF8RXlGRd312T5lgLJRcRunjMi6E8zFjSVGFDz1NQ6ecw7AM/ZrF6VQ5TzyPis6GcoNkobqFDH0UhDutaGvsnTRPFNhBAX3KNPDM5TY4Bw5KLIt5EQQkISkq+/LGGfrlVh8XN3qZNxwA3HzAe0yW3EKYqRmhB3dEhp4HJBlRK984VGX/hIYp79bo+N0kqYDG0DXsapi+5DK6hj4OEBPtx5Kiyfa5eWyLHrscB3Mm0HnNDjnDMwvVLWbowyI+uBrYmZLLjnk3jaoesKdRGDoQHXJhOcnOegBgGtFeLlSkc8P+7ID+/CPzqFd0vOL6fGXRMzUj1NA3ofdEraJjYboaJDDj8BpzpQV0+Ri6c+tdWA6P2AUXpk1cbPYKsS2OA5rfjqAX86EnbKwDaOjz9Xz6/TAgDX1c/mqyLab1cdmOCAdXe9fRdb2K30lbmLKwYzR0sdPeKAwdiA469gJ68qKbsV4ueSyLhL2zNTzwS9+eezzYbL2CZ3y736iSS14c2lVXB/SulZpwa5i6NKEaWhbDBZcKnboFFBaNC6awGws19Oi5EjNOY3yLM1XM1oxc35FhQec1rl0AMd1JC4RxyUXVdG/SsWMCupgs2T8iQ68Lc0V7Cp2tomuREXSPnW9ipmZIBxPLMMisx5maAcqFbZYscXhXHQ+fWZf+bb1jKScCAZ42Hp9dCggBPcLQq7jnpDd3cjsydMCTuSgQWK7chx5KLur3MFU1cN+7vn1MZ+mfhzZuhr51tsVRkLSfyutLJh075t1M+d7pmZox8qTzhqnjQrOLj953Bg+dXpduoyv+EFy60R8/v4Eb9s2MRbcUB0lsZkA/vdKRzmNtKqYVEY7sakSmEhFOrXTAWLQtw8J0NWjXsB01dMDvpW2lM/Q8laKA109nnEOJRQ19HNhK2+IooNxTfM7BpL2PLOyYd0MMPd6UaxhMmQbufnoFP/nBb+DUShuv8SvURLz46t2wXY6f/5sHwDnHE+ebY9tKi2xrsySXw7u80vzl2FBpIDspenhXAxd8XVzE0uU29s/WIix2Ycbc9N3HoKhWQnnNVjD0oFJ0k66PCuPW0GnBGoeDZpxQSS6TttPIwo6RXIihy5pyDYr/87U34oGlVXzTlbtx84EZqTxy+4178c7X3IDf/fTjmKkZWGlbuRwuw0Assd+MpCgQ9kNZWmljnyAjuS7HRi89KUoa+enVTjAcGwCevtTClXuiUg1ViwKjDbcYJ8TWq8TQk719siWXzUCgoY+JoRu6Bo1NXiBMSC6KpnuTjlzvhjH2OsbYY4yx44yxn1c85n9njD3MGHuIMfaXxZ5mNoihj+pwAYBvunIX3vryq3Hb4blUrfsd33odvucFh/AXXz0JIL3kfxREGfrmBIwjQUCPJkabPRucp1chki0xLrs8famd6EG+MB06PrajDx2I9u0JK0VVPvQtZuiBhj4+a6RpaBMXCOMuF2qLPWnvIwuZDJ0xpgN4N4DXAFgCcDdj7KOc84eFx1wP4D8BeDnnfIUxtld+tPGBZngeVEyqGQcYY/j1f3MbllY6uPvk5bEFdJENb1bAoPYJiYCe0gudQEnPU8Jz1zoWLrf6uGpPPKBPAkMXkqI+Q0/40HO4XDYDgYaekuMYFVUjuy/7dkN88hQx9Z02sSjPVX8RgOOc8xMAwBj7KwBvAPCw8JgfBfBuzvkKAHDOLxR9olmgcvUDI/RxGQZVQ8efv/UoHj6zjsWZavYThgAxdNPQCusvnYWGaWDPlJkI6GmNuQh7Z6owdQ1LKyFDf/piC4DXj0OEGNC3q4Yu2hYdVaXodpFc9PEmRQHgit2NzJYV2w2qJmtbvQAXjTwB/RCAU8LPSwBeHHvMDQDAGPsyAB3AL3HOPxE/EGPsDgB3AMAVV1wxzPkqcdOBGfwfL70Srxqhj8uwmKlVBuoPMihI3hi10+KgOLyrHgnKQPq0IoKmMc/HfjlcDJ6+5AX0uOSyZwIkl2pFQ6sVbdaWtC1uF8nFe/1xDm74ux9/GbQxOnXGgXilaOBy0bfnd25YFPXtMwBcD+B2AG8C8N8YY/PxB3HO38c5P8o5P7q4WGzgrRo6/usbnoM90+NhyVsJ0kM3m8Ee3tVIlP+vd9SdFqPPrQfDLADgqYstMIYEs6saYe/67Sq5yG2Lim6LW+xyCSWX8TF0Q9+8nWJRCDX0aFvsUYaSb0fk+fadBnBE+Pmw/zsRSwA+yjm3OOdPAXgcXoAvUQCIbW12sDi8q46l1agXvZkyIFrEkd1RL/rTF1s4OFeXLkoLvlS1XQN6taIHSVGSXJJDoreJ5DLmwqJJRTDs24kVFu0wySXPu7kbwPWMsasZYyaA7wfw0dhjPgKPnYMxtgBPgjlR4Hk+qxEw9E0OFod21dG33WBaEpBPcgG8xWClbQUteJ+61MZVC3LdlXT02jYt/fdsi9FK0fiQaGJ6Wx0gaKEZp4Y+iQg0dH+n1bN3pg89891wzm0A7wDwSQCPAPgw5/whxtgv///t3X9s1Hcdx/Hnq3dtaUuBljLKSgNsY0OGUwibkKmYzWUMCcxoJkoiiUv2j0bURTNjYqL7a4mZzqhLCJuiMcOIZCNzcVHEmCwBLUImA7YBbvwQRsv4ZcOPdrz94/v9Xo+21x5t7673ufcjufTue9/rfT59X9/3/X6+nx+SVsW7vQqckbQf2AF828zOFKrQlSbThl70Jpe4p8u5vmaX5KLosEfocU+XpA3+na7uAT1cEsm86OP2omh2t8UcKxal8lgkuhjSBR4pWq4kUZPuW1c0iWep4zXW8jovM7NXgFf6bft+1n0DvhXf3Bjra0MvdpNLkpQvsShe9PfC5R4aalLDzkXT1xf9EtMbJ3D+Us+AC6KJpC/6uG1ySVfxvyu9vHaoi/PxNYSB0+eOj4uiU+qraahJZRakdn2yl6HrCXTov0e9DNSkq6hNVxX9CLZtSt+lMU12AAAHVElEQVRo0cSFSz15DVppzxppmvRkyXWEvvTWFo50dY/b098PtU7ics811m7cldk2cKTo8AtcFMPaj83igfnTB3zhuOsXivah/66kGidUFz1ZNNSmae7XF324ibkSzQ011FWnOPb+pcwKPf37oCeWL2hl+YLWsSl0ATxydzsPLmhl77Fz7Dl6FqEBS6FlmlxK3MulribFrBxfnJUuewqHUCfn8oReJmZNrR/1PO8j0Tbl+nnRh5uYKyGJ9uao6+LE2hRVg3RZLCeT66pZdvs0lt0+eHfb+vjsqWGEa166wruuySXQXi7+6SsTm75yz4C+z8Uws6kusxoTRAn9psb8vliSaXQnVKdoa6oL7p8n27I7pvGLtYsKuniFG53si6J9Q//D+kyGVZuATaxNl6QXyMx45aLounfUyyXfPs7tzdHApKF6uISiOlXFig/PKOhc5250ptRXZ8ZG+AIXriLNbKrnSu+1zApEF/NscoleW8fFK728eepizh4uzhXLg3e2cvDURd567yJXe6+RrlLZjXgdjid0N6SkL/q7Z7rZc/QsF/K8KBq9Nmozv/rBteCP0N34t/Kum0lViRf3nOBqjqUly523obshJUn5Cxt2Zoa9z2udlNdrsxeD9iN0V2rTGmv5xNwWXtr7X+6bd1OQ13Q8obshzWlpYOVdM5hcV83SW6ey5Jap1015O5TshaRzdVl0rpg+u7CN9Zv38trhLk/orvLUpKv42ZcWjei1k+uqmTQhTffVDzJNN86V0gPzp1Nfk+JIZ3dm4FxIwvuKcuNKe3M97U11QbZXuvJTX5Nm+Z3RIDY/QnfuBq1bOjszmMO58eDhhW1s3XMiuC6L4AndFdgjd7cPv5NzRXTvbS1Ma6ylOh1Wl0XwhO6cqzCpKvHk6gWZCbpC4gndOVdxxvNkcKMRXiOSc85VqLwSuqTlkt6UdEjSE0Ps9zlJJmnx2BXROedcPoZN6JJSwM+Bh4D5wBclzR9kv0ZgPbCr/3POOecKL58j9HuAQ2Z2xMyuApuB1YPs9yTwFHB5DMvnnHMuT/kk9DbgWNbj4/G2DEmLgHYz++NQv0jSY5I6JHV0dnbecGGdc87lNuqLopKqgKeBx4fb18w2mNliM1s8bdrgK78455wbmXwS+gkge3TIzHhbohFYAPxN0jvAEmCbXxh1zrniyieh/xOYK2mOpBpgDbAtedLMzptZi5nNNrPZwE5glZl1FKTEzjnnBjXswCIz65X0NeBVIAU8b2ZvSPoh0GFm24b+DYPbvXt3l6R3b+AlLUDXSN6rzFVivSuxzlCZ9a7EOsPo6j0r1xNK1ooc7yR1mFnFNeNUYr0rsc5QmfWuxDpD4ertI0Wdcy4QntCdcy4Q5ZTQN5S6ACVSifWuxDpDZda7EusMBap32bShO+ecG1o5HaE755wbgid055wLRFkk9Hyn7y1nktol7ZC0X9IbktbH25sl/VnS2/HPplKXtRAkpSTtkfRy/HiOpF1xzH8XD2oLhqQpkrZIOijpgKSllRBrSd+MP9/7JL0gaUJosZb0vKTTkvZlbRs0tor8NK776/G8WCM27hN6vtP3BqAXeNzM5hNNn/DVuJ5PANvNbC6wPX4covXAgazHTwE/NrPbgLPAoyUpVeE8A/zJzOYBHyGqe9CxltQGfB1YbGYLiAYqriG8WP8KWN5vW67YPgTMjW+PAc+O5o3HfUIn/+l7y5qZnTSzf8X3LxL9g7cR1XVTvNsm4OHSlLBwJM0EPgNsjB8LuA/YEu8SVL0lTQY+CTwHYGZXzewcFRBrotHpdZLSQD1wksBibWZ/B97vtzlXbFcDv7bITmCKpBkjfe9ySOjDTt8bGkmzgYVEi4VMN7OT8VOngOklKlYh/QT4DpCs2jsVOGdmvfHj0GI+B+gEfhk3M22U1EDgsTazE8CPgKNEifw8sJuwY53IFdsxzW/lkNAriqSJwB+Ab5jZheznLOpjGlQ/U0krgdNmtrvUZSmiNLAIeNbMFgLd9GteCTTWTURHpHOAm4EGBjZNBK+QsS2HhD7c9L3BkFRNlMx/a2Zb483vJadg8c/TpSpfgdwLrIqnXt5MdPr9DNGpZzJ5XGgxPw4cN7NkucYtRAk+9Fh/GviPmXWaWQ+wlSj+Icc6kSu2Y5rfyiGhDzl9byjiduPngANm9nTWU9uAdfH9dcBLxS5bIZnZd81sZjz18hrgr2a2FtgBfD7eLah6m9kp4JikO+JN9wP7CTzWRE0tSyTVx5/3pN7BxjpLrthuA74c93ZZApzPapq5cWY27m/ACuAt4DDwvVKXp0B1/DjRadjrwN74toKoPXk78DbwF6C51GUt4N/gU8DL8f1bgH8Ah4DfA7WlLt8Y1/WjQEcc7xeBpkqINfAD4CCwD/gNUBtarIEXiK4R9BCdjT2aK7aAiHrxHQb+TdQDaMTv7UP/nXMuEOXQ5OKccy4PntCdcy4QntCdcy4QntCdcy4QntCdcy4QntCdcy4QntCdcy4Q/wfHBXJOXAQT3wAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eG1dRTLarj1T",
        "outputId": "0d45d1ca-481d-4c7d-add1-b81e3d7de7fd"
      },
      "source": [
        "net = SiameseNetwork().cuda()\n",
        "model_save_name = 'classifier3.pt'\n",
        "# model_save_name = 'classifier_body11.pt'\n",
        "path = F\"/content/gdrive/My Drive/{model_save_name}\"\n",
        "net.load_state_dict(torch.load(path))\n",
        "net.eval()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "SiameseNetwork(\n",
              "  (cnn1): Sequential(\n",
              "    (0): ReflectionPad2d((1, 1, 1, 1))\n",
              "    (1): Conv2d(1, 4, kernel_size=(3, 3), stride=(1, 1))\n",
              "    (2): ReLU(inplace=True)\n",
              "    (3): BatchNorm2d(4, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "    (4): ReflectionPad2d((1, 1, 1, 1))\n",
              "    (5): Conv2d(4, 8, kernel_size=(3, 3), stride=(1, 1))\n",
              "    (6): ReLU(inplace=True)\n",
              "    (7): BatchNorm2d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "    (8): ReflectionPad2d((1, 1, 1, 1))\n",
              "    (9): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1))\n",
              "    (10): ReLU(inplace=True)\n",
              "    (11): BatchNorm2d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "  )\n",
              "  (fc1): Sequential(\n",
              "    (0): Linear(in_features=80000, out_features=500, bias=True)\n",
              "    (1): ReLU(inplace=True)\n",
              "    (2): Linear(in_features=500, out_features=500, bias=True)\n",
              "    (3): ReLU(inplace=True)\n",
              "    (4): Linear(in_features=500, out_features=5, bias=True)\n",
              "  )\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 38
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 231
        },
        "id": "mkolQWSXSjDK",
        "outputId": "bfb86190-5474-4e1a-85aa-70a88b1a90bd"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-17-a969ad783903>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0mtest_dataloader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDataLoader\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msiamese_dataset\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mnum_workers\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m6\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m \u001b[0mdataiter1\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0miter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtest_dataloader\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m \u001b[0mM\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mzeros\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdataiter1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdataiter2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     11\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdataiter1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     12\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mNameError\u001b[0m: name 'dataiter2' is not defined"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "75rDQ-9qrFpo",
        "outputId": "6b83a7c5-a30b-4d07-b3d0-3cde3bafdc8e"
      },
      "source": [
        "#@title\n",
        "folder_dataset_test = dset.ImageFolder(root=Config.testing_dir)\n",
        "folder_dataset_test = dset.ImageFolder(root=Config.training_dir)\n",
        "siamese_dataset = SiameseNetworkDataset(imageFolderDataset=folder_dataset_test,\n",
        "                                        # percent_same=0.25,\n",
        "                                        transform=transforms.Compose([transforms.Resize((100,100)),\n",
        "                                                                      transforms.ToTensor()\n",
        "                                                                      ])\n",
        "                                       ,should_invert=False)\n",
        "best_thresh = None\n",
        "best_accuracy = -1\n",
        "\n",
        "\n",
        "for test_it in range(30):\n",
        "    counter = Counter()\n",
        "    threshold = (test_it + 1) * 0.1\n",
        "    test_dataloader = DataLoader(siamese_dataset,num_workers=6,batch_size=1,shuffle=False)\n",
        "    dataiter1 = iter(test_dataloader)\n",
        "\n",
        "    correct = 0\n",
        "    wrong = 0\n",
        "    max_distance = 0\n",
        "    min_distance = 55\n",
        "    for i in range(len(dataiter1)):\n",
        "        x0,x1,label = next(dataiter1)\n",
        "        \n",
        "        with torch.no_grad():\n",
        "            output1,output2 = net(Variable(x0).cuda(),Variable(x1).cuda())\n",
        "        euclidean_distance = F.pairwise_distance(output1, output2)\n",
        "\n",
        "        the_distance = euclidean_distance.item()\n",
        "\n",
        "        if the_distance > max_distance:\n",
        "            max_distance = the_distance\n",
        "        if the_distance < min_distance:\n",
        "            min_distance = the_distance\n",
        "\n",
        "        if the_distance > threshold:\n",
        "            same = 0\n",
        "        else:\n",
        "            same = 1\n",
        "\n",
        "        if same == 0 and label.item() == 0:\n",
        "          counter[\"TN\"] += 1\n",
        "        elif same == 0 and label.item() == 1:\n",
        "          counter[\"FN\"] += 1\n",
        "        elif same == 1 and label.item() == 0:\n",
        "          counter[\"FP\"] += 1\n",
        "        else:\n",
        "          counter[\"TP\"] += 1\n",
        "\n",
        "        if same == label.item():\n",
        "            correct += 1\n",
        "        else:\n",
        "            wrong += 1\n",
        "    acc = correct / (correct + wrong)\n",
        "    print(\"Th: {:.2f}  Accuracy: {:.4f}\".format(threshold, acc))\n",
        "    print(counter)\n",
        "    if acc > best_accuracy:\n",
        "        print(\"\\tnew best acc!!!!\")\n",
        "        best_accuracy = acc\n",
        "        best_thresh = threshold\n",
        "        \n",
        "        \n",
        "          \n",
        "\n",
        "print(\"the best threshold is:\", best_thresh, \"with accuracy: \", best_accuracy)\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "None\n",
            "Th: 0.10  Accuracy: 0.3834\n",
            "Counter({'FN': 838, 'TN': 648, 'FP': 204})\n",
            "\tnew best acc!!!!\n",
            "Th: 0.20  Accuracy: 0.4012\n",
            "Counter({'FN': 815, 'TN': 677, 'FP': 197, 'TP': 1})\n",
            "\tnew best acc!!!!\n",
            "Th: 0.30  Accuracy: 0.3846\n",
            "Counter({'FN': 833, 'TN': 645, 'FP': 207, 'TP': 5})\n",
            "Th: 0.40  Accuracy: 0.3527\n",
            "Counter({'FN': 868, 'TN': 589, 'FP': 226, 'TP': 7})\n",
            "Th: 0.50  Accuracy: 0.3562\n",
            "Counter({'FN': 855, 'TN': 578, 'FP': 233, 'TP': 24})\n",
            "Th: 0.60  Accuracy: 0.3716\n",
            "Counter({'FN': 799, 'TN': 587, 'FP': 263, 'TP': 41})\n",
            "Th: 0.70  Accuracy: 0.3805\n",
            "Counter({'FN': 750, 'TN': 544, 'FP': 297, 'TP': 99})\n",
            "Th: 0.80  Accuracy: 0.3509\n",
            "Counter({'FN': 760, 'TN': 462, 'FP': 337, 'TP': 131})\n",
            "Th: 0.90  Accuracy: 0.3680\n",
            "Counter({'FN': 697, 'TN': 445, 'FP': 371, 'TP': 177})\n",
            "Th: 1.00  Accuracy: 0.3627\n",
            "Counter({'FN': 651, 'FP': 426, 'TN': 393, 'TP': 220})\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kJtYihaov22_"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dY1344_d4IFk"
      },
      "source": [
        "\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}