GoogleColabQuickDrawFastAi_1_0_27.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "GoogleColabQuickDrawFastAi-1.0.27",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "q9eVib2GzZTd",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# QuickDraw Kaggle competition using FastAi and GoogleCollab\n",
        "This notebook is based on [Radek Osmulski](https://www.linkedin.com/in/radek-osmulski-6b935794/) entry for the [QuickDraw Kaggle competition](https://www.kaggle.com/c/quickdraw-doodle-recognition).  The original files are available in this [Github repository](https://github.com/radekosmulski/quickdraw).\n",
        "\n",
        "**Welcome to Google Colab**\n",
        "https://colab.research.google.com/\n",
        "\n",
        "Google Colab provides a free Jupyter Notebook environment that includes 4 NVIDIA Tesla K80 GPUs.\n",
        "\n",
        "**Pricing**\n",
        "Free\n",
        "\n",
        "**Steps**\n",
        "\n",
        "1.   Navigate your browser to https://colab.research.google.com\n",
        "2.   Sign in to your Google account or create a new one\n",
        "3.   File->New Python 3 notebook Click the title to rename it to whatever you like.\n",
        "4.   Runtime->Change runtime type Select GPU from the dropdown menu and click Save\n",
        "5.   Install PyTorch and FastAI version 1.0.27 as shown below"
      ]
    },
    {
      "metadata": {
        "id": "GYpw9mv5DFay",
        "colab_type": "code",
        "outputId": "288f043c-57cc-4925-f030-c708fb87753b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 961
        }
      },
      "cell_type": "code",
      "source": [
        "from os.path import exists\n",
        "from wheel.pep425tags import get_abbr_impl, get_impl_ver, get_abi_tag\n",
        "\n",
        "platform = '{}{}-{}'.format(get_abbr_impl(), get_impl_ver(), get_abi_tag())\n",
        "cuda_output = !ldconfig -p|grep cudart.so|sed -e 's/.*\\.\\([0-9]*\\)\\.\\([0-9]*\\)$/cu\\1\\2/'\n",
        "accelerator = cuda_output[0] if exists('/dev/nvidia0') else 'cpu'\n",
        "\n",
        "!pip install torch_nightly -f https://download.pytorch.org/whl/nightly/{accelerator}/torch_nightly.html\n",
        "!pip install fastai==1.0.27\n",
        "!pip install torchtext==0.2.3\n",
        "\n",
        "import torch\n",
        "\n",
        "print(torch.__version__)\n",
        "print(torch.cuda.is_available())\n",
        "print(torch.backends.cudnn.enabled)"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Looking in links: https://download.pytorch.org/whl/nightly/cu92/torch_nightly.html\n",
            "Collecting torch_nightly\n",
            "  Using cached https://download.pytorch.org/whl/nightly/cu92/torch_nightly-1.0.0.dev20181206-cp36-cp36m-linux_x86_64.whl\n",
            "tcmalloc: large alloc 1073750016 bytes == 0x62b3e000 @  0x7fdb2a21a2a4 0x591a07 0x5b5d56 0x502e9a 0x506859 0x502209 0x502f3d 0x506859 0x504c28 0x502540 0x502f3d 0x506859 0x504c28 0x502540 0x502f3d 0x506859 0x504c28 0x502540 0x502f3d 0x507641 0x502209 0x502f3d 0x506859 0x504c28 0x502540 0x502f3d 0x507641 0x504c28 0x502540 0x502f3d 0x507641\n",
            "Installing collected packages: torch-nightly\n",
            "Successfully installed torch-nightly-1.0.0.dev20181206\n",
            "Requirement already satisfied: fastai==1.0.27 in /usr/local/lib/python3.6/dist-packages (1.0.27)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (2.18.4)\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (3.13)\n",
            "Requirement already satisfied: Pillow in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (4.0.0)\n",
            "Requirement already satisfied: thinc==6.12.0 in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (6.12.0)\n",
            "Requirement already satisfied: bottleneck in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (1.2.1)\n",
            "Requirement already satisfied: torchvision-nightly in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (0.2.1)\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (1.1.0)\n",
            "Requirement already satisfied: numpy>=1.12 in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (1.14.6)\n",
            "Requirement already satisfied: pandas in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (0.22.0)\n",
            "Requirement already satisfied: dataclasses in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (0.6)\n",
            "Requirement already satisfied: regex in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (2018.1.10)\n",
            "Requirement already satisfied: fastprogress>=0.1.15 in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (0.1.18)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (2.1.2)\n",
            "Requirement already satisfied: typing in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (3.6.6)\n",
            "Requirement already satisfied: cymem==2.0.2 in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (2.0.2)\n",
            "Requirement already satisfied: numexpr in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (2.6.8)\n",
            "Requirement already satisfied: spacy==2.0.16 in /usr/local/lib/python3.6/dist-packages (from fastai==1.0.27) (2.0.16)\n",
            "Requirement already satisfied: urllib3<1.23,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->fastai==1.0.27) (1.22)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->fastai==1.0.27) (2018.11.29)\n",
            "Requirement already satisfied: idna<2.7,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->fastai==1.0.27) (2.6)\n",
            "Requirement already satisfied: chardet<3.1.0,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->fastai==1.0.27) (3.0.4)\n",
            "Requirement already satisfied: olefile in /usr/local/lib/python3.6/dist-packages (from Pillow->fastai==1.0.27) (0.46)\n",
            "Requirement already satisfied: msgpack-numpy<0.4.4.0 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (0.4.3.2)\n",
            "Requirement already satisfied: plac<1.0.0,>=0.9.6 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (0.9.6)\n",
            "Requirement already satisfied: murmurhash<1.1.0,>=0.28.0 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (1.0.1)\n",
            "Requirement already satisfied: wrapt<1.11.0,>=1.10.0 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (1.10.11)\n",
            "Requirement already satisfied: tqdm<5.0.0,>=4.10.0 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (4.28.1)\n",
            "Requirement already satisfied: msgpack<1.0.0,>=0.5.6 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (0.5.6)\n",
            "Requirement already satisfied: dill<0.3.0,>=0.2.7 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (0.2.8.2)\n",
            "Requirement already satisfied: six<2.0.0,>=1.10.0 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (1.11.0)\n",
            "Requirement already satisfied: preshed<3.0.0,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (2.0.1)\n",
            "Requirement already satisfied: cytoolz<0.10,>=0.9.0 in /usr/local/lib/python3.6/dist-packages (from thinc==6.12.0->fastai==1.0.27) (0.9.0.1)\n",
            "Requirement already satisfied: python-dateutil>=2 in /usr/local/lib/python3.6/dist-packages (from pandas->fastai==1.0.27) (2.5.3)\n",
            "Requirement already satisfied: pytz>=2011k in /usr/local/lib/python3.6/dist-packages (from pandas->fastai==1.0.27) (2018.7)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib->fastai==1.0.27) (0.10.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->fastai==1.0.27) (2.3.0)\n",
            "Requirement already satisfied: ujson>=1.35 in /usr/local/lib/python3.6/dist-packages (from spacy==2.0.16->fastai==1.0.27) (1.35)\n",
            "Requirement already satisfied: toolz>=0.8.0 in /usr/local/lib/python3.6/dist-packages (from cytoolz<0.10,>=0.9.0->thinc==6.12.0->fastai==1.0.27) (0.9.0)\n",
            "Requirement already satisfied: torchtext==0.2.3 in /usr/local/lib/python3.6/dist-packages (0.2.3)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from torchtext==0.2.3) (4.28.1)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.6/dist-packages (from torchtext==0.2.3) (2.18.4)\n",
            "Requirement already satisfied: idna<2.7,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext==0.2.3) (2.6)\n",
            "Requirement already satisfied: urllib3<1.23,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext==0.2.3) (1.22)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext==0.2.3) (2018.11.29)\n",
            "Requirement already satisfied: chardet<3.1.0,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext==0.2.3) (3.0.4)\n",
            "1.0.0.dev20181206\n",
            "True\n",
            "True\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "U9vuHMxtCwwK",
        "colab_type": "code",
        "outputId": "c5cd7970-9e83-44fe-bb5f-8101f5728d52",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "device_name = tf.test.gpu_device_name()\n",
        "if device_name != '/device:GPU:0':\n",
        "  raise SystemError('GPU device not found')\n",
        "print('Found GPU at: {}'.format(device_name))"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found GPU at: /device:GPU:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "AjpDy_OoDQaa",
        "colab_type": "code",
        "outputId": "655c156d-2086-46b3-d8b5-b84692b12425",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "import fastai\n",
        "print(fastai.__version__)\n",
        "from fastai import *\n",
        "from fastai.vision import *"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1.0.27\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "xXwOLZrRwYpE",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "**Upload utils.py library** \n",
        "1.   Press the > symbol below Google Collab icon.\n",
        "2.   Click on Files\n",
        "3.   Click on upload and select utils.py\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "id": "XdtXE9oNEeq_",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "%matplotlib inline\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "from pathlib import Path\n",
        "\n",
        "import json\n",
        "\n",
        "from utils import *"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "SEXY85gSDdBz",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "NUM_SAMPLES_PER_CLASS = 1_000\n",
        "NUM_VAL = 50 * 340"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "text",
        "id": "nzbWdOUQw6Ep"
      },
      "cell_type": "markdown",
      "source": [
        "1.   Press the > symbol below Google Collab icon\n",
        "2.   Click on upload\n",
        "3.   Upload test_simplified.csv\n",
        "4.   Upload train_simplified.zip\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "id": "7uCMiM8xDwfg",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "!mkdir data\n",
        "PATH = Path('data/txt')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "_-XEER0sDgzK",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "PATH.mkdir(exist_ok=True)\n",
        "(PATH/'train').mkdir(exist_ok=True)\n",
        "(PATH/'test').mkdir(exist_ok=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "OE5_wRnV0U_U",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "941faf2d-4f84-4f77-9fec-78e673eda5ff"
      },
      "cell_type": "code",
      "source": [
        "shutil.move(\"test_simplified.csv\", \"data\")"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'data/test_simplified.csv'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "metadata": {
        "id": "cCImkuMW4kY4",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        ""
      ]
    },
    {
      "metadata": {
        "id": "TtK_CXxp19XE",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "1.   Upload kaggle.json using Google Collab\n",
        "2.   kaggle.json can be found in your Kaggle.com account profile"
      ]
    },
    {
      "metadata": {
        "id": "AQvW0tsi18O-",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 101
        },
        "outputId": "838b876c-d9de-4e4f-ffcb-ec64290ad8ad"
      },
      "cell_type": "code",
      "source": [
        "PATH2 = Path('/root/.kaggle')\n",
        "PATH2.mkdir(exist_ok=True)\n",
        "%ls -a /root\n",
        "%cd /\n",
        "%cd content\n",
        "%pwd\n",
        "shutil.move(\"kaggle.json\", \"/root/.kaggle\")\n",
        "!chmod 600 /root/.kaggle/kaggle.json\n",
        "%ls /root/.kaggle"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[0m\u001b[01;34m.\u001b[0m/   .bashrc  \u001b[01;34m.config\u001b[0m/   \u001b[01;34m.gsutil\u001b[0m/   \u001b[01;34m.jupyter\u001b[0m/  \u001b[01;34m.keras\u001b[0m/  \u001b[01;34m.node-gyp\u001b[0m/  \u001b[01;34m.nv\u001b[0m/\n",
            "\u001b[01;34m..\u001b[0m/  \u001b[01;34m.cache\u001b[0m/  \u001b[01;34m.forever\u001b[0m/  \u001b[01;34m.ipython\u001b[0m/  \u001b[01;34m.kaggle\u001b[0m/   \u001b[01;34m.local\u001b[0m/  \u001b[01;34m.npm\u001b[0m/       .profile\n",
            "/\n",
            "/content\n",
            "kaggle.json\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "sO5rqoP_2aYU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        },
        "outputId": "1cb3a420-4cc8-46ad-ffc7-4e299122f357"
      },
      "cell_type": "code",
      "source": [
        "%more /root/.kaggle/kaggle.json\n",
        "!kaggle competitions download -c quickdraw-doodle-recognition -f train_simplified.zip -p data"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading train_simplified.zip to data\n",
            "100% 7.37G/7.37G [02:59<00:00, 32.3MB/s]\n",
            "100% 7.37G/7.37G [02:59<00:00, 44.1MB/s]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "kVVcmIJ_35iw",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 5746
        },
        "outputId": "7307038d-b603-4601-9b11-fe124fae6bab"
      },
      "cell_type": "code",
      "source": [
        "!unzip data/train_simplified.zip -d train"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Archive:  data/train_simplified.zip\n",
            "  inflating: train/fence.csv         \n",
            "  inflating: train/yoga.csv          \n",
            "  inflating: train/horse.csv         \n",
            "  inflating: train/sandwich.csv      \n",
            "  inflating: train/cat.csv           \n",
            "  inflating: train/camouflage.csv    \n",
            "  inflating: train/mosquito.csv      \n",
            "  inflating: train/see saw.csv       \n",
            "  inflating: train/onion.csv         \n",
            "  inflating: train/grapes.csv        \n",
            "  inflating: train/chandelier.csv    \n",
            "  inflating: train/stitches.csv      \n",
            "  inflating: train/sink.csv          \n",
            "  inflating: train/trumpet.csv       \n",
            "  inflating: train/cell phone.csv    \n",
            "  inflating: train/skyscraper.csv    \n",
            "  inflating: train/hourglass.csv     \n",
            "  inflating: train/donut.csv         \n",
            "  inflating: train/ceiling fan.csv   \n",
            "  inflating: train/bowtie.csv        \n",
            "  inflating: train/lipstick.csv      \n",
            "  inflating: train/suitcase.csv      \n",
            "  inflating: train/moon.csv          \n",
            "  inflating: train/dishwasher.csv    \n",
            "  inflating: train/mermaid.csv       \n",
            "  inflating: train/crocodile.csv     \n",
            "  inflating: train/tiger.csv         \n",
            "  inflating: train/trombone.csv      \n",
            "  inflating: train/hot air balloon.csv  \n",
            "  inflating: train/bread.csv         \n",
            "  inflating: train/bottlecap.csv     \n",
            "  inflating: train/mountain.csv      \n",
            "  inflating: train/pool.csv          \n",
            "  inflating: train/snake.csv         \n",
            "  inflating: train/dog.csv           \n",
            "  inflating: train/ambulance.csv     \n",
            "  inflating: train/backpack.csv      \n",
            "  inflating: train/dumbbell.csv      \n",
            "  inflating: train/elbow.csv         \n",
            "  inflating: train/frog.csv          \n",
            "  inflating: train/screwdriver.csv   \n",
            "  inflating: train/fire hydrant.csv  \n",
            "  inflating: train/squiggle.csv      \n",
            "  inflating: train/dresser.csv       \n",
            "  inflating: train/axe.csv           \n",
            "  inflating: train/mailbox.csv       \n",
            "  inflating: train/cannon.csv        \n",
            "  inflating: train/bear.csv          \n",
            "  inflating: train/hand.csv          \n",
            "  inflating: train/toilet.csv        \n",
            "  inflating: train/garden hose.csv   \n",
            "  inflating: train/tennis racquet.csv  \n",
            "  inflating: train/cup.csv           \n",
            "  inflating: train/drums.csv         \n",
            "  inflating: train/hat.csv           \n",
            "  inflating: train/bucket.csv        \n",
            "  inflating: train/stop sign.csv     \n",
            "  inflating: train/remote control.csv  \n",
            "  inflating: train/key.csv           \n",
            "  inflating: train/penguin.csv       \n",
            "  inflating: train/cookie.csv        \n",
            "  inflating: train/blackberry.csv    \n",
            "  inflating: train/shoe.csv          \n",
            "  inflating: train/popsicle.csv      \n",
            "  inflating: train/cruise ship.csv   \n",
            "  inflating: train/rabbit.csv        \n",
            "  inflating: train/spreadsheet.csv   \n",
            "  inflating: train/microwave.csv     \n",
            "  inflating: train/ant.csv           \n",
            "  inflating: train/parrot.csv        \n",
            "  inflating: train/wine bottle.csv   \n",
            "  inflating: train/bat.csv           \n",
            "  inflating: train/traffic light.csv  \n",
            "  inflating: train/whale.csv         \n",
            "  inflating: train/wheel.csv         \n",
            "  inflating: train/hockey stick.csv  \n",
            "  inflating: train/squirrel.csv      \n",
            "  inflating: train/panda.csv         \n",
            "  inflating: train/line.csv          \n",
            "  inflating: train/picture frame.csv  \n",
            "  inflating: train/finger.csv        \n",
            "  inflating: train/sailboat.csv      \n",
            "  inflating: train/compass.csv       \n",
            "  inflating: train/mouth.csv         \n",
            "  inflating: train/eraser.csv        \n",
            "  inflating: train/ladder.csv        \n",
            "  inflating: train/bench.csv         \n",
            "  inflating: train/frying pan.csv    \n",
            "  inflating: train/goatee.csv        \n",
            "  inflating: train/school bus.csv    \n",
            "  inflating: train/camel.csv         \n",
            "  inflating: train/flamingo.csv      \n",
            "  inflating: train/baseball bat.csv  \n",
            "  inflating: train/floor lamp.csv    \n",
            "  inflating: train/hedgehog.csv      \n",
            "  inflating: train/stereo.csv        \n",
            "  inflating: train/bridge.csv        \n",
            "  inflating: train/broccoli.csv      \n",
            "  inflating: train/blueberry.csv     \n",
            "  inflating: train/bathtub.csv       \n",
            "  inflating: train/cooler.csv        \n",
            "  inflating: train/coffee cup.csv    \n",
            "  inflating: train/grass.csv         \n",
            "  inflating: train/pickup truck.csv  \n",
            "  inflating: train/toe.csv           \n",
            "  inflating: train/monkey.csv        \n",
            "  inflating: train/string bean.csv   \n",
            "  inflating: train/soccer ball.csv   \n",
            "  inflating: train/zigzag.csv        \n",
            "  inflating: train/helicopter.csv    \n",
            "  inflating: train/smiley face.csv   \n",
            "  inflating: train/pig.csv           \n",
            "  inflating: train/octagon.csv       \n",
            "  inflating: train/skull.csv         \n",
            "  inflating: train/hospital.csv      \n",
            "  inflating: train/firetruck.csv     \n",
            "  inflating: train/leaf.csv          \n",
            "  inflating: train/megaphone.csv     \n",
            "  inflating: train/apple.csv         \n",
            "  inflating: train/stove.csv         \n",
            "  inflating: train/lightning.csv     \n",
            "  inflating: train/fireplace.csv     \n",
            "  inflating: train/banana.csv        \n",
            "  inflating: train/fork.csv          \n",
            "  inflating: train/foot.csv          \n",
            "  inflating: train/saw.csv           \n",
            "  inflating: train/purse.csv         \n",
            "  inflating: train/sword.csv         \n",
            "  inflating: train/saxophone.csv     \n",
            "  inflating: train/washing machine.csv  \n",
            "  inflating: train/parachute.csv     \n",
            "  inflating: train/alarm clock.csv   \n",
            "  inflating: train/anvil.csv         \n",
            "  inflating: train/birthday cake.csv  \n",
            "  inflating: train/snowman.csv       \n",
            "  inflating: train/beach.csv         \n",
            "  inflating: train/boomerang.csv     \n",
            "  inflating: train/snorkel.csv       \n",
            "  inflating: train/underwear.csv     \n",
            "  inflating: train/pillow.csv        \n",
            "  inflating: train/keyboard.csv      \n",
            "  inflating: train/sleeping bag.csv  \n",
            "  inflating: train/pear.csv          \n",
            "  inflating: train/church.csv        \n",
            "  inflating: train/ocean.csv         \n",
            "  inflating: train/headphones.csv    \n",
            "  inflating: train/lighthouse.csv    \n",
            "  inflating: train/star.csv          \n",
            "  inflating: train/campfire.csv      \n",
            "  inflating: train/rake.csv          \n",
            "  inflating: train/shark.csv         \n",
            "  inflating: train/camera.csv        \n",
            "  inflating: train/guitar.csv        \n",
            "  inflating: train/rain.csv          \n",
            "  inflating: train/calendar.csv      \n",
            "  inflating: train/vase.csv          \n",
            "  inflating: train/flower.csv        \n",
            "  inflating: train/sweater.csv       \n",
            "  inflating: train/mug.csv           \n",
            "  inflating: train/jail.csv          \n",
            "  inflating: train/telephone.csv     \n",
            "  inflating: train/pencil.csv        \n",
            "  inflating: train/speedboat.csv     \n",
            "  inflating: train/scorpion.csv      \n",
            "  inflating: train/flying saucer.csv  \n",
            "  inflating: train/jacket.csv        \n",
            "  inflating: train/microphone.csv    \n",
            "  inflating: train/cloud.csv         \n",
            "  inflating: train/marker.csv        \n",
            "  inflating: train/shovel.csv        \n",
            "  inflating: train/raccoon.csv       \n",
            "  inflating: train/steak.csv         \n",
            "  inflating: train/face.csv          \n",
            "  inflating: train/brain.csv         \n",
            "  inflating: train/passport.csv      \n",
            "  inflating: train/basketball.csv    \n",
            "  inflating: train/diamond.csv       \n",
            "  inflating: train/train.csv         \n",
            "  inflating: train/tent.csv          \n",
            "  inflating: train/beard.csv         \n",
            "  inflating: train/flip flops.csv    \n",
            "  inflating: train/kangaroo.csv      \n",
            "  inflating: train/tornado.csv       \n",
            "  inflating: train/house plant.csv   \n",
            "  inflating: train/lantern.csv       \n",
            "  inflating: train/canoe.csv         \n",
            "  inflating: train/castle.csv        \n",
            "  inflating: train/diving board.csv  \n",
            "  inflating: train/barn.csv          \n",
            "  inflating: train/bed.csv           \n",
            "  inflating: train/nail.csv          \n",
            "  inflating: train/hockey puck.csv   \n",
            "  inflating: train/peanut.csv        \n",
            "  inflating: train/bush.csv          \n",
            "  inflating: train/bus.csv           \n",
            "  inflating: train/animal migration.csv  \n",
            "  inflating: train/book.csv          \n",
            "  inflating: train/river.csv         \n",
            "  inflating: train/pizza.csv         \n",
            "  inflating: train/sheep.csv         \n",
            "  inflating: train/lion.csv          \n",
            "  inflating: train/postcard.csv      \n",
            "  inflating: train/sea turtle.csv    \n",
            "  inflating: train/roller coaster.csv  \n",
            "  inflating: train/ice cream.csv     \n",
            "  inflating: train/cello.csv         \n",
            "  inflating: train/strawberry.csv    \n",
            "  inflating: train/fan.csv           \n",
            "  inflating: train/drill.csv         \n",
            "  inflating: train/calculator.csv    \n",
            "  inflating: train/octopus.csv       \n",
            "  inflating: train/pineapple.csv     \n",
            "  inflating: train/hexagon.csv       \n",
            "  inflating: train/dragon.csv        \n",
            "  inflating: train/bicycle.csv       \n",
            "  inflating: train/square.csv        \n",
            "  inflating: train/tractor.csv       \n",
            "  inflating: train/toaster.csv       \n",
            "  inflating: train/butterfly.csv     \n",
            "  inflating: train/harp.csv          \n",
            "  inflating: train/crayon.csv        \n",
            "  inflating: train/t-shirt.csv       \n",
            "  inflating: train/necklace.csv      \n",
            "  inflating: train/paintbrush.csv    \n",
            "  inflating: train/bandage.csv       \n",
            "  inflating: train/arm.csv           \n",
            "  inflating: train/paint can.csv     \n",
            "  inflating: train/hot dog.csv       \n",
            "  inflating: train/swing set.csv     \n",
            "  inflating: train/door.csv          \n",
            "  inflating: train/dolphin.csv       \n",
            "  inflating: train/mouse.csv         \n",
            "  inflating: train/piano.csv         \n",
            "  inflating: train/mushroom.csv      \n",
            "  inflating: train/nose.csv          \n",
            "  inflating: train/bird.csv          \n",
            "  inflating: train/fish.csv          \n",
            "  inflating: train/table.csv         \n",
            "  inflating: train/light bulb.csv    \n",
            "  inflating: train/crown.csv         \n",
            "  inflating: train/duck.csv          \n",
            "  inflating: train/The Eiffel Tower.csv  \n",
            "  inflating: train/house.csv         \n",
            "  inflating: train/palm tree.csv     \n",
            "  inflating: train/cactus.csv        \n",
            "  inflating: train/hot tub.csv       \n",
            "  inflating: train/bee.csv           \n",
            "  inflating: train/bracelet.csv      \n",
            "  inflating: train/wine glass.csv    \n",
            "  inflating: train/The Mona Lisa.csv  \n",
            "  inflating: train/bulldozer.csv     \n",
            "  inflating: train/scissors.csv      \n",
            "  inflating: train/moustache.csv     \n",
            "  inflating: train/owl.csv           \n",
            "  inflating: train/eyeglasses.csv    \n",
            "  inflating: train/helmet.csv        \n",
            "  inflating: train/teapot.csv        \n",
            "  inflating: train/streetlight.csv   \n",
            "  inflating: train/candle.csv        \n",
            "  inflating: train/swan.csv          \n",
            "  inflating: train/map.csv           \n",
            "  inflating: train/angel.csv         \n",
            "  inflating: train/sock.csv          \n",
            "  inflating: train/rhinoceros.csv    \n",
            "  inflating: train/wristwatch.csv    \n",
            "  inflating: train/garden.csv        \n",
            "  inflating: train/leg.csv           \n",
            "  inflating: train/golf club.csv     \n",
            "  inflating: train/hurricane.csv     \n",
            "  inflating: train/matches.csv       \n",
            "  inflating: train/umbrella.csv      \n",
            "  inflating: train/pants.csv         \n",
            "  inflating: train/laptop.csv        \n",
            "  inflating: train/pond.csv          \n",
            "  inflating: train/spoon.csv         \n",
            "  inflating: train/asparagus.csv     \n",
            "  inflating: train/stairs.csv        \n",
            "  inflating: train/eye.csv           \n",
            "  inflating: train/submarine.csv     \n",
            "  inflating: train/cow.csv           \n",
            "  inflating: train/couch.csv         \n",
            "  inflating: train/broom.csv         \n",
            "  inflating: train/flashlight.csv    \n",
            "  inflating: train/potato.csv        \n",
            "  inflating: train/baseball.csv      \n",
            "  inflating: train/tree.csv          \n",
            "  inflating: train/circle.csv        \n",
            "  inflating: train/radio.csv         \n",
            "  inflating: train/clock.csv         \n",
            "  inflating: train/hammer.csv        \n",
            "  inflating: train/sun.csv           \n",
            "  inflating: train/stethoscope.csv   \n",
            "  inflating: train/ear.csv           \n",
            "  inflating: train/watermelon.csv    \n",
            "  inflating: train/cake.csv          \n",
            "  inflating: train/paper clip.csv    \n",
            "  inflating: train/windmill.csv      \n",
            "  inflating: train/violin.csv        \n",
            "  inflating: train/belt.csv          \n",
            "  inflating: train/pliers.csv        \n",
            "  inflating: train/elephant.csv      \n",
            "  inflating: train/knee.csv          \n",
            "  inflating: train/triangle.csv      \n",
            "  inflating: train/feather.csv       \n",
            "  inflating: train/computer.csv      \n",
            "  inflating: train/rainbow.csv       \n",
            "  inflating: train/lollipop.csv      \n",
            "  inflating: train/police car.csv    \n",
            "  inflating: train/shorts.csv        \n",
            "  inflating: train/skateboard.csv    \n",
            "  inflating: train/rollerskates.csv  \n",
            "  inflating: train/zebra.csv         \n",
            "  inflating: train/motorbike.csv     \n",
            "  inflating: train/spider.csv        \n",
            "  inflating: train/snail.csv         \n",
            "  inflating: train/toothbrush.csv    \n",
            "  inflating: train/binoculars.csv    \n",
            "  inflating: train/waterslide.csv    \n",
            "  inflating: train/truck.csv         \n",
            "  inflating: train/chair.csv         \n",
            "  inflating: train/airplane.csv      \n",
            "  inflating: train/peas.csv          \n",
            "  inflating: train/envelope.csv      \n",
            "  inflating: train/tooth.csv         \n",
            "  inflating: train/The Great Wall of China.csv  \n",
            "  inflating: train/crab.csv          \n",
            "  inflating: train/van.csv           \n",
            "  inflating: train/clarinet.csv      \n",
            "  inflating: train/car.csv           \n",
            "  inflating: train/carrot.csv        \n",
            "  inflating: train/basket.csv        \n",
            "  inflating: train/snowflake.csv     \n",
            "  inflating: train/power outlet.csv  \n",
            "  inflating: train/hamburger.csv     \n",
            "  inflating: train/lobster.csv       \n",
            "  inflating: train/toothpaste.csv    \n",
            "  inflating: train/teddy-bear.csv    \n",
            "  inflating: train/television.csv    \n",
            "  inflating: train/oven.csv          \n",
            "  inflating: train/giraffe.csv       \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "QZOYpNuy52YG",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "2987e3ea-fe86-4234-f511-6bbeef68546a"
      },
      "cell_type": "code",
      "source": [
        "shutil.move(\"train\", \"data\")\n",
        "!ls data/train | wc -l"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "340\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "c4JpHGoC6P0K",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "There should be 340 folders under data/train"
      ]
    },
    {
      "metadata": {
        "id": "12n2R-04DpPS",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def create_train_txts_from_df(path):\n",
        "    df = pd.read_csv(path)\n",
        "    klass = '_'.join(path.stem.split())\n",
        "    (PATH/'train'/klass).mkdir(exist_ok=True)\n",
        "    for row in df.sample(NUM_SAMPLES_PER_CLASS).iterrows():\n",
        "        example = {\n",
        "            'countrycode': row[1].countrycode,\n",
        "            'drawing': json.loads(row[1].drawing),\n",
        "            'key_id': row[1].key_id,\n",
        "            'recognized': row[1].recognized\n",
        "        }\n",
        "        with open(PATH/'train'/klass/f'{example[\"key_id\"]}.txt', mode='w') as f: json.dump(example, f)\n",
        "\n",
        "def create_test_txts_from_df(path):\n",
        "    df = pd.read_csv(path)\n",
        "    for row in df.iterrows():\n",
        "        example = {\n",
        "            'countrycode': row[1].countrycode,\n",
        "            'drawing': json.loads(row[1].drawing),\n",
        "            'key_id': row[1].key_id\n",
        "        }\n",
        "        with open(PATH/'test'/f'{example[\"key_id\"]}.txt', mode='w') as f: json.dump(example, f)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "7VYUrkp6D5L6",
        "colab_type": "code",
        "outputId": "6993f059-a3da-45cb-ab7e-0e3efada6974",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        }
      },
      "cell_type": "code",
      "source": [
        "%time create_test_txts_from_df('data/test_simplified.csv')"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CPU times: user 54.5 s, sys: 7.1 s, total: 1min 1s\n",
            "Wall time: 1min 2s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Z9n_D_06e7Zw",
        "colab_type": "code",
        "outputId": "e19b7002-51ed-479b-9843-f69ca9c7a1d5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        }
      },
      "cell_type": "code",
      "source": [
        "!ls data/txt/test/9000003627287624.txt\n",
        "!ls data/txt/test | wc -l"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data/txt/test/9000003627287624.txt\n",
            "112199\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "T0EpFOn-7q70",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "There should be 112199 entries in the data/txt/test folder.  These are the samples that will be predict on and they will be used for the submission below"
      ]
    },
    {
      "metadata": {
        "id": "2SAYpy6JD6LY",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        },
        "outputId": "2dfb237e-eec7-428f-f7d1-0ec93259ff3e"
      },
      "cell_type": "code",
      "source": [
        "%time for p in Path('data/train').iterdir(): create_train_txts_from_df(p)"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CPU times: user 8min 9s, sys: 37.3 s, total: 8min 46s\n",
            "Wall time: 9min 19s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "CFulNnfM-WJO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        },
        "outputId": "7a3981b4-6d02-4671-fd74-47e66716c74d"
      },
      "cell_type": "code",
      "source": [
        "%pwd\n",
        "%ls -l data/txt/train/crab | wc -l\n",
        "%ls -l data/txt/train/horse | wc -l\n",
        "%ls -l data/txt/train/sheep | wc -l"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1001\n",
            "1001\n",
            "1001\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "FPCIzgcR_H7S",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "There should be 1001 samples in each folder for each of the pictures.  Note that the label for the figure is in the folder (e.g. crab, horse, sheep, etc.)"
      ]
    },
    {
      "metadata": {
        "id": "2TmcCFJLEMxv",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "sz = 128\n",
        "bs = 640"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "FHniu9IKEM0Z",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def create_func(path):\n",
        "    with open(path) as f: j = json.load(f)\n",
        "    drawing = list2drawing(j['drawing'], size=sz)\n",
        "    tensor = drawing2tensor(drawing)\n",
        "    return Image(tensor.div_(255))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "eno-kaNtET73",
        "colab_type": "code",
        "outputId": "63118c08-59a3-46d3-b448-06313e7b75ed",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 87
        }
      },
      "cell_type": "code",
      "source": [
        "item_list = ItemList.from_folder(PATH/'train', create_func=create_func)\n",
        "print(item_list)"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ItemList (340000 items)\n",
            "[Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128)]...\n",
            "Path: data/txt/train\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "GrX2wdXe_h8K",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "there should be 340,000 images that will be use for our training."
      ]
    },
    {
      "metadata": {
        "id": "QKChw2B1ET_H",
        "colab_type": "code",
        "outputId": "825abe22-6b20-4563-b1ce-4829d508a669",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        }
      },
      "cell_type": "code",
      "source": [
        "idxs = np.arange(item_list.items.shape[0])\n",
        "print(idxs)\n",
        "np.random.shuffle(idxs)\n",
        "val_idxs = idxs[:NUM_VAL]\n",
        "print(val_idxs)\n",
        "print(len(val_idxs))"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[     0      1      2      3 ... 339996 339997 339998 339999]\n",
            "[313783 295016  24344 138824 ...   1281 269085  18742  84822]\n",
            "17000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "YToiuoEP_qVF",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "We select 17,000 samples randomly.  This is set with the NUM_VAL constant above.  17,000 = 50 * 340. 340 is the number of labels."
      ]
    },
    {
      "metadata": {
        "id": "U5p7lmzgE5vN",
        "colab_type": "code",
        "outputId": "675bb104-faf9-443a-f1aa-8b06fa98d110",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 171
        }
      },
      "cell_type": "code",
      "source": [
        "item_lists = item_list.split_by_idx(val_idxs)\n",
        "print(item_lists)"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ItemLists;\n",
            "Train: ItemList (323000 items)\n",
            "[Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128)]...\n",
            "Path: data/txt/train;\n",
            "Valid: ItemList (17000 items)\n",
            "[Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128), Image (3, 128, 128)]...\n",
            "Path: data/txt/train;\n",
            "Test: None\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "mmOZYvTWFL9V",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "label_lists = item_lists.label_from_folder()\n",
        "pd.to_pickle(label_lists.train.y.classes, 'data/classes.pkl')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "pgwOEZzzAaAD",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Refresh the files on the right.  You see a new file created: `data/classess.pkl`"
      ]
    },
    {
      "metadata": {
        "id": "BuSSfuFEFrgj",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "classes = pd.read_pickle('data/classes.pkl')\n",
        "\n",
        "label_lists = item_lists.label_from_folder(classes=classes)\n",
        "\n",
        "test_items = ItemList.from_folder(PATH/'test', create_func=create_func)\n",
        "label_lists.add_test(test_items);\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "2BQBDg3OJ6_o",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "\n",
        "train_dl = DataLoader(label_lists.train, bs, True, num_workers=12)\n",
        "valid_dl = DataLoader(label_lists.valid, 2*bs, False, num_workers=12)\n",
        "test_dl = DataLoader(label_lists.test, 2*bs, False, num_workers=12)\n",
        "\n",
        "data_bunch = ImageDataBunch(train_dl, valid_dl, test_dl)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ZeZIv4fjA_Zu",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Dataloader is a FastAi class to load the labels for the train and test set."
      ]
    },
    {
      "metadata": {
        "id": "g6FIoPBYKET1",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "pd.to_pickle(data_bunch.batch_stats(), f'data/batch_stats_{sz}.pkl')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "n9eF-l4gBMi9",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Refresh files on right panel and check that a new file called `batch_stats_128.pkl` is created."
      ]
    },
    {
      "metadata": {
        "id": "3wxMJwF9KSt8",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "batch_stats = pd.read_pickle(f'data/batch_stats_{sz}.pkl')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "SHsqXlx1Kivk",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "data_bunch.normalize(batch_stats);"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "FmW3k5ckKmpV",
        "colab_type": "code",
        "outputId": "972c8b34-f359-4ab2-f0e4-39b4bcbdc59f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 719
        }
      },
      "cell_type": "code",
      "source": [
        "data_bunch.show_batch(rows=4)"
      ],
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAoAAAAK+CAYAAAAyva/+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzsvXm4LUlV5v3u4iI2IFBUWWoLJSIQ\nOPMJ4tDAyW2LqLSFIIMCimIhoOKEQztUn70ZRFREERkUEZC2GimFYtAP5XPvC3wO0CAqIgElgkwC\n1WChiFgU2X9k5r5x4mRkRo4rIuP9Pc966tbZU2bEmytWrJhWeZ6DEEIIIYSkw3nSF0AIIYQQQuaF\nASAhhBBCSGIwACSEEEIISQwGgIQQQgghicEAkBBCCCEkMRgAEkIIIYQkBgNAD5RSt1JKfaLjZ75C\nKfUlPX7rKqVU1vVzhPRBKbVXSj1Y+jpIeiil3qGUuov0dRCilHqeUuqb+rT1McMAcDq+C0DnAJAQ\nQggh86G1/g6t9Uulr2NuzkhfQEwopR4K4IcAnA/gxwG8AMBjAXxL+ZY/B/B9AL4dwHcAuEQpdZHW\n+pcavvOOAJ4H4PoAXm69dj8Axyjq6b0AHqa1/nul1M0BXAHgNgD+AsA1AN6ttd6Mc6ckZJRSZwA8\nA8BdAVwPwF8D2AB4BYAnAHgYgJsD+BGt9QuUUucB+FUAXwvgUwC8BsBDtdbXWt/79QCeAuC/ALi2\n/MxXoNDfY7XWvzX5zZGgUErdCsCfAXgygO8GsELh2y4DcAcAr9BaP7TBV20AXAjgswF8KYCrAdxL\na/0+63cuRaHbDMAFAJ4OQJUv/6DW+g+VUq8D8ESt9RXlZ/4bgMdpre8wzd2T0Cl182gUunsfirZ3\nDeCeAD6Cwkd+AsD9AHwOCv18sfH5NwL476U9C4VvTAZmAP05D8CnaK2/BMAPA3gcgPsD+AYAdwTw\nhQBuBuCHtdbPAPBaAD/eFPyVPB3Ar2itbwfgTwF8LgAopS4G8BsAvllrfXsUweEzy8/8FIAPaq0v\nBvBzAL5ttLskMXAPFDq5PYDbAvhbAF+FoqH9ZOngfgiFRgHg3igc4RcB+HwUen2A+YVKKYVCi/fS\nWn8QwJMAfLL8ja8AsFVKfdG0t0UC5UIA/6S1Vig6Gy8A8BAUIxwPVEodwe2rgKLx/SEAnwfgAwAe\nan55OQz8EwAu0Vp/DMBzAbyx9InfCOD5SqkLAFwO4IHGR+8N4H+NfK8kEpRSFwF4KoC7a61vC+Aq\nFB0ToNDN00oN7VDo75UAbqGUqtrYzwVwi/LvScIA0J8VikwdAPwlCuHcE8BztdYf1VpfB+C3AHyd\n7xcqpT4VwJejcKhAkdX7aPnvuwPYaa2vKv//WQDWZfbnriicIbTWr0eRBSTp8EEAX4CiAbyh1voy\nFNm/Myg0CABvAHAxAGitfw/AnbTW12qt/x3A6wDc2vi+mwL4PRRZm78r//ZNKDomnywDwt8HcJ9p\nb4sEyhkALyz//TcAXqe1vlpr/X9QZF0yuH0VALxKa/1OrXWOwndebHz3xQCeA+A+Wuv3K6VuhCKD\n82QAKL/z1Sh87QsAfL1S6qZKqeuh0OjvTnHDJHy01h8AcBOt9bvLP70a5/zam8u2ESh9odb6PwC8\nFMAl5d/vDeDFWutk5vzZMAD05zqt9b9V/0Yx9PbpAD5svOfDAC7q8J03L//7EQAoHeQ/l3878d1a\n62tQBKEXohiC/pDxPe/p8JskcrTWrwXwqNL+SSn1Oyiyz9dprasORKVRKKU+HcDzlFJvVUq9BcC9\ncPLZfyyKDs17jb/dDMDvKqXeUn7m3gBuMuFtkXC5rszMAYWu/tV8DcDH4fZVQDFFxXz/9Yz//2UA\nN0aRGQSKzsgKwJ8a2rsTgJtprd+DYmTlPiimKbxDa/324bdHYqTsBDxGKfVmpZQG8Hic82suzV2B\ncwHgN+Nc8iVJOAdwGO9HMV+l4oLyb75UTvMmAK4p52pVQeH7UQzrAQCUUuejGJK7GkXAeGPjez4L\nwN93unISNeU8qCvK+aDPBvBjDW9/PIo5fV+stf64Uup/Wq8/BUUm53lKqa8se8TvRTGk96YJLp8s\niyZf1caPo5jr90wUDfIHUDTYd9Ja/2vN+y9HMaR8FRJvvAkegCKYu5vW+mql1MMAPKjlM68A8FtK\nqdsCuB2AP5n4GoOGGcBhvAzAg5VSNyyHO74b5xZyXIsii+Kk7FX/FYrsCgB8K4BPLf/9xwDuppSq\nUtqPAPBHZeP8WhROEEqpOwC48zi3Q2JAKfVdSqnLAEBr/SEAbwGQN3zkIgB/UwZ/X4oie2J2IK7S\nWj8TRVb5p8u/XYlCc1BKnVFKPVkp9WUj3wpZBtfC7avauArF4pHbKKUeUn7m5TinvRsqpZ6tlLpl\n+f4XArgLgPuCw7+pcxGKLPDV5RzR++OkXzuF1vrjKILAnwdwZTl1K1kYAA7jCgB/AOD1AN4E4F0o\nsikA8CIAT1RKtS0CeSSAn1BKvRVFIPdmACjnNVwK4MpyGORuAB5efubxKObtX4ViBdSVaA4AyLK4\nEsAdlVJvU0r9HYr5gE06exKAR5Tv/T4Umrm0XLlp8t0AHlUGepcBuGk5tPK3OLfamBCbJl/VSjk3\n6yEAflEpdQsUPvGo/K43AHi71vpd5Xs/BOBVAP6h+htJlssBXFC2g5cD+BkAt0Th75q4AkW2OfkO\nxCrPGTfEiFJqVc4ZhFLqhQBeo7X+FeHLIoSQSVFKPQ3Am7TWT5O+FkJihhnACFFKfT+AlyilziuX\nwmco9uoihJDFUs7d+kYA9jxWQkhHuAhkYpRS34Fi3746nqu1fkKPr30OiqDvbSgmWz+pXBlKCCGL\nRCn1GBQb/X5/udKYEDIADgETQgghhCQGh4AJIYQQQhJDdAh4tVqJpB+zLEOWZTg+Pu782e12CwDY\nbDYjX1Ua5Hm+kr6GkhPa2+/3WK/XUtdCZiAg7Yn5vpDYbDYnfPDSn8FQ9EftzYMZIxwdHSHLMpHr\n2O/3yLKsVnuiQ8CmELMsw36/n+y3hgR9TTAg7EYoThCObXNWq1Auj4xNQNpjIwygru1Zr9eTtgOS\nhKK//X6fSwUjRIx67eV5LmYA8s1mk1dsNpscRcM8mm02m3y32+Vt7Ha7w+9vNpsT19WF6rNj38dS\nTFJvltWSZZl4GdEWrz1Il4W0ZVnm9J/S15aA/kggVHHHFHGDFffU+yHXC3NYXWDme3NmYGcXXJ+g\nr+23+gaEY7Db7RYTmOTyzq/VCS6lrGnBag/SZSFtLn/KAFDW95FpMIO8udoX6xmr1YL0KuDaH28b\nBrDnjgDFOPfZs2dbh3ir9w0Zsq0+O/ZwchtLmCOTBzIMslqt8qoe6+Zn+Ghwv98vdrhqiYSiPYBD\nwLvdzjknaqnTMELR33q95hDwDEhPC7OesfCGgM3w1M7YNUXJPtm9ugh8qsg7yzLvrOMYxNxLzuV7\nv7UZmLohKVeG2Pd9tLBMWnNN+kvNmpC+tqXrT7ocaNNbTSxSq4UgMoDb7Rb7/R673e7Ei64sTLWY\no1o44urNVN87d5amWnAyBj6rh6qsZvXvkLNSeSC94LoMTJZlpzRYYZZxU+Z3jAwzmYZQtAeknQGs\nW/1r+jhmAKclZe0tnYY2LNwMYJU9qcvC2Fm7LMvy3W53Yl6cGe3OOcY+l7kmTPsw1QTTviast9Ze\n8JCyNtntduJlTQtTe036S8Hs+X/2/0tf39L1J10OtGnMNQJZtmn1WhAW46kHvi0ItG+yCgQrk66E\nqSt4rGFmc/XRnOU2o7Z6O8Gqk+Fbjk0sXZMxmbTmfPW3dLOfH9vnL/WZkdYctbdMa1qgWiV+nFoQ\nFmOeF4o8YU1BYIqrx+qsmnc45tzDOeayjayfyZ1gW/mamT6XNpfaoMVm0prro7+lme3bjQbq1N+W\nZtKaS117S7OmRIW9c4hLC0HMAayb81E3ll2tgK3m/9Wx1PkjXTDnn/VZqbzf7w/zJ8cmj3geTLXy\n19afPVe1boXjkje4jYVQtAekOw/Lnv9X+WuzHarm0TbNZzbngZtUG/Ob3xXKcxeK/lLVXqx0bc/r\n2hqn9oR7I4doFY4I16aKal1ztZa0Z97Y1jVrOMU1COttlF5wXabP1lxd+VKXsiatubH0F7tV1A3/\n+tB11COUjKK05iqTLgdavZntcx+a5p07tSAsxgOuh7RtTiCHhMexunKcYiGDsN5Gc4I+m5hLBIGV\nE2GwGa72xtBf7FbN2Z4L6fsNSX/S5ZC6jT19yyfp5dSCsBhP4LqJtiCwaSycDWE3YU4dtAjrbVQn\naJdVXafDtb/g0N82V3e7YDY8TO2Npb/YbazGz+dZkL7XkPQnXQ6pWRXwjUHfXT2cWpAUos+2L2Yh\ntr2Xw8LjmM8QZ1+T1JtpY5VV3ap0H+36BIFmT3GIA2E2PCztjam/WK1r8NeUMalbvBai/qU1R+3N\na112k6jT85jbtzm1IC3ELucB1zWk9nCXT0MZ0r54oZpd1mMNB0vqzdbenGXVFAROsaK77rdS7wSN\noZuxTLosJK3vvL/q865nxHyP6+/UX9ram8O6dNbn2o7NqYUQhOiTRamszXn0aUgl9sOLwewynFKI\nUtoby/oGgWNg9xabep7SmpK0PjqZyqTLQrgeTum3Dft5cn2m7jcYAFJ7U5vPMO8c26x11V4wQhwz\nCByKGZVLC0vSptihv49OprCJHrJT+E5T6KJJX11yX8IwtTeV/mKwNj/fNN3B51mqnhMT6XsOTX/S\n5bAk80k2hRBHOLUQihC7zAd0vX9KUgwIGQB2N99FNE0beI6ZkXZlA6dY4R26SWtuDv2Fbm0abGtM\nfYNAE+l7Dk1/0uWwBAs529dFe0EJsWsQWH1m6CT5PlQVvORsCgPAfuYbBFbalajLCkknVffcTjEB\nOjTtTa2/kK3pmfCdv92mbfP/Y2iEqb04zGdRR6gxgVMLoQlx6ApU6YAwJIcz1BgA9rdQs24u7c61\nP+HQ59IMDrs6W2nNmSatA2n91flJlzZ9Mnq2RirNSd9viPqTLocYzWftgfQ19tFesEJsi7SHplgl\nAsXYsoYm3Aamv85c2MGM+bk+K4K7BHJNQ9B96rp6noZuezD2Kmgz6M4D0F1l0s+2lFUMqeu2zGGo\n/lVac6lrr6/VzVsNVWNdtSd6FnDbmYS7mjNVXVRn1w459zHLssPv9TlDdwj2GZbAyTMA+zDk80dH\nR4ey2O/3h3OYh5IneB5m3bnWU+J7lrPrPNWp8TnrtcLWsKnLLqzXa+x2uyC0B6R7HmvV3lT+ro+f\n3W63B11srLOFgXDP3U7R98WOra8x28I5cWovhp5Il97i2Bk2qSHlkOBJIOPoaGh2y84YtuHbU51a\n21Nlvns8l+K6q2xu/YViFeYwralpn2ekSbMhD8VJay517fUxW4+xZf7atBd0BrCOqufX1nP0zYL0\nobqGvtmImBi7x5Mn3gvummX2zWzXZUIqfJ6FMbOBVXbHNwNtv6/uPqqsYRNHR0cAgLNnz7qezSC0\nB4yrv0pTQ0cM5qBqb8wsXhvm+9p87moVTBWfInXfFyNmfNRFs6Hh0l50AaBJ5fiaGtKq4Zi64ro4\nqRiYKoCmE3RjaqjvdIa2QLAtmDcD1D6/7fOdgs9HENoDpvN9oTdSfQPANs2Eft8AfV+MmNPQYh3+\nBRq0t6RUdNtwkNTwwFhnuvoS2h5EptXpQMKky2Fqc+lsjoVIPrviCyGuu8qGlK0PofqACvva+kyR\niG0yvrTmhmovRZtiJ4yQtBd1BtBFUxYECKe3OFdmxFxgMnShzFBy9oJnpe1ZAM7pY4xFVGMMI9ct\niAJGWZgVhPaA/vrr46/HWCA3FtX1mz64y2K/6rOh3E8X6Pviw/afIU8xaMKlvWACwKbGw9Ug1GEG\ndrEEgi7mGFa251ZNXR50gjL4BIImXYLCrt89tebqvm+z2QSjPaCf/upWk5tTXNoCqbmmwzRhB4Bd\nd3oI2V+3EYr+UvN9Q2AAOCGmEKe6jqrxin3+iI2dPZxi25opMod0grLM+bx32e5lDkLRHtB/AVxb\nY+QzLxqQy6KZAWBdp7ausx+KfoYSiv5S9X19sDtdoW4x1IZTeyHMRZhjGwqf3wlxzkwf67O1QlfM\n74/xNAbpOgpFH1Mw9RzDviatuaH6M/E5VcZ3Xt2cfs+8fpsQNbNE/UmXQ2wm9azMob0ghGhinwU6\nRiNlV1oqgaDLpg4OmxYa5AE4wJxO8JSNERSG/txIa26o/oaUtU+9zrGowrz+OX83BJPW3BDtpWxD\nnrtQzKWFIIaAzWsYutTaHsptGj6IfY7gFJj3O8awsl2fOYdBosF3S5hYhuhC0R7QXX+2r+o7FOW7\nUKeL7zP3tWybw13dQ/W3o6Mj8XmJcxGK/uj7umHGJ7HGBC7tBREASk+0bAsEq7lMde9xLVCJUSRN\nDJlzaNYnnSCRIhTtAd31Zy+WGMNH+uyv17ZwpOsK3q7EuuK3jlD0R9/XDTM+iHUvQKf2QkhF28MT\nUmnWKYZEq2E1ifuZu+xcw8rm/c+tMZdJlxdtfpPW3BD9mfjM/+tivvs22tM6fPckHIsp55ZW8yWn\nHI7OA9BdTt/X2ZawF6BTC6EI0X7QJQurz6akvkjf24yCq73nKXTUx6TLhyaiSXHd9dGfHWhN6UN8\nOsFVAFp3Tqo9f3uKhUZjbnJd5+vHDrBD05/0cxibMQCcQYj2QyhdYMC0gWCeLzs7aGI61KGaGcuk\ny4cmoklx3fXRn0QD1JYVHHvUpi5w9PG9Q4LBpu9fsv6kn8PYjAHgDEK0C3mqXlgfy7LsxLBAnZMy\nnVDlPLsGj0sKCF0BvaTeXNqjpWHSmuurP+nOsStLNtc1+fpS36Fbn6zkkvUn/RzGaCYxttFOLYQk\nxLlS8XNb34DQFJzLpO/NZa5ek6TemrRHW75Ja66v/mxfIFmG1Tw8qWtqC96agsAuIzpL1p+kfmK1\nkJ7BMbUXnBCXGgSaNvX+e9L3V92jSXVdU2vK16TLhza/SWuuj/7s5yiE/fKkM5JA8xB13Xub/K3h\nmya9J2nNddUe7UTdndJLTObUQohCTCEIrGxodrAOyfJyOVsGgDRpk9ZcH/2FNv9IMvvnsqb2oilj\naPvJqctZWnNdtUer10YImh9Le8EKMaUg0LS6lXR9mTNb0HadbUKc26TrmTa/SWuuj/5MQvCBIWT/\nfK6rqUPt2upl6vuS1lxX7dHOWWgdsbG0F7QQUw0CuwqzKVCcsrfis3+Y7Wwl9dZFe7TlmbTmuuov\ntGxbaNfTdG0umq7ZZMn6k66rGG2pAWAQJ4E0UbfT/H6/P+wQT87hOuap2s1/TNpOEHDt4J9zN3wi\nRCjaA/z0N9bxb2MxxWkkY9J0KonPEV7m56e4t1D0R9/XHenTyobi1F4MPRFXOj+kRQ8h2ZR7F7bR\nVh95AD3gnL3gJE1ac131F9Jwa8jZvyErg13fs2T9SddXjGbrP4TFWGNoLxohtgU1ITmkEGzOY5q6\nHJ+UB+AAczrBJE1ac131ZyLt30IKRpvKqQ7fsmMASPPVmfTzOJb2zkMkrNfrxiGQ4+Nj5Hk+6cHk\nMVEdWr3dbif7je1221ovhJDutA1XzkmWZSf86n6/D+b66vy97fPqpsUQQoBoeyJty/tji9BTMene\nL3vB6Zq05rroL6RJ502jCV2y/1NZtY2WuUiw7pq7lPmS9SdZVzGbSWzxhVMLsQuxbQ89c5WstKOi\n0QnSqD0f/ZmEsPuBz7zi0Hysfc1t12YGjVPch7TmfLVHa9dTCM/kGNoLfhWwL64VsF1oGi4NZcgj\ndnKuhCNChKI9oF1/pl/2WcE6B9VQsLka0kUo19y1HKv3T3H9oeiPvq8fMa8EdmnvzNwXMhXVliN2\nJXWh6XNNrzFwJISMhe0zQplfa/rYtmDw+Pj48FoowSAhxGKpqeipztsdg2pIesr7D9XyAIZAcg6D\nJGnSmvPVX8grbuvM19dK+Lyuvz/ltUprzkd7NLeFNC93LO0tZgjYl6aeaN/M4RiYWcQl95ZzDoMQ\nIULRHtCsP9Mnx5Y922w2vTeJH5O6DCWHgAvo+/pjPpsx7X7h0l5yAWBfpALHpQWGdIJEilC0B7j1\nl2UZdrvd4f9jCwArmk7lqKhOKBrz/prmgvs02NV1V9tojUko+qPv60+snTMGgELYe2gB4wSMsQaG\ndIJEilC0B7j1F/NEcxOzXamCrqaA0DcYbHq9KfPoG9CZ5T922YeiP/q+/iwtAORcBEHLsuwwH3Cs\n+YrmtjfS91dnknqj9tI2ac356M9Gusz6mD1XytxSZe552V33KZxyL8A8AN3l9H1D6/BAqG1sF+1R\niAHakgNDSb1Re2mbtOZ89Gc/s9Jl1rOcD7j2S2s7v3cofTeoZgBI89XHFBqZylxaWMw2MEvCNUHa\nHE5um2htYw4r2ROjK2JJZxOyRELd/qULtk86e/Zs7fs2m43XdjJdiWlYjhBpOAdwAQwJDF1UgSFX\nwpGlEIr2gHr9LWH+35B7qJsvDRQ+7ezZs5OvHK6uoVqEM/Yqz1D0R9/Xn1ifUZf2GAAumLEDw6YN\nrys8tlkI4omh9tIjFO0B9fozffEUq1DnYOx7qL5vzvKofnPsbGIo+qPv68/SAkAOAS+YsYeSfYZp\nJPdSJCRWfIdOY2LoPZjBV+WzYhwWJ8vBDgDzPD+VGIlpCgIzgOTAFEPJNQTRZaL20iOUDAxwWn+x\nZhZszPZkjAyaRFaUGUDSRNeYyQ4Q55jKYMMhYDILrnk8FcfHx3SCRIRQGmDgtP7GDpwkmCqInTsI\nrDaDZgBIXJjt3NijXlN0/pza43J02pwmvQUCtZeuSWvOpT97a4lYt3+ZaouMubfemGorGGnN0ffN\nY+aWa322cuuzfVFf7XEOICGEBESM2T9CSEHb82uPktnTrY6Pj2cbImYASAghgphDSDEvcjg6Ojr8\nO+b7IGRKXHMAq6kHWZYd9smcmvMm/wVCCCG1LGn1r3kvMd8Hg1cigblYxOxMTQkDQEIIEcIOABl8\nyGPWAYfjyVyYmcGJduA4BQNAQggRwu7pMwAkJF2qwM/n0IUxYABICCFCmD39uZz+HDCQJaQbc2X9\nTBgAEkKIABIOfyrsodKlBIBLPNmoWmxAwsKsk7meHwaAhBAigN0Ic74ZmYMsy7Db7aQvgzTAAJAQ\nQhbMUrZ/IXHCLGBYVP5gTl/AAJAQQoSJeduUpbPUzCwDwDCZ0xcwACSEkJlZ2pw5M5u5pMUshMyB\nVCeDASAhhMwMt3/xZ65NcU1SqI8lLnBZAnMGgwwACSFkZiRW/MWKRFmZw3AMlMjUSHRyAAaAhBAi\nCuf/+cOyGpelzm+Mic1mIzYfkwEgIYQIEnsjPOVxdlJlw6wsmYPdbieaYWYASAghpDdznmc8V0Bo\n30PsQboLqaFHUr8h93q9nvUaGAASQogQS1wxO+ZwFgOU8TGDW24FI4Mr+Js788wAkBBChFjCUKOd\nHRvzqDEulhkfzqOUozqFJYTgDwCQ57mYAchpaZmk3qi9tE1ac5bleSHERViWZblNlmWDvnO32534\nvs1mM+s9bTabE7+/FP3ZdTV3uaZqdc/Ibreb5bddWmAGkBBCZma/388+32dK6u5nSCbQ/ux+v1/s\nPLy5YSZ1furOXw7CB0j2RBBAVE6b16R7v9ReuiatuRT0N0Ym0M78zZUlsW2pGcDyWpJms9kMzlC3\nPQebzeaUhirm1nTu0MIqLwQhwmq1kvtxIkKe5yvpawCovRQJRXvAsvVXl+2wF7vs9/tDJirLMmRZ\ndtgPzf6s2PwoAGb7uN1uB2UhQ9HfarXKN5sNN7hGocPtdjtIX5V+Ab9NwyUyfy7tMQAksxKSE5S+\nBjIvoWgPWL7+6gK5PkgGf8ByA8Cx6mdJ+NSv+XqfAFpKzwwASRCE5ASlr4HMSyjaA9LQ39AgQzr4\nA5YbAFb/TnFe5dHR0ezb31RZRsFMdr32XGPDcxgE5nXQZE1Sb9Re2iatuRT0Z8+ryrLs1Jw+H6ac\nn9XFxpwHKK25pWuvq05d8/OGsNvtDnP/pO/RR3vMAJJZyQPsBZM0CEV7wDL1Z67cXa3ai9o1nDY0\n0zYm9lw5n/tyEYr+lqi9IfSdD7nf7w97KoaiVxcu7TEAJLNCJ0ikCEV7wPL0Zw/3dg3i7G1fhgRa\nY8IAMB3MxRxNhB7s1eHUHlPRtDlNUm/UXtomrbkl688eTus6hDv2lisj6+bAkKE9ac0tVXu0/trj\nRtCEEEIGYQ+hSS/eIIS0wwCQEEJIb+xhM3vPP0JImDAAJIQQ0hs7AOyT/Ts6Ojr8O7QA0rwebp5M\nlgQDQEIIIb0ZY/h37n3ZCCEMAAkhhPSEw7+ExAsDQEIIIb0YY/g3dOxtP2LcBoSQOhgAEkII6QVX\n/xISLwwACSGE9MIM+Dj8S0hkcENK2pwmvQkqtZeuSWtuifqrzvodskFylmW5ifQ91dkYG1VLa25p\n2qMN194ZEEIIIT3Y7/eDh325ApgQGTgETAghhBCSGAwACSGEiGFmEENdRMKVwGSJcAiYEEKIGPv9\nHqvVSvoyCEkOZgAJIYSQFngkHFkaDAAJIYQQQhKDASAhhBBCSGIwACSEEEJa4EIQsjQYABJCCCGE\nJAYDQEIIIYSQxGAASAghhHhgrgQ+OjoSvBJChsMAkBBCCOkIj7AjscMAkBBCCCEkMVZ5nsv9+Gol\n9+NEhDzPg9jyn9pLj1C0B1B/MWO2mev12vv4ulD0R+2lh0t7zAASQgghPeAwMIkZBoCEEEIIIYnB\nAJAQQgjxhCuByVJgAEgIIYT0gEPAJGYYABJCCCGEJAZXAZNZ4Uo4IkUo2gOov9jpsxI4FP1Re+nB\nVcCEEELIyHAYmMQKA0BCCCGEkMRgAEgIIYR0gCuByRJgAEgIIYT0hEPAJFZEF4EQQgghhJD5YQaQ\nEEIIISQxGAASQgghhCQGA0BCCCGEkMRgAEgIIYQQkhgMAAkhhBBCEoMBICGEEEJIYjAAJIQQQghJ\nDAaAhBBCCCGJwQCQEEIIISQxGAASQgghhCQGA0BCCCGEkMRgAEgIIYQQkhgMAAkhhBBCEoMBICGE\nEEJIYjAAJIQQQghJDAaAhBCQUVl2AAAgAElEQVRCCCGJwQCQEEIIISQxGAASQgghhCQGA0BCCCGE\nkMRgAEgIIYQQkhgMAAkhhBBCEoMBICGEEEJIYjAAJIQQQghJDAaAhBBCCCGJwQCQEEIIISQxGAAS\nQgghhCQGA0BCCCGEkMRgAEgIIYQQkhgMAAkhhBBCEoMBICGEEEJIYjAAJIQQQghJDAaAhBBCCCGJ\nwQCQEEIIISQxGAASQgghhCQGA0BCCCGEkMRgAEgIIYQQkhgMAAkhhBBCEoMBYE+UUg9L4TfJ+Cil\nbqWU+oT0dXRBKfWdSqlXDvwO6jcxlFLvUErdxffvJD2UUplS6iql1BOUUo/wea/jtc9QSl1S/js6\nHysBA8AeKKWuB+AXlv6bhIwF9UsIaUJr/ZNa62cM+Io1gEvGup4UOCN9AZHyxwBuqpR6C4CHAfhx\nAKp87Qe11n8IAEqpSwE8GkU5vw/At2ut36mU2gC4FYALAXwxgHcDuLfW+gNKqYsB/Eb5+rUAfl5r\n/TzrN79Ba/0Pc9womQ6l1EMB/BCA81Fo6AUAHgvgW8q3/DmA79Naf1Qp9Q4AD9Zav6b87DsAPLh8\nzzMA3BXA9QD8NYDv1Fp/RCl1LwCPA3AjAFcBeKDW+uqWa8oB/CCAhwL4zwD+h+2UlVKfAeC5KDR6\nAwC/qrX+JeO6ngDguwHcEsDvaK0fDeo3GpRStwLwZyjq8WEAbg7gRwD8LoDLADwIwKcCeDGAH9Fa\nX6eUujWA56DQzIcBPFxr/Qbrey8tvy+r+Xudn1wBeBKAe6Pwhb+htf6F8u+11zFqQZBZUUo9B8BV\nWuvHKaXuAeBZAP4VwJMB/CKALzHe+9Mo/N+nALgUwDUAngrgjFLqxgD+e/m+Ez5Wa325Uuo8CPvZ\nUGAGsB8PBXCd1vr2ADYA3qi1vh2AbwTwfKXUBUqpi1AI8u5a69uiEMZlxnfcB8CjtNafA+DtAH6y\n/PuvA9hrrRWAewJ4SumQD7/JxnMRnAfgU7TWXwLgh1E4kPsD+AYAdwTwhQBuVr7WxD0AfC6A2wO4\nLYC/BfBVZYP82wC+TWt9awA7FA7Mh9tqre+Awtn9slLqAuv1nwHwD6X+/yuAJyilbmm8fjcAX1Xe\nx6OUUrcA9RsbFwL4pNb6i1E0oI9D0RDeH8CdAXxeaY8s3//rAC7XWt8GwONRaO9AOdz7EwAu0Vp/\nzPh7k598UPlbtwNwJxRaunPLdZDIKUcLngvge7TWn4/Cr93IeMstAPxN+drTAfxM2dl4KoArtNbf\nWr6vzscC4fhZcRgADkApdSMUaecnA4DW+ioArwZwT631BwDcRGv97vLtrwZwa+PjO6Mh/H0AX62U\nuj6AuwN4Wvl970QhqK+Z+l7I7KwAPK/891+icGr3BPBcrfVHy2zGbwH4upbv+SCAL0CRJbmh1voy\nrfUrAHw9io7Em8r3PQPAJaVzbePZAKC11gA0iobW5AcAPKp8z9sB/BMK51jxO1rr67TW7wXwfhSZ\nQBIXZ1DoDwDeAOBiAN8E4Nla62u01p9AkaG5j1LqU1H4wcvL918J4CuM77oYRXbwPlrr95s/0uIn\nvxFFg36t1vojAD4fwOtc1zHObZMAuB2AG1QjaQB+FSdjlY9orV9S/rvynXXU+VggHD8rDoeAh3FT\nFCL7U6WqEWDcGMCflAJ4TDkp9XoAPg3AW43Pfsj494dRpKgvALDSWl9jvXbRNJdPBLlOa/1v1b9R\naOTTUdR3RWvda61fq5R6FIqA7LlKqZcC+F4Uvdq7lUOuFdeg0NgHWq6tTpsmX44i63dxee2fhZMO\n2tRvdW8kLq7TWn+0+jeKOrwZgB9VSn1P+fczKBrGm6Oo/2sAQGudoxi6q/hlAJ9Eje5a/OSFAP65\nem91PUop13WQZXA+TvrB91qvf8T4d5N/qfOxQDh+VhwGgMP4AAph3UlrbTo8KKUeiGJC6t201leX\nKyAfZLzlQuPfN0fR6F4N4JNKqfO11pVAL0CRRSHL5/0o6rvCrHvb0R2CMq31FQCuUErdHEX27scA\nvA3AK7XW9+1xHRcCeGf570qbn2m8/nwUWe9naK1zpdR7evwGiY/3AniJ1vqp5h+VUjcAkKPQ69Xl\nHL3PA/D35VuqOdLPBPDN1nc+AG4/eTUMP1nOPf2Y6zrIYvgIikRKxWe63tiTUPysOBwC7se1KMru\nPwF4OYBHAIBS6oZKqWeX86EuAvCO0qldgGLegSnquxjzpu4L4NXlcMYrADy8/L7PQzGf6pXVbyql\nPm3yuyNSvAzAg0sdnUGxkOLl5WvvA/ClAKCUegCKye9QSn2XUuoyANBafwjAW1A0xq8AcNdyjgqU\nUndWSv2K53V8W/mZav7NX1ivXwTg9WXw9xAU83NujGao3/i5EsC3K6VuCABKqYcrpR6itf44gD8C\n8J3l++4B4A/KTCBQzOs7BnCbUi8mTX7yJQC+TSl1g3K6zWsAfJHrOia4XyLD2wBcXymVlf//CBQ+\nrY1rUWTk2gjFz4rDALAf70PhjP4RwBMBHJUp4DcAeLvW+l0o5sNcUO5ZdDmKifO3VEo9qfyOPwbw\na0qpd6GYI/PE8u+PAJCV3/ciAJeW33f4TaXUV89xk2R2rgDwBwBeD+BNAN4F4Cnla48F8CNKqTeh\nmAv15vLvVwK4o1LqbUqpv0MxT+WXtNbvQ7Hi8kXl35+KYpWxDx9QSr0RwKsA/ICRja64rPzev0bR\nWD8TwG+UHRYX1G/8vBjASwG8ofRPl6BoAIFiJeY3KaXejmKy/QPND2qt/wPAQwD8YrkoqKLJT76g\n/P63oZjD9Zta6z9tuQ4SOWWH4pEAnlP6obeimELQFgT+EYCvUUq9ruV9ofhZcVZ57hNYkzFRxTYw\nt9BaXyp9LYSYqGIbmFsak/IJIUSMMvv7rwBuZs2PJwNhBpAQQgghwaCUel05BAsU80T/jsHf+HAR\nCCEJoZT6SRRDcXX87JzXQgghDn4YxRSpx6JYFMI5nhPAIWBCCCGEkMTgEDAhhBBCSGIwACSEEEII\nSY08z8UMxbLuZCzLsryNzWYjfp1TmqTeQtfeZrNp1UdfzbR999J1F5L2QtVfitb3mYtZf77X29Ze\npeAzlmJOLcQgxCWYT/BXkWWZ+PXOLcTUtbfb7bz1Ucdut3PqxqeRW7LmQtNeiPqz9dKkp6VYnU82\n73mz2Rys7nmLVX999OCiTznQ5jenFmISYqzWJfhb+kMlqbdQtefbwJgNkitgrOuVp6y3ELUXmv5M\nM3W1dF3Yz1BbRst+f9cAWVpzQ7XnCgSXrpMlmFMLMQoxNuuT3ZG+5rmFmLL2bHwbliaHbH5HEyk5\nb2nNhao/l56WPMRn36vPc1DXkY9Rf0PLLsuyU22adH3S+mmP+wC2sNlsWt9zfHxc+/ftdgsAyLKs\n1+/6/DaJn/1+f0Ij1b/3+33j5zabDfb7PY6Pj099Psuyg/5crNfr1t8gyyHLsoOv2m63J+refK1i\nydo4Ojo68f9tzwpQlMd2uz1RTlmWLbqc6qju1/Y5qZXDIoi5JzKm1fVq5qKuZ7nUzEweQA84D0x7\nvpm8Pt/hYqn6ikF7c+vP5dtMbXUdDo3dzOely1Cu7au7lJO05sbWnl0WS58zGrM5tbAEIQ61rnP0\n5mKJTjgPwAHmAWmvsqbOh28g2KcTs0SNha69OfXXpoc6zaTSOegTsNgdrRj1N3YZ2nqSrleav/ZE\nTwJZrVZyP16SZRl2u92g73ANH7iGhof+TsxDw3mer6SvAQhDezbVMJxrykA1BOUzNFynPXuo2WS7\n3UatKx9C0R4wj/52u12v6SerVTDFFBx2e9mlrELR39jas9tQ6ic8nNpbSk+kj9Vl/syVlkOyI3UT\nqttWcHZht9vlm80muh6XpN5C0l6bLtsygn0+m2VZ0vsBSmtuTv3V1XNVt00jHrH5EwEN9X5WpDU3\npfZMUskgx2ROLSxNiL7WtgfUCAV+AvO1vpuPNhFLwz2WdoaadDn4arQpmLM11TaVwd7jbAl6ilF7\nU+vPx7fVwYa72exnJrVtYJos1WkEsZhTC0sTYocCmVSwddsM+AR+9nV0yRrGsHmrpN5C0V5fDZlU\n7+k6f7XLb4SupVi1N7X+bHz2hmSD3W5Dtz6R1tzU2mMQGK45tbBEIbbZXELtgu9wbuzDeJJ6C0F7\nPta2877PHn+uDkeTrpbuwKU1N4f+fOvQfF/I/iIkG/psSGtuau3ZZURthWNOLSxViC6bs5HzOUtx\naJZljC1EQhBiCtrrW5emXnzf79rio+vzsSQHLq25OfRn46s16boJ3YZs/xKa/uYspzxf3khCjObU\nwlKFWGd9dn8fanM8EE1zxUJrwCX1Jqm9rhrxqcO6Iak6bZm6D0GvqWtvKv11maPGANDfhp4AEpr+\n5i6v0NqgFM2phSUL0TbJ42vmWLEbw1mN0s5PSnu+muziOH3f2yUArNNRSPpZgvam0l+XoI4BoJ/V\nPZ99nwdpzU2pvbZyk67H1M2phaUL0SqEA0vtlTRlA6WvrUmIS9denbkCdh9t9s322H+r5gran1ni\nULC05qbWX5egzmQpAf6Y5srKDykrac1Nqb02LS5lFKGvlqTv36mFpQvRrASTJTRoTda0D1iIQlyy\n9lzWJciqAvtqbmffbE/1XXWYwWCdfqSd2FK0N4X+6vYdbdKS73tTtTqGlpO05qbSHjXmtlAy7U4t\nLF2IoVXEnBbifC5p5yehvYayaHWQrtW5Lj1XWwa5VgrX/c0Xae0sRXv5iPpzBfNNDS6H55qtbgTF\n1H7VAeNG0K33eyC1LLPruZTyoU4tpCBEIF2nF1oqXtr5SWivoSwO+AzD2u+39Wz+rXK4Lt37bCdk\nI11eS9FePoL+2k6LqdNTXX2n1jC3lWld+fhuveUxZ1dcd2Noz9dSTbo0PZdS1+XSQjJnAZv3mcK5\npyYh3Xu+0PMwu9J0Xu/Zs2dbz5Hebrcn3rNer0+dI7xer0+dc113TmeWZYdziF2/td/vW88gDp1Q\ntAcM19/O45zfqq6bzpher9fR1+sYuM6Ebzo/u4m6ZyYU/c3l+2wf52p7Kv8zxu9J4vLpQMDtLnsi\nyzcT6bkYknqT0F4XXbqosgu2hrucAsJMTzjaG6K/prmZ1EJ/8z2lycz0VZnBtlOajPoR110+UHtd\nR5CaRiBCpM8Imav+pUfbTMtdWnC9MIfNdfM+E1JDqqwJKr/x3kMQ4lK152tVw2KuzjUbalcnxteR\nLlnfsWmvr/7a5vS2Eerm8CFY03Pk6zNdAWH1HEtrboj26srER0t101KA5uktknRtH+vuQ7qNddRf\nrRaSGQK2h0zMoTAzdbvf77Fer+e6rFkw6zjYVPTMSA8Bd8UeXjD16xq+As4NKUsPj4RAKNoDuuuv\nro7t4du6IbeKJQzhT4Xr+RniK+32phxKDkJ/Y2jPpK2cNpsNjo6ODsPiwMk2aaw2t09dHR0dnZo2\n4/uc1A35hjqlwun7XJHhHAbBHp4ZpbuieNtcmEMDIfawJYbAXb0gSb1JaW8sc+mXWy54l5+47vrq\nz+cIy9SnuvQx18KPKb5XWnN9tVenv6FlZiLpr/o+MyHusNFS3vVaiE2IQ8x1JupU8xHM4FGy8udq\nGHz2p+urlbFN+oHsY7a25q7f2E1ac331Z/snV2NLHXSzqRvxmu8X111X7dn305YI8f0uE8n28VQF\nBaCbie6zXgsxCnGImZvp2pU6x7yEKls4p2Cmbhi6nEWcB+AAcyHtjfAQ15arrS/p6wzVpDXXV3++\nzy8DwG5m+6wpfLL1G+K666o9l7mCQV//YxJKAOhzHTEGf+V91mrhDBKibl6G+e/9fn+Yr2AvS6/m\nUtXRtmWHSd2WG+ZcnZjmajVtL1ER0/3EiF32Lo2SeDF9RYjzi5bAVHO31ut1FXQtCrO9NOcIVv9e\n2jz6Crutj35ubew9kTarhmDnONvU3CF+aDZxzHmFY2cGumxfYn/WRxdz2BzaG9vqyjbVDc57lp+4\n7rrqr8tG7swA+pudyZkyi2NWSQg2xT36zFF1lIlYBrDrIQm+UzFCNKcWlibEtgUbEhXns5DElz5z\nCscKErpefxchzm3SD2TPh/hA3XFwMTkkofIT111X/ZnPblv9MgD0tzkbc7NKQrAZ7rO1TE1CCQC7\n3GNsvtaphdiF6LsZp43kuP1YmULffb3szwx9UGyqLKVP+eYBOMB8Qic4pfluOEurN2nNddVfn9Xd\nXd+fss0xKmTVibjufLXXx+rmxzXsBiGu076JkRj9rFMLsQmxb8BnE2IEPyRT6HqI7O8aY6PLuu/y\nzVTkATjAfEInOKWNsVltyiatua7669NAURP+Vvc8TaS7w9eHYHOXaV3AFIJOQ7iGGe+1XgsxCLFP\nwNf2/lgqvGu2sBoarD4/ZPjX5/urazSpe+Crz+QBOMB8Yic4tR66zrehHfQsrrsu+jPx9Vd9PpOy\nTZ0FtDrg4rqbw/f5BIEmUv6rT72POZ2ry++OcK/1WhhDUH2tj5CasIdE6wo6y7IoU7h2ufgIcMgc\nsS7L3V2/4bhOcQeYz+AEp7aqQ8BG3t+kNddFf10nqBv3eIDa8NZFr7L2sRQDQKA9uyo9X9Xn+Rpr\ntLGNOeKR3KUF1wtzWJcKqqNplWxfBxqbdRGob4NQN0HaVX515dyyp6K4A8xncoJT1PNSdTyHSWuu\ni/76NpAmDAD9bMrFVKkGgEDzyNDQqUlj13nde+ZijvvNHVoIch9Ae7++imq/vD57y0W9V08DVVnU\n7S9oc3x87LVv0c4497HtnEb799r2BSTd2VhnTi5Vy2QY9nPHPTj9sPd+pf8ah6rtOD4+xtmzZxvP\nrZ6bo6Ojxtd9NNC0N7Av4s+oKzKcw9DSY6jo0yMb+vmYre1Uk6bPDtl2whPxHnA+Yy94DEtZyyOX\no7jufPXXJwMoPawWs9ltz1jfm3IGsM0k9zA18VlAac79ky63nvdbr4VQhOgK/voMefXZPmFp1hSc\nucpj6KpDT8QdYB6QE+xajxwC7m/SmuuiPwaA85vp/8ZqM5Z6FNwYJqlXkxTiA6cWQhCiK1vVt7Fj\no1m/M3vTwzZkSwR7VVTdvoDGNYg7QFN7oZtdftLXE7NJa66L/vo0jlMEMCnZ2KeD1PhAcd35+r62\nUaQ+tO0eMZdmU4wPnFqYUmhtBtQHf0MbOsnUcihml2dTmYx9wLUr+GsSooT2QrcUHdWUJq05l/58\nplJ03fCdAWA/G3NxQk3bJq47W3sdrn0UmraDkQoApTU3hzm1MERIQ63O8Y0hAglRhWZm2bb1tMbc\nC6sp+GsS4twmXT9ddczs3yjlKa47W3++Da1P/Zuk6vdG0smpsu/T+bK/Q1pztvaabKx97mzs35FI\n1qSYIHJpQXQVsL0KaL1eD17hyBVcBZvN5rDi115pZJaxveK67v1dsOt0u93Kr3SKELvMqhXwZDm4\ndjuoI8sy7Ha7xhX5Jnzm+rNarapACcC5HRa6+LK65zemtqlqP8a85rqyO3v27InfyLJslF0Oqu+s\n+y67vUuaPj2IEe3A0KyTa3fuvt+5JHP1eKbY/8rOPNqvC+utUy9Y0sauF1o42qv0Z2OuNHSd/tNy\nzOIB6bKO3VwZMN9sYF2dSWvO1F4oNsU8QLPu6jbLH/v3YjCnFoTFeMBXLF02PmbDeaj8WsHbjDXH\nrOl7hPUWpBOsK78p6iV1k9acaV3md/oGgSbSZR2bVW2LWQ9NCyGaAgdX3UprrrIQytql2ympqxsG\ngLJi9K6EPpNSU6ncJnP1sKQWGAjrLRgn2GRd9mKkxae9PD8dAHbRhEsX9jNO87OmtqVpLpydDawb\nheL8Z3dZm+U2F13n/1XXFvszlbu04HphJsvzvD346HomcEXslTaGuRoakzmDDGG9BeEEm4zZv+Vr\nL89z9NmypSkItIe9pMs6FhtjtWvTd1hn04vrLhf0fXY5VWXTt30fSttzUnddscYUuUMLq7wQhBQ5\nUEy6baJuoYJ5BIs5udS8Hy5AOFke1dE89jE8Yyy+6XA9zZU9E6vVSlT4Lkyttx3DR7oRivZKDvrr\n4qdcvtBefEXttNNlEU4f7DoIRX8Svq9Ot7Y+fZ8B89i+Ctciuc1m4zx2zqfdc8VHc7aZY+DUnisy\nnMnyvChh7yi8LQJfQrQ+lrnKrksvaGwT1pt4L7hPfdGWpb3c8H196plzoIdb096n5iKcPtQtPAhJ\nf3OXdZeFTFN+X5+pNU3Zyb7bA0mYUwvCYszzQpHele1xowdSb0Drhn+lNxcW1puYE/SxFPenSlF7\nueH7+vqpuiCwbsN3BoGnbcjG966FIT7nxAagudl9X10QPbTNGaLvLMt6/X6XuaAhmlMLwmI8PDyu\nC3fNG/ARBwPAk6K1y0eicRDWm4gT9DFm/9LRXm74vuq57GOu+b0MAt021qlHfco3AM3N6vumCP6k\n/WSXrHCVCbZNQvdOLQiLsbUS206WaHowU29E65a8m0j0WoT1NrsT9DVm/9LRXm74vjHqui6rwSDQ\nqYFRfGAfXxqA5ib1fW3btI3R3oRwjNsY5yQL6L5eC1IizA0n2JZC7RIE2u+pLPQU7RTWlraWuCZh\nvU3uBPuYdK82FZPWnGUH5tCU5DMfio0dFHf9rgA0N7rvq4K+NsZqf0MIAM177xsIhjL1SlqM3oXi\n69BckzZTdH5ND6ZUQCyst0mc4FBj9i8t7eWW75vqfhkEnrMpMqJds4ABaG4U39d1gcyYbY2NtK6A\nkyeRNZ3iUyHRwXdqQVKIdQU0NAgMqdClzYVkIyCpN9Ok68al6RR1mpr2Spulvsea8xazTTkc3mVl\naQCa6+X7+qyItk9VGcvs35DWVizm1IK0EMcIAvP83IRL12upOb2y0p0PZ2hClNBeCBbSkMbSTVpz\nls32LKYcBE59PnyXDlwAmvPyfV2PXK3uew4t+5Y1zU97QQixaxDo+kwT0hUgVOm1MAA8pz1pM0l1\neC417ZU267OYahA4xz37TuEIQHO1vq9PwJfn7v0Op7JUz/Edw5xaCEWIfearNAmWYjnt/EIoC0m9\nubQXSh2lqNEUtVdanufzBmGpBYF12b8pRoN8s4ABaO7g+/oM61YBn5RmOFrS35xakBZi04NUia7t\n4TN7L+b7UxeMC0mnL6m3Ju1JGDspaWqvNBGflFIQ6EoQTJFp98kCBqC5E9rzQTLgsy319nyIubQQ\nXCPsCgL7itAmFDHPVOknynDulH0XIYagvbmNDi1N7eVGIyxRDnU+VrpuJqrvWqbygSZ1QWYAmjuh\nPVfZSLcRvvUpfT0xmUsLQTbCTefvmSL1Ceam2I08FrPLTPp6moQYivbmNG7/kqb2cqMRlvJFoaw+\nt7fQmGrLkDmGL9u2hQlAcye0Z5aLRN0Pqc9YrjkUc2kh2Ea4yyaL5pEr1Wfb9uJJYcJ9iA+MpN58\ntZdy/SzZpDVnmXi9S3ZA+s4/s80V0Ell15ue6QA0hzzPEdKw7lhlS2stu/p2UFKIPhc+9MiVJpYe\nBJqE8sDMrTGXSZdDKBmYlExac5aJ+yDfjYzNzrjd2e5qU/pz+/pMZtTYgVADQOnncEqt0py6rNXC\nqhSECKvVqvePZ1mGLMsAAMfHx8737fd7nD179vBvANjtdqfet16vD68vgc1mc6JcVquV4NWcI8/z\nIC5kiPbGYLfbHfQLhFM/SyYU7ZUc9Cfpe2w/YV5LlmW1vrKJ7XZ74v+Pjo5O6Nyk8s2bzebwN9Ov\n+36PD/v9Huv1uvfnfTHbU7teQ9GftO/rQ6jtWSw4tbeknkg19Ovzvrqe6JKyMKEuMJDu/YbSCzZZ\neiY6FJPWnGXB+B2XFqdizCFIew5hE1M/Z20+N5fXXBC+b4qypTWbUwspC7EuCFxKYxzqAyPt/ELR\nnol0AJCKSWvOsmCezzpf0XZ8WttCvTrm8q1VUDj3NTAAnFejNH9zaiF1Ic61WahAhc/i9MYSYkra\n43yWtLWX58VEfBPJcmkL5lz+w3chh9Sig6ZjQ6f4PQaA01lIz0uM5tQCheh2FLE2zFMefj6VEFPS\nHp1Z2trL89MBoHQWuGlxhnS9DazzWsYub7sNCXkfQOk66WP0mcPMpYXzQA4ThO2J2Lvd7sQE5Rio\nm0Q9x+Rn4o85mXlJC4+IP6H5lWqhnE3svsNelFJxfHw8aFFJ3feZuMqTkKBgT+SkxTwvMIZjnqR7\nv321Z082H7IVhol05iclk9acrT8bybJZ8ob5TYxxj74jLtKaM7UXmzEDOMycWqAQ28VWPdQhO0TX\nNUtfl68QQ9Velw3JXZgBo/1dDADT016lv1AatRg6jmPf31j32uVYPWnNmdqLzUJ5VmI1pxYoxHqL\naV6gK2sZ4rVKOz9f7Y0R+PkydHNdWlzay0v9hbIZ+NK3xKora5sBmjpBk8+V1pypvdiMAeAwc2qB\nQnRb6PsFzr3KbUohhqS9PltcjM0UZ6OmbnkAuqvMuKYTdT53mbg6OaH4uDnuNc/7+cyui+2kNWdr\nLyZjADjMXFqI9iSQObFPbQDm21nehb0zekXoJ5rkEeyG76rv7XZ7qmx9JvM3nVTji3miTWgLCGIh\nFO0B5/QneSJMnc4rttvtInU21j3b3+PTHoSiv1jaXROeBDIMp/bYE/GzkIZZQ7qWribd+/XR3lTl\nWrdVhO9eanXEUN8hmbTm6vQnldmYcl5c6NY2taNpSobrefUpL2nN2dqLyZgBHGZOLVCI/tY05Dq1\nw6xbQGD+vnTZDBViSNozGXMorG0DaNdCkSZiqntpk9Zcnf6k5gG2aUy6rqa2ptNCuuLr+6fQUR+T\nLvs+xgBwmDm1QCF2s6ae89iBoE8wEFtPXVJvPtqz63fs+jTx0ZrPGad5vsw5W0vVnq0/m5nKwkls\nPmWIDQkCu/p7Ka3ZJl3mY9ST9PXEZk4tUIj9rG1Ccd8J/L4901iGfH2FGIr27Hod+d5P1F+f72ha\nnRyrJlLTnq0/+3mfOr2m1VsAACAASURBVKvb1IlNsSNhl30bfZ8zac3VaS8WYwA4zJxaoBD7W9et\nQsy5Jaaj7TIcEfvKUEm9tWmvrg5GvvcT9ThUe00aka7nEE1ac036s/3IlM/4EqaSTFke5mv2BvBD\n6kVacy7txWAMAIeZUwsU4nCbes+42IM+HyGGoL06xir3tvl/fW2pZ7guWXt1+ptyM+YqiBl7G5Sl\n2FTPZqj6ky7vMepI+npiM6cWKMTxbMxAsMoWSt/TXEKU1p4rozZWHUzpwFy6W0qnYWnaq9NfnUaG\n6K/rqIJ03UjbHOUhrbkm7YVuDACn0R6FOKHVnR/bxFKDPh8hSmuvqW7GCKRspijbKecvLsGkNdek\nP1cd5rn/nLOuK8jznMFfVW5zlIm05tq0F7IxAJxGexSioNXNB1y6STs/l/ZMpthn0WSq+qaTjEN7\nbb7P93SOpjOmXcwV7MRkcz030prz0V6oRt82jfYoRFoQQpTUXt1ebGOutJ1rrzcb6boOzaQ118X3\njXUcob0jgf290nUSgjEADN8YAE6jPQqRFoQQJbXnci4+W/343POMDcwBZnbC1V4X39dnTnGTNtmQ\n1uriRNktXX/S5d3HqNtptHceCCG1rNdr59nKWZbh+Pj4xMO02Wyw2WxOnTV6dHQ0+bW6zjclcVNp\n0Pe9q9UK6/XaeaZtn7OsU6I6b5uQFDgjfQGEhMx+v8d+vz8EfE2BVnVYuXloefXZiu12O8l1mr9Z\n/S5ZBvv9HqvVClmWHbRU6bLPd5nM0TkJGXacSNIwFU2b06SHP+q0Zw6ztQ2dTrHVz5B9Hl1zxaTr\nOUST1pxpkuXA4bRzNucG3NKaC0F7fY2anUZ7FCItCCFKas+k69w53y1+uuKzQpwnOsSpvVzY9821\nKCl0q3t+UtCfdLn3MQaA02iPQ8CEDMCcQ1X92xyqOzo66jXMZA7p2kPKZ8+ePTXka8J5TKQJDgMD\nu93u1HPpO9eSkKXAAJCQkTHnZ5U97sPfzeCsT3BoBpdNv09IE9vt9tCJSG0enCv443NDUoMBICET\nYi4COXv2bOOqS/O1pgyf/f3AyUacDRlpY7/fn9BYtYJ9yWRZht1ud+rvDP5IqjAAJGRCqlW/bcEf\nUD+cDNQPKVfZxM1mU9uoEdJEasPAdcHffr/nsC9JG05Gpc1p0hOgbe0t4WQEk1Qn9MekvVB8XyoT\n611HO6aqP+n6oFbnN5cWuBE0SZrY5z/Ffv1EDjsLuLRM8mazQZ7np54RZv4IKWAASEjE1DVuhPhg\nbyadZdki5gFWw71182i32y2Dvwgx63KqzfRThAEgIQuCASDpgt2Ytp12EzrVnNi6e9hut4sIcAkZ\nCwaAhESMvUcgIV2oGw51BVAhk2UZ8jyvzfpVR+kx+CPkJAwACVkI3ACa9KHuXGHfbYikqYZ76+Yv\nVsEth3wJqYcBIEkac/uL2OaWMKNBxsLeC8+1Z15ItA33cn8/QpphAEiSJrahriYYEJIh2JmyKggM\n7RnhcC8h48AAkJBIWfrmvWR+XEFgCIEgh3sJGRcGgIREitkgxzZ8TcLEtUeeRCBYHU9XbVrL4V5C\nxoVHwZFk4TARIaephlHrAq7qWML9fo/tdjtq0FU9j13OwR77GghJCQaAhJTEFBDa18pGkIzNer1G\nlmW1ewOagWCfYdfq89XZ1n2ujZonZBir8mxAmR9freR+nIiQ5/lK+hqAQnum9mPbJNZ+blerIIo1\naELRHhCv73MFhGOy3+8PWxrF9Ey2EYr+YtRezL46BFzaYwaQkMjh/D8yF9WegWMGglXAV7cfISFk\nOhgAkiSJeQhVejUmIWYg6NKjay5f1WFhwEeILAwACUHcASCHQ4gUTUEcdUlI2HAbGJIksRx1VQf3\n/yOEEDIUBoAkeWKbQ2dmAGPKXBJCCAkHBoCEREy1WpIQQgjpAgNAkjwxzVWKefEKIYSQcGAASEjE\nMAAkhBDSBwaAJGliC6C4AIQQQsgYMAAkSRPbHDpzAUhsi1cIIYSEAwNAQiKBG0ATQggZCwaAJGli\nGlK1A8DYhq8JIYSEAwNAkjRZlmG320WZXWMASAghpC8MAEnyxBIEmqeXMPgjhBAyBAaAJEnW6/Wp\nIGq320WzJ2Bsi1cIIYSExRnpCyBEguoQezvzd3x8jKOjI6zXa7mLK8my7HBtMc1VJIQQEj4MAEnS\nrNdrbDabE8OrWZYhz/PDNitTZwXtQM9nKDqWTCUhhJAwWeV5Lvfjq5XcjxMR8jxfSV8DcFp71TzA\nJsYICM3PmkFnF/b7fRAZytgIRXsAfV+KhKK/GLVnxinb7ZYd4I64tMcAkMxKyE4wyzIcHx8Huxhk\nrozkUglFewB9X4qEor8YtccAcBgu7XEImJCSal7gZrPxHoqdAvOEDzo6QgghUyCaASSEEEIIIfPD\nbWAIIYQQQhKDASAhhBBCSGIwACSEEEIISQwGgIQQQgghicEAkBBCCCEkMRgAEkIIIYQkBgNAQggh\nhJDEYABICCGEEJIYDAAJIYQQQhKDASAhhBBCSGIwACSEEEIISQwGgIQQQgghicEAkBBCCCEkMRgA\nEkIIIYQkBgNAQgghhJDEYABICCGEEJIYDAAJIYQQQhKDASAhhBBCSGIwACSEEEIISQwGgIQQQggh\nicEAkBBCCCEkMRgAEkIIIYQkBgNAQgghhJDEYABICCGEEJIYDAAJIYQQQhKDASAhhBBCSGIwACSE\nEEIISQwGgIQQQgghicEAkBBCCCEkMRgAEkIIIYQkBgNAQgghhJDEYABICCGEEJIYDAAJIYQQQhKD\nASAhhBBCSGIwACSEEEIISQwGgIQQQgghicEAkBBCCCEkMRgAEkIIIYQkBgNAQgiUUhul1LOkr4PE\nj1IqU0pd1fEzz1FK/Uz571wpdYtpro6khFLqYdLXEDIMAC1CawiVUt+plHplz8/eSin1ibGviRBC\nCAkZpdT1APyC9HWEzBnpCyAkRZRStwLwZwCeDOC7AawAfAeAywDcAcArtNYPVUrdC8DjANwIwFUA\nHqi1vlopdXMAVwC4DYC/AHANgHdrrTdKqa8C8NTyM58E8ANa61cqpTIATwHwxwD+G4BPAfBtWus/\nLy/rBkqpywF8JYD3A/gWrfV7lFIXA/gNALcCcC2An9daP8+4hycAeBiAmwP4Ea31C5RSq/JeHgTg\nUwG8uHztunFLkkiilDoD4BkA7grgegD+GsBzjNevD+CPALwMwOsBPEtrfZvytcz8f0L64tDhzQDc\nVCn1FgDfgMLf/SaACwBcH8BlWuvLy8/nAH4KwHcC+IJU/FTSAaBS6lIAj0ZRDu8D8O3W6wqeggHw\n9wCeBOC7AHw2gEcC+K8Avh7ABwF8g9b6wy3X8xMAHg7gEygc5qOt12+OQuRfCuA6AM/VWj+xfO3r\ny9+/PoC3oggm7O9/PoAPa60f1VY2ZBYuBPBPWmullLoCwAsA3BFADuA9SqnHA/htAF+ttX6TUuon\nUdT/fVFo74Na669RSt0RwKtwrrf76wAer7X+X0qp7yg/UzWyX1C+9uhS/08H8P+Ur30tgK/UWr9T\nKfVSAA8F8Njy+/Za6ycopT4HwF8ppV5l3MMntdZfrJS6H4CfLe/jwQDuD+DOAD6KIgB8JIrAlCyH\newD4XAC3L///MQA+brz+FABv1Vo/qQz4CJmCOh2+G8DdtNa3B4DSp71Ma/1zSqm7Afh/lVJXaK2v\nLT+z0lqruS9ckmSHgJVSF6FojO6utb4tiuzKZdbbfhGFYD4fRWP4m2WPtmKltVZGb+GLtNZfhqLR\n/G0AL0TR8J4H4D4t13MXAJeiCO6+CMBdUDT0Jj+LIoBT5evfq5S6i1LqRgD+J4AHaK1vV97LY63v\n/wkA5wP4oabrILNyBoVGAOBvALxOa3211vr/oOiQPBhF4PWm8j3PAHBJObRxVwCXA4DW+vUosoAV\ndwDwu+W/Xw3g1sZr/2q89nsA7qCUumH1Xq31O8t/vxHALUq93x3A08rfeieAHYCvMe7ht8p/vwHA\nxeW/vwnAs7XW12itPwHgWWh5BkiUfBBFp+LeAG6otb4MZQColHokCv/3fXKXRxKhToevsN5zL5zr\nJL8GxcjEZxmvv2zqiwyNZANArfUHANxEa/3u8k92Qwl0F8yLy//+DYCPaa33WuscwN8C+M8tl/SN\nAF6utf4XrfV/AMgA/L71nnviXEP8ofL1rwPwXwC8ywgUfhzAD1cfUkrdE8C3AvjWVFLbkXCd1vpj\n1b9RBGcw/v9aAHdTSr2lHMb4MxRDvRegCOY/ZLz/Pca/HwTgtUopjWK4d2W89uFSkwDwz+V/b1b+\n9yPW71+v/K2V1voa8zsAXGTcw0etz1Tf+aPGtf8igP9UXwwkVrTWrwXwqNL+SSn1Oyjq/jMB/ByK\nDDfnIZNJadChyT0AvEop9VYAb0bhF80Y6ENIjGSHgMssymOUUpegaLQ+DcXQqck9APyMUurTUcyl\nahPMv5T/rWvMr4dmLgTw3up/tNb/Vl6n+Z5PR9H4VnwYRWB5Ic415igDyOqz56EYxtbWNZHweS+A\nV2qt7UwwlFIfAXBj40+fBeDvlVKfjWK+3ldord+olLotTur6AuPf55f/bXJ8VwP4pFLqfGMKwwUo\n5gi2XftLtNYc8l04WusrAFxRTlF5NoAfA/DvAL4MwJ8ope6ttX4RTvvB8099GSE9cegQwGEu6gsB\n3F9r/QdKqRsA+Fj9N6VDshlAAA8AcAmKOQIKwLH5oiGYx5fDql+KYm7WVFyNIpCrfv8CpdQF1nve\nj5MNeNUQ25+9obWNwl1QOF4O/8bFxwHcVSl1awBQSt1ZKfUr5WuvBXC/8u93QDHXDig6CR8F8JZy\nYvT3lO+pgsUbKqW+ufz3fQH8b631v7suoMzevALF3FQopT4PwN0AtK1MvxLAt1fDy0qphyulHuJ1\n1yQalFLfpZS6DDiMSrwFhZ/8Z631P6KYE/20shP9PgCfpZS6qOyAP0jqusmycOjwPwCcp5T6NBQL\n4m4E4H+XH/nB8vUb13xdMqQcAF4E4B3lisoLUExYN8Uwt2BegmJ+1/llw/1iFBlIk5fhXIN+IYo5\nVS9HMTz9mUqpLy/fdxmA/1H++5Na66tQOOKfVlZKkQTN+1Csrn2RUurvUMxZfUH52uNRrFO6CsVi\noStRNLx/BeAPUGT9/gzASwH8OYCz5efeAeAu5TDITwH4Xo/reASArBzKfRGAS7XW72r5zIvL335D\n+blLcHpODomfKwHcUSn1tlKjXwDgl6oXtdavRjFX9emlH3o2gL9E4bP+P4HrJcukToePRqGzfyz/\n/+cB/KVS6i9RLNp8MYCXlXPok2SV51MmtcJFKfUZKIKuCwC8HUXQ9BIAzwdwU631pUqpJ6KYiP8B\nFFtx3A/FFhlfiGI49ZbVHEKl1DsAPFhr/ZpyQcfztda3Kl97FsotOlqu6VEAfhRF5ucPUWTsHlJ+\n79cqpc5HsWrzDiiGpH9Na/1r5WePUKzWXAF4W/m5GwO4Smt9pnzPD6DodX815wLGj1JqVc3nU0q9\nEMBrtNa/0vD+DNx2gxBCCBIOAAmJGaXU96PIEN8LxfD/3wK4ZzkZ2vWZDAwACSGEIO0hYEJi5jko\nMsVvA/D/A3hSU/BHCCGEmDADOCNKqdcCuInj5S/XWv+L4zVCCCGEkNFgAEgIIYQQkhp5nosZilWL\nUVuWZbnJbrcTv6aQTVJvS9NearbZbE48a7Fqby79bTabU2U2N7vdLt9sNuLaCcFsDUjZXPdra486\nCE97SQhxSqtzsFmWiV9XqCbt/JakvdTMpmuDIq25qfW32Wzy3W53qpxCIfXOcR6A7vIZfV8dbBvD\n0l6yJ4FMyfHxMfb7vfRlEEISIcsyHB8fe79/v9/j7Nlia8jNZjPKNVTf47qOLMuQ5/nht/f7Pf1k\nYrBtDIwUeiJTG3s6w3si1B6tyeypFhUxam9s/fkM81bDwXP5pSzLvIef57wuSZPW3BTa66NJDgWH\no73FC3EKq5xb0xybFBzamEKk9mhN5nrOYtTemPqrG/Kt5t2F4oN8g8HdbhfMNU9h0pobW3tN1lbf\nS67nEM2phaULcQyrHFjX+TXs6fgLkdqjNZmrQenyjElrbkz91WVEY5hjl3IgKK25sbTnY3Zbadd7\nDFpdkjm1sHQh9rUxJ1Qv1aGNKURqj9ZkrmcxxQBwjOFwafPJCi4tSJDW3Bja63Cvp57RuqBQuk5S\nMacWli7ErpZl2WQr6RgIpuUEaaPqxkls2huiP1fmL2a/0uZzlxIoSGtuqPb6arSqvzrtxqzbmMyp\nhSULcahw66jm/dV91pwX2OTQUha9pN5C1h6tVTdOYtPeEP3ZLClLtvRAUFpzQ7Xna037dXIoOCzt\nLVqIXcwV/LkCPt/vdDm02HvtYwsxZe3Rmq1uQ1n7/2PSXl/92b5kqY1nU0c8Zr8prbm5fJ+t07bX\nYw/sQzAz+eRIQNVrYclC7CvYytGM9f1NgWCszqyvSeotRO3R2q0uo2CSQgBYFxRJ18vc9W4SYyAo\nrbm+2utxn43PJoeC/cwO6gauS6jXwpKF6GNTB3+28Ot+LyXxS+otNO3R/Kwuo9CWZQhZe330l7LP\naAoEYyoHac311V4Xc83/a6vTpWazfW2GYxvrtdBFOGObdKFLBWOpDOXUmaTeQtIerZNmDlQNSp9z\ngaU111d/KfuLypYwiiKtuTl8n/1cNtVNykPBc5zTbe1VXK+FJqFMbdIVYDOnI0lV/JJ6C0V7tM6a\nOfWc+GYaQtReF/2lOPTbVh51gWAM/lNac3P4vi4ds5SGgqcM+KqN3/tob7FC9HgYTxTg3MJLSfw+\nQkxJezR/a8oomCw1AEzRR/hYjBthS2tuDt/XtT6WOhTcN+Crgjkfupz249TCGILqa1KV0yd7MJVI\n7MqXFu7UJqm3ELRH62ZNGYWuz7C05rrqj0O/zRZbECituTl8X9dnEljGaNjQgM8M5JoWevRNVrm0\ncAYJkmXZif/fbDYi17HZbHB0dHS4nizLsNlsxK6HkNA4OjqSvgQRsiw75afW67XMxQTKfr/Her3G\nbrc7/C3LMux2O5ZVRGy321NaH0LdszMVx8fHnd6/3+9x9uxZ7Pd77Pf7U6+7rn273U4TF8zZ87AN\ncj2xExG11HUA6Q0FS+otBO3ROuvlgJ0ZaHotZO356C8lnzDUYskESmtuat9n10MXzfZZ1W//9tSL\nKvpQl+HzLYM8H++5z11acL0wh0k8hENEOpWlNBQsqTdp7dF66eVAKgEgh367W93ikNDKTVpzU/u+\nPivzh3w2xKCvS8Bn3kfd94wZm+QuLbhemMMkHsKhPY3Urmtsk9SbtPZo3aztxI+ujYa05nz0x1W/\nw8z2oyF08EPT31T3N2YA6Kq30IK+PgFf033n+TRzIHOXFlwvzGFCD+GkBT3WwxPStc0hxBS0R+tm\nbQ3KEgNAm5ACmFjMJKQsoLTmpvZ9QwJAu97M9q9L0OezJUoo5lrsMaf2VqUgRFitVrP++G63OzHB\ncrVa9f6upommrgmebZh1UU1wXhp5nvcv9BGZW3ukG9Vk/oq652Gz2ZyYhN32PIeiPeC0/uz7Ld8z\n6zUtiRB9aUD6i973VYspYlswaccgFZMt8ihxam+JPRGXuXoYdTZGqrlranhoDyoGk9SbpPZonXXS\n+rwuKQNow8zfMOtyIsWMmhbXXWlRElN2r7IqjnDFEnPtQZw7tJDkNjBA8/YSdb3xPhwfHx8yFNvt\ntndmkJBUqHv29vt9dD39LtTdL/3EMOzs8PHxMcvUg+122/uzZnlXGbqxviOW57+6Tp/tYYLITLsi\nwzkMwr3Cuvc0bcI4Fk0TXE1i6+34mI8u5jDpcqCdtD5HfC0hA8hFH9NZaFlAac2Z2mtr56qsVZcy\nsz/fs4wGf8dcVmX3+sQMc2vRpYVkM4DA6Xl85qbMFa6eTFMv3d7g2abq8dvj/uyhkhQwnztXT3m/\n3x+y5kvGzv6JZwQWBLOAbtbrdeNIV1Vu1X/bNjAei/1+f8I3hJT565LdqyPIeYtT9jTaDDNH7EPm\n9PUdq/fpIZhzG+y/z11GU5uk3izL8/xcT3fIUn5as1Xl26Wn7Pu9XT4TgOZO+D47+7fE513aQsoC\nSmvO1J5pfTJY9vYnY81fD2Ue/JDsXkXl96SfgSbtJbMK2O4N9mHomH1bZrCOpa0EzCNaCddlHktQ\nvTpBfLJ7TXTJ/MW+Crjr9ZN+mG3c1KstW64jiAp2tbvVs9u1jQJOZu7K3+h1bfYzsV6vZ8naVvc+\nNLsX6vxdl/YWHQBWFTrWuYBjOY8uwaikw5qCUJwgPALAORky+bqir/PpenZmnwbCxdCJ3l0a94C0\nxwBwRhgAnsS33R3SmRuy6HHO+uobI1T+Opa22ak9V2pwDsOEqds+VEMw5tJt1/BgNWl9SIq3y7WG\nkkoeapJ6s4wIMPZwu/3dkWgPQDjDXUu3LhqZ+DrEdVdpr681bWlSR5/fmONUrC73MPS0jxDMqYVY\nhVhndSvq7Ip0jel3rVzze8ZwKr7BYOyBoKTeLDtg10GfOWvkNJXjnFKzJgwAaUM1MvF1iOuu0t5Y\n1hYQ9ilvux0fM/Bqa2Pn8FkhaW8xQ8BNw6pmOtreibvvakO73MYcvvFJS8c6NJwHMgyC4sEAME5Z\nDv38mMOqU9M2N3JOXZrPIYeASR2hlHMo+pti6pUrjug7h8/8vrH2y2uKdWJtT31xaW8R28DUHa/i\nCuzOnj17eG8fYbmWzmdZNtrkT3PuhCsYrDaZXrpw56BpU3BfQqiDrnP5KkKduEwIiZu+fmW73R6C\n9sqv9f2utuRQCL5bjJhT0XVDvj5bKfRNKTcNMU+dMnZtlDvX749lknozbY55JrRZ9OT9DEhrzjSA\nQ8BzWSjlLK05U3sT3FstQ7Y2Gvo9Yw9Nx2xOLcQqRFcwNHEhNjKHqNoCwdAnqkrqzTTbOaTmEJZi\nXepQWnOmAeEEJku3UMpZWnOm9ia4twNdT/QZu94Y+PlrLzohurJwUx+q7LtqaK7DnZsCwZBFLqk3\nW3t2vc1Vd7RR9eSte2nN2foLJTBZuoVSztKaM7U3wb0d/CjgTtB0bZu6frYpOSKtQ0lzaiEmIbqC\nsDl20K/7vaYh4bl29XcFgqGeKiCpN1t7MZUbzamnAwwAaXUWSjlLa87U3lTla3agx5iy5DtVRyox\nFIs5tRCDEJuyXXNVrku8bVuFzFXB0uUzVIgS2nN1KKTLiOZvpu7bgndpzbXpT7osl2qhlLO05kzt\njXxfzudv6Nx1n7pz+fHQ2j5Jc2ohdCE2RfYzF2CtaNuGhucUoausQhoSltRbnfbqkC4jmr91adyl\nNWfrL9RndIlmIhUYSGvO9n1jmNnmNGm4rZ10fbbt+XaN4jD489Ne0EJ0BTRzV26bCG3xNb13agt9\nSFhSb3Xaqysr6TKi+RsDQJqPhVDW0pqzfd8YZvpPn/d3DQRdm0KHkhiKxZxaCFWIri1eJCL7rr2Q\nLMvEeyChBoGSeqvTXp3OpMuI5m8MAGk+FkJZS2vO9n1jlmvXMu0SCJpUJ3XUId3mhmwuLZyHQLE3\nbqw2bQ5xw9q6UxGkr3O9Xp/a5Nq1iXXKVBuGE0II8cfcQLlre7fZbLBarZy+9/j4uPaAh+pgBJOQ\nY4PgCbEnYmevAslcNfZ0XL0XaavLcEn2lCT15tJeKJPEad0t1gyg/VyG5DOWaCGUtbTm6nzfWGU6\n9Lt8t1mzCSE2iMFcWgguA1h3nNUY5wAOZbVaYb1eR3dsTN1xd3U9K0LIfDATT2Jm7PajLSNoU7Vr\nIcQGMRNUAFg3RBlSBTelmE3hjnG27JjUDXO6zkYkhBBCmhi7/dhsNtjtdq3fu9/vD8kYDvkOJ6gA\n0A7+ttttlJUcYnZts9mcKEvOBySEpILp+0LroMeGPUrXdw51lmXYbDbI8xzHx8de7WaIbWvMBBMA\n2sHIfr+Pbrg1dOoWhfCBIoQsHXOhHn3eMOzy69NO73a7xoxfNWoVYwIoJs5IXwBQCMgUVd28NTIO\n6/X6RLC92+2YTifRst/vTzQiWZZFoeX9fn/C5zEr5U+fjivLdzzM563Ps9Y0B70K+szvLReukCmQ\nXo0U2irVvubasDJEs1dczbmSSlJvtvZc5SFdP7TOmjrQtMJTWnO2/mx4gkGztR0r1gUhnYrrrtJe\nX7N9ZVe99jm5w/xNrpYfV3viQrSJ2QGahC7Uus2rJYUoob3KGADGbb7PnbTmbP25CN13SFjfbULq\nkNo6RFpzlQ25hyEnXQ3d3i3m2EDaXFoQnQNozx2wU79kOurmAxJC5HFtgpsi1WK1sVadcuP3/gxZ\n/GHruc80L8YGEyDZE5E+N3dsM4mhFy9R/pJ6M828JmYA4zbf505ac7b+fLJaKWc9moZ7YywXac2Z\n2hujPvp+jps3h6M90UUgYywlDwlzYncMk4632+2JOthsNlx5HSlT1ltMmfkYnjsX2+32VKarWrBV\nNzl+yZSByim22y19lBBd2+u6rca4wDMwhHsjB2LImLVZjJmkuXtm0r3fyqTqbcy5TKGw2WxOmeRz\n58oO5QHorjKX7toWOux2u0X4yibrs1AgBpPWnKm9qZ6xJg0z8xee9qTFeEC6gMawGAPAoau6xhLi\n3CZRb2OtYIyJKmCZQVcnfjNk7eWl/pp059NRWEJQ5POMLCXgldacqb0hz1hTIOfycQz+wtSetBgP\nSBfQGBZjAFiK48DUzlZYb7VOcI56SzH4q6MKCMfWmc82THkAuqus7tmru+bNZuOlnSUESXVzxZYU\n4EprztZe3+es7u9NHZal1WOM5tLCqhSEFDmwnHkd9pyH1WoleDX+2Cu0przuPM+DKJTVanUQ/maz\nOTH3auz7r1sBZ55MMAZTPj9tG+8OXaE51vw2s5zr5hqFoj3gnP5M/9vmB22d2lS6itGX1q16jsV/\n+hKK/kzf1wVbPpbrrQAAIABJREFUVz7PPQ8ZCAOn9oR7I4vpvVZmEst92dmTKa9buvdbmXlNU2YA\n7bJNYSgkyzLvzNWYZdOWBZTWXJ3+zDLyzZJU5dvEHMPuUz0jXcoiJpPWnK09H+s7ZzmWti8Vc2pB\nWIyLE0usD4HZEE0ZpAjrrdYJThkAxjotYGyrhn3bGDpc1LRVhbTm6vRXTZjv6yt8Au2Q/VAqwV9I\n+htSN03sdrskOrgxmlMLwmJc3ANvNnLS19L3uqesE2G9OZ3gVPfOANBdLk3BS9/vbcpmS2uuSX9D\nra2xDrFhTin4C0l/Ha/ZqSepVf+08bQnLcY8z5fXKMbqxOZoMKSdn8sJmozp0BgAtltdIDBEf64s\noLTmmvQ3Zlm6guqQ/FJqwV9I+ut63Qzy4jenFoTFmOc5G8WpzdexzhGsSDs/lxM0YQAoY2OtAnVl\ns6U116S/Kcwuh1Aa8RSDv5D0J10OtHC0Jy3GZB5+KavLBjSVt/n+OYUo7QRNGADK2hgZwbrPSmuu\nSX9zlaV0EJjysWDSmptbe7RwzKWF8xAAPPR8OuqO7NntdqeO6KkwtydhvRAJ6raEqbZY8tWkqfuU\ndWyXo/RRefZ2SDwWjBBBhHsjBxBAlDy22RPd+/S+zQySfdxWl20juswLmrJ3Lt37dfWCTZgBDMNc\nq4Z9dF83/CmtuSb9zVmOkiMuFSll/ox7F9fd3NqjhWFOLQiLMQinNLY1bcvQ5T77LMNvChBdDart\njKdcyTyDpno5QRMGgOGY6xnweY7sOpXWXJP+prap9N33OqSHooXuXVx3EtqjyZtTC8JiXJxDaAva\nelTcKFTl68oGmuVv3sPY9ZIH4ADzGifoKosR7neyskzJus5ltcu+RFx3Lv3NXX5S9ZjysyCtOSnt\n0eTNpYUg5gACw4+TCgXXsTfb7bbXfJfVaoXtdns4Lqsvx8fHyLLsMO+maW4Qj+4hobFer0/NZ22b\nE1g3/zVV7LKI8bg4Qsi4nJH88e12u5jAz2S1WmGz2Yxyvingdtb2Ga1HR0eNDWJ1LXVnu7KxJKFT\nPQemz9jtds7zRvf7/SL9Sx9CWwxCCAkAyVR0aFsULNHsTTx99+Caqk6khz9cwyBT6ZD6Ht+6bCVi\nIa47l/7mMOn5qLGekjSWSWtOUnu0MLUnOgTMXun0bDabQ+ak2krDxJU9ISRUbM3W6bqC2naT8vY4\nhBDIzwHkfl3zwOCPLAnfINDc1zJ17Kkk9LeEpI14AGgHIJycPD4M/sgS6ZIJJIQQcpLgAsBqtSoZ\nD3siPIM/shTslfUMApsxn3sukCEkbcQDQOD0ClQ6pnExh8EY/JGlURcEshNZD4fECSEVQQSA1ZYp\nFezFj8tms8F6vcZqtWLwRxZH3ZmyXc4NJoSQFAkiAATYi58aBn5kybiCQO4scBLOuSaEVAQTAAKn\ng0D24gkhvtQFgfQfJ2FHkBBSEVQAuN/vOR+QENKbOh9C3MyZIa18OeuHkDAIKgAEOB+QEDIM24eQ\nk3DvVUIIEGAACHA+oDTmvCA2pCRGuNqdEEKaCTIABDgfUIosyzjsThYBg0A/6FcJSZNgA0DOB5TB\nLOPtdptEA2o3gCnccyqs1+tkdOwLj4QjhAAA8jwXMwB5m+12u9zk/7Z3/8H3tHVdx1/n9qsYGArZ\nSBOO2A8virRGrcm0+ewpJiOKEbCkIgwzQyZLndIC6exJqz8indLEmYjQIayG8gcjkVh7PqPDGBgy\nRcECFkVhBWncg1TCzfbHOfu591yfa3ev/Xldu/t8zLxn7u/9OT/27L732vde17W7WZa1vofoF0mS\nXK3rKb4jZL7V5V6appP97qo0TYNv4y1H6Jzr2vZNvC5mz8tqWx769285/0KvByKe3Iu2B7DEfMD5\ncLHNuOyeFu65hliEeCRc9SkktOFAeNEXgBLzAedgFycMmQHrFeKRcNU2hvYbCG8RBSDzAafluvBj\nS/fqsuc9ApgW7TcQ3iIKQIn7A07JVfzRAwisVwyPhKMXEAhrMQWg5J4PyHDwMK45lcxVA9YthhM8\negGBsBZVAEr1RSBFSz92L6q9fteOW8AA8z0SjqeQAPFYXAFY96zPw+HAkHBHrgs/tlYAUQBiq0IU\nYzEMPQM4W1wBKJ0bDddd/pkX6G/rF34AmJ/dZs/V8wjgvkUWgNK5IXENVyZJoqIoGF5o4CqUufAD\n2La52kyGgYE4LLYAlJobjyzL6A10qFsvWx2K4RYw2CoeCQds26oKQPsATm/go+rWxel00m63C7NQ\nK1ctLuldBQDEZNEFoH2Ate8VWCp7vbZaCKZp6uz1Ox6Pm7vqN5QQT14A2oR4JByAOCy2ALSHL8re\nv/1+Xzs3MMsyFUWhNE03UwxmWeZs2Pf7/WaHfUtTXgG8lfzCsnFiAmzXYgtA++qx6sG7HNasO6CX\nt4wpimK19xBk+LvdnAXgGnMMALBciywA7adX1E3er+sNtD/rcDhcFYNLLpqSJKkd8q1iuGda3N4C\nABCzxRaAVU29K2Vv4G63a73VSVkMLnGouCz6XEO+5S1z7NsvLOW3LZHPCQoAAMEURREsJBV9oirL\nsl6fkaZpkaZp0UX5njRNe33nmJEkSZGmaZFlWeMy2+tnjHU3JELmm5179rYd+XdO9tnEsnOvzL8Y\nwm4D1/qdMUTonIst94jwube4RLQbjyRJBq+cspjqK8uyu8JwjOWpW74uy5hlmXNZplh/YyTi3HFZ\nlqv1NdZvDL2Oibhzr8y/GIICcHv5F3o9EPHk3u6SEEHsdrvOX27fzmWKe9ilaaqbm5tRhkh9hv/G\nmo93Op10e3vbesFBdZu7Hqk3paIoorjp4G63K9I0vXcrobarIn2el2zvU9xnMQ6x5J7Ur+2bgr0P\nzJGrc7ThMYol/2LJPcynNveWdCaSJMnVmeMcQ2vVYd9Yde15TJLkrtdy6vVnRxHBGXBxyT07n6YQ\nYpidiDv3ioh6YUL0xlVtaXpE6JyLLfeI8Lm3qB7A6rLWPQs4BtUeuLbePZ9ep/J1a3iaRBHhWXB5\n8c/YF8Ucj0du/xKRWHJPiqcXZu4eQPv75h6BCCmW/Isl9zCfutx7MPeCjCXmG5hWD/oUAPHrWlz7\nbFO2O9BuK8UfEKPFFID2AZWGA6FQ3AEAlm6R9wGUKAABAAD6WkwByJMVAGBc1XaVG5YD27KYArA6\nQZ/ePwAYjqcBAdu1mAKwKuYLQAAAAGK3iAKQs1QAAIDxLLIAZAgYAJaHthuIxyIKQBuNCAAsj912\nc0slIJxFFIBcAQwAADCeRRSA1SFgblUAAAAwzCIKQAAAAIwn+gKQOSIAAADjir4AtFEQAgAADLO4\nAhAAsEycwAPxeBB6AdocDofQiwAA6CBJknv3b3W15dzSCwhnVxRFuC/f7Vq/vLp8x+ORM8iFK4pi\nF3oZJL/cw7rEkntSPPmXZdlVobbb+a0iu8C7ubnp/MSm0+mk/X7f6T1LFkv+xZJ7mE9d7kXfAwgA\nmMbxeLwq3NI0vYuqMUdiTqeTjscjvX9AYFH3AKZpetXw7Pd7Go2F4ywYocSSe1Jc+Wf3Ak6lLPq2\n2obHkn8x5R7msYoewK02HAAwFbsXsI/T6aTb29urf9NeA3EL2gMIAACA+XEbGAAAgI2hAAQAANgY\nCkAAAICNoQAEAADYGApAAACAjaEABAAA2BgKQAAAgI2hAAQAANgYCkAAAICNoQAEAADYGApAAACA\njaEABAAA2BgKQAAAgI2hAAQAANgYCkAAAICNoQAEAADYGApAAACAjaEABAAA2BgKQAAAgI2hAAQA\nANgYCkAAAICNoQAEAADYGApAAACAjaEABAAA2BgKQAAAgI2hAAQAANgYCkAAAICNoQAEAADYGApA\nAACAjaEABAAA2BgKQAAAgI2hAAQAANgYCkAAAICNoQAEAADYGApAAACAjaEABAAA2BgKQAAAgI2h\nAByJMeYpxpiPh14ObI8x5k+P9Dnk8IYZYxJjzHuNMX/DGPMin9fW/O2zjDHP8vi+Vxtjvq3jMv5J\nY8xPXP77ZIx5fpf3Y/mMMX/WGPPtM35faox5Zc3f/qUx5gvnWpaxPQi9AAD6M8Z8kqS/KenvhV4W\nrEOe53954EfsJT1d0o+OsDjAlTzPvyf0MpTyPP+9oZdhCArAHowxDyR9n6TfLemTJP1bSenlb18j\n6RslPUHSt+R5/oPGmJ2kl0n645I+VdIPS/rmPM8fMcY8WdIrJJnLx//5PM//uTEmkfR3JL1J0h+U\n9CmS/mie5z89y4/EUrxJ0qcbY94l6VdIeo2k50j6U5L+uqRX5nn+GuncY1L+2xjz+yX9LUmfLOnd\nkl5gf7Ax5jWSfjHP82+Y44cgDsaYV0t6b57n32GM+XJJr5T0EUnfJenlkr6g8tqXSnq+zu3T10r6\nsKTvkfTAGPNpeZ4/r+Xrfq0x5lbSUyS9TdLz8zz/JWNMIemz8zz/r5fvKSR9tuP9n2+MeYukXyPp\njZJelOf5I/1+OeZijHmBpLL3919L+g5JmaR/LOkLJX21zjn44PL6p5T/Nsakkp6c5/nXGmP+sKSD\nzsfhj0n6c3men+qOqy3L9Ft0PpF+vM75/LcrxeZjjDE/KOl3Svofkp6b5/l/M8a8T+f8f6AFHq8Z\nAu7nyyV9rqSnSvqNkv69pC/ReX1+Sp7nXyDpm3ROaumcIH9E0u+Q9Osv8fWXv32/pLfnef55kv6A\npNcYY37V5W+/WdJb8jw3kv6azgkNVH2NpEfyPH+qpP8k6YskPS3P8zfXvcEY8zhJ/1DSV13y7r2S\nvt16zbfqfBLzjVMtOOJ26V3+fklfl+f5b9K5rXtc5SVPlvTvLn97haRvy/P8bToXgK/zKP4k6RmS\nvlLSr5P0RJ2LyC72khKdD/Q3Oh98EbFLMfdyPbrdHifpWZI+U+dj4U2Hj/teSc+85OCLL58jNR9X\n6xwkfV+e50/T+Xj+dGPMYy5/e7qkv5Tn+edK+qDO7a5tccdrCsB+Pqjzxn62pMfmef4ySf9C0k7S\nD1xe87M6N5CS9IckvSrP8w/nef5xnc+on3M5EO91PrNWnufvlfSTkp55ed9HJP2Ty3//U0m/zRjz\n2Cl/GBbvDXmef6LlNV8q6f15nr/j8u9v0fmERZJkjHmmpOdJeh69KZv2eZIeU+k5+W5dHzMezvO8\nHOattnddvCHP8w9e8uyf6Xzg7eJ1eZ5/NM/zj0r6sR7vx/x+n6Q353n+gTzPC0l/TNIP6Twa8UMd\nP+t/SnqRMeZz8jz/qTzPv9njuNr0Wc+9zOn7X3mef0We5//v8refzPP8P1/+++1y5/rijtcUgD3k\nef4WSd9wif9ujHmtpM/QuSfmo5eXPaJzt7Quf/sLxph3XYbqXq7zcN2n61w0vrnyty++vF46D78V\nl//+35XPAur8gsdrPlOP5pPyPP/lPM9/+fLPhyT9fUkP69ygYbueIOkXK//+gPX3hyv/XW3vuvhg\n5b8/fPnOOd+P+dntz/+V9HGdj58P177L7VmSniTp3xhjftYYc6P242qdb5X0Dp2LuPcbY15c+ZtP\nri/ueM0cwJ7yPH+dpNcZY54o6VWS/mLDyz8g6UftyauXuYSPSPriPM8/Yv0tkVTtsi4bNp8DPCDd\nb6jKHPqQzo2wJOlylvrEyuu+TNKrdR7+/a5pFxERe1jSp1X+/aQJvqOad0/Qo+3bJ3TJXWNMU1FX\n937E60OSflf5D2PM4yV9lvWaRyQ9ZIzZXYoqZw7kef5zkl5ojHlI53nMr5X0Oao5rja5vPYlkl5i\njPntkt5YXnHuaXHHa3oAezDGvNAY8zJJyvP8FyS9S1LR8JYfkfQnyu5gY8yfMcZ89WU4+Mckvejy\n/x9rjHmVMaac7PxYY8xXXP77KyX9zOVsCSh9TOeG8lc6/vbzkn6rJBljvkTnIT1J+ilJT7o0ctL5\nAqW/cvnvT1yGTF4o6aXGGCNs1XskffLlZFQ6t1NN7VzpY/Lv+XiGMeYJl/mGz9Z5qE6q5K7O863q\npjU8xxjzqZdhv2dU3o94vUHSl15uO7XT+YJK+2raD+lcxH3+5d+ui9R+tTHmTcaYx1+mvfy0pMLj\nuOpkjHm9MeZpl3++Q+ceZZ98Ly3ueE0B2M+PSPoiY8x7jDHv1Hk+4Hc2vP6HJb1e0tsu3dHP0nnO\noHS+GOTm8v/fJuk/5nn+/svf3ifpy4wx79b5zOTFAq79vM4F3X9R5az64jslPfOSoy+Q9OOSdJmm\n8FydJ0a/W+erOl9SfWOe5++R9Fcl/cDl4IyNucx/+npJrzbGvF3nq8U/ofaD4o9L+j3GmLd6fM3r\ndZ4v9XM6X135Dy7//6WSXnH53l/S9RBc1U/ofPXoOy///UaP70RAlyu7v07Sv9I5pwqd539WX/N/\ndL4o443GmJ/Red6d/Tkf1Hl7v9UY8x8k/SOd734gNR9X63y3pNde2su3SfreSzvo631a2PF6VxRd\nClzM5XLW/co8z39D6GUBgEsv20ckfUae5x8OvTxALJZ6vKYHEADgZIx5qzHmqy7//CpJ76T4A9aB\ni0AAAHW+SdLfvTx662Gdb9DrxRjzVJ2nv7i8M8/zZ4+wfIAX8vE+hoABAAA2hiFgAACArSmKIljo\nfPVPtJFlWeGSZVmRpmmRJEnrZyRJcvc55ftcr0vT1PldfWRZFnzd1UXIfKtGyHWQpmmn7V2XM1uP\n6r7lKXjejZl/5MuyInTOjZl7MYVvWxp6OWPMPRJxhMSqK+ySJGl9X8cDmLdYi8AiggawCJB7aZoO\n3tYc2K/XZw/B866MqX+/z8kpMW+Ezrmxcs83XJ0fY+dllzY19PaPMfeCzgHc7XbhvtxTlmVKksTr\ntcfjUWmaSpKSJFGWZb2/93g86nQ66XQ63f2/8rNth8Oh02efTifd3t7Wft6UiqLYzf6lDnPkXpqm\nurm58cqf6jZJ07Rxm+73+6u86GKM/f14PErSvfycQ5IkOhwOznVarsMqa/1HkXvS8PzzaZeG5AnG\nt6W2T6pva06nk/b7/eDP73qM3e2iWP1B1ObeFs5EhobvUFPZ6+bq+fN9/5BenrYex7rlnTNceRAi\nxs6Pcli3yxlp2/Zu6uXps5xjTjNw/Y6peyib1m1dz4L1m4PnXRkj7Ede6AmMJ0Ln3Fi55xM+o18j\nrM8rZRtUbYeq+3/o7R9j7q0+EcdO6qaDUJnUrsSsfo5dKIzZNd61CJx7WDFkvvXNvWqBN3Qot0+h\n5Crc+my3qaYbuIxZFDYVrm2fv8YCsGshTxEYR4TOuTFyr2+OutqeIXlpH+d85taH3v4x5h5DwB6S\nJLkbbimHVHy7nsfq7u6qaXjXHl7s0zVuD8W5ht/K/1+us0uXfRT98E251zTM2IdrOL+P6r5anW7Q\n9f1128pH1+kGVdXv9Vn2tuHecr02sYbTo8g9aVjbZ08R2O/3re0Rw8HhFRsaArZzdLfbOYds+w7L\nuj7fpfqdfdrMtajNvbWfiXQNn+He8myj7Uw84gsxnMva5crmAYKfARcNuTfGhRpljN3zYn9P3/eP\n0StX7RUdoq6nsM9wryvW2ANorz/ffZKewLAROufGyL0+OVqcv7SQ7u/XfY+RdrvjsyxbvoiuNhfW\nnoi+0eWWEnbSug6CsRZ/kl+R0zQ/bWCRFLwBLBy51/W2LK75JjPsxFfL0DW/+763y3eMVRTWrfeu\ny7S2ArBp6MtnvYZsd7YeoXNuaO51/K3OHHXlaZ8TEwrAcXJv9YnoE10n7rdNOI890fr04JU9WvZ7\nqz04noI3gIUj91yGXpQzwU58tT26vDfUXJgxisKmfa7L754zv9qi7/psOvDVrbtSTLm8xQidc0Nz\nr+Nvrc073/l7TdGlAKzuA6FzILbcW30itkXbzZ7X3GiWB+euBbBrx/PtVZwql7pGdT24ipMYh8uq\nllIANuWdT1E4dP9bWwFYt//Z+WG/JsZ83lqEzrmhudfxtzbuw015PNZ3lBFT2xdb7q0+EZui7nYt\nW24su/TWlDtdl1vkhMy3utzrcjYZMlzrvs97Q/+Opryr3jh2rM+t/vRYYoocqLPmk9glReicG5p7\nXaJ6THDty2O0ub45Xv2urR7ba3Nh7YnYFHYSbjU5GpKmkWsdlju8rVy3IfOtLve2VABurRhYUwHY\n1l7VYe5fHBE654bk3tBcbfv7kIvafN4/5HvWEHW58JA2JkkSpWmqoiiuLiMP8VSDJTsej3e36aja\n7/f3btvBLSiGs9cp6xO+OVC9jRUQg63ejiU2myoAsyxTlmXOe5n1vS/altn3dNrv9/fWLYX1OIYU\ngNXGdmvbYsh9C2Njn7A2KR/X53ovEJubm5tVfc9SbKYAbDsLJjH8nU4nZ8+f66a9FNZx2VoBuBZ9\n9qtqEUgvIGJj5+cc38U+cC1oAVgOxQ4J367ktg3f9cHSW+bqjTqdTrVPbMC80jRVlmV3+wcnN8tn\n71tt7V6apvdeQy8g5mK3+z6FF8VZAIEnpI7C5+IN16TUpdz+I1T4XA3c9iQC+zOLCCZBF9ZE6CVc\nBOIzqdrnauzQv2PusH9+LDHkd9Rd1OHazlzoFk+Ezrm+uTfg996pu/jC5zVDPr+MOW6EH3PU5cIq\nhoD79jKlaXrvOb30WD2qrZeh+pzj8tmsmI99QRNn0NvQNvxbbcPsfZgcwVpUjzdtoxwc190ehF4A\nadjD6X2HgO15a+UQyel0untQNUlyrW3d2gVf+fpyPbM+p3M5k8cG2Pth235lt6XH4/FunzwcDlyB\niWhUpw4NyU1ObPqJogC0e+GmYM9Ts5ONYuW+PnOGqu/hApD5HY/Hqyuv0zRl7tfC2b0brraq6cBZ\nd/ILhHZ7ezt78TbXSVDdhVc3NzeT/uayQ83rNwaejzD7mPyQOQdbCtczf12qc4p8nvEYON+c82DG\neCzR1NE0H9Pn6TVbzXVL8Lxz5V+X31C3Hdsed7WEHF97hM65Prk38Pe25u2Q5wJ3nbs9NP/LpxXZ\nEbPy2FDU5MKuOCdEKFdfPtcNg+3fvIYbFZdnFGP8jur6qc7zK2VZ1noGczwenWcgRVHsBi/gOJyJ\n7zMdIcS9Dcuzyerwetnbh3rWvh5L7mm323k1vHYP7m7n/gnl73TtryV7v63bRzGNWNo+39wbqmvu\ntr3O9/PtnrfYRkCGTHnzUfN7nSs1aAF4Op2KEE+NcA2Ljf295e03pj5I27+l6QDgo3qQaPqspqHF\npvfF0giqpgAci+8FMa4DcLUBa2q8hm7rLVh6Adh2Mma/rqmoc93qag0nv0sRS9sXWwHo+zqf94WY\n8lLX1od8CIJjX6+vvkOFdH9ooiiGD80mSVJkWdb4Oa7vLd/jM6TWFPZn9/mspmVvu+VH3+XvOkxU\nrudym/lst5D5ZsXibXVYt0tYQudcp2E43+GxLre4cN2uKfQ22kqEzrkuuTdG+A7R9r1VkevzfW6F\n1aSsAexY4u2TrPXjzoWxk6tL1G3IckP0+dF2A9e04dqSpc/3+8yV67L8ZUImSdJ5vkGXpLWXe6oH\nyBcRNIBFUahcn9UY2njMpe3khnDvS0UEeVeGz/L7Hhjb5v913dftfWENB8MYInTOdcm9MaLLHL2q\nts6Pvu119b2hcyHA+nfmQtAh4GpXtGt4os8Ql/172rqTm7qMuw6P9B0WLfnMrav77NvbW+fVRT6/\nwf7eKYcWixUMg/g8VmusYYjqfBHma3Xj2B+jyD3JL//s/dJnCM13+MxuJ4/HY++rE8lRP2to+3zZ\nc5al+rlvU18VW/LdN9bCqqnimwNoJ+IYRWD19/i+t+4xcF0mSdctu29h1af4s2/5Ufc5TUXgnMWf\ntK1G0EddfoWcP7IWSy8Aq21ZU1tUvq5t3y0Pyjc3N7q9vZ1srhTzCt1ibvt8TmpLsV1UUYcTk6s2\nJM45gHbUPVKs6+Peug5h1g2xurqL7WVxLXP5Gpu9XNV5dPb31g2Nt62Lus+z3+f6/KmHd4oIhkCK\nGYdBiHDhEDzvfPOvy7yo6j5e95q6drXJkKG20Ns+xhgrd0aIVWJqzHUsZg6gK1yNTlMj2GXuX4eV\ndsc1L8b1vfZ31zW8TY1qkiRXF7KUxWnd7/Zdf/bvaVruqaII3/h5HYCJ5Ycj34PnnW/++c6f8rkA\npMvcYZ8TTNf8WfszQm/7GCN0zlUiKnUXXriibR8PvY1jikUXgFK3ItBVnPWNIRcDuBo/37PvauNb\nXYa639z0nV2+t+k7xo4ifOPnlXvE8qNqaQWgre51bReA+LRlWZZdva5PAddlwv9WI3TOVWIQn0LN\n7gSZap12vQBqS2Ht++52KGQi9vgRzg09tPFyRd3QbJOm7277PPu9Va6C1neIyKcInPOqviJ84+ed\ne8Ryw7V/hM65LvlX1XZVZMn+W1t7U/1ce311veKXArA9QudcJa628dKv7F7ysk+cb1e7pLMdCpmI\nPj+irRfQLnCmSC7fs+g+G6Zu56v7vU2f0/a9dXN65uw6nzKfukTonZOYNlwFSeic65J/vvun/Rvr\n/lblaquahol92jbfexZuOULnXCWKopj3xJ+YN3xvgRVtI9g0d676OruYmWJl2obON2xbXt+z6b6/\nfYoe0w6/PXTj53UAJpYdrn0jdM755l+fC0Ds/b+uoGva35v4zAusogB0rt/geXeJ1lwglh2O/d/d\nDoVMxLqFryv+QkwCrbtp6pDvvdoqHhvPdz35LMdYF8v0jVC5ZkfoHZSYPM/u7Rehc843//ru/02f\nURTtB/wmbdNbbKG3f4wROucqcYdewHXGogtAn7krrqHMsZO5rmHrcobuCvvz2jZe02d17QW01+Xc\niTl2DvWN0DsoMXme3VlaAei7Tze1Ezaffd01NaRtOdrugkBcbZPgeXeJq20cer0Qk+Sazd0OhUzE\nDgvvZexktpUN29AJz23v7TJE67oFg+/tIEI01K48CBGhd1BiuqjL89A555t/Vb4XgBTFo22J63GS\nfddb07LH9xJOAAAM6klEQVRQ/HWL0DlXxhzTpojguXa139blQpSNYJf7VlWN2aNVN/Rrr9w+31nl\ns/G6fF7de11X+QVKzOANYNFyACaWHXWFUeic880/3/20qtoO9T3Rc73PVSzU3TSf4q85QudcGczX\nXHe4tm9dLkTbCLpuNtp0V3qfG5h2CVvTyh3y2W1/7zOvz0eo5OyaI1NF6J10SNRdyd1209StRF0P\nR+ic88m/vheANJ3sdVl3rpyqqmt/Kf681m3wvCtzj17A9UaXOyA8UKTanoV6SeS71475/Fr7mYjH\n47H2bzE88/J0Omm32ym9/+xTp+rvwTKkadr40HTXdnf9P3vbpyt8RmZ1HS091+val67PDR+L63t5\n9u/yHI/Hq22Zpukq24Iturm58X9x6DORvlE1dm9H0xn0GGdOTe8f476GTT2loXuGQubbWLk3VzRt\nxyktvTfR/i2x5V7RkH9jXAAyVg9gm7FHXdYeoXPOzj16AdcZ9j7alHvR9gC2OR6Pdz0ch8OhsefL\ntxegPANq+qyxexeKotDxeNTpdFKSJF49eG3K35Ekyd3ycna3DG09fVXV/PN9j4++vYlS+Dyzv39p\nPVOhey/LdqjJ8XgMvp0xHL2A63d7e9v498UWgF34FlWHw0G73e7q/51OJ++h1b7qCtihB6+2YXTE\nIUkSZVnm9dryZMFnu1ZPAEpj5vESCsW15n91Pbf9xiRJ7l5Tt959TiAo/NalbEfK7X44HNi+C9f5\nBDiGrui+MYUuQ259l9v3O9Y4vFJEMARSjJB7Y4VPLkz5rE7XcG8I1ft7Dvm9TcNaRQR5V0bdtqjy\nvQK4+ro+zzD3FXpfWXpMkUd9oinn1njM2VJ0vQBud0mIIHa73eAv9zlj8e31KHtXfHpjhl54UvbO\n1PWYrPVMrCiKXfurpjdG7g3R1uvXpadvDq58nLJX3HY6nXR7e9vp4jB7P4ol9yR3/tkjDfZoRMnO\nnervzLJs8FSAcl2naXrXTq21PZpTLPln5151nxn7gkrMq679q829WM5EYoq6mysP7Z2oi/KWN1s4\n+wp99htD7s11K6MQMVePYt3FKfZrYsy9oib/prgApOs6XXr+xRyhc64u9+gFXEc0bce6XFh8DyCW\npYj0LHgOTXNJ19zr6zJ2j+LxeLw3j21pPYDVtripJ6atp9Ce+1k3v5iennnFkn9Dcg/xsnv/q+1C\nbe7FdCZCrD9Cn/2GyL2meVkhnse8lBjaixhr7tXlX5Xv/L8uj3kr1+MaepqXGKFzrin36AVcfjS1\nH7W5EFsiEuuO0I1fiNzj6QnDo8/FDbHmniv/ujxhyPd1RFwROufqcs+VV5yYLivaCvi6XHhIACbj\nmpRfPrkllgs8YlNedJBl2V1DVbce6yxt3dq/rW46QKgngGD97CdekWvLYT/9w7v9i/FMhFhvhD77\nnTP3XM9optfvfoxxsYjr2cix5p4r/3yf3jHkKR9E2Aidc3W5Zy3j1T4Vep0R3rnVq+2LNhGJdUbo\nxm+u3KP467eOxlA3LBoiz+rCsWx3mg68FIDLjdA5V5d75Neyw2f+Zm0uxJqIxDojdOM3V+759Eht\nOXx7/Kq3YKq+t2k+YOy558o/31yx103o7Uj4R+icq8u9anSZi0rEET63j6rNhVgTkVhnhG785sg9\ne4fse6Au78u2tp7Dpiui6+7vVxflPTR9Dlqhc64u/7gAZBsROudcuecK3/tREuHDt+2ozYWYE5FY\nX4Ru/KbOPVfPlu/7mnq21nCwrxvyHaMny+fzQudcXf75DrvRO7PsCJ1zrtwjz5YdtrrOgtpciDkR\nifVF6MZvytzznfdX9lr1ufBhqb2BdcXfWL/Hp4gKnXN1+df3ApCl5sJWI3TOuXKvLuwTUXItvugy\n0lSbC7EnIrGuCN34TZl7NvvMeazHoi3tRr6u4m/s3zBkInSIqMsbLgBZb4TOOVfu1QUXscUdrsfV\n9sm96BORWFeEbvymzL26HbLrTYyrz5yuKxqXdAHAWHMiu6z/Jd0Gpm25Xa8rzh9CLChC55wr95qC\nIjDO6LNdanNhCYlIrCdCN35T5V61UKsWOG29fj4XPtQVkEtojLueqQ6JtsnroXPOlX9cALKdCJ1z\ndu75RN85zcQ00bcor82FpSQisY4I3fhNkXv2TlleuTv2BR1Lm5ztaqym/L62IdLQOefKP995fUvb\n9sT9CJ1zdu75xlw9+ES37dBlW9TlwgMBGCTLsrv/3u/3SpJEh8Ph3utOp5OOx2Pvx5SdTiedTqe7\nRzQdDofaR4bFwH6U1H6/n+27l/YouFLdcnd5DB4wpv1+f/UoxiRJVBTFXVtGLvrr217XHU+GtqkU\ngMAA9oH5cDg4n6F5PB5HKdZub2+vPj9Jkmgb4GqjNceBovp9t7e3k35XaLFuc6yTXQRK5/3tcDgM\nLkRCnMS6CqolKTsTBltSVzSx/Ag9/DFF7jWZ4ordqpiHAudezrbvC51zrvzzvbLXFnrbEr3yM3je\nVXOvT3S5mA3T6HM8KWpygR5AYABXb9/Qod42x+Px7gw29mHgudjrYCnrpNoTUXdGb+fYKGf+QA/V\nnr4kSWqnu8Qu1D4UW7tEAQh0lKbp3ZBmdf6fNN5QL7q5ubkJvQiTYf4fYlS2gWmaDmrzaC/DoQAE\nOijOQyjOs979fs/BOYCyJ6K09h4ycgyxoYhbpodCLwCwFK7h3ioOzGHYF5twMAKAdhSAgKe2K1nb\nCsQxpGm6yDk3Uw3Rpmm6qd4/ABhN3dUhc4QiuDKLmDdC5tuYuZem6exP53B9X8xXAc/x7Nqqvs/D\nDJV/rhuIh1qPxPQROuequUdsK2pzgUQkYkjEpeaey9h3ync9UWOK7xk7pn56he9TNGLLvTL/fAu7\nauEf+zYn4s+/0OuBiCf3SEQiikRcau7VGetAXfcs4Zh7/qrR9nzesda9z/oOnXN2/vW5B+BStjsR\nb/6FXg9EPLnHHEBgAuXjkvrOC0ySRFmW3ZvvV951fykXOthP5LBvm9OX/TlzPmYOANaAAhAYiasI\nybLMu1hLkuTunlr2Y5ekR4u/JV1tbP/2srAdYmu3fQGASdAVTcwZoYc/xs696jCedJ735rpYI8uy\nIk3T2otH2kx5ccnU4ZrD2Hco0zUkvrTcKy75xxDwtiJ0zlVzj9hW1OYCiUjEkIhLzj1XcTbWMzOn\neJZwiLCLwD5zJIdedR0656ph/56m9VFFAbjcCJ1z1dwjthV1ucCTQICBXEOy+/1+0HDn6XTS7e3t\nYub62arPCC1/R7lObm5uOg/buobElzYcbqv+HnuuJABMjQIQmEg5Z89+YHpd8bPUYs/lcDjcFTjl\n7z8ej52f1JGmqW5ublZX/AFAaBSAwISqD0zfktvb23tF2+Fw0OFwuNe7aV/U0fakE4o/ABiOAhDA\n6MrizlXMlQVf10falT2qW+K6EhwAxsBtYABMIk1T7Xa7UXrsjsfj5oo/iQIQwHToAQQwqXIYXDoX\nhW09f+Ucyer71obCDkBoFIAAZlPe6Lo6729r8yMlCkAA4VEAApjdmnv3AGAJmAMIAACwMRSAAAAA\nG0MBCAAAsDEUgAAws5ubm7v/Zi4kgBAoAAFgZr7PAfZ5hCAA9EEBCAAAsDEUgAAAABtDAQgAALAx\nFIAAAAAbQwEIAAFxFTCAECgAASAgCkAAIVAAAgAAbAwFIAAAwMZQAAJAhNI0bfw3AAxBAQgAALAx\nFIAAEAgXgAAIhQIQAAJpeg4wAEyJAhAAIpMkiQ6Hw92/6SkEMDYKQACITJZlV/8+Ho+BlgTAWlEA\nAkAEkiRRlmUqiuLq/59OJ3oAAYxuZzc2s375bhfuyxFEURS70MsgkXtbFEvuXRTSubi7vb29Gu6t\nOp1O2u/3sy4YphFL/tH2bU9d7lEAYlY0gggllty7aM0/ir91iSX/aPu2py73GAIGgMDKYm+3290F\nxR+AKT0IvQAAsFXH45EnfAAIgh5AAAiA4g9ASPQAAsDM9vs9V/YCCIqLQDArJkIjlFhyTyL/tiiW\n/CP3toeLQAAAACCJAhAAAGBzKAABAAA2hgIQAABgYygAAQAANoYCEAAAYGMoAAEAADaGAhAAAGBj\nKAABAAA2hgIQAABgYygAAQAANoYCEAAAYGMoAAEAADaGAhAAAGBjKAABAHfSNFWWZUqSJPSiAJjQ\ng9ALAACIQ5IkOhwOd/8+nU7hFgbApHZFUYReBgAAAMyIIWAAAICNoQAEAADYGApAAACAjaEABAAA\n2BgKQAAAgI2hAAQAANgYCkAAAICNoQAEAADYGApAAACAjaEABAAA2BgKQAAAgI2hAAQAANgYCkAA\nAICNoQAEAADYGApAAACAjaEABAAA2BgKQAAAgI2hAAQAANgYCkAAAICNoQAEAADYGApAAACAjaEA\nBAAA2BgKQAAAgI35/7KBm8oerNK2AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f9741c77710>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "4fZr4NXEkJJP",
        "colab_type": "code",
        "outputId": "5aa76b13-ebe2-434e-bddd-1beb22495b28",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1445
        }
      },
      "cell_type": "code",
      "source": [
        "%ls data/txt/train"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[0m\u001b[01;34mairplane\u001b[0m/          \u001b[01;34mcruise_ship\u001b[0m/      \u001b[01;34mline\u001b[0m/            \u001b[01;34mshoe\u001b[0m/\n",
            "\u001b[01;34malarm_clock\u001b[0m/       \u001b[01;34mcup\u001b[0m/              \u001b[01;34mlion\u001b[0m/            \u001b[01;34mshorts\u001b[0m/\n",
            "\u001b[01;34mambulance\u001b[0m/         \u001b[01;34mdiamond\u001b[0m/          \u001b[01;34mlipstick\u001b[0m/        \u001b[01;34mshovel\u001b[0m/\n",
            "\u001b[01;34mangel\u001b[0m/             \u001b[01;34mdishwasher\u001b[0m/       \u001b[01;34mlobster\u001b[0m/         \u001b[01;34msink\u001b[0m/\n",
            "\u001b[01;34manimal_migration\u001b[0m/  \u001b[01;34mdiving_board\u001b[0m/     \u001b[01;34mlollipop\u001b[0m/        \u001b[01;34mskateboard\u001b[0m/\n",
            "\u001b[01;34mant\u001b[0m/               \u001b[01;34mdog\u001b[0m/              \u001b[01;34mmailbox\u001b[0m/         \u001b[01;34mskull\u001b[0m/\n",
            "\u001b[01;34manvil\u001b[0m/             \u001b[01;34mdolphin\u001b[0m/          \u001b[01;34mmap\u001b[0m/             \u001b[01;34mskyscraper\u001b[0m/\n",
            "\u001b[01;34mapple\u001b[0m/             \u001b[01;34mdonut\u001b[0m/            \u001b[01;34mmarker\u001b[0m/          \u001b[01;34msleeping_bag\u001b[0m/\n",
            "\u001b[01;34marm\u001b[0m/               \u001b[01;34mdoor\u001b[0m/             \u001b[01;34mmatches\u001b[0m/         \u001b[01;34msmiley_face\u001b[0m/\n",
            "\u001b[01;34masparagus\u001b[0m/         \u001b[01;34mdragon\u001b[0m/           \u001b[01;34mmegaphone\u001b[0m/       \u001b[01;34msnail\u001b[0m/\n",
            "\u001b[01;34maxe\u001b[0m/               \u001b[01;34mdresser\u001b[0m/          \u001b[01;34mmermaid\u001b[0m/         \u001b[01;34msnake\u001b[0m/\n",
            "\u001b[01;34mbackpack\u001b[0m/          \u001b[01;34mdrill\u001b[0m/            \u001b[01;34mmicrophone\u001b[0m/      \u001b[01;34msnorkel\u001b[0m/\n",
            "\u001b[01;34mbanana\u001b[0m/            \u001b[01;34mdrums\u001b[0m/            \u001b[01;34mmicrowave\u001b[0m/       \u001b[01;34msnowflake\u001b[0m/\n",
            "\u001b[01;34mbandage\u001b[0m/           \u001b[01;34mduck\u001b[0m/             \u001b[01;34mmonkey\u001b[0m/          \u001b[01;34msnowman\u001b[0m/\n",
            "\u001b[01;34mbarn\u001b[0m/              \u001b[01;34mdumbbell\u001b[0m/         \u001b[01;34mmoon\u001b[0m/            \u001b[01;34msoccer_ball\u001b[0m/\n",
            "\u001b[01;34mbaseball\u001b[0m/          \u001b[01;34mear\u001b[0m/              \u001b[01;34mmosquito\u001b[0m/        \u001b[01;34msock\u001b[0m/\n",
            "\u001b[01;34mbaseball_bat\u001b[0m/      \u001b[01;34melbow\u001b[0m/            \u001b[01;34mmotorbike\u001b[0m/       \u001b[01;34mspeedboat\u001b[0m/\n",
            "\u001b[01;34mbasket\u001b[0m/            \u001b[01;34melephant\u001b[0m/         \u001b[01;34mmountain\u001b[0m/        \u001b[01;34mspider\u001b[0m/\n",
            "\u001b[01;34mbasketball\u001b[0m/        \u001b[01;34menvelope\u001b[0m/         \u001b[01;34mmouse\u001b[0m/           \u001b[01;34mspoon\u001b[0m/\n",
            "\u001b[01;34mbat\u001b[0m/               \u001b[01;34meraser\u001b[0m/           \u001b[01;34mmoustache\u001b[0m/       \u001b[01;34mspreadsheet\u001b[0m/\n",
            "\u001b[01;34mbathtub\u001b[0m/           \u001b[01;34meye\u001b[0m/              \u001b[01;34mmouth\u001b[0m/           \u001b[01;34msquare\u001b[0m/\n",
            "\u001b[01;34mbeach\u001b[0m/             \u001b[01;34meyeglasses\u001b[0m/       \u001b[01;34mmug\u001b[0m/             \u001b[01;34msquiggle\u001b[0m/\n",
            "\u001b[01;34mbear\u001b[0m/              \u001b[01;34mface\u001b[0m/             \u001b[01;34mmushroom\u001b[0m/        \u001b[01;34msquirrel\u001b[0m/\n",
            "\u001b[01;34mbeard\u001b[0m/             \u001b[01;34mfan\u001b[0m/              \u001b[01;34mnail\u001b[0m/            \u001b[01;34mstairs\u001b[0m/\n",
            "\u001b[01;34mbed\u001b[0m/               \u001b[01;34mfeather\u001b[0m/          \u001b[01;34mnecklace\u001b[0m/        \u001b[01;34mstar\u001b[0m/\n",
            "\u001b[01;34mbee\u001b[0m/               \u001b[01;34mfence\u001b[0m/            \u001b[01;34mnose\u001b[0m/            \u001b[01;34msteak\u001b[0m/\n",
            "\u001b[01;34mbelt\u001b[0m/              \u001b[01;34mfinger\u001b[0m/           \u001b[01;34mocean\u001b[0m/           \u001b[01;34mstereo\u001b[0m/\n",
            "\u001b[01;34mbench\u001b[0m/             \u001b[01;34mfire_hydrant\u001b[0m/     \u001b[01;34moctagon\u001b[0m/         \u001b[01;34mstethoscope\u001b[0m/\n",
            "\u001b[01;34mbicycle\u001b[0m/           \u001b[01;34mfireplace\u001b[0m/        \u001b[01;34moctopus\u001b[0m/         \u001b[01;34mstitches\u001b[0m/\n",
            "\u001b[01;34mbinoculars\u001b[0m/        \u001b[01;34mfiretruck\u001b[0m/        \u001b[01;34monion\u001b[0m/           \u001b[01;34mstop_sign\u001b[0m/\n",
            "\u001b[01;34mbird\u001b[0m/              \u001b[01;34mfish\u001b[0m/             \u001b[01;34moven\u001b[0m/            \u001b[01;34mstove\u001b[0m/\n",
            "\u001b[01;34mbirthday_cake\u001b[0m/     \u001b[01;34mflamingo\u001b[0m/         \u001b[01;34mowl\u001b[0m/             \u001b[01;34mstrawberry\u001b[0m/\n",
            "\u001b[01;34mblackberry\u001b[0m/        \u001b[01;34mflashlight\u001b[0m/       \u001b[01;34mpaintbrush\u001b[0m/      \u001b[01;34mstreetlight\u001b[0m/\n",
            "\u001b[01;34mblueberry\u001b[0m/         \u001b[01;34mflip_flops\u001b[0m/       \u001b[01;34mpaint_can\u001b[0m/       \u001b[01;34mstring_bean\u001b[0m/\n",
            "\u001b[01;34mbook\u001b[0m/              \u001b[01;34mfloor_lamp\u001b[0m/       \u001b[01;34mpalm_tree\u001b[0m/       \u001b[01;34msubmarine\u001b[0m/\n",
            "\u001b[01;34mboomerang\u001b[0m/         \u001b[01;34mflower\u001b[0m/           \u001b[01;34mpanda\u001b[0m/           \u001b[01;34msuitcase\u001b[0m/\n",
            "\u001b[01;34mbottlecap\u001b[0m/         \u001b[01;34mflying_saucer\u001b[0m/    \u001b[01;34mpants\u001b[0m/           \u001b[01;34msun\u001b[0m/\n",
            "\u001b[01;34mbowtie\u001b[0m/            \u001b[01;34mfoot\u001b[0m/             \u001b[01;34mpaper_clip\u001b[0m/      \u001b[01;34mswan\u001b[0m/\n",
            "\u001b[01;34mbracelet\u001b[0m/          \u001b[01;34mfork\u001b[0m/             \u001b[01;34mparachute\u001b[0m/       \u001b[01;34msweater\u001b[0m/\n",
            "\u001b[01;34mbrain\u001b[0m/             \u001b[01;34mfrog\u001b[0m/             \u001b[01;34mparrot\u001b[0m/          \u001b[01;34mswing_set\u001b[0m/\n",
            "\u001b[01;34mbread\u001b[0m/             \u001b[01;34mfrying_pan\u001b[0m/       \u001b[01;34mpassport\u001b[0m/        \u001b[01;34msword\u001b[0m/\n",
            "\u001b[01;34mbridge\u001b[0m/            \u001b[01;34mgarden\u001b[0m/           \u001b[01;34mpeanut\u001b[0m/          \u001b[01;34mtable\u001b[0m/\n",
            "\u001b[01;34mbroccoli\u001b[0m/          \u001b[01;34mgarden_hose\u001b[0m/      \u001b[01;34mpear\u001b[0m/            \u001b[01;34mteapot\u001b[0m/\n",
            "\u001b[01;34mbroom\u001b[0m/             \u001b[01;34mgiraffe\u001b[0m/          \u001b[01;34mpeas\u001b[0m/            \u001b[01;34mteddy-bear\u001b[0m/\n",
            "\u001b[01;34mbucket\u001b[0m/            \u001b[01;34mgoatee\u001b[0m/           \u001b[01;34mpencil\u001b[0m/          \u001b[01;34mtelephone\u001b[0m/\n",
            "\u001b[01;34mbulldozer\u001b[0m/         \u001b[01;34mgolf_club\u001b[0m/        \u001b[01;34mpenguin\u001b[0m/         \u001b[01;34mtelevision\u001b[0m/\n",
            "\u001b[01;34mbus\u001b[0m/               \u001b[01;34mgrapes\u001b[0m/           \u001b[01;34mpiano\u001b[0m/           \u001b[01;34mtennis_racquet\u001b[0m/\n",
            "\u001b[01;34mbush\u001b[0m/              \u001b[01;34mgrass\u001b[0m/            \u001b[01;34mpickup_truck\u001b[0m/    \u001b[01;34mtent\u001b[0m/\n",
            "\u001b[01;34mbutterfly\u001b[0m/         \u001b[01;34mguitar\u001b[0m/           \u001b[01;34mpicture_frame\u001b[0m/   \u001b[01;34mThe_Eiffel_Tower\u001b[0m/\n",
            "\u001b[01;34mcactus\u001b[0m/            \u001b[01;34mhamburger\u001b[0m/        \u001b[01;34mpig\u001b[0m/             \u001b[01;34mThe_Great_Wall_of_China\u001b[0m/\n",
            "\u001b[01;34mcake\u001b[0m/              \u001b[01;34mhammer\u001b[0m/           \u001b[01;34mpillow\u001b[0m/          \u001b[01;34mThe_Mona_Lisa\u001b[0m/\n",
            "\u001b[01;34mcalculator\u001b[0m/        \u001b[01;34mhand\u001b[0m/             \u001b[01;34mpineapple\u001b[0m/       \u001b[01;34mtiger\u001b[0m/\n",
            "\u001b[01;34mcalendar\u001b[0m/          \u001b[01;34mharp\u001b[0m/             \u001b[01;34mpizza\u001b[0m/           \u001b[01;34mtoaster\u001b[0m/\n",
            "\u001b[01;34mcamel\u001b[0m/             \u001b[01;34mhat\u001b[0m/              \u001b[01;34mpliers\u001b[0m/          \u001b[01;34mtoe\u001b[0m/\n",
            "\u001b[01;34mcamera\u001b[0m/            \u001b[01;34mheadphones\u001b[0m/       \u001b[01;34mpolice_car\u001b[0m/      \u001b[01;34mtoilet\u001b[0m/\n",
            "\u001b[01;34mcamouflage\u001b[0m/        \u001b[01;34mhedgehog\u001b[0m/         \u001b[01;34mpond\u001b[0m/            \u001b[01;34mtooth\u001b[0m/\n",
            "\u001b[01;34mcampfire\u001b[0m/          \u001b[01;34mhelicopter\u001b[0m/       \u001b[01;34mpool\u001b[0m/            \u001b[01;34mtoothbrush\u001b[0m/\n",
            "\u001b[01;34mcandle\u001b[0m/            \u001b[01;34mhelmet\u001b[0m/           \u001b[01;34mpopsicle\u001b[0m/        \u001b[01;34mtoothpaste\u001b[0m/\n",
            "\u001b[01;34mcannon\u001b[0m/            \u001b[01;34mhexagon\u001b[0m/          \u001b[01;34mpostcard\u001b[0m/        \u001b[01;34mtornado\u001b[0m/\n",
            "\u001b[01;34mcanoe\u001b[0m/             \u001b[01;34mhockey_puck\u001b[0m/      \u001b[01;34mpotato\u001b[0m/          \u001b[01;34mtractor\u001b[0m/\n",
            "\u001b[01;34mcar\u001b[0m/               \u001b[01;34mhockey_stick\u001b[0m/     \u001b[01;34mpower_outlet\u001b[0m/    \u001b[01;34mtraffic_light\u001b[0m/\n",
            "\u001b[01;34mcarrot\u001b[0m/            \u001b[01;34mhorse\u001b[0m/            \u001b[01;34mpurse\u001b[0m/           \u001b[01;34mtrain\u001b[0m/\n",
            "\u001b[01;34mcastle\u001b[0m/            \u001b[01;34mhospital\u001b[0m/         \u001b[01;34mrabbit\u001b[0m/          \u001b[01;34mtree\u001b[0m/\n",
            "\u001b[01;34mcat\u001b[0m/               \u001b[01;34mhot_air_balloon\u001b[0m/  \u001b[01;34mraccoon\u001b[0m/         \u001b[01;34mtriangle\u001b[0m/\n",
            "\u001b[01;34mceiling_fan\u001b[0m/       \u001b[01;34mhot_dog\u001b[0m/          \u001b[01;34mradio\u001b[0m/           \u001b[01;34mtrombone\u001b[0m/\n",
            "\u001b[01;34mcello\u001b[0m/             \u001b[01;34mhot_tub\u001b[0m/          \u001b[01;34mrain\u001b[0m/            \u001b[01;34mtruck\u001b[0m/\n",
            "\u001b[01;34mcell_phone\u001b[0m/        \u001b[01;34mhourglass\u001b[0m/        \u001b[01;34mrainbow\u001b[0m/         \u001b[01;34mtrumpet\u001b[0m/\n",
            "\u001b[01;34mchair\u001b[0m/             \u001b[01;34mhouse\u001b[0m/            \u001b[01;34mrake\u001b[0m/            \u001b[01;34mt-shirt\u001b[0m/\n",
            "\u001b[01;34mchandelier\u001b[0m/        \u001b[01;34mhouse_plant\u001b[0m/      \u001b[01;34mremote_control\u001b[0m/  \u001b[01;34mumbrella\u001b[0m/\n",
            "\u001b[01;34mchurch\u001b[0m/            \u001b[01;34mhurricane\u001b[0m/        \u001b[01;34mrhinoceros\u001b[0m/      \u001b[01;34munderwear\u001b[0m/\n",
            "\u001b[01;34mcircle\u001b[0m/            \u001b[01;34mice_cream\u001b[0m/        \u001b[01;34mriver\u001b[0m/           \u001b[01;34mvan\u001b[0m/\n",
            "\u001b[01;34mclarinet\u001b[0m/          \u001b[01;34mjacket\u001b[0m/           \u001b[01;34mroller_coaster\u001b[0m/  \u001b[01;34mvase\u001b[0m/\n",
            "\u001b[01;34mclock\u001b[0m/             \u001b[01;34mjail\u001b[0m/             \u001b[01;34mrollerskates\u001b[0m/    \u001b[01;34mviolin\u001b[0m/\n",
            "\u001b[01;34mcloud\u001b[0m/             \u001b[01;34mkangaroo\u001b[0m/         \u001b[01;34msailboat\u001b[0m/        \u001b[01;34mwashing_machine\u001b[0m/\n",
            "\u001b[01;34mcoffee_cup\u001b[0m/        \u001b[01;34mkey\u001b[0m/              \u001b[01;34msandwich\u001b[0m/        \u001b[01;34mwatermelon\u001b[0m/\n",
            "\u001b[01;34mcompass\u001b[0m/           \u001b[01;34mkeyboard\u001b[0m/         \u001b[01;34msaw\u001b[0m/             \u001b[01;34mwaterslide\u001b[0m/\n",
            "\u001b[01;34mcomputer\u001b[0m/          \u001b[01;34mknee\u001b[0m/             \u001b[01;34msaxophone\u001b[0m/       \u001b[01;34mwhale\u001b[0m/\n",
            "\u001b[01;34mcookie\u001b[0m/            \u001b[01;34mladder\u001b[0m/           \u001b[01;34mschool_bus\u001b[0m/      \u001b[01;34mwheel\u001b[0m/\n",
            "\u001b[01;34mcooler\u001b[0m/            \u001b[01;34mlantern\u001b[0m/          \u001b[01;34mscissors\u001b[0m/        \u001b[01;34mwindmill\u001b[0m/\n",
            "\u001b[01;34mcouch\u001b[0m/             \u001b[01;34mlaptop\u001b[0m/           \u001b[01;34mscorpion\u001b[0m/        \u001b[01;34mwine_bottle\u001b[0m/\n",
            "\u001b[01;34mcow\u001b[0m/               \u001b[01;34mleaf\u001b[0m/             \u001b[01;34mscrewdriver\u001b[0m/     \u001b[01;34mwine_glass\u001b[0m/\n",
            "\u001b[01;34mcrab\u001b[0m/              \u001b[01;34mleg\u001b[0m/              \u001b[01;34msea_turtle\u001b[0m/      \u001b[01;34mwristwatch\u001b[0m/\n",
            "\u001b[01;34mcrayon\u001b[0m/            \u001b[01;34mlight_bulb\u001b[0m/       \u001b[01;34msee_saw\u001b[0m/         \u001b[01;34myoga\u001b[0m/\n",
            "\u001b[01;34mcrocodile\u001b[0m/         \u001b[01;34mlighthouse\u001b[0m/       \u001b[01;34mshark\u001b[0m/           \u001b[01;34mzebra\u001b[0m/\n",
            "\u001b[01;34mcrown\u001b[0m/             \u001b[01;34mlightning\u001b[0m/        \u001b[01;34msheep\u001b[0m/           \u001b[01;34mzigzag\u001b[0m/\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "tXc4CF0wBi_y",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "There should be 340 folders for each of the labels.  Each label has 1,000 samples, for a total of 340,000 samples."
      ]
    },
    {
      "metadata": {
        "id": "3yU2H4LJK4Ms",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "name = f'res34-{sz}'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "wjOI3YfEK8Bb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 70
        },
        "outputId": "d7ebe7d3-ed84-4bcd-c91a-c4de9a7c17f4"
      },
      "cell_type": "code",
      "source": [
        "learn = create_cnn(data_bunch, models.resnet34, metrics=[accuracy, map3])\n"
      ],
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading: \"https://download.pytorch.org/models/resnet34-333f7ec4.pth\" to /root/.torch/models/resnet34-333f7ec4.pth\n",
            "100%|██████████| 87306240/87306240 [00:01<00:00, 76844800.42it/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "bU2TSEofHohK",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "[Pythorch Model](https://pytorch.org/docs/stable/torchvision/models.html) using this [paper ](https://arxiv.org/abs/1512.03385)from Microsoft Research."
      ]
    },
    {
      "metadata": {
        "id": "ERF0TZ-kK8EA",
        "colab_type": "code",
        "outputId": "f1807006-6868-4fbe-fc17-8d9986e75105",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 131
        }
      },
      "cell_type": "code",
      "source": [
        "learn.fit_one_cycle(2)"
      ],
      "execution_count": 52,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "Total time: 24:46 <p><table style='width:3375px; margin-bottom:10px'>\n",
              "  <tr>\n",
              "    <th>e</th>\n",
              "    <th>p</th>\n",
              "    <th>o</th>\n",
              "    <th>c</th>\n",
              "    <th>h</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>t</th>\n",
              "    <th>r</th>\n",
              "    <th>a</th>\n",
              "    <th>i</th>\n",
              "    <th>n</th>\n",
              "    <th>_</th>\n",
              "    <th>l</th>\n",
              "    <th>o</th>\n",
              "    <th>s</th>\n",
              "    <th>s</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>v</th>\n",
              "    <th>a</th>\n",
              "    <th>l</th>\n",
              "    <th>i</th>\n",
              "    <th>d</th>\n",
              "    <th>_</th>\n",
              "    <th>l</th>\n",
              "    <th>o</th>\n",
              "    <th>s</th>\n",
              "    <th>s</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>a</th>\n",
              "    <th>c</th>\n",
              "    <th>c</th>\n",
              "    <th>u</th>\n",
              "    <th>r</th>\n",
              "    <th>a</th>\n",
              "    <th>c</th>\n",
              "    <th>y</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>m</th>\n",
              "    <th>a</th>\n",
              "    <th>p</th>\n",
              "    <th>3</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>1</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>9</th>\n",
              "    <th>3</th>\n",
              "    <th>5</th>\n",
              "    <th>1</th>\n",
              "    <th>3</th>\n",
              "    <th>3</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>0</th>\n",
              "    <th>4</th>\n",
              "    <th>7</th>\n",
              "    <th>9</th>\n",
              "    <th>2</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>0</th>\n",
              "    <th>7</th>\n",
              "    <th>1</th>\n",
              "    <th>1</th>\n",
              "    <th>8</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>9</th>\n",
              "    <th>2</th>\n",
              "    <th>2</th>\n",
              "    <th>6</th>\n",
              "    <th>5</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>2</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>6</th>\n",
              "    <th>4</th>\n",
              "    <th>6</th>\n",
              "    <th>3</th>\n",
              "    <th>7</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>4</th>\n",
              "    <th>3</th>\n",
              "    <th>6</th>\n",
              "    <th>4</th>\n",
              "    <th>2</th>\n",
              "    <th>3</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>4</th>\n",
              "    <th>4</th>\n",
              "    <th>8</th>\n",
              "    <th>8</th>\n",
              "    <th>2</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>2</th>\n",
              "    <th>6</th>\n",
              "    <th>8</th>\n",
              "    <th>9</th>\n",
              "    <th>2</th>\n",
              "  </tr>\n",
              "</table>\n"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "yFpqsBBSSk2g",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "learn.save(f'{name}-stage-1')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Sr0M2G5pH7f4",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "The file `res34-128-stage-1.pth` should be now in `models` folder"
      ]
    },
    {
      "metadata": {
        "id": "r6fj8IATSzoP",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "learn.unfreeze()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "sR64TqVwS38G",
        "colab_type": "code",
        "outputId": "1355b5f2-cd2c-40f3-cff4-8718c0f8ca6f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 211
        }
      },
      "cell_type": "code",
      "source": [
        "learn.fit_one_cycle(6, max_lr=6e-4)\n"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "Total time: 1:40:27 <p><table style='width:3375px; margin-bottom:10px'>\n",
              "  <tr>\n",
              "    <th>e</th>\n",
              "    <th>p</th>\n",
              "    <th>o</th>\n",
              "    <th>c</th>\n",
              "    <th>h</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>t</th>\n",
              "    <th>r</th>\n",
              "    <th>a</th>\n",
              "    <th>i</th>\n",
              "    <th>n</th>\n",
              "    <th>_</th>\n",
              "    <th>l</th>\n",
              "    <th>o</th>\n",
              "    <th>s</th>\n",
              "    <th>s</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>v</th>\n",
              "    <th>a</th>\n",
              "    <th>l</th>\n",
              "    <th>i</th>\n",
              "    <th>d</th>\n",
              "    <th>_</th>\n",
              "    <th>l</th>\n",
              "    <th>o</th>\n",
              "    <th>s</th>\n",
              "    <th>s</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>a</th>\n",
              "    <th>c</th>\n",
              "    <th>c</th>\n",
              "    <th>u</th>\n",
              "    <th>r</th>\n",
              "    <th>a</th>\n",
              "    <th>c</th>\n",
              "    <th>y</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>m</th>\n",
              "    <th>a</th>\n",
              "    <th>p</th>\n",
              "    <th>3</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>1</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>4</th>\n",
              "    <th>4</th>\n",
              "    <th>8</th>\n",
              "    <th>8</th>\n",
              "    <th>3</th>\n",
              "    <th>5</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>3</th>\n",
              "    <th>2</th>\n",
              "    <th>9</th>\n",
              "    <th>9</th>\n",
              "    <th>0</th>\n",
              "    <th>0</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>6</th>\n",
              "    <th>4</th>\n",
              "    <th>7</th>\n",
              "    <th>0</th>\n",
              "    <th>6</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>4</th>\n",
              "    <th>6</th>\n",
              "    <th>3</th>\n",
              "    <th>3</th>\n",
              "    <th>3</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>2</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>3</th>\n",
              "    <th>5</th>\n",
              "    <th>8</th>\n",
              "    <th>4</th>\n",
              "    <th>1</th>\n",
              "    <th>0</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>3</th>\n",
              "    <th>0</th>\n",
              "    <th>7</th>\n",
              "    <th>2</th>\n",
              "    <th>7</th>\n",
              "    <th>9</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>7</th>\n",
              "    <th>1</th>\n",
              "    <th>8</th>\n",
              "    <th>2</th>\n",
              "    <th>4</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>5</th>\n",
              "    <th>2</th>\n",
              "    <th>6</th>\n",
              "    <th>6</th>\n",
              "    <th>7</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>3</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>1</th>\n",
              "    <th>5</th>\n",
              "    <th>4</th>\n",
              "    <th>8</th>\n",
              "    <th>8</th>\n",
              "    <th>8</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>1</th>\n",
              "    <th>7</th>\n",
              "    <th>4</th>\n",
              "    <th>4</th>\n",
              "    <th>1</th>\n",
              "    <th>1</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>0</th>\n",
              "    <th>0</th>\n",
              "    <th>8</th>\n",
              "    <th>8</th>\n",
              "    <th>2</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>7</th>\n",
              "    <th>6</th>\n",
              "    <th>6</th>\n",
              "    <th>2</th>\n",
              "    <th>7</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>4</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>9</th>\n",
              "    <th>1</th>\n",
              "    <th>8</th>\n",
              "    <th>9</th>\n",
              "    <th>4</th>\n",
              "    <th>0</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>0</th>\n",
              "    <th>6</th>\n",
              "    <th>6</th>\n",
              "    <th>1</th>\n",
              "    <th>9</th>\n",
              "    <th>1</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>2</th>\n",
              "    <th>9</th>\n",
              "    <th>9</th>\n",
              "    <th>4</th>\n",
              "    <th>1</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>8</th>\n",
              "    <th>0</th>\n",
              "    <th>0</th>\n",
              "    <th>0</th>\n",
              "    <th>0</th>\n",
              "    <th>0</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>5</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>6</th>\n",
              "    <th>4</th>\n",
              "    <th>4</th>\n",
              "    <th>1</th>\n",
              "    <th>0</th>\n",
              "    <th>4</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>0</th>\n",
              "    <th>5</th>\n",
              "    <th>1</th>\n",
              "    <th>9</th>\n",
              "    <th>9</th>\n",
              "    <th>0</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>3</th>\n",
              "    <th>8</th>\n",
              "    <th>7</th>\n",
              "    <th>6</th>\n",
              "    <th>5</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>8</th>\n",
              "    <th>0</th>\n",
              "    <th>7</th>\n",
              "    <th>1</th>\n",
              "    <th>6</th>\n",
              "    <th>7</th>\n",
              "  </tr>\n",
              "  <tr>\n",
              "    <th>6</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>4</th>\n",
              "    <th>5</th>\n",
              "    <th>0</th>\n",
              "    <th>9</th>\n",
              "    <th>1</th>\n",
              "    <th>5</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>1</th>\n",
              "    <th>.</th>\n",
              "    <th>0</th>\n",
              "    <th>7</th>\n",
              "    <th>2</th>\n",
              "    <th>0</th>\n",
              "    <th>9</th>\n",
              "    <th>8</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>7</th>\n",
              "    <th>3</th>\n",
              "    <th>9</th>\n",
              "    <th>6</th>\n",
              "    <th>4</th>\n",
              "    <th>7</th>\n",
              "    <th> </th>\n",
              "    <th> </th>\n",
              "    <th>0</th>\n",
              "    <th>.</th>\n",
              "    <th>8</th>\n",
              "    <th>0</th>\n",
              "    <th>6</th>\n",
              "    <th>9</th>\n",
              "    <th>8</th>\n",
              "    <th>0</th>\n",
              "  </tr>\n",
              "</table>\n"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "m-9x9Zd5UeWe",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "learn.save(f'{name}-stage-2')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "7Pmeq4s2fINg",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "You should see `res34-128-stage-2.pth` in the `models` folder"
      ]
    },
    {
      "metadata": {
        "id": "-yHfsxUQU8og",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Predict"
      ]
    },
    {
      "metadata": {
        "id": "VwW1WMTxVE-X",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "learn.load(f'{name}-stage-2');"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "3XT4hCB8VQQp",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "preds, _ = learn.get_preds(ds_type=DatasetType.Test)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5gtPIiB6V00r",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "PATH1 = Path('subs')\n",
        "PATH1.mkdir(exist_ok=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "1Te83T2-VQTO",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "create_submission(preds, data_bunch.test_dl, name, classes)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Ylu1nyGIWVll",
        "colab_type": "code",
        "outputId": "aca308ec-fc07-44b1-fc5b-6acc6d1314f6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 195
        }
      },
      "cell_type": "code",
      "source": [
        "pd.read_csv(f'subs/{name}.csv.gz').head()"
      ],
      "execution_count": 61,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>key_id</th>\n",
              "      <th>word</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>9307630548749285</td>\n",
              "      <td>grapes blackberry blueberry</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>9702020527625626</td>\n",
              "      <td>watermelon pizza hourglass</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>9893924504621355</td>\n",
              "      <td>computer laptop television</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>9813687402308814</td>\n",
              "      <td>spider fan crab</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>9174447586176505</td>\n",
              "      <td>car police_car school_bus</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "             key_id                         word\n",
              "0  9307630548749285  grapes blackberry blueberry\n",
              "1  9702020527625626   watermelon pizza hourglass\n",
              "2  9893924504621355   computer laptop television\n",
              "3  9813687402308814              spider fan crab\n",
              "4  9174447586176505    car police_car school_bus"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 61
        }
      ]
    },
    {
      "metadata": {
        "id": "3-HjzEHigaME",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "You should see `subs/res34-128.csv.gz` in the files"
      ]
    },
    {
      "metadata": {
        "id": "FuEKfSmwWlyD",
        "colab_type": "code",
        "outputId": "c6a0328f-d4b9-46bb-8c33-dcd6d0464534",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        }
      },
      "cell_type": "code",
      "source": [
        "!kaggle competitions submit -c quickdraw-doodle-recognition -f subs/{name}.csv.gz -m \"Explorers\"\n",
        "\n",
        "\n"
      ],
      "execution_count": 62,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100% 1.52M/1.52M [00:09<00:00, 172kB/s]\n",
            "Successfully submitted to Quick, Draw! Doodle Recognition Challenge"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "ScBlC76spvfa",
        "colab_type": "code",
        "outputId": "31c17c5b-4ed5-4633-a9c3-a1d62a8f540d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "\n",
        "!ls sample_data/txt/test | wc -l"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "112199\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}