flappyBirdsTrain.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "flappyBirdsTrain.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [
        "7PVx00g90FFN",
        "yWri_QCi0f_i"
      ]
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3lq4SQq60Rli",
        "colab_type": "text"
      },
      "source": [
        "# Imports and settings"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OLGYBFtJ36ob",
        "colab_type": "code",
        "outputId": "47876b06-5b41-4278-f731-23729d2c9b1c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "source": [
        "#test if current transfer training works\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "from sklearn import preprocessing\n",
        "from collections import deque\n",
        "import random\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras.models import Sequential, Model\n",
        "from tensorflow.keras.layers import Dense, Dropout, LSTM, CuDNNLSTM, BatchNormalization, Flatten, Bidirectional, Input, RepeatVector, Conv1D, MaxPooling1D, LeakyReLU\n",
        "from tensorflow.keras import layers\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.utils import shuffle\n",
        "\n",
        "\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "#from numpy.random import seed\n",
        "#seed(42)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JPC1zsd937r2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#in frames\n",
        "SEQ_LEN = 100\n",
        "EPOCHS = 100\n",
        "LEARNING_RATE = 0.0001\n",
        "BATCH_SIZE = 2000\n",
        "LOSS =  tf.keras.losses.mse#tf.keras.losses.logcosh\n",
        "LOSS_NAME = \"MSE\"\n",
        "OPT =  tf.keras.optimizers.Adam(lr=LEARNING_RATE, decay=1e-10)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7PVx00g90FFN",
        "colab_type": "text"
      },
      "source": [
        "# Preprocessing functions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "U7b08NnH3-R8",
        "colab_type": "code",
        "cellView": "code",
        "colab": {}
      },
      "source": [
        "max_vals = [600, 1, 18, 500]\n",
        "min_vals = [170, 0, -10, 0]\n",
        "\n",
        "\n",
        "def min_max_scale(_df):\n",
        "    #x1 = (x - min_x) / (max_x - min_x)\n",
        "    for col, max_val, min_val in zip(_df.columns, max_vals, min_vals):\n",
        "        _df[col] = (_df[col] - min_val) / (max_val - min_val)\n",
        "\n",
        "    _df.dropna(inplace=True)\n",
        "\n",
        "    return _df\n",
        "\n",
        "def build_seq(_df):\n",
        "    out_df = _df.copy()\n",
        "\n",
        "    out_df = out_df.drop([\"birdJump\", \"birdVelocity\", \"pipeY\"], axis=1)#drops all the non x and y columns\n",
        "\n",
        "    sequential_data = []\n",
        "    for i in range(len(_df.values) - (SEQ_LEN + SEQ_LEN)):\n",
        "\n",
        "        prev_end = i+SEQ_LEN;\n",
        "\n",
        "        _inp = _df.values[i:prev_end]\n",
        "        _out = out_df.values[prev_end:prev_end+SEQ_LEN]\n",
        "        \n",
        "        sequential_data.append([np.array(_inp), np.array(_out)])\n",
        "\n",
        "    random.shuffle(sequential_data)\n",
        "\n",
        "    X = []\n",
        "    y = []\n",
        "\n",
        "    for seq, target in sequential_data:\n",
        "        X.append(seq)\n",
        "        y.append(target)\n",
        "\n",
        "    return np.array(X), np.array(y)\n",
        "  \n",
        "  \n",
        "def load_and_split_runs(train_runs, test_run):\n",
        "    train_input = np.empty((0, SEQ_LEN, 4))\n",
        "    train_output = np.empty((0, SEQ_LEN, 1))\n",
        "    \n",
        "    \n",
        "    for train_run in train_runs:\n",
        "      _df = pd.read_csv(\"drive/My Drive/Data Sets/\"+train_run+\".csv\")\n",
        "      _df = min_max_scale(_df)\n",
        "      _input, _output = build_seq(_df)\n",
        "      \n",
        "      \n",
        "      train_input = np.concatenate((train_input, _input))\n",
        "      train_output = np.concatenate((train_output, _output))\n",
        "\n",
        "    test_df = pd.read_csv(\"drive/My Drive/Data Sets/\"+test_run+\".csv\")\n",
        "    test_df = min_max_scale(test_df)\n",
        "    test_input, test_output = build_seq(test_df)\n",
        "    \n",
        "    #shuffle\n",
        "    train_input, train_output = shuffle(train_input, train_output)\n",
        "    test_input, test_output = shuffle(test_input, test_output)\n",
        "    \n",
        "    \n",
        "    return train_input, test_input, train_output, test_output"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yWri_QCi0f_i",
        "colab_type": "text"
      },
      "source": [
        "# load and preprocess data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "02te9ATH4RgN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_runs = [\"flappyBird (2)\",\n",
        "             \"flappyBird (3)\",\n",
        "             \"flappyBird (4)\",\n",
        "             \"flappyBird (5)\",\n",
        "             \"flappyBird (6)\",\n",
        "             \"flappyBird (7)\",\n",
        "             \"flappyBird (8)\",\n",
        "             \"flappyBird (9)\",\n",
        "             \"flappyBird (10)\",\n",
        "             \"flappyBird (11)\",\n",
        "             \"flappyBird (12)\",\n",
        "             \"flappyBird (13)\",\n",
        "             \"flappyBird (14)\",\n",
        "             \"flappyBird (15)\",\n",
        "             \"flappyBird (16)\",\n",
        "             \"flappyBird (17)\",\n",
        "             \"flappyBird (18)\",\n",
        "             \"flappyBird (19)\",\n",
        "             \"flappyBird (20)\",\n",
        "             ]\n",
        "\n",
        "train_input, test_input, train_output, test_output = load_and_split_runs(train_runs, \"flappyBird\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-7-4T8cl0lhP",
        "colab_type": "text"
      },
      "source": [
        "# Train model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EztyF-iPSIpZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def plot_prediction(test_input, test_output, pred):\n",
        "  x_in_seq_vals = np.linspace(start=0, stop=SEQ_LEN, num=SEQ_LEN)\n",
        "  x_out_seq_vals = np.linspace(start=SEQ_LEN, stop=2*SEQ_LEN, num=SEQ_LEN)\n",
        "\n",
        "  y_inp = []\n",
        "  y_out = []\n",
        "  y_pred = []\n",
        "\n",
        "\n",
        "  for i in range(SEQ_LEN):\n",
        "      y_inp.append(test_input[0][i][0])\n",
        "      y_out.append(test_output[0][i][0])\n",
        "      y_pred.append(pred[0][i][0])\n",
        "\n",
        "  plt.scatter(x_in_seq_vals, y_inp, color=\"red\", label=\"input\")\n",
        "  plt.scatter(x_out_seq_vals, y_out, color=\"blue\", label=\"output\")\n",
        "  plt.scatter(x_out_seq_vals, y_pred, color=\"green\", label=\"output\")\n",
        "\n",
        "  #plt.scatter(x_vals, main_df[\"birdY\"], color=\"green\")\n",
        "  #plt.scatter(x_vals, main_df[\"pipeY\"], color=\"black\")\n",
        "\n",
        "  plt.show()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pVK8kMUNNeCN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def lstmModel():\n",
        "    inputs = Input(shape=(train_input.shape[1:]))\n",
        "\n",
        "    lstm_1 = LSTM(256, return_sequences=True)(inputs)\n",
        "    lstm_1 = BatchNormalization()(lstm_1)\n",
        "    lstm_1 = Dropout(0.5)(lstm_1)\n",
        "\n",
        "    lstm_2 = LSTM(256, return_sequences=True)(lstm_1)\n",
        "    lstm_2 = BatchNormalization()(lstm_2)\n",
        "    lstm_2 = Dropout(0.5)(lstm_2)\n",
        "    \n",
        "    lstm_3 = LSTM(256, return_sequences=True)(lstm_2)\n",
        "    lstm_3 = BatchNormalization()(lstm_3)\n",
        "    lstm_3 = Dropout(0.5)(lstm_3)\n",
        "   \n",
        "    lstm_4 = LSTM(128, return_sequences=True)(lstm_3)\n",
        "\n",
        "    outputs = Dense(1, activation=\"linear\")(lstm_4)\n",
        "\n",
        "    return Model(inputs=inputs, outputs=outputs)\n",
        "\n",
        "def seq2seqModel():\n",
        "    inputs = Input(shape=(train_input.shape[1:]))\n",
        "\n",
        "    lstm_1 = LSTM(256, return_sequences=True)(inputs)\n",
        "    lstm_1 = LSTM(256, return_sequences=False)(lstm_1)\n",
        "\n",
        "    repeat_vec = RepeatVector(SEQ_LEN)(lstm_1)\n",
        "\n",
        "    lstm_2 = LSTM(256, return_sequences=True)(repeat_vec)\n",
        "    lstm_2 = LSTM(256, return_sequences=True)(lstm_2)\n",
        "    \n",
        "    dense_1 = Dense(128, activation=\"relu\")(lstm_2)\n",
        "    \n",
        "    outputs = Dense(1, activation='linear')(dense_1)\n",
        "    \n",
        "    return Model(inputs=inputs, outputs=outputs) "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "s49Ojuuh4Uiy",
        "colab_type": "code",
        "outputId": "5c41a758-f232-4ec4-b786-57c1376f028d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "model = lstmModel()\n",
        "\n",
        "model.compile(loss=LOSS, \n",
        "                  optimizer=OPT,\n",
        "                  metrics=['mse']) \n",
        "\n",
        "# Train model\n",
        "history = model.fit(\n",
        "    train_input, train_output,\n",
        "    batch_size=BATCH_SIZE,\n",
        "    epochs=EPOCHS,\n",
        "    validation_data=(test_input, test_output)\n",
        ")\n",
        "\n",
        "pred = model.predict(np.array([test_input[0]]))\n",
        "\n",
        "plot_prediction(test_input, test_output, pred)\n",
        "\n",
        "model.save_weights(\"model_1.h5\")"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 21463 samples, validate on 1248 samples\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
            "Epoch 1/100\n",
            "21463/21463 [==============================] - 52s 2ms/sample - loss: 0.1893 - mean_squared_error: 0.1893 - val_loss: 0.0479 - val_mean_squared_error: 0.0479\n",
            "Epoch 2/100\n",
            "21463/21463 [==============================] - 62s 3ms/sample - loss: 0.0551 - mean_squared_error: 0.0551 - val_loss: 0.0325 - val_mean_squared_error: 0.0325\n",
            "Epoch 3/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0466 - mean_squared_error: 0.0466 - val_loss: 0.0302 - val_mean_squared_error: 0.0302\n",
            "Epoch 4/100\n",
            "21463/21463 [==============================] - 42s 2ms/sample - loss: 0.0410 - mean_squared_error: 0.0410 - val_loss: 0.0285 - val_mean_squared_error: 0.0285\n",
            "Epoch 5/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0389 - mean_squared_error: 0.0389 - val_loss: 0.0279 - val_mean_squared_error: 0.0279\n",
            "Epoch 6/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0373 - mean_squared_error: 0.0373 - val_loss: 0.0262 - val_mean_squared_error: 0.0262\n",
            "Epoch 7/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0361 - mean_squared_error: 0.0361 - val_loss: 0.0252 - val_mean_squared_error: 0.0252\n",
            "Epoch 8/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0350 - mean_squared_error: 0.0350 - val_loss: 0.0241 - val_mean_squared_error: 0.0241\n",
            "Epoch 9/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0338 - mean_squared_error: 0.0338 - val_loss: 0.0230 - val_mean_squared_error: 0.0230\n",
            "Epoch 10/100\n",
            "21463/21463 [==============================] - 42s 2ms/sample - loss: 0.0325 - mean_squared_error: 0.0325 - val_loss: 0.0220 - val_mean_squared_error: 0.0220\n",
            "Epoch 11/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0312 - mean_squared_error: 0.0312 - val_loss: 0.0210 - val_mean_squared_error: 0.0210\n",
            "Epoch 12/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0297 - mean_squared_error: 0.0297 - val_loss: 0.0196 - val_mean_squared_error: 0.0196\n",
            "Epoch 13/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0273 - mean_squared_error: 0.0273 - val_loss: 0.0176 - val_mean_squared_error: 0.0176\n",
            "Epoch 14/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0243 - mean_squared_error: 0.0243 - val_loss: 0.0164 - val_mean_squared_error: 0.0164\n",
            "Epoch 15/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0232 - mean_squared_error: 0.0232 - val_loss: 0.0155 - val_mean_squared_error: 0.0155\n",
            "Epoch 16/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0224 - mean_squared_error: 0.0224 - val_loss: 0.0153 - val_mean_squared_error: 0.0153\n",
            "Epoch 17/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0218 - mean_squared_error: 0.0218 - val_loss: 0.0149 - val_mean_squared_error: 0.0149\n",
            "Epoch 18/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0213 - mean_squared_error: 0.0213 - val_loss: 0.0149 - val_mean_squared_error: 0.0149\n",
            "Epoch 19/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0208 - mean_squared_error: 0.0208 - val_loss: 0.0152 - val_mean_squared_error: 0.0152\n",
            "Epoch 20/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0205 - mean_squared_error: 0.0205 - val_loss: 0.0151 - val_mean_squared_error: 0.0151\n",
            "Epoch 21/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0203 - mean_squared_error: 0.0203 - val_loss: 0.0140 - val_mean_squared_error: 0.0140\n",
            "Epoch 22/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0190 - mean_squared_error: 0.0190 - val_loss: 0.0134 - val_mean_squared_error: 0.0134\n",
            "Epoch 23/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0188 - mean_squared_error: 0.0188 - val_loss: 0.0131 - val_mean_squared_error: 0.0131\n",
            "Epoch 24/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0184 - mean_squared_error: 0.0184 - val_loss: 0.0133 - val_mean_squared_error: 0.0133\n",
            "Epoch 25/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0182 - mean_squared_error: 0.0182 - val_loss: 0.0146 - val_mean_squared_error: 0.0146\n",
            "Epoch 26/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0181 - mean_squared_error: 0.0181 - val_loss: 0.0132 - val_mean_squared_error: 0.0132\n",
            "Epoch 27/100\n",
            "21463/21463 [==============================] - 46s 2ms/sample - loss: 0.0175 - mean_squared_error: 0.0175 - val_loss: 0.0129 - val_mean_squared_error: 0.0129\n",
            "Epoch 28/100\n",
            "21463/21463 [==============================] - 44s 2ms/sample - loss: 0.0173 - mean_squared_error: 0.0173 - val_loss: 0.0127 - val_mean_squared_error: 0.0127\n",
            "Epoch 29/100\n",
            "21463/21463 [==============================] - 45s 2ms/sample - loss: 0.0174 - mean_squared_error: 0.0174 - val_loss: 0.0133 - val_mean_squared_error: 0.0133\n",
            "Epoch 30/100\n",
            "21463/21463 [==============================] - 42s 2ms/sample - loss: 0.0172 - mean_squared_error: 0.0172 - val_loss: 0.0140 - val_mean_squared_error: 0.0140\n",
            "Epoch 31/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0170 - mean_squared_error: 0.0170 - val_loss: 0.0125 - val_mean_squared_error: 0.0125\n",
            "Epoch 32/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0166 - mean_squared_error: 0.0166 - val_loss: 0.0132 - val_mean_squared_error: 0.0132\n",
            "Epoch 33/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0162 - mean_squared_error: 0.0162 - val_loss: 0.0119 - val_mean_squared_error: 0.0119\n",
            "Epoch 34/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0159 - mean_squared_error: 0.0159 - val_loss: 0.0115 - val_mean_squared_error: 0.0115\n",
            "Epoch 35/100\n",
            "21463/21463 [==============================] - 47s 2ms/sample - loss: 0.0162 - mean_squared_error: 0.0162 - val_loss: 0.0121 - val_mean_squared_error: 0.0121\n",
            "Epoch 36/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0161 - mean_squared_error: 0.0161 - val_loss: 0.0122 - val_mean_squared_error: 0.0122\n",
            "Epoch 37/100\n",
            "21463/21463 [==============================] - 45s 2ms/sample - loss: 0.0157 - mean_squared_error: 0.0157 - val_loss: 0.0123 - val_mean_squared_error: 0.0123\n",
            "Epoch 38/100\n",
            "21463/21463 [==============================] - 44s 2ms/sample - loss: 0.0154 - mean_squared_error: 0.0154 - val_loss: 0.0120 - val_mean_squared_error: 0.0120\n",
            "Epoch 39/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0155 - mean_squared_error: 0.0155 - val_loss: 0.0115 - val_mean_squared_error: 0.0115\n",
            "Epoch 40/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0150 - mean_squared_error: 0.0150 - val_loss: 0.0112 - val_mean_squared_error: 0.0112\n",
            "Epoch 41/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0160 - mean_squared_error: 0.0160 - val_loss: 0.0123 - val_mean_squared_error: 0.0123\n",
            "Epoch 42/100\n",
            "21463/21463 [==============================] - 43s 2ms/sample - loss: 0.0153 - mean_squared_error: 0.0153 - val_loss: 0.0114 - val_mean_squared_error: 0.0114\n",
            "Epoch 43/100\n",
            "21463/21463 [==============================] - 45s 2ms/sample - loss: 0.0148 - mean_squared_error: 0.0148 - val_loss: 0.0113 - val_mean_squared_error: 0.0113\n",
            "Epoch 44/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0149 - mean_squared_error: 0.0149 - val_loss: 0.0116 - val_mean_squared_error: 0.0116\n",
            "Epoch 45/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0149 - mean_squared_error: 0.0149 - val_loss: 0.0116 - val_mean_squared_error: 0.0116\n",
            "Epoch 46/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0148 - mean_squared_error: 0.0148 - val_loss: 0.0119 - val_mean_squared_error: 0.0119\n",
            "Epoch 47/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0145 - mean_squared_error: 0.0145 - val_loss: 0.0116 - val_mean_squared_error: 0.0116\n",
            "Epoch 48/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0143 - mean_squared_error: 0.0143 - val_loss: 0.0111 - val_mean_squared_error: 0.0111\n",
            "Epoch 49/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0142 - mean_squared_error: 0.0142 - val_loss: 0.0118 - val_mean_squared_error: 0.0118\n",
            "Epoch 50/100\n",
            "21463/21463 [==============================] - 44s 2ms/sample - loss: 0.0141 - mean_squared_error: 0.0141 - val_loss: 0.0119 - val_mean_squared_error: 0.0119\n",
            "Epoch 51/100\n",
            "21463/21463 [==============================] - 44s 2ms/sample - loss: 0.0141 - mean_squared_error: 0.0141 - val_loss: 0.0114 - val_mean_squared_error: 0.0114\n",
            "Epoch 52/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0148 - mean_squared_error: 0.0148 - val_loss: 0.0135 - val_mean_squared_error: 0.0135\n",
            "Epoch 53/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0146 - mean_squared_error: 0.0146 - val_loss: 0.0110 - val_mean_squared_error: 0.0110\n",
            "Epoch 54/100\n",
            "21463/21463 [==============================] - 42s 2ms/sample - loss: 0.0140 - mean_squared_error: 0.0140 - val_loss: 0.0115 - val_mean_squared_error: 0.0115\n",
            "Epoch 55/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0138 - mean_squared_error: 0.0138 - val_loss: 0.0112 - val_mean_squared_error: 0.0112\n",
            "Epoch 56/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0139 - mean_squared_error: 0.0139 - val_loss: 0.0114 - val_mean_squared_error: 0.0114\n",
            "Epoch 57/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0134 - mean_squared_error: 0.0134 - val_loss: 0.0112 - val_mean_squared_error: 0.0112\n",
            "Epoch 58/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0132 - mean_squared_error: 0.0132 - val_loss: 0.0113 - val_mean_squared_error: 0.0113\n",
            "Epoch 59/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0135 - mean_squared_error: 0.0135 - val_loss: 0.0118 - val_mean_squared_error: 0.0118\n",
            "Epoch 60/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0138 - mean_squared_error: 0.0138 - val_loss: 0.0111 - val_mean_squared_error: 0.0111\n",
            "Epoch 61/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0133 - mean_squared_error: 0.0133 - val_loss: 0.0110 - val_mean_squared_error: 0.0110\n",
            "Epoch 62/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0132 - mean_squared_error: 0.0132 - val_loss: 0.0109 - val_mean_squared_error: 0.0109\n",
            "Epoch 63/100\n",
            "21463/21463 [==============================] - 42s 2ms/sample - loss: 0.0133 - mean_squared_error: 0.0133 - val_loss: 0.0115 - val_mean_squared_error: 0.0115\n",
            "Epoch 64/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0128 - mean_squared_error: 0.0128 - val_loss: 0.0113 - val_mean_squared_error: 0.0113\n",
            "Epoch 65/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0131 - mean_squared_error: 0.0131 - val_loss: 0.0156 - val_mean_squared_error: 0.0156\n",
            "Epoch 66/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0143 - mean_squared_error: 0.0143 - val_loss: 0.0113 - val_mean_squared_error: 0.0113\n",
            "Epoch 67/100\n",
            "21463/21463 [==============================] - 42s 2ms/sample - loss: 0.0130 - mean_squared_error: 0.0130 - val_loss: 0.0105 - val_mean_squared_error: 0.0105\n",
            "Epoch 68/100\n",
            "21463/21463 [==============================] - 45s 2ms/sample - loss: 0.0126 - mean_squared_error: 0.0126 - val_loss: 0.0106 - val_mean_squared_error: 0.0106\n",
            "Epoch 69/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0126 - mean_squared_error: 0.0126 - val_loss: 0.0107 - val_mean_squared_error: 0.0107\n",
            "Epoch 70/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0124 - mean_squared_error: 0.0124 - val_loss: 0.0109 - val_mean_squared_error: 0.0109\n",
            "Epoch 71/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0125 - mean_squared_error: 0.0125 - val_loss: 0.0113 - val_mean_squared_error: 0.0113\n",
            "Epoch 72/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0122 - mean_squared_error: 0.0122 - val_loss: 0.0116 - val_mean_squared_error: 0.0116\n",
            "Epoch 73/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0123 - mean_squared_error: 0.0123 - val_loss: 0.0135 - val_mean_squared_error: 0.0135\n",
            "Epoch 74/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0130 - mean_squared_error: 0.0130 - val_loss: 0.0116 - val_mean_squared_error: 0.0116\n",
            "Epoch 75/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0126 - mean_squared_error: 0.0126 - val_loss: 0.0104 - val_mean_squared_error: 0.0104\n",
            "Epoch 76/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0124 - mean_squared_error: 0.0124 - val_loss: 0.0128 - val_mean_squared_error: 0.0128\n",
            "Epoch 77/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0124 - mean_squared_error: 0.0124 - val_loss: 0.0111 - val_mean_squared_error: 0.0111\n",
            "Epoch 78/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0119 - mean_squared_error: 0.0119 - val_loss: 0.0110 - val_mean_squared_error: 0.0110\n",
            "Epoch 79/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0120 - mean_squared_error: 0.0120 - val_loss: 0.0116 - val_mean_squared_error: 0.0116\n",
            "Epoch 80/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0119 - mean_squared_error: 0.0119 - val_loss: 0.0104 - val_mean_squared_error: 0.0104\n",
            "Epoch 81/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0117 - mean_squared_error: 0.0117 - val_loss: 0.0101 - val_mean_squared_error: 0.0101\n",
            "Epoch 82/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0118 - mean_squared_error: 0.0118 - val_loss: 0.0105 - val_mean_squared_error: 0.0105\n",
            "Epoch 83/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0117 - mean_squared_error: 0.0117 - val_loss: 0.0106 - val_mean_squared_error: 0.0106\n",
            "Epoch 84/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0117 - mean_squared_error: 0.0117 - val_loss: 0.0115 - val_mean_squared_error: 0.0115\n",
            "Epoch 85/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0120 - mean_squared_error: 0.0120 - val_loss: 0.0108 - val_mean_squared_error: 0.0108\n",
            "Epoch 86/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0115 - mean_squared_error: 0.0115 - val_loss: 0.0106 - val_mean_squared_error: 0.0106\n",
            "Epoch 87/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0114 - mean_squared_error: 0.0114 - val_loss: 0.0104 - val_mean_squared_error: 0.0104\n",
            "Epoch 88/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0112 - mean_squared_error: 0.0112 - val_loss: 0.0108 - val_mean_squared_error: 0.0108\n",
            "Epoch 89/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0115 - mean_squared_error: 0.0115 - val_loss: 0.0103 - val_mean_squared_error: 0.0103\n",
            "Epoch 90/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0116 - mean_squared_error: 0.0116 - val_loss: 0.0105 - val_mean_squared_error: 0.0105\n",
            "Epoch 91/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0114 - mean_squared_error: 0.0114 - val_loss: 0.0119 - val_mean_squared_error: 0.0119\n",
            "Epoch 92/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0118 - mean_squared_error: 0.0118 - val_loss: 0.0105 - val_mean_squared_error: 0.0105\n",
            "Epoch 93/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0116 - mean_squared_error: 0.0116 - val_loss: 0.0107 - val_mean_squared_error: 0.0107\n",
            "Epoch 94/100\n",
            "21463/21463 [==============================] - 42s 2ms/sample - loss: 0.0114 - mean_squared_error: 0.0114 - val_loss: 0.0107 - val_mean_squared_error: 0.0107\n",
            "Epoch 95/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0112 - mean_squared_error: 0.0112 - val_loss: 0.0103 - val_mean_squared_error: 0.0103\n",
            "Epoch 96/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0112 - mean_squared_error: 0.0112 - val_loss: 0.0105 - val_mean_squared_error: 0.0105\n",
            "Epoch 97/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0110 - mean_squared_error: 0.0110 - val_loss: 0.0104 - val_mean_squared_error: 0.0104\n",
            "Epoch 98/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0108 - mean_squared_error: 0.0108 - val_loss: 0.0107 - val_mean_squared_error: 0.0107\n",
            "Epoch 99/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0111 - mean_squared_error: 0.0111 - val_loss: 0.0105 - val_mean_squared_error: 0.0105\n",
            "Epoch 100/100\n",
            "21463/21463 [==============================] - 41s 2ms/sample - loss: 0.0116 - mean_squared_error: 0.0116 - val_loss: 0.0105 - val_mean_squared_error: 0.0105\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tl7ZQIc83GRT",
        "colab_type": "code",
        "outputId": "2f5cfdca-6919-44a1-8cae-5daf27571d07",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "# summarize history for accuracy\n",
        "plt.plot(history.history['mean_squared_error'])\n",
        "plt.plot(history.history['val_mean_squared_error'])\n",
        "plt.title('model mse')\n",
        "plt.ylabel('mse')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()\n",
        "# summarize history for loss\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3XmcXGWd7/HPr5bel3Q6nX2FsIUt\nQAggiiCCARRQWQX3MXJHr8wd9Qqj4sh1wXFe4jgiigrDJsuAjJkhCIJsypYQIkkIIQtZOmun0/te\nXb/7xzmdVDrV3dXprnSn832/XvXqqrPVc7qgv/k9z6nnmLsjIiKyvyJD3QARETm4KUhERGRAFCQi\nIjIgChIRERkQBYmIiAyIgkRERAZEQSKSRWb2H2b2vQy3XW9mHxzocUQONAWJiIgMiIJEREQGREEi\nh7ywS+nrZvammTWZ2W/NbJyZPWFmDWb2tJmVpWx/sZmtMLNaM3vOzI5JWXeSmS0J93sIyOv2Xh82\ns6Xhvi+Z2Qn72eYvmNkaM9tlZgvMbGK43MzsVjPbYWb1ZrbMzI4L111oZm+FbdtsZl/br1+YSDcK\nEpHAx4HzgCOBjwBPAP8EVBD8f/IVADM7EngA+Idw3ULgv80sx8xygP8C7gVGA/8ZHpdw35OAO4Ev\nAuXAr4AFZpbbn4aa2QeAHwJXABOADcCD4erzgbPC8ygNt6kO1/0W+KK7FwPHAX/uz/uK9ERBIhL4\nd3ff7u6bgReBV939DXdvBR4DTgq3uxJ43N3/5O4dwL8C+cB7gNOBOPBTd+9w90eARSnvMR/4lbu/\n6u6d7n430Bbu1x/XAHe6+xJ3bwNuBM4ws+lAB1AMHA2Yu690963hfh3ALDMrcfcad1/Sz/cVSUtB\nIhLYnvK8Jc3rovD5RIIKAAB3TwKbgEnhus2+90yoG1KeTwO+GnZr1ZpZLTAl3K8/urehkaDqmOTu\nfwZ+DtwG7DCzO8ysJNz048CFwAYze97Mzujn+4qkpSAR6Z8tBIEABGMSBGGwGdgKTAqXdZma8nwT\n8H13H5XyKHD3BwbYhkKCrrLNAO7+M3c/BZhF0MX19XD5Ine/BBhL0AX3cD/fVyQtBYlI/zwMXGRm\n55pZHPgqQffUS8DLQAL4ipnFzexjwNyUfX8NXGdmp4WD4oVmdpGZFfezDQ8AnzWz2eH4yg8IuuLW\nm9mp4fHjQBPQCiTDMZxrzKw07JKrB5ID+D2I7KYgEekHd18FXAv8O7CTYGD+I+7e7u7twMeAzwC7\nCMZTfp+y72LgCwRdTzXAmnDb/rbhaeDbwKMEVdDhwFXh6hKCwKoh6P6qBn4crvsksN7M6oHrCMZa\nRAbMdGMrEREZCFUkIiIyIAoSEREZEAWJiIgMiIJEREQGJDbUDTgQxowZ49OnTx/qZoiIHFRef/31\nne5e0dd2h0SQTJ8+ncWLFw91M0REDipmtqHvrdS1JSIiA6QgERGRAVGQiIjIgBwSYyTpdHR0UFlZ\nSWtr61A3Javy8vKYPHky8Xh8qJsiIiPUIRsklZWVFBcXM336dPaerHXkcHeqq6uprKxkxowZQ90c\nERmhDtmurdbWVsrLy0dsiACYGeXl5SO+6hKRoXXIBgkwokOky6FwjiIytA7pIOlLTVM71Y1tQ90M\nEZFhTUHSi9qWDnY1tWfn2LW1/OIXv+j3fhdeeCG1tbVZaJGIyP5RkPTCgGzdraWnIEkkEr3ut3Dh\nQkaNGpWlVomI9N8he9VWJswgW/f9uuGGG1i7di2zZ88mHo+Tl5dHWVkZb7/9Nu+88w6XXnopmzZt\norW1leuvv5758+cDe6Z7aWxs5IILLuC9730vL730EpMmTeIPf/gD+fn52WmwiEgPFCTAd/97BW9t\nqd9neVsiSWfSKciJ9vuYsyaW8J2PHNvj+ltuuYXly5ezdOlSnnvuOS666CKWL1+++zLdO++8k9Gj\nR9PS0sKpp57Kxz/+ccrLy/c6xurVq3nggQf49a9/zRVXXMGjjz7Ktdde2++2iogMhIJkmJg7d+5e\n3/X42c9+xmOPPQbApk2bWL169T5BMmPGDGbPng3AKaecwvr16w9Ye0VEuihIoMfKobKmmfrWBLMm\nlGS9DYWFhbufP/fcczz99NO8/PLLFBQUcPbZZ6f9Lkhubu7u59FolJaWlqy3U0SkOw2298LI3iBJ\ncXExDQ0NadfV1dVRVlZGQUEBb7/9Nq+88kpW2iAiMhhUkfQim4Pt5eXlnHnmmRx33HHk5+czbty4\n3evmzZvHL3/5S4455hiOOuooTj/99Ow0QkRkEJhn6y/lMDJnzhzvfmOrlStXcswxx/S639a6Fqob\n2zluUmk2m5d1mZyriEh3Zva6u8/pazt1bfXCyF5FIiIyUihIemFmOM6hULWJiOwvBUkvuqY7VIyI\niPRMQdKLrolzVZCIiPQsq0FiZvPMbJWZrTGzG9KsP8vMlphZwswuS1l+jpktTXm0mtml4br/MLN3\nU9bNzlr7w5pEXVsiIj3L2uW/ZhYFbgPOAyqBRWa2wN3fStlsI/AZ4Gup+7r7s8Ds8DijgTXAUymb\nfN3dH8lW27vsrkiy/UYiIgexbFYkc4E17r7O3duBB4FLUjdw9/Xu/iaQ7OU4lwFPuHtz9pragyx2\nbe3vNPIAP/3pT2luPvC/DhGRdLIZJJOATSmvK8Nl/XUV8EC3Zd83szfN7FYzy023k5nNN7PFZra4\nqqpqP952T9dWNmoSBYmIjBTD+pvtZjYBOB54MmXxjcA2IAe4A/gGcHP3fd39jnA9c+bM2a8kiIQ5\nksxCRZI6jfx5553H2LFjefjhh2lra+OjH/0o3/3ud2lqauKKK66gsrKSzs5Ovv3tb7N9+3a2bNnC\nOeecw5gxY3j22WcHv3EiIv2QzSDZDExJeT05XNYfVwCPuXtH1wJ33xo+bTOzu+g2vrJfnrgBti3b\nZ3FRMslhHUlycqJ7BkwyNf54uOCWHlenTiP/1FNP8cgjj/Daa6/h7lx88cW88MILVFVVMXHiRB5/\n/HEgmIOrtLSUn/zkJzz77LOMGTOmf20SEcmCbHZtLQKOMLMZZpZD0EW1oJ/HuJpu3VphlYKZGXAp\nsHwQ2tqrbA+2P/XUUzz11FOcdNJJnHzyybz99tusXr2a448/nj/96U984xvf4MUXX6S09OCeqkVE\nRqasVSTunjCzLxN0S0WBO919hZndDCx29wVmdirwGFAGfMTMvuvuxwKY2XSCiub5boe+38wqCIbC\nlwLXDbixPVQOzS0drK9uYmZFEQW52Sve3J0bb7yRL37xi/usW7JkCQsXLuRb3/oW5557LjfddFPW\n2iEisj+yOkbi7guBhd2W3ZTyfBFBl1e6fdeTZnDe3T8wuK3sWTYv/02dRv5DH/oQ3/72t7nmmmso\nKipi8+bNxONxEokEo0eP5tprr2XUqFH85je/2WtfdW2JyHAwrAfbh1o2v5CYOo38BRdcwCc+8QnO\nOOMMAIqKirjvvvtYs2YNX//614lEIsTjcW6//XYA5s+fz7x585g4caIG20VkyGka+V40tSVYW9XI\njDGFFOfFs9nErNI08iKyPzSN/CDQXFsiIn1TkPRCs/+KiPTtkA6Svrr1zA7+SRsP5raLyMHhkA2S\nvLw8qqure/1De7BXJO5OdXU1eXl5Q90UERnBDtmrtiZPnkxlZSW9zcOVSCbZXtdG+844hVn8Hkk2\n5eXlMXly2iusRUQGxcH513EQxONxZsyY0es22+tb+ci9z/CDjx7PJ2ZPPUAtExE5uByyXVuZiIWz\nNnZ09jbLvYjIoU1B0otYNPj1KEhERHqmIOlFThgkiWzMIy8iMkIoSHoRi4ZdWwlVJCIiPVGQ9GL3\nGIkqEhGRHilIemFmxKNGQmMkIiI9UpD0IRaJaLBdRKQXCpI+xKJGR6e6tkREeqIg6UNONEIiqYpE\nRKQnCpI+xKJGR0IViYhIT7IaJGY2z8xWmdkaM7shzfqzzGyJmSXM7LJu6zrNbGn4WJCyfIaZvRoe\n8yEzy8nmOcQiETpUkYiI9ChrQWJmUeA24AJgFnC1mc3qttlG4DPA79IcosXdZ4ePi1OW/wi41d1n\nAjXA5we98SlyYhESGiMREelRNiuSucAad1/n7u3Ag8AlqRu4+3p3fxPI6J/8Ftwg5APAI+Giu4FL\nB6/J+4pFTFdtiYj0IptBMgnYlPK6MlyWqTwzW2xmr5hZV1iUA7XuntjPY/ZbLBrRVVsiIr0YztPI\nT3P3zWZ2GPBnM1sG1GW6s5nNB+YDTJ26/1PAx6Omq7ZERHqRzYpkMzAl5fXkcFlG3H1z+HMd8Bxw\nElANjDKzrgDs8Zjufoe7z3H3ORUVFf1vfSge1RcSRUR6k80gWQQcEV5llQNcBSzoYx8AzKzMzHLD\n52OAM4G3PLgv7rNA1xVenwb+MOgtTxGMkahrS0SkJ1kLknAc48vAk8BK4GF3X2FmN5vZxQBmdqqZ\nVQKXA78ysxXh7scAi83sbwTBcYu7vxWu+wbwj2a2hmDM5LfZOgcIKhLNtSUi0rOsjpG4+0JgYbdl\nN6U8X0TQPdV9v5eA43s45jqCK8IOiGCMRBWJiEhP9M32PsSiEdp1PxIRkR4pSPqgikREpHcKkj5o\njEREpHcKkj4E9yNRRSIi0hMFSR/iUU2RIiLSGwVJH+LRiMZIRER6oSDpQ3A/ElUkIiI9UZD0IR7V\n/UhERHqjIOlDPGq6H4mISC8UJH2IRYIxkmCaLxER6U5B0od41AB0CbCISA8UJH2IR4Nfke5JIiKS\nnoKkD7EwSDoSqkhERNJRkPRhd9eWKhIRkbQUJH3Y3bWlMRIRkbQUJH2IRboG21WRiIikoyDpQ1dF\noiAREUlPQdKHWDhGovm2RETSU5D0QRWJiEjvshokZjbPzFaZ2RozuyHN+rPMbImZJczsspTls83s\nZTNbYWZvmtmVKev+w8zeNbOl4WN2Ns9BX0gUEeldLFsHNrMocBtwHlAJLDKzBe7+VspmG4HPAF/r\ntnsz8Cl3X21mE4HXzexJd68N13/d3R/JVttTxSJdV22pIhERSSdrQQLMBda4+zoAM3sQuATYHSTu\nvj5ct9dfaXd/J+X5FjPbAVQAtRxge7q2VJGIiKSTza6tScCmlNeV4bJ+MbO5QA6wNmXx98Mur1vN\nLLeH/eab2WIzW1xVVdXft91tT9eWKhIRkXSG9WC7mU0A7gU+6+5df8lvBI4GTgVGA99It6+73+Hu\nc9x9TkVFxX63Iaa5tkREepXNINkMTEl5PTlclhEzKwEeB77p7q90LXf3rR5oA+4i6ELLGg22i4j0\nLptBsgg4wsxmmFkOcBWwIJMdw+0fA+7pPqgeVimYmQGXAssHtdXd6PJfEZHeZS1I3D0BfBl4ElgJ\nPOzuK8zsZjO7GMDMTjWzSuBy4FdmtiLc/QrgLOAzaS7zvd/MlgHLgDHA97J1DrBnihTNtSUikl42\nr9rC3RcCC7stuynl+SKCLq/u+90H3NfDMT8wyM3slSoSEZHeDevB9uFAl/+KiPROQdKHPXNtqSIR\nEUlHQdIHVSQiIr1TkPRBX0gUEemdgqQPmmtLRKR3CpI+6AuJIiK9U5D0wcyIRUxdWyIiPVCQZCAW\nNd0hUUSkBwqSDMSjEVUkIiI9UJBkIB6NaIoUEZEeKEgyoDESEZGeKUgyEHRtqSIREUlHQZKBeNQ0\nRYqISA8UJBmIabBdRKRHCpIMBGMk6toSEUlHQZKBnFhEU6SIiPQgoyAxs+vNrMQCvzWzJWZ2frYb\nN1yoIhER6VmmFcnn3L0eOB8oAz4J3JK1Vg0zGiMREelZpkFi4c8LgXvdfUXKshEvJxrRFCkiIj3I\nNEheN7OnCILkSTMrBvr8J7qZzTOzVWa2xsxuSLP+rLCbLGFml3Vb92kzWx0+Pp2y/BQzWxYe82dm\nlvVAi0X1hUQRkZ5kGiSfB24ATnX3ZiAOfLa3HcwsCtwGXADMAq42s1ndNtsIfAb4Xbd9RwPfAU4D\n5gLfMbOycPXtwBeAI8LHvAzPYb/FIvpCoohITzINkjOAVe5ea2bXAt8C6vrYZy6wxt3XuXs78CBw\nSeoG7r7e3d9k3+rmQ8Cf3H2Xu9cAfwLmmdkEoMTdX3F3B+4BLs3wHPZbTsx01ZaISA8yDZLbgWYz\nOxH4KrCW4I94byYBm1JeV4bLMtHTvpPC530e08zmm9liM1tcVVWV4dumF1QkChIRkXQyDZJEWAFc\nAvzc3W8DirPXrIFz9zvcfY67z6moqBjQsYIxEnVtiYikk2mQNJjZjQSX/T5uZhGCcZLebAampLye\nHC7LRE/7bg6f788x91tw1ZYqEhGRdDINkiuBNoLvk2wj+AP+4z72WQQcYWYzzCwHuApYkOH7PQmc\nb2Zl4SD7+cCT7r4VqDez08OrtT4F/CHDY+43VSQiIj3LKEjC8LgfKDWzDwOt7t7rGIm7J4AvE4TC\nSuBhd19hZjeb2cUAZnaqmVUClwO/MrMV4b67gP9HEEaLgJvDZQB/D/wGWEMwVvNEf054f2iMRESk\nZ7FMNjKzKwgqkOcIvoj472b2dXd/pLf93H0hsLDbsptSni9i766q1O3uBO5Ms3wxcFwm7R4swVxb\nqkhERNLJKEiAbxJ8h2QHgJlVAE8DvQbJSBGL6H4kIiI9yXSMJNIVIqHqfux70IuFd0gMLlwTEZFU\nmVYkfzSzJ4EHwtdX0q3LaiTLiQazsCSSTjx6yEwxJiKSkYyCxN2/bmYfB84MF93h7o9lr1nDSywa\nFF+JTiceHeLGiIgMM5lWJLj7o8CjWWzLsBWLBFVIe2eSfJQkIiKpeg0SM2sA0g0MGODuXpKVVg0z\nObGuikQD7iIi3fUaJO4+rKdBOVBikTBIdE8SEZF9HDJXXg1ELBxgb0+oIhER6U5BkoF4ylVbIiKy\nNwVJBuJRjZGIiPREQZKBrjGSdgWJiMg+FCQZ2N21pfm2RET2oSDJwO6uLc23JSKyDwVJBvZctaWK\nRESkOwVJBlSRiIj0TEGSgXjKXFsiIrI3BUkGUufaEhGRvSlIMqCKRESkZ1kNEjObZ2arzGyNmd2Q\nZn2umT0Urn/VzKaHy68xs6Upj6SZzQ7XPRces2vd2GyeA6R+s10ViYhId1kLEjOLArcBFwCzgKvN\nbFa3zT4P1Lj7TOBW4EcA7n6/u89299nAJ4F33X1pyn7XdK3vdufGrOiqSDpUkYiI7CObFclcYI27\nr3P3duBB4JJu21wC3B0+fwQ418y634Lw6nDfIdN1+W+HxkhERPaRzSCZBGxKeV0ZLku7jbsngDqg\nvNs2V7LnFr9d7gq7tb6dJngAMLP5ZrbYzBZXVVXt7zkAmmtLRKQ3w3qw3cxOA5rdfXnK4mvc/Xjg\nfeHjk+n2dfc73H2Ou8+pqKgYUDviEXVtiYj0JJtBshmYkvJ6crgs7TZmFgNKgeqU9VfRrRpx983h\nzwbgdwRdaFmlri0RkZ5lM0gWAUeY2QwzyyEIhQXdtlkAfDp8fhnwZ3d3ADOLAFeQMj5iZjEzGxM+\njwMfBpaTZXu+2a6KRESku15vtTsQ7p4wsy8DTwJR4E53X2FmNwOL3X0B8FvgXjNbA+wiCJsuZwGb\n3H1dyrJc4MkwRKLA08Cvs3UOXeKqSEREepS1IAFw94XAwm7Lbkp53gpc3sO+zwGnd1vWBJwy6A3t\ng5kRjZiCREQkjWE92D6cxKOmb7aLiKShIMlQPBLRVVsiImkoSDIUi6prS0QkHQVJhmLRiObaEhFJ\nQ0GSoZyourZERNJRkGRIXVsiIukpSDIUi+iqLRGRdBQkGYpHI6pIRETSUJBkSEEiIpKegiRDsahp\nri0RkTQUJBlSRSIikp6CJEOaIkVEJD0FSYZiEVUkIiLpKEgyFNcXEkVE0lKQZCgeNU2RIiKShoIk\nQzFVJCIiaSlIMlRRlMvWuhaa2xND3RQRkWFFQZKh82aNo7UjyfOrqoa6KSIiw4qCJENzZ4ymvDCH\nhcu3DXVTRESGlawGiZnNM7NVZrbGzG5Isz7XzB4K179qZtPD5dPNrMXMloaPX6bsc4qZLQv3+ZmZ\nWTbPoUs0Ypx/7Hj+vHI7rR2dB+ItRUQOClkLEjOLArcBFwCzgKvNbFa3zT4P1Lj7TOBW4Ecp69a6\n++zwcV3K8tuBLwBHhI952TqH7i48fjxN7Z28uHrngXpLEZFhL5sVyVxgjbuvc/d24EHgkm7bXALc\nHT5/BDi3twrDzCYAJe7+irs7cA9w6eA3Pb3TDyunND/OE8u2Hqi3FBEZ9rIZJJOATSmvK8Nlabdx\n9wRQB5SH62aY2Rtm9ryZvS9l+8o+jgmAmc03s8VmtriqanAGyOPRCOfPGsefVm6nPaHvlIiIwPAd\nbN8KTHX3k4B/BH5nZiX9OYC73+Huc9x9TkVFxf614vW74aWf77XowuMn0NCa4K9r1b0lIgLZDZLN\nwJSU15PDZWm3MbMYUApUu3ubu1cDuPvrwFrgyHD7yX0cc/CsfQYW/WavRe+ZWU5xbkzdWyIioWwG\nySLgCDObYWY5wFXAgm7bLAA+HT6/DPizu7uZVYSD9ZjZYQSD6uvcfStQb2anh2MpnwL+kLUzGHc8\n1LwLbY27F+XGonxw1jj+uHwbjW36cqKISNaCJBzz+DLwJLASeNjdV5jZzWZ2cbjZb4FyM1tD0IXV\ndYnwWcCbZraUYBD+OnffFa77e+A3wBqCSuWJbJ0D444Nfu54a6/FnzpjGvWtCe57ZUPW3lpE5GAR\ny+bB3X0hsLDbsptSnrcCl6fZ71Hg0R6OuRg4bnBb2oOuINm+HKbM3b34pKllnHVkBb9+YR2fOmMa\nBTlZ/TWKiAxrw3WwfXgYNRVyS2D7in1WXX/uEVQ3tasqEZFDnoKkN2ZBVZImSE6ZVsZ7Z47hjhfW\n0dKub7qLyKFLQdKXriDxfaeQv/6DR7CzsZ37X1VVIiKHLgVJX8YdC231ULtxn1WnTh/New4v5xfP\nreWNjTVD0DgRkaGnIOnLuHBcP033FsA/X3ws+fEol//yZX75/FqSSd38SkQOLQqSvowN55nsIUiO\nHFfMwuvfx/nHjuOWJ97m03e9xvqdTQewgSIiQ0tB0pfcIiibEVwC3IPS/Di3feJkfvDR41myoYbz\nbn2eHy5cSUNrxwFsqIjI0FCQZGLcsb0GCYCZ8YnTpvLs187m0tmTuOPFdZz94+f46dPvUNXQdoAa\nKiJy4ClIMjH+eKheC+3NfW46tiSPH19+Igu+9F5OmFzKT59ezZm3/Jl/fHgpr2+owdNc/SUicjDT\nV7IzMe5YwKFqJUw6JaNdjp9cyl2fncu6qkbueXkD/7l4E79fspmjxxdz9dypfPTkSZTkxbPbbhGR\nA0AVSSZ2T5WSfsC9N4dVFPHPFx/Lq9/8ID/82PHkxCJ8Z8EKTv/BM3zrv5axenvDIDdWROTAUkWS\niVHTIadov4KkS1FujKvnTuXquVN5s7KWe17ewMOLK7nvlY2cdWQF173/MM44rJwDdAt6EZFBY4dC\nn/2cOXN88eLFAzvIb84Lpkz53JPBz0Gwq6mdB17byF1/Xc/OxjZOmFzKde8/nA8dO55oRIEiIkPL\nzF539zl9baeurUzN/CBsehUeuw46WgblkKMLc/jSOTP5yzfO4QcfPZ76lg7+/v4lnPeT53l40Sbd\nzldEDgqqSDLlDi/8Kzz7PZh4Mlx1P5RMHJwGhjqTzhPLt3L7c2tZsaWeKaPz+ep5R3HxiROJqEIR\nkQMs04pEQdJfbz8Ov58PsVz40A/hhCsGrauri7vz3KoqfvzkKt7aWs/R44v53qXHMWf66EF9HxGR\n3qhrK1uOvgi+8GcYfRg8Nh/uuQSq3hnUtzAzzjl6LP/zv9/Lz64+ica2BJ/49ass+NuWQX0fEZHB\noIpkfyWT8Ppd8PR3oa0uuAnW5Lkw431w4tVBxTJI6po7+MK9i3nt3V18Y97RXPf+w3R1l4hk3bCo\nSMxsnpmtMrM1ZnZDmvW5ZvZQuP5VM5seLj/PzF43s2Xhzw+k7PNceMyl4WNsNs+hR5EInPp5+PIi\nOP/7MPEk2PAS/Pf18IvT4Z0nB+2tSgvi3Pv5uXzkxIn86I9v873HV+ob8iIybGTteyRmFgVuA84D\nKoFFZrbA3d9K2ezzQI27zzSzq4AfAVcCO4GPuPsWMzsOeBKYlLLfNeG924de8Th4z5f3vF7zDPzx\nBvjdFXD4uXDaF4Of0YH9qnNjUf7tytmUF+bw27+8S2l+nK+ce8QAGy8iMnDZrEjmAmvcfZ27twMP\nApd02+YS4O7w+SPAuWZm7v6Gu3cNCKwA8s1s8PqKsmnmuXDdX4MqZcsbQaD85Bh48puwrfeJH/sS\niRg3fXgWHzt5Ej/50zvcq/vFi8gwkM1vtk8CNqW8rgRO62kbd0+YWR1QTlCRdPk4sMTdU6fQvcvM\nOoFHge95mn4eM5sPzAeYOnXqAE+ln2I5QZUydz6sfgr+9gC8+it4+ecw/gSYfU1wtVdB/6/CikSM\nH338BOqaO7jpD8tJdCa58tQpFORokgIRGRpZG2w3s8uAee7+d+HrTwKnufuXU7ZZHm5TGb5eG26z\nM3x9LLAAON/d14bLJrn7ZjMrJgiS+9z9nt7akpXB9v5qqoblj8DS+2Hr3yCWB8d+DOZ8Fiaf2u9L\niFs7OvnMXa/xyrpdFOZEmXfcBC49aSKnH1ZOPKqL8URk4DIdbM/mP2M3A1NSXk8Ol6XbptLMYkAp\nUA1gZpOBx4BPdYUIgLtvDn82mNnvCLrQeg2SYaGwPBgvOe2LsG0ZLL4L3nwI/vY7qDg6qFJOvAqK\nMrt2IC8e5Xd/dzqL1u/isTc28/ibW3l0SSUleTHOPWYcZx9VwYmTRzGtvEBXeIlIVmWzIokB7wDn\nEgTGIuAT7r4iZZsvAce7+3XhYPvH3P0KMxsFPA98191/3+2Yo9x9p5nFgQeAp939l721ZVhUJOm0\nNcDyR+GN+6ByEVg0GGM54Uo46gLIKcz4UK0dnby4eidPrtjG0yu3U9sc3J2xJC/GrIklHD2+hCPH\nFXPU+CKOGl9CUa66wkSkd8Pim+1mdiHwUyAK3Onu3zezm4HF7r7AzPKAe4GTgF3AVe6+zsy+BdwI\nrE453PlAE/ACEA+P+TTwj+7AlbUNAAATjElEQVTe2Vs7hm2QpKpaFXR7LXsE6jdDvDCoUM78CpRN\n79ehEp1J3t7WwPLNdSzbXMeKLfWs3t5AU/ueX9O08gKOHFfMlLICJo7KY+KofMoKchhVEKe8KIex\nxXmDfIIicrAZFkEyXBwUQdIlmYSNL8HSB2DZw5DshOM+BmffCOWHD+CwzubaFlZta2Dl1npWbqvn\nne2NbK5poaVj3xx+z+HlfOGswzj7yAp1jYkcohQkKQ6qIElVvwVevi0YTwH48K1w4pWD+hbuTm1z\nB1vrWqltaaeuuYN1O5u49+UNbKtv5YixRZwweRRjS3IZV5zL5LICppUXMGV0AXnx6KC2RUSGFwVJ\nioM2SLrUbYZH/y6oVE66Fi74l36Nn+yP9kSS/3lzCw8u2kTlrmZ2NLSRSO7930pJXozSgjij8nMo\nyo1RkBOlIDfG5LJ8TphUynGTSplclq+KRuQgpSBJcdAHCUBnAp6/JZjKvnAMnPwpOOWzMGpK3/sO\ngmTSqW5qZ1NNMxurm9m4q5ldTe3UNrdT29JBU1uCprZOmtsTVNa07A6deNQozQ/GXnJjEdoTSTo6\nk0Qixqj8OKMKchiVH6ckP05pfjA+M3V0AdPKC5lclq9LmUWGkIIkxYgIki4bXoaXfgbv/DF4PfOD\nMOsSOOrC/fqCYza0dnSyalsDb26uY3NNC3Ut7dQ2d9CeSJITi5ATi5DodOpaOqhtaaemqYP6lg4a\n2hJ7HScaMaaU5TNjTCHTyguZNCqf8aV5jC7MYWdjG9vqWtnR0IYBObEIefEoU0bnc8TYYg6vKCI/\nJ7OuN3dn1fYGmts7mTWhRF12IiEFSYoRFSRdajcGYyfLHwmeR2Iw7jgoKIf8suD7KGXTg0fROLDw\nX/adHdC0Axp3QEtNsDwaD+5Jf+Q8KKoYslNKdCapbmpnQ3UzG6qb2FDdzLvVTbxb1cT66iaa2/e9\nKKAgDIv2RHKfrrfywhwqinMZW5KHAS0dnbR2dDK6MIejxhVzxLhiNlQ38fibW1m3swkIKqhZE0s5\ncXIpsyaUMGtiCRNK82nt6KS5vZP2RBKzIOTiUaMgJ+jSK8yNjfzqKdEOj34OjrwATrpmqFsjB4CC\nJMWIDJIu7rB1Kbz1h+CLji010LwLGrdDR3P/jhWJw9EXwkmfgunvhXgflwC7w4a/wut3B11sp38p\n+OJlfyTaoLMdcov7eCunoS3BtrpWqhvbGVOUw/jSPIrz4ru3aU8k2biridXbG1m9o5Gtda1UNbRS\n1RDMrpMXj5IXj7KjoY21Oxpp70wSMTjj8HIuPH4C5YW5LN1Uy5KNNazYXLfX5dKZKM2PM6YoCK+K\n4jzGFudSUZxLYU7wvvk5UfJiUXLjEXJjUZLudHQmae1IsnJrPYs31PDGxhraE0lK8uMU58WYXl7I\nyVNHcfLUMmaOLaIkP75XxdTRmSRiRvRA3EHz2R8G3auxPJo/9zwLKgsYXZjDB48Zpzt4jlAKkhQj\nOkh64g5NVVCzPqg+ukRiQdVRNC6oXDwJyURwhdgb9wfzgrXsgmhOMDX+lLnBTbyKJ0DhWOhsg5Za\naNgKb9wbTEyZWwpt9RAvCKbWP/bSYNuisfvel6WlFja/DhtfCabdr1wULD/9Onjv/wnaBFC/Feoq\nYdIpwZT9gyzRmWTDrubwj/++84Emk87GXc2s3FpPVWMb+fGg6siJRuh0J5l0OpJOS3swNtTQmqC6\nqY2djW1UNQSPHQ1taauodMzgmIp8vlT4DLWlx7A85wTqWjpYta2BtVVNe22bG4sQj0Zo7egkkXRy\nYhGOGlfMrAklzBxbRFlhMO4Uj0XYXt/KjvpWttW3sq2ujW31LdS1dDB5VAGHjy1kxpgiygrilOTF\nKcyN0d6ZpLktQUtHJxNK8zlyXBHlRbmwbTl+x/tpmXoOtulVVnZO4LLWb5MkwmEVhVx31uFcPHvi\nwLsFq1YF/yia83koLKehtYOi3Jgu2BgiCpIUh2SQ7K9EWzAV/saXYOOrQVAkO9JvO/rwYHLKE6+G\nmg3w4r8G39T35J5tckuDKqWwIgiRnauC5RaBCSfC1PdAc3UwXUxeKRzzYah8HapWBttVHA3v+Qoc\nf3kwGWZ3Ha1BFdZWD631QXXjnUEbCsqhdDLkjep7LrOa9UG335jBnZq/qS1Bc3vQpdbS0UlbR5LW\nRPAzEoGcaBAKM3LrKfmf+bDx5WCGg4/8NLigAqhtbueNTbVU1rRQ3xKMJ3V0Ovk5EfLjUepbE7y1\npZ4VW+qoaU7/WZUVxBlfms/4klyK8+JsqmlmzY5GGloTabdPNSY/wt3+T4z1nZzX9mM+EF3KT+K3\ns2nuTSyddDW3P7eWt7bWYwbjS/KYOrqA8qKc3ZVSfjzK6MIcRhfmBGGVSO7uIiwvyqGiKI/SXGP0\nm79iwpJbiSTbaYiN5nvR/8VDdcdSmBNlWnkh08oLGFeSx9G+jnPf/Re2jD+Pd2Z8Eo/GmDQqn5lj\nixhbnKvQGUQKkhQKkgHoTARjKg1bg8omlhv8Yc4vg1HT9q0WatbD9rfCcZiqoCpq3hn8jBfC5FOC\nSSonngx5JXv227YMnv7noEqZchocfk4QBK/8ErYvC4Jo/PFQPjOopqpWBV16O1cDffw3nFMUfJlz\n7KzgUTYtOEZhBVQuDiqr9S8G2445Eo65GKadEVRY0dwgnOoqoW5j0G0YzQke3hmMT9WsD8KweMKe\ncakxRwQhWDIxCON3X4TK14IwTbQGj7LpQdVXMgmeuwU6WuCCH8GK38PaP8P7vxFUaVWrYMdbQTtK\nJgWPnAJoa4T2xiA080rx3BIaOuM01NfR1FiPN1ZR0bKWkvp3iLVUw9TT4LBzgrG0SAR3p6a5g7qW\nDhpaO2hsS5AbgbKmNRRVL6O6Fda3FJC35WXO2XEvTxx9CzUzLuLMw0cz7cnPwbsvwHUv4uUz+eua\nal7fUMOGXU1srG6mtqWDZNLpdKe5vZNdTe107h7DcsZRw4zINsZSwzir4aLoK8yOrOOJzlO5p/N8\n/jl+H0fZBlaOv5gnx3yavzWUsGFXM8c2/JUf+b/hQKG1sSw5nRs6vsAKnwFAcV6MkmiCwzvXcGRy\nHZtsIm/ETyIvJ05RbozS8OrAeCxCR3gFoVlwv5/ceIR4JEIymWBa0zIKW7bwl8QxLGsooqWjk8Mr\nijh6fPHuLsaSvBiFueEjJ0Y8alQ3tbO1LqgCO1PG7PJixuE1L3DEpkfZkX8YT5VczqrGoOu4qxuz\na0yvojiXsoIcSsOrGXPCtiaSTjS82jHWy3hca0cnq7c38tbWOt7aUs//nXc0hfs5JZKCJIWC5CDi\nvnf14B5USG8+GIRG9Vpobwj+aE88KZiWv3hcUM3klgQXDlg0OEbTzjAAKoNKaMfKIBC7K5sRfD8n\ntwRWLgjGfVKrqlSx/KArMNkRVFUlk4JAKBgddA/WbAhCtDuLBH/Ai8cHMz9H48H57HgrOF7F0XDF\nPVBxVFAZ/fc/wNL7AKPPoOxLfhnkj4Zd4dynuSV7flex3KA98fzgvbYvD6q77o7+MFx5357Ppn4L\n3HZ6sO3kOXDE+UFYJ9qDwOtoCUKuvQnaG/G2Rjpa6qFhO7Fdq4h0e4/WvArWnPwt6mdcRFF+nGPG\n5hF/8V/gLz8NPouZH4Txx8Ff/43k+BPZ+ZF7iG5+hVHPfpNIy046cstpJ0Z70iht306UPV2Ku3Im\n8GLpxSyKz2Vdopyq1giJziRjI3UczibKOmuwzjYs2c6Mzg28P/kq5dTu3n977jQ2FM1mS1seG5ui\n7GjPpc4LqaGYVo8zyXYyzXYwzmrY4uWs8Yms9/HkkGCUNTLFqvhM9I8cGdnMNi+jglraifNE7odY\nH51BXvsuChK1we8tVO0lrPIpvO1TqfcCKqyOsVZDAW20EyeWm09ZrjE1XsukSA1F3kRVIo+tbXls\nacuhJRmnnRjRWA5f+8KnOGrKhP36T0dBkkJBMoK4B3+ccov2b//mXcFcZo3boWE7jJoK087cu7Jq\n2hn8ke9sC7r6IlEonRJ0k3V9ETSZBDxY111bI1SvDiqJ2k3BH8CpZ0D+qH237WgN/sCPPnzvixvc\nYcndQQiOOzYIoVhe0Pa6yqCiyS0Oqi2zoFuvtTZob7wgqFjyRgUVWPH4YJv6rbDuOdi8OHjfzraw\nOmoL/vAnEzD2GJhyehAO7kE12dYA09+378UX1Wth2X8G99zZvIS0gReJBb+znOLgMysYA2OPDoKz\nfGYQxMXjg3NJ1yVVVwlL7gku6GjcBkddBB//9Z7PoaUGXrkdGrYFAZzsCCrlSafAhBNg02uw6Lew\n4S97jlk4Nvij3Vq77/vFC4JQnHVJ0L53X4C1zwRVZVtD8DvqQXu8lJyOurTrEmOOoe6U/82u6RdS\n3rGVsiX/jv3toaCqBYjl4/E83CGZTBJtr8f68Q+INnLIpT39yi8tgoojMz5WKgVJCgWJSJY17Qwe\nu6uc/CA4ojn9vtdOWp0dsPVNmDg7fXj3Zedq2LIUatcHVWMkFoRmxVFQMjlsc25Q2Xa/QKSLexC6\nrXVBgLXUBP+oKZ0cdJfG84Ow2bkaat4Nfgf5ZUEXbfnMfbuBG3cEAV44Zt+ZKtoaoertoEJsbw6q\n7qLxwXad7cE/ACwSdJ0WTwxCPhFeCNNWH14N2RZUiBNnhxVn/ylIUihIRET6L9MgGeHfoBIRkWxT\nkIiIyIAoSEREZEAUJCIiMiAKEhERGRAFiYiIDIiCREREBkRBIiIiA3JIfCHRzKqADfu5+xhg5yA2\n52BxKJ73oXjOcGiet845M9Pcvc+73R0SQTIQZrY4k292jjSH4nkfiucMh+Z565wHl7q2RERkQBQk\nIiIyIAqSvt0x1A0YIofieR+K5wyH5nnrnAeRxkhERGRAVJGIiMiAKEhERGRAFCS9MLN5ZrbKzNaY\n2Q1D3Z5sMLMpZvasmb1lZivM7Ppw+Wgz+5OZrQ5/lg11WwebmUXN7A0z+5/w9QwzezX8vB8ys5yh\nbuNgM7NRZvaImb1tZivN7IyR/lmb2f8J/9tebmYPmFneSPyszexOM9thZstTlqX9bC3ws/D83zSz\nkwfy3gqSHphZFLgNuACYBVxtZrOGtlVZkQC+6u6zgNOBL4XneQPwjLsfATwTvh5prgdWprz+EXCr\nu88EaoDPD0mrsuvfgD+6+9HAiQTnP2I/azObBHwFmOPuxwFR4CpG5mf9H8C8bst6+mwvAI4IH/OB\n2wfyxgqSns0F1rj7OndvBx4ELhniNg06d9/q7kvC5w0Ef1gmEZzr3eFmdwOXDk0Ls8PMJgMXAb8J\nXxvwAeCRcJOReM6lwFnAbwHcvd3daxnhnzUQA/LNLAYUAFsZgZ+1u78A7Oq2uKfP9hLgHg+8Aowy\nswn7+94Kkp5NAjalvK4Ml41YZjYdOAl4FRjn7lvDVduAcUPUrGz5KfB/gWT4uhyodfdE+Hokft4z\ngCrgrrBL7zdmVsgI/qzdfTPwr8BGggCpA15n5H/WXXr6bAf175uCRAAwsyLgUeAf3L0+dZ0H14iP\nmOvEzezDwA53f32o23KAxYCTgdvd/SSgiW7dWCPwsy4j+Nf3DGAiUMi+3T+HhGx+tgqSnm0GpqS8\nnhwuG3HMLE4QIve7++/Dxdu7St3w546hal8WnAlcbGbrCbosP0AwdjAq7P6Akfl5VwKV7v5q+PoR\ngmAZyZ/1B4F33b3K3TuA3xN8/iP9s+7S02c7qH/fFCQ9WwQcEV7dkUMwQLdgiNs06MKxgd8CK939\nJymrFgCfDp9/GvjDgW5btrj7je4+2d2nE3yuf3b3a4BngcvCzUbUOQO4+zZgk5kdFS46F3iLEfxZ\nE3RpnW5mBeF/613nPKI/6xQ9fbYLgE+FV2+dDtSldIH1m77Z3gszu5CgLz0K3Onu3x/iJg06M3sv\n8CKwjD3jBf9EME7yMDCVYAr+K9y9+0DeQc/Mzga+5u4fNrPDCCqU0cAbwLXu3jaU7RtsZjab4AKD\nHGAd8FmCf1CO2M/azL4LXElwheIbwN8RjAeMqM/azB4AziaYLn478B3gv0jz2Yah+nOCbr5m4LPu\nvni/31tBIiIiA6GuLRERGRAFiYiIDIiCREREBkRBIiIiA6IgERGRAVGQiAxzZnZ21wzFIsORgkRE\nRAZEQSIySMzsWjN7zcyWmtmvwvudNJrZreH9MJ4xs4pw29lm9kp4L4jHUu4TMdPMnjazv5nZEjM7\nPDx8Ucp9RO4Pv1AmMiwoSEQGgZkdQ/Dt6TPdfTbQCVxDMEngYnc/Fnie4NvGAPcA33D3EwhmFeha\nfj9wm7ufCLyHYMZaCGZl/geCe+McRjBflMiwEOt7ExHJwLnAKcCisFjIJ5ggLwk8FG5zH/D78L4g\no9z9+XD53cB/mlkxMMndHwNw91aA8HivuXtl+HopMB34S/ZPS6RvChKRwWHA3e5+414Lzb7dbbv9\nnZModR6oTvT/rgwj6toSGRzPAJeZ2VjYfa/saQT/j3XNMvsJ4C/uXgfUmNn7wuWfBJ4P71BZaWaX\nhsfINbOCA3oWIvtB/6oRGQTu/paZfQt4yswiQAfwJYKbR80N1+0gGEeBYErvX4ZB0TULLwSh8isz\nuzk8xuUH8DRE9otm/xXJIjNrdPeioW6HSDapa0tERAZEFYmIiAyIKhIRERkQBYmIiAyIgkRERAZE\nQSIiIgOiIBERkQH5/zbGIcphOA7tAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wE8wxBM88QiK",
        "colab_type": "text"
      },
      "source": [
        "# Save model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HbmJ3twe8S8v",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from google.colab import files\n",
        "\n",
        "model.save('keras.h5')\n",
        "files.download('keras.h5')\n",
        "#!pip install \"tensorflow>=1.7.0\"\n",
        "#!pip install tensorflowjs\n",
        "\n",
        "#!mkdir model\n",
        "#tfjs.converters.save_keras_model(model, \"./model\")\n",
        "#!tensorflowjs_converter --input_format keras keras.h5 model/\n",
        "#!zip -r model.zip model \n",
        "\n",
        "#files.download('model.zip')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7DsV3IEcsbaJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}