train_test_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "oUFdX44RdE88"
   },
   "source": [
    "This notebook provides a comprehensive guide for building and training a neural network using the methodology we developed during the competition. In addition, the final section of the notebook demonstrates how to perform inference and generate a valid log for submission."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "4bjpCRsteiGi"
   },
   "source": [
    "# Download dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9pH-6M5NdsFj"
   },
   "source": [
    "Download the dataset in TFRecord file format. Alternatively, create your own train/test files exploiting the notebook `convert_dataset_to_TFRecords.ipynb` in the folder `convert_to_TFRecord`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 532660,
     "status": "ok",
     "timestamp": 1679176201230,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "zEJarcCYQUlO",
    "outputId": "2b9267b1-b1ae-456d-9b72-b529715c73c9"
   },
   "outputs": [],
   "source": [
    "%cd\n",
    "%cd hyperspectral-cnn-soil-estimation/dataset\n",
    "\n",
    "#Challenge train set: will be used for training and validation\n",
    "!gdown https://drive.google.com/uc?id=1wD3vKqKEFh6OfrfLNtOENF-lbe4auQDb\n",
    "\n",
    "#Challenge test set: no labels are provided, will be used for making inference\n",
    "!gdown https://drive.google.com/uc?id=1XPaSRHsGHGkLieZnR6cU8-Bwg6_1kCTj\n",
    "\n",
    "%cd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KfsHjNylePr7"
   },
   "source": [
    "# Define dataset paths"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2D5NU0ZSJ3Xf"
   },
   "source": [
    "Navigate to the working directory."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1679259061935,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "OE9gS6QyJ2zt",
    "outputId": "72ccb365-0593-4bb3-a5e7-518df6c757c6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/microsat\n",
      "/home/microsat/hyperspectral-cnn-soil-estimation\n"
     ]
    }
   ],
   "source": [
    "%cd\n",
    "%cd hyperspectral-cnn-soil-estimation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "executionInfo": {
     "elapsed": 3,
     "status": "ok",
     "timestamp": 1679259062933,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "o2h0RGrVPVIe"
   },
   "outputs": [],
   "source": [
    "train_set_path = 'dataset/train_tfrecords0.record'\n",
    "test_set_path = 'dataset/test_tfrecords0.record'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "r-Y9IEKUevT6"
   },
   "source": [
    "# Import required libraries and functions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "nktdY6WpeYai"
   },
   "source": [
    "Import libraries and set random seeds."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1153,
     "status": "ok",
     "timestamp": 1679259065221,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "Ker0Hmm5PtR9",
    "outputId": "ef38bb83-0efd-4a29-eb4a-219b15548de0"
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import os, zipfile, logging\n",
    "logging.disable(logging.WARNING)\n",
    "os.environ[\"TF_CPP_MIN_LOG_LEVEL\"] = \"3\"\n",
    "import random as rn\n",
    "rn.seed(2)\n",
    "import numpy as np\n",
    "np.random.seed(3)\n",
    "SEED = 7231\n",
    "os.environ['PYTHONHASHSEED'] = str(SEED)\n",
    "os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0'\n",
    "\n",
    "#Disable GPU\n",
    "#os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"-1\"\n",
    "\n",
    "#Enable full deterministic operations\n",
    "#This options slows down the training process remarkably but it allows to get \n",
    "#deterministic results through different runs of the code\n",
    "#note that results might differ on different hardware (cpu vs gpu) and on different computers\n",
    "os.environ['TF_DETERMINISTIC_OPS'] = '1'\n",
    "\n",
    "import tensorflow as tf\n",
    "from dataset_processing import *\n",
    "from efficientnet_lite import EfficientNetLiteB0mod\n",
    "import tensorflow.keras.layers as layers\n",
    "\n",
    "tf.random.set_seed(1)\n",
    "\n",
    "AUTO = tf.data.AUTOTUNE "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "JUzhdRa5eZhT"
   },
   "source": [
    "# Define training hyperparameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "executionInfo": {
     "elapsed": 1176,
     "status": "ok",
     "timestamp": 1679259127319,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "SOc1Fc64L9u-"
   },
   "outputs": [],
   "source": [
    "noise_std = 0.05\n",
    "epochs = 300\n",
    "batch_size = 32\n",
    "target_image_size = 32\n",
    "start_lr = 0.005 \n",
    "network_name = 'cnn_final'\n",
    "\n",
    "label_normalization_mode = 0 # 0 for minmax normalization, everything else for standard normalization\n",
    "\n",
    "###########################################################\n",
    "max_reflectance = 6315\n",
    "test_set_len = 1153\n",
    "max_labels = [325., 625., 400., 14.]\n",
    "std_labels = [29.496254, 61.874084, 39.860474, 0.2602235]\n",
    "mean_labels = [70.30264, 227.98851, 159.28125, 6.782706]\n",
    "train_set_len = len(list(load_tf_records(train_set_path)))\n",
    "\n",
    "num_train_images = (train_set_len//batch_size)*batch_size"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "PtNPZGe4fjAr"
   },
   "source": [
    "Learning rate scheduling."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 447
    },
    "executionInfo": {
     "elapsed": 238,
     "status": "ok",
     "timestamp": 1679176251539,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "xgq4Q20lfbia",
    "outputId": "1d254fdc-3cc6-4da7-dc04-d616c9f48bdc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Learning rate per epoch:\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "alpha = 0.0001\n",
    "decay_steps = epochs\n",
    "\n",
    "def lr_scheduler(epoch, start_lr):\n",
    "    if epoch < 1:\n",
    "      lr = (start_lr-0.001)/1*epoch+0.001\n",
    "    else:\n",
    "      cosine_decay = 0.5 * (1 + tf.cos(np.pi * epoch / (decay_steps)))\n",
    "      decayed = (1 - alpha) * cosine_decay + alpha\n",
    "      lr = start_lr * decayed\n",
    "    return lr\n",
    "\n",
    "lr_callback = tf.keras.callbacks.LearningRateScheduler(lambda epoch: lr_scheduler(epoch, start_lr), verbose=0)\n",
    "\n",
    "#Plot learning rate\n",
    "\n",
    "x_range = np.arange(epochs)\n",
    "y = [lr_scheduler(x, start_lr) for x in x_range]\n",
    "plt.plot(x_range, y)\n",
    "print('Learning rate per epoch:')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "razlKuZYhNPK"
   },
   "source": [
    "# Define saving paths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "executionInfo": {
     "elapsed": 2,
     "status": "ok",
     "timestamp": 1679259138603,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "pemlGtg8hMFy"
   },
   "outputs": [],
   "source": [
    "base_path = 'networks/'\n",
    "\n",
    "saved_model_filepath = base_path + network_name \n",
    "h5_filepath = saved_model_filepath + '/'+network_name+'.h5'\n",
    "submission_name = 'submission_files/sumbission_' + network_name + '.csv'\n",
    "submission_zip = 'submission_files/sumbission_' + network_name + '.zip'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "FW0ms1FTfBMs"
   },
   "source": [
    "# Build the neural network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hFitZcYSkkEZ"
   },
   "source": [
    "Define a custom metric reflecting the competition scoring."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "executionInfo": {
     "elapsed": 2,
     "status": "ok",
     "timestamp": 1679259142147,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "tEzCX8v9kjg6"
   },
   "outputs": [],
   "source": [
    "def custom_metric(y_true, y_pred):\n",
    "    y_true = tf.cond(tf.math.equal(label_normalization_mode,0), \n",
    "                     lambda: tf.multiply(y_true, max_labels),\n",
    "                     lambda: tf.multiply(y_true, std_labels)+mean_labels)\n",
    "    y_pred = tf.cond(tf.math.equal(label_normalization_mode,0), \n",
    "                     lambda: tf.multiply(y_pred, max_labels),\n",
    "                     lambda: tf.multiply(y_pred, std_labels)+mean_labels)\n",
    "    \n",
    "    mse = tf.reduce_mean((y_true-y_pred)**2, axis=0)\n",
    "    mse_baseline = [870.02899169921875, 3828.40234375, 1588.857421875, 0.0677162706851959228515625]\n",
    "    score = tf.reduce_mean(mse/mse_baseline)\n",
    "\n",
    "    return score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 386,
     "status": "ok",
     "timestamp": 1679259143654,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "5Gpt_v7ErF3M",
    "outputId": "a13bf889-8648-4d99-ef5b-1cc0825c871e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " efficientnetlite0mod (Funct  (None, 1, 1, 1280)       728896    \n",
      " ional)                                                          \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 1280)              0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 4)                 5124      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 734,020\n",
      "Trainable params: 719,780\n",
      "Non-trainable params: 14,240\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "backbone = EfficientNetLiteB0mod(input_shape=(target_image_size, target_image_size, 150),\n",
    "                                   width_coefficient=0.5,\n",
    "                                   depth_coefficient=0.5,\n",
    "                                   dropout_rate=0.1\n",
    "                                   )\n",
    "\n",
    "model = tf.keras.Sequential([backbone,  \n",
    "                                layers.Flatten(),\n",
    "                                layers.Dense(4, kernel_initializer=tf.keras.initializers.GlorotUniform(seed=1509))])\n",
    "\n",
    "\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.Adam(clipnorm=1.),\n",
    "    loss='mse',\n",
    "    metrics=[custom_metric],\n",
    "  )\n",
    "\n",
    "#Initialize the model with the same weights we employed during the competition\n",
    "model.load_weights('efficientnet_lite/initialization_weigths.h5')\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ROfL1mCYfXhS"
   },
   "source": [
    "# Dataset preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "QZSuhqwiwrbU"
   },
   "outputs": [],
   "source": [
    "steps_per_epoch = num_train_images//batch_size\n",
    "\n",
    "ds = load_tf_records(train_set_path).map(decode_dataset_train_val, num_parallel_calls=AUTO,deterministic=True)\n",
    "ds = ds.shuffle(train_set_len, seed=1860)\n",
    "\n",
    "#Train\n",
    "train_data = ds.take(num_train_images)\n",
    "train_data = train_data.map(lambda patch, label, height, width: normalize_train_val(patch, label, height, width, max_reflectance, max_labels, mean_labels, std_labels, label_normalization_mode), num_parallel_calls=AUTO,deterministic=True).cache()    #normalize train dataset\n",
    "train_data = train_data.shuffle(num_train_images, seed=1866)    #shuffle train dataset\n",
    "train_data = train_data.map(lambda patch, label, height, width: augment_train(patch, label, height, width,target_image_size,noise_std), num_parallel_calls=AUTO,deterministic=True)    \n",
    "train_data = train_data.batch(batch_size=batch_size, drop_remainder=True)    #batch train dataset\n",
    "train_data = train_data.prefetch(AUTO)    #prefetch train dataset\n",
    "\n",
    "#Validation\n",
    "val_data_raw = ds.skip(num_train_images).cache()    #create a small validation dataset from the last elements of the train dataset that are discarded during batching\n",
    "val_data = val_data_raw.map(lambda patch, label, height, width: normalize_train_val(patch, label, height, width, max_reflectance, max_labels, mean_labels, std_labels, label_normalization_mode), num_parallel_calls=AUTO).cache()    #normalize validation dataset \n",
    "val_data = val_data.map(lambda patch, label, height, width: preprocess_val(patch, label, height, width,target_image_size), num_parallel_calls=AUTO)    #pad validation dataset\n",
    "val_data = val_data.repeat()\n",
    "val_data = val_data.batch(batch_size=batch_size, drop_remainder=True)    #batch validation dataset\n",
    "val_data = val_data.prefetch(AUTO)    #prefetch validation dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "91TlJzx4fu9L"
   },
   "source": [
    "Display few samples from train dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "executionInfo": {
     "elapsed": 2464,
     "status": "ok",
     "timestamp": 1679176263525,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "bNHLlotAX40k",
    "outputId": "430d1c08-e4d9-415c-a9d9-17fbde76e4cc"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaAAAAGdCAYAAABU0qcqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAxdElEQVR4nO3de3TU9Z038PfcJ5eZyf1GLoSLgNxUlJhVKUoKpF0Wld310meLXR890uCuUmtNH6uVthtrn9PaehDPc2phu6dItSt6dCtWo4S1BZQURLxEwCBBkgCBZJJJ5v57/qBmGwX5fCDhm4T365w5hyQfPvnO75J3fpmZz9gsy7JARER0jtlNL4CIiM5PDCAiIjKCAUREREYwgIiIyAgGEBERGcEAIiIiIxhARERkBAOIiIiMcJpewGclk0kcOnQIPp8PNpvN9HKIiEjJsix0d3ejqKgIdvupr3OGXQAdOnQIJSUlppdBRERnqaWlBcXFxaf8+pAF0KpVq/CTn/wEbW1tmDlzJh577DHMnj37tP/P5/MBAK685FtwOjyi7xUakyJelzOUFNcCgDOcENdGA8rNqRiC5OmKqVo7QvL63qJUVW/Pcd1awjluca33SETVO5buEtf2FOv2T+oR+b73dOq2SSxNtxbP0bC4NunR9Y755PX2uG5yl6X4I0bCo3tEwNkr3z+OiO68D42R/ez5VKCpW967NE3V29knX3vU51D1tsfk+1OzvePxCLb998P9P89P2VPcUeG3v/0tVqxYgSeeeAIVFRV49NFHsWDBAjQ1NSEvL+8L/++nf3ZzOjxwOr2i7+d0yepO1CoDKC7f6EnX0AWQ06k7sBwO+cms2X5nshanSx5ATqfuz66WSx5ADrdu/zhd8n2v3SaW8lhxKsqTmmLlWuyagxa6ALK5lAHkVARQQnfeO1y6AHI6ovJa7fkWk6896VIGkGJ/arb3p073MMqQPAnhpz/9KW677TZ84xvfwIUXXognnngCqamp+NWvfjUU346IiEagQQ+gaDSKxsZGVFVV/c83sdtRVVWFLVu2fK4+EokgGAwOuBER0eg36AF09OhRJBIJ5OfnD/h8fn4+2traPldfV1eHQCDQf+MTEIiIzg/GXwdUW1uLrq6u/ltLS4vpJRER0Tkw6E9CyMnJgcPhQHt7+4DPt7e3o6Cg4HP1Ho8HHo/uAT8iIhr5Bv0KyO12Y9asWaivr+//XDKZRH19PSorKwf72xER0Qg1JE/DXrFiBZYuXYpLL70Us2fPxqOPPopQKIRvfOMbQ/HtiIhoBBqSALrhhhtw5MgRPPDAA2hra8NFF12EjRs3fu6JCUREdP6yWZale2XZEAsGgwgEAqj425XiF2w5FK8U9nTIX1EOAOF8+ZQFb3ufqnc8Xf4CTcuunIunKHe3h1StLeWL3WJZ8hfeuTt02zDplq/FpnhBHwAk/PL9g4TuNHJ16LZ5X1mGuDaWpvvLetIlP1gC73epesd98sd3E6lDNxks6tcdsynt8heWAkBfvvxYSTuoO8Yj2fJtmHKwR9W7rzhdXOs9LP/ZGU+E8XpjHbq6uuD3+09ZZ/xZcEREdH5iABERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkRFDN/viLKW/3wGnQziC4jTvO/7XwmMzVetI/bBDXNt1Ua6qd2D3MXFtMl33PvLHJ8tHbGS36cZ3JLy6wybhlv+eE82Sjz4CAMsp3/fe/cdVvfuK08S1ru64qreW56h8fEvCm6rq7QrJRxTZ+nQjamJj5Meh56huTNbRi+W9M/bq1g3lhLJYivwYtxy63/s1549lV15TKO5mx3T59k5EnUDj6et4BUREREYwgIiIyAgGEBERGcEAIiIiIxhARERkBAOIiIiMYAAREZERDCAiIjKCAUREREYwgIiIyAgGEBERGTFsZ8FFiwJIOmXzzxIeeY4mFXOVAKBvXJa4NuOtVlXvaLG8t+tYr6q3PS6fYxbL1M1f6xqvqw/sk88x6ysQzv/7C0dYPswqNClb1TuU5xDX+iPyeWoAEJqgm0momTWX0qabqWbZ5fP0YoV+Ve94qvx8i5fqZtjlvB0S11qKeZEAkEjR/WhMa4+Ja8O5blXv9OZucW0iTdfbHpUft76D8vsYj8tqeQVERERGMICIiMgIBhARERnBACIiIiMYQEREZAQDiIiIjGAAERGREQwgIiIyggFERERGMICIiMiIYTuKJ1TohsMtGyvhb5aPeolk60a9pDYdFtceqyxS9U5rjYprLafudwVHVD6iRiu9VT6SAwDiafLDzLcnqOrdM04+Gsb/druqd6igUFzrfXOPrvdVk1T1sXT5NoxkulS9XaGEuDblQ/n5AADhbMU2PBxR9XYcl4+nOj4rR9XbHtOdP4G3j4prj1yRp+oNpIsrQ/ny8VEAkNuoGPPjlR+DVlw2OopXQEREZAQDiIiIjGAAERGREQwgIiIyggFERERGMICIiMgIBhARERnBACIiIiMYQEREZAQDiIiIjGAAERGREcN2FlzmO51wOmRz2xLvNon7plw8VbWORG5AXOv7WD6TDgBiPvnMLndCN5sq6pP/bpHweFW9fYrZewBg//MH4trjf3+xqnc8xSautc3QzeDKelc+aywZUu77NN3vfgmX/H5qZgwCQNKtWEtcPjcOALwd8rXE/LoZdonUDHGts093/qR/pJtJ2HG5/Njyf6zbPzG/fL5b0ik/TgAAlny7hHPlczTjwll6vAIiIiIjBj2Avv/978Nmsw24TZ48ebC/DRERjXBD8ie4qVOn4tVXX/2fb+Ictn/pIyIiQ4YkGZxOJwoKCoaiNRERjRJD8hjQnj17UFRUhHHjxuFrX/saDhw4cMraSCSCYDA44EZERKPfoAdQRUUF1q5di40bN2L16tVobm7GVVddhe7uk7/zXl1dHQKBQP+tpKRksJdERETD0KAHUHV1Nf7hH/4BM2bMwIIFC/D73/8enZ2dePrpp09aX1tbi66urv5bS0vLYC+JiIiGoSF/dkBGRgYuuOAC7N2796Rf93g88Hjkzy8nIqLRYchfB9TT04N9+/ahsLBwqL8VERGNIIMeQPfccw8aGhqwf/9+/OlPf8J1110Hh8OBm266abC/FRERjWCD/ie4gwcP4qabbkJHRwdyc3Nx5ZVXYuvWrcjNzVX1iWWnwHLKRsT0LakQ99WOqnCFkuJad6duxEbqnqPi2kS2T9U7e2eXuLavME3VO56mO2xiX71IXJvx3smfrHIqyVTd+BYNW1I+psS6eJKqt2+/fMwPAMT8bnFtOFu3TVIPy4/b7lljVL29hyPy2h7dNtEct66QboRQNCdVVZ/5fo+4NuGRj9YBAJfiicEZYd3IoaMX+cW1OTvlC4knZPt90ANo/fr1g92SiIhGIc6CIyIiIxhARERkBAOIiIiMYAAREZERDCAiIjKCAUREREYwgIiIyAgGEBERGcEAIiIiIxhARERkxJC/HcMZS/7lJpDWIp8h1TNWN/cMutFxKpGyLHFt1K/bVY6IfCaUs083J8vZJZ/vBQDetw+La0OXl6t6hzPkc7UymuTzugAgVCqfBxbx636Xy2zqU9W7j2jqU1S9e8bI58xlbZPvSwAIl2WKaz3duuNKNXvRppwBeVg3k9DWGxbXxibmq3r35stn+/n2646rzJ6YuDbpkf8MSsbjojpeARERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMmLYjuKJpzoAp2zMisMpH7ORckQxvgOAMygfDxLOl49uAQD3cfla7KnykTMAYCnKk27d7yGOY0FVfe8lZeJaW1I+QggAsre2i2tDk3NUvR1h+VoyP5GPgwIAR6/uOAwXyEdIhbN1p7W/WTFGpiig6p3yUYe4tufCPFVvV49s3AsAWHbdKJ5ooV9V7+z2iGu9zUdVvY9NGSOuzXxLd272jc8W1yZS5D8n4jHZucMrICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERjCAiIjIiGE7Cy7psiPpkuVjX478brh6dbPG4l7FUDXduCkEx6WIa73HE6relkO+mKhP93uIqzBTVR9Ll29DT6d8vhcAJDPkM9K6x+jm6eVv7RLXarY3AITKfar6WKp8H/k/0s2lsxzy3n05blXv3vwCca23I6bq7TlwTFwbK8xQ9XYeC6nqk+lecW20OEvVO2dXn7g2PE4+2w0A7HH5z8OuQvn5k4jKankFRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERjCAiIjICAYQEREZMWxnwcH6y03AdyAibhvNcKmWEU+TZ3Tcq5sH5j0mn+9mS6paw9knn6mWVG6TSKZHVe/b1y2ujafrZo1J5wUCQOoR3UaM5KWKa6M+3Zw5Z1i3lvQW+TFu742qeie98v2fdOqOcXe3/BgPZ+l+HCXcueJab5tuPl4yVXccJrzytVvKbeg6oli7JZ9JBwBJt/y4zXtLcR4nwqI6XgEREZER6gDavHkzFi1ahKKiIthsNjz33HMDvm5ZFh544AEUFhYiJSUFVVVV2LNnz2Ctl4iIRgl1AIVCIcycOROrVq066dcfeeQR/OIXv8ATTzyBbdu2IS0tDQsWLEA4LLskIyKi84P6MaDq6mpUV1ef9GuWZeHRRx/F/fffj8WLFwMAfv3rXyM/Px/PPfccbrzxxrNbLRERjRqD+hhQc3Mz2traUFVV1f+5QCCAiooKbNmy5aT/JxKJIBgMDrgREdHoN6gB1NbWBgDIz88f8Pn8/Pz+r31WXV0dAoFA/62kpGQwl0RERMOU8WfB1dbWoqurq//W0tJieklERHQODGoAFRSceP/39vb2AZ9vb2/v/9pneTwe+P3+ATciIhr9BjWAysvLUVBQgPr6+v7PBYNBbNu2DZWVlYP5rYiIaIRTPwuup6cHe/fu7f+4ubkZO3fuRFZWFkpLS3HXXXfhhz/8ISZOnIjy8nJ873vfQ1FREa699trBXDcREY1w6gDavn07rr766v6PV6xYAQBYunQp1q5di3vvvRehUAi33347Ojs7ceWVV2Ljxo3wenUjIhyRBBwJ2RiPWLr8bjgiynEsyzrFtW6HfOwIAOSkykdbRJO6XdXSmSGuvab4Q1Xv95ZeoKqP5qWJax2KEUIAECqWH1f5d36k6q3Zn+8dyT990V8pzehU1WvkentU9e3fOPmfx08mY7duzA+S8vPNFgypWndfViyuPT5N96f9jCbdNrTH5MdK0hq6h97DuboxWb158rW4g/KRTYmYE/jz6evUATR37lxY1qmHtNlsNqxcuRIrV67UtiYiovOI8WfBERHR+YkBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERqhH8ZwrjlgSDks2R8p26slAn3Pb48+q1vH4/rni2imZJ3/TvVO5wr9HXLum5QpV73yffM7czLQDqt7vPyafHQYAx/4zS1yb1qo7JL/6vU3i2p1B+ewwAChLPSauvXbKDlXvP3VPUNXHLYe4tthzXNW78tl94tq1D/2dqrcjKj85fU2634ddQcX8NadN1Tvp0R2HtoR85p3lHLrf++1xxQ9DAL4D8tmL3iN94tp4Iiyq4xUQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjhu0onoTLDptwZMXFP5aPQXmrp1y1jhtKtotr/xwsU/X+fccMca3bLh87AgBX534orv2PTypVvV3KtWS9HxHX2uPykSYA0NyXI64tT+tQ9W4N+8W1H/derOo9O7BfVZ/v6hLXbusep+rdFpXfz8X/p17V+9VvXimu7R0rXwcAJF3y8TraETXBsV5VfSxdXpuxL6bq3TMxIK51RHTnj/eobGQOoBtPlIzLankFRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERjCAiIjICAYQEREZMWxnwfWMccPhdotq89zd4r5vB4t160h4xLULst5R9e5Nynv/7IN5qt5lxUfFtRdlHlT1DsZ1c7LSH2kT1zrtullWDsjr2yO6WWPfKXxZXLv6yFxVb63/OiqfG7gkt1HV+/fH5L27ncoZaX75jxhPh3xmIADYFHMDe8rSVL2z3j6uqo8UyIfBpTS1q3o7SrLFtfF0l6p3PE1eH091yGtjsnmRvAIiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGTEsB3Fk/FhL5xO2aiN3+y7VNx35dQXVOsIJ+WjKl7smKnqnevuEdf+74l/UvWe6JaP+9jdpxtP9NXMt1X1ScXvOc93XKzq3d7nE9f+TfZHqt5Pdc4W15Z4j6l6r8jSreUfO8vFtS8e0x2HN+S8Ka7d3D1J1dsWt8S1xyfpxuVkNoXEtWktvarefWPkxxUAuLpj4tpkhnxsDwCExsjHH7l6daOs7DGbuNYRlve2hGOSeAVERERGMICIiMgIdQBt3rwZixYtQlFREWw2G5577rkBX7/llltgs9kG3BYuXDhY6yUiolFCHUChUAgzZ87EqlWrTlmzcOFCtLa29t+eeuqps1okERGNPuonIVRXV6O6uvoLazweDwoKCs54UURENPoNyWNAmzZtQl5eHiZNmoRly5aho6PjlLWRSATBYHDAjYiIRr9BD6CFCxfi17/+Nerr6/HjH/8YDQ0NqK6uRiJx8nfIq6urQyAQ6L+VlJQM9pKIiGgYGvTXAd144439/54+fTpmzJiB8ePHY9OmTZg37/NvK11bW4sVK1b0fxwMBhlCRETngSF/Gva4ceOQk5ODvXv3nvTrHo8Hfr9/wI2IiEa/IQ+ggwcPoqOjA4WFhUP9rYiIaARR/wmup6dnwNVMc3Mzdu7ciaysLGRlZeGhhx7CkiVLUFBQgH379uHee+/FhAkTsGDBgkFdOBERjWzqANq+fTuuvvrq/o8/ffxm6dKlWL16NXbt2oV///d/R2dnJ4qKijB//nz84Ac/gMfjUX2fUHEKnC7ZDKSVU/9T3PfZo5eo1pHlls+Q8jsjqt6t4YC4NhhPUfVOtcvXUt96gap3PXT1C4veF9fu68pR9U5Y8llWbx4fq+pdnNoprg04dbPGftwxUVWf45HPPZufsVvVe5Lr1M9S/az9HnktAOz4RD7v0B7XzYLrLZKfE+6uuKp3JMOhqk/9Y5O4Nj59nKq3Iyafp2cp/6Z1bJL853LWB7qfbxLqAJo7dy4s69Qb5OWXXz6rBRER0fmBs+CIiMgIBhARERnBACIiIiMYQEREZAQDiIiIjGAAERGREQwgIiIyggFERERGMICIiMgIBhARERkx6O8HNFgC7x6D0yGbU7Sjt0zcN8URU60jmpRvoo+6s1W9L8lsEde+3qabHRZJymdZzco5qOo9I12+bgD43a1fFtd65WOvAAA5//eAuPZQSD57DwA6nfJZY12uVFXvLUfLVfVj0jrFtQ++u0jVe2GZYlZfj25WXzLVJa51hE/+ppWnEiuU905p1c2C8x6TzxgEAOsUb7h5MqEi3VxM/4fyd4m2Hz6u6o2KUnFpNEP+szAek9XyCoiIiIxgABERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkRHDdhRPMt2DpMMrqv3vb10u7vuPv9ioWkdzJFdcuzD/XVXv1miGuPbinE9UvY9F5aNhpqQ1q3o39Rao6h29ivFHyaSq941528S1O3rHqnpXpu0R1z7TMVvV+5Is3Tij1rB8jNCisbtVvcs9R8S1f/jl36h6F4SPiWutpG4Ok6dTfqxEs9yq3q6gbnRP4iLdqCyN7gl+ca0/phtn5OmQn5tJl/x6xR6XbT9eARERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERgzbWXDHLkyHwy2bBZfxYZ+471P3flW1jpse+S9x7bF4uqp3Zfpece3bvaWq3pf75L3HuI6rev/X1VNU9bEpsv0IADZLNw/sia9Wi2v/5fcvqnrH4BDX/lPuH1W9/+PIFar60hT5TLU56R+oev/bv9wirvVk6Wb1JXfJ15K86mJV71ia/Pdnd7duRlpojEdVH/HbxLV5b3apevcVpsmLY7oZduEc+Yw8T6d8bpz0POYVEBERGcEAIiIiIxhARERkBAOIiIiMYAAREZERDCAiIjKCAUREREYwgIiIyAgGEBERGcEAIiIiI2yWpZx9MsSCwSACgQCuuPpBOJ2yES7BsfJxEjmNQd2CnPKMDufKR84AQPq9B8W1szIPqHpvWXapuDaaKd9+AJDycbeqvnesX1xrj+sOR3dnRN67N6rqHctKFddGa3XjjFKc8rEmABB+tEhcG/fKx8IAgCMq3+beI/LtDQC2pLx3NKA7Dt1d8v0ZydSN1vEe7lXVx33y/rF05QQ0xe50RHSjkjyHesS18ewUeW08jM1v/ABdXV3w+099/vMKiIiIjFAFUF1dHS677DL4fD7k5eXh2muvRVNT04CacDiMmpoaZGdnIz09HUuWLEF7e/ugLpqIiEY+VQA1NDSgpqYGW7duxSuvvIJYLIb58+cjFAr119x999144YUX8Mwzz6ChoQGHDh3C9ddfP+gLJyKikU31x8iNGzcO+Hjt2rXIy8tDY2Mj5syZg66uLjz55JNYt24drrnmGgDAmjVrMGXKFGzduhWXX3754K2ciIhGtLN6DKir68T7WmRlZQEAGhsbEYvFUFVV1V8zefJklJaWYsuWLSftEYlEEAwGB9yIiGj0O+MASiaTuOuuu3DFFVdg2rRpAIC2tja43W5kZGQMqM3Pz0dbW9tJ+9TV1SEQCPTfSkpKznRJREQ0gpxxANXU1GD37t1Yv379WS2gtrYWXV1d/beWlpaz6kdERCPDGb0l9/Lly/Hiiy9i8+bNKC4u7v98QUEBotEoOjs7B1wFtbe3o6Cg4KS9PB4PPB7dc/SJiGjkU10BWZaF5cuXY8OGDXjttddQXl4+4OuzZs2Cy+VCfX19/+eamppw4MABVFZWDs6KiYhoVFBdAdXU1GDdunV4/vnn4fP5+h/XCQQCSElJQSAQwK233ooVK1YgKysLfr8fd955JyorK/kMOCIiGkAVQKtXrwYAzJ07d8Dn16xZg1tuuQUA8LOf/Qx2ux1LlixBJBLBggUL8Pjjjw/KYomIaPQYtrPgrr7kPjgdstlqfYXymV025awxDXenbtZYNEs++yoScKh6p7fIZ3Y5+nRzyY7M8qnq/R/HxbXOHt1aHGF575hf91ijPSafqxXJcql69+bo9megWX5s2ZSndNwrX4u3TTcjrXui/FhJ3x86fdFfcR6Rv2QjUpat6m2L62aq2RLybd49Vj5TDQDSW8LiWs35AAAJr/waJOmRHyfxeBhvbHqIs+CIiGh4YgAREZERDCAiIjKCAUREREYwgIiIyAgGEBERGcEAIiIiIxhARERkBAOIiIiMYAAREZERZ/R2DOdCT0kanC7ZKB5HRD4Gw6acxJP53f3i2oBbPjIDAPZ05srX4dH1HpfeoarX6P3hTFV9OFM+wsPfKR8hBADRLNkxAgC3PrZB1fv5IxeJa4tSulS9N38yXlWP1D5xaUn6cVXr5ocmi2ttsYSqt6dTMYZJMVoHAHovkJ8/jqhutE7nRN24HJuifcaHunFGlsMmrk2k6EZCaUb39JTIz7WE8DjhFRARERnBACIiIiMYQEREZAQDiIiIjGAAERGREQwgIiIyggFERERGMICIiMgIBhARERnBACIiIiMYQEREZMSwnQVnj1uwSwe3yUclIff+j1Tr6I27xbVvHy5S9f5K6Xvi2r6EbsZTX1K+7lhSPqsNAGbc/7aq/u1/u0hcG0/T3c+ylU3i2vZYQNX7mqwPxLXPHJql6l3g61bVV2TtF9cWu4+peuNBeem7T0xTtc7ZckRcGxmbrertfeN9cW34qgtVvTP2yGfvAUA0Q36+hUp0c+acffJBc7aEbthlJFt+vvmb5TPs4nHZ7EpeARERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMmLYjuLxHo3A6ZTN2Jn+2DvivsG4bgzG4tyd4tpt6eNUvV8+OEW+jtJdqt4fd2WJa7uium1SmNqlqrcUk37skYSqd2mKfOzM/rBu1MuxaJq49uYxb6p6v9wxVVV/PJ4qro1pNjiAS3wHxLWR23Q/MlqCF4hr3Z1xVe/o5ZPFtY6w7rhKenTbMOqT/y5vk0/WAQCkfXBUXNs5K0/V23s0Jq5NpMj3fSIuq+UVEBERGcEAIiIiIxhARERkBAOIiIiMYAAREZERDCAiIjKCAUREREYwgIiIyAgGEBERGcEAIiIiIxhARERkxLCdBacxxtMprtXOyepMyGdwVfg+UvW22yxxbXfCq+pdlf2+uNbnCKt6bwlOUNVPvVc+xy5p6X4n6oimi2urMt5V9f7dkUvFta2xDFXvPE+Pqn6sVz4P7HhcPsMOAN7vLRTXtvf5VL39O9rEtZGxull97vaQuNZyyOZKfqp7YkBV7z0mn2Pn6pbPXwOA45fK57tlvNOp6h3Lkf9802xDyy6r5RUQEREZoQqguro6XHbZZfD5fMjLy8O1116LpqamATVz586FzWYbcLvjjjsGddFERDTyqQKooaEBNTU12Lp1K1555RXEYjHMnz8fodDAS+HbbrsNra2t/bdHHnlkUBdNREQjn+oxoI0bNw74eO3atcjLy0NjYyPmzJnT//nU1FQUFBQMzgqJiGhUOqvHgLq6TrwxWVbWwDc/+81vfoOcnBxMmzYNtbW16O3tPWWPSCSCYDA44EZERKPfGT8LLplM4q677sIVV1yBadOm9X/+5ptvRllZGYqKirBr1y585zvfQVNTE5599tmT9qmrq8NDDz10pssgIqIR6owDqKamBrt378Ybb7wx4PO33357/7+nT5+OwsJCzJs3D/v27cP48eM/16e2thYrVqzo/zgYDKKkpORMl0VERCPEGQXQ8uXL8eKLL2Lz5s0oLi7+wtqKigoAwN69e08aQB6PBx6P50yWQUREI5gqgCzLwp133okNGzZg06ZNKC8vP+3/2blzJwCgsFD+YjciIhr9VAFUU1ODdevW4fnnn4fP50Nb24lXOQcCAaSkpGDfvn1Yt24dvvKVryA7Oxu7du3C3XffjTlz5mDGjBlDcgeIiGhkUgXQ6tWrAZx4selfW7NmDW655Ra43W68+uqrePTRRxEKhVBSUoIlS5bg/vvvH7QFExHR6KD+E9wXKSkpQUNDw1ktqP972W3ieUKvHpks7vv/xj+tWsfvQxeIaz8M61779JXA2+LaPVFdb82cub3hfFXv/aGs0xf9lSl++TywHJduRlqOs1tc2xTW/Rl4WcFr4tqWmG6O2RTvIVX9hqOXiGuvy/mzqnenWz4P7O8yd6h6/7TwRnGt+2CnqnesSD6vLenUveLEv7tDVd89Vb7/Ixm6h94De+TnhL1Tfj4AQO/UDHGtK5QU18ZjspmbnAVHRERGMICIiMgIBhARERnBACIiIiMYQEREZAQDiIiIjGAAERGREQwgIiIyggFERERGMICIiMiIM34/oKHm+qAFTptbVHtJ5lFx3+9+8hXVOv4+Z7u41gH5qAoA2NwjHyH0dwHdeBWfLSau/cGxr6p6t3RmqOrzvfLxIHu681S9NbxO+TYBgJglGycCABM87areLltcVX9lxh5xbUciXdVbcz87E/KxPQDQV+AV16bGdedP3Ctfd0qL7p2Wk/4UVb0jIl+7o0/VGpFs+Ta0B3Qju3rz5dcgKfIfs0jEZH15BUREREYwgIiIyAgGEBERGcEAIiIiIxhARERkBAOIiIiMYAAREZERDCAiIjKCAUREREYwgIiIyAgGEBERGTFsZ8HFJ5UATtkMpPZIVNz3WCRNtY43Q+PFta3hgKp3e9gnrt3cPkHV+7Kcj8W1TntC1fvH055V1f/o27eIa10h3Vou+7F8Vt91gUZV75UfLxLXvucuVPWuKaxX1W/tke//I1H5cQUAF/kOiGtfOT5V1dvVI9+fPWN156azVz5/LZar623ZbKp6WPLSmE8+ww4A7DF5c++RiKp36mH5NkxrlfeOx2W1vAIiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjGEBERGTEsB3FY9lO3CT2PHShuG/WfftV64gk5ZsoCd34jrtL/iCufWDPYlXvcSlHxLV/49ur6l0flG9vQDcexLF1t6r3eO9hce2TR+aoet9evFlc+0ksU9X7jrf/l6r+6xO3iWvn+t5X9f5V+1Xi2s67xqh6x4rlY2e8R2Oq3rakfERNLE33o859XD7eCwDsTvm574jqfk44+uTjctyHOlW94ZAftz1jPOLahHB8EK+AiIjICAYQEREZwQAiIiIjGEBERGQEA4iIiIxgABERkREMICIiMoIBRERERjCAiIjICAYQEREZwQAiIiIjhu0suEiWGwmXW1Sbti8o77s0VbWOvqdlawCAv816W9X7SNwvrv2nUvksMK0dvWWq+u3fvVRV7+nrFdf2VV+i6r3unovFta1fl8+kA4AJqfI5c9f5d6p6506VH7MA8GLHTHHtl9I+UPVurZsgru2dLp/tBgCeLvkcM3tMXnviP8hL3Z262W6xgPy8B4CUA13i2mh+uqq3PSrfLolsn6q3LS6fp5exq1NcG0/IzjVeARERkRGqAFq9ejVmzJgBv98Pv9+PyspKvPTSS/1fD4fDqKmpQXZ2NtLT07FkyRK0t7cP+qKJiGjkUwVQcXExHn74YTQ2NmL79u245pprsHjxYrz77rsAgLvvvhsvvPACnnnmGTQ0NODQoUO4/vrrh2ThREQ0sqkeA1q0aNGAj3/0ox9h9erV2Lp1K4qLi/Hkk09i3bp1uOaaawAAa9aswZQpU7B161Zcfvnlg7dqIiIa8c74MaBEIoH169cjFAqhsrISjY2NiMViqKqq6q+ZPHkySktLsWXLllP2iUQiCAaDA25ERDT6qQPonXfeQXp6OjweD+644w5s2LABF154Idra2uB2u5GRkTGgPj8/H21tbafsV1dXh0Ag0H8rKSlR3wkiIhp51AE0adIk7Ny5E9u2bcOyZcuwdOlSvPfee2e8gNraWnR1dfXfWlpazrgXERGNHOrXAbndbkyYcOJ1A7NmzcJbb72Fn//857jhhhsQjUbR2dk54Cqovb0dBQUFp+zn8Xjg8cjfa5yIiEaHs34dUDKZRCQSwaxZs+ByuVBfX9//taamJhw4cACVlZVn+22IiGiUUV0B1dbWorq6GqWlpeju7sa6deuwadMmvPzyywgEArj11luxYsUKZGVlwe/3484770RlZSWfAUdERJ+jCqDDhw/j61//OlpbWxEIBDBjxgy8/PLL+PKXvwwA+NnPfga73Y4lS5YgEolgwYIFePzxx89oYQm3DTaXTVR7fGaGuG9Gk+6vjs3XJ8S1H5ZOUfVOr/tEXFuQ0q3q/UHtVHGt5ZBt508l3br6RLp8rIlmNAgAxFPlF/Glj+ku+F/yzRXXtqzMUvUuSzmqqj/8L/JxSfeMqVH19nbIRxTZYy5Vb7tif8Z8unPT1R0X10aydH/m78vWjRxyH5P3j6fqetsV52e0QHc/HVH5/vFY8tqEcL+r9viTTz75hV/3er1YtWoVVq1apWlLRETnIc6CIyIiIxhARERkBAOIiIiMYAAREZERDCAiIjKCAUREREYwgIiIyAgGEBERGcEAIiIiI9TTsIea9ZdxD4lYWPx/kpCPqogn5H0BwErKx5TE47resVBUXBtNyGu1a7Es5Sgem+73Fns8Jq6Nx3SHZDwmX0s8Lt+XJ3rLxzBFe+T3EQDCCfkYGUB33MZjuv2p2S7xuHybALpRPAm77riyxeXbULMvASAR1Y3L0ZxvitMBAGCPJxW9deePFZPvH4fqODlRa51mfI/NOl3FOXbw4EG+KR0R0SjQ0tKC4uLiU3592AVQMpnEoUOH4PP5YLP9z29ywWAQJSUlaGlpgd/vN7jCocX7OXqcD/cR4P0cbQbjflqWhe7ubhQVFcH+BVe2w+5PcHa7/QsT0+/3j+qd/ynez9HjfLiPAO/naHO29zMQCJy2hk9CICIiIxhARERkxIgJII/HgwcffBAej+4Nl0Ya3s/R43y4jwDv52hzLu/nsHsSAhERnR9GzBUQERGNLgwgIiIyggFERERGMICIiMiIERNAq1atwtixY+H1elFRUYE333zT9JIG1fe//33YbLYBt8mTJ5te1lnZvHkzFi1ahKKiIthsNjz33HMDvm5ZFh544AEUFhYiJSUFVVVV2LNnj5nFnoXT3c9bbrnlc/t24cKFZhZ7hurq6nDZZZfB5/MhLy8P1157LZqamgbUhMNh1NTUIDs7G+np6ViyZAna29sNrfjMSO7n3LlzP7c/77jjDkMrPjOrV6/GjBkz+l9sWllZiZdeeqn/6+dqX46IAPrtb3+LFStW4MEHH8Sf//xnzJw5EwsWLMDhw4dNL21QTZ06Fa2trf23N954w/SSzkooFMLMmTOxatWqk379kUcewS9+8Qs88cQT2LZtG9LS0rBgwQKEw7qhrqad7n4CwMKFCwfs26eeeuocrvDsNTQ0oKamBlu3bsUrr7yCWCyG+fPnIxQK9dfcfffdeOGFF/DMM8+goaEBhw4dwvXXX29w1XqS+wkAt91224D9+cgjjxha8ZkpLi7Gww8/jMbGRmzfvh3XXHMNFi9ejHfffRfAOdyX1ggwe/Zsq6ampv/jRCJhFRUVWXV1dQZXNbgefPBBa+bMmaaXMWQAWBs2bOj/OJlMWgUFBdZPfvKT/s91dnZaHo/HeuqppwyscHB89n5almUtXbrUWrx4sZH1DJXDhw9bAKyGhgbLsk7sO5fLZT3zzDP9Ne+//74FwNqyZYupZZ61z95Py7KsL33pS9a//uu/mlvUEMnMzLR++ctfntN9OeyvgKLRKBobG1FVVdX/ObvdjqqqKmzZssXgygbfnj17UFRUhHHjxuFrX/saDhw4YHpJQ6a5uRltbW0D9msgEEBFRcWo268AsGnTJuTl5WHSpElYtmwZOjo6TC/prHR1dQEAsrKyAACNjY2IxWID9ufkyZNRWlo6ovfnZ+/np37zm98gJycH06ZNQ21tLXp7e00sb1AkEgmsX78eoVAIlZWV53RfDrthpJ919OhRJBIJ5OfnD/h8fn4+PvjgA0OrGnwVFRVYu3YtJk2ahNbWVjz00EO46qqrsHv3bvh8PtPLG3RtbW0AcNL9+unXRouFCxfi+uuvR3l5Ofbt24fvfve7qK6uxpYtW+Bw6N53ZjhIJpO46667cMUVV2DatGkATuxPt9uNjIyMAbUjeX+e7H4CwM0334yysjIUFRVh165d+M53voOmpiY8++yzBler984776CyshLhcBjp6enYsGEDLrzwQuzcufOc7cthH0Dni+rq6v5/z5gxAxUVFSgrK8PTTz+NW2+91eDK6GzdeOON/f+ePn06ZsyYgfHjx2PTpk2YN2+ewZWdmZqaGuzevXvEP0Z5Oqe6n7fffnv/v6dPn47CwkLMmzcP+/btw/jx48/1Ms/YpEmTsHPnTnR1deF3v/sdli5dioaGhnO6hmH/J7icnBw4HI7PPQOjvb0dBQUFhlY19DIyMnDBBRdg7969ppcyJD7dd+fbfgWAcePGIScnZ0Tu2+XLl+PFF1/E66+/PuBtUwoKChCNRtHZ2TmgfqTuz1Pdz5OpqKgAgBG3P91uNyZMmIBZs2ahrq4OM2fOxM9//vNzui+HfQC53W7MmjUL9fX1/Z9LJpOor69HZWWlwZUNrZ6eHuzbtw+FhYWmlzIkysvLUVBQMGC/BoNBbNu2bVTvV+DEu/52dHSMqH1rWRaWL1+ODRs24LXXXkN5efmAr8+aNQsul2vA/mxqasKBAwdG1P483f08mZ07dwLAiNqfJ5NMJhGJRM7tvhzUpzQMkfXr11sej8dau3at9d5771m33367lZGRYbW1tZle2qD51re+ZW3atMlqbm62/vjHP1pVVVVWTk6OdfjwYdNLO2Pd3d3Wjh07rB07dlgArJ/+9KfWjh07rI8//tiyLMt6+OGHrYyMDOv555+3du3aZS1evNgqLy+3+vr6DK9c54vuZ3d3t3XPPfdYW7ZssZqbm61XX33VuuSSS6yJEyda4XDY9NLFli1bZgUCAWvTpk1Wa2tr/623t7e/5o477rBKS0ut1157zdq+fbtVWVlpVVZWGly13unu5969e62VK1da27dvt5qbm63nn3/eGjdunDVnzhzDK9e57777rIaGBqu5udnatWuXdd9991k2m836wx/+YFnWuduXIyKALMuyHnvsMau0tNRyu93W7Nmzra1bt5pe0qC64YYbrMLCQsvtdltjxoyxbrjhBmvv3r2ml3VWXn/9dQvA525Lly61LOvEU7G/973vWfn5+ZbH47HmzZtnNTU1mV30Gfii+9nb22vNnz/fys3NtVwul1VWVmbddtttI+6Xp5PdPwDWmjVr+mv6+vqsb37zm1ZmZqaVmppqXXfddVZra6u5RZ+B093PAwcOWHPmzLGysrIsj8djTZgwwfr2t79tdXV1mV240j//8z9bZWVlltvttnJzc6158+b1h49lnbt9ybdjICIiI4b9Y0BERDQ6MYCIiMgIBhARERnBACIiIiMYQEREZAQDiIiIjGAAERGREQwgIiIyggFERERGMICIiMgIBhARERnBACIiIiP+P+O0PE1qAS56AAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaAAAAGdCAYAAABU0qcqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAxOUlEQVR4nO3de3TV5Zkv8O++JzuXHZKQG7kY7igEWwqYUSlKhktXXVhZHa2eKVhHlzZ4qrSdNl2tVts5cfQctXUhznQcaM8SqXaJVKdqFUs4tgRLKsVbU5IGEyQXLiY72cm+v+cPh7RRkOcJCW8Svp+19lokeXjy/i57P/ll73y3wxhjQEREdI45bS+AiIjOTxxARERkBQcQERFZwQFERERWcAAREZEVHEBERGQFBxAREVnBAURERFa4bS/go5LJJI4cOYKMjAw4HA7byyEiIiVjDHp7e1FUVASn8/TXOWNuAB05cgQlJSW2l0FERGepra0NxcXFp/36qA2gjRs34oEHHkBHRwfmz5+PRx55BIsWLTrj/8vIyAAAXHbxBrhdPtH3csaS4nUNFPrFtQDgOxYW18ayZOs9KeXQB+LaSHGWqnfc7xLXugcSqt7eY/26tWSmiGsjkzyq3mltIXGts1e37oELssW12n3YWyrfJwAQaOoT18Yydeehr71XXGtc8vMKAIxHXh8uSFX1dvfL97m3S77/AP129pdmiGtT25X3n4BXXJvw6p5VcYfl+zDpkveOxyOo/+19g4/np/3+4o4KP//5z7FhwwY89thjWLx4MR5++GGsWLECjY2NyMvL+8T/e/LXbm6XD2637E7qTMp3otuju+O7FXvIuHV3fOmABYCEcF8MUtzx3THdg6fbpauHYu0Jj24AuV1xca3TqdxOxbrdbl1vl1d5Hiq2U38eRuW9tQNIUa+/byru966Yqrd2OzVr199/5MfT4VYOIMU+1Aygk870NMqovAjhwQcfxM0334wbb7wRF154IR577DH4/X7853/+52h8OyIiGodGfABFo1E0NDSgqqrqr9/E6URVVRX27NnzsfpIJIJgMDjkRkREE9+ID6Bjx44hkUggPz9/yOfz8/PR0dHxsfra2loEAoHBG1+AQER0frD+d0A1NTXo6ekZvLW1tdleEhERnQMj/iKE3NxcuFwudHZ2Dvl8Z2cnCgoKPlbv8/ng8+meNCUiovFvxK+AvF4vFixYgJ07dw5+LplMYufOnaisrBzpb0dEROPUqLwMe8OGDVi7di0+85nPYNGiRXj44YcRCoVw4403jsa3IyKicWhUBtC1116Lo0eP4q677kJHRwcuvvhivPjiix97YQIREZ2/Ri0JYf369Vi/fv2w/3883Qu4ZX8B7IrI/5jKNSBPTQCAS37yB3FtZVqTqndrLEdce2lqs6r3q6HZ4tr/WrdE1TtcJP+rbwBwxOX7PL2pR9U7li1PtvD0R1S9b330F6p6jf6k7nlPv1O+9stT3lf1vmmF/DcT0bw0VW9f6wlxbUqX7hmBWKY8IUCTyAAAzj55AgoAOBLy+0RCsW4A8HbJ0z4SabreyVT5CHBG5Y+zzris1vqr4IiI6PzEAURERFZwABERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWjFoUz9lyRRJwJYRxDj84Ku7rd8VV6yj3yXt3xXURNc3hPHFtJOlR9S7yfCCu3fSLTare//SVr6nqvccHxLXxzBRV7xkPviuu3XPkAlXvFEdMXHtgoFTV+0gkS1V/XU69uPaPUXnEEwDctOMlce2Pv36dqne0Qn6Op7X1q3o7o/KIp0S6LvoooowcSjkqP8cTKbqHXUdI3jupiKYCAM8x+T7XxF7B4RCV8QqIiIis4AAiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrOAAIiIiKziAiIjIirGbBdcfg8slm4+HewLivh6XLF/upKZAvri2wt+q6r0k40/i2iynLidrf7hMXLtloETVu/L/vK6qr7v378S1Kcejqt4DCXlGnt+n6/2WYr98xt+i6t3t02V2HYlNGpVaAEhAltsFADUP/VTV+6Hr/0Fc6xiQZ+8BgPdoj7jWpKWqerv65PsEAKK58uw4X0evqjec8usEz/GQqrVxK3q3d4trHYmIqI5XQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGQFBxAREVkxZqN4ch96H540r6g22JMr7jspZUC1jqqMt8W1TxyrVPW+JLNZXHvITFb1znDKt7N6cpuq96/75fE3APB6x6fFtdkP6NYS8Mi3c0Gurne+Rx718seBUlXvpv48Vf3CTHnUj8+pi7S5wHtUXLuvv1zVO54uuw8DgDesW3e0LEdcG0vTPdSltOuir9x98pin2GR5bA8AOKK6GCGN6CT58fG/Z+SNE7JaXgEREZEVHEBERGQFBxAREVnBAURERFZwABERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZMWaz4DSmBY6Jayd7+1S9m6PyzK6rst9Q9X49NE1cm+EKq3pPdgfFtXe0f0bVe3Zqu6o+6XWJa6/O1e3DX3QtENfOTO9S9Q4beebdp1IPqXq3hbNV9WnOiLi2xHNc1ftoPFNc+/mMP6p67w4vFtcm/fJcMgAYyFXkmLXr7j/aH81dvZr+KarejlhSXBsu8qt6u/sT4tpYtrx3PC7bgbwCIiIiK0Z8AH3/+9+Hw+EYcps9e/ZIfxsiIhrnRuVXcBdddBFeeeWVv34T94T4TR8REY2gUZkMbrcbBQUFo9GaiIgmiFF5DujgwYMoKirC1KlTccMNN6C1tfW0tZFIBMFgcMiNiIgmvhEfQIsXL8aWLVvw4osvYtOmTWhpacHll1+O3t7eU9bX1tYiEAgM3kpKSkZ6SURENAaN+ABatWoVvvjFL6KiogIrVqzAr371K3R3d+Opp546ZX1NTQ16enoGb21turdNJiKi8WnUXx2QlZWFmTNnoqmp6ZRf9/l88Pl8o70MIiIaY0b974D6+vrQ3NyMwsLC0f5WREQ0joz4APrGN76Buro6HDp0CL/73e/whS98AS6XC1/60pdG+lsREdE4NuK/gjt8+DC+9KUv4fjx45g8eTIuu+wy1NfXY/Lkyao+k7198PlkUSi/P1Ym7usLyKMnAMDjiItrUxwxVe9VmfJYk7ZYjqp3U1j+MvgL/UdUvVMcUVW9r+WouPa14ExVb028TlNIdw4G3APiWu3xmZOm2+ehpPzX1CcS6areGU75dnoc8lgYAIhOksflpL0rP08AINUnj3iCUbUGnLqfzUPTJolrfcflsUoA4Gp+X1zrD+vOQyN8jAWARJr8WDqSsh0+4gNo27ZtI92SiIgmIGbBERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGQFBxAREVnBAURERFZwABERkRUcQEREZMWovx3DcP1pXRncLln+1XdeeE7c91c981Xr+EPoAnGtNlMtzSnPhOpO+FW9s9194tqeRKqqd45X3huAKlfrL2vluX4AcMuOX8l79+eqendFM8S1s1LaVb2d0GWqvdhdIa5dGnhX1Ttq5OfWnp4Zqt7ufnn2oknVvS2LIyrfh+7uflVv41HkzAHwt8pzILW9IxeXi2uTboeqd0qnfL94/tIhrnUkZXmRvAIiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrOAAIiIiKziAiIjIijEbxRMpzkLCnSKq3ROSx4PsP16sWkdF9vvi2mneTlXv92OTxLUZzgFVb69DHoFS4OlR9d74lS+q6qHY5a4BeaQJADzVtVBcuyhwSNV723sLxLXaGKYMl+549sW94tqY0d2t9wSni2sb75yj6u3tkp9bxq+M4jFGXDtQlqXqnfpet6o+mS57rAIAxHUxTBrOuHyfAIAjJn+ciF2QL66Nx8OAILmHV0BERGQFBxAREVnBAURERFZwABERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWjNksOE93BG6XrHbvcnnY2MIX3lOt42Bvnri2LT1H1TuUlGdfPddRoer9vQueE9fec8ONqt4Ohy7LytMiz8gbmKfL6uv5x4C49rUt01S9rytrENe2RnXHPtfdp6q/LKtJXKs5rwDgL/+jRFzrjXWreif98oy0WJYiTw2AOyTPDUx6HKrexudR1XfPTBPXTno7qOqd8mdFxmRSd980mfJ1e9o/ENc6khFRHa+AiIjICg4gIiKyggOIiIis4AAiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrHAYY4ztRfytYDCIQCCAqvyb4XZ6Rf/HZKaL+zviCdV6PlhUIK5dd/cvVb1nejvEtfdde72qN0bxsEZzUlX13qMD4lrj0f1MlEyVxxkOTJadTyd5euXnSu2mf1P1/nM0X1X/82WLxLWJvEmq3pp9HsnV5bX5D/UoFqI7Z/vLs8S1zvjoPsyltCm2M6lbS7QwU1zr7ouqeidS5Zl3nmPy/MJ4IoKdjQ+ip6cHmZmnXz+vgIiIyAr1ANq9ezeuuuoqFBUVweFw4Nlnnx3ydWMM7rrrLhQWFiI1NRVVVVU4ePDgSK2XiIgmCPUACoVCmD9/PjZu3HjKr99///348Y9/jMceewx79+5FWloaVqxYgXA4fNaLJSKiiUP9fkCrVq3CqlWrTvk1YwwefvhhfPe738Xq1asBAD/72c+Qn5+PZ599Ftddd93ZrZaIiCaMEX0OqKWlBR0dHaiqqhr8XCAQwOLFi7Fnz55T/p9IJIJgMDjkRkREE9+IDqCOjg9f1ZWfP/QVPvn5+YNf+6ja2loEAoHBW0mJ/N0ZiYho/LL+Kriamhr09PQM3tra2mwviYiIzoERHUAFBR/+zUxn59D3MO/s7Bz82kf5fD5kZmYOuRER0cQ3ogOovLwcBQUF2Llz5+DngsEg9u7di8rKypH8VkRENM6pXwXX19eHpqamwY9bWlqwf/9+ZGdno7S0FHfccQd++MMfYsaMGSgvL8f3vvc9FBUV4eqrrx7JdRMR0TinjuLZtWsXrrjiio99fu3atdiyZQuMMbj77rvx7//+7+ju7sZll12GRx99FDNnzhT1PxnFs3Thd+B2y2I/nP3y+AlHTBfFg6MnxKWJ6VNUrZ2K2Ix4li4CJTzZJ671v9+v6h3LkvcGAF+XvH88oOvtUESsuEK6mJLuC+W/Ds54T/d3bq6+iKo+VJ4hL1amzqQ1y1956uzXbWdozmRxreZYAoBxO8S1/vpmVe9keZGqXvO4EpukjLLq7JWvo1d3Xw4uKhbXeoJxcW08HsZvf3PPGaN41FdAS5cuxSfNLIfDgXvvvRf33nuvtjUREZ1HrL8KjoiIzk8cQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGSFOornXHGG43C6YqJa45VvRlyZw3TxE01nLvpvt+fuUPV+sudT4tpZKe2q3lHjEtf+5KZrVL2THnkGFwAYl7zeOHW9XYocQK3Z//Ntca1TGcBWmirPGASA1oFsce01uftUvR+5do241hhdVp//oHw7+2fItxEAUt8PiWujFReoejtiSV29T/4Y5ArLM9UAAB3HxKWmIFfVOqOxW14cl+fdxROyrENeARERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGTFmI3iSaa4kXR7RLUzHmkU9204VqxaR8A9IK59pneuqndMEZejidYBgKWpR8S1nY+9qur9y69eqapPamJKBnQxJVM2vSeuLU79QNU7nJSdfwDwD1mvq3q/Gpqjql+b95q41gVdjMw3n9omrn3gH65T9TYZ8uge14Ay/mZAHsOUKEhT9U7tksf8AIDxKO6fceV2+uXxYUYR2wMA8QvLxLWeVnlvR5JRPERENIZxABERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGTFmM2Ci6d5AGEWXGMwT9x3SUGzah1t4Uni2kUZLareRR55Ntkvj31K1TuWIz+0nbFMVe+bf/KMqn7L6uXi2qTfq+rd2C0/9n/o1OUALi/9k7jW49Dle302Td4bANpiOeLaBByq3kdi8nP8ss0Nqt57Vs8S10ZnyY8lAOC4/P7jy5TnqQEAjFGV907LENemvS/PlwQAR0KeY2cm6e7L7s4ecW0iP0temwgDh89cxysgIiKyggOIiIis4AAiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrBizUTw3PrgD/gyXqHZ3z2xx3xNReawFAJSlHhfXLkw9pOr9//pniGtnpHWpek9xy2NKoEwp2ds7TVXv6O0X1/r/rVvVe3pqUFyb6dZFoLRHAuLa+oGpqt7vR+XxNwAwxSs/ni7oYoHeDRWKa5cE/qzq/dtMeYRUwqeLEOq7VH4e+g+HVL2NR/bYc5ImXseR0MX8RArl8Tq+jl5V7/hkee/oJHlMVjwmO5a8AiIiIis4gIiIyAr1ANq9ezeuuuoqFBUVweFw4Nlnnx3y9XXr1sHhcAy5rVy5cqTWS0REE4R6AIVCIcyfPx8bN248bc3KlSvR3t4+eHvyySfPapFERDTxqF+EsGrVKqxateoTa3w+HwoKCoa9KCIimvhG5TmgXbt2IS8vD7NmzcJtt92G48dP/0qySCSCYDA45EZERBPfiA+glStX4mc/+xl27tyJf/3Xf0VdXR1WrVqFRCJxyvra2loEAoHBW0lJyUgviYiIxqAR/zug6667bvDf8+bNQ0VFBaZNm4Zdu3Zh2bJlH6uvqanBhg0bBj8OBoMcQkRE54FRfxn21KlTkZubi6amplN+3efzITMzc8iNiIgmvlEfQIcPH8bx48dRWCj/a2siIpr41L+C6+vrG3I109LSgv379yM7OxvZ2dm45557sGbNGhQUFKC5uRn//M//jOnTp2PFihUjunAiIhrf1ANo3759uOKKKwY/Pvn8zdq1a7Fp0yYcOHAAP/3pT9Hd3Y2ioiIsX74cP/jBD+Dz+VTf50gsGylR2fJm+TvEfQ+Fc1Tr6FDkgb3QO0/V+1OK7Lik0V2s1vXJ8/Hmph5W9Z6Xpqv/4/SLxbWfy31Z1fvt0BRxbczo8r1KU0+Iaz+I6zIG8zy6V3u2RuTn7WXpury26X752i9PfU/V+6d+eX5YWrNunziS8sy74Bxd9l7my++q6geukN/f/K26XDpfu3y/RAsyVL3dwYi8dkBx/4mf+kVnH+sp7/ihpUuXwpjTh+m99NJL2pZERHQeYhYcERFZwQFERERWcAAREZEVHEBERGQFBxAREVnBAURERFZwABERkRUcQEREZAUHEBERWcEBREREVoz4+wGNlGAiBZGER1Sb6+4T983z9qrWkVBksM1NbVP1TnPKc5g+ndqi6v2nSJG49r1orqp3Ag5VvcZFvvdV9e3RLHHtn/vyVL2/NeUFce3BaL6q9y+PXayqXzJJnu/WFdflgUWSsvsZAPyqb5aqtyskP8cjBbo8PXd/XFyb2ilfBwCgUHeu+N/vF9cany6TMJItf4saV1iWwXaSs7VLXBu9tFxcG4/JRguvgIiIyAoOICIisoIDiIiIrOAAIiIiKziAiIjICg4gIiKyggOIiIis4AAiIiIrOICIiMgKDiAiIrJizEbxvPrg38HtSRHV3vXDzeK+e0PTVOvwu+QRHvtCU1W956TKY2feHihW9fY45JEcV6S/o+qd4Yyq6ncdukhce//n16h6f/nZV8S1+7t1+/Ch9uXi2sqsZlXvyqy/qOqzXPKol4Ty58rPZ/xRXPvbgemq3sYnj/lxD+hiZJKKSJukSxcf5Y3GVPWh0hxxrSNhVL19x+X3N3dXUNU7WSaPkHJG5Ot2xmS1vAIiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrOAAIiIiKziAiIjICg4gIiKyYsxmwQX+3Au3S5aB9E54irjvfx2W55IBgMclz6dakNum6u31y3uHk/JMLQCYltopru2IZ6l6/6+VV6rqjd8rro1npap6v9FfJq69PKdJ1fupQ58S194z5XlV7/qwfN0A8LteeQbbZ9JbVL1DRn5uPXfDElVvRzIurtX+NOzqk+c0Ok/0qnpHp05W1fsPK7L6/Lr7sjOqyMg79oGqtyssz5nzpMjHhSMuy9LjFRAREVnBAURERFZwABERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWjNkonp4ZGXB7UkS1L9z2WXHf2fc3q9ZxIpImrr0g5biqtyZeJ9sdUvU+kUgX125Zv0zV250vj1cBANeALJYDANxHdZEpB/5xtrg2+L/lsSMAMCv7qLj2ub65qt4pDvk+AYCpqfK1tEZzVb3/7z99Xlzr8KtaI+mV/4zrPRJU9Y4VZMiLs2SPJSe5Qrrj44jJ43KMx6Xq7e7slhdnZap6hy7ME9emtinumwlZTBKvgIiIyArVAKqtrcXChQuRkZGBvLw8XH311WhsbBxSEw6HUV1djZycHKSnp2PNmjXo7JQHYxIR0flBNYDq6upQXV2N+vp6vPzyy4jFYli+fDlCob/+eujOO+/Ec889h6effhp1dXU4cuQIrrnmmhFfOBERjW+q54BefPHFIR9v2bIFeXl5aGhowJIlS9DT04PHH38cW7duxZVXfhjZv3nzZsyZMwf19fW45JJLRm7lREQ0rp3Vc0A9PT0AgOzsbABAQ0MDYrEYqqqqBmtmz56N0tJS7Nmz55Q9IpEIgsHgkBsREU18wx5AyWQSd9xxBy699FLMnfvhK4A6Ojrg9XqRlZU1pDY/Px8dHR2n7FNbW4tAIDB4KykpGe6SiIhoHBn2AKqursZbb72Fbdu2ndUCampq0NPTM3hra9O9qygREY1Pw/o7oPXr1+P555/H7t27UVxcPPj5goICRKNRdHd3D7kK6uzsREFBwSl7+Xw++Hy+4SyDiIjGMdUVkDEG69evx/bt2/Hqq6+ivLx8yNcXLFgAj8eDnTt3Dn6usbERra2tqKysHJkVExHRhKC6AqqursbWrVuxY8cOZGRkDD6vEwgEkJqaikAggJtuugkbNmxAdnY2MjMzcfvtt6OyspKvgCMioiFUA2jTpk0AgKVLlw75/ObNm7Fu3ToAwEMPPQSn04k1a9YgEolgxYoVePTRR0dksURENHE4jDHG9iL+VjAYRCAQwLI534DbJXtuqHdWlrh/Wlu/aj1Jrzy3afVPdp65aJiev/5yVb0mb8o4dGtx9Qyo6pN++XN8rh5d5p1xy7fTEdFlwfVeXCiuXXvfL1W9Y0aXB7b9K/K8PuPQHdCkT74WZ0SeeQYAno5uca0mlwwA/Ifkf7IRm5Sq6u3u050rsUnyrDnPB2FVb6fiPmFSvKrejl7542GsbLK4Nh4Po67+h+jp6UFm5unz6ZgFR0REVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGQFBxAREVnBAURERFZwABERkRXDejuGcyFckAa3WxZv4YokxX2dPbooHkeqPNril+uWqnprYn5crpiqtzMoj8uJ56aresOpi3pJpMv3oSOmi3pxKJKkEgG/qnf6O8fEtU9fL4/KAQBnny7OKDFFcVdVRislPfKfQz3duhiZZJo8AkcTrQMAxi1ft7ftuKp3rDhHVe8aiItrk/vfUfUOr1wornX3KR8nAvLj4+qVH3uTiMi+v7gjERHRCOIAIiIiKziAiIjICg4gIiKyggOIiIis4AAiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyIoxmwXnPRGBWxiVlkyVb4ZJ8ajW4eiXZRoBgCsqz4MCAJMlzyZL+OV5agDgOtEnL07K89QAIJkuy+g7ydvSJa7t/cwUVe+0ll5x7UCBPPcKANK7usW1yZQMVe9EZkBV7z0aEtf2l2aqevubT4hrIyVZqt6eYFRcGyxPU/XOev2IuNak+FS9tRwReYahc+Y0VW/fMXluYCJV9/gWzZLvl5Sw4vHNyK5teAVERERWcAAREZEVHEBERGQFBxAREVnBAURERFZwABERkRUcQEREZAUHEBERWcEBREREVnAAERGRFWM2isf1QRAupywmwpGVLu4bKZDXAoCrXxE/4XCoejsSSXFt3C/MJfpvLsU+cYXkcUMAEA/oIm1McY641v+ePHIGAOKZ8lgg/3tBVe9EXpa41nh1P8vFU3THE9ny2CZnQhet5AjKY5u8x3SRUM5eeYxMmke3D2NTssW17i7dsQ9PVkZfDcgfSlM65Pd7AHC2yaOsnFm6GCZM1j0ejjReARERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGQFBxAREVkxZrPgolOykXTLcr68LfKspOQUXfZRyol+cW3fzCxVb9eAPBPKoYv3grNXvm54Parenk5drtbAVHlml5b/z0fFtclMeZ4aALiO9ohrTV5A1Vt7xzNuXc6gRmTOFHGt93C3qnf4AnkOoK/tA1Vv45PntWlqASD1iDzDDgDcbfLzMF6Wp+odn1sirlVlVwJwRhPi2kSq/HEiEZf15RUQERFZoRpAtbW1WLhwITIyMpCXl4err74ajY2NQ2qWLl0Kh8Mx5HbrrbeO6KKJiGj8Uw2guro6VFdXo76+Hi+//DJisRiWL1+OUGhohP7NN9+M9vb2wdv9998/oosmIqLxT/Wr6BdffHHIx1u2bEFeXh4aGhqwZMmSwc/7/X4UFBSMzAqJiGhCOqvngHp6PnySNjt76JPMTzzxBHJzczF37lzU1NSgv//0T4hHIhEEg8EhNyIimviG/Sq4ZDKJO+64A5deeinmzp07+Pnrr78eZWVlKCoqwoEDB/Ctb30LjY2NeOaZZ07Zp7a2Fvfcc89wl0FEROPUsAdQdXU13nrrLbz22mtDPn/LLbcM/nvevHkoLCzEsmXL0NzcjGnTpn2sT01NDTZs2DD4cTAYREmJ/GWHREQ0Pg1rAK1fvx7PP/88du/ejeLi4k+sXbx4MQCgqanplAPI5/PB5/MNZxlERDSOqQaQMQa33347tm/fjl27dqG8vPyM/2f//v0AgMLCwmEtkIiIJibVAKqursbWrVuxY8cOZGRkoKOjAwAQCASQmpqK5uZmbN26FZ/73OeQk5ODAwcO4M4778SSJUtQUVExKhtARETjk2oAbdq0CcCHf2z6tzZv3ox169bB6/XilVdewcMPP4xQKISSkhKsWbMG3/3ud0dswURENDGofwX3SUpKSlBXV3dWCxqOeEmuuNbbHdX1niTPD0vtCKt6Owdi4lpHTJ7ZBACJ3AxxresDRW4cAJzhPPgod598O+Ppuly64MX54tqUroiqt9Mrz3cLlehy5jKae1X10exUca2nW7edA0Xy3t435fl4AODOkWcvxgp0eXpOxX3CkdCds85++TkLAMmcLHGtcepy/dy98scsV0j3+Obo6RPXmj55LYxsHcyCIyIiKziAiIjICg4gIiKyggOIiIis4AAiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyIphvx/QaItmeZH0eEW1jrg8ZsN/qFu1jqRftgYAiOTKI00AwJkmj51xDcR1vaPymBKToou/Sfp0p4330FFxbWxekap35t42cW2oYoqqt18RlxPo0kXURC6Qx0cBgK9Dvpb+Ml2kTcIr/znUFMujjwAgpohWcg/o4qaMS75ud+cHqt6xgixVfThP/pYy6e8cU/WOlEwS1yZTdfdNT0Qe3ROd8fG30zmdeDwM7D5zHa+AiIjICg4gIiKyggOIiIis4AAiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrBizWXC+ExG43Q5RrSb3LDRNnqsEAGkHT8iLtVlwEUX2lTzuDgDgCMuz48JT0lW9PX26XLpY2WRxbdobrarefQvLxLWazEAASGSniWuNU3au/nUtSVV90ifPVHMkdduZ2SjPsUumKnMDPfKfcV3HBlS94+nynMZkly5/DYVZqvKMN7vEtYmA/LwCAO8J+X4xbt01RXxyprjWGZY/XjnjslpeARERkRUcQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGTFmI3iSaS44XDLlufpkEeJoNCvWodJlcd9pLYq1gEgliOP5Ej6XKre7qMRcW1Kp+7nkIQyjgWKmJrorCJVa39bn7jWeHT7MBqQH/toQHdXCtS3qeoH5hSKa7XnoXHJ90tSGfWS0tUvrnX2hFS9EfDJa2fII5uAYUQlZcofVxxGF5XkiMqjryK5Gare3g/C4tp4pnx/J+Oyc4pXQEREZAUHEBERWcEBREREVnAAERGRFRxARERkBQcQERFZwQFERERWcAAREZEVHEBERGQFBxAREVnBAURERFaM2Sw4d38MbrcsT8ikKPLa2uXZVAAQLpDntaUc6VX1doVi4tp4liL3CoBJSxHXOo8HVb2ThZN0a3HIf85xRHU5WdGcVHFtSmOHqrcX2fLeymOfnJylqvcEo/LioydUvQcWThXXpr3VruptMuX3H9Onu296jsnP8Uh+uqp3ysFOVX0iNyCudSqy3QCge578PMxslmcjAoBxye+bmscrE5fV8gqIiIisUA2gTZs2oaKiApmZmcjMzERlZSVeeOGFwa+Hw2FUV1cjJycH6enpWLNmDTo7dT9JEBHR+UE1gIqLi3HfffehoaEB+/btw5VXXonVq1fj7bffBgDceeedeO655/D000+jrq4OR44cwTXXXDMqCyciovFN9RzQVVddNeTjf/mXf8GmTZtQX1+P4uJiPP7449i6dSuuvPJKAMDmzZsxZ84c1NfX45JLLhm5VRMR0bg37OeAEokEtm3bhlAohMrKSjQ0NCAWi6GqqmqwZvbs2SgtLcWePXtO2ycSiSAYDA65ERHRxKceQG+++SbS09Ph8/lw6623Yvv27bjwwgvR0dEBr9eLrKysIfX5+fno6Dj9q49qa2sRCAQGbyUlJeqNICKi8Uc9gGbNmoX9+/dj7969uO2227B27Vq88847w15ATU0Nenp6Bm9tbbq3KiYiovFJ/XdAXq8X06dPBwAsWLAAv//97/GjH/0I1157LaLRKLq7u4dcBXV2dqKgoOC0/Xw+H3w+3d+4EBHR+HfWfweUTCYRiUSwYMECeDwe7Ny5c/BrjY2NaG1tRWVl5dl+GyIimmBUV0A1NTVYtWoVSktL0dvbi61bt2LXrl146aWXEAgEcNNNN2HDhg3Izs5GZmYmbr/9dlRWVvIVcERE9DGqAdTV1YUvf/nLaG9vRyAQQEVFBV566SX8/d//PQDgoYcegtPpxJo1axCJRLBixQo8+uijw1qYI5qEI5EQ1SbS5b/C65/iV60j453j8mKji5ExabKoIQBw9yqiWAA4u+WRHJq4FACIp8ujjwDA16qIhvF6VL2NV772WGmuqre7S/6KzGSWbh8mPfJjDwChEnnkUFan7hxPORYW18ZKclS9NRwZhbr6iOzxAQC8XSFV77gybsoRS4prY7m64xN4t1tcm0jTPZ3hCkXEtaHyTHFtPOYQ1akG0OOPP/6JX09JScHGjRuxceNGTVsiIjoPMQuOiIis4AAiIiIrOICIiMgKDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrFCnYY82899xNvGEPCLCOGSxDwAQj+lmrmYd2iieZEK+bi2TVOy/hC4WJh7XnTYuzT5MyCNNPlyLfO2OuK43FOtOKvdh0qHc58JoEwCIK449ACQS8iiepGP0fmZ1JHX3H4fiXJHGep2UjOvqHQn52hOjeB4m4so4sIQ84isek58nJ2vNGR4THeZMFefY4cOH+aZ0REQTQFtbG4qLi0/79TE3gJLJJI4cOYKMjAw4/ubKJhgMoqSkBG1tbcjMlIfijTfczonjfNhGgNs50YzEdhpj0Nvbi6KiIjidp79qHnO/gnM6nZ84MTMzMyf0wT+J2zlxnA/bCHA7J5qz3c5AIHDGGr4IgYiIrOAAIiIiK8bNAPL5fLj77rvh8+necGm84XZOHOfDNgLczonmXG7nmHsRAhERnR/GzRUQERFNLBxARERkBQcQERFZwQFERERWjJsBtHHjRlxwwQVISUnB4sWL8frrr9te0oj6/ve/D4fDMeQ2e/Zs28s6K7t378ZVV12FoqIiOBwOPPvss0O+bozBXXfdhcLCQqSmpqKqqgoHDx60s9izcKbtXLdu3ceO7cqVK+0sdphqa2uxcOFCZGRkIC8vD1dffTUaGxuH1ITDYVRXVyMnJwfp6elYs2YNOjs7La14eCTbuXTp0o8dz1tvvdXSiodn06ZNqKioGPxj08rKSrzwwguDXz9Xx3JcDKCf//zn2LBhA+6++2784Q9/wPz587FixQp0dXXZXtqIuuiii9De3j54e+2112wv6ayEQiHMnz8fGzduPOXX77//fvz4xz/GY489hr179yItLQ0rVqxAOCwPPRwLzrSdALBy5cohx/bJJ588hys8e3V1daiurkZ9fT1efvllxGIxLF++HKFQaLDmzjvvxHPPPYenn34adXV1OHLkCK655hqLq9aTbCcA3HzzzUOO5/33329pxcNTXFyM++67Dw0NDdi3bx+uvPJKrF69Gm+//TaAc3gszTiwaNEiU11dPfhxIpEwRUVFpra21uKqRtbdd99t5s+fb3sZowaA2b59++DHyWTSFBQUmAceeGDwc93d3cbn85knn3zSwgpHxke30xhj1q5da1avXm1lPaOlq6vLADB1dXXGmA+PncfjMU8//fRgzbvvvmsAmD179tha5ln76HYaY8xnP/tZ87Wvfc3eokbJpEmTzH/8x3+c02M55q+AotEoGhoaUFVVNfg5p9OJqqoq7Nmzx+LKRt7BgwdRVFSEqVOn4oYbbkBra6vtJY2alpYWdHR0DDmugUAAixcvnnDHFQB27dqFvLw8zJo1C7fddhuOHz9ue0lnpaenBwCQnZ0NAGhoaEAsFhtyPGfPno3S0tJxfTw/up0nPfHEE8jNzcXcuXNRU1OD/v5+G8sbEYlEAtu2bUMoFEJlZeU5PZZjLoz0o44dO4ZEIoH8/Pwhn8/Pz8ef/vQnS6saeYsXL8aWLVswa9YstLe345577sHll1+Ot956CxkZGbaXN+I6OjoA4JTH9eTXJoqVK1fimmuuQXl5OZqbm/Gd73wHq1atwp49e+By6d4XaCxIJpO44447cOmll2Lu3LkAPjyeXq8XWVlZQ2rH8/E81XYCwPXXX4+ysjIUFRXhwIED+Na3voXGxkY888wzFler9+abb6KyshLhcBjp6enYvn07LrzwQuzfv/+cHcsxP4DOF6tWrRr8d0VFBRYvXoyysjI89dRTuOmmmyyujM7WddddN/jvefPmoaKiAtOmTcOuXbuwbNkyiysbnurqarz11lvj/jnKMznddt5yyy2D/543bx4KCwuxbNkyNDc3Y9q0aed6mcM2a9Ys7N+/Hz09PfjFL36BtWvXoq6u7pyuYcz/Ci43Nxcul+tjr8Do7OxEQUGBpVWNvqysLMycORNNTU22lzIqTh678+24AsDUqVORm5s7Lo/t+vXr8fzzz+M3v/nNkLdNKSgoQDQaRXd395D68Xo8T7edp7J48WIAGHfH0+v1Yvr06ViwYAFqa2sxf/58/OhHPzqnx3LMDyCv14sFCxZg586dg59LJpPYuXMnKisrLa5sdPX19aG5uRmFhYW2lzIqysvLUVBQMOS4BoNB7N27d0IfV+DDd/09fvz4uDq2xhisX78e27dvx6uvvory8vIhX1+wYAE8Hs+Q49nY2IjW1tZxdTzPtJ2nsn//fgAYV8fzVJLJJCKRyLk9liP6koZRsm3bNuPz+cyWLVvMO++8Y2655RaTlZVlOjo6bC9txHz96183u3btMi0tLea3v/2tqaqqMrm5uaarq8v20oatt7fXvPHGG+aNN94wAMyDDz5o3njjDfPee+8ZY4y57777TFZWltmxY4c5cOCAWb16tSkvLzcDAwOWV67zSdvZ29trvvGNb5g9e/aYlpYW88orr5hPf/rTZsaMGSYcDtteuthtt91mAoGA2bVrl2lvbx+89ff3D9bceuutprS01Lz66qtm3759prKy0lRWVlpctd6ZtrOpqcnce++9Zt++faalpcXs2LHDTJ061SxZssTyynW+/e1vm7q6OtPS0mIOHDhgvv3tbxuHw2F+/etfG2PO3bEcFwPIGGMeeeQRU1paarxer1m0aJGpr6+3vaQRde2115rCwkLj9XrNlClTzLXXXmuamppsL+us/OY3vzEAPnZbu3atMebDl2J/73vfM/n5+cbn85lly5aZxsZGu4sehk/azv7+frN8+XIzefJk4/F4TFlZmbn55pvH3Q9Pp9o+AGbz5s2DNQMDA+arX/2qmTRpkvH7/eYLX/iCaW9vt7foYTjTdra2tpolS5aY7Oxs4/P5zPTp0803v/lN09PTY3fhSl/5yldMWVmZ8Xq9ZvLkyWbZsmWDw8eYc3cs+XYMRERkxZh/DoiIiCYmDiAiIrKCA4iIiKzgACIiIis4gIiIyAoOICIisoIDiIiIrOAAIiIiKziAiIjICg4gIiKyggOIiIis4AAiIiIr/j/KIxpHtXWA5QAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for image, label in train_data.unbatch().take(10):\n",
    "    plt.imshow(image[:,:,149])\n",
    "    plt.show()    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WbZBaR-AgE9R"
   },
   "source": [
    "# Train neural network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "54/54 [==============================] - 14s 163ms/step - loss: 0.3752 - custom_metric: 290.2100 - val_loss: 0.0939 - val_custom_metric: 97.7853 - lr: 0.0010\n",
      "Epoch 2/300\n",
      "54/54 [==============================] - 7s 121ms/step - loss: 0.2181 - custom_metric: 148.3953 - val_loss: 0.0455 - val_custom_metric: 5.2579 - lr: 0.0050\n",
      "Epoch 3/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0392 - custom_metric: 27.7009 - val_loss: 0.0096 - val_custom_metric: 2.5025 - lr: 0.0050\n",
      "Epoch 4/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0400 - custom_metric: 22.6821 - val_loss: 0.0195 - val_custom_metric: 5.0404 - lr: 0.0050\n",
      "Epoch 5/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0358 - custom_metric: 25.1825 - val_loss: 0.0066 - val_custom_metric: 1.1639 - lr: 0.0050\n",
      "Epoch 6/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0348 - custom_metric: 20.5259 - val_loss: 0.0106 - val_custom_metric: 8.6398 - lr: 0.0050\n",
      "Epoch 7/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0285 - custom_metric: 12.3327 - val_loss: 0.0093 - val_custom_metric: 1.7717 - lr: 0.0050\n",
      "Epoch 8/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0285 - custom_metric: 14.1943 - val_loss: 0.0109 - val_custom_metric: 2.3395 - lr: 0.0050\n",
      "Epoch 9/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0539 - custom_metric: 10.7343 - val_loss: 0.0107 - val_custom_metric: 2.1760 - lr: 0.0050\n",
      "Epoch 10/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0607 - custom_metric: 31.1109 - val_loss: 0.0224 - val_custom_metric: 3.4224 - lr: 0.0050\n",
      "Epoch 11/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0274 - custom_metric: 9.4843 - val_loss: 0.0109 - val_custom_metric: 2.1253 - lr: 0.0050\n",
      "Epoch 12/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0314 - custom_metric: 27.1009 - val_loss: 0.0091 - val_custom_metric: 1.7813 - lr: 0.0050\n",
      "Epoch 13/300\n",
      "54/54 [==============================] - 6s 104ms/step - loss: 0.0317 - custom_metric: 15.6470 - val_loss: 0.0178 - val_custom_metric: 2.7916 - lr: 0.0050\n",
      "Epoch 14/300\n",
      "54/54 [==============================] - 6s 108ms/step - loss: 0.0227 - custom_metric: 6.2435 - val_loss: 0.0120 - val_custom_metric: 1.6338 - lr: 0.0050\n",
      "Epoch 15/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0283 - custom_metric: 9.5632 - val_loss: 0.0091 - val_custom_metric: 1.2465 - lr: 0.0050\n",
      "Epoch 16/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0299 - custom_metric: 17.7923 - val_loss: 0.0127 - val_custom_metric: 2.1005 - lr: 0.0050\n",
      "Epoch 17/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0266 - custom_metric: 20.5445 - val_loss: 0.0126 - val_custom_metric: 3.0797 - lr: 0.0050\n",
      "Epoch 18/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0486 - custom_metric: 14.2930 - val_loss: 0.0233 - val_custom_metric: 3.6714 - lr: 0.0050\n",
      "Epoch 19/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0326 - custom_metric: 9.4583 - val_loss: 0.0145 - val_custom_metric: 2.5658 - lr: 0.0050\n",
      "Epoch 20/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0293 - custom_metric: 10.1822 - val_loss: 0.0076 - val_custom_metric: 0.9013 - lr: 0.0050\n",
      "Epoch 21/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0258 - custom_metric: 10.0686 - val_loss: 0.0086 - val_custom_metric: 5.8259 - lr: 0.0049\n",
      "Epoch 22/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0244 - custom_metric: 4.9882 - val_loss: 0.0129 - val_custom_metric: 2.3877 - lr: 0.0049\n",
      "Epoch 23/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0176 - custom_metric: 5.2661 - val_loss: 0.0123 - val_custom_metric: 2.0026 - lr: 0.0049\n",
      "Epoch 24/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0201 - custom_metric: 12.3685 - val_loss: 0.0106 - val_custom_metric: 1.3074 - lr: 0.0049\n",
      "Epoch 25/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0158 - custom_metric: 13.8327 - val_loss: 0.0087 - val_custom_metric: 1.8364 - lr: 0.0049\n",
      "Epoch 26/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0104 - custom_metric: 3.2765 - val_loss: 0.0110 - val_custom_metric: 2.2580 - lr: 0.0049\n",
      "Epoch 27/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0142 - custom_metric: 5.3669 - val_loss: 0.0093 - val_custom_metric: 1.3406 - lr: 0.0049\n",
      "Epoch 28/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0103 - custom_metric: 3.4787 - val_loss: 0.0104 - val_custom_metric: 1.8289 - lr: 0.0049\n",
      "Epoch 29/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0095 - custom_metric: 2.0225 - val_loss: 0.0122 - val_custom_metric: 1.6038 - lr: 0.0049\n",
      "Epoch 30/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0100 - custom_metric: 2.4416 - val_loss: 0.0112 - val_custom_metric: 2.1491 - lr: 0.0049\n",
      "Epoch 31/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0104 - custom_metric: 2.4274 - val_loss: 0.0088 - val_custom_metric: 1.4186 - lr: 0.0049\n",
      "Epoch 32/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0091 - custom_metric: 1.7607 - val_loss: 0.0082 - val_custom_metric: 1.3385 - lr: 0.0049\n",
      "Epoch 33/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0080 - custom_metric: 1.3723 - val_loss: 0.0101 - val_custom_metric: 1.4342 - lr: 0.0049\n",
      "Epoch 34/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0090 - custom_metric: 1.5803 - val_loss: 0.0120 - val_custom_metric: 2.9320 - lr: 0.0049\n",
      "Epoch 35/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0081 - custom_metric: 1.8364 - val_loss: 0.0116 - val_custom_metric: 1.8260 - lr: 0.0048\n",
      "Epoch 36/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0085 - custom_metric: 1.6454 - val_loss: 0.0138 - val_custom_metric: 1.8625 - lr: 0.0048\n",
      "Epoch 37/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0095 - custom_metric: 3.8117 - val_loss: 0.0105 - val_custom_metric: 1.7852 - lr: 0.0048\n",
      "Epoch 38/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0095 - custom_metric: 1.9015 - val_loss: 0.0112 - val_custom_metric: 2.4421 - lr: 0.0048\n",
      "Epoch 39/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0091 - custom_metric: 1.7733 - val_loss: 0.0118 - val_custom_metric: 1.7309 - lr: 0.0048\n",
      "Epoch 40/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0082 - custom_metric: 1.4123 - val_loss: 0.0116 - val_custom_metric: 1.5287 - lr: 0.0048\n",
      "Epoch 41/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0089 - custom_metric: 1.7043 - val_loss: 0.0095 - val_custom_metric: 1.1850 - lr: 0.0048\n",
      "Epoch 42/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0081 - custom_metric: 1.6659 - val_loss: 0.0091 - val_custom_metric: 1.2602 - lr: 0.0048\n",
      "Epoch 43/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0082 - custom_metric: 1.4798 - val_loss: 0.0099 - val_custom_metric: 1.6592 - lr: 0.0048\n",
      "Epoch 44/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0075 - custom_metric: 1.3007 - val_loss: 0.0120 - val_custom_metric: 1.5892 - lr: 0.0048\n",
      "Epoch 45/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0077 - custom_metric: 1.2356 - val_loss: 0.0099 - val_custom_metric: 1.3483 - lr: 0.0047\n",
      "Epoch 46/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0074 - custom_metric: 1.2984 - val_loss: 0.0104 - val_custom_metric: 1.4254 - lr: 0.0047\n",
      "Epoch 47/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0075 - custom_metric: 1.5028 - val_loss: 0.0101 - val_custom_metric: 1.4224 - lr: 0.0047\n",
      "Epoch 48/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0075 - custom_metric: 1.3387 - val_loss: 0.0096 - val_custom_metric: 1.3636 - lr: 0.0047\n",
      "Epoch 49/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0072 - custom_metric: 1.3202 - val_loss: 0.0112 - val_custom_metric: 1.4161 - lr: 0.0047\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 50/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0071 - custom_metric: 1.1518 - val_loss: 0.0119 - val_custom_metric: 1.5348 - lr: 0.0047\n",
      "Epoch 51/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0073 - custom_metric: 1.2738 - val_loss: 0.0101 - val_custom_metric: 1.5155 - lr: 0.0047\n",
      "Epoch 52/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0071 - custom_metric: 1.1444 - val_loss: 0.0107 - val_custom_metric: 1.3891 - lr: 0.0047\n",
      "Epoch 53/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0072 - custom_metric: 1.1631 - val_loss: 0.0120 - val_custom_metric: 1.5469 - lr: 0.0046\n",
      "Epoch 54/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0072 - custom_metric: 1.2391 - val_loss: 0.0112 - val_custom_metric: 1.4080 - lr: 0.0046\n",
      "Epoch 55/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0071 - custom_metric: 1.1457 - val_loss: 0.0100 - val_custom_metric: 1.2580 - lr: 0.0046\n",
      "Epoch 56/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0082 - custom_metric: 1.3765 - val_loss: 0.0097 - val_custom_metric: 1.6503 - lr: 0.0046\n",
      "Epoch 57/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0074 - custom_metric: 1.1783 - val_loss: 0.0118 - val_custom_metric: 1.5565 - lr: 0.0046\n",
      "Epoch 58/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0071 - custom_metric: 1.1105 - val_loss: 0.0103 - val_custom_metric: 1.4164 - lr: 0.0046\n",
      "Epoch 59/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0070 - custom_metric: 1.1195 - val_loss: 0.0114 - val_custom_metric: 1.4829 - lr: 0.0046\n",
      "Epoch 60/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0069 - custom_metric: 1.0944 - val_loss: 0.0102 - val_custom_metric: 1.3721 - lr: 0.0045\n",
      "Epoch 61/300\n",
      "54/54 [==============================] - 6s 119ms/step - loss: 0.0069 - custom_metric: 1.0973 - val_loss: 0.0104 - val_custom_metric: 1.3601 - lr: 0.0045\n",
      "Epoch 62/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0069 - custom_metric: 1.0794 - val_loss: 0.0110 - val_custom_metric: 1.4598 - lr: 0.0045\n",
      "Epoch 63/300\n",
      "54/54 [==============================] - 7s 121ms/step - loss: 0.0069 - custom_metric: 1.0770 - val_loss: 0.0098 - val_custom_metric: 1.3384 - lr: 0.0045\n",
      "Epoch 64/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0067 - custom_metric: 1.0180 - val_loss: 0.0114 - val_custom_metric: 1.4908 - lr: 0.0045\n",
      "Epoch 65/300\n",
      "54/54 [==============================] - 6s 119ms/step - loss: 0.0068 - custom_metric: 1.0044 - val_loss: 0.0090 - val_custom_metric: 1.2677 - lr: 0.0045\n",
      "Epoch 66/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0070 - custom_metric: 1.1177 - val_loss: 0.0108 - val_custom_metric: 1.4752 - lr: 0.0044\n",
      "Epoch 67/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0068 - custom_metric: 1.0193 - val_loss: 0.0098 - val_custom_metric: 1.3565 - lr: 0.0044\n",
      "Epoch 68/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0068 - custom_metric: 1.0329 - val_loss: 0.0100 - val_custom_metric: 1.3023 - lr: 0.0044\n",
      "Epoch 69/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0066 - custom_metric: 0.9926 - val_loss: 0.0104 - val_custom_metric: 1.4069 - lr: 0.0044\n",
      "Epoch 70/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0068 - custom_metric: 1.0074 - val_loss: 0.0090 - val_custom_metric: 1.2688 - lr: 0.0044\n",
      "Epoch 71/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0066 - custom_metric: 0.9788 - val_loss: 0.0091 - val_custom_metric: 1.3061 - lr: 0.0044\n",
      "Epoch 72/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0066 - custom_metric: 1.0177 - val_loss: 0.0091 - val_custom_metric: 1.3591 - lr: 0.0043\n",
      "Epoch 73/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0066 - custom_metric: 0.9718 - val_loss: 0.0112 - val_custom_metric: 1.4967 - lr: 0.0043\n",
      "Epoch 74/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0067 - custom_metric: 0.9823 - val_loss: 0.0100 - val_custom_metric: 1.4797 - lr: 0.0043\n",
      "Epoch 75/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0067 - custom_metric: 1.0964 - val_loss: 0.0110 - val_custom_metric: 1.5556 - lr: 0.0043\n",
      "Epoch 76/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0066 - custom_metric: 0.9773 - val_loss: 0.0099 - val_custom_metric: 1.3516 - lr: 0.0043\n",
      "Epoch 77/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0065 - custom_metric: 0.9641 - val_loss: 0.0102 - val_custom_metric: 1.4363 - lr: 0.0042\n",
      "Epoch 78/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0065 - custom_metric: 0.9600 - val_loss: 0.0103 - val_custom_metric: 1.3997 - lr: 0.0042\n",
      "Epoch 79/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0065 - custom_metric: 0.9357 - val_loss: 0.0104 - val_custom_metric: 1.4137 - lr: 0.0042\n",
      "Epoch 80/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0066 - custom_metric: 0.9945 - val_loss: 0.0099 - val_custom_metric: 1.3744 - lr: 0.0042\n",
      "Epoch 81/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0066 - custom_metric: 0.9439 - val_loss: 0.0112 - val_custom_metric: 1.4360 - lr: 0.0042\n",
      "Epoch 82/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0066 - custom_metric: 0.9479 - val_loss: 0.0100 - val_custom_metric: 1.3852 - lr: 0.0042\n",
      "Epoch 83/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0067 - custom_metric: 0.9548 - val_loss: 0.0092 - val_custom_metric: 1.2688 - lr: 0.0041\n",
      "Epoch 84/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0067 - custom_metric: 0.9681 - val_loss: 0.0101 - val_custom_metric: 1.4734 - lr: 0.0041\n",
      "Epoch 85/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0068 - custom_metric: 0.9845 - val_loss: 0.0093 - val_custom_metric: 1.2872 - lr: 0.0041\n",
      "Epoch 86/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0067 - custom_metric: 0.9737 - val_loss: 0.0112 - val_custom_metric: 1.5022 - lr: 0.0041\n",
      "Epoch 87/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0066 - custom_metric: 0.9582 - val_loss: 0.0106 - val_custom_metric: 1.4568 - lr: 0.0041\n",
      "Epoch 88/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0065 - custom_metric: 0.9467 - val_loss: 0.0109 - val_custom_metric: 1.4777 - lr: 0.0040\n",
      "Epoch 89/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0065 - custom_metric: 0.9420 - val_loss: 0.0108 - val_custom_metric: 1.5436 - lr: 0.0040\n",
      "Epoch 90/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0065 - custom_metric: 0.9342 - val_loss: 0.0095 - val_custom_metric: 1.2592 - lr: 0.0040\n",
      "Epoch 91/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0065 - custom_metric: 0.9383 - val_loss: 0.0104 - val_custom_metric: 1.4180 - lr: 0.0040\n",
      "Epoch 92/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0065 - custom_metric: 0.9384 - val_loss: 0.0107 - val_custom_metric: 1.4888 - lr: 0.0039\n",
      "Epoch 93/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0065 - custom_metric: 0.9269 - val_loss: 0.0104 - val_custom_metric: 1.4095 - lr: 0.0039\n",
      "Epoch 94/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0065 - custom_metric: 0.9315 - val_loss: 0.0092 - val_custom_metric: 1.2404 - lr: 0.0039\n",
      "Epoch 95/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0064 - custom_metric: 0.9207 - val_loss: 0.0094 - val_custom_metric: 1.3151 - lr: 0.0039\n",
      "Epoch 96/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0065 - custom_metric: 0.9237 - val_loss: 0.0096 - val_custom_metric: 1.3755 - lr: 0.0039\n",
      "Epoch 97/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0064 - custom_metric: 0.9173 - val_loss: 0.0100 - val_custom_metric: 1.3787 - lr: 0.0038\n",
      "Epoch 98/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0064 - custom_metric: 0.9150 - val_loss: 0.0093 - val_custom_metric: 1.3715 - lr: 0.0038\n",
      "Epoch 99/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0064 - custom_metric: 0.9111 - val_loss: 0.0108 - val_custom_metric: 1.4580 - lr: 0.0038\n",
      "Epoch 100/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0063 - custom_metric: 0.9080 - val_loss: 0.0105 - val_custom_metric: 1.4408 - lr: 0.0038\n",
      "Epoch 101/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0064 - custom_metric: 0.9119 - val_loss: 0.0105 - val_custom_metric: 1.4136 - lr: 0.0038\n",
      "Epoch 102/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0065 - custom_metric: 0.9300 - val_loss: 0.0106 - val_custom_metric: 1.3930 - lr: 0.0037\n",
      "Epoch 103/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0064 - custom_metric: 0.9073 - val_loss: 0.0106 - val_custom_metric: 1.4745 - lr: 0.0037\n",
      "Epoch 104/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0064 - custom_metric: 0.9109 - val_loss: 0.0109 - val_custom_metric: 1.4632 - lr: 0.0037\n",
      "Epoch 105/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0064 - custom_metric: 0.9162 - val_loss: 0.0106 - val_custom_metric: 1.4349 - lr: 0.0037\n",
      "Epoch 106/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0063 - custom_metric: 0.8950 - val_loss: 0.0112 - val_custom_metric: 1.5463 - lr: 0.0036\n",
      "Epoch 107/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0063 - custom_metric: 0.9023 - val_loss: 0.0100 - val_custom_metric: 1.3210 - lr: 0.0036\n",
      "Epoch 108/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0063 - custom_metric: 0.8993 - val_loss: 0.0105 - val_custom_metric: 1.3730 - lr: 0.0036\n",
      "Epoch 109/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0064 - custom_metric: 0.9091 - val_loss: 0.0100 - val_custom_metric: 1.3510 - lr: 0.0036\n",
      "Epoch 110/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0063 - custom_metric: 0.9052 - val_loss: 0.0107 - val_custom_metric: 1.4193 - lr: 0.0035\n",
      "Epoch 111/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0065 - custom_metric: 0.9400 - val_loss: 0.0108 - val_custom_metric: 1.4969 - lr: 0.0035\n",
      "Epoch 112/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0064 - custom_metric: 0.9208 - val_loss: 0.0115 - val_custom_metric: 1.5503 - lr: 0.0035\n",
      "Epoch 113/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0063 - custom_metric: 0.9051 - val_loss: 0.0112 - val_custom_metric: 1.5012 - lr: 0.0035\n",
      "Epoch 114/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0063 - custom_metric: 0.9064 - val_loss: 0.0102 - val_custom_metric: 1.3333 - lr: 0.0034\n",
      "Epoch 115/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8769 - val_loss: 0.0114 - val_custom_metric: 1.5211 - lr: 0.0034\n",
      "Epoch 116/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0062 - custom_metric: 0.8931 - val_loss: 0.0101 - val_custom_metric: 1.3336 - lr: 0.0034\n",
      "Epoch 117/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0062 - custom_metric: 0.8835 - val_loss: 0.0105 - val_custom_metric: 1.3510 - lr: 0.0034\n",
      "Epoch 118/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0063 - custom_metric: 0.9081 - val_loss: 0.0102 - val_custom_metric: 1.3493 - lr: 0.0033\n",
      "Epoch 119/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0063 - custom_metric: 0.9018 - val_loss: 0.0102 - val_custom_metric: 1.3934 - lr: 0.0033\n",
      "Epoch 120/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0062 - custom_metric: 0.8894 - val_loss: 0.0090 - val_custom_metric: 1.2024 - lr: 0.0033\n",
      "Epoch 121/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0063 - custom_metric: 0.9003 - val_loss: 0.0091 - val_custom_metric: 1.1466 - lr: 0.0033\n",
      "Epoch 122/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0063 - custom_metric: 0.8983 - val_loss: 0.0095 - val_custom_metric: 1.2876 - lr: 0.0032\n",
      "Epoch 123/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0065 - custom_metric: 0.9262 - val_loss: 0.0120 - val_custom_metric: 1.6334 - lr: 0.0032\n",
      "Epoch 124/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0063 - custom_metric: 0.8973 - val_loss: 0.0107 - val_custom_metric: 1.4875 - lr: 0.0032\n",
      "Epoch 125/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0063 - custom_metric: 0.9066 - val_loss: 0.0077 - val_custom_metric: 1.0176 - lr: 0.0032\n",
      "Epoch 126/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0064 - custom_metric: 0.9317 - val_loss: 0.0107 - val_custom_metric: 1.4877 - lr: 0.0031\n",
      "Epoch 127/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0067 - custom_metric: 0.9595 - val_loss: 0.0096 - val_custom_metric: 1.2742 - lr: 0.0031\n",
      "Epoch 128/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0067 - custom_metric: 0.9558 - val_loss: 0.0089 - val_custom_metric: 1.1213 - lr: 0.0031\n",
      "Epoch 129/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0067 - custom_metric: 0.9554 - val_loss: 0.0093 - val_custom_metric: 1.2016 - lr: 0.0031\n",
      "Epoch 130/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0066 - custom_metric: 0.9490 - val_loss: 0.0101 - val_custom_metric: 1.3774 - lr: 0.0030\n",
      "Epoch 131/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0065 - custom_metric: 0.9301 - val_loss: 0.0094 - val_custom_metric: 1.2420 - lr: 0.0030\n",
      "Epoch 132/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0064 - custom_metric: 0.9170 - val_loss: 0.0091 - val_custom_metric: 1.1241 - lr: 0.0030\n",
      "Epoch 133/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0064 - custom_metric: 0.9218 - val_loss: 0.0094 - val_custom_metric: 1.2555 - lr: 0.0030\n",
      "Epoch 134/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0064 - custom_metric: 0.9235 - val_loss: 0.0104 - val_custom_metric: 1.3710 - lr: 0.0029\n",
      "Epoch 135/300\n",
      "54/54 [==============================] - 6s 119ms/step - loss: 0.0064 - custom_metric: 0.9199 - val_loss: 0.0083 - val_custom_metric: 1.1191 - lr: 0.0029\n",
      "Epoch 136/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0064 - custom_metric: 0.9118 - val_loss: 0.0095 - val_custom_metric: 1.3042 - lr: 0.0029\n",
      "Epoch 137/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0063 - custom_metric: 0.9101 - val_loss: 0.0102 - val_custom_metric: 1.3679 - lr: 0.0029\n",
      "Epoch 138/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0064 - custom_metric: 0.9154 - val_loss: 0.0096 - val_custom_metric: 1.3649 - lr: 0.0028\n",
      "Epoch 139/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0063 - custom_metric: 0.8982 - val_loss: 0.0099 - val_custom_metric: 1.3510 - lr: 0.0028\n",
      "Epoch 140/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0062 - custom_metric: 0.8943 - val_loss: 0.0097 - val_custom_metric: 1.3209 - lr: 0.0028\n",
      "Epoch 141/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0063 - custom_metric: 0.9000 - val_loss: 0.0089 - val_custom_metric: 1.1123 - lr: 0.0028\n",
      "Epoch 142/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0062 - custom_metric: 0.8928 - val_loss: 0.0088 - val_custom_metric: 1.1222 - lr: 0.0027\n",
      "Epoch 143/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0062 - custom_metric: 0.8888 - val_loss: 0.0091 - val_custom_metric: 1.1939 - lr: 0.0027\n",
      "Epoch 144/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0062 - custom_metric: 0.8940 - val_loss: 0.0093 - val_custom_metric: 1.2108 - lr: 0.0027\n",
      "Epoch 145/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0062 - custom_metric: 0.8893 - val_loss: 0.0088 - val_custom_metric: 1.1508 - lr: 0.0027\n",
      "Epoch 146/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8722 - val_loss: 0.0108 - val_custom_metric: 1.3881 - lr: 0.0026\n",
      "Epoch 147/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8855 - val_loss: 0.0094 - val_custom_metric: 1.2554 - lr: 0.0026\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 148/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8850 - val_loss: 0.0098 - val_custom_metric: 1.2230 - lr: 0.0026\n",
      "Epoch 149/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8777 - val_loss: 0.0100 - val_custom_metric: 1.2972 - lr: 0.0026\n",
      "Epoch 150/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0062 - custom_metric: 0.8875 - val_loss: 0.0101 - val_custom_metric: 1.3645 - lr: 0.0025\n",
      "Epoch 151/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0062 - custom_metric: 0.8914 - val_loss: 0.0097 - val_custom_metric: 1.2683 - lr: 0.0025\n",
      "Epoch 152/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0062 - custom_metric: 0.8973 - val_loss: 0.0102 - val_custom_metric: 1.3802 - lr: 0.0025\n",
      "Epoch 153/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8802 - val_loss: 0.0103 - val_custom_metric: 1.3633 - lr: 0.0024\n",
      "Epoch 154/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0062 - custom_metric: 0.8909 - val_loss: 0.0112 - val_custom_metric: 1.5008 - lr: 0.0024\n",
      "Epoch 155/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0061 - custom_metric: 0.8809 - val_loss: 0.0107 - val_custom_metric: 1.4729 - lr: 0.0024\n",
      "Epoch 156/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0061 - custom_metric: 0.8847 - val_loss: 0.0101 - val_custom_metric: 1.3865 - lr: 0.0024\n",
      "Epoch 157/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0061 - custom_metric: 0.8840 - val_loss: 0.0096 - val_custom_metric: 1.3243 - lr: 0.0023\n",
      "Epoch 158/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0060 - custom_metric: 0.8631 - val_loss: 0.0090 - val_custom_metric: 1.2766 - lr: 0.0023\n",
      "Epoch 159/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8805 - val_loss: 0.0099 - val_custom_metric: 1.2911 - lr: 0.0023\n",
      "Epoch 160/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0060 - custom_metric: 0.8672 - val_loss: 0.0083 - val_custom_metric: 1.1481 - lr: 0.0023\n",
      "Epoch 161/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0060 - custom_metric: 0.8610 - val_loss: 0.0097 - val_custom_metric: 1.2619 - lr: 0.0022\n",
      "Epoch 162/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0060 - custom_metric: 0.8689 - val_loss: 0.0089 - val_custom_metric: 1.2342 - lr: 0.0022\n",
      "Epoch 163/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0059 - custom_metric: 0.8551 - val_loss: 0.0095 - val_custom_metric: 1.2157 - lr: 0.0022\n",
      "Epoch 164/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0060 - custom_metric: 0.8660 - val_loss: 0.0094 - val_custom_metric: 1.2468 - lr: 0.0022\n",
      "Epoch 165/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0060 - custom_metric: 0.8678 - val_loss: 0.0094 - val_custom_metric: 1.2822 - lr: 0.0021\n",
      "Epoch 166/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0061 - custom_metric: 0.8759 - val_loss: 0.0088 - val_custom_metric: 1.1932 - lr: 0.0021\n",
      "Epoch 167/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0061 - custom_metric: 0.8867 - val_loss: 0.0092 - val_custom_metric: 1.2382 - lr: 0.0021\n",
      "Epoch 168/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0059 - custom_metric: 0.8654 - val_loss: 0.0092 - val_custom_metric: 1.2870 - lr: 0.0021\n",
      "Epoch 169/300\n",
      "54/54 [==============================] - 6s 112ms/step - loss: 0.0058 - custom_metric: 0.8453 - val_loss: 0.0095 - val_custom_metric: 1.2668 - lr: 0.0020\n",
      "Epoch 170/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0059 - custom_metric: 0.8604 - val_loss: 0.0104 - val_custom_metric: 1.3421 - lr: 0.0020\n",
      "Epoch 171/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0060 - custom_metric: 0.8669 - val_loss: 0.0095 - val_custom_metric: 1.2244 - lr: 0.0020\n",
      "Epoch 172/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0059 - custom_metric: 0.8546 - val_loss: 0.0090 - val_custom_metric: 1.1993 - lr: 0.0020\n",
      "Epoch 173/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0059 - custom_metric: 0.8477 - val_loss: 0.0086 - val_custom_metric: 1.1729 - lr: 0.0019\n",
      "Epoch 174/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0058 - custom_metric: 0.8420 - val_loss: 0.0076 - val_custom_metric: 0.9471 - lr: 0.0019\n",
      "Epoch 175/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0059 - custom_metric: 0.8533 - val_loss: 0.0090 - val_custom_metric: 1.1662 - lr: 0.0019\n",
      "Epoch 176/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0059 - custom_metric: 0.8513 - val_loss: 0.0077 - val_custom_metric: 0.9521 - lr: 0.0019\n",
      "Epoch 177/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0058 - custom_metric: 0.8487 - val_loss: 0.0081 - val_custom_metric: 1.0645 - lr: 0.0018\n",
      "Epoch 178/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0058 - custom_metric: 0.8365 - val_loss: 0.0079 - val_custom_metric: 1.0489 - lr: 0.0018\n",
      "Epoch 179/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0058 - custom_metric: 0.8377 - val_loss: 0.0079 - val_custom_metric: 1.0691 - lr: 0.0018\n",
      "Epoch 180/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0057 - custom_metric: 0.8320 - val_loss: 0.0076 - val_custom_metric: 0.9333 - lr: 0.0018\n",
      "Epoch 181/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0057 - custom_metric: 0.8345 - val_loss: 0.0080 - val_custom_metric: 1.0030 - lr: 0.0017\n",
      "Epoch 182/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0057 - custom_metric: 0.8262 - val_loss: 0.0075 - val_custom_metric: 0.9451 - lr: 0.0017\n",
      "Epoch 183/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0057 - custom_metric: 0.8298 - val_loss: 0.0086 - val_custom_metric: 1.1032 - lr: 0.0017\n",
      "Epoch 184/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0058 - custom_metric: 0.8342 - val_loss: 0.0084 - val_custom_metric: 1.0957 - lr: 0.0017\n",
      "Epoch 185/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0057 - custom_metric: 0.8308 - val_loss: 0.0079 - val_custom_metric: 1.0050 - lr: 0.0016\n",
      "Epoch 186/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0056 - custom_metric: 0.8283 - val_loss: 0.0078 - val_custom_metric: 1.1329 - lr: 0.0016\n",
      "Epoch 187/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0056 - custom_metric: 0.8180 - val_loss: 0.0084 - val_custom_metric: 1.1139 - lr: 0.0016\n",
      "Epoch 188/300\n",
      "54/54 [==============================] - 6s 119ms/step - loss: 0.0056 - custom_metric: 0.8206 - val_loss: 0.0082 - val_custom_metric: 1.0816 - lr: 0.0016\n",
      "Epoch 189/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0056 - custom_metric: 0.8203 - val_loss: 0.0074 - val_custom_metric: 0.9719 - lr: 0.0015\n",
      "Epoch 190/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0057 - custom_metric: 0.8282 - val_loss: 0.0083 - val_custom_metric: 1.1336 - lr: 0.0015\n",
      "Epoch 191/300\n",
      "54/54 [==============================] - 6s 112ms/step - loss: 0.0056 - custom_metric: 0.8205 - val_loss: 0.0082 - val_custom_metric: 1.0427 - lr: 0.0015\n",
      "Epoch 192/300\n",
      "54/54 [==============================] - 6s 120ms/step - loss: 0.0056 - custom_metric: 0.8155 - val_loss: 0.0074 - val_custom_metric: 0.8778 - lr: 0.0015\n",
      "Epoch 193/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0056 - custom_metric: 0.8192 - val_loss: 0.0074 - val_custom_metric: 0.9194 - lr: 0.0014\n",
      "Epoch 194/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0056 - custom_metric: 0.8050 - val_loss: 0.0073 - val_custom_metric: 0.8622 - lr: 0.0014\n",
      "Epoch 195/300\n",
      "54/54 [==============================] - 7s 121ms/step - loss: 0.0056 - custom_metric: 0.8140 - val_loss: 0.0080 - val_custom_metric: 1.0107 - lr: 0.0014\n",
      "Epoch 196/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0056 - custom_metric: 0.8200 - val_loss: 0.0085 - val_custom_metric: 1.1509 - lr: 0.0014\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 197/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0054 - custom_metric: 0.7979 - val_loss: 0.0087 - val_custom_metric: 1.1431 - lr: 0.0013\n",
      "Epoch 198/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0055 - custom_metric: 0.8111 - val_loss: 0.0074 - val_custom_metric: 0.9205 - lr: 0.0013\n",
      "Epoch 199/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0055 - custom_metric: 0.8106 - val_loss: 0.0078 - val_custom_metric: 0.9303 - lr: 0.0013\n",
      "Epoch 200/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0054 - custom_metric: 0.7931 - val_loss: 0.0085 - val_custom_metric: 1.1663 - lr: 0.0013\n",
      "Epoch 201/300\n",
      "54/54 [==============================] - 6s 120ms/step - loss: 0.0055 - custom_metric: 0.8064 - val_loss: 0.0085 - val_custom_metric: 1.1330 - lr: 0.0013\n",
      "Epoch 202/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0055 - custom_metric: 0.8022 - val_loss: 0.0087 - val_custom_metric: 1.1571 - lr: 0.0012\n",
      "Epoch 203/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0054 - custom_metric: 0.7966 - val_loss: 0.0088 - val_custom_metric: 1.2088 - lr: 0.0012\n",
      "Epoch 204/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0054 - custom_metric: 0.7967 - val_loss: 0.0075 - val_custom_metric: 0.9437 - lr: 0.0012\n",
      "Epoch 205/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0054 - custom_metric: 0.7951 - val_loss: 0.0076 - val_custom_metric: 1.0071 - lr: 0.0012\n",
      "Epoch 206/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0054 - custom_metric: 0.7946 - val_loss: 0.0079 - val_custom_metric: 0.9656 - lr: 0.0011\n",
      "Epoch 207/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0053 - custom_metric: 0.7827 - val_loss: 0.0081 - val_custom_metric: 1.0159 - lr: 0.0011\n",
      "Epoch 208/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0054 - custom_metric: 0.7957 - val_loss: 0.0080 - val_custom_metric: 1.0022 - lr: 0.0011\n",
      "Epoch 209/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0054 - custom_metric: 0.7833 - val_loss: 0.0077 - val_custom_metric: 0.9696 - lr: 0.0011\n",
      "Epoch 210/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0053 - custom_metric: 0.7833 - val_loss: 0.0075 - val_custom_metric: 0.9069 - lr: 0.0011\n",
      "Epoch 211/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0053 - custom_metric: 0.7713 - val_loss: 0.0073 - val_custom_metric: 0.8844 - lr: 0.0010\n",
      "Epoch 212/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0053 - custom_metric: 0.7696 - val_loss: 0.0072 - val_custom_metric: 0.8558 - lr: 0.0010\n",
      "Epoch 213/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0052 - custom_metric: 0.7630 - val_loss: 0.0080 - val_custom_metric: 1.0726 - lr: 9.8890e-04\n",
      "Epoch 214/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0053 - custom_metric: 0.7785 - val_loss: 0.0078 - val_custom_metric: 0.9773 - lr: 9.6814e-04\n",
      "Epoch 215/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0052 - custom_metric: 0.7660 - val_loss: 0.0075 - val_custom_metric: 0.9053 - lr: 9.4754e-04\n",
      "Epoch 216/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0052 - custom_metric: 0.7638 - val_loss: 0.0074 - val_custom_metric: 0.8746 - lr: 9.2711e-04\n",
      "Epoch 217/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0052 - custom_metric: 0.7716 - val_loss: 0.0073 - val_custom_metric: 0.8812 - lr: 9.0685e-04\n",
      "Epoch 218/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0051 - custom_metric: 0.7550 - val_loss: 0.0074 - val_custom_metric: 0.9369 - lr: 8.8677e-04\n",
      "Epoch 219/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0052 - custom_metric: 0.7602 - val_loss: 0.0077 - val_custom_metric: 1.0234 - lr: 8.6686e-04\n",
      "Epoch 220/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0052 - custom_metric: 0.7620 - val_loss: 0.0078 - val_custom_metric: 0.9933 - lr: 8.4714e-04\n",
      "Epoch 221/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0051 - custom_metric: 0.7506 - val_loss: 0.0076 - val_custom_metric: 0.9407 - lr: 8.2759e-04\n",
      "Epoch 222/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0051 - custom_metric: 0.7560 - val_loss: 0.0076 - val_custom_metric: 0.9344 - lr: 8.0823e-04\n",
      "Epoch 223/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0050 - custom_metric: 0.7423 - val_loss: 0.0078 - val_custom_metric: 0.9546 - lr: 7.8905e-04\n",
      "Epoch 224/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0051 - custom_metric: 0.7521 - val_loss: 0.0079 - val_custom_metric: 1.0512 - lr: 7.7006e-04\n",
      "Epoch 225/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0052 - custom_metric: 0.7593 - val_loss: 0.0076 - val_custom_metric: 0.9646 - lr: 7.5127e-04\n",
      "Epoch 226/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0052 - custom_metric: 0.7640 - val_loss: 0.0074 - val_custom_metric: 0.8780 - lr: 7.3266e-04\n",
      "Epoch 227/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0052 - custom_metric: 0.7628 - val_loss: 0.0074 - val_custom_metric: 0.9031 - lr: 7.1425e-04\n",
      "Epoch 228/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0050 - custom_metric: 0.7418 - val_loss: 0.0076 - val_custom_metric: 0.9045 - lr: 6.9603e-04\n",
      "Epoch 229/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0051 - custom_metric: 0.7442 - val_loss: 0.0074 - val_custom_metric: 0.9198 - lr: 6.7801e-04\n",
      "Epoch 230/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0050 - custom_metric: 0.7375 - val_loss: 0.0074 - val_custom_metric: 0.9126 - lr: 6.6019e-04\n",
      "Epoch 231/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0050 - custom_metric: 0.7398 - val_loss: 0.0073 - val_custom_metric: 0.8751 - lr: 6.4257e-04\n",
      "Epoch 232/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0050 - custom_metric: 0.7369 - val_loss: 0.0074 - val_custom_metric: 0.9082 - lr: 6.2516e-04\n",
      "Epoch 233/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0050 - custom_metric: 0.7381 - val_loss: 0.0073 - val_custom_metric: 0.9338 - lr: 6.0795e-04\n",
      "Epoch 234/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0050 - custom_metric: 0.7349 - val_loss: 0.0074 - val_custom_metric: 1.0060 - lr: 5.9095e-04\n",
      "Epoch 235/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0049 - custom_metric: 0.7316 - val_loss: 0.0078 - val_custom_metric: 0.9909 - lr: 5.7416e-04\n",
      "Epoch 236/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0050 - custom_metric: 0.7309 - val_loss: 0.0077 - val_custom_metric: 0.9873 - lr: 5.5758e-04\n",
      "Epoch 237/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0050 - custom_metric: 0.7407 - val_loss: 0.0073 - val_custom_metric: 0.9202 - lr: 5.4121e-04\n",
      "Epoch 238/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0050 - custom_metric: 0.7334 - val_loss: 0.0072 - val_custom_metric: 0.9022 - lr: 5.2506e-04\n",
      "Epoch 239/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0049 - custom_metric: 0.7374 - val_loss: 0.0073 - val_custom_metric: 0.9463 - lr: 5.0912e-04\n",
      "Epoch 240/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0050 - custom_metric: 0.7362 - val_loss: 0.0077 - val_custom_metric: 1.0234 - lr: 4.9341e-04\n",
      "Epoch 241/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0049 - custom_metric: 0.7319 - val_loss: 0.0073 - val_custom_metric: 0.9558 - lr: 4.7791e-04\n",
      "Epoch 242/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0048 - custom_metric: 0.7207 - val_loss: 0.0078 - val_custom_metric: 1.0495 - lr: 4.6263e-04\n",
      "Epoch 243/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0048 - custom_metric: 0.7180 - val_loss: 0.0074 - val_custom_metric: 0.9400 - lr: 4.4758e-04\n",
      "Epoch 244/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0048 - custom_metric: 0.7190 - val_loss: 0.0077 - val_custom_metric: 1.0281 - lr: 4.3276e-04\n",
      "Epoch 245/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0048 - custom_metric: 0.7159 - val_loss: 0.0079 - val_custom_metric: 1.0462 - lr: 4.1816e-04\n",
      "Epoch 246/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0048 - custom_metric: 0.7176 - val_loss: 0.0079 - val_custom_metric: 1.0452 - lr: 4.0378e-04\n",
      "Epoch 247/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0048 - custom_metric: 0.7208 - val_loss: 0.0079 - val_custom_metric: 1.0569 - lr: 3.8964e-04\n",
      "Epoch 248/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0048 - custom_metric: 0.7148 - val_loss: 0.0079 - val_custom_metric: 1.0553 - lr: 3.7573e-04\n",
      "Epoch 249/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0048 - custom_metric: 0.7137 - val_loss: 0.0076 - val_custom_metric: 1.0145 - lr: 3.6205e-04\n",
      "Epoch 250/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0047 - custom_metric: 0.7056 - val_loss: 0.0077 - val_custom_metric: 1.0490 - lr: 3.4861e-04\n",
      "Epoch 251/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0048 - custom_metric: 0.7095 - val_loss: 0.0074 - val_custom_metric: 0.9554 - lr: 3.3540e-04\n",
      "Epoch 252/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0048 - custom_metric: 0.7113 - val_loss: 0.0073 - val_custom_metric: 0.9432 - lr: 3.2243e-04\n",
      "Epoch 253/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0048 - custom_metric: 0.7116 - val_loss: 0.0072 - val_custom_metric: 0.9092 - lr: 3.0970e-04\n",
      "Epoch 254/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0048 - custom_metric: 0.7186 - val_loss: 0.0074 - val_custom_metric: 0.9593 - lr: 2.9721e-04\n",
      "Epoch 255/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0047 - custom_metric: 0.7017 - val_loss: 0.0075 - val_custom_metric: 0.9489 - lr: 2.8496e-04\n",
      "Epoch 256/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0048 - custom_metric: 0.7116 - val_loss: 0.0073 - val_custom_metric: 0.9280 - lr: 2.7296e-04\n",
      "Epoch 257/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0047 - custom_metric: 0.7076 - val_loss: 0.0075 - val_custom_metric: 0.9770 - lr: 2.6119e-04\n",
      "Epoch 258/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0047 - custom_metric: 0.7032 - val_loss: 0.0076 - val_custom_metric: 1.0383 - lr: 2.4968e-04\n",
      "Epoch 259/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0047 - custom_metric: 0.7012 - val_loss: 0.0077 - val_custom_metric: 1.0446 - lr: 2.3841e-04\n",
      "Epoch 260/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0048 - custom_metric: 0.7074 - val_loss: 0.0076 - val_custom_metric: 1.0381 - lr: 2.2739e-04\n",
      "Epoch 261/300\n",
      "54/54 [==============================] - 7s 124ms/step - loss: 0.0047 - custom_metric: 0.6997 - val_loss: 0.0075 - val_custom_metric: 1.0533 - lr: 2.1661e-04\n",
      "Epoch 262/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0047 - custom_metric: 0.6980 - val_loss: 0.0072 - val_custom_metric: 0.9307 - lr: 2.0609e-04\n",
      "Epoch 263/300\n",
      "54/54 [==============================] - 6s 119ms/step - loss: 0.0047 - custom_metric: 0.7053 - val_loss: 0.0073 - val_custom_metric: 0.9720 - lr: 1.9582e-04\n",
      "Epoch 264/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0046 - custom_metric: 0.6966 - val_loss: 0.0071 - val_custom_metric: 0.9294 - lr: 1.8580e-04\n",
      "Epoch 265/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0046 - custom_metric: 0.6951 - val_loss: 0.0071 - val_custom_metric: 0.8918 - lr: 1.7604e-04\n",
      "Epoch 266/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0047 - custom_metric: 0.6954 - val_loss: 0.0071 - val_custom_metric: 0.9073 - lr: 1.6653e-04\n",
      "Epoch 267/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0047 - custom_metric: 0.7005 - val_loss: 0.0071 - val_custom_metric: 0.9083 - lr: 1.5728e-04\n",
      "Epoch 268/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0047 - custom_metric: 0.6987 - val_loss: 0.0071 - val_custom_metric: 0.9109 - lr: 1.4828e-04\n",
      "Epoch 269/300\n",
      "54/54 [==============================] - 6s 118ms/step - loss: 0.0047 - custom_metric: 0.7030 - val_loss: 0.0071 - val_custom_metric: 0.9197 - lr: 1.3955e-04\n",
      "Epoch 270/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0047 - custom_metric: 0.6999 - val_loss: 0.0071 - val_custom_metric: 0.9160 - lr: 1.3107e-04\n",
      "Epoch 271/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0046 - custom_metric: 0.6953 - val_loss: 0.0072 - val_custom_metric: 0.9189 - lr: 1.2285e-04\n",
      "Epoch 272/300\n",
      "54/54 [==============================] - 6s 119ms/step - loss: 0.0046 - custom_metric: 0.6948 - val_loss: 0.0072 - val_custom_metric: 0.9001 - lr: 1.1489e-04\n",
      "Epoch 273/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0046 - custom_metric: 0.6932 - val_loss: 0.0072 - val_custom_metric: 0.9074 - lr: 1.0719e-04\n",
      "Epoch 274/300\n",
      "54/54 [==============================] - 6s 119ms/step - loss: 0.0047 - custom_metric: 0.7017 - val_loss: 0.0071 - val_custom_metric: 0.9144 - lr: 9.9756e-05\n",
      "Epoch 275/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0047 - custom_metric: 0.6945 - val_loss: 0.0071 - val_custom_metric: 0.9239 - lr: 9.2584e-05\n",
      "Epoch 276/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0046 - custom_metric: 0.6966 - val_loss: 0.0071 - val_custom_metric: 0.9232 - lr: 8.5677e-05\n",
      "Epoch 277/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0046 - custom_metric: 0.6844 - val_loss: 0.0071 - val_custom_metric: 0.9155 - lr: 7.9034e-05\n",
      "Epoch 278/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0046 - custom_metric: 0.6888 - val_loss: 0.0071 - val_custom_metric: 0.9257 - lr: 7.2657e-05\n",
      "Epoch 279/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6842 - val_loss: 0.0071 - val_custom_metric: 0.9300 - lr: 6.6546e-05\n",
      "Epoch 280/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6906 - val_loss: 0.0071 - val_custom_metric: 0.9313 - lr: 6.0702e-05\n",
      "Epoch 281/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6906 - val_loss: 0.0071 - val_custom_metric: 0.9318 - lr: 5.5125e-05\n",
      "Epoch 282/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0046 - custom_metric: 0.6892 - val_loss: 0.0071 - val_custom_metric: 0.9232 - lr: 4.9817e-05\n",
      "Epoch 283/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0046 - custom_metric: 0.6895 - val_loss: 0.0071 - val_custom_metric: 0.9232 - lr: 4.4777e-05\n",
      "Epoch 284/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6894 - val_loss: 0.0071 - val_custom_metric: 0.9316 - lr: 4.0007e-05\n",
      "Epoch 285/300\n",
      "54/54 [==============================] - 6s 117ms/step - loss: 0.0046 - custom_metric: 0.6914 - val_loss: 0.0071 - val_custom_metric: 0.9234 - lr: 3.5506e-05\n",
      "Epoch 286/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6838 - val_loss: 0.0071 - val_custom_metric: 0.9222 - lr: 3.1276e-05\n",
      "Epoch 287/300\n",
      "54/54 [==============================] - 6s 112ms/step - loss: 0.0046 - custom_metric: 0.6925 - val_loss: 0.0071 - val_custom_metric: 0.9155 - lr: 2.7316e-05\n",
      "Epoch 288/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0045 - custom_metric: 0.6821 - val_loss: 0.0071 - val_custom_metric: 0.9160 - lr: 2.3628e-05\n",
      "Epoch 289/300\n",
      "54/54 [==============================] - 6s 115ms/step - loss: 0.0046 - custom_metric: 0.6887 - val_loss: 0.0071 - val_custom_metric: 0.9088 - lr: 2.0211e-05\n",
      "Epoch 290/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6833 - val_loss: 0.0071 - val_custom_metric: 0.9108 - lr: 1.7066e-05\n",
      "Epoch 291/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0046 - custom_metric: 0.6835 - val_loss: 0.0071 - val_custom_metric: 0.9106 - lr: 1.4194e-05\n",
      "Epoch 292/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6859 - val_loss: 0.0071 - val_custom_metric: 0.9170 - lr: 1.1594e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 293/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6892 - val_loss: 0.0071 - val_custom_metric: 0.9060 - lr: 9.2670e-06\n",
      "Epoch 294/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0046 - custom_metric: 0.6824 - val_loss: 0.0071 - val_custom_metric: 0.9059 - lr: 7.2132e-06\n",
      "Epoch 295/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6874 - val_loss: 0.0071 - val_custom_metric: 0.9117 - lr: 5.4327e-06\n",
      "Epoch 296/300\n",
      "54/54 [==============================] - 6s 113ms/step - loss: 0.0046 - custom_metric: 0.6871 - val_loss: 0.0071 - val_custom_metric: 0.9106 - lr: 3.9259e-06\n",
      "Epoch 297/300\n",
      "54/54 [==============================] - 6s 116ms/step - loss: 0.0045 - custom_metric: 0.6791 - val_loss: 0.0071 - val_custom_metric: 0.9167 - lr: 2.6926e-06\n",
      "Epoch 298/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6840 - val_loss: 0.0071 - val_custom_metric: 0.9169 - lr: 1.7335e-06\n",
      "Epoch 299/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0045 - custom_metric: 0.6808 - val_loss: 0.0071 - val_custom_metric: 0.9209 - lr: 1.0483e-06\n",
      "Epoch 300/300\n",
      "54/54 [==============================] - 6s 114ms/step - loss: 0.0046 - custom_metric: 0.6880 - val_loss: 0.0071 - val_custom_metric: 0.9197 - lr: 6.3708e-07\n"
     ]
    }
   ],
   "source": [
    "History = model.fit(train_data,\n",
    "                    epochs=epochs,\n",
    "                    callbacks=[lr_callback],\n",
    "                    validation_data=val_data,\n",
    "                    validation_steps=1,\n",
    "                    verbose=1)\n",
    "\n",
    "#Export trained model\n",
    "model.save(saved_model_filepath)\n",
    "model.save(h5_filepath)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9u7hfBLSkBAS"
   },
   "source": [
    "# Inference"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zTZuJyUvkWgZ"
   },
   "source": [
    "Test set preprocessing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "executionInfo": {
     "elapsed": 207,
     "status": "ok",
     "timestamp": 1679259163288,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "RphZPeJsgiwB"
   },
   "outputs": [],
   "source": [
    "test_data = load_tf_records(test_set_path).map(decode_dataset_test, num_parallel_calls=AUTO)\n",
    "\n",
    "test_data = test_data.map(lambda filename, patch, height, width: normalize_test(filename, patch, height, width, max_reflectance), num_parallel_calls=AUTO)    #normalize test dataset\n",
    "test_data = test_data.map(lambda filename, patch, height, width: preprocess_test(filename, patch, height, width,target_image_size), num_parallel_calls=AUTO).batch(1).prefetch(AUTO)    #batch and prefetch test dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DTwygR8ekZE5"
   },
   "source": [
    "Load exported model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "executionInfo": {
     "elapsed": 1989,
     "status": "ok",
     "timestamp": 1679259168969,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "MuoQ5alMkYU5"
   },
   "outputs": [],
   "source": [
    "loaded_model = tf.keras.models.load_model(saved_model_filepath, custom_objects={'custom_metric': custom_metric})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "D6DT5L5zlyTB"
   },
   "source": [
    "Make inference and generate submission file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "executionInfo": {
     "elapsed": 32452,
     "status": "ok",
     "timestamp": 1679260146955,
     "user": {
      "displayName": "Sistemi Spaziali",
      "userId": "03880471792653564924"
     },
     "user_tz": -60
    },
    "id": "OKgGkloPl1o5"
   },
   "outputs": [],
   "source": [
    "filenames = np.array([],dtype=int)\n",
    "export_predictions =  []\n",
    "\n",
    "for filename, image, height, width in test_data:\n",
    "  filenames = np.append(filenames, int(filename.numpy()[0].decode().replace(\".npz\", \"\")))\n",
    "  predictions = loaded_model.predict(image,verbose=0)\n",
    "  \n",
    "  if label_normalization_mode == 0:\n",
    "        \n",
    "     predictions *= max_labels\n",
    "    \n",
    "  else:\n",
    "     predictions  = (predictions*std_labels)+mean_labels\n",
    "        \n",
    "  export_predictions.append(predictions.reshape([-1]))\n",
    "\n",
    "submission = pd.DataFrame(data=export_predictions, columns=[\"P\", \"K\", \"Mg\", \"pH\"])\n",
    "submission.index = filenames\n",
    "submission.to_csv(submission_name, index_label=\"sample_index\")\n",
    "\n",
    "with zipfile.ZipFile(submission_zip, mode='w') as zipf:\n",
    "    zipf.write(submission_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "authorship_tag": "ABX9TyOaSxuzRRg6XkPDpDv1q+3u",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}