3_keras_mnist_GAN.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Copy of mnist_gan.ipynb",
      "provenance": [],
      "gpuType": "T4",
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/peterjsadowski/keras_tutorial/blob/master/3_keras_mnist_GAN.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jZ5Mkwyxz-yF"
      },
      "source": [
        "# Keras Tutorial for Generative Adversarial Nets\n",
        "This tutorial introduces Generative Adversarial Networks on MNIST.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cUgCKtD4z-yG"
      },
      "source": [
        "import numpy as np\n",
        "import os\n",
        "import numpy.random\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from collections import defaultdict\n",
        "from tqdm import tqdm\n",
        "from IPython import display\n",
        "\n",
        "import keras\n",
        "from keras.datasets import mnist\n",
        "from keras.models import Model, Sequential\n",
        "from keras.layers import Dense, Dropout, Activation, Input, Convolution2D, Flatten, BatchNormalization\n",
        "from keras.utils import to_categorical\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras.optimizers import SGD,Adam\n"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TVd9otiYz-yM"
      },
      "source": [
        "## Define data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bseTbVObz-yO",
        "outputId": "9e83a182-e29d-4db0-ea97-c2d4e269cdeb",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "# Load data.\n",
        "(X_train, Y_train), (X_test, Y_test) = mnist.load_data()\n",
        "print(\"Original X shape\", X_train.shape)\n",
        "print(\"Original Y shape\", Y_train.shape)\n",
        "\n",
        "# Reshape data.\n",
        "X_train = X_train.reshape(60000, 784)\n",
        "X_test  = X_test.reshape(10000, 784)\n",
        "X_train = X_train.astype('float32')\n",
        "X_test  = X_test.astype('float32')\n",
        "X_train /= 255 # Original data is uint8 (0-255). Scale it to range [0,1].\n",
        "X_test  /= 255\n",
        "print(\"Training X matrix shape\", X_train.shape)\n",
        "print(\"Testing X matrix shape\", X_test.shape)\n",
        "\n",
        "# Represent the targets as one-hot vectors: e.g. 2 -> [0, 0, 1, 0, 0, 0, 0, 0, 0].\n",
        "nb_classes = 10\n",
        "Y_train = to_categorical(Y_train, nb_classes)\n",
        "Y_test  = to_categorical(Y_test, nb_classes)\n",
        "print(\"Training Y matrix shape\", Y_train.shape)\n",
        "print(\"Testing Y matrix shape\", Y_test.shape)"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "\u001b[1m11490434/11490434\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 0us/step\n",
            "Original X shape (60000, 28, 28)\n",
            "Original Y shape (60000,)\n",
            "Training X matrix shape (60000, 784)\n",
            "Testing X matrix shape (10000, 784)\n",
            "Training Y matrix shape (60000, 10)\n",
            "Testing Y matrix shape (10000, 10)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fZirAYc7z-yT"
      },
      "source": [
        "# Define Keras Model\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YKugZxVnz-yT"
      },
      "source": [
        "# Model\n",
        "def define_generator(ndense=1, nhid=100, learning_rate=1e-3, act='relu', mom=0.9, dropout=0.0, input_shape=20):\n",
        "    # Define 'generator' that predicts sig vs bg from features.\n",
        "    input_shape = input_shape # Number of inputs to generator (random numbers)\n",
        "    output_shape= 784\n",
        "    ndense      = 2\n",
        "    nhid        = 100\n",
        "    act         = 'relu'\n",
        "    input       = Input(shape=(input_shape,), name='input')\n",
        "    x           = input\n",
        "    for i in range(ndense):\n",
        "        x = Dense(units=nhid, activation=act)(x)\n",
        "        x = BatchNormalization()(x)\n",
        "    output    = Dense(units=output_shape, activation='sigmoid', name='output')(x)\n",
        "    generator = Model(input, output)\n",
        "    opt       = Adam(learning_rate=learning_rate, beta_1=mom)\n",
        "    #generator.compile(loss='binary_crossentropy', optimizer=opt)\n",
        "    return generator\n",
        "\n",
        "def define_discriminator(ndense=1, nhid=100, act='relu', learning_rate=1e-3, mom=0.9, decay=0.0, dropout=0.0):\n",
        "    # Define 'discriminator' that predicts whether input is generated vs. real.\n",
        "    input_shape = 784\n",
        "    output_shape= 1\n",
        "    ndense      = ndense\n",
        "    nhid        = nhid\n",
        "    act         = act\n",
        "    input       = Input(shape=(input_shape,))\n",
        "    x           = input\n",
        "    for i in range(ndense):\n",
        "        x = Dense(units=nhid, activation=act)(x)\n",
        "    output = Dense(units=output_shape, activation='sigmoid')(x)\n",
        "    discriminator = Model(input, output)\n",
        "    #opt       = Adam(lr=lr, beta_1=mom)\n",
        "    opt = SGD(learning_rate=learning_rate,momentum=mom,decay=decay)\n",
        "    discriminator.compile(loss='binary_crossentropy', optimizer=opt)\n",
        "    return discriminator\n",
        "\n",
        "def plot_loss(history):\n",
        "    plt.figure(1)\n",
        "    plt.clf()\n",
        "    for k in history:\n",
        "        plt.plot(history[k], label=k)\n",
        "    plt.legend()\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.ylabel('')\n",
        "    #plt.show()\n",
        "\n",
        "def plot_gen(generator, input_shape):\n",
        "    noisebatch = np.random.uniform(-1, 1, size=[10, input_shape])\n",
        "    generated  = generator.predict(noisebatch)\n",
        "    plt.figure(2, figsize=(14,5))\n",
        "    plt.clf()\n",
        "    for i in range(10):\n",
        "        plt.subplot(2,5, i+1)\n",
        "        plt.imshow(generated[i,:].reshape(28,28), cmap='gray', interpolation='none', vmin=0.0, vmax=1.0)\n",
        "        plt.xticks([])\n",
        "        plt.yticks([])\n",
        "    plt.show()"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CHm8f363z-yX"
      },
      "source": [
        "# Training"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "agHlta29z-yY",
        "outputId": "8ab05d82-1d55-4dae-b85b-f6c1db8d64ed",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "num_epochs  = 10000  *3\n",
        "plt_frq     = 100\n",
        "batchsize   = 100\n",
        "input_shape = 10 # Number of random inputs to generator.\n",
        "learning_rate  = .05\n",
        "momgen      = 0.\n",
        "momdisc     = 0.\n",
        "clipnorm    = 1.\n",
        "decay     =   1e-5\n",
        "\n",
        "generator     = define_generator(ndense=4, nhid=100, act='tanh', input_shape=input_shape)\n",
        "discriminator = define_discriminator(ndense=2, nhid=100, act='tanh',\n",
        "  learning_rate=learning_rate*10, mom=momdisc, decay=decay) #, clipnorm=clipnorm)\n",
        "discriminator.trainable = False\n",
        "gan           = Sequential(layers=[generator, discriminator])\n",
        "optimizer = SGD(learning_rate=learning_rate, momentum=momgen, decay=decay, clipnorm=clipnorm)\n",
        "gan.compile(loss='binary_crossentropy', optimizer=optimizer)\n",
        "\n",
        "%history = defaultdict(list)\n",
        "history = {\"d_loss\":[], \"g_loss\":[]}\n",
        "for epoch in tqdm(range(num_epochs)):\n",
        "\n",
        "    # Generate images.\n",
        "    noisebatch = np.random.uniform(-1, 1, size=[batchsize,input_shape])\n",
        "    noisebatch_data = tf.data.Dataset.from_tensor_slices(noisebatch)\n",
        "    noisebatch_data = noisebatch_data.batch(batch_size=batchsize)\n",
        "    generated  = generator.predict(noisebatch_data)\n",
        "\n",
        "    # Update discriminator.\n",
        "    databatch  = X_train[np.random.randint(0,X_train.shape[0], size=batchsize), :]\n",
        "    X          = np.concatenate([databatch, generated])\n",
        "    T          = np.concatenate([np.ones((databatch.shape[0],)), np.zeros((generated.shape[0],))]) # Real:1, fake:0\n",
        "    discriminator.trainable = True\n",
        "    d_loss  = discriminator.train_on_batch(X,T)\n",
        "    history[\"d_loss\"].append(d_loss)\n",
        "\n",
        "    # Update generator.\n",
        "    noisebatch = np.random.uniform(-1, 1, size=[batchsize,input_shape])\n",
        "    T          = np.ones((noisebatch.shape[0],)) # Try to generate data that looks real.\n",
        "    discriminator.trainable = False\n",
        "    g_loss = gan.train_on_batch(noisebatch, T)\n",
        "    history[\"g_loss\"].append(g_loss)\n",
        "\n",
        "    # Updates plots\n",
        "    if epoch % plt_frq == plt_frq - 1:\n",
        "        display.clear_output(wait=True)\n",
        "        display.display(plt.gcf())\n",
        "        plot_loss(history)\n",
        "        plot_gen(generator, input_shape=input_shape)\n",
        ""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 0 Axes>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 33ms/step\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAiwAAAGwCAYAAACKOz5MAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABfvUlEQVR4nO3deXxU1f3/8dedSWayJ2zZIOw7sskSA641CkhRKlX0SwWp6E+KrRiLFVtxq8WlKtZSqFZArIpaFawLaqOAKLIa9p1IwpKwZl9n5v7+CBkYtmQgmcnyfj4e85C598ydz722zNtzzj3XME3TRERERKQOs/i7ABEREZGqKLCIiIhInafAIiIiInWeAouIiIjUeQosIiIiUucpsIiIiEidp8AiIiIidV6AvwuoCS6XiwMHDhAeHo5hGP4uR0RERKrBNE3y8/OJj4/HYjl/H0qDCCwHDhwgISHB32WIiIjIBcjMzKRVq1bnbdMgAkt4eDhQccIRERF+rkZERESqIy8vj4SEBPfv+Pk0iMBSOQwUERGhwCIiIlLPVGc6hybdioiISJ2nwCIiIiJ1ngKLiIiI1HkNYg6LiIjIhXA6nZSXl/u7jAYtMDAQq9V60cdRYBERkUbHNE2ysrLIycnxdymNQlRUFLGxsRe1VpoCi4iINDqVYSU6OpqQkBAtOlpLTNOkqKiIQ4cOARAXF3fBx1JgERGRRsXpdLrDSrNmzfxdToMXHBwMwKFDh4iOjr7g4SFNuhURkUalcs5KSEiInytpPCqv9cXMF1JgERGRRknDQL5TE9dagUVERETqPAUWERERqfMUWEREROqxq6++msmTJ1fZrm3btsyYMaPW66ktCizVkF+YQ1lZqb/LEBERabQUWKpwPPcwye8N5up/X8rRnCx/lyMiItIoKbBUIXXNexRZLORbLYz+z7X+LkdERGqBaZoUlTl8/jJN06s6CwsLGTt2LGFhYcTFxfHCCy9c8DlnZGRw0003ERYWRkREBLfeeivZ2dnu/evXr+eaa64hPDyciIgI+vXrx5o1awDYu3cvI0aMoEmTJoSGhtKjRw8+++yzC66lOrRwXBWslpML3GQHWsgtOEZkWFM/ViQiIjWtuNxJ92lf+Px7tzw5hBBb9X+Kp0yZwtKlS1m0aBHR0dE88sgjrFu3jj59+nj1vS6Xyx1Wli5disPhYNKkSYwePZolS5YAMGbMGPr27cusWbOwWq2kpaURGBgIwKRJkygrK2PZsmWEhoayZcsWwsLCvKrBWwosVbBYPC/Rig2fM3TQGD9VIyIijVVBQQGvv/46//73v7n22ooe/zfeeINWrVp5fazU1FQ2btxIeno6CQkJAMyfP58ePXqwevVqBgwYQEZGBlOmTKFr164AdOrUyf35jIwMRo0aRc+ePQFo3779xZ5elbwKLNOnT+fDDz9k27ZtBAcHM2jQIJ599lm6dOly3s+9//77PProo/z000906tSJZ599lhtuuMG93zRNHnvsMV577TVycnIYPHgws2bN8rg4/mIxPEfNMg5v8VMlIiJSW4IDrWx5cohfvre6du/eTVlZGYmJie5tTZs2rfI3+Gy2bt1KQkKCO6wAdO/enaioKLZu3cqAAQNISUlhwoQJvPnmmyQnJ3PLLbfQoUMHAH73u98xceJEvvzyS5KTkxk1ahS9evXyug5veDWHZenSpUyaNIkffviBr776ivLycq6//noKCwvP+Znvv/+e22+/nbvuuosff/yRkSNHMnLkSDZt2uRu89xzz/G3v/2N2bNns3LlSkJDQxkyZAglJSUXfmY1xHpaD4vDqceQi4g0NIZhEGIL8PmrLq+2+/jjj7N582aGDx/O119/Tffu3fnoo48AmDBhAnv27OGOO+5g48aN9O/fn1deeaVW6/EqsCxevJg777yTHj160Lt3b+bNm0dGRgZr164952defvllhg4dypQpU+jWrRtPPfUUl156KX//+9+Bit6VGTNm8Kc//YmbbrqJXr16MX/+fA4cOMDChQsv6uRqgtXqGVicCiwiIuIHHTp0IDAwkJUrV7q3HT9+nB07dnh9rG7dupGZmUlmZqZ725YtW8jJyaF79+7ubZ07d+aBBx7gyy+/5Oabb2bu3LnufQkJCdx77718+OGHPPjgg7z22msXeGbVc1F3CeXm5gIVXVLnsmLFCpKTkz22DRkyhBUrVgCQnp5OVlaWR5vIyEgSExPdbU5XWlpKXl6ex6u2GHim33KXAouIiPheWFgYd911F1OmTOHrr79m06ZN3HnnnVgs3v+UJycn07NnT8aMGcO6detYtWoVY8eO5aqrrqJ///4UFxdz3333sWTJEvbu3ct3333H6tWr6datGwCTJ0/miy++ID09nXXr1vHNN9+499WWC55063K5mDx5MoMHD+aSSy45Z7usrCxiYmI8tsXExJCVleXeX7ntXG1ON336dJ544okLLd0rJp63nLlMh0++V0RE5HTPP/88BQUFjBgxgvDwcB588EF354E3DMNg0aJF/Pa3v+XKK6/EYrEwdOhQ97CO1Wrl6NGjjB07luzsbJo3b87NN9/s/u11Op1MmjSJffv2ERERwdChQ3nppZdq9FxPd8GBZdKkSWzatInly5fXZD3VMnXqVFJSUtzv8/LyPCYO1SSXy+nx3qEeFhER8ZOwsDDefPNN3nzzTfe2KVOmVOuzP/30k8f71q1bs2jRorO2tdlsvPPOO+c8Vm3PVzmbCwos9913H5988gnLli2r8naq2NhYj4VoALKzs4mNjXXvr9wWFxfn0eZc95Xb7XbsdvuFlO41l+kZWJynBRgRERGpfV4NfJmmyX333cdHH33E119/Tbt27ar8TFJSEqmpqR7bvvrqK5KSkgBo164dsbGxHm3y8vJYuXKlu40/OZyeQ0AaEhIRkbrm22+/JSws7JyvhsCrHpZJkybx9ttvs2jRIsLDw91zTCIjIwkODgZg7NixtGzZkunTpwNw//33c9VVV/HCCy8wfPhwFixYwJo1a3j11VeBinG0yZMn8+c//5lOnTrRrl07Hn30UeLj4xk5cmQNnuqFMV2n97AosIiISN3Sv39/0tLS/F1GrfIqsMyaNQuoeJT1qebOncudd94JVKx+d+qM5UGDBvH222/zpz/9iUceeYROnTqxcOFCj4m6Dz30EIWFhdxzzz3k5ORw+eWXs3jxYoKCgi7wtGqO8/QhIVNDQiIiUrcEBwfTsWNHf5dRq7wKLNV5SFPlMwhOdcstt3DLLbec8zOGYfDkk0/y5JNPelOOTzhPGxJyakhIRETE5/S05iqcPun29PciIiJS+xRYquAyXR7v1cMiIiLiewosVXC5Tr9LSD0sIiIivqbAUoUz1mFRYBEREfE5BZYqnDGHBdc5WoqIiNRdbdu2ZcaMGf4u44IpsFTB5TptDgvqYREREfE1BZYqnDkkpB4WERERX1NgqYKGhEREGgHThLJC37+qsb7ZqfLz8xkzZgyhoaHExcXx0ksvcfXVVzN58mSvTzkjI4ObbrqJsLAwIiIiuPXWWz2e/bd+/XquueYawsPDiYiIoF+/fqxZswaAvXv3MmLECJo0aUJoaCg9evTgs88+87oGb1zw05obC+dpQ0IKLCIiDVB5Efwl3vff+8gBsIVWu3lKSgrfffcdH3/8MTExMUybNo1169ad82HB5+JyudxhZenSpTgcDiZNmsTo0aPdC8COGTOGvn37MmvWLKxWK2lpaQQGBgIVj+opKytj2bJlhIaGsmXLllp/ZpECSxVMDQmJiEgdkJ+fzxtvvMHbb7/NtddeC1Q8Gic+3vuglZqaysaNG0lPTychIQGA+fPn06NHD1avXs2AAQPIyMhgypQpdO3aFYBOnTq5P5+RkcGoUaPo2bMnAO3bt7/Y06uSAksVTn/YoXpYREQaoMCQit4Of3xvNe3Zs4fy8nIGDhzo3hYZGUmXLl28/tqtW7eSkJDgDisA3bt3Jyoqiq1btzJgwABSUlKYMGECb775JsnJydxyyy106NABgN/97ndMnDiRL7/8kuTkZEaNGkWvXr28rsMbmsNSBfP0lW7xbrxRRETqAcOoGJrx9csw/H3m5/T444+zefNmhg8fztdff0337t356KOPAJgwYQJ79uzhjjvuYOPGjfTv359XXnmlVutRYKnC6Uvzuwz1sIiIiO+1b9+ewMBAVq9e7d6Wm5vLjh07vD5Wt27dyMzMJDMz071ty5Yt5OTk0L17d/e2zp0788ADD/Dll19y8803M3fuXPe+hIQE7r33Xj788EMefPBBXnvttQs8s+pRYKnCmXcJqYdFRER8Lzw8nHHjxjFlyhS++eYbNm/ezF133YXFYsHwsqcmOTmZnj17MmbMGNatW8eqVasYO3YsV111Ff3796e4uJj77ruPJUuWsHfvXr777jtWr15Nt27dAJg8eTJffPEF6enprFu3jm+++ca9r7YosFRBQ0IiIlJXvPjiiyQlJfHzn/+c5ORkBg8eTLdu3QgKCvLqOIZhsGjRIpo0acKVV15JcnIy7du359133wXAarVy9OhRxo4dS+fOnbn11lsZNmwYTzzxBABOp5NJkybRrVs3hg4dSufOnfnHP/5R4+d7Kk26rcIZQ0IKLCIi4ifh4eG89dZb7veFhYU88cQT3HPPPVV+9qeffvJ437p1axYtWnTWtjabjXfeeeecx6rt+Spno8BShcohoQDTxGEYmAosIiLiJz/++CPbtm1j4MCB5Obm8uSTTwJw0003+bmy2qfAUoXKIaEAExwGOA0FFhER8Z+//vWvbN++HZvNRr9+/fj222/ZunUrw4YNO+dnCgoKfFhh7VBgqYJ5Yt2VQNOkBEOrsIiIiN/07duXtWvXnrG9uLiYtLQ03xfkQwosVahcmr/yQimwiIhIXRMcHEzHjh39XUat0l1CVajsYQk4MRLk0pCQiIiIzymwVOHkHJaKe9zVwyIiIuJ7CixVqLyt2XrivdZhERER8T0FlipUPuzQWtnDUncf+yAiItJgKbBUwTQrelQ0JCQiIuI/CixVqFw4zop6WEREpP5q27YtM2bM8HcZF0yBpQqVK9tWBhbn+RqLiIhIrVBgqYLpXpq/4lJpSEhERMT3FFiq4HL3sFRcKqeGhEREGhzTNCkqL/L5q3KeZHXl5+czZswYQkNDiYuL46WXXuLqq69m8uTJXp9zRkYGN910E2FhYURERHDrrbeSnZ3t3r9+/XquueYawsPDiYiIoF+/fqxZswaAvXv3MmLECJo0aUJoaCg9evTgs88+87oGb2il2yqY7tuaK3tYlFhERBqaYkcxiW8n+vx7V/7fSkICQ6rdPiUlhe+++46PP/6YmJgYpk2bxrp16+jTp49X3+tyudxhZenSpTgcDiZNmsTo0aNZsmQJAGPGjKFv377MmjULq9VKWloagYGBAEyaNImysjKWLVtGaGgoW7ZsISwszKsavKXAUgX3bc3qYRERET/Kz8/njTfe4O233+baa68FYO7cucTHx3t9rNTUVDZu3Eh6ejoJCQkAzJ8/nx49erB69WoGDBhARkYGU6ZMoWvXrgB06tTJ/fmMjAxGjRpFz549AWjfvv3Fnl6VFFiqUNldZzUqlo7TpFsRkYYnOCCYlf+30i/fW1179uyhvLycgQMHurdFRkbSpUsXr79369atJCQkuMMKQPfu3YmKimLr1q0MGDCAlJQUJkyYwJtvvklycjK33HILHTp0AOB3v/sdEydO5MsvvyQ5OZlRo0bRq1cvr+vwhtdzWJYtW8aIESOIj4/HMAwWLlx43vZ33nknhmGc8erRo4e7zeOPP37G/spE52/maXNYTMPA4Sj3Z0kiIlLDDMMgJDDE5y/DqLvd9o8//jibN29m+PDhfP3113Tv3p2PPvoIgAkTJrBnzx7uuOMONm7cSP/+/XnllVdqtR6vA0thYSG9e/dm5syZ1Wr/8ssvc/DgQfcrMzOTpk2bcsstt3i069Gjh0e75cuXe1tarXAPCRknO6PKHGX+KkdERBqp9u3bExgYyOrVq93bcnNz2bFjh9fH6tatG5mZmWRmZrq3bdmyhZycHLp37+7e1rlzZx544AG+/PJLbr75ZubOnevel5CQwL333suHH37Igw8+yGuvvXaBZ1Y9Xg8JDRs2jGHDhlW7fWRkJJGRke73Cxcu5Pjx44wfP96zkIAAYmNjq3XM0tJSSktL3e/z8vKqXY+33Cvdup8mBGXlxYQEhdbad4qIiJwuPDyccePGMWXKFJo2bUp0dDSPPfYYFovF656a5ORkevbsyZgxY5gxYwYOh4Pf/OY3XHXVVfTv35/i4mKmTJnCL3/5S9q1a8e+fftYvXo1o0aNAmDy5MkMGzaMzp07c/z4cb755hu6detWG6ft5vPbml9//XWSk5Np06aNx/adO3cSHx9P+/btGTNmDBkZGec8xvTp091BKDIy0mMMrqaZJ3pYLMbJwFKuHhYREfGDF198kaSkJH7+85+TnJzM4MGD6datG0FBQV4dxzAMFi1aRJMmTbjyyitJTk6mffv2vPvuuwBYrVaOHj3K2LFj6dy5M7feeivDhg3jiSeeAMDpdDJp0iS6devG0KFD6dy5M//4xz9q/Hw9aja9vQn81A8bBh999BEjR46sVvsDBw7QunVr3n77bW699Vb39s8//5yCggK6dOnCwYMHeeKJJ9i/fz+bNm0iPDz8jOOcrYclISGB3NxcIiIiLvR0zup3r/2Mb2yHGepIYHFARdfZ58P+S6votjX6PSIi4hslJSWkp6fTrl07r3/o65rCwkJatmzJCy+8wF133eXvcs7pXNc8Ly+PyMjIav1++/QuoTfeeIOoqKgzAs6pQ0y9evUiMTGRNm3a8N577531X4Ddbsdut9d2ucApk25PmcPiVA+LiIj4wY8//si2bdsYOHAgubm5PPnkkwDcdNNNfq6s9vkssJimyZw5c7jjjjuw2WznbRsVFUXnzp3ZtWuXj6o7t8ruJ8M4OXpWWl7in2JERKTR++tf/8r27dux2Wz069ePb7/9lq1bt553fmlBQYEPK6wdPgssS5cuZdeuXdXqsiooKGD37t3ccccdPqisKhWRxcDAapo4DQOnU7c1i4iI7/Xt25e1a9eesb24uJi0tDTfF+RDXgeWgoICj56P9PR00tLSaNq0Ka1bt2bq1Kns37+f+fPne3zu9ddfJzExkUsuueSMY/7+979nxIgRtGnThgMHDvDYY49htVq5/fbbL+CUalblDB8DsJoVK92WO0rP+xkRERFfCg4OpmPHjv4uo1Z5HVjWrFnDNddc436fkpICwLhx45g3bx4HDx484w6f3NxcPvjgA15++eWzHnPfvn3cfvvtHD16lBYtWnD55Zfzww8/0KJFC2/Lq3mGe1DIfUuVFo4TEan/XC6Xv0toNGriWnsdWK6++urzPl1y3rx5Z2yLjIykqKjonJ9ZsGCBt2X4zMkeFgMrJmBQ5lQPi4hIfWWz2bBYLBw4cIAWLVpgs9nq9Iqz9ZlpmpSVlXH48GEsFkuVc1jPR88SqtLJMSHLiYDocDr8V46IiFwUi8VCu3btOHjwIAcOHPB3OY1CSEgIrVu3xmK58OXfFFiqYLon3Vrca906NIdFRKRes9lstG7dGofDgdOpx9rWJqvVSkBAwEX3YimwVOHkbc1gOfFGK92KiNR/hmEQGBhIYGCgv0uRavD50vz1l+HuYXG6NCQkIiLiSwosVTq5DsvJHhYNCYmIiPiSAksVTt7UbGChYvzNpR4WERERn1JgqSbDOHmxyrXSrYiIiE8psFTBPKWPpXJIyOlSYBEREfElBZZqqlg4rmJISIFFRETEtxRYqnCyh+XkxdI9+yIiIr6lwFJdhoHFrOhhcaiHRURExKcUWKpQ+dykU+8S0josIiIivqXAUhWj8h+nBhb1sIiIiPiSAksVTs5gMSqzCy6X5rCIiIj4kgJLVSqHhIyTdwlpDouIiIhvKbBUwbOH5cRKt6Z6WERERHxJgaUqRkVksRhgMSsulybdioiI+JYCSxVMdxeLgcVQD4uIiIg/KLBUU8XTmnVbs4iIiD8osFTBvdKtoR4WERERf1FgqaaKdVgq57AosIiIiPiSAksVKntYjFPvEkKBRURExJcUWKpwcs6tgfXE5dLCcSIiIr6lwFJN6mERERHxHwWWqlQ+S8gwsGAFNIdFRETE1xRYquDxtGZDPSwiIiL+oMBSTaf2sLhMl5+rERERaVwUWKpw6rOELMaJSbdah0VERMSnFFiqVBlZTq7Doh4WERER31JgqUJlXLEYp/SwoMAiIiLiSwosXrAYmsMiIiLiD14HlmXLljFixAji4+MxDIOFCxeet/2SJUswDOOMV1ZWlke7mTNn0rZtW4KCgkhMTGTVqlXellYrTOOUISF3D4vmsIiIiPiS14GlsLCQ3r17M3PmTK8+t337dg4ePOh+RUdHu/e9++67pKSk8Nhjj7Fu3Tp69+7NkCFDOHTokLfl1Tz3sw9PvUvIPM8HREREpKYFePuBYcOGMWzYMK+/KDo6mqioqLPue/HFF7n77rsZP348ALNnz+bTTz9lzpw5PPzww2e0Ly0tpbS01P0+Ly/P63qqy6xcOA7j5JCQelhERER8ymdzWPr06UNcXBzXXXcd3333nXt7WVkZa9euJTk5+WRRFgvJycmsWLHirMeaPn06kZGR7ldCQkItVn5y4TjricBioh4WERERX6r1wBIXF8fs2bP54IMP+OCDD0hISODqq69m3bp1ABw5cgSn00lMTIzH52JiYs6Y51Jp6tSp5Obmul+ZmZm1Vv+pDz+snMPi1F1CIiIiPuX1kJC3unTpQpcuXdzvBw0axO7du3nppZd48803L+iYdrsdu91eUyVWi2FYsBgBYKqHRURExNf8clvzwIED2bVrFwDNmzfHarWSnZ3t0SY7O5vY2Fh/lOfBfY+QYWC1aB0WERERf/BLYElLSyMuLg4Am81Gv379SE1Nde93uVykpqaSlJTkj/I8mB5zWAIBcGodFhEREZ/yekiooKDA3TsCkJ6eTlpaGk2bNqV169ZMnTqV/fv3M3/+fABmzJhBu3bt6NGjByUlJfzrX//i66+/5ssvv3QfIyUlhXHjxtG/f38GDhzIjBkzKCwsdN81VBcYGARYA6EcnBoSEhER8SmvA8uaNWu45ppr3O9TUlIAGDduHPPmzePgwYNkZGS495eVlfHggw+yf/9+QkJC6NWrF//73/88jjF69GgOHz7MtGnTyMrKok+fPixevPiMibj+cHLSrQWrpaKHRUNCIiIivmWYZv1fBS0vL4/IyEhyc3OJiIio0WPf8mofttmd3Bfxcxymg9n5i+leGsC79/xYo98jIiLS2Hjz+13rdwk1GIZBQOUcFvWwiIiI+JQCSxXck24NgwCLDQCX5rCIiIj4lJ7WXG0WAq0Va784DAUWERERX1JgqUJlNLEYBoEBFT0sepKQiIiIbymwVJtBwIkeFqd6WERERHxKgaUKlXNYLIaFwIATgcWfBYmIiDRCCizVZmALUA+LiIiIPyiwVOHUZwkFBgQB4PBfOSIiIo2SAksVTOPkn+2BlXcJ+akYERGRRkqBpSonulgMA+yBIQA4FVhERER8SoGlCqZR+bRmCzZb5ZCQEouIiIgvKbBUk2EY2AODAQ0JiYiI+JoCSxXck26xYA+s6GFxGQYOR7n/ihIREWlkFFiqcOpdQjZ7qHt7cWmhfwoSERFphBRYqskwDIJPzGEBKCkr8mM1IiIijYsCSxVO7WEJDgp3by8pUQ+LiIiIryiwVJuFoMCTPSzFZcV+rEVERKRxUWCpJothYLFaCTAr+lzKyhVYREREfEWBpQonnxpUcS+z9cSGsrISf5QjIiLSKCmwVKHyac2GURFYAk68L3Woh0VERMRXFFiqyThxqQIqe1jKS/1YjYiISOOiwFKFyocfGkbFpaocEirVbc0iIiI+o8BSTSeHhCqUO9XDIiIi4isKLFU4dR0WAOuJLpdyDQmJiIj4jAJLFSoDi6VySOjE+zKHAouIiIivKLBUk4FnD4vDpcAiIiLiKwosVThjSOjE+3JHmV/qERERaYwUWKrLfZfQiTksTgUWERERX1FgqcKZPSwVl0yBRURExHcUWKpgnrI4P4D1xFwWh25rFhER8RkFlmo6Oem24pI5XeX+LEdERKRR8TqwLFu2jBEjRhAfH49hGCxcuPC87T/88EOuu+46WrRoQUREBElJSXzxxRcebR5//HEMw/B4de3a1dvSakXlSrcWS8WlslT2sCiwiIiI+IzXgaWwsJDevXszc+bMarVftmwZ1113HZ999hlr167lmmuuYcSIEfz4448e7Xr06MHBgwfdr+XLl3tbWq1wz2Ghch2Win8qsIiIiPhOQNVNPA0bNoxhw4ZVu/2MGTM83v/lL39h0aJF/Pe//6Vv374nCwkIIDY2tlrHLC0tpbT05BySvLy8atdzoU6fdOt0OWr9O0VERKSCz+ewuFwu8vPzadq0qcf2nTt3Eh8fT/v27RkzZgwZGRnnPMb06dOJjIx0vxISEmq7bPfDDytXvHW4dJeQiIiIr/g8sPz1r3+loKCAW2+91b0tMTGRefPmsXjxYmbNmkV6ejpXXHEF+fn5Zz3G1KlTyc3Ndb8yMzNrrd4zb2uuWDrOaaqHRURExFe8HhK6GG+//TZPPPEEixYtIjo62r391CGmXr16kZiYSJs2bXjvvfe46667zjiO3W7Hbrf7pOaTc1hOCywaEhIREfEZnwWWBQsWMGHCBN5//32Sk5PP2zYqKorOnTuza9cuH1VXtcohIeuJfzpNpz/LERERaVR8MiT0zjvvMH78eN555x2GDx9eZfuCggJ2795NXFycD6o7v5NDQpWBpSLjaUhIRETEd7zuYSkoKPDo+UhPTyctLY2mTZvSunVrpk6dyv79+5k/fz5QMQw0btw4Xn75ZRITE8nKygIgODiYyMhIAH7/+98zYsQI2rRpw4EDB3jsscewWq3cfvvtNXGOF6UysFjOmMOiHhYRERFf8bqHZc2aNfTt29d9S3JKSgp9+/Zl2rRpABw8eNDjDp9XX30Vh8PBpEmTiIuLc7/uv/9+d5t9+/Zx++2306VLF2699VaaNWvGDz/8QIsWLS72/C7eiYXj3JNuLephERER8TWve1iuvvpqTNM85/558+Z5vF+yZEmVx1ywYIG3ZfjM6UNCAUYgAA4UWERERHxFzxKqwumBJdBiA8ChISERERGfUWCppso5LAGVgQUFFhEREV9RYKnC6T0sNmvF+i8OzWERERHxGQWWKrhn61TOYakMLLj8U5CIiEgjpMBSTZVDQvbAYAAchgKLiIiIryiwVOH0ZwkFWoMA9bCIiIj4kgJLFUz3OiwVlypIPSwiIiI+p8BShdNXnLEFVASW8jP2iIiISG1RYKkmy4lLFWwLBcBh+LMaERGRxkWBpQruOSyWE5NuTwSWckM9LCIiIr6iwFJNxomHHgbZQgAoVw+LiIiIzyiwVME88fTDytuaQ+zhgAKLiIiILymwVME98GM5cZdQUBgAZYYSi4iIiK8osFRTZQ9LWFBFD4vDMHA4yv1ZkoiISKOhwFKFyh4Wy4l1WEKCI9z7Copz/VCRiIhI46PAUoXKheMqnyUUHhLl3ldQXOD7gkRERBohBZYqnOxhqUguwfYQ976i4hzfFyQiItIIKbBUk2FU3NZssVqxuypiTFGJelhERER8QYGlCqf3sAAEmhVbi0sVWERERHxBgaUKp0+6BQg88c9i9bCIiIj4hAJLFU5fmh8g8MTGkrIi3xckIiLSCCmwVMF0DwWdvFQBJ24dKi5TD4uIiIgvKLBUk8eQ0InAUlpe7K9yREREGhUFlvNwOZ3uP5866TbgxPOFyhwaEhIREfEFBZbzcJku958Ni9X950Cz4rKph0VERMQ3FFjOw+U6GVhOHRI62cNS4vOaREREGiMFlvNwmecYEjIrelvKnQosIiIivqDAch4u0zz5xjg5JBRAxZ/LnKW+LklERKRRUmC5ANYTl61cgUVERMQnFFjOw3SdOun2lDksRgAA5S4FFhEREV9QYDmPc81hsRkVi/OXu8p8XpOIiEhj5HVgWbZsGSNGjCA+Ph7DMFi4cGGVn1myZAmXXnopdrudjh07Mm/evDPazJw5k7Zt2xIUFERiYiKrVq3ytrQaF2wP5ZUeT/K37o/TPDLWvT3QsAGawyIiIuIrXgeWwsJCevfuzcyZM6vVPj09neHDh3PNNdeQlpbG5MmTmTBhAl988YW7zbvvvktKSgqPPfYY69ato3fv3gwZMoRDhw55W16NCggI5Or+v+CaAaOw2ezu7TZrxZ/LTfWwiIiI+EKAtx8YNmwYw4YNq3b72bNn065dO1544QUAunXrxvLly3nppZcYMmQIAC+++CJ3330348ePd3/m008/Zc6cOTz88MNnHLO0tJTS0pO9G3l5ed6exkWxWYIAKDPLffq9IiIijVWtz2FZsWIFycnJHtuGDBnCihUrACgrK2Pt2rUebSwWC8nJye42p5s+fTqRkZHuV0JCQu2dwFnYrMEAlCuwiIiI+EStB5asrCxiYmI8tsXExJCXl0dxcTFHjhzB6XSetU1WVtZZjzl16lRyc3Pdr8zMzFqr/2yCAioCSxkOn36viIhIY+X1kFBdYLfbsdvtVTesre8PDIVSKDecVTcWERGRi1brgSU2Npbs7GyPbdnZ2URERBAcHIzVasVqtZ61TWxsLHVRUGAoAOW4qmgpIiIiNaHWh4SSkpJITU312PbVV1+RlJQEgM1mo1+/fh5tXC4Xqamp7jZ1TYg9DIAyQ4FFRETEF7wOLAUFBaSlpZGWlgZU3LaclpZGRkYGUDG/ZOzYse729957L3v27OGhhx5i27Zt/OMf/+C9997jgQcecLdJSUnhtdde44033mDr1q1MnDiRwsJC911DdU2ILQKAMsOsoqWIiIjUBK+HhNasWcM111zjfp+SkgLAuHHjmDdvHgcPHnSHF4B27drx6aef8sADD/Dyyy/TqlUr/vWvf7lvaQYYPXo0hw8fZtq0aWRlZdGnTx8WL158xkTcuiIkqCKwlCqwiIiI+IRhmma9/9XNy8sjMjKS3NxcIiIiav37vl//Of8v7SHCnS6+//XmWv8+ERGRhsib3289S+gChIU0AaDEYlTRUkRERGqCAssFiAxtCkC5YVBSWuTnakRERBo+BZYLEBne1P3n3ILjfqxERESkcVBguQARJ4aEAHILjvqxEhERkcZBgeUCWKxWglwVc5ULCtXDIiIiUtsUWC6Q/cTNVfnFOf4tREREpBFQYLlA9hM3gxcW5/q3EBERkUZAgeUC2cyKW5oLS/L8XImIiEjDp8BygSoDS3GZAouIiEhtU2C5QDbTCkBxWYGfKxEREWn4FFgukAKLiIiI7yiwXCDbiedGlpQrsIiIiNQ2BZYLZDcCASh2FPq5EhERkYZPgeUC2Q07ACVOPUtIRESktimwXCC7JQiAUmexnysRERFp+BRYLlBQQAgAJWapnysRERFp+BRYLlBwQCgApWaZnysRERFp+BRYLlBwYDgApUa5nysRERFp+BRYLlCIPQKAUpx+rkRERKThU2C5QGFBTQEoNVx+rkRERKThU2C5QOHBTQAoUWARERGpdQosFygqrDkAxbqCIiIitU4/txcoMjwagCKL4edKREREGj4FlgvUNDIWAIdhkF+Y499iREREGjgFlgvUPDLa/ecjOVl+rERERKThU2C5QEH2EOwuE4Djudl+rkZERKRhU2C5CCFmRWA5ln/Iz5WIiIg0bAosFyHIVTHhNr/oqJ8rERERadgUWC5CsFkRWPIUWERERGqVAstFCDYDAMgvOe7nSkRERBo2BZaLEGQGAlBQcszPlYiIiDRsFxRYZs6cSdu2bQkKCiIxMZFVq1ads+3VV1+NYRhnvIYPH+5uc+edd56xf+jQoRdSmk8FG3YACspz/VyJiIhIwxbg7QfeffddUlJSmD17NomJicyYMYMhQ4awfft2oqOjz2j/4YcfUlZW5n5/9OhRevfuzS233OLRbujQocydO9f93m63e1uaz4VYQoAcih2F/i5FRESkQfO6h+XFF1/k7rvvZvz48XTv3p3Zs2cTEhLCnDlzztq+adOmxMbGul9fffUVISEhZwQWu93u0a5JkyYXdkY+FGwNA6DYVeTnSkRERBo2rwJLWVkZa9euJTk5+eQBLBaSk5NZsWJFtY7x+uuvc9tttxEaGuqxfcmSJURHR9OlSxcmTpzI0aPnvvOmtLSUvLw8j5c/hNoiACg2S/3y/SIiIo2FV4HlyJEjOJ1OYmJiPLbHxMSQlVX18vSrVq1i06ZNTJgwwWP70KFDmT9/PqmpqTz77LMsXbqUYcOG4XQ6z3qc6dOnExkZ6X4lJCR4cxo1JswWBUCxUXb+hiIiInJRvJ7DcjFef/11evbsycCBAz2233bbbe4/9+zZk169etGhQweWLFnCtddee8Zxpk6dSkpKivt9Xl6eX0JLREhzKIJiw+Hz7xYREWlMvOphad68OVarlexsz2fnZGdnExsbe97PFhYWsmDBAu66664qv6d9+/Y0b96cXbt2nXW/3W4nIiLC4+UPUWEVPU1FhumX7xcREWksvAosNpuNfv36kZqa6t7mcrlITU0lKSnpvJ99//33KS0t5Ve/+lWV37Nv3z6OHj1KXFycN+X5XLOIipBWaFFgERERqU1e3yWUkpLCa6+9xhtvvMHWrVuZOHEihYWFjB8/HoCxY8cyderUMz73+uuvM3LkSJo1a+axvaCggClTpvDDDz/w008/kZqayk033UTHjh0ZMmTIBZ6Wb0Q3bQVAgcXA4Sj3czUiIiINl9dzWEaPHs3hw4eZNm0aWVlZ9OnTh8WLF7sn4mZkZGCxeOag7du3s3z5cr788sszjme1WtmwYQNvvPEGOTk5xMfHc/311/PUU0/V+bVYYpq1BsA0DLKO7adVdFv/FiQiItJAGaZp1vvxjLy8PCIjI8nNzfX5fJb+cy+h1GLwRuI/uLTrFT79bhERkfrMm99vPUvoIoW5KvLe4eOZfq5ERESk4VJguUhhropLeCR3n58rERERabgUWC5SmFkxDehY/kE/VyIiItJwKbBcpFCCAMgtOeLnSkRERBouBZaLFGoEA5BfnuPfQkRERBowBZaLFBYQCUCBI9/PlYiIiDRcCiwXKcLWBIAis9jPlYiIiDRcCiwXKTKkYsG8AqPUz5WIiIg0XAosF6npiQcgFlj0xGYREZHaosBykWKatAEgXw9AFBERqTUKLBcpvkUHAPIsBmVlGhYSERGpDQosF6l1bEeg4gGIGVk7/VyNiIhIw6TAcpGC7CFEOl0AZGRt83M1IiIiDZMCSw2IclZcxoPH9vi5EhERkYZJgaUGRJiBABzO0xObRUREaoMCSw0IJwSAY8XZfq5ERESkYVJgqQER1ggA8sqO+bkSERGRhkmBpQZE2poDkOcq8HMlIiIiDZMCSw1oGhoLQL5R4udKREREGiYFlhoQHVmx2m2elucXERGpFQosNaBVdCcAjltNXE6nn6sRERFpeBRYakCX1n0BKLZY2H94r5+rERERaXgUWGpAk8gW7tVut/202s/ViIiINDwKLDWkucMKQOZhLc8vIiJS0xRYakiUGQTAwdx0P1ciIiLS8Ciw1JAm1igAjpUc9G8hIiIiDZACSw1pZo8B4Lgz18+ViIiINDwKLDUkOqItADlGkX8LERERaYAUWGpI6xZdAThi1TosIiIiNU2BpYZ0adMPgOMBFnIL9BBEERGRmqTAUkPaxHYiyGUCsDV9jZ+rERERaVguKLDMnDmTtm3bEhQURGJiIqtWrTpn23nz5mEYhscrKCjIo41pmkybNo24uDiCg4NJTk5m586dF1Ka31isVlqceJTQngMb/VuMiIicIfvofqa8fgN3/LM/d/yzPw/N+TmvLvojG3Z8r8eq1AMB3n7g3XffJSUlhdmzZ5OYmMiMGTMYMmQI27dvJzo6+qyfiYiIYPv27e73hmF47H/uuef429/+xhtvvEG7du149NFHGTJkCFu2bDkj3NRlTV12Milj37HtVTcWERGfWbFhMU+t/D2ZNsP9y5fGXsjZyysrPqbJchety4NJsLWic4t+DO55E53b9PJv0eLB68Dy4osvcvfddzN+/HgAZs+ezaeffsqcOXN4+OGHz/oZwzCIjY096z7TNJkxYwZ/+tOfuOmmmwCYP38+MTExLFy4kNtuu83bEv2mubUpkEV2oZ4nJCJSV3z+3Zs8seMZCm0WmjlcDA2+DNN0kVm4h/3GMTICTY5bLRy3lrKe3XB0Ny8ueY8WDhetHaEk2NvQLW4gV/QaSUJcJ3+fTqPlVWApKytj7dq1TJ061b3NYrGQnJzMihUrzvm5goIC2rRpg8vl4tJLL+Uvf/kLPXr0ACA9PZ2srCySk5Pd7SMjI0lMTGTFihVnDSylpaWUlpa63+fl5XlzGrUmNqQ1lGdxyHnU36WIiAjw/frPeWnrsxQGWuhcamH6kDfp3KaPR5uc/CN8++PHbMpcxt7CXWQYOewLhMMBFg4HFLOWbSzM2sb0rPnElZskuMJpHdSWDi36UFJeREl5Ie1iLqFXh8EKNLXIq8By5MgRnE4nMTExHttjYmLYtu3sz9Dp0qULc+bMoVevXuTm5vLXv/6VQYMGsXnzZlq1akVWVpb7GKcfs3Lf6aZPn84TTzzhTek+0bZ5Dzi4ikPWEn+XIiLS6C1P+4yp66aQE2ihhcPF88MW0D6hxxntosKbM+LKXzOCX7u3HT5+gKXrPmLLge/JKN5DpiWfA4EGBwMNDlLAKnMTHNp08iAFX8DuF4hwuohxWIkmkvigBNo2v4Q+Ha/mkg6JWKxWX5x2g+X1kJC3kpKSSEpKcr8fNGgQ3bp145///CdPPfXUBR1z6tSppKSkuN/n5eWRkJBw0bVerEvaD4aDc8kOgKKSQkKCQv1dkohIo/TW4ueYdeANcq0W2pfB89e/c9awci4tmsTzy2snAZPc2zKz9vDt+o/YlvUDmSUZHLYUEYBBoGnhsLWcowEW8qwW8qwmO8kBMwcOb4TD7xDynYuW5Z5BpnubQfTqlITNZq/x82+IvAoszZs3x2q1kp2d7bE9Ozv7nHNUThcYGEjfvn3ZtWsXgPtz2dnZxMXFeRyzT58+Zz2G3W7Hbq97/4K7t+tP0LcmJRaDtO3LGNR7mL9LEhFpdD7/7k1eOjifUquFNmXw0tCz96x4KyG2Pf8X++A59x/NyWL9juVs37+WzJxtZJUdJMtSwMFAKLJY2Gk/M8jYVpnEOgxauIJpEdCCuLB2dIjtw6Wdr9Lw0mm8Ciw2m41+/fqRmprKyJEjAXC5XKSmpnLfffdV6xhOp5ONGzdyww03ANCuXTtiY2NJTU11B5S8vDxWrlzJxIkTvSnP7yxWK3EOg3QbbMtYpcAiIuJjR3OymL3lOUptBr1KbPz9tsU0iWzhk+9uFhXLzwb+kp/xS4/tRSWFrN2Sypa9K9ibs43ssoNkWwo5GGBSZjHIsEEGxUAGlGTAT0vhp5eJdLqIdQTQgghi7C1p1aQLXVsn0q3tpZQ7yoluEt+ohpm8HhJKSUlh3Lhx9O/fn4EDBzJjxgwKCwvddw2NHTuWli1bMn36dACefPJJLrvsMjp27EhOTg7PP/88e/fuZcKECUDFHUSTJ0/mz3/+M506dXLf1hwfH+8ORfVJCzOMdArIPH72OT0iIo2Nw1HOoeMHOHAone2ZqwiyhREd1YpWMZ1JiOlAQEBgjXxP1pFM/vjhreyxQ4jLxZ+ue9VnYeV8QoJCueLSG7ni0hs9tpeVlbJp1w9s/mkFe49uIbskk8NmDtnWMo4EWMi1Wsi1uthODpADxzfD8Q9hfcXnA0yT5g6IdtppbokiOjiB1k270qpFJ1pGd6B9yx5VXtvso/v5YdPnHM7NpEVESxJiu9EpoSfhoVG1cSkuiteBZfTo0Rw+fJhp06aRlZVFnz59WLx4sXvSbEZGBhbLyfXojh8/zt13301WVhZNmjShX79+fP/993Tv3t3d5qGHHqKwsJB77rmHnJwcLr/8chYvXlyv1mCpFB0YDRSQXXLA36WIiPiMy+nkix/e5qdDWyhzFGG1BHK0cD9bS7aRHlhGkeW0dUozgY1gNU0inCZ206CF00bLgDjaN+1F/87J9Ot6dbV6EFZt/B+frPsn35dvIdtuwTBN7o+/k27t+9XOydYQm83Opd2v4tLuV52x7/DxA6zbtoSdB9axP28Xh8uzOWwUcDDARfGJa+kwDLICISuwDDgEzkNweC0cBraAzWXS3AlNXYFEEkbTwOY0D2lJfJMO5BUfY3nWYrbbik7+uzkO7AVWQpTTRROnhTAzkAPWUkwgyDT4fPx6v/XqGKZpmn755hqUl5dHZGQkubm5RERE+LWWmR9OYXb+YjqXWvjgnvV+rUVEfGtXxibeWfoM2SX7sRk2QgLCSIjszIBuN9Cn06AG232/Y+8G/vjFWLbZz79abIBpElcOTgMKLCZ51vMvth7mdNHaYaOlJZp2UT3o2zGZgd2v5Xj+Ed786in2F+xhu3mgYjG4E5o4XIxtMYoJNz5ZI+dW17icTjKzdxMR2oTMQ7vZkbGWvYc3kVW4l8OOIxyxFJNndZFrMXCdtkjruTR1uGjqtJJvcXIswKD8HJ8LdrlYNX5zTZ6OV7/ftX6XUGPTq+0VsHExmYEOHI7yGuvqFJGasydzM5+tmsuxwgPklR+nyFlEsCWIpkGxBAeG0zw8nit630y7ll2rdbwDh/fy7MK7+CHgYMV/rdpO2Zm3A1Z+QrPvXHRwhNM+tCv9O1zPVZeOJMgecs5jFpUUAhAUGITFasXldLJ22xJWbPkvhwoyiYtsT6+2V5B4yRC/3mWyPf1HJqfewT67gd1l0rbcig0rLkxcmHQIaM21Pe5gQI9rCQ+O9AhthUX5ZGTtYN+hXRQUH2dn1joyCnaw3zxGRqCLAquFLVYHWzgAhQdg/VfYfzQptZz4QQ0EMAgwTbqWBdI1qAt3XP9ojUywrassVitt4jsD0CSyBb06XXbWdkUlhWz7aR179m9g/7EdHC7Yx3HHEY6b+Ry3lOEALiGe5K7/x5DLxrh/q1xOJ3uzdrJ730b2H9nB0YL9NA2Lo2WzzpSUFfrqNM9KPSw1rKS0iMvfHkipxeC1S1/gsp7X+7UeETnJ4Sjnkfm/4EvLTzir8V+fMeUu2jjD6RTWncHdRjK49/AzeknS929j8me3sOdESGlbBl0tCZimSYGrgCxy2Wtz4Tjt+0JcLtqX22kX2IZeLa8gOqo1H62fiYFBriufDfZiHIZBuNNFvCOAUsPFTzbOUHmctoEJ9IgdxNX9RtMqum21r0na9uV8supVjpZmEWINw2a1ERwQTrdWSVzee8R554C4nE5+/a8k1gYV08Lh4i8DX6qxv/OKSgr5fv1nbPhpCXvztrHfPMLeQCclJ8JKc4eL3q44OjTpxS+u/J1X5yx1hze/3wosteAXr/Zil93kNxE/Z+Ivpvu7HJF6b/PuNazf9Q22gBDaxvbg0q5XeD28UlRSyO/nD+Nb+3EAWpabJLgiCLOEEWINo8hZQK4rlyLKyLGUsz/wzEDTxOmiY3k4ncN7cHm3X9C78+X8+q2r2WZ3Eul0MTpiCBNHPntGz+rx3MP8b/W7bNi3lPTSPewOLKGgiuGQswkwTTqUWWlGOEfIJzPQ4Z7PUMkwTVqXG7Qxm9EpqieXdRvBwO7XnnG9Zn44hdQjX7DTfu6fAItpklBu0MqMom1YZ3q3uZrL+4xwT8h8+f37+VfR11hMk5e6P87PBv7ynMeqCWVlpWzYuQITJz06XKa1rhoABRY/+82rV/Gt/Rg3ONvy7K//6+9yROqtz797k/9smslae4FHj0jl7Z7Fhosg00JTQukS0Yvr+9151i5yl9PJ/XOSWWI7AsBY2yCm3P7P8373vkM/sTztQzbuX86e8nR2BpafHIo4jcU0+WvXR7nustHVOi+Ho5xvf/yYlTs/Y3fBVtItuWQHVgSPq8uaE2wJ4fpL7qRL6z5sz0hja+YKCkqPMzLpPo+JpGVlpSxP+y+rdy1md8FWfrLkcvAsQSvS6aK9I4S29nb0anUVB4/v5tXCr4CKgNO1LIBWlhhKzRJKXWUUGaXsCyjl+FlCVaBp0rrcQrzZhPUBR8izWrjR1Ymnx39YrXMXOZUCi59Nf2s8bzvW0KfEzpv/b42/yxGpd1xOJ1PmDuergH2YJ4JKq/KKv6qyAzjnpECA1mXQxYinb/w13DDoTppFxfLAv67nf4EHMUyTCaHJ/O6WGV7XlF+Yw5cr32bd3v+xuzydXScCjNU0mRA+lPtG/fWCzrXS5t1riAiJvOjFwnbs3cCy9f9h2+HV7HUeZE+gg7JzBK1LS4L449DXzni2zqk1rdj8X3YeWce+8gP8FFhyxkTZuHKTd375P5pFVW/xUJFTKbD42UffzGZaxkxaOFx8fVfNzqgWaQye/vc4FjjXAdC/JIQxfX5PcuItQEVwWLL2Q9bvXYJhGDQPjWdvzja2l+9hl83pcWdEgGkS7jQ5HlDxIzsmYAAPj5lTIzXmFhzjf6sW0CQsutaHQi5GYVE+X695n7S9X5NevJM91gKOWw0uL2/Gc2M+ITQkvNrHcjmdpG1fzsptn7H72AbKzXLGXfEYl3a9ohbPQBoyBRY/yzqSyfWfDMM0DD64+i06t+nl75JE6o2vfniXKduewmkYXF8ez/PjP6v2fJXMgzv5ZMW/2HDke3ZajrmHWQCuKG3KP+5ZWltl1yt61pnUFbqt2c9imyfQqhwybbBs/QcKLCLV9PYXL/DygTk4LRb6lNh5fkL1wwpAQlwnJt78LID7NuAlG96juDyfX4/8S22VXe8orEh9pMBSS9qazcjkGNsPr/Z3KSL1wtQ5N/KJNR0sFjqVGvzlxncuaqE1i9XKgB7XMqDHtTVYpYj4i/f31Um1dIysWLjoJ6eW6Bepyp/fHFsRVoBeJTZe+vl/9KRaEfGgwFJLBnb5OQDpgQ73ipUicqbNu9fwX8daAH5WFsNb/2+teyVPEZFKCiy15LJLriPc6aLUYrB0rdYnEDmbTbtWcv834yg6MQz0wvjP/V2SiNRRCiy1JCAgkPblwQCs2fOFn6sRqXsKi/J5+uuJZAdaiCl3MeWy5/TsLRE5JwWWWtQpuGIMfnvRNj9XIlL3PLHgdjbZy7GaJvd3fpCkXkP9XZKI1GEKLLVocJebAdhhK6awKN/P1YjUHY+9MZrPrXsBuC1wICOu/LWfKxKRuk6BpRZd3W8kUU4XxRYLn6+Y7+9yROqE/6TO5EO2AHB9eUseuu01P1ckIvWBAkstCggIpEt5xcp9a/dqHotISWkRr6bPAmBASSjPj//0otZaEZHGQ4GllnWPuhSAbY6f/FuISB0w99MnOBhoEOl08cRNbyusiEi1KbDUsp9fdjeGabLLbrJ5t57cLI1XUUkhi459CsAVRicSYtv7uSIRqU8UWGpZ5zZ96FRW8V+R/105y8/ViPjP3z74LfsDDaKcLu4f8Yq/yxGRekaBxQd62rsBsKHgRz9XIuIfZWWlfFWyEoAb7AOIbZ7g54pEpL5RYPGBIX3vBGCrrYzMrD3+LUbEDxb87wUOBVgId7qYeONf/V2OiNRDCiw+kNRrKG3KwGEYvPP1dH+XI+JzX2Z+AMBAVzxR4c39XI2I1EcKLD4ywH4JACsLVvm5EhHf+nbdx6wPKsNqmtyWOMXf5YhIPaXA4iO3XfEQFtNkh93Fig2L/V2OiM98tn4OAD1Lg7is5/V+rkZE6isFFh/p0q4vPUvtACxa8w8/VyPiO+sduwDo12SQnysRkfpMgcWHklr8DIBV5m5KSov8XI1I7Vu6dhGZNoMA0+SXV6f4uxwRqccUWHzoV9dPJdLp4nCAhbmfPeXvckRq3eINcwG4pDSIVtFt/VuMiNRrAf4uoDGJDGvKFXTgE9L53+HPmIjuGJL6a+uetXy78UN2HUnDYTqICUngktaXM7j3z4kKb47L6awYDrIZ9G862N/likg9Z5imafq7iIuVl5dHZGQkubm5RERE+Luc89qxN43bvvkV5YbBX9o9wIgrf+3vkkSqLb8wh39/MZ1Vh5ew1l6IaRhntLGaJq3KDYJNK9vsTgJNk0+Gf0p8izZ+qFhE6jJvfr/Vw+Jjndv0YWBZFN/Zc/lo2z8VWKTOKCkt4r/fzmH/se0UluVxsDiDUrOUHIrYG1juHj8utFggCMCgbRm0pilBRhBZziNkBJaQY7Ww1wbgBKBPaZjCiohctAsKLDNnzuT5558nKyuL3r1788orrzBw4MCztn3ttdeYP38+mzZtAqBfv3785S9/8Wh/55138sYbb3h8bsiQISxe3DBv/73t0gf5ftOjrLYX8dUP73LdZaP9XZI0UkUlhbz+yaOsPbqcrbZCiiynTGuzndry5PbmDhf9acPPut7OsMF3eBzP5XSyac9q1mz7gj1HN2CaLib+4qXaPQkRaRS8HhJ69913GTt2LLNnzyYxMZEZM2bw/vvvs337dqKjo89oP2bMGAYPHsygQYMICgri2Wef5aOPPmLz5s20bNkSqAgs2dnZzJ071/05u91OkyZNqlVTfRoSqjThn0msDCqgT4mdN/+fnuIsvvfN6g94ev00sgNPhpEwp4t2DjtBBBJpiaSpPZpAi42eba4kMiya/MKjXNN/FEH2ED9WLiINhTe/314HlsTERAYMGMDf//53AFwuFwkJCfz2t7/l4YcfrvLzTqeTJk2a8Pe//52xY8cCFYElJyeHhQsXelOKW30MLD9s/JJ716bgNAz+3Oa33HT1Pf4uSRqRw8cPMOaD6zkYaNDE6WKw0ZErO/+S6xJvIyAg0N/liUgj4c3vt1e3NZeVlbF27VqSk5NPHsBiITk5mRUrVlTrGEVFRZSXl9O0aVOP7UuWLCE6OpouXbowceJEjh49es5jlJaWkpeX5/Gqby7reT2DypoBMHfn3ykrK/VzRdKYPPmfMRwMNIh2uHhr2EKmj1/EsMF3KKyISJ3lVWA5cuQITqeTmJgYj+0xMTFkZWVV6xh/+MMfiI+P9wg9Q4cOZf78+aSmpvLss8+ydOlShg0bhtPpPOsxpk+fTmRkpPuVkFA/H1U/ecjLhDld7LaZvPj+RH+XI43E9+s/Z4ntCAB3J9xFQlwnP1ckIlI1ny4c98wzz7BgwQI++ugjgoKC3Ntvu+02brzxRnr27MnIkSP55JNPWL16NUuWLDnrcaZOnUpubq77lZmZ6aMzqFmd2/ThRnvF5ONFpSvZsTfNvwVJo/Dx2lkA9C6xc9v1Wn1WROoHrwJL8+bNsVqtZGdne2zPzs4mNjb2vJ/961//yjPPPMOXX35Jr169ztu2ffv2NG/enF27dp11v91uJyIiwuNVXz14y2w6lBoUWC08tXgCrnP0KomcrqS0iPT921ixYTGLlv6Lr1f9h8Ki/PN+5vDxA6wydwOQ1OIaX5QpIlIjvLqt2Waz0a9fP1JTUxk5ciRQMek2NTWV++6775yfe+6553j66af54osv6N+/f5Xfs2/fPo4ePUpcXJw35dVLNpud+/s+xoObHyMtqJQX35/E72+b7e+ypA7KPrqf1z57mAPFGeSbhewOLCbf6vnfHIFbHqd1mYWWNKV1WEe6txrE5b1H0CSyBQB///gBDgdYiHK6+L9rH/LHaYiIXBCv12FJSUlh3Lhx9O/fn4EDBzJjxgwKCwsZP348AGPHjqVly5ZMn16x7Pyzzz7LtGnTePvtt2nbtq17rktYWBhhYWEUFBTwxBNPMGrUKGJjY9m9ezcPPfQQHTt2ZMiQITV4qnXXNQNGMXLrB7xvbuS94m9JXPcxV1x6o7/LkhpSVFLIf75+mZ+ObqZpSBzdEy7jsl7DCAkKrdbnl6d9xuyVj7LFXkq5YYC9ck9FWAkwTZo7oMBiUmC1sNtuspujUH4U0ldi2fMiLcuhpRnBD/aKHpjrA/u6Q4yISH3gdWAZPXo0hw8fZtq0aWRlZdGnTx8WL17snoibkZGB5ZTFp2bNmkVZWRm//OUvPY7z2GOP8fjjj2O1WtmwYQNvvPEGOTk5xMfHc/311/PUU09ht9tpLB75vzfYOmcgm+wO/rzuEV5v1UsPi2sA/pM6k3/89A8OB5z4/0TBBtj6BYFbHieh3CDOjCIhpANd4weS1HO4x4qwDkc5D837Ocus+ykNMgCDVuUmfawdaRoUQ9sWPRl++XhsATYCAgJxOZ2s2/Yta3d8we5jG9jvyCIzsITjVguZNsjk5HDRFd1G+fhKiIhcHD1LqA7Zk7mZe764lexAC71LbMwZ/z02W+MJbQ3NK+8/wL8Kv8JlGEQ5XXRyhJFPKZmB5RXL259FXLlJS2coLW2tcJrlfGJNB6BnSSATBzzO4N7DsVit1a7B5XSyfW8aK7d+xq5DaWSXH6RjWHf+8H//qpFzFBG5GLW6cFxd1FACC8CXK97hkW1PU2oxuKqsGX/7dapXP1BSNyxa8ipP/PQ3yg2DpNIIpt/6Ac2iKiamu5xO0rYvZ83O/5F+bAP7yw6wP6CIQwFnDzEDS0J5bcJ3+t+BiDQ4Ciz13D8/msrM3P9iGgZDHQk8f9dn/i5JvPDtuo95JG0qOVYL/UqC+ddd31VrQba9B3awYtOn7Mhew77inzhg5FFiuHjs0uma0yQiDZICSwPw/Dv/j/ll3wPwc1cHnh77gf4Lux5I276cSd/9P/KsFtqVweu//IIWTeL9XZaISJ1Ua0vzi+9Muf2f3GL0BOATy25S5gzB4Sj3c1VSlX9/+zR5VgvhThd/vvo1hRURkRqiwFKHTRv7NrdZ+wGQastm4pyryMk/4ueq5Fx27N3AMmsGABNajKZXp8v8XJGISMOhwFLH/fFX87g75FqspskP9nzufOdnbNjxvb/LkrN4ZvE9FFssdCw1uPOGP/q7HBGRBkWBpR743S0zeDj+LiKdLnbbTSYtv5u3v3jB32XJKd787BlWBxUSYJqM73q/5huJiNQwBZZ64rbrH+Bvg2bTrgxyrBamZ81j8r+u0xBRHbA9/UdezXoTgCvKo7nxyrv8XJGISMOjwFKPXNr1Ct649WuuLmsOQGpgFre/ezX/SZ3p58oar/zCHB7631hyrBYSykyevO09f5ckItIgKbDUM00iW/DK3d8wpcXtNHW42Bdo8MS+2dz96iC2p//o7/IanSlvjWCPDUJcLh7p/wxR4c39XZKISIOkwFJPjb3hEf59w0KuKmuOcWJC7tglv+KPc28m60imv8trFJauXcQK23EA7oz8OZf3/bmfKxIRabi0cFwD8Pl3b/La5ufZaa/4VxnpdHFdQE/uu3GGezl4qXkTXk1ipb2AfiXBzPt/q/xdjohIvaOF4xqZYYPv4L1fr2Vi+DBalpvkWi38x9zMiA+v5aHXh7N1z1p/l9jg5BYcY2NgHgCjut3r52pERBo+9bA0MCWlRcxe9Ac+y/uGg4EGAAGmSf+yCK5v/3+MumaibrmtAXP++wQvHfsPLRwuvhyXVq1nBYmIiCf1sDRiQfYQJt/6Cp+MXcv9TW6ma6kVh2Hwgz2fJ/f/k6Fze/OnuaPYsPMHf5dar608+BUAvVxxCisiIj6gHpZG4PPv3uSTLf9iTcARiiwVGdUwTTqXWekVfAk3DpxIny6X+7nK+sPhKOfK+X3It1p4rNW9/PLaSf4uSUSkXtLTmuWsjuZk8eaXf+H7nGVstTs99nUoM+gR0JFBnW7kuoG3Y7PZ/VRl3bd07SLu2/Qngl0ulv3faoLsIf4uSUSkXlJgkSpt2rWSRT/MYn1RGtttDlyG4d4X4XTR1RFBj8hL+Vmf2+nVMUnzXk7xzNt38Vb5KrqXBvDuPVr7RkTkQnnz+x3go5qkjrmkYyKXdEwEKp4yvPD7v7M1bz1bbQXkWS2sshawqmQZc39YRovlLto7I+gQ2pXETsO5vM+IRt0Ds6ZgDdihe1AXf5ciItJoqIdFPJSUFrH4+zdZmf4p2x0/kW5z4Til9wUqVnVtUx5IS0s07SN70LfTdQzsfm2jCDEORzmXvdmXUovBq32fJ6nXUH+XJCJSb2lISGrM0Zws/rd6ARv2LSO9/Cd2B5a6J+6eKsTlonV5IDFGE1qGtKFdi95c2uUaOrbq2aCGk9Zt+5ZxK39DgGmy4vZVmr8iInIRNCQkNaZZVCyjr5vMaCYDFT0wKzZ8zo97vmFv3jb2m4fZG+ikyGJhm93JNo6A4wgcXAsH5xDhdBHvCCTGiCLaHk98VEc6xveld6fLaRLZwr8ndwHSdn4DQFw5CisiIj6kwCJeCbKHcM2AUVwzYJR7W1lZKSs2Lmb9nm/Yl7eTLMchsq3FHAyAPKuFPKuTbRwF8ygc3wjHP4LN0MzhooUzkGaE0TwwhhZhrYiL6kC7+B50btOXyLCmfjzTs/v2wKcQBB2of2FLRKQ+U2CRi2az2bmq301c1e8mj+05+UdYvTmVbftWsi9vJ0ccRzhqFHLI6iTfauFogIWjAU4gt+JVtAOKvoYDwBqIcrpo6rQQ6QoiyhJOlK0ZUUHRNAuLJ7ZpO1pFd6Zdy26EBIX67Fz3WwsBgyvbjaqyrYiI1BwFFqk1UeHNue6y0VzH6DP2ZR7cycbdK0g/tJGDeXs4UnaIXAo4binniBVKLQY5Vgs5VoCSE6/DULKt4o9HgB0Vx4p0uohwGkSYgYSZQYRbwwkLiCTcFkVkcHOiwmJoEdmKmGZtaRXT/oJ7bsrKSjl84v8xXdr0v6BjiIjIhVFgEb9IiOtEQlyns+5zOZ3sP7SH7Rlp7DuyjezcvRwtySLXkUOBWUy+pYw8i4scq4HTMMi1Wsi1AjiAghOvg1BOxSuPil6bE4JcJuEuk1CXhVAzgCAzEJsRSJBhx2axE2QJISgglODAUEJsEYTYIwgLbsLR/P04DIMA06Rrm761fIVERORUCixS51is1vMGmkoORzkZB3eQkb2DrGPpHM7bx/HibPLKjlHoLKDILKHIKKfQcFBgNcm3GDgMgxJLxetwxVFOvIo9D+4CSk+88j13xZXTKG7hFhGpSxRYpN4KCAikfUIP2if0qFZ7l9NJ1rF9HDiUzqHjGRzNO0BO0SGKyvIocRRS4iyi1FlCqVlCqVlGqVlOqeGgDBclFhclhondNBiTMLaWz0xERE6nwCKNhsVqJb5FG+JbtPF3KSIi4qUzVwATERERqWMUWERERKTOu6DAMnPmTNq2bUtQUBCJiYmsWrXqvO3ff/99unbtSlBQED179uSzzz7z2G+aJtOmTSMuLo7g4GCSk5PZuXPnhZQmIiIiDZDXgeXdd98lJSWFxx57jHXr1tG7d2+GDBnCoUOHztr++++/5/bbb+euu+7ixx9/ZOTIkYwcOZJNmza52zz33HP87W9/Y/bs2axcuZLQ0FCGDBlCSUnJhZ+ZiIiINBheP/wwMTGRAQMG8Pe//x0Al8tFQkICv/3tb3n44YfPaD969GgKCwv55JNP3Nsuu+wy+vTpw+zZszFNk/j4eB588EF+//vfA5Cbm0tMTAzz5s3jtttuq7ImPfxQRESk/vHm99urHpaysjLWrl1LcnLyyQNYLCQnJ7NixYqzfmbFihUe7QGGDBnibp+enk5WVpZHm8jISBITE895zNLSUvLy8jxeIiIi0nB5FViOHDmC0+kkJibGY3tMTAxZWVln/UxWVtZ521f+05tjTp8+ncjISPcrISHBm9MQERGReqZe3iU0depUcnNz3a/MzEx/lyQiIiK1yKvA0rx5c6xWK9nZ2R7bs7OziY2NPetnYmNjz9u+8p/eHNNutxMREeHxEhERkYbLq8Bis9no168fqamp7m0ul4vU1FSSkpLO+pmkpCSP9gBfffWVu327du2IjY31aJOXl8fKlSvPeUwRERFpXLxemj8lJYVx48bRv39/Bg4cyIwZMygsLGT8+PEAjB07lpYtWzJ9+nQA7r//fq666ipeeOEFhg8fzoIFC1izZg2vvvoqAIZhMHnyZP785z/TqVMn2rVrx6OPPkp8fDwjR46suTMVERGResvrwDJ69GgOHz7MtGnTyMrKok+fPixevNg9aTYjIwOL5WTHzaBBg3j77bf505/+xCOPPEKnTp1YuHAhl1xyibvNQw89RGFhIffccw85OTlcfvnlLF68mKCgoBo4RREREanvvF6HpS7SOiwiIiL1T62twyIiIiLiD14PCdVFlZ1EWkBORESk/qj83a7OYE+DCCz5+fkAWkBORESkHsrPzycyMvK8bRrEHBaXy8WBAwcIDw/HMIwaPXZeXh4JCQlkZmZqfkwt0nX2HV1r39B19g1dZ9+oretsmib5+fnEx8d73LBzNg2ih8VisdCqVata/Q4tUOcbus6+o2vtG7rOvqHr7Bu1cZ2r6lmppEm3IiIiUucpsIiIiEidp8BSBbvdzmOPPYbdbvd3KQ2arrPv6Fr7hq6zb+g6+0ZduM4NYtKtiIiINGzqYREREZE6T4FFRERE6jwFFhEREanzFFhERESkzlNgqcLMmTNp27YtQUFBJCYmsmrVKn+XVG9Mnz6dAQMGEB4eTnR0NCNHjmT79u0ebUpKSpg0aRLNmjUjLCyMUaNGkZ2d7dEmIyOD4cOHExISQnR0NFOmTMHhcPjyVOqVZ555BsMwmDx5snubrnPN2b9/P7/61a9o1qwZwcHB9OzZkzVr1rj3m6bJtGnTiIuLIzg4mOTkZHbu3OlxjGPHjjFmzBgiIiKIiorirrvuoqCgwNenUmc5nU4effRR2rVrR3BwMB06dOCpp57yeN6MrrP3li1bxogRI4iPj8cwDBYuXOixv6au6YYNG7jiiisICgoiISGB5557rmZOwJRzWrBggWmz2cw5c+aYmzdvNu+++24zKirKzM7O9ndp9cKQIUPMuXPnmps2bTLT0tLMG264wWzdurVZUFDgbnPvvfeaCQkJZmpqqrlmzRrzsssuMwcNGuTe73A4zEsuucRMTk42f/zxR/Ozzz4zmzdvbk6dOtUfp1TnrVq1ymzbtq3Zq1cv8/7773dv13WuGceOHTPbtGlj3nnnnebKlSvNPXv2mF988YW5a9cud5tnnnnGjIyMNBcuXGiuX7/evPHGG8127dqZxcXF7jZDhw41e/fubf7www/mt99+a3bs2NG8/fbb/XFKddLTTz9tNmvWzPzkk0/M9PR08/333zfDwsLMl19+2d1G19l7n332mfnHP/7R/PDDD03A/Oijjzz218Q1zc3NNWNiYswxY8aYmzZtMt955x0zODjY/Oc//3nR9SuwnMfAgQPNSZMmud87nU4zPj7enD59uh+rqr8OHTpkAubSpUtN0zTNnJwcMzAw0Hz//ffdbbZu3WoC5ooVK0zTrPg/mMViMbOystxtZs2aZUZERJilpaW+PYE6Lj8/3+zUqZP51VdfmVdddZU7sOg615w//OEP5uWXX37O/S6Xy4yNjTWff/5597acnBzTbreb77zzjmmaprllyxYTMFevXu1u8/nnn5uGYZj79++vveLrkeHDh5u//vWvPbbdfPPN5pgxY0zT1HWuCacHlpq6pv/4xz/MJk2aePy98Yc//MHs0qXLRdesIaFzKCsrY+3atSQnJ7u3WSwWkpOTWbFihR8rq79yc3MBaNq0KQBr166lvLzc4xp37dqV1q1bu6/xihUr6NmzJzExMe42Q4YMIS8vj82bN/uw+rpv0qRJDB8+3ON6gq5zTfr444/p378/t9xyC9HR0fTt25fXXnvNvT89PZ2srCyPax0ZGUliYqLHtY6KiqJ///7uNsnJyVgsFlauXOm7k6nDBg0aRGpqKjt27ABg/fr1LF++nGHDhgG6zrWhpq7pihUruPLKK7HZbO42Q4YMYfv27Rw/fvyiamwQDz+sDUeOHMHpdHr8BQ4QExPDtm3b/FRV/eVyuZg8eTKDBw/mkksuASArKwubzUZUVJRH25iYGLKystxtzvbvoHKfVFiwYAHr1q1j9erVZ+zTda45e/bsYdasWaSkpPDII4+wevVqfve732Gz2Rg3bpz7Wp3tWp56raOjoz32BwQE0LRpU13rEx5++GHy8vLo2rUrVqsVp9PJ008/zZgxYwB0nWtBTV3TrKws2rVrd8YxKvc1adLkgmtUYBGfmDRpEps2bWL58uX+LqXByczM5P777+err74iKCjI3+U0aC6Xi/79+/OXv/wFgL59+7Jp0yZmz57NuHHj/Fxdw/Hee+/x1ltv8fbbb9OjRw/S0tKYPHky8fHxus6NmIaEzqF58+ZYrdYz7qTIzs4mNjbWT1XVT/fddx+ffPIJ33zzDa1atXJvj42NpaysjJycHI/2p17j2NjYs/47qNwnFUM+hw4d4tJLLyUgIICAgACWLl3K3/72NwICAoiJidF1riFxcXF0797dY1u3bt3IyMgATl6r8/29ERsby6FDhzz2OxwOjh07pmt9wpQpU3j44Ye57bbb6NmzJ3fccQcPPPAA06dPB3Sda0NNXdPa/LtEgeUcbDYb/fr1IzU11b3N5XKRmppKUlKSHyurP0zT5L777uOjjz7i66+/PqObsF+/fgQGBnpc4+3bt5ORkeG+xklJSWzcuNHj/yRfffUVERERZ/xwNFbXXnstGzduJC0tzf3q378/Y8aMcf9Z17lmDB48+Ixb83fs2EGbNm0AaNeuHbGxsR7XOi8vj5UrV3pc65ycHNauXetu8/XXX+NyuUhMTPTBWdR9RUVFWCyeP09WqxWXywXoOteGmrqmSUlJLFu2jPLycnebr776ii5dulzUcBCg25rPZ8GCBabdbjfnzZtnbtmyxbznnnvMqKgojzsp5NwmTpxoRkZGmkuWLDEPHjzofhUVFbnb3HvvvWbr1q3Nr7/+2lyzZo2ZlJRkJiUlufdX3m57/fXXm2lpaebixYvNFi1a6HbbKpx6l5Bp6jrXlFWrVpkBAQHm008/be7cudN86623zJCQEPPf//63u80zzzxjRkVFmYsWLTI3bNhg3nTTTWe9NbRv377mypUrzeXLl5udOnVq1Lfbnm7cuHFmy5Yt3bc1f/jhh2bz5s3Nhx56yN1G19l7+fn55o8//mj++OOPJmC++OKL5o8//mju3bvXNM2auaY5OTlmTEyMeccdd5ibNm0yFyxYYIaEhOi2Zl945ZVXzNatW5s2m80cOHCg+cMPP/i7pHoDOOtr7ty57jbFxcXmb37zG7NJkyZmSEiI+Ytf/MI8ePCgx3F++uknc9iwYWZwcLDZvHlz88EHHzTLy8t9fDb1y+mBRde55vz3v/81L7nkEtNut5tdu3Y1X331VY/9LpfLfPTRR82YmBjTbreb1157rbl9+3aPNkePHjVvv/12MywszIyIiDDHjx9v5ufn+/I06rS8vDzz/vvvN1u3bm0GBQWZ7du3N//4xz963Cqr6+y9b7755qx/J48bN840zZq7puvXrzcvv/xy0263my1btjSfeeaZGqnfMM1Tlg4UERERqYM0h0VERETqPAUWERERqfMUWERERKTOU2ARERGROk+BRUREROo8BRYRERGp8xRYREREpM5TYBEREZE6T4FFRBoswzBYuHChv8sQkRqgwCIiteLOO+/EMIwzXkOHDvV3aSJSDwX4uwARabiGDh3K3LlzPbbZ7XY/VSMi9Zl6WESk1tjtdmJjYz1elY+YNwyDWbNmMWzYMIKDg2nfvj3/+c9/PD6/ceNGfvaznxEcHEyzZs245557KCgo8GgzZ84cevTogd1uJy4ujvvuu89j/5EjR/jFL35BSEgInTp14uOPP67dkxaRWqHAIiJ+8+ijjzJq1CjWr1/PmDFjuO2229i6dSsAhYWFDBkyhCZNmrB69Wref/99/ve//3kEklmzZjFp0iTuueceNm7cyMcff0zHjh09vuOJJ57g1ltvZcOGDdxwww2MGTOGY8eO+fQ8RaQG1Mgzn0VETjNu3DjTarWaoaGhHq+nn37aNE3TBMx7773X4zOJiYnmxIkTTdM0zVdffdVs0qSJWVBQ4N7/6aefmhaLxczKyjJN0zTj4+PNP/7xj+esATD/9Kc/ud8XFBSYgPn555/X2HmKiG9oDouI1JprrrmGWbNmeWxr2rSp+89JSUke+5KSkkhLSwNg69at9O7dm9DQUPf+wYMH43K52L59O4ZhcODAAa699trz1tCrVy/3n0NDQ4mIiODQoUMXekoi4icKLCJSa0JDQ88YoqkpwcHB1WoXGBjo8d4wDFwuV22UJCK1SHNYRMRvfvjhhzPed+vWDYBu3bqxfv16CgsL3fu/++47LBYLXbp0ITw8nLZt25KamurTmkXEP9TDIiK1prS0lKysLI9tAQEBNG/eHID333+f/v37c/nll/PWW2+xatUqXn/9dQDGjBnDY489xrhx43j88cc5fPgwv/3tb7njjjuIiYkB4PHHH+fee+8lOjqaYcOGkZ+fz3fffcdvf/tb356oiNQ6BRYRqTWLFy8mLi7OY1uXLl3Ytm0bUHEHz4IFC/jNb35DXFwc77zzDt27dwcgJCSEL774gvvvv58BAwYQEhLCqFGjePHFF93HGjduHCUlJbz00kv8/ve/p3nz5vzyl7/03QmKiM8Ypmma/i5CRBofwzD46KOPGDlypL9LEZF6QHNYREREpM5TYBEREZE6T3NYRMQvNBotIt5QD4uIiIjUeQosIiIiUucpsIiIiEidp8AiIiIidZ4Ci4iIiNR5CiwiIiJS5ymwiIiISJ2nwCIiIiJ13v8HNfk8nIGPHDQAAAAASUVORK5CYII=\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1400x500 with 10 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1000/30000 [06:30<5:47:30,  1.39it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1001/30000 [06:30<5:05:49,  1.58it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1002/30000 [06:30<4:34:16,  1.76it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1003/30000 [06:31<4:12:41,  1.91it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1004/30000 [06:31<3:57:15,  2.04it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1005/30000 [06:32<3:48:10,  2.12it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1006/30000 [06:32<3:48:34,  2.11it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1007/30000 [06:33<4:11:32,  1.92it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1008/30000 [06:33<4:22:29,  1.84it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1009/30000 [06:34<4:34:09,  1.76it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 10ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1010/30000 [06:35<4:42:33,  1.71it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 21ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1011/30000 [06:35<4:52:28,  1.65it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 10ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1012/30000 [06:36<4:39:05,  1.73it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1013/30000 [06:36<4:16:10,  1.89it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1014/30000 [06:37<4:00:41,  2.01it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1015/30000 [06:37<3:50:50,  2.09it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1016/30000 [06:37<3:42:34,  2.17it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1017/30000 [06:38<3:39:03,  2.21it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1018/30000 [06:38<3:35:12,  2.24it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1019/30000 [06:39<3:34:53,  2.25it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1020/30000 [06:39<3:33:13,  2.27it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1021/30000 [06:40<3:29:16,  2.31it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1022/30000 [06:40<3:28:27,  2.32it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1023/30000 [06:41<3:33:31,  2.26it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1024/30000 [06:41<3:35:43,  2.24it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 11ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1025/30000 [07:18<91:23:34, 11.36s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 71ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1026/30000 [07:19<65:55:47,  8.19s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 13ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1027/30000 [07:19<47:49:27,  5.94s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 12ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1028/30000 [07:20<34:33:06,  4.29s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1029/30000 [07:20<25:14:30,  3.14s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1030/30000 [07:21<18:42:35,  2.33s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1031/30000 [07:21<14:07:59,  1.76s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1032/30000 [07:21<10:55:07,  1.36s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1033/30000 [07:22<8:39:06,  1.08s/it] "
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1034/30000 [07:22<7:06:45,  1.13it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1035/30000 [07:23<6:01:26,  1.34it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1036/30000 [07:23<5:16:40,  1.52it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1037/30000 [07:24<4:43:46,  1.70it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1038/30000 [07:24<4:41:39,  1.71it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 12ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1039/30000 [07:25<4:53:56,  1.64it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 11ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1040/30000 [07:26<4:58:28,  1.62it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1041/30000 [07:26<5:03:43,  1.59it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 15ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1042/30000 [07:27<5:06:46,  1.57it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1043/30000 [07:27<4:39:54,  1.72it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1044/30000 [07:28<4:16:16,  1.88it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1045/30000 [07:28<4:03:48,  1.98it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1046/30000 [07:29<3:52:20,  2.08it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1047/30000 [07:29<3:44:00,  2.15it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1048/30000 [07:29<3:44:05,  2.15it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  3%|▎         | 1049/30000 [07:30<3:38:33,  2.21it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1050/30000 [07:30<3:37:48,  2.22it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1051/30000 [07:31<3:34:20,  2.25it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1052/30000 [07:31<3:29:29,  2.30it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1053/30000 [07:32<3:30:28,  2.29it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1054/30000 [07:32<3:28:06,  2.32it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1055/30000 [07:32<3:29:40,  2.30it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1056/30000 [07:33<3:27:36,  2.32it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1057/30000 [07:33<3:26:08,  2.34it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1058/30000 [07:34<3:28:45,  2.31it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1059/30000 [07:34<3:27:50,  2.32it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1060/30000 [07:35<3:28:22,  2.31it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1061/30000 [07:35<3:26:34,  2.33it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1062/30000 [07:35<3:24:57,  2.35it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1063/30000 [07:36<3:25:49,  2.34it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1064/30000 [07:36<3:24:13,  2.36it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1065/30000 [07:37<3:26:07,  2.34it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1066/30000 [07:37<3:52:54,  2.07it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 10ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1067/30000 [07:38<4:12:07,  1.91it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 11ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1068/30000 [07:39<4:28:16,  1.80it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 18ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1069/30000 [07:39<4:46:23,  1.68it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 11ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1070/30000 [07:40<4:43:03,  1.70it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1071/30000 [07:40<4:19:45,  1.86it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1072/30000 [07:41<4:05:15,  1.97it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1073/30000 [07:41<3:52:48,  2.07it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1074/30000 [07:42<3:43:57,  2.15it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1075/30000 [07:42<3:38:39,  2.20it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1076/30000 [07:42<3:34:25,  2.25it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1077/30000 [07:43<3:33:18,  2.26it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1078/30000 [07:43<3:31:32,  2.28it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1079/30000 [07:44<3:30:29,  2.29it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1080/30000 [07:44<3:29:54,  2.30it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1081/30000 [07:45<3:29:52,  2.30it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1082/30000 [07:45<3:29:00,  2.31it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1083/30000 [07:45<3:27:59,  2.32it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1084/30000 [07:46<3:27:51,  2.32it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1085/30000 [07:46<3:24:44,  2.35it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1086/30000 [07:47<3:25:13,  2.35it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1087/30000 [07:47<3:26:49,  2.33it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 10ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1088/30000 [07:48<3:27:06,  2.33it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1089/30000 [07:48<3:27:06,  2.33it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1090/30000 [07:48<3:29:34,  2.30it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1091/30000 [07:49<3:27:34,  2.32it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 7ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1092/30000 [07:49<3:28:51,  2.31it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 6ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1093/30000 [07:50<3:43:02,  2.16it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\r  4%|▎         | 1094/30000 [08:01<29:33:13,  3.68s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3s/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "43udmjgIz-yc"
      },
      "source": [
        "# Attempt 2: Convolution"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rCeHJEaZz-yd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 383
        },
        "outputId": "559dc7f8-b697-4c94-b20b-53d0a8d1ecce"
      },
      "source": [
        "from keras.models import Sequential\n",
        "from keras.layers import Dense\n",
        "from keras.layers import Reshape\n",
        "from keras.layers.core import Activation\n",
        "from keras.layers.normalization import BatchNormalization\n",
        "from keras.layers.convolutional import UpSampling2D\n",
        "from keras.layers.convolutional import Convolution2D, MaxPooling2D\n",
        "from keras.layers.core import Flatten\n",
        "from keras.optimizers import SGD\n",
        "from keras.datasets import mnist\n",
        "import numpy as np\n",
        "#from PIL import Image\n",
        "import argparse\n",
        "import math\n",
        "\n",
        "def generator_model():\n",
        "    model = Sequential()\n",
        "    model.add(Dense(input_dim=100, units=1024))\n",
        "    model.add(Activation('tanh'))\n",
        "    model.add(Dense(128*7*7))\n",
        "    model.add(BatchNormalization())\n",
        "    model.add(Activation('tanh'))\n",
        "    model.add(Reshape((128, 7, 7), input_shape=(128*7*7,)))\n",
        "    model.add(UpSampling2D(size=(2, 2)))\n",
        "    model.add(Convolution2D(64, 5, 5, border_mode='same'))\n",
        "    model.add(Activation('tanh'))\n",
        "    model.add(UpSampling2D(size=(2, 2)))\n",
        "    model.add(Convolution2D(1, 5, 5, border_mode='same'))\n",
        "    model.add(Activation('tanh'))\n",
        "    return model\n",
        "\n",
        "\n",
        "def discriminator_model():\n",
        "    model = Sequential()\n",
        "    model.add(Convolution2D(\n",
        "                        64, 5, 5,\n",
        "                        border_mode='same',\n",
        "                        input_shape=(1, 28, 28)))\n",
        "    model.add(Activation('tanh'))\n",
        "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "    model.add(Convolution2D(128, 5, 5))\n",
        "    model.add(Activation('tanh'))\n",
        "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(1024))\n",
        "    model.add(Activation('tanh'))\n",
        "    model.add(Dense(1))\n",
        "    model.add(Activation('sigmoid'))\n",
        "    return model\n",
        "\n",
        "def generator_containing_discriminator(generator, discriminator):\n",
        "    model = Sequential()\n",
        "    model.add(generator)\n",
        "    discriminator.trainable = False\n",
        "    model.add(discriminator)\n",
        "    return model\n",
        "\n",
        "num_epochs  = 10000\n",
        "plt_frq     = 500\n",
        "batchsize   = 100\n",
        "input_shape = 10 # Number of random inputs to generator.\n",
        "lr          = .01\n",
        "momgen      = 0.0\n",
        "momdisc     = 0.0\n",
        "clipnorm    = 1.\n",
        "decay     =   1e-5\n",
        "\n",
        "discriminator_on_generator = generator_containing_discriminator(generator, discriminator)\n",
        "d_optim = SGD(lr=0.0005, momentum=0.9, nesterov=True)\n",
        "g_optim = SGD(lr=0.0005, momentum=0.9, nesterov=True)\n",
        "generator.compile(loss='binary_crossentropy', optimizer=\"SGD\")\n",
        "discriminator_on_generator.compile(loss='binary_crossentropy', optimizer=g_optim)\n",
        "discriminator.trainable = True\n",
        "discriminator.compile(loss='binary_crossentropy', optimizer=d_optim)\n",
        "\n",
        "history = defaultdict(list)\n",
        "for epoch in tqdm(range(num_epochs)):\n",
        "\n",
        "    # Generate images.\n",
        "    noisebatch = np.random.uniform(-1, 1, size=[batchsize,input_shape])\n",
        "    generated  = generator.predict(noisebatch)\n",
        "\n",
        "    # Update discriminator.\n",
        "    databatch  = X_train[np.random.randint(0,X_train.shape[0], size=batchsize), :]\n",
        "    X          = np.concatenate([databatch, generated])\n",
        "    T          = np.concatenate([np.ones((databatch.shape[0],)), np.zeros((generated.shape[0],))]) # Real:1, fake:0\n",
        "    discriminator.trainable = True\n",
        "    d_loss  = discriminator.train_on_batch(X,T)\n",
        "    history[\"d_loss\"].append(d_loss)\n",
        "\n",
        "    # Update generator.\n",
        "    noisebatch = np.random.uniform(-1, 1, size=[batchsize,input_shape])\n",
        "    T          = np.ones((noisebatch.shape[0],)) # Try to generate data that looks real.\n",
        "    discriminator.trainable = False\n",
        "    g_loss = gan.train_on_batch(noisebatch, T)\n",
        "    history[\"g_loss\"].append(g_loss)\n",
        "\n",
        "    # Updates plots\n",
        "    if epoch % plt_frq == plt_frq - 1:\n",
        "        display.clear_output(wait=True)\n",
        "        display.display(plt.gcf())\n",
        "        plot_loss(history)\n",
        "        plot_gen(generator, input_shape=input_shape)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "error",
          "ename": "ModuleNotFoundError",
          "evalue": "No module named 'keras.layers.core'",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-10-b2966c1e3a68>\u001b[0m in \u001b[0;36m<cell line: 4>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mDense\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mReshape\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcore\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mActivation\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnormalization\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mBatchNormalization\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvolutional\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mUpSampling2D\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'keras.layers.core'",
            "",
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0;32m\nNOTE: If your import is failing due to a missing package, you can\nmanually install dependencies using either !pip or !apt.\n\nTo view examples of installing some common dependencies, click the\n\"Open Examples\" button below.\n\u001b[0;31m---------------------------------------------------------------------------\u001b[0m\n"
          ],
          "errorDetails": {
            "actions": [
              {
                "action": "open_url",
                "actionText": "Open Examples",
                "url": "/notebooks/snippets/importing_libraries.ipynb"
              }
            ]
          }
        }
      ]
    }
  ]
}