Conv_2_LTH_CIFAR10.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Lottery Ticket Hypothesis: _Conv-2_ CNN for CIFAR10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import math\n",
    "import tensorflow_model_optimization as tfmot\n",
    "from tensorflow_model_optimization.sparsity import keras as sparsity\n",
    "# from tensorflow.keras import datasets, layers, models\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras.layers import AveragePooling2D, Conv2D, MaxPooling2D, ReLU\n",
    "from tensorflow.keras import models, layers, datasets\n",
    "from tensorflow.keras.layers import Dense, Flatten, Reshape, Input, InputLayer\n",
    "from tensorflow.keras.models import Sequential, Model\n",
    "from tensorflow.keras.initializers import RandomNormal\n",
    "# import math\n",
    "from sklearn.metrics import accuracy_score, precision_score, recall_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.0.0'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: CUDA_DEVICE_ORDER=PCI_BUS_ID\n",
      "env: CUDA_VISIBLE_DEVICES=2\n"
     ]
    }
   ],
   "source": [
    "%env CUDA_DEVICE_ORDER=PCI_BUS_ID\n",
    "%env CUDA_VISIBLE_DEVICES=2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 60\n",
    "num_classes = 10\n",
    "num_epochs = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Data preprocessing and cleaning:\n",
    "# input image dimensions\n",
    "img_rows, img_cols = 32, 32\n",
    "\n",
    "# Load CIFAR-10 dataset-\n",
    "(X_train, y_train), (X_test, y_test) = tf.keras.datasets.cifar10.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "'input_shape' which will be used = (32, 32, 3)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "if tf.keras.backend.image_data_format() == 'channels_first':\n",
    "    X_train = X_train.reshape(X_train.shape[0], 3, img_rows, img_cols)\n",
    "    X_test = X_test.reshape(X_test.shape[0], 3, img_rows, img_cols)\n",
    "    input_shape = (3, img_rows, img_cols)\n",
    "else:\n",
    "    X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 3)\n",
    "    X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 3)\n",
    "    input_shape = (img_rows, img_cols, 3)\n",
    "\n",
    "print(\"\\n'input_shape' which will be used = {0}\\n\".format(input_shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Convert datasets to floating point types-\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "\n",
    "# Normalize the training and testing datasets-\n",
    "X_train /= 255.0\n",
    "X_test /= 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert class vectors/target to binary class matrices or one-hot encoded values-\n",
    "y_train = tf.keras.utils.to_categorical(y_train, num_classes)\n",
    "y_test = tf.keras.utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Dimensions of training and testing sets are:\n",
      "X_train.shape = (50000, 32, 32, 3), y_train.shape = (50000, 10)\n",
      "X_test.shape = (10000, 32, 32, 3), y_test.shape = (10000, 10)\n"
     ]
    }
   ],
   "source": [
    "print(\"\\nDimensions of training and testing sets are:\")\n",
    "print(\"X_train.shape = {0}, y_train.shape = {1}\".format(X_train.shape, y_train.shape))\n",
    "print(\"X_test.shape = {0}, y_test.shape = {1}\".format(X_test.shape, y_test.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prepare CIFAR10 dataset for _GradientTape_ training:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create training and testing datasets-\n",
    "train_dataset = tf.data.Dataset.from_tensor_slices((X_train, y_train))\n",
    "test_dataset = tf.data.Dataset.from_tensor_slices((X_test, y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dataset = train_dataset.shuffle(buffer_size = 20000, reshuffle_each_iteration = True).batch(batch_size = batch_size, drop_remainder = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_dataset = test_dataset.batch(batch_size=batch_size, drop_remainder=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Choose an optimizer and loss function for training-\n",
    "loss_fn = tf.keras.losses.CategoricalCrossentropy()\n",
    "optimizer = tf.keras.optimizers.Adam(lr = 0.0002)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Select metrics to measure the error & accuracy of model.\n",
    "# These metrics accumulate the values over epochs and then\n",
    "# print the overall result-\n",
    "train_loss = tf.keras.metrics.Mean(name = 'train_loss')\n",
    "train_accuracy = tf.keras.metrics.CategoricalAccuracy(name = 'train_accuracy')\n",
    "\n",
    "test_loss = tf.keras.metrics.Mean(name = 'test_loss')\n",
    "test_accuracy = tf.keras.metrics.CategoricalAccuracy(name = 'test_accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "'end_step parameter' for this dataset =  83400\n"
     ]
    }
   ],
   "source": [
    "# The model is first trained without any pruning for 'num_epochs' epochs-\n",
    "epochs = num_epochs\n",
    "\n",
    "num_train_samples = X_train.shape[0]\n",
    "\n",
    "end_step = np.ceil(1.0 * num_train_samples / batch_size).astype(np.int32) * epochs\n",
    "\n",
    "print(\"'end_step parameter' for this dataset =  {0}\".format(end_step))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Specify the parameters to be used for layer-wise pruning, NO PRUNING is done here:\n",
    "pruning_params_unpruned = {\n",
    "    'pruning_schedule': sparsity.ConstantSparsity(\n",
    "        target_sparsity=0.0, begin_step=0,\n",
    "        end_step = end_step, frequency=100\n",
    "    )\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "l = tf.keras.layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def pruned_nn(pruning_params_conv, pruning_params_fc):\n",
    "    \"\"\"\n",
    "    Function to define the architecture of a neural network model\n",
    "    following Conv-2 architecture for CIFAR-10 dataset and using\n",
    "    provided parameter which are used to prune the model.\n",
    "    \n",
    "    Conv-2 architecture-\n",
    "    64, 64, pool  -- convolutions\n",
    "    256, 256, 10  -- fully connected layers\n",
    "    \n",
    "    Input: 'pruning_params' Python 3 dictionary containing parameters which are used for pruning\n",
    "    Output: Returns designed and compiled neural network model\n",
    "    \"\"\"\n",
    "    \n",
    "    pruned_model = Sequential()\n",
    "    \n",
    "    pruned_model.add(sparsity.prune_low_magnitude(\n",
    "        Conv2D(\n",
    "            filters = 64, kernel_size = (3, 3),\n",
    "            activation='relu', kernel_initializer = tf.initializers.GlorotUniform(),\n",
    "            strides = (1, 1), padding = 'same',\n",
    "            input_shape=(32, 32, 3)\n",
    "        ),\n",
    "        **pruning_params_conv)\n",
    "    )\n",
    "        \n",
    "    pruned_model.add(sparsity.prune_low_magnitude(\n",
    "        Conv2D(\n",
    "            filters = 64, kernel_size = (3, 3),\n",
    "            activation='relu', kernel_initializer = tf.initializers.GlorotUniform(),\n",
    "            strides = (1, 1), padding = 'same'\n",
    "        ),\n",
    "        **pruning_params_conv)\n",
    "    )\n",
    "    \n",
    "    pruned_model.add(sparsity.prune_low_magnitude(\n",
    "        MaxPooling2D(\n",
    "            pool_size = (2, 2),\n",
    "            strides = (2, 2)\n",
    "        ),\n",
    "        **pruning_params_conv)\n",
    "    )\n",
    "    \n",
    "    pruned_model.add(Flatten())\n",
    "    \n",
    "    pruned_model.add(sparsity.prune_low_magnitude(\n",
    "        Dense(\n",
    "            units = 256, activation='relu',\n",
    "            kernel_initializer = tf.initializers.GlorotUniform()\n",
    "        ),\n",
    "        **pruning_params_fc)\n",
    "    )\n",
    "    \n",
    "    pruned_model.add(sparsity.prune_low_magnitude(\n",
    "        Dense(\n",
    "            units = 256, activation='relu',\n",
    "            kernel_initializer = tf.initializers.GlorotUniform()\n",
    "        ),\n",
    "        **pruning_params_fc)\n",
    "    )\n",
    "    \n",
    "    pruned_model.add(sparsity.prune_low_magnitude(\n",
    "        Dense(\n",
    "            units = 10, activation='softmax'\n",
    "        ),\n",
    "        **pruning_params_fc)\n",
    "    )\n",
    "    \n",
    "\n",
    "    # Compile pruned CNN-\n",
    "    pruned_model.compile(\n",
    "        loss=tf.keras.losses.categorical_crossentropy,\n",
    "        # optimizer='adam',\n",
    "        optimizer=tf.keras.optimizers.Adam(lr = 0.0002),\n",
    "        metrics=['accuracy']\n",
    "    )\n",
    "    \n",
    "    \n",
    "    return pruned_model\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Add a pruning step callback to peg the pruning step to the optimizer's\n",
    "# step. Also add a callback to add pruning summaries to tensorboard\n",
    "callback = [\n",
    "             sparsity.UpdatePruningStep(),\n",
    "             # sparsity.PruningSummaries(log_dir = logdir, profile_batch=0),\n",
    "             tf.keras.callbacks.EarlyStopping(\n",
    "                 monitor='val_loss', patience = 3,\n",
    "                 min_delta=0.001\n",
    "             )\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow_model_optimization/python/core/sparsity/keras/pruning_wrapper.py:183: Layer.add_variable (from tensorflow.python.keras.engine.base_layer) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `layer.add_weight` method instead.\n"
     ]
    }
   ],
   "source": [
    "# Initialize a CNN model-\n",
    "orig_model = pruned_nn(pruning_params_unpruned, pruning_params_unpruned)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Strip model of it's pruning parameters\n",
    "orig_model_stripped = sparsity.strip_pruning(orig_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save random weights-\n",
    "orig_model.save_weights(\"Conv_2_CIFAR_Ramdom_Weights.h5\", overwrite=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save random weights-\n",
    "orig_model.save_weights(\"Conv_2_CIFAR_Winning_Ticket.h5\", overwrite=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 32, 32, 64)        1792      \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 32, 32, 64)        36928     \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 16, 16, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 16384)             0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 256)               4194560   \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 256)               65792     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                2570      \n",
      "=================================================================\n",
      "Total params: 4,301,642\n",
      "Trainable params: 4,301,642\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Get CNN summary-\n",
    "orig_model_stripped.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Layer-wise pruning:\n",
    "The Lottery Ticket Hypothesis follows a _layer-wise pruning_ heuristic, where the fully-connected (fc) layers are pruned at the rate of 20% per iterative\n",
    "pruning round and the convolutional (conv) layers are pruned at the rate of 10% per iterative pruning round.\n",
    "\n",
    "Therefore, there is a need to compute the percentage of weights being pruned for each layer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Total number of trainable parameters = 4301642\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# number of convolutional parameters-\n",
    "conv1 = 1792\n",
    "conv2 = 36928\n",
    "\n",
    "# number of fully-connected dense parameters-\n",
    "dense1 = 4194560\n",
    "dense2 = 65792\n",
    "op_layer = 2570\n",
    "\n",
    "\n",
    "# total number of parameters-\n",
    "total_params = conv1 + conv2 + dense1 + dense2 + op_layer\n",
    "\n",
    "print(\"\\nTotal number of trainable parameters = {0}\\n\".format(total_params))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# maximum pruning performed is till 0.5% of all parameters-\n",
    "max_pruned_params = 0.005 * total_params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "loc_tot_params = total_params\n",
    "loc_conv1 = conv1\n",
    "loc_conv2 = conv2\n",
    "loc_dense1 = dense1\n",
    "loc_dense2 = dense2\n",
    "loc_op_layer = op_layer\n",
    "\n",
    "# variable to count number of pruning rounds-\n",
    "n = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Lists to hold percentage of weights pruned in each round for all layers in CNN-\n",
    "conv1_pruning = []\n",
    "conv2_pruning = []\n",
    "dense1_pruning = []\n",
    "dense2_pruning = []\n",
    "op_layer_pruning = []\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Conv1 = 1612.800, Conv2 = 33235.200\n",
      "Dense1 = 3355648.000, Dense2 = 52633.600 & O/p layer = 2056.000\n",
      "Total number of parameters = 3445185.600\n",
      "\n",
      "\n",
      "Conv1 = 1451.520, Conv2 = 29911.680\n",
      "Dense1 = 2684518.400, Dense2 = 42106.880 & O/p layer = 1644.800\n",
      "Total number of parameters = 2759633.280\n",
      "\n",
      "\n",
      "Conv1 = 1306.368, Conv2 = 26920.512\n",
      "Dense1 = 2147614.720, Dense2 = 33685.504 & O/p layer = 1315.840\n",
      "Total number of parameters = 2210842.944\n",
      "\n",
      "\n",
      "Conv1 = 1175.731, Conv2 = 24228.461\n",
      "Dense1 = 1718091.776, Dense2 = 26948.403 & O/p layer = 1052.672\n",
      "Total number of parameters = 1771497.043\n",
      "\n",
      "\n",
      "Conv1 = 1058.158, Conv2 = 21805.615\n",
      "Dense1 = 1374473.421, Dense2 = 21558.723 & O/p layer = 842.138\n",
      "Total number of parameters = 1419738.054\n",
      "\n",
      "\n",
      "Conv1 = 952.342, Conv2 = 19625.053\n",
      "Dense1 = 1099578.737, Dense2 = 17246.978 & O/p layer = 673.710\n",
      "Total number of parameters = 1138076.820\n",
      "\n",
      "\n",
      "Conv1 = 857.108, Conv2 = 17662.548\n",
      "Dense1 = 879662.989, Dense2 = 13797.582 & O/p layer = 538.968\n",
      "Total number of parameters = 912519.196\n",
      "\n",
      "\n",
      "Conv1 = 771.397, Conv2 = 15896.293\n",
      "Dense1 = 703730.391, Dense2 = 11038.066 & O/p layer = 431.174\n",
      "Total number of parameters = 731867.322\n",
      "\n",
      "\n",
      "Conv1 = 694.258, Conv2 = 14306.664\n",
      "Dense1 = 562984.313, Dense2 = 8830.453 & O/p layer = 344.940\n",
      "Total number of parameters = 587160.627\n",
      "\n",
      "\n",
      "Conv1 = 624.832, Conv2 = 12875.997\n",
      "Dense1 = 450387.451, Dense2 = 7064.362 & O/p layer = 275.952\n",
      "Total number of parameters = 471228.594\n",
      "\n",
      "\n",
      "Conv1 = 562.349, Conv2 = 11588.398\n",
      "Dense1 = 360309.960, Dense2 = 5651.490 & O/p layer = 220.761\n",
      "Total number of parameters = 378332.958\n",
      "\n",
      "\n",
      "Conv1 = 506.114, Conv2 = 10429.558\n",
      "Dense1 = 288247.968, Dense2 = 4521.192 & O/p layer = 176.609\n",
      "Total number of parameters = 303881.441\n",
      "\n",
      "\n",
      "Conv1 = 455.502, Conv2 = 9386.602\n",
      "Dense1 = 230598.375, Dense2 = 3616.953 & O/p layer = 141.287\n",
      "Total number of parameters = 244198.720\n",
      "\n",
      "\n",
      "Conv1 = 409.952, Conv2 = 8447.942\n",
      "Dense1 = 184478.700, Dense2 = 2893.563 & O/p layer = 113.030\n",
      "Total number of parameters = 196343.186\n",
      "\n",
      "\n",
      "Conv1 = 368.957, Conv2 = 7603.148\n",
      "Dense1 = 147582.960, Dense2 = 2314.850 & O/p layer = 90.424\n",
      "Total number of parameters = 157960.338\n",
      "\n",
      "\n",
      "Conv1 = 332.061, Conv2 = 6842.833\n",
      "Dense1 = 118066.368, Dense2 = 1851.880 & O/p layer = 72.339\n",
      "Total number of parameters = 127165.481\n",
      "\n",
      "\n",
      "Conv1 = 298.855, Conv2 = 6158.550\n",
      "Dense1 = 94453.094, Dense2 = 1481.504 & O/p layer = 57.871\n",
      "Total number of parameters = 102449.874\n",
      "\n",
      "\n",
      "Conv1 = 268.970, Conv2 = 5542.695\n",
      "Dense1 = 75562.475, Dense2 = 1185.203 & O/p layer = 46.297\n",
      "Total number of parameters = 82605.640\n",
      "\n",
      "\n",
      "Conv1 = 242.073, Conv2 = 4988.425\n",
      "Dense1 = 60449.980, Dense2 = 948.163 & O/p layer = 37.038\n",
      "Total number of parameters = 66665.678\n",
      "\n",
      "\n",
      "Conv1 = 217.865, Conv2 = 4489.583\n",
      "Dense1 = 48359.984, Dense2 = 758.530 & O/p layer = 29.630\n",
      "Total number of parameters = 53855.593\n",
      "\n",
      "\n",
      "Conv1 = 196.079, Conv2 = 4040.624\n",
      "Dense1 = 38687.987, Dense2 = 606.824 & O/p layer = 23.704\n",
      "Total number of parameters = 43555.219\n",
      "\n",
      "\n",
      "Conv1 = 176.471, Conv2 = 3636.562\n",
      "Dense1 = 30950.390, Dense2 = 485.459 & O/p layer = 18.963\n",
      "Total number of parameters = 35267.845\n",
      "\n",
      "\n",
      "Conv1 = 158.824, Conv2 = 3272.906\n",
      "Dense1 = 24760.312, Dense2 = 388.367 & O/p layer = 15.171\n",
      "Total number of parameters = 28595.580\n",
      "\n",
      "\n",
      "Conv1 = 142.941, Conv2 = 2945.615\n",
      "Dense1 = 19808.250, Dense2 = 310.694 & O/p layer = 12.136\n",
      "Total number of parameters = 23219.637\n",
      "\n",
      "\n",
      "Conv1 = 128.647, Conv2 = 2651.054\n",
      "Dense1 = 15846.600, Dense2 = 248.555 & O/p layer = 9.709\n",
      "Total number of parameters = 18884.565\n",
      "\n"
     ]
    }
   ],
   "source": [
    "while loc_tot_params >= max_pruned_params:\n",
    "    loc_conv1 *= 0.9    # 10% weights are pruned\n",
    "    loc_conv2 *= 0.9    # 10% weights are pruned\n",
    "    loc_dense1 *= 0.8   # 20% weights are pruned\n",
    "    loc_dense2 *= 0.8   # 20% weights are pruned\n",
    "    loc_op_layer *= 0.8 # 20% weights are pruned\n",
    "    \n",
    "    conv1_pruning.append(((conv1 - loc_conv1) / conv1) * 100)\n",
    "    conv2_pruning.append(((conv2 - loc_conv2) / conv2) * 100)\n",
    "    dense1_pruning.append(((dense1 - loc_dense1) / dense1) * 100)\n",
    "    dense2_pruning.append(((dense2 - loc_dense2) / dense2) * 100)\n",
    "    op_layer_pruning.append(((op_layer - loc_op_layer) / op_layer) * 100)\n",
    "\n",
    "    loc_tot_params = loc_conv1 + loc_conv2 + loc_dense1 + loc_dense2 + loc_op_layer\n",
    "\n",
    "    n += 1\n",
    "\n",
    "    \n",
    "    print(\"\\nConv1 = {0:.3f}, Conv2 = {1:.3f}\".format(loc_conv1, loc_conv2))\n",
    "    print(\"Dense1 = {0:.3f}, Dense2 = {1:.3f} & O/p layer = {2:.3f}\".format(\n",
    "        loc_dense1, loc_dense2, loc_op_layer))\n",
    "    print(\"Total number of parameters = {0:.3f}\\n\".format(loc_tot_params))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "number of pruning rounds = 25\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"\\nnumber of pruning rounds = {0}\\n\\n\".format(n))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_pruning_rounds = n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Convert from list to np.array-\n",
    "conv1_pruning = np.array(conv1_pruning)\n",
    "conv2_pruning = np.array(conv2_pruning)\n",
    "dense1_pruning = np.array(dense1_pruning)\n",
    "dense2_pruning = np.array(dense2_pruning)\n",
    "op_layer_pruning = np.array(op_layer_pruning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Round off numpy arrays to 3 decimal digits-\n",
    "conv1_pruning = np.round(conv1_pruning, decimals=3)\n",
    "conv2_pruning = np.round(conv2_pruning, decimals=3)\n",
    "dense1_pruning = np.round(dense1_pruning, decimals=3)\n",
    "dense2_pruning = np.round(dense2_pruning, decimals=3)\n",
    "op_layer_pruning = np.round(op_layer_pruning, decimals=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([10.   , 19.   , 27.1  , 34.39 , 40.951, 46.856, 52.17 , 56.953,\n",
       "       61.258, 65.132, 68.619, 71.757, 74.581, 77.123, 79.411, 81.47 ,\n",
       "       83.323, 84.991, 86.491, 87.842, 89.058, 90.152, 91.137, 92.023,\n",
       "       92.821])"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conv1_pruning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### NOTE:\n",
    "\n",
    "For _tensorflow_model_optimization_ to prune weights, the weights have to be between 0 and 1. Therefore, the weights have to be divided by 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "conv1_pruning = conv1_pruning / 100\n",
    "conv2_pruning = conv2_pruning / 100\n",
    "dense1_pruning = dense1_pruning / 100\n",
    "dense2_pruning = dense2_pruning / 100\n",
    "op_layer_pruning = op_layer_pruning / 100\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.2    , 0.36   , 0.488  , 0.5904 , 0.67232, 0.73786, 0.79028,\n",
       "       0.83223, 0.86578, 0.89263, 0.9141 , 0.93128, 0.94502, 0.95602,\n",
       "       0.96482, 0.97185, 0.97748, 0.98199, 0.98559, 0.98847, 0.99078,\n",
       "       0.99262, 0.9941 , 0.99528, 0.99622])"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dense1_pruning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create a mask:\n",
    "A mask is created where all parameters equal to one.\n",
    "\n",
    "This will be used later by _GradientTape_ for training the defined neural network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Instantiate a new neural network model for which, the mask is to be created,\n",
    "# according to the paper-\n",
    "mask_model = pruned_nn(pruning_params_unpruned, pruning_params_unpruned)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Strip the model of its pruning parameters-\n",
    "mask_model_stripped = sparsity.strip_pruning(mask_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Assign all masks to one-\n",
    "\n",
    "for wts in mask_model_stripped.trainable_weights:\n",
    "    wts.assign(\n",
    "        tf.ones_like(\n",
    "            input = wts,\n",
    "            dtype = tf.float32\n",
    "        )\n",
    "\n",
    "    )\n",
    "    # wts.assign(1.)\n",
    "    # wts.assign(tf.where(tf.equal(wts, 0.), 0., 1.))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Mask model metrics:\n",
      "layer-wise number of nonzero parameters in each layer are: \n",
      "\n",
      "1728\n",
      "64\n",
      "36864\n",
      "64\n",
      "4194304\n",
      "256\n",
      "65536\n",
      "256\n",
      "2560\n",
      "10\n",
      "\n",
      "Total number of trainable parameters = 4301642\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"\\nMask model metrics:\")\n",
    "print(\"layer-wise number of nonzero parameters in each layer are: \\n\")\n",
    "\n",
    "masked_sum_params = 0\n",
    "\n",
    "for layer in mask_model_stripped.trainable_weights:\n",
    "    print(tf.math.count_nonzero(layer, axis = None).numpy())\n",
    "    masked_sum_params += tf.math.count_nonzero(layer, axis = None).numpy()\n",
    "\n",
    "print(\"\\nTotal number of trainable parameters = {0}\\n\".format(masked_sum_params))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "number of pruning rounds for Conv-2 CNN = 25 and number of epochs = 100\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"\\nnumber of pruning rounds for Conv-2 CNN = {0} and number of epochs = {1}\\n\".format(num_pruning_rounds, num_epochs))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Python 3 dictionary to hold model training metrics for each of _n_ rounds:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Method - 1: Nested Python 3 dictionaries:\n",
    "history_main = {}\n",
    "\n",
    "# for x in range(num_pruning_rounds + 1):\n",
    "for x in range(num_pruning_rounds):\n",
    "    history = {}\n",
    "    history['accuracy'] = np.zeros(shape = num_epochs)\n",
    "    history['val_accuracy'] = np.zeros(shape = num_epochs)\n",
    "    history['loss'] = np.zeros(shape = num_epochs)\n",
    "    history['val_loss'] = np.zeros(shape = num_epochs)\n",
    "    history['percentage_wts_pruned'] = 90\n",
    "\n",
    "    history_main[x + 1] = history\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25])"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sanity check-\n",
    "history_main.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(100,)"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sanity check-\n",
    "history_main[10]['accuracy'].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "90"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sanity check-\n",
    "history_main[19]['percentage_wts_pruned']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Total number of parameters in original network:  4301642\n"
     ]
    }
   ],
   "source": [
    "orig_sum_params = total_params\n",
    "print(\"\\nTotal number of parameters in original network: \", orig_sum_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# User input parameters for Early Stopping in manual implementation-\n",
    "minimum_delta = 0.001\n",
    "patience = 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "best_val_loss = 100\n",
    "loc_patience = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 1\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.4986, Accuracy: 46.0360, Test Loss: 1.2745, Test Accuracy: 54.410000\n",
      "Total number of trainable parameters = 4301634\n",
      "\n",
      "Epoch 2, Loss: 1.1331, Accuracy: 59.8000, Test Loss: 1.0784, Test Accuracy: 61.549995\n",
      "Total number of trainable parameters = 4301636\n",
      "\n",
      "Epoch 3, Loss: 0.9582, Accuracy: 66.4500, Test Loss: 0.9968, Test Accuracy: 65.520004\n",
      "Total number of trainable parameters = 4301638\n",
      "\n",
      "Epoch 4, Loss: 0.8405, Accuracy: 70.6820, Test Loss: 0.9206, Test Accuracy: 67.849998\n",
      "Total number of trainable parameters = 4301638\n",
      "\n",
      "Epoch 5, Loss: 0.7453, Accuracy: 74.0280, Test Loss: 0.9052, Test Accuracy: 68.139999\n",
      "Total number of trainable parameters = 4301639\n",
      "\n",
      "Epoch 6, Loss: 0.6573, Accuracy: 77.3400, Test Loss: 0.8949, Test Accuracy: 69.330002\n",
      "Total number of trainable parameters = 4301640\n",
      "\n",
      "Epoch 7, Loss: 0.5691, Accuracy: 80.4040, Test Loss: 0.9186, Test Accuracy: 69.279999\n",
      "Total number of trainable parameters = 4301640\n",
      "\n",
      "Epoch 8, Loss: 0.4787, Accuracy: 83.5600, Test Loss: 0.9423, Test Accuracy: 69.830002\n",
      "Total number of trainable parameters = 4301640\n",
      "\n",
      "Epoch 9, Loss: 0.3909, Accuracy: 86.8440, Test Loss: 0.9571, Test Accuracy: 70.190002\n",
      "Total number of trainable parameters = 4301640\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 120s 2ms/sample - loss: 0.3098 - accuracy: 0.8975 - val_loss: 1.0757 - val_accuracy: 0.6908\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 123s 2ms/sample - loss: 0.2131 - accuracy: 0.9336 - val_loss: 1.1341 - val_accuracy: 0.6965\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 125s 2ms/sample - loss: 0.1485 - accuracy: 0.9552 - val_loss: 1.2722 - val_accuracy: 0.6909\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 124s 2ms/sample - loss: 0.1054 - accuracy: 0.9699 - val_loss: 1.3773 - val_accuracy: 0.6892\n",
      "\n",
      "Round = 1, total number of trainable parameters = 3445302\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 2\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.3177, Accuracy: 52.4380, Test Loss: 1.1160, Test Accuracy: 60.639999\n",
      "Total number of trainable parameters = 3445277\n",
      "\n",
      "Epoch 2, Loss: 0.9534, Accuracy: 66.6840, Test Loss: 0.9550, Test Accuracy: 66.189995\n",
      "Total number of trainable parameters = 3445282\n",
      "\n",
      "Epoch 3, Loss: 0.8235, Accuracy: 71.1840, Test Loss: 0.9100, Test Accuracy: 68.400002\n",
      "Total number of trainable parameters = 3445285\n",
      "\n",
      "Epoch 4, Loss: 0.7251, Accuracy: 74.7720, Test Loss: 0.9274, Test Accuracy: 68.010002\n",
      "Total number of trainable parameters = 3445287\n",
      "\n",
      "Epoch 5, Loss: 0.6368, Accuracy: 78.0080, Test Loss: 0.8838, Test Accuracy: 69.959999\n",
      "Total number of trainable parameters = 3445289\n",
      "\n",
      "Epoch 6, Loss: 0.5583, Accuracy: 80.6220, Test Loss: 0.8916, Test Accuracy: 70.470001\n",
      "Total number of trainable parameters = 3445291\n",
      "\n",
      "Epoch 7, Loss: 0.4718, Accuracy: 83.9940, Test Loss: 0.9596, Test Accuracy: 69.340004\n",
      "Total number of trainable parameters = 3445291\n",
      "\n",
      "Epoch 8, Loss: 0.3893, Accuracy: 86.9600, Test Loss: 0.9600, Test Accuracy: 70.599998\n",
      "Total number of trainable parameters = 3445293\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 120s 2ms/sample - loss: 0.3138 - accuracy: 0.8980 - val_loss: 1.0090 - val_accuracy: 0.7011\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 125s 2ms/sample - loss: 0.2316 - accuracy: 0.9277 - val_loss: 1.1012 - val_accuracy: 0.7018\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 127s 3ms/sample - loss: 0.1736 - accuracy: 0.9475 - val_loss: 1.1992 - val_accuracy: 0.6997\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 102s 2ms/sample - loss: 0.1256 - accuracy: 0.9647 - val_loss: 1.3424 - val_accuracy: 0.6946\n",
      "\n",
      "Round = 2, total number of trainable parameters = 2759843\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 3\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.2697, Accuracy: 54.7140, Test Loss: 1.0475, Test Accuracy: 63.410004\n",
      "Total number of trainable parameters = 2759722\n",
      "\n",
      "Epoch 2, Loss: 0.9345, Accuracy: 67.1820, Test Loss: 0.9604, Test Accuracy: 66.849998\n",
      "Total number of trainable parameters = 2759737\n",
      "\n",
      "Epoch 3, Loss: 0.8210, Accuracy: 71.4840, Test Loss: 0.9034, Test Accuracy: 68.269997\n",
      "Total number of trainable parameters = 2759738\n",
      "\n",
      "Epoch 4, Loss: 0.7408, Accuracy: 74.2800, Test Loss: 0.9005, Test Accuracy: 69.099998\n",
      "Total number of trainable parameters = 2759744\n",
      "\n",
      "Epoch 5, Loss: 0.6710, Accuracy: 76.9300, Test Loss: 0.8875, Test Accuracy: 69.360001\n",
      "Total number of trainable parameters = 2759746\n",
      "\n",
      "Epoch 6, Loss: 0.6028, Accuracy: 79.2800, Test Loss: 0.9148, Test Accuracy: 69.239998\n",
      "Total number of trainable parameters = 2759747\n",
      "\n",
      "Epoch 7, Loss: 0.5400, Accuracy: 81.4420, Test Loss: 0.9199, Test Accuracy: 69.880005\n",
      "Total number of trainable parameters = 2759751\n",
      "\n",
      "Epoch 8, Loss: 0.4745, Accuracy: 83.7500, Test Loss: 0.9387, Test Accuracy: 69.700005\n",
      "Total number of trainable parameters = 2759755\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 74s 1ms/sample - loss: 0.4116 - accuracy: 0.8633 - val_loss: 0.9634 - val_accuracy: 0.7064\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.3524 - accuracy: 0.8843 - val_loss: 0.9932 - val_accuracy: 0.7043\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.2927 - accuracy: 0.9068 - val_loss: 1.0797 - val_accuracy: 0.7000\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.2437 - accuracy: 0.9232 - val_loss: 1.1450 - val_accuracy: 0.6958\n",
      "\n",
      "Round = 3, total number of trainable parameters = 2211100\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 4\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.2105, Accuracy: 57.0060, Test Loss: 1.0131, Test Accuracy: 64.709999\n",
      "Total number of trainable parameters = 2210985\n",
      "\n",
      "Epoch 2, Loss: 0.9019, Accuracy: 68.6920, Test Loss: 0.9399, Test Accuracy: 67.079994\n",
      "Total number of trainable parameters = 2211003\n",
      "\n",
      "Epoch 3, Loss: 0.7932, Accuracy: 72.6540, Test Loss: 0.9032, Test Accuracy: 68.709999\n",
      "Total number of trainable parameters = 2211012\n",
      "\n",
      "Epoch 4, Loss: 0.7072, Accuracy: 75.4840, Test Loss: 0.9049, Test Accuracy: 69.019997\n",
      "Total number of trainable parameters = 2211018\n",
      "\n",
      "Epoch 5, Loss: 0.6361, Accuracy: 78.0000, Test Loss: 0.9112, Test Accuracy: 69.480003\n",
      "Total number of trainable parameters = 2211020\n",
      "\n",
      "Epoch 6, Loss: 0.5678, Accuracy: 80.4860, Test Loss: 0.8997, Test Accuracy: 69.599998\n",
      "Total number of trainable parameters = 2211024\n",
      "\n",
      "Epoch 7, Loss: 0.5010, Accuracy: 82.9320, Test Loss: 0.9339, Test Accuracy: 70.300003\n",
      "Total number of trainable parameters = 2211024\n",
      "\n",
      "Epoch 8, Loss: 0.4396, Accuracy: 85.0700, Test Loss: 0.9435, Test Accuracy: 69.330002\n",
      "Total number of trainable parameters = 2211027\n",
      "\n",
      "Epoch 9, Loss: 0.3769, Accuracy: 87.5720, Test Loss: 0.9847, Test Accuracy: 69.980003\n",
      "Total number of trainable parameters = 2211027\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.3179 - accuracy: 0.8962 - val_loss: 1.0552 - val_accuracy: 0.6935\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.2643 - accuracy: 0.9153 - val_loss: 1.0723 - val_accuracy: 0.7033\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.2089 - accuracy: 0.9372 - val_loss: 1.1976 - val_accuracy: 0.6934\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.1660 - accuracy: 0.9526 - val_loss: 1.2706 - val_accuracy: 0.6924\n",
      "\n",
      "Round = 4, total number of trainable parameters = 1771816\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 5\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.1834, Accuracy: 58.1840, Test Loss: 1.0167, Test Accuracy: 64.780006\n",
      "Total number of trainable parameters = 1771734\n",
      "\n",
      "Epoch 2, Loss: 0.8790, Accuracy: 69.3280, Test Loss: 0.9319, Test Accuracy: 67.540001\n",
      "Total number of trainable parameters = 1771746\n",
      "\n",
      "Epoch 3, Loss: 0.7632, Accuracy: 73.6920, Test Loss: 0.9008, Test Accuracy: 69.169998\n",
      "Total number of trainable parameters = 1771750\n",
      "\n",
      "Epoch 4, Loss: 0.6756, Accuracy: 76.7380, Test Loss: 0.9070, Test Accuracy: 68.570000\n",
      "Total number of trainable parameters = 1771753\n",
      "\n",
      "Epoch 5, Loss: 0.5975, Accuracy: 79.5200, Test Loss: 0.8770, Test Accuracy: 70.370003\n",
      "Total number of trainable parameters = 1771756\n",
      "\n",
      "Epoch 6, Loss: 0.5256, Accuracy: 82.1920, Test Loss: 0.8937, Test Accuracy: 70.120003\n",
      "Total number of trainable parameters = 1771757\n",
      "\n",
      "Epoch 7, Loss: 0.4553, Accuracy: 84.8680, Test Loss: 0.9364, Test Accuracy: 70.220001\n",
      "Total number of trainable parameters = 1771758\n",
      "\n",
      "Epoch 8, Loss: 0.3909, Accuracy: 87.0140, Test Loss: 1.0131, Test Accuracy: 69.110001\n",
      "Total number of trainable parameters = 1771761\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 75s 2ms/sample - loss: 0.3278 - accuracy: 0.8947 - val_loss: 1.0457 - val_accuracy: 0.6964\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.2790 - accuracy: 0.9133 - val_loss: 1.0580 - val_accuracy: 0.6983\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.2225 - accuracy: 0.9343 - val_loss: 1.1414 - val_accuracy: 0.6995\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.1788 - accuracy: 0.9474 - val_loss: 1.2385 - val_accuracy: 0.6903\n",
      "\n",
      "Round = 5, total number of trainable parameters = 1420104\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 6\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.1635, Accuracy: 58.7840, Test Loss: 1.0047, Test Accuracy: 65.090004\n",
      "Total number of trainable parameters = 1420025\n",
      "\n",
      "Epoch 2, Loss: 0.8508, Accuracy: 70.4780, Test Loss: 0.9370, Test Accuracy: 67.470001\n",
      "Total number of trainable parameters = 1420029\n",
      "\n",
      "Epoch 3, Loss: 0.7345, Accuracy: 74.6560, Test Loss: 0.9111, Test Accuracy: 68.330002\n",
      "Total number of trainable parameters = 1420031\n",
      "\n",
      "Epoch 4, Loss: 0.6405, Accuracy: 77.9540, Test Loss: 0.9137, Test Accuracy: 68.879997\n",
      "Total number of trainable parameters = 1420032\n",
      "\n",
      "Epoch 5, Loss: 0.5621, Accuracy: 80.7320, Test Loss: 0.8953, Test Accuracy: 69.840004\n",
      "Total number of trainable parameters = 1420032\n",
      "\n",
      "Epoch 6, Loss: 0.4882, Accuracy: 83.3840, Test Loss: 0.9063, Test Accuracy: 70.500000\n",
      "Total number of trainable parameters = 1420033\n",
      "\n",
      "Epoch 7, Loss: 0.4225, Accuracy: 85.9080, Test Loss: 0.9619, Test Accuracy: 69.889999\n",
      "Total number of trainable parameters = 1420034\n",
      "\n",
      "Epoch 8, Loss: 0.3558, Accuracy: 88.3440, Test Loss: 1.0076, Test Accuracy: 69.270004\n",
      "Total number of trainable parameters = 1420034\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.3011 - accuracy: 0.9034 - val_loss: 1.0618 - val_accuracy: 0.6980\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.2611 - accuracy: 0.9205 - val_loss: 1.0625 - val_accuracy: 0.7032\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.2076 - accuracy: 0.9394 - val_loss: 1.1423 - val_accuracy: 0.6970\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.1662 - accuracy: 0.9541 - val_loss: 1.2353 - val_accuracy: 0.6895\n",
      "\n",
      "Round = 6, total number of trainable parameters = 1138444\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 7\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.1564, Accuracy: 59.3440, Test Loss: 1.0063, Test Accuracy: 65.200005\n",
      "Total number of trainable parameters = 1138362\n",
      "\n",
      "Epoch 2, Loss: 0.8493, Accuracy: 70.6060, Test Loss: 0.9302, Test Accuracy: 67.860001\n",
      "Total number of trainable parameters = 1138366\n",
      "\n",
      "Epoch 3, Loss: 0.7284, Accuracy: 74.7720, Test Loss: 0.8915, Test Accuracy: 69.309998\n",
      "Total number of trainable parameters = 1138369\n",
      "\n",
      "Epoch 4, Loss: 0.6361, Accuracy: 78.1460, Test Loss: 0.8856, Test Accuracy: 69.300003\n",
      "Total number of trainable parameters = 1138372\n",
      "\n",
      "Epoch 5, Loss: 0.5572, Accuracy: 81.0460, Test Loss: 0.9011, Test Accuracy: 69.950005\n",
      "Total number of trainable parameters = 1138376\n",
      "\n",
      "Epoch 6, Loss: 0.4866, Accuracy: 83.5720, Test Loss: 0.9013, Test Accuracy: 70.459999\n",
      "Total number of trainable parameters = 1138378\n",
      "\n",
      "Epoch 7, Loss: 0.4215, Accuracy: 86.1460, Test Loss: 0.9495, Test Accuracy: 70.380005\n",
      "Total number of trainable parameters = 1138380\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 75s 2ms/sample - loss: 0.3606 - accuracy: 0.8833 - val_loss: 0.9717 - val_accuracy: 0.6998\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.3413 - accuracy: 0.8935 - val_loss: 1.0009 - val_accuracy: 0.6985\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.2816 - accuracy: 0.9129 - val_loss: 1.0615 - val_accuracy: 0.6937\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.2358 - accuracy: 0.9293 - val_loss: 1.0877 - val_accuracy: 0.6991\n",
      "\n",
      "Round = 7, total number of trainable parameters = 912931\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 8\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.1638, Accuracy: 59.0560, Test Loss: 0.9941, Test Accuracy: 65.779999\n",
      "Total number of trainable parameters = 912873\n",
      "\n",
      "Epoch 2, Loss: 0.8490, Accuracy: 70.6400, Test Loss: 0.9376, Test Accuracy: 67.470001\n",
      "Total number of trainable parameters = 912878\n",
      "\n",
      "Epoch 3, Loss: 0.7229, Accuracy: 75.0920, Test Loss: 0.9004, Test Accuracy: 69.029999\n",
      "Total number of trainable parameters = 912880\n",
      "\n",
      "Epoch 4, Loss: 0.6297, Accuracy: 78.5580, Test Loss: 0.8858, Test Accuracy: 69.610001\n",
      "Total number of trainable parameters = 912881\n",
      "\n",
      "Epoch 5, Loss: 0.5540, Accuracy: 81.1440, Test Loss: 0.8949, Test Accuracy: 69.459999\n",
      "Total number of trainable parameters = 912884\n",
      "\n",
      "Epoch 6, Loss: 0.4870, Accuracy: 83.7700, Test Loss: 0.9145, Test Accuracy: 69.700005\n",
      "Total number of trainable parameters = 912886\n",
      "\n",
      "Epoch 7, Loss: 0.4233, Accuracy: 85.9100, Test Loss: 0.9394, Test Accuracy: 70.169998\n",
      "Total number of trainable parameters = 912887\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 75s 2ms/sample - loss: 0.3716 - accuracy: 0.8781 - val_loss: 0.9937 - val_accuracy: 0.6955\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.3597 - accuracy: 0.8866 - val_loss: 0.9741 - val_accuracy: 0.6996\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.3016 - accuracy: 0.9056 - val_loss: 1.0605 - val_accuracy: 0.6905\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.2589 - accuracy: 0.9212 - val_loss: 1.0763 - val_accuracy: 0.6972\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.2205 - accuracy: 0.9345 - val_loss: 1.1493 - val_accuracy: 0.6940\n",
      "\n",
      "Round = 8, total number of trainable parameters = 732265\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 9\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.1920, Accuracy: 58.3280, Test Loss: 1.0302, Test Accuracy: 64.080002\n",
      "Total number of trainable parameters = 732218\n",
      "\n",
      "Epoch 2, Loss: 0.8707, Accuracy: 70.1000, Test Loss: 0.9438, Test Accuracy: 66.860001\n",
      "Total number of trainable parameters = 732221\n",
      "\n",
      "Epoch 3, Loss: 0.7420, Accuracy: 74.3560, Test Loss: 0.9033, Test Accuracy: 68.750000\n",
      "Total number of trainable parameters = 732222\n",
      "\n",
      "Epoch 4, Loss: 0.6496, Accuracy: 77.7500, Test Loss: 0.9018, Test Accuracy: 68.839996\n",
      "Total number of trainable parameters = 732224\n",
      "\n",
      "Epoch 5, Loss: 0.5749, Accuracy: 80.3680, Test Loss: 0.9045, Test Accuracy: 69.400002\n",
      "Total number of trainable parameters = 732224\n",
      "\n",
      "Epoch 6, Loss: 0.5095, Accuracy: 83.0120, Test Loss: 0.9209, Test Accuracy: 69.760002\n",
      "Total number of trainable parameters = 732226\n",
      "\n",
      "Epoch 7, Loss: 0.4499, Accuracy: 85.0600, Test Loss: 0.9352, Test Accuracy: 69.809998\n",
      "Total number of trainable parameters = 732227\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.3984 - accuracy: 0.8679 - val_loss: 0.9579 - val_accuracy: 0.7015\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.3922 - accuracy: 0.8759 - val_loss: 0.9677 - val_accuracy: 0.6936\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.3351 - accuracy: 0.8945 - val_loss: 1.0199 - val_accuracy: 0.6971\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.2936 - accuracy: 0.9084 - val_loss: 1.0547 - val_accuracy: 0.6965\n",
      "\n",
      "Round = 9, total number of trainable parameters = 587589\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 10\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.2088, Accuracy: 57.7160, Test Loss: 1.0390, Test Accuracy: 63.849998\n",
      "Total number of trainable parameters = 587538\n",
      "\n",
      "Epoch 2, Loss: 0.8840, Accuracy: 69.5460, Test Loss: 0.9658, Test Accuracy: 66.239998\n",
      "Total number of trainable parameters = 587539\n",
      "\n",
      "Epoch 3, Loss: 0.7627, Accuracy: 73.9480, Test Loss: 0.9135, Test Accuracy: 68.370003\n",
      "Total number of trainable parameters = 587541\n",
      "\n",
      "Epoch 4, Loss: 0.6711, Accuracy: 77.1380, Test Loss: 0.9072, Test Accuracy: 68.940002\n",
      "Total number of trainable parameters = 587541\n",
      "\n",
      "Epoch 5, Loss: 0.5974, Accuracy: 79.8740, Test Loss: 0.9117, Test Accuracy: 68.870003\n",
      "Total number of trainable parameters = 587542\n",
      "\n",
      "Epoch 6, Loss: 0.5336, Accuracy: 82.0840, Test Loss: 0.9141, Test Accuracy: 69.500000\n",
      "Total number of trainable parameters = 587543\n",
      "\n",
      "Epoch 7, Loss: 0.4793, Accuracy: 83.9240, Test Loss: 0.9516, Test Accuracy: 69.320000\n",
      "Total number of trainable parameters = 587544\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 75s 1ms/sample - loss: 0.4282 - accuracy: 0.8588 - val_loss: 0.9654 - val_accuracy: 0.6978\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.4445 - accuracy: 0.8599 - val_loss: 0.9526 - val_accuracy: 0.6958\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.3776 - accuracy: 0.8804 - val_loss: 0.9966 - val_accuracy: 0.6950\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.3386 - accuracy: 0.8933 - val_loss: 1.0404 - val_accuracy: 0.6919\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.3055 - accuracy: 0.9038 - val_loss: 1.0861 - val_accuracy: 0.6936\n",
      "\n",
      "Round = 10, total number of trainable parameters = 471624\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 11\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.2337, Accuracy: 57.0020, Test Loss: 1.0631, Test Accuracy: 62.669998\n",
      "Total number of trainable parameters = 471568\n",
      "\n",
      "Epoch 2, Loss: 0.9187, Accuracy: 68.2960, Test Loss: 0.9945, Test Accuracy: 65.190002\n",
      "Total number of trainable parameters = 471571\n",
      "\n",
      "Epoch 3, Loss: 0.7955, Accuracy: 72.6240, Test Loss: 0.9415, Test Accuracy: 67.479996\n",
      "Total number of trainable parameters = 471573\n",
      "\n",
      "Epoch 4, Loss: 0.7084, Accuracy: 75.8260, Test Loss: 0.9097, Test Accuracy: 68.699997\n",
      "Total number of trainable parameters = 471573\n",
      "\n",
      "Epoch 5, Loss: 0.6340, Accuracy: 78.4360, Test Loss: 0.9291, Test Accuracy: 68.110001\n",
      "Total number of trainable parameters = 471575\n",
      "\n",
      "Epoch 6, Loss: 0.5729, Accuracy: 80.5900, Test Loss: 0.9171, Test Accuracy: 69.090004\n",
      "Total number of trainable parameters = 471576\n",
      "\n",
      "Epoch 7, Loss: 0.5196, Accuracy: 82.7180, Test Loss: 0.9299, Test Accuracy: 69.349998\n",
      "Total number of trainable parameters = 471576\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.4712 - accuracy: 0.8436 - val_loss: 0.9519 - val_accuracy: 0.6938\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.4875 - accuracy: 0.8422 - val_loss: 0.9520 - val_accuracy: 0.6926\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.4246 - accuracy: 0.8636 - val_loss: 0.9756 - val_accuracy: 0.6941\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.3896 - accuracy: 0.8749 - val_loss: 1.0121 - val_accuracy: 0.6898\n",
      "\n",
      "Round = 11, total number of trainable parameters = 378739\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 12\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.2730, Accuracy: 55.4300, Test Loss: 1.1001, Test Accuracy: 61.059998\n",
      "Total number of trainable parameters = 378684\n",
      "\n",
      "Epoch 2, Loss: 0.9504, Accuracy: 67.1060, Test Loss: 0.9915, Test Accuracy: 65.250000\n",
      "Total number of trainable parameters = 378686\n",
      "\n",
      "Epoch 3, Loss: 0.8285, Accuracy: 71.5700, Test Loss: 0.9482, Test Accuracy: 66.849998\n",
      "Total number of trainable parameters = 378690\n",
      "\n",
      "Epoch 4, Loss: 0.7446, Accuracy: 74.2940, Test Loss: 0.9313, Test Accuracy: 67.510002\n",
      "Total number of trainable parameters = 378692\n",
      "\n",
      "Epoch 5, Loss: 0.6749, Accuracy: 76.8100, Test Loss: 0.9164, Test Accuracy: 68.559998\n",
      "Total number of trainable parameters = 378692\n",
      "\n",
      "Epoch 6, Loss: 0.6140, Accuracy: 79.1220, Test Loss: 0.9161, Test Accuracy: 68.830002\n",
      "Total number of trainable parameters = 378693\n",
      "\n",
      "Epoch 7, Loss: 0.5648, Accuracy: 80.8220, Test Loss: 0.9198, Test Accuracy: 69.160004\n",
      "Total number of trainable parameters = 378694\n",
      "\n",
      "Epoch 8, Loss: 0.5191, Accuracy: 82.6120, Test Loss: 0.9358, Test Accuracy: 69.000000\n",
      "Total number of trainable parameters = 378696\n",
      "\n",
      "Epoch 9, Loss: 0.4790, Accuracy: 84.0120, Test Loss: 0.9629, Test Accuracy: 68.910004\n",
      "Total number of trainable parameters = 378696\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 75s 2ms/sample - loss: 0.4414 - accuracy: 0.8525 - val_loss: 0.9879 - val_accuracy: 0.6887\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.4579 - accuracy: 0.8511 - val_loss: 0.9553 - val_accuracy: 0.6897\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.4071 - accuracy: 0.8686 - val_loss: 1.0079 - val_accuracy: 0.6883\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.3743 - accuracy: 0.8791 - val_loss: 1.0139 - val_accuracy: 0.6925\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.3459 - accuracy: 0.8883 - val_loss: 1.0516 - val_accuracy: 0.6886\n",
      "\n",
      "Round = 12, total number of trainable parameters = 304283\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 13\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.3364, Accuracy: 53.5760, Test Loss: 1.1325, Test Accuracy: 60.600002\n",
      "Total number of trainable parameters = 304233\n",
      "\n",
      "Epoch 2, Loss: 1.0120, Accuracy: 65.0480, Test Loss: 1.0246, Test Accuracy: 64.029999\n",
      "Total number of trainable parameters = 304235\n",
      "\n",
      "Epoch 3, Loss: 0.8960, Accuracy: 69.0720, Test Loss: 0.9896, Test Accuracy: 65.300003\n",
      "Total number of trainable parameters = 304238\n",
      "\n",
      "Epoch 4, Loss: 0.8170, Accuracy: 71.8180, Test Loss: 0.9628, Test Accuracy: 66.689995\n",
      "Total number of trainable parameters = 304239\n",
      "\n",
      "Epoch 5, Loss: 0.7517, Accuracy: 74.2020, Test Loss: 0.9428, Test Accuracy: 67.229996\n",
      "Total number of trainable parameters = 304239\n",
      "\n",
      "Epoch 6, Loss: 0.6943, Accuracy: 76.2740, Test Loss: 0.9427, Test Accuracy: 68.129997\n",
      "Total number of trainable parameters = 304240\n",
      "\n",
      "Epoch 7, Loss: 0.6447, Accuracy: 78.0360, Test Loss: 0.9560, Test Accuracy: 67.619995\n",
      "Total number of trainable parameters = 304240\n",
      "\n",
      "Epoch 8, Loss: 0.5992, Accuracy: 79.6600, Test Loss: 0.9387, Test Accuracy: 68.529999\n",
      "Total number of trainable parameters = 304241\n",
      "\n",
      "Epoch 9, Loss: 0.5585, Accuracy: 81.1980, Test Loss: 0.9343, Test Accuracy: 69.120003\n",
      "Total number of trainable parameters = 304241\n",
      "\n",
      "Epoch 10, Loss: 0.5205, Accuracy: 82.4900, Test Loss: 0.9614, Test Accuracy: 68.559998\n",
      "Total number of trainable parameters = 304241\n",
      "\n",
      "Epoch 11, Loss: 0.4856, Accuracy: 83.9340, Test Loss: 0.9833, Test Accuracy: 68.970001\n",
      "Total number of trainable parameters = 304241\n",
      "\n",
      "Epoch 12, Loss: 0.4557, Accuracy: 84.8720, Test Loss: 0.9915, Test Accuracy: 68.839996\n",
      "Total number of trainable parameters = 304241\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.4247 - accuracy: 0.8602 - val_loss: 1.0050 - val_accuracy: 0.6889\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.4553 - accuracy: 0.8528 - val_loss: 0.9887 - val_accuracy: 0.6821\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.4037 - accuracy: 0.8713 - val_loss: 1.0379 - val_accuracy: 0.6836\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.3756 - accuracy: 0.8799 - val_loss: 1.0369 - val_accuracy: 0.6878\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.3504 - accuracy: 0.8880 - val_loss: 1.0621 - val_accuracy: 0.6882\n",
      "\n",
      "Round = 13, total number of trainable parameters = 244621\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 14\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.3955, Accuracy: 51.4200, Test Loss: 1.1818, Test Accuracy: 58.270000\n",
      "Total number of trainable parameters = 244558\n",
      "\n",
      "Epoch 2, Loss: 1.0618, Accuracy: 63.2680, Test Loss: 1.0725, Test Accuracy: 62.919998\n",
      "Total number of trainable parameters = 244561\n",
      "\n",
      "Epoch 3, Loss: 0.9481, Accuracy: 67.1980, Test Loss: 1.0187, Test Accuracy: 64.510002\n",
      "Total number of trainable parameters = 244565\n",
      "\n",
      "Epoch 4, Loss: 0.8691, Accuracy: 69.9640, Test Loss: 0.9814, Test Accuracy: 65.900002\n",
      "Total number of trainable parameters = 244567\n",
      "\n",
      "Epoch 5, Loss: 0.8087, Accuracy: 72.0140, Test Loss: 0.9584, Test Accuracy: 66.720001\n",
      "Total number of trainable parameters = 244567\n",
      "\n",
      "Epoch 6, Loss: 0.7511, Accuracy: 74.3080, Test Loss: 0.9438, Test Accuracy: 67.279999\n",
      "Total number of trainable parameters = 244568\n",
      "\n",
      "Epoch 7, Loss: 0.7035, Accuracy: 75.8200, Test Loss: 0.9321, Test Accuracy: 68.010002\n",
      "Total number of trainable parameters = 244568\n",
      "\n",
      "Epoch 8, Loss: 0.6582, Accuracy: 77.6320, Test Loss: 0.9277, Test Accuracy: 68.300003\n",
      "Total number of trainable parameters = 244570\n",
      "\n",
      "Epoch 9, Loss: 0.6174, Accuracy: 79.0360, Test Loss: 0.9403, Test Accuracy: 68.349998\n",
      "Total number of trainable parameters = 244573\n",
      "\n",
      "Epoch 10, Loss: 0.5805, Accuracy: 80.4040, Test Loss: 0.9421, Test Accuracy: 68.989998\n",
      "Total number of trainable parameters = 244576\n",
      "\n",
      "Epoch 11, Loss: 0.5467, Accuracy: 81.6780, Test Loss: 0.9602, Test Accuracy: 68.980003\n",
      "Total number of trainable parameters = 244577\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.5153 - accuracy: 0.8266 - val_loss: 0.9554 - val_accuracy: 0.6895\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.6540 - accuracy: 0.7873 - val_loss: 0.9431 - val_accuracy: 0.6878\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.5325 - accuracy: 0.8212 - val_loss: 0.9585 - val_accuracy: 0.6853\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.4957 - accuracy: 0.8349 - val_loss: 0.9923 - val_accuracy: 0.6816\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.4672 - accuracy: 0.8465 - val_loss: 1.0043 - val_accuracy: 0.6842\n",
      "\n",
      "Round = 14, total number of trainable parameters = 196748\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 15\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.4612, Accuracy: 48.8520, Test Loss: 1.2210, Test Accuracy: 57.389999\n",
      "Total number of trainable parameters = 196672\n",
      "\n",
      "Epoch 2, Loss: 1.1151, Accuracy: 61.4360, Test Loss: 1.1088, Test Accuracy: 61.519997\n",
      "Total number of trainable parameters = 196674\n",
      "\n",
      "Epoch 3, Loss: 0.9993, Accuracy: 65.4780, Test Loss: 1.0546, Test Accuracy: 63.160004\n",
      "Total number of trainable parameters = 196678\n",
      "\n",
      "Epoch 4, Loss: 0.9219, Accuracy: 68.2660, Test Loss: 1.0272, Test Accuracy: 63.900002\n",
      "Total number of trainable parameters = 196678\n",
      "\n",
      "Epoch 5, Loss: 0.8642, Accuracy: 70.3280, Test Loss: 0.9928, Test Accuracy: 65.440002\n",
      "Total number of trainable parameters = 196679\n",
      "\n",
      "Epoch 6, Loss: 0.8162, Accuracy: 71.9300, Test Loss: 0.9661, Test Accuracy: 66.839996\n",
      "Total number of trainable parameters = 196680\n",
      "\n",
      "Epoch 7, Loss: 0.7736, Accuracy: 73.4940, Test Loss: 0.9637, Test Accuracy: 66.529999\n",
      "Total number of trainable parameters = 196682\n",
      "\n",
      "Epoch 8, Loss: 0.7361, Accuracy: 74.6540, Test Loss: 0.9604, Test Accuracy: 67.009995\n",
      "Total number of trainable parameters = 196683\n",
      "\n",
      "Epoch 9, Loss: 0.7019, Accuracy: 76.1380, Test Loss: 0.9445, Test Accuracy: 67.949997\n",
      "Total number of trainable parameters = 196683\n",
      "\n",
      "Epoch 10, Loss: 0.6688, Accuracy: 77.1960, Test Loss: 0.9438, Test Accuracy: 68.339996\n",
      "Total number of trainable parameters = 196685\n",
      "\n",
      "Epoch 11, Loss: 0.6385, Accuracy: 78.1100, Test Loss: 0.9665, Test Accuracy: 67.409996\n",
      "Total number of trainable parameters = 196689\n",
      "\n",
      "Epoch 12, Loss: 0.6117, Accuracy: 79.2320, Test Loss: 0.9632, Test Accuracy: 67.750000\n",
      "Total number of trainable parameters = 196690\n",
      "\n",
      "Epoch 13, Loss: 0.5825, Accuracy: 80.2880, Test Loss: 0.9609, Test Accuracy: 68.360001\n",
      "Total number of trainable parameters = 196691\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.5571 - accuracy: 0.8117 - val_loss: 0.9795 - val_accuracy: 0.6792\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.6031 - accuracy: 0.8012 - val_loss: 0.9849 - val_accuracy: 0.6729\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.5494 - accuracy: 0.8172 - val_loss: 0.9777 - val_accuracy: 0.6772\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.5205 - accuracy: 0.8274 - val_loss: 0.9835 - val_accuracy: 0.6821\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 80s 2ms/sample - loss: 0.4975 - accuracy: 0.8348 - val_loss: 1.0046 - val_accuracy: 0.6795\n",
      "Epoch 6/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.4775 - accuracy: 0.8423 - val_loss: 1.0109 - val_accuracy: 0.6826\n",
      "\n",
      "Round = 15, total number of trainable parameters = 158345\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 16\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.5498, Accuracy: 45.5400, Test Loss: 1.3145, Test Accuracy: 54.160000\n",
      "Total number of trainable parameters = 158279\n",
      "\n",
      "Epoch 2, Loss: 1.1984, Accuracy: 58.3520, Test Loss: 1.1621, Test Accuracy: 59.590000\n",
      "Total number of trainable parameters = 158286\n",
      "\n",
      "Epoch 3, Loss: 1.0735, Accuracy: 62.7940, Test Loss: 1.0955, Test Accuracy: 61.580002\n",
      "Total number of trainable parameters = 158291\n",
      "\n",
      "Epoch 4, Loss: 0.9956, Accuracy: 65.5700, Test Loss: 1.0530, Test Accuracy: 63.489998\n",
      "Total number of trainable parameters = 158293\n",
      "\n",
      "Epoch 5, Loss: 0.9391, Accuracy: 67.4060, Test Loss: 1.0264, Test Accuracy: 64.120003\n",
      "Total number of trainable parameters = 158294\n",
      "\n",
      "Epoch 6, Loss: 0.8913, Accuracy: 69.1240, Test Loss: 1.0084, Test Accuracy: 64.980003\n",
      "Total number of trainable parameters = 158294\n",
      "\n",
      "Epoch 7, Loss: 0.8497, Accuracy: 70.7460, Test Loss: 0.9912, Test Accuracy: 65.650002\n",
      "Total number of trainable parameters = 158294\n",
      "\n",
      "Epoch 8, Loss: 0.8136, Accuracy: 71.9640, Test Loss: 0.9761, Test Accuracy: 66.159996\n",
      "Total number of trainable parameters = 158294\n",
      "\n",
      "Epoch 9, Loss: 0.7801, Accuracy: 73.0800, Test Loss: 0.9766, Test Accuracy: 66.229996\n",
      "Total number of trainable parameters = 158295\n",
      "\n",
      "Epoch 10, Loss: 0.7497, Accuracy: 74.2300, Test Loss: 0.9896, Test Accuracy: 66.089996\n",
      "Total number of trainable parameters = 158295\n",
      "\n",
      "Epoch 11, Loss: 0.7210, Accuracy: 75.2280, Test Loss: 0.9680, Test Accuracy: 67.099998\n",
      "Total number of trainable parameters = 158296\n",
      "\n",
      "Epoch 12, Loss: 0.6958, Accuracy: 76.1640, Test Loss: 0.9641, Test Accuracy: 67.250000\n",
      "Total number of trainable parameters = 158297\n",
      "\n",
      "Epoch 13, Loss: 0.6706, Accuracy: 77.0620, Test Loss: 0.9609, Test Accuracy: 67.659996\n",
      "Total number of trainable parameters = 158299\n",
      "\n",
      "Epoch 14, Loss: 0.6459, Accuracy: 78.0180, Test Loss: 0.9629, Test Accuracy: 67.769997\n",
      "Total number of trainable parameters = 158299\n",
      "\n",
      "Epoch 15, Loss: 0.6230, Accuracy: 78.7780, Test Loss: 0.9789, Test Accuracy: 67.669998\n",
      "Total number of trainable parameters = 158299\n",
      "\n",
      "Epoch 16, Loss: 0.6010, Accuracy: 79.5740, Test Loss: 0.9807, Test Accuracy: 67.690002\n",
      "Total number of trainable parameters = 158300\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.5803 - accuracy: 0.8038 - val_loss: 1.0044 - val_accuracy: 0.6735\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.6594 - accuracy: 0.7829 - val_loss: 0.9731 - val_accuracy: 0.6744\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.5940 - accuracy: 0.8015 - val_loss: 0.9914 - val_accuracy: 0.6708\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.5666 - accuracy: 0.8089 - val_loss: 0.9876 - val_accuracy: 0.6782\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.5462 - accuracy: 0.8159 - val_loss: 1.0062 - val_accuracy: 0.6750\n",
      "\n",
      "Round = 16, total number of trainable parameters = 127541\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 17\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.6278, Accuracy: 43.1200, Test Loss: 1.3778, Test Accuracy: 51.160004\n",
      "Total number of trainable parameters = 127495\n",
      "\n",
      "Epoch 2, Loss: 1.2690, Accuracy: 56.0580, Test Loss: 1.2278, Test Accuracy: 56.720001\n",
      "Total number of trainable parameters = 127496\n",
      "\n",
      "Epoch 3, Loss: 1.1416, Accuracy: 60.4040, Test Loss: 1.1459, Test Accuracy: 59.810001\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 4, Loss: 1.0613, Accuracy: 63.3260, Test Loss: 1.1032, Test Accuracy: 61.479996\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 5, Loss: 1.0053, Accuracy: 65.1440, Test Loss: 1.0834, Test Accuracy: 61.919998\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 6, Loss: 0.9604, Accuracy: 66.6680, Test Loss: 1.0575, Test Accuracy: 62.639999\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 7, Loss: 0.9233, Accuracy: 67.8960, Test Loss: 1.0386, Test Accuracy: 63.590004\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 8, Loss: 0.8901, Accuracy: 69.0520, Test Loss: 1.0106, Test Accuracy: 64.900002\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 9, Loss: 0.8586, Accuracy: 70.1660, Test Loss: 1.0035, Test Accuracy: 65.220001\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 10, Loss: 0.8322, Accuracy: 71.0940, Test Loss: 0.9979, Test Accuracy: 65.349998\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 11, Loss: 0.8089, Accuracy: 71.8040, Test Loss: 1.0141, Test Accuracy: 65.270004\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 12, Loss: 0.7860, Accuracy: 72.7480, Test Loss: 0.9852, Test Accuracy: 66.110001\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 13, Loss: 0.7628, Accuracy: 73.5320, Test Loss: 0.9813, Test Accuracy: 66.189995\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 14, Loss: 0.7421, Accuracy: 74.3300, Test Loss: 0.9814, Test Accuracy: 66.250000\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 15, Loss: 0.7233, Accuracy: 74.9820, Test Loss: 0.9720, Test Accuracy: 66.769997\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 16, Loss: 0.7046, Accuracy: 75.6480, Test Loss: 0.9689, Test Accuracy: 67.150002\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 17, Loss: 0.6858, Accuracy: 76.3940, Test Loss: 0.9734, Test Accuracy: 67.299995\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 18, Loss: 0.6702, Accuracy: 77.0360, Test Loss: 0.9808, Test Accuracy: 66.930000\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "Epoch 19, Loss: 0.6515, Accuracy: 77.6960, Test Loss: 0.9812, Test Accuracy: 67.220001\n",
      "Total number of trainable parameters = 127497\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.6358 - accuracy: 0.7818 - val_loss: 0.9773 - val_accuracy: 0.6742\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 1.1431 - accuracy: 0.6346 - val_loss: 1.3127 - val_accuracy: 0.5848\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 1.0768 - accuracy: 0.6799 - val_loss: 1.2464 - val_accuracy: 0.6088\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.9917 - accuracy: 0.7117 - val_loss: 1.2011 - val_accuracy: 0.6229\n",
      "\n",
      "Round = 17, total number of trainable parameters = 102785\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 18\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.8507, Accuracy: 35.0040, Test Loss: 1.6614, Test Accuracy: 42.259998\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 2, Loss: 1.5616, Accuracy: 47.9080, Test Loss: 1.5035, Test Accuracy: 49.980000\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 3, Loss: 1.4245, Accuracy: 53.8260, Test Loss: 1.3942, Test Accuracy: 54.590000\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 4, Loss: 1.3260, Accuracy: 57.5120, Test Loss: 1.3243, Test Accuracy: 56.510002\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 5, Loss: 1.2465, Accuracy: 60.1600, Test Loss: 1.2765, Test Accuracy: 57.889999\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 6, Loss: 1.1785, Accuracy: 62.4260, Test Loss: 1.2197, Test Accuracy: 60.290001\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 7, Loss: 1.1203, Accuracy: 64.3860, Test Loss: 1.1778, Test Accuracy: 61.470001\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 8, Loss: 1.0684, Accuracy: 65.9100, Test Loss: 1.1426, Test Accuracy: 62.840004\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 9, Loss: 1.0241, Accuracy: 67.2320, Test Loss: 1.1125, Test Accuracy: 63.209999\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 10, Loss: 0.9813, Accuracy: 68.2740, Test Loss: 1.1015, Test Accuracy: 63.139999\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 11, Loss: 0.9468, Accuracy: 69.4220, Test Loss: 1.0686, Test Accuracy: 64.370003\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 12, Loss: 0.9142, Accuracy: 70.4120, Test Loss: 1.0664, Test Accuracy: 64.389999\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 13, Loss: 0.8853, Accuracy: 71.2960, Test Loss: 1.0434, Test Accuracy: 65.130005\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 14, Loss: 0.8594, Accuracy: 71.9660, Test Loss: 1.0323, Test Accuracy: 65.699997\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 15, Loss: 0.8340, Accuracy: 72.7920, Test Loss: 1.0328, Test Accuracy: 65.629997\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 16, Loss: 0.8127, Accuracy: 73.4080, Test Loss: 1.0154, Test Accuracy: 65.989998\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 17, Loss: 0.7931, Accuracy: 74.0460, Test Loss: 1.0284, Test Accuracy: 65.849998\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 18, Loss: 0.7741, Accuracy: 74.5540, Test Loss: 1.0102, Test Accuracy: 66.129997\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 19, Loss: 0.7544, Accuracy: 75.3460, Test Loss: 1.0066, Test Accuracy: 66.610001\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 20, Loss: 0.7371, Accuracy: 75.7660, Test Loss: 0.9965, Test Accuracy: 66.239998\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 21, Loss: 0.7206, Accuracy: 76.3780, Test Loss: 1.0025, Test Accuracy: 66.439995\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 22, Loss: 0.7058, Accuracy: 76.7920, Test Loss: 1.0059, Test Accuracy: 66.399994\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "Epoch 23, Loss: 0.6905, Accuracy: 77.2300, Test Loss: 1.0002, Test Accuracy: 66.599998\n",
      "Total number of trainable parameters = 102760\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.6776 - accuracy: 0.7771 - val_loss: 0.9955 - val_accuracy: 0.6689\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.9047 - accuracy: 0.7027 - val_loss: 1.1287 - val_accuracy: 0.6117\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 80s 2ms/sample - loss: 0.8535 - accuracy: 0.7220 - val_loss: 1.0828 - val_accuracy: 0.6337\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.8073 - accuracy: 0.7400 - val_loss: 1.0602 - val_accuracy: 0.6456\n",
      "\n",
      "Round = 18, total number of trainable parameters = 82892\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 19\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.9521, Accuracy: 31.6060, Test Loss: 1.7894, Test Accuracy: 39.910000\n",
      "Total number of trainable parameters = 82864\n",
      "\n",
      "Epoch 2, Loss: 1.6992, Accuracy: 44.4560, Test Loss: 1.6264, Test Accuracy: 46.049999\n",
      "Total number of trainable parameters = 82864\n",
      "\n",
      "Epoch 3, Loss: 1.5514, Accuracy: 50.6260, Test Loss: 1.5146, Test Accuracy: 52.369999\n",
      "Total number of trainable parameters = 82864\n",
      "\n",
      "Epoch 4, Loss: 1.4385, Accuracy: 55.2300, Test Loss: 1.4222, Test Accuracy: 54.430000\n",
      "Total number of trainable parameters = 82864\n",
      "\n",
      "Epoch 5, Loss: 1.3433, Accuracy: 58.3500, Test Loss: 1.3585, Test Accuracy: 57.040001\n",
      "Total number of trainable parameters = 82864\n",
      "\n",
      "Epoch 6, Loss: 1.2616, Accuracy: 61.3260, Test Loss: 1.2933, Test Accuracy: 58.789997\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 7, Loss: 1.1906, Accuracy: 63.5000, Test Loss: 1.2404, Test Accuracy: 60.639999\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 8, Loss: 1.1283, Accuracy: 65.0980, Test Loss: 1.2083, Test Accuracy: 61.580002\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 9, Loss: 1.0751, Accuracy: 66.4740, Test Loss: 1.1587, Test Accuracy: 61.979996\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 10, Loss: 1.0284, Accuracy: 67.8420, Test Loss: 1.1390, Test Accuracy: 63.010002\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 11, Loss: 0.9858, Accuracy: 69.1040, Test Loss: 1.1047, Test Accuracy: 63.720001\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 12, Loss: 0.9482, Accuracy: 69.9980, Test Loss: 1.0853, Test Accuracy: 64.060005\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 13, Loss: 0.9173, Accuracy: 70.8700, Test Loss: 1.0722, Test Accuracy: 64.360001\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 14, Loss: 0.8872, Accuracy: 71.8300, Test Loss: 1.0596, Test Accuracy: 64.870003\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 15, Loss: 0.8607, Accuracy: 72.2900, Test Loss: 1.0614, Test Accuracy: 64.860001\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 16, Loss: 0.8379, Accuracy: 72.8940, Test Loss: 1.0500, Test Accuracy: 65.139999\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 17, Loss: 0.8158, Accuracy: 73.6940, Test Loss: 1.0347, Test Accuracy: 65.529999\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 18, Loss: 0.7968, Accuracy: 74.1200, Test Loss: 1.0362, Test Accuracy: 65.760002\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 19, Loss: 0.7775, Accuracy: 74.7580, Test Loss: 1.0173, Test Accuracy: 65.739998\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 20, Loss: 0.7601, Accuracy: 75.3180, Test Loss: 1.0169, Test Accuracy: 65.650002\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 21, Loss: 0.7433, Accuracy: 75.7040, Test Loss: 1.0253, Test Accuracy: 65.940002\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 22, Loss: 0.7280, Accuracy: 76.2860, Test Loss: 1.0195, Test Accuracy: 66.159996\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 23, Loss: 0.7148, Accuracy: 76.7320, Test Loss: 1.0124, Test Accuracy: 66.059998\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 24, Loss: 0.6996, Accuracy: 77.2100, Test Loss: 1.0178, Test Accuracy: 66.339996\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 25, Loss: 0.6869, Accuracy: 77.6040, Test Loss: 1.0250, Test Accuracy: 66.299995\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 26, Loss: 0.6756, Accuracy: 77.8580, Test Loss: 1.0092, Test Accuracy: 66.529999\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 27, Loss: 0.6619, Accuracy: 78.4540, Test Loss: 1.0102, Test Accuracy: 66.459999\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 28, Loss: 0.6490, Accuracy: 78.8140, Test Loss: 1.0272, Test Accuracy: 66.339996\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "Epoch 29, Loss: 0.6380, Accuracy: 79.1500, Test Loss: 1.0222, Test Accuracy: 66.360001\n",
      "Total number of trainable parameters = 82865\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 75s 2ms/sample - loss: 0.6276 - accuracy: 0.7963 - val_loss: 1.0341 - val_accuracy: 0.6621\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.6463 - accuracy: 0.7893 - val_loss: 1.0154 - val_accuracy: 0.6652\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.6266 - accuracy: 0.7953 - val_loss: 1.0179 - val_accuracy: 0.6651\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.6143 - accuracy: 0.8001 - val_loss: 1.0196 - val_accuracy: 0.6656\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.6005 - accuracy: 0.8052 - val_loss: 1.0238 - val_accuracy: 0.6640\n",
      "\n",
      "Round = 19, total number of trainable parameters = 66941\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 20\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.9771, Accuracy: 29.8640, Test Loss: 1.8245, Test Accuracy: 36.860001\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 2, Loss: 1.7246, Accuracy: 42.8980, Test Loss: 1.6628, Test Accuracy: 45.100002\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 3, Loss: 1.5761, Accuracy: 49.9040, Test Loss: 1.5327, Test Accuracy: 50.690002\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 4, Loss: 1.4607, Accuracy: 54.3260, Test Loss: 1.4419, Test Accuracy: 53.670002\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 5, Loss: 1.3654, Accuracy: 57.8600, Test Loss: 1.3746, Test Accuracy: 56.040001\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 6, Loss: 1.2807, Accuracy: 60.7980, Test Loss: 1.3132, Test Accuracy: 58.840000\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 7, Loss: 1.2091, Accuracy: 62.8420, Test Loss: 1.2495, Test Accuracy: 60.110004\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 8, Loss: 1.1447, Accuracy: 64.7280, Test Loss: 1.2140, Test Accuracy: 60.900002\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 9, Loss: 1.0881, Accuracy: 66.3460, Test Loss: 1.1815, Test Accuracy: 61.720001\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 10, Loss: 1.0401, Accuracy: 67.5740, Test Loss: 1.1390, Test Accuracy: 62.680000\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 11, Loss: 0.9977, Accuracy: 68.8560, Test Loss: 1.1131, Test Accuracy: 63.470001\n",
      "Total number of trainable parameters = 66926\n",
      "\n",
      "Epoch 12, Loss: 0.9608, Accuracy: 69.6780, Test Loss: 1.0968, Test Accuracy: 64.209999\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 13, Loss: 0.9299, Accuracy: 70.3680, Test Loss: 1.0715, Test Accuracy: 64.680000\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 14, Loss: 0.9006, Accuracy: 71.2800, Test Loss: 1.0616, Test Accuracy: 64.850006\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 15, Loss: 0.8721, Accuracy: 71.9780, Test Loss: 1.0536, Test Accuracy: 65.020004\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 16, Loss: 0.8476, Accuracy: 72.7400, Test Loss: 1.0527, Test Accuracy: 65.029999\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 17, Loss: 0.8254, Accuracy: 73.2600, Test Loss: 1.0356, Test Accuracy: 65.240005\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 18, Loss: 0.8040, Accuracy: 74.1360, Test Loss: 1.0307, Test Accuracy: 65.720001\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 19, Loss: 0.7859, Accuracy: 74.6260, Test Loss: 1.0208, Test Accuracy: 65.869995\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 20, Loss: 0.7675, Accuracy: 75.1700, Test Loss: 1.0146, Test Accuracy: 66.199997\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 21, Loss: 0.7523, Accuracy: 75.6480, Test Loss: 1.0180, Test Accuracy: 66.129997\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 22, Loss: 0.7360, Accuracy: 75.9860, Test Loss: 1.0169, Test Accuracy: 66.029999\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 23, Loss: 0.7213, Accuracy: 76.5840, Test Loss: 1.0107, Test Accuracy: 66.199997\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 24, Loss: 0.7094, Accuracy: 76.8660, Test Loss: 1.0125, Test Accuracy: 66.449997\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 25, Loss: 0.6954, Accuracy: 77.3960, Test Loss: 1.0133, Test Accuracy: 66.360001\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "Epoch 26, Loss: 0.6835, Accuracy: 77.7520, Test Loss: 1.0281, Test Accuracy: 66.339996\n",
      "Total number of trainable parameters = 66927\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.6729 - accuracy: 0.7811 - val_loss: 1.0239 - val_accuracy: 0.6648\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.8100 - accuracy: 0.7345 - val_loss: 1.0945 - val_accuracy: 0.6372\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.7818 - accuracy: 0.7415 - val_loss: 1.0774 - val_accuracy: 0.6450\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.7534 - accuracy: 0.7514 - val_loss: 1.0535 - val_accuracy: 0.6519\n",
      "\n",
      "Round = 20, total number of trainable parameters = 54131\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 21\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 1.9991, Accuracy: 29.1300, Test Loss: 1.8610, Test Accuracy: 37.200001\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 2, Loss: 1.7771, Accuracy: 40.9680, Test Loss: 1.7175, Test Accuracy: 43.180000\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 3, Loss: 1.6418, Accuracy: 47.1140, Test Loss: 1.5964, Test Accuracy: 48.750000\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 4, Loss: 1.5315, Accuracy: 51.7480, Test Loss: 1.5110, Test Accuracy: 51.639999\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 5, Loss: 1.4395, Accuracy: 55.1860, Test Loss: 1.4343, Test Accuracy: 54.540001\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 6, Loss: 1.3601, Accuracy: 57.7840, Test Loss: 1.3691, Test Accuracy: 56.239998\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 7, Loss: 1.2907, Accuracy: 60.2120, Test Loss: 1.3128, Test Accuracy: 57.980000\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 8, Loss: 1.2266, Accuracy: 62.1540, Test Loss: 1.2745, Test Accuracy: 58.600002\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 9, Loss: 1.1717, Accuracy: 63.7200, Test Loss: 1.2335, Test Accuracy: 60.149998\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 10, Loss: 1.1226, Accuracy: 64.9620, Test Loss: 1.2025, Test Accuracy: 60.650002\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 11, Loss: 1.0791, Accuracy: 66.1020, Test Loss: 1.1739, Test Accuracy: 61.519997\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 12, Loss: 1.0408, Accuracy: 67.2040, Test Loss: 1.1488, Test Accuracy: 62.260002\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 13, Loss: 1.0069, Accuracy: 67.9640, Test Loss: 1.1350, Test Accuracy: 62.720001\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 14, Loss: 0.9770, Accuracy: 68.7780, Test Loss: 1.1173, Test Accuracy: 63.029999\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 15, Loss: 0.9501, Accuracy: 69.5580, Test Loss: 1.1052, Test Accuracy: 63.559998\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 16, Loss: 0.9245, Accuracy: 70.0640, Test Loss: 1.0899, Test Accuracy: 63.800003\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 17, Loss: 0.9029, Accuracy: 70.8280, Test Loss: 1.0824, Test Accuracy: 63.739998\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 18, Loss: 0.8818, Accuracy: 71.4820, Test Loss: 1.0988, Test Accuracy: 62.959999\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 19, Loss: 0.8632, Accuracy: 72.0380, Test Loss: 1.0779, Test Accuracy: 63.980000\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 20, Loss: 0.8466, Accuracy: 72.2820, Test Loss: 1.0597, Test Accuracy: 64.830002\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 21, Loss: 0.8302, Accuracy: 72.9540, Test Loss: 1.0641, Test Accuracy: 64.400002\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 22, Loss: 0.8137, Accuracy: 73.4140, Test Loss: 1.0545, Test Accuracy: 64.790001\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 23, Loss: 0.8007, Accuracy: 73.8160, Test Loss: 1.0453, Test Accuracy: 64.980003\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 24, Loss: 0.7865, Accuracy: 74.2160, Test Loss: 1.0578, Test Accuracy: 65.050003\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 25, Loss: 0.7733, Accuracy: 74.7640, Test Loss: 1.0470, Test Accuracy: 65.240005\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "Epoch 26, Loss: 0.7610, Accuracy: 75.1200, Test Loss: 1.0495, Test Accuracy: 65.570000\n",
      "Total number of trainable parameters = 54113\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.7507 - accuracy: 0.7524 - val_loss: 1.0461 - val_accuracy: 0.6528\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.8007 - accuracy: 0.7360 - val_loss: 1.0512 - val_accuracy: 0.6472\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.7731 - accuracy: 0.7452 - val_loss: 1.0663 - val_accuracy: 0.6422\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.7577 - accuracy: 0.7516 - val_loss: 1.0416 - val_accuracy: 0.6541\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.7467 - accuracy: 0.7541 - val_loss: 1.0465 - val_accuracy: 0.6562\n",
      "Epoch 6/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 0.7364 - accuracy: 0.7577 - val_loss: 1.0467 - val_accuracy: 0.6535\n",
      "Epoch 7/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 0.7269 - accuracy: 0.7603 - val_loss: 1.0445 - val_accuracy: 0.6554\n",
      "\n",
      "Round = 21, total number of trainable parameters = 43798\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 22\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 2.0226, Accuracy: 27.9000, Test Loss: 1.8903, Test Accuracy: 34.670002\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 2, Loss: 1.8113, Accuracy: 39.2020, Test Loss: 1.7478, Test Accuracy: 42.340000\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 3, Loss: 1.6815, Accuracy: 45.2500, Test Loss: 1.6318, Test Accuracy: 46.799999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 4, Loss: 1.5721, Accuracy: 49.8840, Test Loss: 1.5391, Test Accuracy: 50.379997\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 5, Loss: 1.4807, Accuracy: 53.1460, Test Loss: 1.4636, Test Accuracy: 52.999996\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 6, Loss: 1.4004, Accuracy: 56.0540, Test Loss: 1.4003, Test Accuracy: 54.949997\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 7, Loss: 1.3319, Accuracy: 58.1540, Test Loss: 1.3519, Test Accuracy: 56.209999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 8, Loss: 1.2706, Accuracy: 60.0480, Test Loss: 1.3035, Test Accuracy: 57.930000\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 9, Loss: 1.2155, Accuracy: 61.7420, Test Loss: 1.2673, Test Accuracy: 58.749996\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 10, Loss: 1.1673, Accuracy: 63.1520, Test Loss: 1.2312, Test Accuracy: 59.820004\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 11, Loss: 1.1242, Accuracy: 64.3000, Test Loss: 1.1988, Test Accuracy: 60.650002\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 12, Loss: 1.0859, Accuracy: 65.3060, Test Loss: 1.1803, Test Accuracy: 61.020000\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 13, Loss: 1.0517, Accuracy: 66.2480, Test Loss: 1.1597, Test Accuracy: 61.580002\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 14, Loss: 1.0202, Accuracy: 67.1920, Test Loss: 1.1376, Test Accuracy: 62.410004\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 15, Loss: 0.9937, Accuracy: 67.8620, Test Loss: 1.1230, Test Accuracy: 62.599998\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 16, Loss: 0.9687, Accuracy: 68.5840, Test Loss: 1.1115, Test Accuracy: 63.020004\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 17, Loss: 0.9454, Accuracy: 69.2880, Test Loss: 1.1017, Test Accuracy: 63.389999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 18, Loss: 0.9253, Accuracy: 69.9020, Test Loss: 1.0927, Test Accuracy: 63.639999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 19, Loss: 0.9062, Accuracy: 70.4220, Test Loss: 1.0786, Test Accuracy: 63.959999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 20, Loss: 0.8890, Accuracy: 70.9120, Test Loss: 1.0821, Test Accuracy: 63.639999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 21, Loss: 0.8718, Accuracy: 71.3020, Test Loss: 1.0668, Test Accuracy: 64.070000\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 22, Loss: 0.8575, Accuracy: 71.7420, Test Loss: 1.0610, Test Accuracy: 64.360001\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 23, Loss: 0.8419, Accuracy: 72.2880, Test Loss: 1.0673, Test Accuracy: 64.270004\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 24, Loss: 0.8313, Accuracy: 72.5120, Test Loss: 1.0566, Test Accuracy: 64.389999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 25, Loss: 0.8177, Accuracy: 72.9360, Test Loss: 1.0586, Test Accuracy: 64.820000\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 26, Loss: 0.8053, Accuracy: 73.2860, Test Loss: 1.0501, Test Accuracy: 64.889999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 27, Loss: 0.7952, Accuracy: 73.6960, Test Loss: 1.0723, Test Accuracy: 64.720001\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 28, Loss: 0.7857, Accuracy: 74.0880, Test Loss: 1.0584, Test Accuracy: 65.160004\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 29, Loss: 0.7754, Accuracy: 74.2740, Test Loss: 1.0471, Test Accuracy: 65.130005\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 30, Loss: 0.7652, Accuracy: 74.5460, Test Loss: 1.0473, Test Accuracy: 65.389999\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 31, Loss: 0.7568, Accuracy: 74.7580, Test Loss: 1.0445, Test Accuracy: 65.240005\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 32, Loss: 0.7480, Accuracy: 75.2300, Test Loss: 1.0420, Test Accuracy: 65.650002\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 33, Loss: 0.7392, Accuracy: 75.4180, Test Loss: 1.0459, Test Accuracy: 65.419998\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 34, Loss: 0.7319, Accuracy: 75.7520, Test Loss: 1.0491, Test Accuracy: 65.629997\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "Epoch 35, Loss: 0.7256, Accuracy: 75.8800, Test Loss: 1.0475, Test Accuracy: 65.540001\n",
      "Total number of trainable parameters = 43778\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 0.7167 - accuracy: 0.7630 - val_loss: 1.0481 - val_accuracy: 0.6580\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 1.2672 - accuracy: 0.6376 - val_loss: 1.5893 - val_accuracy: 0.5322\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 1.2948 - accuracy: 0.6202 - val_loss: 1.5299 - val_accuracy: 0.5381\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 79s 2ms/sample - loss: 1.2269 - accuracy: 0.6282 - val_loss: 1.4674 - val_accuracy: 0.5491\n",
      "\n",
      "Round = 22, total number of trainable parameters = 35512\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 23\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 2.0902, Accuracy: 24.8620, Test Loss: 1.9793, Test Accuracy: 31.230000\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 2, Loss: 1.9161, Accuracy: 35.3200, Test Loss: 1.8671, Test Accuracy: 38.029999\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 3, Loss: 1.8138, Accuracy: 40.2040, Test Loss: 1.7809, Test Accuracy: 41.530003\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 4, Loss: 1.7289, Accuracy: 43.3880, Test Loss: 1.7039, Test Accuracy: 44.670002\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 5, Loss: 1.6525, Accuracy: 46.0740, Test Loss: 1.6397, Test Accuracy: 45.770000\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 6, Loss: 1.5813, Accuracy: 47.8600, Test Loss: 1.5762, Test Accuracy: 48.230000\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 7, Loss: 1.5180, Accuracy: 50.7140, Test Loss: 1.5259, Test Accuracy: 49.110001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 8, Loss: 1.4612, Accuracy: 52.4240, Test Loss: 1.4820, Test Accuracy: 50.470001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 9, Loss: 1.4104, Accuracy: 52.3720, Test Loss: 1.4439, Test Accuracy: 51.160004\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 10, Loss: 1.3644, Accuracy: 54.3740, Test Loss: 1.4113, Test Accuracy: 52.060001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 11, Loss: 1.3231, Accuracy: 55.7280, Test Loss: 1.3868, Test Accuracy: 52.249996\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 12, Loss: 1.2879, Accuracy: 56.5800, Test Loss: 1.3622, Test Accuracy: 52.329998\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 13, Loss: 1.2555, Accuracy: 57.4180, Test Loss: 1.3407, Test Accuracy: 53.509998\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 14, Loss: 1.2278, Accuracy: 57.7020, Test Loss: 1.3242, Test Accuracy: 52.700001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 15, Loss: 1.2008, Accuracy: 58.1580, Test Loss: 1.3148, Test Accuracy: 53.740002\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 16, Loss: 1.1780, Accuracy: 58.7800, Test Loss: 1.2963, Test Accuracy: 53.310001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 17, Loss: 1.1577, Accuracy: 58.7540, Test Loss: 1.2922, Test Accuracy: 53.119999\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 18, Loss: 1.1378, Accuracy: 59.3580, Test Loss: 1.2774, Test Accuracy: 54.060001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 19, Loss: 1.1211, Accuracy: 60.1820, Test Loss: 1.2687, Test Accuracy: 53.730000\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 20, Loss: 1.1039, Accuracy: 59.9700, Test Loss: 1.2629, Test Accuracy: 54.030003\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 21, Loss: 1.0889, Accuracy: 59.8920, Test Loss: 1.2587, Test Accuracy: 54.009998\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 22, Loss: 1.0758, Accuracy: 60.4640, Test Loss: 1.2573, Test Accuracy: 54.280003\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 23, Loss: 1.0639, Accuracy: 61.0160, Test Loss: 1.2508, Test Accuracy: 54.739998\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 24, Loss: 1.0520, Accuracy: 61.1900, Test Loss: 1.2551, Test Accuracy: 53.930004\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 25, Loss: 1.0405, Accuracy: 61.5560, Test Loss: 1.2502, Test Accuracy: 54.170002\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 26, Loss: 1.0310, Accuracy: 61.4460, Test Loss: 1.2467, Test Accuracy: 55.500000\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 27, Loss: 1.0219, Accuracy: 62.2220, Test Loss: 1.2381, Test Accuracy: 56.080002\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 28, Loss: 1.0108, Accuracy: 62.2340, Test Loss: 1.2540, Test Accuracy: 55.110001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 29, Loss: 1.0045, Accuracy: 62.8040, Test Loss: 1.2375, Test Accuracy: 56.040001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 30, Loss: 0.9957, Accuracy: 62.5740, Test Loss: 1.2403, Test Accuracy: 55.980003\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 31, Loss: 0.9871, Accuracy: 63.4520, Test Loss: 1.2462, Test Accuracy: 55.860001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "Epoch 32, Loss: 0.9815, Accuracy: 63.5720, Test Loss: 1.2397, Test Accuracy: 56.110001\n",
      "Total number of trainable parameters = 35487\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 0.9740 - accuracy: 0.6403 - val_loss: 1.2375 - val_accuracy: 0.5515\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.4596 - accuracy: 0.5418 - val_loss: 1.7060 - val_accuracy: 0.4653\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.4498 - accuracy: 0.5390 - val_loss: 1.6109 - val_accuracy: 0.4811\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 78s 2ms/sample - loss: 1.3636 - accuracy: 0.5496 - val_loss: 1.5553 - val_accuracy: 0.4806\n",
      "\n",
      "Round = 23, total number of trainable parameters = 28798\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 24\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 2.1286, Accuracy: 24.8280, Test Loss: 2.0329, Test Accuracy: 31.779999\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 2, Loss: 1.9691, Accuracy: 34.5920, Test Loss: 1.9233, Test Accuracy: 36.559998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 3, Loss: 1.8758, Accuracy: 37.9480, Test Loss: 1.8493, Test Accuracy: 40.049999\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 4, Loss: 1.8010, Accuracy: 41.0300, Test Loss: 1.7811, Test Accuracy: 42.500000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 5, Loss: 1.7357, Accuracy: 44.3500, Test Loss: 1.7245, Test Accuracy: 41.690002\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 6, Loss: 1.6760, Accuracy: 45.7640, Test Loss: 1.6727, Test Accuracy: 43.110001\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 7, Loss: 1.6199, Accuracy: 45.8120, Test Loss: 1.6232, Test Accuracy: 44.260002\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 8, Loss: 1.5699, Accuracy: 46.9700, Test Loss: 1.5832, Test Accuracy: 45.270000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 9, Loss: 1.5260, Accuracy: 48.5180, Test Loss: 1.5494, Test Accuracy: 46.009998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 10, Loss: 1.4849, Accuracy: 48.6720, Test Loss: 1.5158, Test Accuracy: 48.509998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 11, Loss: 1.4483, Accuracy: 49.9260, Test Loss: 1.4918, Test Accuracy: 49.309998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 12, Loss: 1.4167, Accuracy: 50.2320, Test Loss: 1.4665, Test Accuracy: 49.480000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 13, Loss: 1.3863, Accuracy: 52.1140, Test Loss: 1.4470, Test Accuracy: 49.419998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 14, Loss: 1.3605, Accuracy: 51.9160, Test Loss: 1.4322, Test Accuracy: 49.730000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 15, Loss: 1.3371, Accuracy: 52.5020, Test Loss: 1.4216, Test Accuracy: 48.020000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 16, Loss: 1.3154, Accuracy: 52.3820, Test Loss: 1.4062, Test Accuracy: 50.029999\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 17, Loss: 1.2965, Accuracy: 53.4200, Test Loss: 1.4086, Test Accuracy: 47.350002\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 18, Loss: 1.2784, Accuracy: 53.6480, Test Loss: 1.3863, Test Accuracy: 50.279999\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 19, Loss: 1.2634, Accuracy: 53.3280, Test Loss: 1.3814, Test Accuracy: 49.149998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 20, Loss: 1.2489, Accuracy: 54.3620, Test Loss: 1.3721, Test Accuracy: 50.570000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 21, Loss: 1.2350, Accuracy: 53.4800, Test Loss: 1.3686, Test Accuracy: 50.849998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 22, Loss: 1.2231, Accuracy: 54.9660, Test Loss: 1.3605, Test Accuracy: 50.840004\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 23, Loss: 1.2109, Accuracy: 54.3280, Test Loss: 1.3625, Test Accuracy: 49.340000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 24, Loss: 1.2012, Accuracy: 54.2320, Test Loss: 1.3539, Test Accuracy: 50.840004\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 25, Loss: 1.1910, Accuracy: 55.6100, Test Loss: 1.3496, Test Accuracy: 50.919998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 26, Loss: 1.1813, Accuracy: 56.1620, Test Loss: 1.3470, Test Accuracy: 51.169998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 27, Loss: 1.1732, Accuracy: 55.3120, Test Loss: 1.3444, Test Accuracy: 49.639999\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 28, Loss: 1.1649, Accuracy: 55.1600, Test Loss: 1.3496, Test Accuracy: 51.059998\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 29, Loss: 1.1570, Accuracy: 56.4660, Test Loss: 1.3429, Test Accuracy: 51.330002\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 30, Loss: 1.1482, Accuracy: 56.2640, Test Loss: 1.3434, Test Accuracy: 51.450001\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 31, Loss: 1.1423, Accuracy: 55.7580, Test Loss: 1.3413, Test Accuracy: 51.470001\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 32, Loss: 1.1366, Accuracy: 56.2840, Test Loss: 1.3467, Test Accuracy: 49.810001\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 33, Loss: 1.1290, Accuracy: 56.7020, Test Loss: 1.3406, Test Accuracy: 51.440002\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 34, Loss: 1.1227, Accuracy: 56.8500, Test Loss: 1.3411, Test Accuracy: 51.430000\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 35, Loss: 1.1174, Accuracy: 56.6280, Test Loss: 1.3602, Test Accuracy: 51.090004\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "Epoch 36, Loss: 1.1120, Accuracy: 56.6260, Test Loss: 1.3485, Test Accuracy: 51.310001\n",
      "Total number of trainable parameters = 28782\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 1.1062 - accuracy: 0.5741 - val_loss: 1.3447 - val_accuracy: 0.5149\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.2549 - accuracy: 0.5373 - val_loss: 1.3832 - val_accuracy: 0.5046\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.1785 - accuracy: 0.5551 - val_loss: 1.3679 - val_accuracy: 0.5086\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.1596 - accuracy: 0.5549 - val_loss: 1.3550 - val_accuracy: 0.5099\n",
      "\n",
      "Round = 24, total number of trainable parameters = 23411\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Iterative pruning round: 25\n",
      "\n",
      "\n",
      "Epoch 1, Loss: 2.1616, Accuracy: 22.1240, Test Loss: 2.0794, Test Accuracy: 29.830000\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 2, Loss: 2.0259, Accuracy: 31.4660, Test Loss: 1.9902, Test Accuracy: 33.770000\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 3, Loss: 1.9414, Accuracy: 35.0480, Test Loss: 1.9138, Test Accuracy: 36.239998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 4, Loss: 1.8704, Accuracy: 37.5280, Test Loss: 1.8477, Test Accuracy: 39.369999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 5, Loss: 1.8082, Accuracy: 40.2460, Test Loss: 1.7937, Test Accuracy: 40.779999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 6, Loss: 1.7541, Accuracy: 41.7780, Test Loss: 1.7426, Test Accuracy: 42.010002\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 7, Loss: 1.7043, Accuracy: 42.3740, Test Loss: 1.6965, Test Accuracy: 42.070000\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 8, Loss: 1.6567, Accuracy: 44.4420, Test Loss: 1.6557, Test Accuracy: 43.309998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 9, Loss: 1.6131, Accuracy: 45.3660, Test Loss: 1.6172, Test Accuracy: 44.310001\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 10, Loss: 1.5737, Accuracy: 46.6860, Test Loss: 1.5854, Test Accuracy: 44.840000\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 11, Loss: 1.5376, Accuracy: 46.9560, Test Loss: 1.5582, Test Accuracy: 45.529999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 12, Loss: 1.5059, Accuracy: 47.6280, Test Loss: 1.5355, Test Accuracy: 45.730000\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 13, Loss: 1.4762, Accuracy: 48.3940, Test Loss: 1.5185, Test Accuracy: 48.030003\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 14, Loss: 1.4502, Accuracy: 48.7360, Test Loss: 1.4945, Test Accuracy: 46.570000\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 15, Loss: 1.4255, Accuracy: 49.8660, Test Loss: 1.4754, Test Accuracy: 48.669998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 16, Loss: 1.4037, Accuracy: 50.1280, Test Loss: 1.4605, Test Accuracy: 48.899998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 17, Loss: 1.3837, Accuracy: 50.2300, Test Loss: 1.4483, Test Accuracy: 49.049999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 18, Loss: 1.3651, Accuracy: 51.7180, Test Loss: 1.4434, Test Accuracy: 48.989998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 19, Loss: 1.3493, Accuracy: 51.2580, Test Loss: 1.4297, Test Accuracy: 48.919998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 20, Loss: 1.3331, Accuracy: 51.6220, Test Loss: 1.4180, Test Accuracy: 49.380001\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 21, Loss: 1.3198, Accuracy: 52.3680, Test Loss: 1.4097, Test Accuracy: 49.540001\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 22, Loss: 1.3070, Accuracy: 52.7220, Test Loss: 1.4006, Test Accuracy: 49.689999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 23, Loss: 1.2954, Accuracy: 52.4940, Test Loss: 1.4015, Test Accuracy: 49.669998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 24, Loss: 1.2840, Accuracy: 52.5840, Test Loss: 1.3954, Test Accuracy: 50.089996\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 25, Loss: 1.2744, Accuracy: 53.0860, Test Loss: 1.3920, Test Accuracy: 48.689999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 26, Loss: 1.2643, Accuracy: 52.7260, Test Loss: 1.3892, Test Accuracy: 50.089996\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 27, Loss: 1.2557, Accuracy: 53.4840, Test Loss: 1.3826, Test Accuracy: 50.099998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 28, Loss: 1.2466, Accuracy: 53.5740, Test Loss: 1.3816, Test Accuracy: 50.040001\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 29, Loss: 1.2397, Accuracy: 53.5180, Test Loss: 1.3735, Test Accuracy: 50.150002\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 30, Loss: 1.2326, Accuracy: 54.2520, Test Loss: 1.3774, Test Accuracy: 50.559998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 31, Loss: 1.2247, Accuracy: 53.8160, Test Loss: 1.3728, Test Accuracy: 50.400002\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 32, Loss: 1.2175, Accuracy: 54.4740, Test Loss: 1.3663, Test Accuracy: 50.730003\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 33, Loss: 1.2120, Accuracy: 54.1760, Test Loss: 1.3661, Test Accuracy: 50.660004\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 34, Loss: 1.2050, Accuracy: 54.2420, Test Loss: 1.3691, Test Accuracy: 50.459999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 35, Loss: 1.1991, Accuracy: 54.7420, Test Loss: 1.3655, Test Accuracy: 50.629997\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 36, Loss: 1.1939, Accuracy: 54.8640, Test Loss: 1.3597, Test Accuracy: 50.680000\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 37, Loss: 1.1878, Accuracy: 54.9400, Test Loss: 1.3585, Test Accuracy: 50.840004\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 38, Loss: 1.1838, Accuracy: 55.2680, Test Loss: 1.3612, Test Accuracy: 50.950005\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 39, Loss: 1.1782, Accuracy: 55.3420, Test Loss: 1.3559, Test Accuracy: 50.779999\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 40, Loss: 1.1740, Accuracy: 55.0300, Test Loss: 1.3583, Test Accuracy: 50.919998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 41, Loss: 1.1676, Accuracy: 55.7800, Test Loss: 1.3607, Test Accuracy: 50.770004\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 42, Loss: 1.1636, Accuracy: 55.6960, Test Loss: 1.3539, Test Accuracy: 51.120003\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 43, Loss: 1.1591, Accuracy: 55.2560, Test Loss: 1.3645, Test Accuracy: 50.840004\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 44, Loss: 1.1553, Accuracy: 55.7160, Test Loss: 1.3559, Test Accuracy: 51.160004\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "Epoch 45, Loss: 1.1503, Accuracy: 55.7800, Test Loss: 1.3585, Test Accuracy: 50.989998\n",
      "Total number of trainable parameters = 23399\n",
      "\n",
      "\n",
      "'EarlyStopping' called!\n",
      "\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 1.1468 - accuracy: 0.5616 - val_loss: 1.3530 - val_accuracy: 0.4986\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.1824 - accuracy: 0.5472 - val_loss: 1.3569 - val_accuracy: 0.4916\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.1711 - accuracy: 0.5504 - val_loss: 1.3583 - val_accuracy: 0.4921\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.1624 - accuracy: 0.5508 - val_loss: 1.3517 - val_accuracy: 0.4939\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.1566 - accuracy: 0.5540 - val_loss: 1.3548 - val_accuracy: 0.5015\n",
      "Epoch 6/100\n",
      "50000/50000 [==============================] - 77s 2ms/sample - loss: 1.1507 - accuracy: 0.5579 - val_loss: 1.3583 - val_accuracy: 0.4909\n",
      "Epoch 7/100\n",
      "50000/50000 [==============================] - 76s 2ms/sample - loss: 1.1470 - accuracy: 0.5546 - val_loss: 1.3598 - val_accuracy: 0.4948\n",
      "\n",
      "Round = 25, total number of trainable parameters = 19084\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for i in range(1, num_pruning_rounds + 1):\n",
    "    \n",
    "    print(\"\\n\\n\\nIterative pruning round: {0}\\n\\n\".format(i))\n",
    "    \n",
    "    # Define 'train_one_step()' and 'test_step()' functions here-\n",
    "    @tf.function\n",
    "    def train_one_step(model, mask_model, optimizer, x, y):\n",
    "        '''\n",
    "        Function to compute one step of gradient descent optimization\n",
    "        '''\n",
    "        with tf.GradientTape() as tape:\n",
    "            # Make predictions using defined model-\n",
    "            y_pred = model(x)\n",
    "\n",
    "            # Compute loss-\n",
    "            loss = loss_fn(y, y_pred)\n",
    "        \n",
    "        # Compute gradients wrt defined loss and weights and biases-\n",
    "        grads = tape.gradient(loss, model.trainable_variables)\n",
    "    \n",
    "        # type(grads)\n",
    "        # list\n",
    "    \n",
    "        # List to hold element-wise multiplication between-\n",
    "        # computed gradient and masks-\n",
    "        grad_mask_mul = []\n",
    "    \n",
    "        # Perform element-wise multiplication between computed gradients and masks-\n",
    "        for grad_layer, mask in zip(grads, mask_model.trainable_weights):\n",
    "            grad_mask_mul.append(tf.math.multiply(grad_layer, mask))\n",
    "    \n",
    "        # Apply computed gradients to model's weights and biases-\n",
    "        optimizer.apply_gradients(zip(grad_mask_mul, model.trainable_variables))\n",
    "\n",
    "        # Compute accuracy-\n",
    "        train_loss(loss)\n",
    "        train_accuracy(y, y_pred)\n",
    "\n",
    "        return None\n",
    "    \n",
    "    \n",
    "    @tf.function\n",
    "    def test_step(model, optimizer, data, labels):\n",
    "        \"\"\"\n",
    "        Function to test model performance\n",
    "        on testing dataset\n",
    "        \"\"\"\n",
    "    \n",
    "        predictions = model(data)\n",
    "        t_loss = loss_fn(labels, predictions)\n",
    "\n",
    "        test_loss(t_loss)\n",
    "        test_accuracy(labels, predictions)\n",
    "\n",
    "        return None\n",
    "\n",
    "\n",
    "    # Instantiate a model\n",
    "    model_gt = pruned_nn(pruning_params_unpruned, pruning_params_unpruned)\n",
    "    \n",
    "    # Load winning ticket (from above)-\n",
    "    model_gt.load_weights(\"Conv_2_CIFAR_Winning_Ticket.h5\")\n",
    "    \n",
    "    # Strip model of pruning parameters-\n",
    "    model_gt_stripped = sparsity.strip_pruning(model_gt)\n",
    "    \n",
    "    \n",
    "    # Train model using 'GradientTape'-\n",
    "    \n",
    "    # Initialize parameters for Early Stopping manual implementation-\n",
    "    best_val_loss = 100\n",
    "    loc_patience = 0\n",
    "    \n",
    "    for epoch in range(num_epochs):\n",
    "    \n",
    "        if loc_patience >= patience:\n",
    "            print(\"\\n'EarlyStopping' called!\\n\")\n",
    "            break\n",
    "        \n",
    "        # Reset the metrics at the start of the next epoch\n",
    "        train_loss.reset_states()\n",
    "        train_accuracy.reset_states()\n",
    "        test_loss.reset_states()\n",
    "        test_accuracy.reset_states()\n",
    "        \n",
    "        \n",
    "        '''\n",
    "        # Initialize 'grad_mask_mul' list-\n",
    "        grad_mask_mul = []\n",
    "    \n",
    "        # Initialize all values to one-\n",
    "        for wts in mask_model_stripped.trainable_weights:\n",
    "            grad_mask_mul.append(wts.assign(tf.ones_like(input = wts,dtype = tf.float32)))\n",
    "    \n",
    "        # Convert from Python list to tf.Tensor-\n",
    "        grad_mask_mul = tf.convert_to_tensor(grad_mask_mul, dtype=tf.float32)\n",
    "    \n",
    "        print(\"type(grad_mask_mul): {0}\".format(type(grad_mask_mul)))\n",
    "        '''\n",
    "    \n",
    "        for x, y in train_dataset:\n",
    "            # train_one_step(model_gt_stripped, mask_model, optimizer, x, y, grad_mask_mul)\n",
    "            train_one_step(model_gt_stripped, mask_model_stripped, optimizer, x, y)\n",
    "\n",
    "\n",
    "        for x_t, y_t in test_dataset:\n",
    "            # test_step(x_t, y_t)\n",
    "            test_step(model_gt_stripped, optimizer, x_t, y_t)\n",
    "\n",
    "        template = 'Epoch {0}, Loss: {1:.4f}, Accuracy: {2:.4f}, Test Loss: {3:.4f}, Test Accuracy: {4:4f}'\n",
    "    \n",
    "        # 'i' is the index for number of pruning rounds-\n",
    "        history_main[i]['accuracy'][epoch] = train_accuracy.result() * 100\n",
    "        history_main[i]['loss'][epoch] = train_loss.result()\n",
    "        history_main[i]['val_loss'][epoch] = test_loss.result()\n",
    "        history_main[i]['val_accuracy'][epoch] = test_accuracy.result() * 100\n",
    "        \n",
    "        print(template.format(epoch + 1, \n",
    "                              train_loss.result(), train_accuracy.result()*100,\n",
    "                              test_loss.result(), test_accuracy.result()*100))\n",
    "    \n",
    "        # Count number of non-zero parameters in each layer and in total-\n",
    "        # print(\"layer-wise manner model, number of nonzero parameters in each layer are: \\n\")\n",
    "\n",
    "        model_sum_params = 0\n",
    "    \n",
    "        for layer in model_gt_stripped.trainable_weights:\n",
    "            # print(tf.math.count_nonzero(layer, axis = None).numpy())\n",
    "            model_sum_params += tf.math.count_nonzero(layer, axis = None).numpy()\n",
    "    \n",
    "        print(\"Total number of trainable parameters = {0}\\n\".format(model_sum_params))\n",
    "\n",
    "    \n",
    "        # Code for manual Early Stopping:\n",
    "        if np.abs(test_loss.result() < best_val_loss) >= minimum_delta:\n",
    "            # update 'best_val_loss' variable to lowest loss encountered so far-\n",
    "            best_val_loss = test_loss.result()\n",
    "        \n",
    "            # reset 'loc_patience' variable-\n",
    "            loc_patience = 0\n",
    "        \n",
    "        else:  # there is no improvement in monitored metric 'val_loss'\n",
    "            loc_patience += 1  # number of epochs without any improvement\n",
    "\n",
    "    \n",
    "    # Resize numpy arrays according to the epoch when 'EarlyStopping' was called-\n",
    "    for metrics in history_main[i].keys():\n",
    "        history_main[i][metrics] = np.resize(history_main[i][metrics], new_shape = epoch)\n",
    "        # history[metrics] = np.resize(history[metrics], new_shape=epoch)\n",
    "     \n",
    "    \n",
    "    # Save trained model weights-\n",
    "    model_gt.save_weights(\"Conv_2_CIFAR_Trained_Weights.h5\", overwrite=True)\n",
    "    \n",
    "    \n",
    "    \n",
    "    \n",
    "    # Prune trained model:\n",
    "    \n",
    "    # print(\"\\n% of weights to be pruned in round = {0} is: {1:.4f}\\n\".format(i, wts_np[i - 1]))\n",
    "    \n",
    "    # Specify the parameters to be used for layer-wise pruning, Conv layer-\n",
    "    pruning_params_conv = {\n",
    "        'pruning_schedule': sparsity.ConstantSparsity(\n",
    "            target_sparsity=conv1_pruning[i - 1], begin_step = 1000,\n",
    "            end_step = end_step, frequency=100\n",
    "        )\n",
    "    }\n",
    "\n",
    "    \n",
    "    # Specify the parameters to be used for layer-wise pruning, Fully-Connected layer pruning-\n",
    "    pruning_params_fc = {\n",
    "        'pruning_schedule': sparsity.ConstantSparsity(\n",
    "            target_sparsity=dense1_pruning[i - 1], begin_step = 1000,\n",
    "            end_step = end_step, frequency=100\n",
    "        )\n",
    "    }\n",
    "    \n",
    "    \n",
    "    # Instantiate a Nueal Network model to be pruned using parameters from above-\n",
    "    pruned_model = pruned_nn(pruning_params_conv, pruning_params_fc)\n",
    "    \n",
    "    # Load weights from original trained and unpruned model-\n",
    "    pruned_model.load_weights(\"Conv_2_CIFAR_Trained_Weights.h5\")\n",
    "    \n",
    "    # Train pruned NN-\n",
    "    history_pruned = pruned_model.fit(\n",
    "        x = X_train, y = y_train,\n",
    "        batch_size = batch_size,\n",
    "        epochs = epochs,\n",
    "        verbose = 1,\n",
    "        callbacks = callback,\n",
    "        validation_data = (X_test, y_test),\n",
    "        shuffle = True\n",
    "    )\n",
    "    \n",
    "    # Strip the pruning wrappers from pruned model-\n",
    "    pruned_model_stripped = sparsity.strip_pruning(pruned_model)\n",
    "    \n",
    "    # print(\"\\nIn pruned model, number of nonzero parameters in each layer are: \\n\")\n",
    "    pruned_sum_params = 0\n",
    "    \n",
    "    for layer in pruned_model_stripped.trainable_weights:\n",
    "        # print(tf.math.count_nonzero(layer, axis = None).numpy())\n",
    "        pruned_sum_params += tf.math.count_nonzero(layer, axis = None).numpy()\n",
    "    \n",
    "    print(\"\\nRound = {0}, total number of trainable parameters = {1}\\n\".format(i, pruned_sum_params))\n",
    "    # print(\"\\nTotal number of trainable parameters = {0}\\n\".format(pruned_sum_params))\n",
    "    \n",
    "    '''\n",
    "    # Sanity-check: confirm that the weights are actually pruned away from the network-\n",
    "    print(\"\\nRound = {0}, % of weights pruned away = {1:.2f}%\\n\".format( \\\n",
    "                                                i, (orig_sum_params - pruned_sum_params) / orig_sum_params * 100))\n",
    "    '''\n",
    "\n",
    "    # 'i' is the index for number of pruning rounds-\n",
    "    history_main[i]['percentage_wts_pruned'] = ((orig_sum_params - pruned_sum_params) / orig_sum_params) * 100\n",
    " \n",
    "    \n",
    "    # Save weights of PRUNED and Trained model BEFORE stripping-\n",
    "    pruned_model.save_weights(\"Conv_2_CIFAR_Pruned_Weights.h5\", overwrite = True)\n",
    "    \n",
    "    \n",
    "    # Create a mask:\n",
    "    \n",
    "    # Instantiate a new neural network model for which, the mask is to be created,\n",
    "    mask_model = pruned_nn(pruning_params_unpruned, pruning_params_unpruned)\n",
    "    \n",
    "    # Load weights of PRUNED model-\n",
    "    mask_model.load_weights(\"Conv_2_CIFAR_Pruned_Weights.h5\")\n",
    "    \n",
    "    # Strip the model of its pruning parameters-\n",
    "    mask_model_stripped = sparsity.strip_pruning(mask_model)\n",
    "    \n",
    "    # For each layer, for each weight which is 0, leave it, as is.\n",
    "    # And for weights which survive the pruning,reinitialize it to ONE (1)-\n",
    "    for wts in mask_model_stripped.trainable_weights:\n",
    "        wts.assign(tf.where(tf.equal(wts, 0.), 0., 1.))\n",
    "    \n",
    "    \n",
    "    \n",
    "    \n",
    "    # Extract Winning Ticket:\n",
    "    \n",
    "    # Instantiate a new neural network model for which, the weights are to be extracted-\n",
    "    winning_ticket_model = pruned_nn(pruning_params_unpruned, pruning_params_unpruned)\n",
    "    \n",
    "    # Load weights of PRUNED model-\n",
    "    winning_ticket_model.load_weights(\"Conv_2_CIFAR_Pruned_Weights.h5\")\n",
    "    \n",
    "    # Strip the model of its pruning parameters-\n",
    "    winning_ticket_model_stripped = sparsity.strip_pruning(winning_ticket_model)\n",
    "    \n",
    "    # For each layer, for each weight which is 0, leave it, as is. And for weights which survive the pruning,\n",
    "    # reinitialize it to the value, the model received BEFORE it was trained and pruned-\n",
    "    for orig_wts, pruned_wts in zip(orig_model_stripped.trainable_weights,\n",
    "                                    winning_ticket_model_stripped.trainable_weights):\n",
    "        pruned_wts.assign(tf.where(tf.equal(pruned_wts, 0), pruned_wts, orig_wts))\n",
    "    \n",
    "    \n",
    "    # Save the weights (with pruning parameters) extracted to a file-\n",
    "    winning_ticket_model.save_weights(\"Conv_2_CIFAR_Winning_Ticket.h5\", overwrite=True)\n",
    "    \n",
    "    \n",
    "    # Write message to log file:\n",
    "    with open(\"/home/majumdar/The_Lottery_Ticket_Hypothesis-Resources/Conv2_log_file.txt\", \"a+\") as f:\n",
    "        f.write(\"Completed iterative pruning round: {0}\\n\".format(i))\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Iterative-pruning for Conv-2 CNN using Lottery Ticket Hypothesis is now complete.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"\\nIterative-pruning for Conv-2 CNN using Lottery Ticket Hypothesis is now complete.\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pickle\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/home/majumdar/The_Lottery_Ticket_Hypothesis-Resources'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "os.getcwd()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Pickle/Save Python 3 dictionary for later use-\n",
    "with open(\"/home/majumdar/The_Lottery_Ticket_Hypothesis-Resources/Conv2_history_main_Experiment_3.pkl\", \"wb\") as f:\n",
    "    pickle.dump(history_main, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load saved Python 3 dictionary-\n",
    "with open(\"/home/majumdar/The_Lottery_Ticket_Hypothesis-Resources/Conv2_history_main_Experiment_3.pkl\", \"rb\") as f:\n",
    "    history_main = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Finding _winning ticket_:\n",
    "\n",
    "A _winning ticket_ is a sub-network which has:\n",
    "\n",
    "1. equivalent or higher _validation accuracy_\n",
    "1. equivalent or lower _number of epochs (using early stopping_)\n",
    "1. maximum possible percentage of weights pruned\n",
    "\n",
    "as compared to the original, over-parametrized, unpruned convolutional neural network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "base accuracy = 70.1900% & base number of epochs = 9\n",
      "\n"
     ]
    }
   ],
   "source": [
    "base_accuracy = 0\n",
    "base_num_epochs = 0\n",
    "\n",
    "\n",
    "base_num_epochs = len(history_main[1]['loss'])\n",
    "base_accuracy = history_main[1]['val_accuracy'][base_num_epochs - 1]\n",
    "\n",
    "base_accuracy\n",
    "\n",
    "print(\"\\nbase accuracy = {0:.4f}% & base number of epochs = {1}\\n\".format(base_accuracy, base_num_epochs))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "best_accuracy = base_accuracy\n",
    "best_num_epochs = base_num_epochs\n",
    "best_wts = 100\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "best accuracy = 70.6000%, best number of epochs = 8 & % of weights pruned = 35.8421%\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# NOTE: A winning ticket has 'maximum' amount of % of weights pruned-\n",
    "\n",
    "for k in history_main.keys():\n",
    "    num_epochs = len(history_main[k]['loss'])\n",
    "    if history_main[k]['val_accuracy'][num_epochs - 1] >= best_accuracy and num_epochs <= best_num_epochs:\n",
    "        best_accuracy = history_main[k]['val_accuracy'][num_epochs - 1]\n",
    "        best_num_epochs = num_epochs\n",
    "        best_wts = history_main[k]['percentage_wts_pruned']\n",
    "\n",
    "\n",
    "print(\"\\nbest accuracy = {0:.4f}%, best number of epochs = {1} & % of weights pruned = {2:.4f}%\\n\".format(best_accuracy, best_num_epochs, best_wts))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Round = 1, % of wts pruned = 19.9073%\n",
      "val_accuracy: [54.40999985 61.54999542 65.52000427 67.84999847 68.13999939 69.33000183\n",
      " 69.27999878 69.83000183 70.19000244]\n",
      "\n",
      "\n",
      "Round = 2, % of wts pruned = 35.8421%\n",
      "val_accuracy: [60.63999939 66.18999481 68.40000153 68.01000214 69.95999908 70.47000122\n",
      " 69.34000397 70.59999847]\n",
      "\n",
      "\n",
      "Round = 3, % of wts pruned = 48.5987%\n",
      "val_accuracy: [63.41000366 66.84999847 68.26999664 69.09999847 69.36000061 69.23999786\n",
      " 69.88000488 69.70000458]\n",
      "\n",
      "\n",
      "Round = 4, % of wts pruned = 58.8107%\n",
      "val_accuracy: [64.70999908 67.0799942  68.70999908 69.01999664 69.48000336 69.59999847\n",
      " 70.30000305 69.33000183 69.98000336]\n",
      "\n",
      "\n",
      "Round = 5, % of wts pruned = 66.9869%\n",
      "val_accuracy: [64.78000641 67.54000092 69.16999817 68.56999969 70.37000275 70.12000275\n",
      " 70.22000122 69.11000061]\n",
      "\n",
      "\n",
      "Round = 6, % of wts pruned = 73.5347%\n",
      "val_accuracy: [65.09000397 67.47000122 68.33000183 68.87999725 69.84000397 70.5\n",
      " 69.88999939 69.27000427]\n",
      "\n",
      "\n",
      "Round = 7, % of wts pruned = 78.7772%\n",
      "val_accuracy: [65.20000458 67.86000061 69.30999756 69.30000305 69.95000458 70.45999908\n",
      " 70.38000488]\n",
      "\n",
      "\n",
      "Round = 8, % of wts pruned = 82.9771%\n",
      "val_accuracy: [65.77999878 67.47000122 69.02999878 69.61000061 69.45999908 69.70000458\n",
      " 70.16999817]\n",
      "\n",
      "\n",
      "Round = 9, % of wts pruned = 86.3404%\n",
      "val_accuracy: [64.08000183 66.86000061 68.75       68.83999634 69.40000153 69.76000214\n",
      " 69.80999756]\n",
      "\n",
      "\n",
      "Round = 10, % of wts pruned = 89.0362%\n",
      "val_accuracy: [63.84999847 66.23999786 68.37000275 68.94000244 68.87000275 69.5\n",
      " 69.31999969]\n",
      "\n",
      "\n",
      "Round = 11, % of wts pruned = 91.1955%\n",
      "val_accuracy: [62.66999817 65.19000244 67.47999573 68.69999695 68.11000061 69.09000397\n",
      " 69.34999847]\n",
      "\n",
      "\n",
      "Round = 12, % of wts pruned = 92.9264%\n",
      "val_accuracy: [61.05999756 65.25       66.84999847 67.51000214 68.55999756 68.83000183\n",
      " 69.16000366 69.         68.91000366]\n",
      "\n",
      "\n",
      "Round = 13, % of wts pruned = 94.3133%\n",
      "val_accuracy: [60.60000229 64.02999878 65.30000305 66.68999481 67.22999573 68.12999725\n",
      " 67.61999512 68.52999878 69.12000275 68.55999756 68.97000122 68.83999634]\n",
      "\n",
      "\n",
      "Round = 14, % of wts pruned = 95.4262%\n",
      "val_accuracy: [58.27000046 62.91999817 64.51000214 65.90000153 66.72000122 67.27999878\n",
      " 68.01000214 68.30000305 68.34999847 68.98999786 68.98000336]\n",
      "\n",
      "\n",
      "Round = 15, % of wts pruned = 96.3190%\n",
      "val_accuracy: [57.38999939 61.51999664 63.16000366 63.90000153 65.44000244 66.83999634\n",
      " 66.52999878 67.00999451 67.94999695 68.33999634 67.40999603 67.75\n",
      " 68.36000061]\n",
      "\n",
      "\n",
      "Round = 16, % of wts pruned = 97.0351%\n",
      "val_accuracy: [54.15999985 59.59000015 61.58000183 63.48999786 64.12000275 64.98000336\n",
      " 65.65000153 66.15999603 66.22999573 66.08999634 67.09999847 67.25\n",
      " 67.65999603 67.76999664 67.66999817 67.69000244]\n",
      "\n",
      "\n",
      "Round = 17, % of wts pruned = 97.6106%\n",
      "val_accuracy: [51.16000366 56.72000122 59.81000137 61.47999573 61.91999817 62.63999939\n",
      " 63.59000397 64.90000153 65.22000122 65.34999847 65.27000427 66.11000061\n",
      " 66.18999481 66.25       66.76999664 67.15000153 67.29999542 66.93000031\n",
      " 67.22000122]\n",
      "\n",
      "\n",
      "Round = 18, % of wts pruned = 98.0730%\n",
      "val_accuracy: [42.25999832 49.97999954 54.59000015 56.51000214 57.88999939 60.29000092\n",
      " 61.47000122 62.84000397 63.20999908 63.13999939 64.37000275 64.38999939\n",
      " 65.13000488 65.69999695 65.62999725 65.98999786 65.84999847 66.12999725\n",
      " 66.61000061 66.23999786 66.43999481 66.3999939  66.59999847]\n",
      "\n",
      "\n",
      "Round = 19, % of wts pruned = 98.4438%\n",
      "val_accuracy: [39.90999985 46.04999924 52.36999893 54.43000031 57.04000092 58.7899971\n",
      " 60.63999939 61.58000183 61.97999573 63.01000214 63.72000122 64.06000519\n",
      " 64.36000061 64.87000275 64.86000061 65.13999939 65.52999878 65.76000214\n",
      " 65.73999786 65.65000153 65.94000244 66.15999603 66.05999756 66.33999634\n",
      " 66.29999542 66.52999878 66.45999908 66.33999634 66.36000061]\n",
      "\n",
      "\n",
      "Round = 20, % of wts pruned = 98.7416%\n",
      "val_accuracy: [36.86000061 45.10000229 50.69000244 53.67000198 56.04000092 58.84000015\n",
      " 60.11000443 60.90000153 61.72000122 62.68000031 63.47000122 64.20999908\n",
      " 64.68000031 64.8500061  65.02000427 65.02999878 65.24000549 65.72000122\n",
      " 65.86999512 66.19999695 66.12999725 66.02999878 66.19999695 66.44999695\n",
      " 66.36000061 66.33999634]\n",
      "\n",
      "\n",
      "Round = 21, % of wts pruned = 98.9818%\n",
      "val_accuracy: [37.20000076 43.18000031 48.75       51.63999939 54.54000092 56.23999786\n",
      " 57.97999954 58.60000229 60.14999771 60.65000153 61.51999664 62.26000214\n",
      " 62.72000122 63.02999878 63.55999756 63.80000305 63.73999786 62.95999908\n",
      " 63.97999954 64.83000183 64.40000153 64.79000092 64.98000336 65.05000305\n",
      " 65.24000549 65.56999969]\n",
      "\n",
      "\n",
      "Round = 22, % of wts pruned = 99.1745%\n",
      "val_accuracy: [34.67000198 42.34000015 46.79999924 50.37999725 52.99999619 54.94999695\n",
      " 56.20999908 57.93000031 58.74999619 59.82000351 60.65000153 61.02000046\n",
      " 61.58000183 62.41000366 62.59999847 63.02000427 63.38999939 63.63999939\n",
      " 63.95999908 63.63999939 64.06999969 64.36000061 64.27000427 64.38999939\n",
      " 64.81999969 64.88999939 64.72000122 65.16000366 65.13000488 65.38999939\n",
      " 65.24000549 65.65000153 65.41999817 65.62999725 65.54000092]\n",
      "\n",
      "\n",
      "Round = 23, % of wts pruned = 99.3305%\n",
      "val_accuracy: [31.22999954 38.02999878 41.53000259 44.67000198 45.77000046 48.22999954\n",
      " 49.11000061 50.47000122 51.16000366 52.06000137 52.24999619 52.32999802\n",
      " 53.50999832 52.70000076 53.74000168 53.31000137 53.11999893 54.06000137\n",
      " 53.72999954 54.03000259 54.00999832 54.28000259 54.73999786 53.93000412\n",
      " 54.17000198 55.5        56.08000183 55.11000061 56.04000092 55.98000336\n",
      " 55.86000061 56.11000061]\n",
      "\n",
      "\n",
      "Round = 24, % of wts pruned = 99.4558%\n",
      "val_accuracy: [31.77999878 36.55999756 40.04999924 42.5        41.69000244 43.11000061\n",
      " 44.26000214 45.27000046 46.00999832 48.50999832 49.30999756 49.47999954\n",
      " 49.41999817 49.72999954 48.02000046 50.02999878 47.35000229 50.27999878\n",
      " 49.14999771 50.56999969 50.84999847 50.84000397 49.34000015 50.84000397\n",
      " 50.91999817 51.16999817 49.63999939 51.05999756 51.33000183 51.45000076\n",
      " 51.47000122 49.81000137 51.44000244 51.43000031 51.09000397 51.31000137]\n",
      "\n",
      "\n",
      "Round = 25, % of wts pruned = 99.5564%\n",
      "val_accuracy: [29.82999992 33.77000046 36.23999786 39.36999893 40.77999878 42.01000214\n",
      " 42.06999969 43.30999756 44.31000137 44.84000015 45.52999878 45.72999954\n",
      " 48.03000259 46.56999969 48.66999817 48.89999771 49.04999924 48.98999786\n",
      " 48.91999817 49.38000107 49.54000092 49.68999863 49.66999817 50.08999634\n",
      " 48.68999863 50.08999634 50.09999847 50.04000092 50.15000153 50.55999756\n",
      " 50.40000153 50.73000336 50.66000366 50.45999908 50.62999725 50.68000031\n",
      " 50.84000397 50.95000458 50.77999878 50.91999817 50.77000427 51.12000275\n",
      " 50.84000397 51.16000366 50.98999786]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# For cross-checking, tally results received from above with this, manually-\n",
    "for k in history_main.keys():\n",
    "    print(\"\\nRound = {0}, % of wts pruned = {1:.4f}%\".format(k, history_main[k]['percentage_wts_pruned']))\n",
    "    print(\"val_accuracy: {0}\\n\".format(history_main[k]['val_accuracy']))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualizations of training process:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Python 3 dict for training and testing accuracy visualization-\n",
    "# key is 'percentage_wts_pruned' and value is accuracy\n",
    "plot_accuracy = {}\n",
    "plot_test_accuracy = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "# populate 'plot_accuracy'-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['accuracy'])\n",
    "    plot_accuracy[history_main[k]['percentage_wts_pruned']] = history_main[k]['accuracy'][epoch_length - 1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "# populate 'plot_test_accuracy'-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['accuracy'])\n",
    "    plot_test_accuracy[history_main[k]['percentage_wts_pruned']] = history_main[k]['val_accuracy'][epoch_length - 1]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualization of training and testing accuracy VS percentage of weights\n",
    "# pruned-\n",
    "fig=plt.figure(figsize=(10, 9), dpi= 80, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.plot(list(plot_accuracy.keys()), list(plot_accuracy.values()), label = 'training_accuracy')\n",
    "plt.plot(list(plot_test_accuracy.keys()), list(plot_test_accuracy.values()), label = 'testing_accuracy')\n",
    "\n",
    "plt.title(\"Conv-2 CNN: Percentage of weights pruned VS. Accuracy\")\n",
    "plt.xlabel(\"percentage of weights pruned\")\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.legend(loc = 'best')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Python 3 dict for training and testing loss visualization-\n",
    "plot_loss = {}\n",
    "plot_test_loss = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "# populate 'plot_loss'-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_loss[history_main[k]['percentage_wts_pruned']] = history_main[k]['loss'][epoch_length - 1]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "# populate 'plot_test_loss'-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_test_loss[history_main[k]['percentage_wts_pruned']] = history_main[k]['val_loss'][epoch_length - 1]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqMAAAJhCAYAAAB1v7nXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAAMTQAADE0B0s6tTgAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nOzdeViU9fo/8Pew76CsCuKGgoqKihvu+66Zpqa5laZpWZn6q6zUjh1LzdMpsyw7Zkpq+s1c0lJz19w3lEBAATFxD0UBYeb+/fEwAyPDojIL8H5d11wz86z3DANzcz+fRSUiAiIiIiIiM7AydwBEREREVHExGSUiIiIis2EySkRERERmw2SUiIiIiMyGySgRERERmQ2TUSIiIiIyGyajRERERGQ2TEaJiCzcv//9b/j4+EClUmHPnj1GPdf333+PgICAEm+fmJgIlUqF+Ph4I0ZVtsTHx0OlUiExMdHcoRCVCUxGqdxLTEzEmDFjULVqVTg4OKBu3bqYMmUKUlJSzBbT6dOnMWTIEFStWhXOzs5o0qQJ1q9fX+x+9+/fx3vvvYe6devCwcEB1apVw+DBg3Hs2DEAwJ49e6BSqdCmTRu9/R79cvz++++hUqkwfPhwve127twJlUr1WK+lRo0aUKlUUKlU8PDwQMeOHXH06NHHOoY5BAQE4Pvvvzd3GMVKSkrCe++9h2+++QZXr15FRESEUc83dOhQnDp1qlSP+cILL2DMmDGlesyyKjU1FTY2Nti6dWuBdWq1GlWqVMHChQsBAMeOHUOPHj3g6ekJZ2dnhISE4PXXX3+s86lUKuzcubNUYicyFiajVK7FxsYiPDwct27dwtq1a3HhwgWsWLECOTk5+M9//mO2uE6dOoWAgACsXbsWUVFRGDt2LIYNG1Zk1SsjIwOdOnXChg0b8MknnyA2Nhbbtm1Ds2bNMGPGDL1tT548iV9//bXIGOzt7fHTTz8hKirqqV/Pp59+iqtXr+LQoUPw8PBAnz59cOfOnSc6VlZW1lPHU55cunQJIoIBAwbAz88PdnZ2Rj2fo6MjvL29jXoOU7DUz5Gfnx+6d++OH374ocC67du348aNGxgxYgSuX7+O7t27IygoCHv37sWZM2ewYMECqNVqM0RNZGRCVI516dJFWrRoIRqNpsC6O3fu6B7Pnz9fAgICxM7OTlq2bClHjhzRrVu+fLn4+/vLunXrpEaNGuLu7i5jx46VzMxMERGZMmWK9OrVS+/Y165dE2trazl+/HiJY+3evbu8+eabha7/17/+JS4uLpKamlroa9m9e7cAkDfffFPCwsJ0rzsuLk4AyKVLl/Re03PPPScDBgzQHWfHjh3yuH8WqlevLt9++63u+ZUrVwSAbNu2TURE7t+/L5MmTRIvLy9xd3eXPn366OIQERk9erQMHz5c/t//+3/i6ekpffv2FRHlPRwxYoRUqlRJnJ2dJSIiQuLj43X7/fe//5WaNWuKo6OjhIeHy+7du3XrivuZdejQQQDobh06dBARkY0bN0rLli3FxcVFqlSpIq+88oqkp6frjqvRaGT69Oni7u4uXl5eMn/+fGnTpo3MmjVLt83169dl+PDh4u7uLp6enjJ8+HC5efNmke9hYZ+/5cuX68Vp6Gdz8+ZNUalUcvHiRRERSUpKEgDyzjvv6Lbp2bOnzJs377HeO62MjAwZM2aMODk5SUBAgPzwww/i7+8vy5cvFxGRS5cuCQDZsGGDNG/eXJycnKRDhw6SlJQkIiKzZs0y+BoSEhKkR48e4urqKq6urtKiRQuJi4sz+P5oP9e//vqr1KlTRxwcHGTgwIF6v8MdOnSQt956S8aNGyeurq4yefJk3X7Z2dmFvr7Ro0fLiBEjZObMmVKpUiWpUqWKfPrpp3rnT0hIkL59+4qzs7NUqVJFJk+eLPfv39etT05Ols6dO4u9vb00btxY1q5dq/f79qg1a9aIg4ODpKWl6S1//vnndX9LNmzYIK6urgb3fxwAZMeOHQbXZWdny/Tp08Xb21scHByka9eucuHCBd36EydOSJs2bcTJyUk8PDykffv2uvd8+/btEhYWJg4ODuLp6Sm9e/d+6lip4mIySuXWjRs3RKVSyZo1a4rcLjIyUpycnGTVqlUSHR0t48ePF09PT90XxfLly8XBwUH69u0rZ8+elV27dknlypXl888/FxGRQ4cOia2trdy6dUt3zC+//FKCgoIeK96mTZvKv/71r0LXN2zYUCZOnFjkMbRfvqmpqeLm5iY//fSTiBSejP71119ibW2tS34eTUa1xyvsS1WkYDJ6+/ZtASCbNm0SEZGRI0dKt27d5NixYxITEyNjx46V0NBQycnJERElGXB2dpYpU6ZITEyM7suwTZs20qpVKzlw4IBcuHBBVqxYITExMSIi8t1330mtWrVk27ZtkpCQIJ9//rk4Ojrqvb6ifma3bt2SKlWqyGeffSZXr17V/ezWrl0rmzZtkoSEBNm7d6+EhITI9OnTda/t22+/FTc3N1m/fr2cP39eBg8eLK6urnrJaPv27WX48OFy9uxZiYqKkt69e0vPnj0Lff+K+vw9ePBAfvrpJwEgV69elatXrxo8Rv369WXFihUiIrJy5Urx9PSUtm3bioiIWq0WNzc3OXDgQInfu/zJ2syZM8Xf31927Nghp0+flk6dOomDg0OBZLRx48aya9cuOXfunDRv3lyeffZZERG5d++eDBo0SIYMGaL3Gnr16iWjRo2SmJgYiY2NlVWrVklCQoLB16f9HIaHh8uhQ4fkzz//lHr16sno0aN123To0EFcXFxk3rx5EhcXJxcvXixxMurq6iozZsyQ2NhYWbp0qQCQM2fOiIhIVlaWBAUFyZtvvikxMTFy9OhRadGihd7vYseOHSUiIkJOnz4tv//+uwQFBRX5e5ORkSEeHh6ybNky3bK7d++Ko6Oj7u/VoUOHxMrKSvdP3ZMqKhn96KOPxMfHR7Zs2SJRUVHSr18/CQkJ0f1u1qtXT2bMmCEJCQly7tw5Wbp0qdy5c0eys7PFzc1NPvvsM0lMTJQzZ87If/7zn6eKkyo2JqNUbh0+fFgAyKlTp4rcrmXLlnoJR3Z2tgQEBMjixYtFRPnyUqlUehXJl19+WQYNGiQiSrXs0YSsQ4cOMnPmzBLHun79enF0dNRVtwxxcHAo9g9+/i/fWbNm6b5YCktGRZQv427duolIwWT0yJEjEhwcLCkpKYWeM/9rf/DggUyePFmcnJzk6tWrcunSJbGzs5Pbt2/rtn/48KE4OTnJ/v37deevVauWqNVq3Ta7du0SOzu7Qs9bs2ZN2bx5s96ybt266ZL54n5mIqJX3SvM6tWrpWbNmrrn4eHhej/XO3fuiKOjoy4Z3bt3r/j6+uolP9pK8eXLlw2eo7jPX0mq1RMmTJBx48aJiMj48eNl1qxZ4uTkJBkZGXLixAlxcHDQVYVL8t7lT9a8vLz0PtuxsbECoEAyunbtWt02P/74o3h6euqejxgxQi9xFBEJDQ2VH374ocjXpaX9XOdPzHbs2CE2Nja6Sl2HDh2kY8eOBvcrLhmtX7++3n5169aVL774QkREVqxYIc2aNdNbf/DgQbGzs5OcnByJjo4WAPLXX3/p1n/11VfF/hM3fvx4XUVeROR///ufuLu7S0ZGhm7Z1KlTxcrKSnx9fWXgwIGydOlSefDgQaHHNKSoZNTX11e+/PJL3fNbt26Jo6OjbNmyRUREXFxcZN++fQX2u3nzpgCQ5OTkx4qFqDBsM0oVXmxsLFq1aqV7bmNjg/DwcMTGxuqWeXt7w9fXV/fcz88P169fB6B0EBgyZAjWrl0LQOmgsH//fgwdOhQAMHHiRLi4uMDFxQUNGjQocP5Dhw5h7NixWLZsGWrWrFlqr2vq1Km4ceMGVq1aVeR2s2bNwp49e7B3794C61q0aIGYmBj4+/sXeYxXX31V9xo3btyIyMhI+Pn54fz588jOzka1atV06ytVqoSMjAxcvHhRt3/jxo1hZZX35+jcuXOoU6eOwfOmp6fj0qVLGDp0qO6YLi4u2L17t94xi/qZFSY6OhoDBw5EYGAgXF1dMXbsWFy+fFm3Pi4uDs2aNdM99/DwQFBQkO55VFQUbty4AQ8PD11cdevWBQC92PIryeevOO3atcP+/fsBAPv370ffvn0REhKCo0ePYv/+/WjRogXs7e1L/N5p/fPPP7h586bea65bty5cXV0LbNuwYUPdYz8/P9y6davI9o2TJk3CuHHj0KNHDyxcuFDvfS5MixYt9B7n5OQgISFBt6xJkybFHsOQ0NBQvef5PytRUVE4c+aM3vvVrVs3PHz4EFeuXEFsbCxcXV0REhJiMM7CjB49Gvv27UNSUhIA4IcffsCQIUPg4OCg2+bTTz9FSkoKFixYAB8fH7zzzjto3rw57t+//0SvM7+0tDRcu3ZN77NXuXJlBAcH6z57r776Krp3745nnnkGX375JW7evAkA8PT0xLBhwxAaGophw4Zh+fLlSE9Pf+qYqOKyMXcARMZSu3ZtqFQqxMbGIiws7KmOZWtrq/dcpVJBo9Hong8dOhSLFi3C9evXsW7dOgQHB+u+nD/88ENMmzbN4HGOHTuG3r17Y8GCBQV6tj8qKCjosRIUNzc3zJgxA3PmzCmyM1PNmjUxbtw4zJw5E7Nnzy7x8fObNWsWnnvuOXh4eMDLy0u3PD09HY6Ojjh9+nSBfXx8fHSPnZyc9NaJSKHn0n4R//jjjwWS+/xJUnE/M0P69++PRo0aITIyEj4+Pti3bx9efvnlAscpTHp6OoKCggy+38Ul9E+jXbt2iI2Nxfnz53HlyhU0adIE7dq1w759+3Dq1Cm0a9cOQMnfOy3tz6EkIyzkf7+12xf1c3zllVfQo0cPbN68GZs3b8asWbPw22+/6WI1JH8chmJ69HOk/QcnfxzZ2dlFxq49tvazkp6ejvbt22Pp0qUF9qtSpQpE5LFHoACANm3aoHbt2li1ahVGjhyJvXv3Yu7cuQbPMXLkSIwcORLvvPMOgoKCsHbtWrz44ouPfc7HNW/ePIwYMQJbtmzBypUr8cEHH+Dw4cOoU6cOVq9ejSNHjmDr1q1YuHAh5syZgxMnTsDT09PocVH5w8oolVteXl7o1KkTPvvsM4NfimlpaQCA4OBgHD58WLc8JycHx48f16t0FKdZs2aoWbMm/u///g8//fQThg0bplvn4+ODoKAgBAUFoXr16rrlp06dQo8ePfDee+9hwoQJxZ5jyJAhiIyMNFjd076WR7366qvIyMjAsmXLijz2e++9h5MnT2Lbtm3FxmGIt7c3goKC9BJRQKl4PnjwABkZGbr3QHtzc3Mr9HgNGzZEXFwc/v777wLrfHx84Ofnh+Tk5ALHzF8JLY6tra1e5e7mzZtISEjABx98gHbt2iE4OBipqal6+9SpUwcnTpzQPU9LS9MbX7Nx48ZITk6Gm5tbgdgcHR0NxlEan7/AwEAEBgZi3rx5aN26NaytrdG+fXvs27cPBw4c0CV4j/veVapUCV5eXnqvOS4uDvfu3StxbEDB91qrVq1aeP3117Fz50506NABq1evLvI4+YcMO3r0KGxsbFC7du1Ct9eOCpD/5/i4o0c0btwYMTExCAgIKPCe2draIjg4GHfv3tX7R1E71FpxRo0ahZUrV2LVqlWoVatWgSHZHhUYGAhnZ+dSqYy6u7vD19dX77N3+/ZtxMbG6n32QkND8fbbb+Pw4cPw8/PDhg0bdOtatmyJOXPm4NSpU/jnn3/wxx9/PHVcVDExGaVybfHixYiNjUXXrl2xfft2JCYm4siRI3jttdfw4YcfAgBef/11LFmyBD/++CNiYmIwadIkZGRk4IUXXniscw0dOhRffvklDh06pJeMGnLu3Dl069YNzz//PF544QWkpqYiNTW10KQSAN566y3UqVMHrVq1wpo1a3DhwgWcP38eCxYswDPPPGNwHycnJ8ycORNLliwpMp6qVati0qRJBbY7evQoQkJCcOXKlSL3L0xISAieffZZDBs2DL///jsuXbqEffv24bXXXsOtW7cK3a9Tp05o3rw5Bg0ahIMHDyIhIQGRkZGIjY2FSqXCu+++i/fffx/Lly9HQkICjh8/jo8//hi7du0qcWzVq1fHvn37dO97pUqVUKlSJXz77be4ePEi1q5dW6AaNmHCBHzxxRfYsGED/vrrL7z88suwsbHRVca6d++Ohg0b4tlnn8X+/ftx8eJF7Nixo0B1Nb/S+vy1a9cOa9as0SWe7dq1w+7du3Hr1i20bt0aAJ7ovZswYQJmz56NXbt24ezZs5g0aRIcHBweqxpYvXp1nDp1ComJibpLvW+++SZ27tyJxMRE7N+/H2fPnkVwcHCRx3n//fdx5MgRHDlyBK+//jqGDx8ODw+PQrcPCgpC1apVMXv2bMTHx2PVqlX46aefShw3AIwYMQJ2dnYYOnQojh07hvj4eGzevFl3taN+/fpo3749xo8fjzNnzmDnzp1YtGhRiY49atQoXLhwAQsWLMDo0aP11v3yyy8YO3Ysfv/9dyQkJODs2bMYN24csrKy0K1bNwDAhg0bSvRPS0JCAk6fPq13A5TP3pw5c7B161acP38eY8aMQfXq1dGjRw9kZGRgypQp2L9/P5KSkrB161YkJycjODgYly5dwsyZM3HkyBEkJSVh3bp1SE9PR506dR7nrSXKY84Gq0SmkJCQIKNGjRI/Pz+xt7eXoKAgee211/Q6x8yfP1/8/f2LHNopv1mzZkmbNm30lkVFRQkACQsLKzYmQ8PdACjQyeNR9+7dk3feeUeCgoLEzs5O/P39ZdCgQbohpAx12MjKypLq1asX2oFJ68aNG+Lq6vrUvekflZGRIVOnTpWqVauKnZ2d1KxZUyZMmKDriKEdWudR165dk2HDhombm5u4uLhI27Zt9XpbL126VEJCQsTW1lb8/Pxk4MCBut72JfmZ7d69W4KDg8XGxkbXkWTr1q0SFBQkDg4O0qlTJ/nuu+/03g+1Wi3Tpk0TNzc38fLykgULFkjTpk31hk26deuWvPjii+Ll5SUODg4SHBys10HJkKI+fyUdbuvrr78WALJ3717dspCQEGnSpEmBbR/nvcvIyJDRo0eLk5OT+Pv7y8qVK6Vy5cqyevVqEcnrwJR/WKZHP4cpKSnSrl07cXR01L2WyZMnS82aNcXe3l78/f1l+vTpul7cj9Ieb9OmTVK7dm2xt7eXAQMG6HWMK6zT4K5du6RevXri6OgozzzzjCxatMjg0E75PXqsxMREGTx4sLi7u4uTk5M0atRIFi5cqFuflJQkHTt2FDs7OwkNDZXVq1cX+3uj1bFjR1GpVAW2jYuLkxdffFFq164tDg4O4uXlJZ07d5Zdu3bpttEO/VUUQ39ntPs8OrRTly5ddKNZZGVlydChQ3Wfy5o1a8r8+fNFRCQ1NVX69+8vvr6+Ym9vL/Xq1dON5kD0JFQiRTTqISKiQt2/fx9Vq1bFsmXL8Nxzz5k7HJO4fPkyAgMDcfToUTRv3twk59yzZw86deqE7Oxs2NiwqwNRecPfaiKiEkpLS8PKlSvRrVs3ZGZmYu7cubCzs0PPnj3NHZrRJCQk4MCBA2jTpg1u376NGTNmICQkBOHh4eYOjYjKCbYZJSIqIZVKhXXr1qFFixZo3749bt++jd27dxvsiV5eqFQqLFmyBGFhYejduzc8PDywffv2J+pBTkRkCC/TExEREZHZsDJKRERERGbDZJSIiIiIzKbMd2Cyt7fXDWxMRERERJbnxo0byMrKMrjO6MnolClTsGnTJiQlJSEqKqrAHMD5ZWZmomnTpnBycsLx48dLdHxvb2+kpKSUVrhEREREVMoCAgIKXWf0y/SDBw/GgQMH9KZBLMzMmTN1M4UQERERUfln9GS0ffv2RWbDWvv370dcXBxGjhxp7JCIiIiIyEJYRAem+/fv44033sBXX31V7LaLFi1CQECA7paenm6CCImIiIjIGCyiA9P06dMxefJk+Pv7Iy4urshtp06diqlTp+qeF1d11Wg04FCqlk2lUuluREREVLFYRDJ64MABbN26FR9++CEyMzNx584dNGjQAOfPn3/iYz58+BDJycnIzs4uxUjJWFQqFTw8PODj4wMrK4so2BMREZEJWEQyevbsWd3jPXv2YNq0aSXuTV+Y5ORkuLq6wtPTkxW3MiA7OxvXrl1DUlISatasae5wiIiIyESMnoxOnjwZGzduRGpqKrp27QoXFxfEx8dj3Lhx6N+/P/r371/q59RoNMjOzoanpydsbCwi36ZiWFtb65ppaDQaVkeJiIgqiDI/N31AQECBcUbVajUuXLiAunXrwtra2kyR0ePiz42IiKh8MpSvabH8RERERERmw2SUiIiIiMyGyaiJzJ49Gw8fPnzs/Y4fP44RI0aUaNuwsDBkZGQ89jmKs2fPHoSHh5f6cYmIiIgqRO+ecSuOIenWA6Mcu7qnE5aNbl7sdnPmzMG0adNgZ2entzwnJ6fITlbh4eGIjIwsUSynT58u0XZEREREloKVUROYOHEiACAiIgJhYWHo3bs3pkyZgp49e6Jx48YAgBdeeAHh4eFo1KgR+vbti+vXrwPQr0omJibCy8sLH3zwAZo1a4agoCBs3bpVdx6VSqWbkapGjRqYM2cOIiIiULNmTcydO1e3XXR0NFq2bInQ0FAMHz4crVq1wpYtW0r8elauXImGDRuiUaNG6NOnD65cuQIAOHz4MJo1a4awsDCEhobqZtRatmwZ6tevj7CwMDRs2BBHjhx50reSiIiIyhsp4/z9/Qssy8nJkejoaMnJyTFDRIYBkHv37omIyOjRo6VJkya65yIiN27c0D2eN2+eTJ48WUREdu/eLc2aNRMRkUuXLgkA+eWXX0REZNu2bVK3bl2D56hevbq88cYbIiJy/fp1cXNzk5SUFBERadq0qaxcuVJERI4fPy5WVlayefPmQmPPH0NUVJT4+vrqjjV37lzp3bu3iIj0799fIiMjdfvdvn1bRETc3NzkypUrIiLy8OFDvdednyX+3IiIiOjpGcrXtCrEZXpLNGTIELi4uOieR0ZGYuXKlcjKykJGRgb8/PwM7ufs7IwBAwYAAFq3bo2EhIRCz6Fta+rt7Y1atWrh0qVLcHV1xblz5zB8+HAAQLNmzdCoUaMSx71792707dsX/v7+AIBJkyZh7ty5EBF06tQJc+fORXx8PDp37oy2bdsCADp37oxRo0ahX79+6NWrF+rWrVvi8xEREVH5xsv0ZpI/ET1w4AAWL16Mbdu2ISoqCosWLUJmZqbB/RwcHHSPra2toVarCz3Ho9vm5ORARJ5qHnjt/lr5H7/xxhvYsmULqlSpgnfffReTJk0CAPz888/4+OOPkZ2djd69e2PNmjVPdG4iIiIqf5iMmoirqyvS0tIMrrtz5w7c3NxQuXJlPHz4EEuXLjVaHO7u7qhfvz5Wr14NADh16hSioqJKvH+XLl2wdetWpKamAgC+/vprdOnSBSqVCrGxsahVqxbGjx+Pd999F4cPH0ZOTg4SEhIQHh6OadOmYfDgwTh69KhRXhsRERGVPbxMbyJvvfUWOnfuDEdHR1StWlVvXa9evbBq1SqEhIQgICAAERER+P33340Wyw8//ICxY8fi008/RZMmTdC4cWO4u7uXaN8GDRpg3rx56N69OwCgWrVq+OabbwAAX3zxBXbv3g07OztYW1vj008/hVqtxtixY3Hnzh3Y2NjA29sby5cvN9prIyIiorKF04FWQPfv34eTkxNUKhWio6PRsWNHxMbGolKlSmaNiz83IiKi8qmo6UBZGa2ADh48iOnTp0P7f8i3335r9kSUiIiIKiYmoxVQ9+7ddZfZ8+vfvz+Sk5P1llWqVAm7d+82VWhERERUwTAZJZ1NmzaZOwQiIiIyNo0asLKc5nDsTU9ERERUUaizgX/7A5vfMHckOkxGiYiIiCqKf5KBnAzA3qX4bU2EySgRERFRRXH7onJfubZ548iHySgRERFRRXErdxrxyrXMG0c+TEZNZPbs2Xj48GGp7f/BBx9g7dq1pRFaATVq1MC5c+eMcmwiIiIyo9u5yagnK6MVzpw5c54qGX10/w8//BBDhw4tjdCIiIioorh9EbBxAFyrFr+tiVSMoZ1+HAbcuWScY1eqCQxfU+QmEydOBABERETAysoKGzduxNy5c3HmzBlkZmYiIiICX3zxBWxtbTF37lxERkbC3t4eALBx40bMmzdPb//t27djxowZCA8Px6uvvorZs2fjwoULuHfvHhISEuDn54f169fr5rp/9dVXsWfPHnh7e6Nx48a4fv061q9fX6KXFx8fj4kTJ+L69euwsrLC7Nmz8cwzzyAjIwNjxoxBVFQUbG1t4evri+3btyMuLg5jxoxBeno6NBoNBgwYgLlz5z7FG0xERESlJisdcPAArCynHmk5kZRjX3/9NQDg0KFDOH36ND766CO0b98eR48exZkzZ5CTk4PFixfjzp07WLhwIU6ePInTp0/j0KFD8PX1LbC/j49PgXMcOXIEK1asQHR0NHx8fLB06VIAwNKlS5GcnIzo6Gj88ccfOHny5GPFPmLECAwZMgRnz57FunXr8NJLL+Hy5cv47bffcOfOHURHR+PMmTNYs0ZJyBcvXow+ffrgzJkziIqKwtSpU5/mrSMiIqLSpMkGrCyrFmlZ0RhLMZVLU/vll19w+PBhfPrppwCAjIwM2NnZwc3NDXXq1MELL7yA7t27o0+fPggICCjRMXv16oXKlSsDAFq3bo2oqCgAwO7duzFy5EjY2NjAxsYGzz//PPbv31+iY967dw+nT5/GSy+9BACoU6cO2rZtiwMHDqBly5aIiYnBpEmT0KFDB/Tu3RsA0L59e0yfPh33799Hhw4d0LVr18d6b4iIiMiINDkWNeA9wMqoWYgIfvnlF5w+fRqnT59GbGwslixZAmtraxw+fBhvvPEGrl+/jlatWpU4cXRwcNA9tra2Rk5Oju5cKpXqieMEUGB/lUqFWrVqITo6Gj179sTBgwcRGhqKO3fuYNCgQTh48CCCg4OxePFi9O3b94nOTUREREagUQPWtuaOQg+TUYS6ywUAACAASURBVBNxdXVFWloaAGUO+I8//liXMN65cwfx8fG4d+8erl27hnbt2uH9999H27ZtcerUqQL7P45OnTph1apVyMnJQWZm5mP1wHdzc0NYWBhWrFgBAEhISMDBgwfRpk0bpKSkQKVSoX///li4cCFEBJcvX0ZcXBx8fHwwatQozJ8/H4cPH37smImIiMhINDm8TF9RvfXWW+jcuTMcHR2xceNGfPLJJwgLC4OVlRVsbW3xySefwMHBAYMHD8b9+/ehUqlQp04djB49usD+27dvL/F5J06ciDNnzqBBgwYICAhA06ZNkZGRUeL9IyMjMWHCBHz22WdQqVRYtmwZqlWrhm3btuHtt9+GiECj0WDkyJFo1KgR/v3vfyMyMhJ2dnYQEV17VyIiIrIA6mzA1tHcUehRifZabBkVEBCAlJQUvWVqtRoXLlxA3bp1YW1tWe0izOHevXtwdXVFVlYW+vfvj+eeew7jxo0zd1gF8OdGRERkZJ81Ahw9gAn7THpaQ/maFiujFUDXrl2RlZWFzMxMdO3aFWPGjDF3SERERGQOGjUv05PpHTlypMCyZcuWYfHixQWWf/HFF2jXrp0pwiIiIiJT0+QAVpbVgYnJaAU1btw4i7xUT0REREZkgeOMlsve9NqhiMp4c9gKp7ChpIiIiKiUWOA4o5aVGpcSbQ/1W7duwdPTk8lNGZCdnY1r167BwcEBVhY0RRkREVG5wjajphMYGIjk5GTcvn3b3KFQCahUKnh4eBic6pSIiIhKiSbH4ga9L7fJqJ2dHYKCgqDRaHi53sKpVCrdjYiIiIxIbXltRi0rGiPgJV8iIiIiACKAqC2uzSgzNSIiIqKKQKNW7i2sMspklIiIiKgi0OQo90xGiYiIiMjkdMmoZXVgYjJKREREVBFospV7thklIiIiIpNjm1EiIiIiMhu2GSUiIiIis9EmoxY26D2TUSIiIqKKQM02o0RERERkLmwzSkRERERmwzajRERERGQ2TEaJiIiIyGx044wyGSUiIiIiU2ObUSIiIiIyG16mJyIiIiKzYTJKRERERGbDcUaJiIiIyGy0bUY5AxMRERERmRwv0xMRERGR2TAZJSIiIiKz0SWjbDNKRERERKamS0bZZpSIiIiITI2X6YmIiIjIbJiMEhEREZHZsM0oEREREZmNbtB7VkaJiIiIyNQ46D0RERERmQ3bjBIRERGR2bDNKBERERGZjYZtRomIiIjIXLRtRjnoPRERERGZHNuMEhEREZHZsM0oEREREZkNK6NEREREZDZqJqNEREREZC7ayigHvSciIiIik+NleiIiIiIyG3ZgIiIiIiKzYWWUiIiIiMyGySgRERERmY12BiYVL9MTERERkamJdjpQJqNEREREZGoaNQAVoFKZOxI9TEaJiIiIKgLRWFxVFGAySkRERFQxaNQW114UYDJKREREVDGImpVRIiIiIjITVkaJiIiIyGxEDVhZXupneRERERERUenTaFgZJSIiIiIzYZtRIiIiIjIbthklIiIiIrNhZZSIiIiIzIaVUSIiIiIyG/amJyIiIiKzYW96IiIiIjKbitpmdMqUKahRowZUKhXOnTtncJtdu3ahZcuWqF+/PkJDQzFz5kyIiLFDIyIiIqo4Kmqb0cGDB+PAgQOoXr16odtUqlQJq1evRnR0NI4fP469e/di9erVxg6NiIiIqOKw0MqojbFP0L59+2K3adKkie6xg4MDwsLCcPHiRWOGRURERFSxaNSAyvJaaFpcRKmpqVi/fj169+5tcP2iRYsQEBCgu6Wnp5s4QiIiIqIySDQWWRm1qGT07t276NevH2bMmIGmTZsa3Gbq1KlISUnR3VxcXEwcJREREVEZVFHbjJbUvXv30LNnT/Tv3x9Tp041dzhERERE5YuFthm1iGQ0PT0dPXv2RI8ePfD++++bOxwiIiKi8qeiVkYnT56MgIAApKSkoGvXrggKCgIAjBs3Dps2bQIA/Pe//8XRo0exYcMGhIWFISwsDB999JGxQyMiIiKqOCy0MqqSMj6gpzbRJSIiIqIizAsEqoYBozeZ/NRF5WsWcZmeiIiIiIzMQiujTEaJiIiIKoKK2maUiIiIiCwAK6NEREREZDasjBIRERGR2YgasLK81M/yIiIiIiKi0qXRKPesjBIRERGRyYlauWebUSIiIiIyOU1uMsrKKBERERGZHCujRERERGQ2rIwSERERkdnoKqOWl/pZXkREREREVLrYm56IiIiIzIZtRomIiIjIbNhmlIiIiIjMhpVRIiIiIjIbXWXU8lI/y4uIiIiIiEoXK6NEREREZDa5vekF1vjtXCrO/51m5oDyMBklIiIiKu9yK6PZosLEVSfwvwOJ5o0nHyajREREROVdbpvRHFEBABztLCcFtJxIiIiIiMg48lVGAcDBxnLajjIZJSIiIirvdJVRJfVztGMySkRERESmoq2ManIro7ZMRomIiIjIVHJ70+su0zMZJSIiIiKTya2MPsytjDoyGSUiIiIik9HoX6Znb3oiIiIiMh1dZVR5yt70RERERGQ6j1RGHdibnoiIiIhMJrcymsU2o0RERERkcrm96R+qmYwSERERkalJbjKqbTPKZJSIiIiITEZ7mZ6VUSIiIiIyOY02GVWeOnBoJyIiIiIymUc6MPEyPRERERGZziOVUV6mJyIiIiLTye3AlKlWwcZKBVtry0kBLScSIiIiIjKO3MpoptqyLtEDTEaJiIiIyj9tm9EcwMHWstI/y4qGiIiIiEpfbmU0R1SwtlKZORh9TEaJiIiIyrvcymg2rGBjZVnpn2VFQ0RERESlL19l1MJyUSajREREROVebm/6HA0ro0RERERkarmV0WxRwcKajDIZJSIiIir3JO8yPSujRERERGRa2sqoRgUrCyuNMhklIiIiKu/0etMzGSUiIiIiU9LkdmBSszJKRERERKYmeR2YWBklIiIiItPK15veWsVklIiIiIhMKbcy+lDD6UCJiIiIyNR0lVErJqNEREREZGKSN7QTk1EiIiIiMq3c3vRMRomIiIjI9FgZJSIiIiKzyW0z+pC96YmIiIjI5HIro2qxgrU1k1EiIiIiMqXcyqgaVqyMEhEREZGJ5VZGNZybnoiIiIhMLrc3vRpWnJueiIiIiExM8i7TszJKRERERKalybtMz8ooEREREZkWK6NEREREZDa6yqgKVuxNT0REREQmJRqIygqAipVRIiIiIjIxjRpQWQMApwMlIiIiIhMTNYTJKBERERGZhUYNqJS0j8koEREREZmGCHDwcyDhD4gVK6NEREREZEo3YoAd7wMArB/eAwB2YCIiIiIiE7kRo9w7e+Ne1bYAwEHviYiIiMhEblxQ7kesR3zPVQBYGSUiIiIiU9FWRr3qQCMCABz0noiIiIhM5OYFwD0QsHNGjlpJRlkZJSIiIiLj06iBm3GAdzAAQJ1bGWVveiIiIiIyvjuJgDorLxnVaJNRy0r/LCsaIiIiIiodN3M7L3nVBZA/GTVXQIZZWDhEREREVCpuxCr3rIwSERERkcmxMkpEREREZpGTBST/CTj7AE6VAbAySkRERESm8seHwO2LQPOXdIt0vek5zigRERERGU38TuDPxUC1lkC7abrFeZVRJqNEREREZAwiwJY3AXs34NlvAWsb3Somo0RERERkXLcvAv8kA42HAZWq663K0XAGJiIiIiIypqRDyn31NgVWaSujVkxGiYiIiMgodMloRIFVagutjNoUvwkREdFjyLwLXD4KJB0EbsUBdi5K+zUHNwP37vrPbZ0AC+vpS1SmJB0EPOsALj4FVukqoxb2O8ZklIiIns79W0DyISDpT+WLMPUsIJonO5bKuuhktbhk1t5VSX4t7MuWyCTSUoB/koCmow2u1lVGrS3r94PJKBERPZ60K8pg2kkHlUuCN2Ly1jl5AiF9lPZq1SMAnwZATiaQdVepmOru0x55Xsj6tBTlXtQlj09lpSSlxSayboCDu+Hldi6AhQ0MTlSspD+VewPtRQFWRomIqCwSUXrnJh3KvR1UKi9abv5AwyFK4lk9Qpl28NEvOmsXwN4FcKv65DFkPyhhMnsv93Fa3rJ7V5V7TXbJz2llC7SeBHT+QG9oHCKLlnRAuTfQXhTIG/S+wrUZnTJlCjZt2oSkpCRERUUhNDTU4HbfffcdPv74Y2g0GnTp0gVLliyBjQ3/ABARmZRGA9z4Ky/xTDoEpF/LW+8ZBDQZmVf59Ag0/iVxlQqwc1ZuqPJkxxBRKrQlrc5ePQMc/C+QchwY/D/A1a9UXxKRUSQdAtwDAY9qBldb6jijRs/2Bg8ejBkzZqBt27aFbnPp0iW8//77OHXqFHx8fDBgwAB89913mDBhgrHDIyKq2NTZwNWzSuKZ/KfyZZb5T+5KFeDbAKg/QEk8AyMAV1+zhvvEVCrA1lG5leQ15DwEdrwPHPkaWNoeGLwcqGH40ieRRUi/Ady8ADQaVugmFTYZbd++fbHbrF+/HgMHDoSvr/IHYuLEiZg/fz6TUSKi0padCVw5kVf5vHwUyL6vrFNZA1Wb5F5ybwMEtgQcK5k3XnOxsQN6fQJUawFsfA1Y0Q/oOhuIeI2do8gyJRc+pJNWTkVNRksiOTkZ1avnzRJQo0YNJCcnmzEiIqJyIusecPlIXpvPKycA9UNlnY0DENAcCGytfIEFNFfadlKe0EGAbyiwdqRSKU05Cgz4Uun4RGRJEnYr94V0XgIADZPRoqny/acpuQ1sDVm0aBEWLVqke56enm7UuIiIypT7t/Iutz86zJKdK1CzfV7ls2oTwMbevPGWBd7BwPhdwOYpwLn/A66dB4asBPwM94EgMjl1NhC9EfCuB3jWLnQzXWXUwqr7FpGMBgYGIjExUfc8KSkJgYGBBredOnUqpk6dqnseEBBg7PCIiCzX3b/zdTb6U+l8pOVYGQjundfZyDeUPcOflL0LMOg7oFpL4Pd3gWVdgX6fKfN/E5lbwm4g47YyAkQRiaZGWBkt1KBBg9C2bVt88MEH8PHxwddff41hw/gLTkSkRzvMUv7K553EvPWuVYGGz+V1NvIOZvvG0qRSAS0nKBXln0YDGyYAyYeBnh8Dtg7mjo4slUYNWFkb9xxR65T70MFFbpajrqDJ6OTJk7Fx40akpqaia9eucHFxQXx8PMaNG4f+/fujf//+qFWrFubMmYM2bdpAo9Ggc+fOeOmll4wdGhGRZdMbZin3lp6at75ybaDJC/mGWarO5NMUqrUAJu4H1r8InFgOXD0NPLcCqFS9+H2pYrl2HvjhGaD9NOUfGWN4+ACI+RXwDwcq1yxyU42FjjOqkqIaaJYBAQEBSElJMXcYRERPT50DpJ7RTz51wyxBmc1IO7h89QiOfWluGjWw+9/A/oXKqAPPfgvU6WbuqMhS5GQB33YGrp1TZgN744xxRqc49zOwfizQ8xOg1cQiN33vlyisOpyM4+91hZeLaduLF5WvWcRleiKiCqnYYZbC8jobVWsJOFU2b7ykz8oa6PK+Uin9+WUg8jmg/XSg49vGvyxLlm/3R0oiGtBCGYXhzyVA55mlf56o9coUuA0GFrupOrcvo6VVRpmMEhGZ0s04IGYLcGE7cOV43jBL1vbK0ErVI4DqrZUvMA6zVDbU7QFM2Av8NArYNx9IOaZ0dnL2NHdkZC6JB4GDnwP+zYAxW4Cv2gCHvwJavVK6/1Rm/APE7wBqtCvRZA5qjZKNWjEZJSKqQDQa4O9TSgIa8ytwM1ZZbuusfIFoK5/+TTnMUllWqQbw4nZg2wzg5Apl1qYhK4CAcHNHRqaWeRfYMFEZx3fgN8rvdce3gf97CTj0uTJ5Qmn5a7PyD23D50q0uXZoJ1ZGiYjKO3U2kLhfST5jtgL3/laWO/sAzcYAIX2V8T6ZfJYvtg5A/8+VJhW/TgX+1xPoOQ9oPo4dyyqS394G0pKBPp8CXkHKsgbPAvsWAkeWAq0mAy7eT3bsPR8DCbuUySyy7gH3bwDWdkC9fiXaXTvovZWFfR6ZjBIRlYasdCDhD+CvLUDc70BmmrK8Uk1lCsmQfkqVjG0Jy78mIwC/hspl+63TlBmw+v0XsHM2d2RkbNGbgNORQFBXIDzfqEBWVkCnd5TPxMHPgB4fPf6xr5wA9sxTOkO5+ADO3kDlWkBwL8DRo0SHyM1FK97QTkRE5db9m0DsNuUSfMJuQJ2lLK8SplQ/6/UFvENYFauIqjQCXt4D/PKKMgZkapQya5N3XXNHRsZy7xqw+XWlx/yALwv+3of0A3wbAseWKf+gPu5oGLvmAlABL/0O+NR7ohDVwsooEVHZdydRufz+1xbg8mFlqk2VNVCjjZKABvcGPKqZO0qyBI4ewNBI4NB/gT8+BL7tBAxYXKJez1TGiACbXlVmQRryg+FE08oK6PQusOZ54MB/gF6flPz4iQeVy/MNhzxxIqqEqU1Gn/gQRsFklIioKCJKVSvmV+V2LUpZbuOoJJ4hfZXe1Bx2iQyxsgLavqkMSL7+RWDdGGUIr24fAta25o6OSsuJ5UDcdqDx80D9AYVvF9xLmcHr+P+AiCmAu3/xxxZRqqIqa6Uj1FPQaJSCrYqVUSIiC6dRK9M8xmxRbv8kK8sdKwNhI4CQPkCtToCdk3njpLKjZjtgwj5lcPLDS4ArJ4HnlgNuVc0dGT2tWwnA7zMB92rFVztVKqDTTCByMLD/U6DvouKPn/AHkHwIaDoK8Kz9VKFqRCzuEj3AZJSISJGdAVzco1x+v7ANeHBLWe5eDWj5ipKABrYGrPlnk56QWxVg9GZg52zgz8XK8E+D/6eMrEBlkzoH2DBB+fsx8GvAwb34fYK6KmMKn/wBaPsG4BFY+Lbaqqi1HdB+xlOHqySjT32YUse/qkRUcWXcUQafj9kMxP8BZD9Qlvs0UIbjCekD+DViByQqPda2Sk/qgObAxleBHwYAnd8H2ryhXNKnsuXAf5RJDiJeA2q0Ldk+2uroymeAfQuA/l8Uvm3Mr8o4xS0nlkpbdI1YXuclgMkoEVU0aVeA2K3K5ffEA4AmB4AKCGyltP8M6a0Ml0JkTA2eAXwbAGtHAn/MURKaZ74q8RA9ZAGunAT2fqz889r5/cfbt1ZHIDACOBWptCk29DdHo1amFLVxBNpOLY2IeZmeiMgsRIAbsXkzIP19UllubQ/U7qIMv1S315MPQk30pLzqAOP/ALa8CZxdC3zTQRn+qUojc0dGxclKB35+WZkT/tlvHn8CC5VKmaf++z7A3gXAwK8KbnPuZ+B6tFI1L8FUnyWhVEZL5VCliskolR4RZZgbjRoQtVJx0qhzl+XkW567TrutJiff8nzb6I6heeR4BrbVHaO4bXOUNj3u1QA3f8A9AHCtwnaA5Y1GowwQHbNZSUBvxSvL7d2VafNC+ijttuxdzRsnkZ0zMHApUK0FsO1t4LtuQO+FQNOR5o6MCiOizLB1Kw7o/hHgF/pkx6nRVmkvfHYNkHUXeJieO7NS7v2Dm4C9G9Dm9VILXaNhZbTsy7oHZGcaToDyJ1SGEiDd40e2LSwpK5DA5U/KSjOBe5pzG3jdZZHKCnCtqgyx4R6Qe6um3GsTVsdKbDdo6XIeAon7lA5IsVuB9GvKchc/ZSaUkD7KXPA2duaNk+hRKpXSRrlKE2DdaGW8ystHgN4LAFtHc0dHjzr5g1LJrtsLaD356Y7VZRawvLfyT7O9q3Kzc1FGWfCuCzQZWarDxmlEYGWBpVEmo4/j5wlA7K/mjuIJqZRpCK1slLHKrHJvqvz3NkoDeu1ja1tA5aAs09vPRkng9Pazzl1mo39cvcfaY1g9sl9h2xo6RzHb5o+zwLa522WmAWmXgbQU5XY39/72ReULwBBb59wk1d9wsurmr8xLTaaVdQ+I26Fcgo/boVQXAMCzDhA2XGkDWrUpO4ZQ2RDQDHh5L/DzOODUSuDqGWUA9co1zR0ZaaWeA7bNANwDgWeWPH2RIiAcePdK3nejkbE3fXlQp6syNEeBxO2RRKu4hEj3uKRJ3uMkYI8mhNrHFvjpM6eAZoaXZ2cCd6/kJar5k9W0FGXsSW2P60c5+xSSrOY+dvZmUlQa0q/ndkD6VRmKSf1QWe4frlQ/Q/pyykUqu5w9gRHrgb3zgb2fKO1IBy5VBksn88q6p1SuNWpljNjSqliacPID9qYvD8JfNHcEZGy2DsqgwoUNLCyiDAeUP1lNu6yfwF49ozR1eJS1nXLpRZucPpqsuvuzDWNhbiXkzYB0+QgAUf7RqtFO6YAU3JuDh1P5YWUNdHpHGf7p53HA6mFAu7eU4YCsrM0dneW6fRG4GQ/U7lz6/QBEgM1vKO3Pe8xTKpplkEYjFjf7EsBklOjxqFTKf8NOlQvv8arOAe5dLTxZTT0LJO43vK+DR752q48mqxWos5WIktRre8Bfj1aW2zoD9fsDIf2AOt04DA6Vb3W6KrM2/TRama0n5ZiSkHqH8LOvpdEoc7YfXao01YEoSfyz35TuEG0nlgPn1itXXlq9UnrHNTGNCKwt8AJdBfhWIzIxaxtlcOKiBijOuqeMd2koWU27rAzArskuuJ/KSklI9RLWAP3nZbWzlTpHmfLur9wE9G6KstzJS2nEX68fULMD2+ZSxeIRCLz4G/DbO8Dx74BL+5TlrlUB72DAp56SnPrUU56XZAag8iAzDTj9I3D0W+B2gvK3MaQP4OoHHFsGfN0O6Pkx0OSFp/97ePWsMtKBRyAwYHHZ/Puai5fpiSiPvSvgE6LcDNFogPs38pLTtJTchPVyCTpbORWdrFpSZ6uHD5SqRsyvyhScGXeU5R7VgdavKlWIai14aZIqNht7ZQ7z0EFAylHgegxw4y+lDfvF3frbuvkryal37t8Xb22S6mae2Evb9Rjg6DfAmTVA9n3ln+82bwDNX8qbVrNeP2DDK8qoBBd+A/p9rrTFfRKZd5V2oqIBnvteOV8ZJhz0nohKzMpKGeTY1bfkna3yJ6vFdrbyLiRZraa0XXX2MV5nqwe3lS+ImF+VCnBOhrLcr5EyB3y9voBP/TJdfSAyihptlJuWRg38k5SXnN6IBa7/BSQdBBL+0N/XLSA3OQ3Jq6Z6B5eNduoaNRC7TUlCL+1Vlvk1AlpOUBL0R4e/qtURmHRImUzg/AalecOAJUqzh8chAmyeovzz32s+4F/I3+IyRC1ikX9amYwSlVVP0tkq/8gARXW2srLNHRkg3+QA+UcJeNzOVv9czu2AtAVIOqSMSauyAqq3ye0B3yevqkFEJWNlrbSLrFxLmcZWS6MG7iQCN2KU5PRGjJKwXtoPxO/UP4Z7Nf0qqk8I4BUM2LuY9KUY9OC2Mqbnse+AtGSl02LoIKDFBOWKSVFZlWMlYPByZSzQrdOAyEFAi5eBbh+WfOzW498pyWy9/sq+5YBGA1hb4NhOKhERcwfxNAICApCSkmLuMIjKpkc7Wz2arKalAJn/GN5XO5NV/vFW8yerWfeAmK3KLEhXzyj72DgoU3CG9AHq9nzyS2dE9Pi0Ser1v5RK6vUYJVG9eSFviDQtj8C8y/3ewUpbTGdv5ebkZdzJI66eVTokRa0HcjKVKzXhLwLNxijDKz6uf5KBDROVirFXsNK5qWpY0fv8fVqZDcutqtKJrJy0xe352T5k5Wiwe1pHk5+7qHyNySgRFS1/ZytDyerdKwW/yPJz8FDGSAzpowy5YudsutiJqHjqHODOpdwkNTYvUb0VV/jvtoN7XnLq7JXvsYHnDh7FN/tRZwN/bQKOfANcPqwsC2iuVEHrD3j65FejBg59Duz6SHneeSYQMcVwe/TMNGBpe+Du38BL24GqTZ7u3Bakx3/2IVujwa63Opr83EXla7xMT0RFe5zOVvmTVaiAuj2A6hEmHdSZiB6TtQ3gVUe55afOUdpL3opTJpt4cBO4f1P5fb9/Q3l8K17pTGmouY+WyjpfgupVMGm9+zdwfDmQngpY2wONhwMtxgP+TUvvNVpZA23fBGp1An5+Gdg5G7iwHRj4NVCpet52IsDGV5UKcu+F5SoRBZQ2o+zARETlT/7OVij7DfyJKJe1jTKbWXEzmmnUSvt0XZJ6o2DSqn185WTetL35ufkDXT4Amo5WklRjqRoGTNgL7PhA6RD1dVsl6Ww0RGmDevRbpUJb/xmg+TjjxWEmGhHYWuBMgExGiYiI6MlZaSufXgDqFb99dmZulTU3UVVZKWMIm2pCD1tHoPcCoE53YONkYMPLyggfTUcCv78LVKoJ9P+iXI7oIWKZL4vJKBEREZmOrUPeCB3mVKcb8MqfyvBN539Wbtb2wJAV5Wdc1kdoLPQyveXVaomIiIhMwdkTGLoK6L9YmdWq7yKgSmNzR2U0ao0YbQjpp8HKKBEREVVcKpVyib7pSHNHYnRiodOBWmB+TERERESljZfpiYiIiMhslGTU3FEUxGSUiIiIqAJQa3iZnoiIiIjMRHiZnoiIiIjMRSOW2ZveAkMiIiIiotKmYW96IiIiIjIX9qYnIiIiIrPRaMQipwNlMkpERERUAWgEsLbAsZ2YjBIRERFVALxMT0RERERmo0wHau4oCmIySkRERFQBqEWgYmWUiIiIiMxBIwJrJqNEREREZGoiolymt8DMzwJDIiIiIqLSJKLc8zI9EREREZmcOjcbZW96IiIiIjI5TW4yam15uSiTUSIiIqLyTnuZnpVRIiIiIjI5bWWUbUaJiIiIyOTUGm2bUTMHYgCTUSIiIqJyLjcX5dz0RERERGR6wsv0RERERGQuGl0HJvPGYQiTUSIiIqJyLq/NqOVlo0xGiYiIiMo57WV6thklIiIiIpPT6KYDNW8chpQ4GV26dCnS0tIAAJMnT0Z4eDj27dtntMCIiIiIqHRoysN0oF9++SXc3d1x8OBBnDt3Dh999BGmTZtmzNiIiIiIqBSUi3FGbWxsAAC7du3CqFGj0KNHD+Tk5BgtMCIiJRXl7gAAIABJREFUIiIqHbrpQC0wGy1xMmplZYU1a9Zg7dq16NKlCwDg4cOHRguMiIiIiEpHubhMv3jxYqxZswbjx49HjRo1cOHCBXTq1MmYsRERERFRKchLRs0ciAE2Jd2wVatW+OWXXwAowwNUqVIFX3zxhdECIyIiIqLSUS4qoy+99BL++ecfPHz4EGFhYfD19cWSJUuMGRsRERERlYK8GZjKcDJ64sQJeHh44Pfff0eTJk2QmpqKpUuXGjM2IiIiIioF5aIyqh25f9++fejbty/c3NxgZcUx84mIiIgsnUaj3Ftim9ESZ5N+fn6YOHEi1q1bh65duyI7OxtqtdqYsRERERFRKdBVRi0wGy1xMhoZGYmQkBCsWbMGHh4euHLlCqZOnWrM2IiIiIioFJSLy/ReXl6YMGECVCoVjh49Cl9fX4wZM8aIoRERERFRacjrwGTeOAwp8dBOhw4dwuDBg+Hr6wsRwY0bN7B+/Xq0bt3amPERERER0VOy5MpoiZPRqVOnYt26dWjTpg0AJTl98803cfjwYaMFR0RERERPT5NbGrXAXLTkl+kzMzN1iSgAREREICMjwyhBEREREVHpKRfjjDo5OWHnzp2653v27IGzs7NRgiIiIiKi0qO9TG9tgY1GS3yZ/vPPP8egQYNgb28PlUqFrKwsREZGGjM2IiIiIioF5WJu+vDwcMTHxyM2NhYiguDgYAQFBSE5OdmY8RERERHRU9IOeq+ywMv0JU5GAcDW1hahoaG659pZmYiIiIjIcllyb/qnms/TErNrIiIiItKX12bUzIEYUGxlNDo6utB1OTk5pRoMEREREZU+7cVsSywkFpuM9unTp9B1Dg4OpRoMEREREZU+S75MX2wyeunSJVPEQURERERGotZYbm96C2w5QERERESlSTvovSWOM8pklIiIiKic046AZIltRpmMEhEREZVzedOBmjcOQ5iMEhEREZVzagvuwMRklIiIiKicEyajRERERGQuljw3PZNRIiIionJOOzc9K6NEREREZHK6NqMWmPlZYEhEREREVJrYZpSIiIiIzCZvaKcKmIzGxcUhIiICdevWRYsWLRAdHV1gGxHB9OnT0aBBAzRq1AidOnVCfHy8sUMjIiIiqhAseW56oyejEyZMwMsvv4wLFy5gxowZeOmllwpss2nTJuzbtw+nT5/G2bNn0aVLF7z77rvGDo2IiIioQtBU1Lnpr1+/jpMnT+KFF14AAAwaNAiXLl1CYmJigW2zsrKQmZkJEcHdu3cREBBgzNCIiIiIKgzdZXoLzEZtjHnwy5cvo2rVqrCxUU6jUqkQGBiI5ORk1KhRQ7ddv379sGfPHvj5+cHV1RX+/v7Yu3evwWMuWrQIi/5/e/cdH1WVtwH8mfROCkkgnYQUAqQCIRA6SBGzKogFBERUFl1fF0FRd1nWAmvjddV1AV2pyqs0sayyohBgBQkloSaEFNJIbzNp0877R5IhgYRmJnfK8/18+JA7c+fOb44wPpxzzzlr1uiOFQqFPj8CERERkdEz62F62TUfum02V3snT55ERkYGioqKUFxcjAkTJuCZZ57p9HpLlixBYWGh7peTk5Ne6iYiIiIyFWa7N72/vz8KCwuhVqsBtATRgoICBAQEdDhv48aNGDduHFxdXWFhYYF58+Zh//79+iyNiIiIyGy03TN6bSehIdBrGPXy8kJsbCy2bt0KANi5cyeCgoI6DNEDQHBwMH766SeoVCoAwDfffINBgwbpszQiIiIis9E2TG9pgF2jer1nFADWrVuH+fPnY9WqVXBxccGmTZsAAAsXLkRycjKSk5Px9NNP48KFCxg8eDBsbGzQt29frFu3Tt+lEREREZkFQx6m13sYDQ8Px5EjR657/JNPPtH9bGtri48//ljfpRARERGZJbOewERERERE0rp6z6jEhXSCYZSIiIjIxLUN0xviPaMMo0REREQmjsP0RERERCQZIcx0O1AiIiIikp5GmOk6o0REREQkPd09owyjRERERNTTeM8oEREREUmmNYtyaSciIiIi6nma1nF6CwOcwcQwSkRERGTidHvTG2DXKMMoERERkYkTBrw3PcMoERERkYnTcmknIiIiIpLCzxmlKK1rAmCYPaNWUhdARERERPpRVteEBRuP6465tBMRERER9ZjaRlWHY86mJyIiIqIeU6/UdDg2wCzKMEpERERkqhqU6g7HHKYnIiIioh7TeF3PKMMoEREREfWQ64bpDTD5GWBJRERERNQdGjlMT0RERERSqW/mMD0RERERSaRRxdn0RERERCSRa2fTcztQIiIiIuox1w7TGyKGUSIiIiITde3SToaIYZSIiIjIRNVfM0xviBhGiYiIiEwUe0aJiIiISDINSg2cbK2kLuOGGEaJiIiITFSDUg1nO4ZRIiIiIpJAg1IDBxtLqcu4IYZRIiIiIhPVEkYNu2fUsKsjIiIiojvWoFTDz80emxcMQ7m8WepyOsUwSkRERGSi2obpR4d5Sl1KlzhMT0RERGSCNFqBZrXW4IfpGUaJiIiITFDbvvScwEREREREPa5twXuGUSIiIiLqcfVtYZSL3hMRERFRT9MN01uzZ5SIiIiIelhDa8+oPYfpiYiIiKintYVRRw7TExEREVFPa+RseiIiIiKSSn1z22x69owSERERUQ9rUHFpJyIiIiKSSENzyzA9JzARERERUY/TTWDiMD0RERER9bRGDtMTERERkVTqmzmbnoiIiIh6wPniOrz5Qwa0WqF77Ore9BymJyIiIiI92vhLLv55IBuXyhW6xxqUGshkgJ21Ycc9w66OiIiIiG4qs7QlhJbVNeseq1eq4WBtCZlMJlVZt4RhlIiIiMiIabUCWaVyAEBpXZPu8UalBvYGPkQPMIwSERERGbWimkbdMk6l8qthVNGshqOtYU9eAhhGiYiIiIzaxdZeUaDjMH1lvRIejjZSlHRbGEaJiIiIjFhmuzDaNkyv0QpUKprh6WwrVVm3jGGUiIiIyIhltU5ecrK10oXRqnoltAJGEUYN/65WIiIiIupSZokcPr3s4O5kg9LWYfpyecvvXs52UpZ2S9gzSkRERGSkNFqBS+UKhPVxhrezHcrkTRBCoFzREkbZM0pEREREenO5sh5KtRZh3s6QN6mh0ghUN6h0PaOeToYfRtkzSkRERGSk2mbSh3k7w9ulJXiW1jVdDaPsGSUiIiIifcksaZm8FO7tDJVGC4BhlIiIiIh6yMUyOWQyoL+XE8oVLTPpy+qadfeMejhxnVEiIiIi0pOLJXIEuDvA3sZSN3O+pWe0Ca4O1rC14g5MRERERKQHSrUWuRX1CPN2BgB4u7SGUXnLML0xTF4CGEaJiIiIjFJuRT3UWoEwbycAgIejDSwtZCita24Jo0ZwvyjAMEpERERklDLbzaQHAAsLGbycbVFQ1YC6JjXDKBERERHpz8WSjmEUALxc7JBV1jLDnsP0RERERKQ3F0vlsLSQIdjTUfeYt7MtNFoBwDiWdQIYRomIiIiMUlaZAkEeDh1mzLdNYgIYRomIiIhIT5pUGlyurO8wRA9AtwsTwDBKRERERHqSW1EPrQBCvZw6PO7FnlEiIiIi0re2SUoh14TRDsP0nMBERERERPpwqXVZp1CvzofpLS1kcHMw/K1AAYZRIiIiIqOTVaaAhQwdZtIDgHfrlqC9nWxgYSGTorTbxjBKREREZGSyyhQIcHeAnXXHveddHaxhY2lhNPeLAoCV1AUQERER0a1TqrXIq6jH2HCv656TyWSYmxgIPzd7CSq7MwyjREREREbkcmXLnvSh3k6dPv+n6ZE9XNFvw2F6IiIiIiPSNpP+2mWdjBXDKBEREZERuaQLo843OdM4MIwSERERGZGra4w63uRM48AwSkRERGREskrl8HOzh4ONaUz9YRglIiIiMhJqjRY5FfUmc78owDBKREREZDQKqhuhVGsR6m0a94sCDKNERERERiOrdRvQ/p7sGSUiIiKiHtY2eal/F2uMGiOGUSIiIiIj0basU3/eM0pEREREPS2rTI4+LnZwsbOWupRuwzBKREREZAS0WoFLZYoutwE1VgyjREREREagqKYRTSqtSQ3RAwyjREREREbB1LYBbaP3MJqVlYURI0YgLCwMw4YNw/nz5zs978yZMxg7diwGDBiA8PBw7Nq1S9+lERERERmNrLKWZZ1MbZhe7/tIPfXUU3jyyScxf/587NixA48//jiOHDnS4ZyGhgbce++92LRpE5KSkqBWq1FdXa3v0oiIiIiMRlZp60x6E1pjFNBzz2hZWRlOnjyJOXPmAABmzJiB3Nxc5OXldTjv888/R2JiIpKSkgAAVlZW8PT01GdpREREREYlq0yB3k62cHO0kbqUbqXXMFpQUAAfHx9YWbV0wMpkMgQEBCA/P7/DeefPn4ednR2mT5+OmJgYzJ07F+Xl5fosjYiIiMhoXLhSh/TCGsT495K6lG6n93tGZTJZh2MhxHXnqFQq7N27F+vWrcOpU6fg7++Pp59+utPrrVmzBn5+frpfCoVCL3UTERERGYrV32dACOC5iWFSl9Lt9BpG/f39UVhYCLVaDaAliBYUFCAgIKDDeYGBgRg3bhx8fX0hk8kwe/ZsHDt2rNNrLlmyBIWFhbpfTk6mdd8EERERUXspF8tx8GI57o/1xSBf9ozeFi8vL8TGxmLr1q0AgJ07dyIoKAhBQUEdzps1axZSU1NRV1cHAPjhhx8QHR2tz9KIiIiIDJ5GK7DquwuwtbLA0snhUpejF3qfTb9u3TrMnz8fq1atgouLCzZt2gQAWLhwIZKTk5GcnIyAgAC89NJLSExMhJWVFXx9fbF+/Xp9l0ZERERk0HacKEBmqRxPjwuBj6u91OXohUx0dhOnEfHz80NhYaHUZRARERF1q/pmNca+cwBarcCBZWPhbMT70d8or3EHJiIiIiIDtP5gDsrlzXhuUphRB9GbYRglIiIiMjCldU1YfzAHIZ6OeHiov9Tl6BXDKBEREZGBWfOfi2hUafDytAGwsjTtuGban46IiIjIyFQqmrHjZCGG9XPH+AgvqcvRO4ZRIiIiIgPy7ekr0GgFZicEXLd5kCliGCUiIiIyILtPFcHBxhKTIr2lLqVHMIwSERERGYi8inqkFdRg8sA+cLDR+3LwBoFhlIiIiMhAfJVWBAC4N9ZX4kp6DsMoERERkQEQQmBPWjF6O9lgZIiH1OX0GIZRIiIiIgOQXliL3Ip63BPtY/LLObVnPp+UiIiIyIB9dap1iD7GfIboAYZRIiIiIsmpNVp8e7oYwb0dEeXXS+pyehTDKBEREZHEDl+qQIVCid/F+JrF2qLtMYwSERERSUw3RB/rI3ElPY9hlIiIiEhC9c1q7D1XitgAVwR6OEpdTo9jGCUiIiKS0I/nS9Go0uA+M1pbtD2GUSIiIiIJfZVWBEsLGe4e3FfqUiTBMEpEREQkkQpFMw5lVWBMmCc8nGylLkcSDKNEREREEvk2vRgarcDvYsxv4lIbhlEiIiIiiexOK4ajjSXuiuwjdSmSYRglIiIikkBeRT3SC2oweWAf2NtYSl2OZBhGiYiIiCRw/HI1AGBipLfElUiLYZSIiIhIApkldQCAiD7OElciLYZRIiIiIglklMhhZ21hlgvdt8cwSkRERCSBzBI5Qr2cYWlhXnvRX4thlIiIiKiHVdcrUSZvRriZD9EDDKNEREREPS6jRA6A94sCDKNEREREPa5t8hJ7RhlGiYiIiHpcZmlLzyjDKMMoERERUY/LKJHD3dEGnma6H317DKNEREREPUirFbhYIke4tzNkMvOeSQ8wjBIRERH1qKKaRtQrNRyib8UwSkRERNSDOJO+I4ZRIiIiMmtqjRb/PnMFdU2qHnk/zqTviGGUiIiIzNo7/7mIxZ+dxNIv03vk/dp6RsO8GUYBhlEiIiIyYz+eL8XalGxYWsjwn/Ol+OlCqd7fM7NEjgB3BzjaWun9vYwBwygRERGZpcuV9VjyZRrcHKyxfVEiHG0s8Zevz6FRqdHbezarNcipqOcQfTsMo0RERGR2mlQaLNp6EopmNf7+UCziAtyw5K5wFFY34oOfs/T2vtll9dBoBScvtcMwSkRERGZnxZ6zuHClDv8zIRSjwzwBAPMSAxHZ1wUfH8rBpTK5Xt43s5STl67FMEpERERm5cvUAnx5vBCjwzzx7PhQ3eNWlhZ4/b5BUGsF/vTVWQghuv29uazT9RhGiYiIyGycK67Fn/echU8vO7z3YAwsLDrugBQX4IaHhgbgaE4Vdp8q6vb3zyyRw8bKAkEejt1+bWPFMEpERERmobZRhd9vPQmtEPhoTjzcHW06Pe/FKeHwcLTBG99dQG1D9649mlkiR39PJ1hZMoK1YUsQERGRyRNCYOn2dORXNeDP0yMR4+/a5bmuDjZ4adoAVNYr8dbejG6robZBhSu1TRyivwbDKBEREZm8dQdz8OP5UiRH++DR4YE3PX9GnC+G9XPH58fykVZQ0y01ZJa23C/KyUsdMYwSERGRSTuaU4m3fshAqJcTVt8/GDKZ7KavkclkeP3eQbCUyfDK7jNQa7S/uQ5uA9o5hlEiIiIyWWV1TXjm81Owt7bEP+fE39auR2Hezlg4Khjniuuw9ejl31zL1Zn0Lr/5WqaEYZSIiIhMklqjxTPbTqFC0Yw3Z0ahv5fTbV/j2Qn94etqj3f/cxFldU2/qZ7MEjl62VvD28X2N13H1DCMEhERkUl6e28mjuVWYf6IIEyP8rmjazjYWGFl8kDIm9V47bsLd1yLEAKZpXKE93G+pdsEzAnDKBEREZmcH86WYN3BHMQFuOLlaQN+07UmRXpj4gBvfJNejENZ5Xd0jeLaJsib1JxJ3wmGUSIiIjIpuRX1WLY9He6ONvjH7DjYWP32uLMyORL21pZYseccmlSa2349Jy91jWGUiIiITEajUoPfbz0BhVKNvz8Ug7697Lvlun5uDnh2QihyK+qxLiXntl/PbUC7xjBKREREJkEIgT/vOYuMEjn+ODEMo0I9u/X6jyf1Q6iXE/5x4BLyKupv67WZrWE0zJth9FoMo0RERGQSvkgtwI4ThRgb7olnxvXv9uvbWFng9XsHQanWYsXX5yCEuOXXZpbI4etqD2c7626vy9gxjBIREZHRO55XhRVfn4Ovqz3+d1YMLCz0M2M9IdgDM+L8cPBiOb47c+WWXqPSaJFdruAQfRcYRomIiMhoCSHw6eFcPPzxUcgAfDQ7Dm6ONnp9z5enRcDVwRov7TqDrNYtPm8kp7weKo3g5KUuMIwSUbeTN6lQLm+WugwiMnF1TSos/uwkXv32PPzcHLB78UhE+7vq/X09nGzxj0fi0KjU4LGNqahQ3Pj7LoMz6W+IYZSIuk1JbRNW/fsCElf/jIRV+7D6+wt3tAQKEdHNnCuuRfIHh/H92RJMG9wHXz8zEpE+PbfN5sj+vfHGfYNQWN2IJzYfv+F3XSa3Ab2hW9+glYioC5fKFFh/MBu7TxVBpREI9XKCrbUF1qXkYN/5UrzzQDRiA9ykLpOITIAQAl+kFugmEK28JxLzRgRJsqvRg0MDkFvRgLUp2Xh+ezo+eCi203tVM0vksLaUIdjTscdrNAYMo0R0x07mV2PtgWz8eKEUQgBDAt2waEwIxkd4QSsE1h3Mwd/3ZWHGP3/BE6OC8cdJYbCztpS6bCIyUg1KNf60+yx2nSqCr6s9PnwkVvJ/6L4wORyXK+vx3ekr6OfhiKWTw687J6NEjhBPJ1hbckC6MwyjRHRbhBA4kFmOf6Zk41huFQBgQoQXFo0NwdAgd915FpDh6XH9MSnSG8u2p2PdwRzsu1CKtx+IRhx7SYnoNl0qk+P3W08iq0yB8RFeWDMrGq4O+p2odCssLGRYMysGxbVH8eH+Swj0cMADQ/x1z8ubVCiqaURytI+EVRo2hlEiuiVqjRbfnr6CtSnZyCiRw8pChvvjfPHU6JAb3pQf5u2Mnb8fgfWHcvDej1mY+c9fsHBUMJawl5SIbtGetCK8tOsMmtVavDglAk+NDtbb0k13wt7GEh/Pjcd9//gFL+8+Az83BySGeAAALrbOtufkpa4xjBLRDTUo1fgytQAfH8pFUU0jHGwssWBkPzw+qh98XW9tmz0rSwssHtsfkwZ4Y+n2dKxv6yWdGY34QPaSElHnmlQavPrteXz+az68nG2xYX4sEoI9pC6rU17Odvh0/lDM+OcvWLT1BHYtHoEQTyduA3oLZOJ2tg8wQH5+figsLJS6DCKTU12vxKYjedj0Sx6qG1Rwd7TB/BFBmJsY+JuGxtQaLT4+lIv//fEiVFotFib1w/N3hbOXlIg6yK9swOLPT+BsUR1GhHjg7w/FwtPZVuqybirlYjkWbEyFn5s9di8eiff2XcTmI5dx+MVx8HNzkLo8ydworzGMElEHhdUN+ORQLr5ILUCjSgM/N3s8OToYD8T7w96m+wJjVqkcS3ecRnpBDYJ7O+LtB6IQH+h+8xcSkcn74WwJlu1Ih6JZjT+M64//mRgGSwMalr+ZrUcv409fncXQIDeotQKXShU4vfIuSWb8G4ob5TUO0xMRgJalR9alZGNPejE0WoEBfV2waEww7h7cF1Z6mAEa6u2MnYsS8cnhXKz58SJmrj3CXlIiM6fSaPHm9xn45HAu3B1tsPGxYRgT5il1WbdtzvBA5FXU45PDuQCA+EA3sw6iN8MwSmTGhBBIzavG2pRs/JxRBgAYHuyORWNCMCbMU+9fnlaWFlg0JgQTB3jh+e2n8fGhXPx0oYy9pERmqLimEc98fhIn82sQH+iGDx+JRd9et3ZfuiF6adoAXK5qwI/nSzl56SY4TE9khrRagX0XSrE2JRsn82sgkwGTI/tg0dgQxPTAVnqdUWu0+NfhXLz740WoNFo8PrKll7Q7bw0gIsN0ILMMf/wiDdUNKjwxqh9emBJhEmtyNijV+N8fL2JGvJ/Z777Ee0aJCACgVGvxVVoR1h/MwaUyBWwsLXB/nC+eGB2MEE8nqcsD0LKW4NLtp5FWUIN+vR3x9swoDAliLymRKdJoBd7bdxEf7r8EJ1srvPNANCYP7CN1WaQHDKNEZk7RrMa2X/Pxr8O5KKlrgpOtFWYPD8CCkf3g7WIndXnX0WgFPjmUo+slXTCyH5ayl5TIpBzOqsCqf1/A+St1GOTrgo8eiUeAh/nONjd1DKNEZqpc3oyNv+Riy5HLqGtSo7eTLRYkBWF2QiB62VtLXd5NXSpTYNmOdJzKb+klfWtmVIddnojI+Fy4UofV32fg4MVy2FhZ4LGRQfjjRG6CYeoYRonMTH5lA9Yfysb244VoVmsR5OGAJ0eH4P44X6P7wtdoBT49nIt3/pMJpUaLx0b0w7LJ7CUlMjYltU149z+Z2HGyEEIA98b4YOnkcLNee9OcMIwSmYmzRbVYm5KNf5+5Aq0Aovx6YdGYEEwe2Meo1ujrTHa5Asu2p+Nkfg2CPBzw9gPR7CUlMgLyJhXWpmTjX4dz0aTSYkSIB16eNgCDfHtJXRr1IIZRIhMmhMAv2ZVYm5KNQ1kVAIBRob3x+zEhSAzxMKm17a7tJZ0/IggvTI5gLymRAVJptNh2LB9/35eFynolwryd8NLUARgbrv9l48jwMIwSmSCNVuCHsyVYm5KNM0W1sJABd0f54KnRwSbf45BdrsALO07jxOVqBHo44O2Z0RjWj72kRIZACIG950rw5g+ZyK2oh5ezLZ6/Kwwz4vz0soEGGQeGUSIT0qTSYOfJQnx8MAd5lQ2wtbLArCH+eGJUsFnNRNVoBTb8Nxdv723pJZ2XGIQXpoTDwYZ7eRBJ5cTlaqz+9wUcv1wNRxtLPDUmBAtH9ePfS2IYJTIFtY0qbD16GRv+m4cKRTNc7KwwNzEI80cGobeTrdTlSSanXIFl7XpJ35oRhYRgD6nLIjIruRX1eOuHDHx/tgSWFjI8PMwf/zMhDJ7O5vvdRB0xjBIZsdK6Jnx6OBef/ZoPRbMafVzssHBUPzw0LABOtuxtADr2kjarW+8lZS8pkd5V1Svx/k9Z2Hr0MtRagUmR3nhxSgT6exnGJhpkOBhGiYxQdrkC61NysPtUEZQaLfp7OeGp0cH4XYwvbKx431VncivqsWx7Oo6zl5RIr5pUGvzrcC7WHsiGvFmNGH9XvDxtAO/dpi4xjBIZkVP51Vibko3/nC+FEEB8oBsWjQnBhAgvWBj58kw9QaMV2PhLHt7em4EmFXtJibqTRiuw+1QR3v1PJq7UNiHA3QEvTAnH3YP7coY83RDDKJGBE0LgwMVyrD2QjV9zqwAA4yO88PuxIVxL8w7lVtTjhR3pSM2rRoC7A96aGYXh7CUlumMHL5Zj9fcZuHClDq4O1nh2fCjmDA/kSA3dEoZRIgOl1mjx7ekrWJuSjYwSOawsZEiO9sFTY0IQ3sdZ6vKMnra1l/St1l7SeYmBeGFKBBx5ry3RLdFoBTJK6vC37zNwKKtCt33n4rH9jWJLYTIcDKNEBqZRqcGXxwuw/mAOimoaYW9tiYeG+WPhqGD4utpLXZ7Jad9L6u9uj7dmRCMxhL2kZFxqG1QokzehWa1t/aVBs1oLZduxSgOlRotm1dXnle3PVWnbPX+jc9uuqYFK0xIRZDLgvhhfPD85nN9RdEcYRokMRHW9EpuPXMamI3moqlfCzcEa80f0w9zEQLg52khdnkm7tpd0bmIgXmQvKRmJ705fwZIv09Cs1v7ma8lkgK2VBWwsLWBrbdnys5UFbK1afu5wbG0BFzsrzE4INPnNNEi/GEaJJFZU04hPDuXg/44VoFGlgZ+bPZ4YFYxZQ/y5lWUPy6uoxws7TuNYXhX83e3x5owojAjpLXVZRJ0SQmD9wRys/j4Dns62eHiovy5A2rYGRpu2n60tYGPZEiCve65dwLS2lHGyEfU4hlEiiWSWyLEuJRtfpxdDrRWI6OM0iZiPAAAeUklEQVSM348Nwd2D+3JbPAlptQKbj+ThzR8y0ajS4NHhgVg+lb2kZFjUGi3++s15bDl6GWHeTtjw2DAOkZPRYhgl6mGpeVX454Fs/JxRBgBI6OeORWNDMDbMkz0SBuRyZT2W7TiNY7lV8HOzx1sz2UtKhqFBqcYfPj+FnzLKkBjsgbWPxnPCEBk1hlGiHqDVCvyUUYa1Kdk4cbkaMhlwV6Q3Fo0JQWyAm9TlUReu7SWdMzwAL00dwF5SkkyZvAmPbzyOM0W1uD/WF3+bEcXlk8joMYwS6ZFSrcWetCKsP5iDrDIFrC1luD/WD0+OCUaIJ7fEMxaXK1vuJf21rZd0RhRG9GcvKfWsS2VyzPs0FUU1jXh2Qij+ODGUoylkEhhGifRA0azG/x3Lx78O5+JKbROcbK0wOyEAC5L6wdvFTury6A5otQJbjl7G377P0PWSLp86AE7sJaUecDSnEk9uPo4GpQar7huMWUP9pS6JqNswjBJ1owpFMzb+Nw+bj+ShrkmN3k62WJAUhNkJgbyny0TkVzZg2Y50/JpbBV/XlntJR7KXlPRoT1oRlm5Ph62VJT6aHYfRYZ5Sl0TUrSQNo1lZWZg3bx4qKirg6uqKjRs3IjIystNzm5qaEBcXBwcHBxw/fvyWrs8wSj0lv7IB6w9lY/vxQjSrtQjycMATo4MxI84PdtZcnsnUaLUCW39t6SVtUGowOyEAL01jLyl1LyEEPjqQjbf3ZqKPix02PDYUA/q6SF0WUbeTNIyOHz8ec+fOxfz587Fjxw68++67OHLkSKfnPv/886ipqUF6ejrDKBmMs0W1WJuSjX+fuQKtAAb79sKiMSGYMqgPLC14L5epy69swAs703E0h72k1L3UGi3+vOcsth0rQEQfZ2x4bCj69uLSTWSaJAujZWVlCAsLQ0VFBaysrCCEQN++fXH06FEEBQV1OPfQoUN4++23sWTJEixdupRhlCQlhMCR7Er8MyUbh7IqAACjQntj0ZgQjAjx4IQCM6PVCnz262Wsbu0lfSQhAC9NjYCzHW/LoDujaFbj6c9OIuViOUaF9sZHs+P454lM2o3yml7HmwoKCuDj4wMrq5a3kclkCAgIQH5+focwWl9fj+eeew5ff/01srKybnjNNWvWYM2aNbpjhUKhl9rJPGm0AnvPlWBtSjZOF9bCQgZMj+qLRWNCuBWeGbOwkOHRxCCMDffCCztO4/Nf85GSWY43Z0QhKZS9pHR7Suua8NiGVJy/UodZQ/zwxn2DYc1NMMiM6f3mp2t7kDrriF22bBmefvpp+Pr63jSMLlmyBEuWLNEd+/n5dU+hZNaaVBrsOlmEjw/lILeiHrZWFpgzPABPjApGoIej1OWRgfB3d8BnCxPw2bF8rP73Bcz51694eFgAXp7GXlK6NZklcjy24RiKa5vw/KQwPDO+P0dayOzpfZg+NDQUlZWVNxymj4qKQl1dHYCWSUzV1dXo378/zp07d9P34DA9/RZ1TSpsPXoZG/6bh3J5M1zsrPBoYiDmj+gHT2dbqcsjA1ZQ1YAXdpzGkZxK+Lra428zBmNUKGdAU9f+e6kCi7acQJNagzdnROH+OHamkPmQdALT2LFjMX/+fN0EpnfeeQdHjx7t8vwDBw4Y7D2jzWoNhABkMsBCJmv9dX3vLxm+0romfHo4F5/9mg9Fsxp9XOzweFI/PJwQwNnSdMu0WoHPW3tJ65UaPDzMHy9PG8Be0lZCCGi0AlYcgsaOE4VYvvM07G0ssW5OPDdUILMj2T2jALBu3TrMnz8fq1atgouLCzZt2gQAWLhwIZKTk5GcnKzvErrNs9tOYe+50k6fuxpQW8KpDFePLWQyyFofv3rcdu7VYCvr5Nii3bV0z1kAMlx9r67fo+N7tV2r/XHbtdrXbyGTAbIb12/RGsB151i0noP257S2xQ2O2+pr+Yxt12k9F519ls5/v9oWrdeyuHrc/lpaAXx7uhi7ThZBqdEixNMRK8ZE4t4YX263R7fNwkKGOcMDMSbMEy/uPI1txwqQklmOv82IMot1IlUaLUpqm1BU04jimkYUVTeiuLYRhdUtx8U1TdBoBSZFemNGvC9Gh3qaXTAVQuD9ny7hf/ddhK+rPTY8NhRh3s5Sl0VkULjo/W349HAuzhTVQisEhECnv2tFy5ePVggIoMOxVgsItD/n+t87v2bLtToci47H2tZj0e5YKwTQri7d69Dx2BzFBbhi0ZgQTBzgDQsuz0TdQIiWXtJV37X0kj401B+v3G3cvaSKZnVLwKxpRGH7wFnTiKKaRpTWNUHbyXeInbUFfF3t4evmgCalBsfyqgAAns62uDfGBzPj/RHex/QDmVKtxcu7z2DHiUIM9HHBhvlD4cXd2chMcQcmuqH2QbjroHt94G4J262v1XYWdK+G5JteGy3X6BDQr7lWy+tufO0u/4HQ7h8BkX1dEB/oxtsrSC8KqhqwfNdp/PdSJXx62WH1jCiMMcBeUq1WoELRjKLWYNkWNItqrvZ01jaqOn2th6MNfN3s4dPLvuV3V/uW8OnacuzmYN3h71dBVQN2nyrCjhOFyK9qAAAM8nXBzDg/JMf4wt3Rpkc+c0+qa1Jh8daTOHypAuPCPfHhI3Fw5C1AZMYYRomIetC1vaQPDvHHK9MHwKUHe0mb1RpcaQ2WRdf0aLYNoSs12uteZ2UhQ19XO13QbAuZPq1B06eXPext7mzHMSEEUvOqsfNEIb47cwWKZjWsLWUYH+GFGXF+GBfhZRJLHF2pbcRjG1KRUSLHIwkBeDV5oNndnkB0LYZRIiIJFFY3YPnOMzh8qQJ9e9nhb93USyqEQF2jul3QbEBxbVNrz2bLr3J5c6evdbK10vVg+rjawdfVAT6udvBr7eH0crbrkZ3FGpUa7D1Xgp0nC3H4UgWEaOlxTY7xwcx4Pwz0Mc51fc8V12LBxlSU1jXjxSkRWDQmmKMwRGAYJSKSjBAC244V4I3vzt9yL6lGK1Am7xguO/ZsNkHRrO70tV7Otrqhcz/Xq0PobT2bLnZWBheOimsasftUEXaeKERORT0AIKKPM2bG++F3Mb5Gs8xaysVyLN56AiqNwNsPROF3Mb5Sl0RkMBhGiYgkVljdgJd2ncGhrJZe0j9Pj4STrVWHoNk2SaiktgnqTmYG2VhZtAZLu6sBs929mn162cHW6s6G0A2BEAKnCmqw40QhvkkvhrxJDUsLGcaFe2JGnB/GD/Ay2M/3RWo+Xt59Fk62Vlj/aDwSgj2kLonIoDCMEhEZACEE/i+1AG98d6HTns1e9ta6YHndvZqudujtaGs2qz80qTT48Xwpdp4sxMGL5dAKwNXBGsnRLcP4g317GUQPrxAC7/7nIj7cfwl+bvbY+Ngw9PdykrosIoPDMEpEZECKahrx1amilvDpdjV0csOFzpXWNeGr1tn4WWUKAEColxNmxvvhvlhfyZZLUqq1eHHnaew+VYRov174ZN5Qo7mlgKinMYwSEZHRE0LgTFEtdpwoxNfpxahpUMFCBowOaxnGnxTpDTvrnhnGr21Q4amtx3E0pwoTB3jh/Ydj4WDDf0wQdYVhlIiITEqzWoOfL5Rh58lC7M8sh0Yr4GJnhemtw/ix/q56G8YvrG7A/A2puFSmwLzEQKy4Z2CPrEBAZMwYRomIyGSVy5uxJ61lGD+jRA4ACO7tiBnxfrg/zhd9e9l323udKazFgk2pKJc34093D8DjSf0M4t5VIkPHMEpERGbhXHEtdp4owldpRaiqV0ImA5L698aMOD9MHtjnjhfsB4CfM0rxzOenoNYKvPdgDKYN7tuNlROZNoZRIiIyKyqNFgcyy7HjRAF+ziiDSiPgZGuFuwf3xcwhfhhym1sCbz16GSv2nEUve2t8Mm8I4gPd9Vg9kelhGCUiIrNVVa/E12lF2HmyCGeKagEAgR4OmBHXMozv5+bQ5Wu1WoG39mZibUo2Aj0csPGxYejX27GnSicyGQyjREREADJL5Nh5shC7ThahQtGyZWpisAdmxPth6qA+cGy3vFaTSoNlO07jm/RixAa44pO5Q+DhxKWbiO4EwygREVE7ao0Wh7IqsONEIX48XwqlRgsHG0tMHdQXM+J9EdHHBU9tOY7UvGpMGdgH7z0U02PLRhGZIoZRIiKiLtQ0KPHN6SvYeaIQaQU1AABrSxlUGoHHk/rh5WkDuHQT0W/EMEpERHQLLpUpsPNkIfadL8Wc4YGYNyJI6pKITALDKBERERFJ5kZ5zaKHayEiIiIi0mEYJSIiIiLJMIwSERERkWQYRomIiIhIMgyjRERERCQZhlEiIiIikgzDKBERERFJhmGUiIiIiCTDMEpEREREkmEYJSIiIiLJMIwSERERkWQYRomIiIhIMgyjRERERCQZhlEiIiIikgzDKBERERFJhmGUiIiIiCTDMEpEREREkmEYJSIiIiLJMIwSERERkWQYRomIiIhIMgyjRERERCQZhlEiIiIikgzDKBERERFJRiaEEFIX8VvY2trC09OzR99ToVDAycmpR9/TWLBtOsd26Rrbpmtsm66xbTrHduka26ZrPdE25eXlaG5u7vQ5ow+jUvDz80NhYaHUZRgktk3n2C5dY9t0jW3TNbZN59guXWPbdE3qtuEwPRERERFJhmGUiIiIiCRjuXLlypVSF2GMEhMTpS7BYLFtOsd26Rrbpmtsm66xbTrHduka26ZrUrYN7xklIiIiIslwmJ6IiIiIJMMwSkRERESSYRi9gaamJtx7770ICwtDTEwMpkyZgry8PABAWVkZpkyZgtDQUAwaNAiHDx+WttgedtdddyEqKgoxMTEYNWoU0tLSALBd2vvrX/8KmUyGs2fPAmDbAEBQUBAiIiIQExODmJgYfPHFFwDYNs3NzXjmmWcQGhqKgQMHYs6cOQDYLjU1Nbo/KzExMQgLC4OVlRWqqqrMvm0AYO/evYiPj0dsbCwGDRqETZs2AeCfGwD44YcfMGTIEERFRWH48OFIT08HYH5t8+yzzyIoKKjD/4uAG7dDQ0MDHn74YfTv3x9hYWHYtWuX/gsV1KXGxkbx3XffCa1WK4QQ4oMPPhCTJk0SQgjx2GOPib/85S9CCCGOHTsmAgIChEqlkqrUHlddXa37effu3SI2NlYIwXZpc+LECTFlyhQREBAgzpw5I4Rg2wghRGBgoK492jP3tnnuuefEH/7wB913TXFxsRCC7XKtt99+W0yfPl0IwbbRarXC3d1dpKenCyGEyM3NFba2tqKurs7s26aqqkp4eHiI8+fPCyGEOHDggBg4cKAQwvz+3KSkpIiCgoLrvntv1A5//etfxbx584QQQuTk5Ahvb29RVVWl1zoZRm9DamqqCAkJEUII4ejoKMrKynTPDR06VOzfv1+iyqS1ceNGER8fL4RguwghRFNTkxg+fLjIycnp8AXAtuk6jJpz2ygUCtGrVy8hl8uve86c26UzkZGRYvfu3UIItk1bGE1JSRFCCJGeni58fHxEc3Oz2bdNamqqGDBgQIfHnJycxIkTJ8y2ba797r1RO0RGRopjx47pnnvggQfEhg0b9Fqflf77Xk3H+++/j3vuuQeVlZXQarUdtiENCgpCfn6+hNX1vLlz52L//v0AWoZE2C4tVqxYgTlz5qBfv366x9g2V82ePRtarRYJCQlYvXo1LCwszLptsrOz4eHhgddffx379u2Dvb09Vq5ciZiYGLNul2sdOXIElZWVmD59Ov8+AZDJZPjyyy9x//33w9HREdXV1di1axfkcrnZt01oaCjKy8tx9OhRDB8+HLt374ZCoUBubq7Ztw1w8/8f5efnIzAwsNPn9IX3jN6iVatWISsrC2+88QaAli+C9oQZrpC1efNmFBQU4PXXX8eyZcsAsF2OHDmC1NRULF68+LrnzL1tAODgwYNIT0/HyZMn4eHhgXnz5gEw77ZRqVTIyclBZGQkjh8/jg8//BAPPfQQ1Gq1WbfLtT799FPMnTsXVlYtfSjm3jZqtRqrV6/Gnj17cPnyZfz000/8+9SqV69e2LlzJ5YvX474+HgcOHAAkZGRUCgUZt82bW7WDu2f74k2Yhi9Be+88w527dqF77//Hg4ODvDw8AAAlJeX6865fPkyAgICpCpRUvPmzdP1kALm3S4pKSnIyMhAv379EBQUhMLCQkyePBnHjh0DYN5tA0D3ea2trfHcc8/h0KFDZv/3KTAwEBYWFpg9ezYAIDo6Gv369cOFCxcAmG+7tFdfX48vvvgCCxYsAACz/zMDAGlpaSguLsbIkSMBAEOHDoWPjw9Onz4NwLzbBgBGjx6NAwcO4MSJE3jrrbc6tJW5t83N/v4EBAToJmtf+5y+MIzexJo1a7Bt2zb8+OOPcHV11T3+wAMP4B//+AcAIDU1FSUlJUhKSpKqzB5VV1eH4uJi3fHu3bvh4eEBd3d3s24XAFi+fDmKi4uRl5eHvLw8+Pn5Ye/evZg6darZt019fT1qamp0x9u2bUNsbCwA8/771Lt3b0yYMAF79+4F0PLFn5ubi/DwcLNul/a2b9+OqKgoRERE6B4z97bx9/dHYWEhMjMzAQCXLl1CdnY2wsLCzL5tAODKlSu6n1977TWMHz8e/fv3Z9u0ulE7tH8uNzcXKSkpSE5O1m9Ber0j1cgVFBQIACI4OFhER0eL6OhoMWzYMCGEECUlJWLSpEmif//+IjIyUhw4cEDiantOfn6+GDp0qBg0aJCIiooSEyZMEKdOnRJCmHe7dKb9TePm3jbZ2dkiJiZGDB48WAwaNEgkJyeL3NxcIQTbJjs7W4wZM0YMGjRIREdHi127dgkh2C5tkpKSxKefftrhMbaNEJ9//rnue3jw4MFi27ZtQgi2jRBCPP744yI8PFyEhISIOXPm6FaAMbe2Wbx4sfD19RWWlpbC29tbNwn7Ru2gUCjErFmzREhIiAgNDRXbt2/Xe53cDpSIiIiIJMNheiIiIiKSDMMoEREREUmGYZSIiIiIJMMwSkRERESSYRglIiIiIskwjBKRSUtLS8OXX34pdRk669atQ0REBGJiYlBZWdmt116xYgW++OKLm563cuVKLF26tNPnDK29fguZTAaFQiF1GUR0EwyjRGQw1Gp1t1/T0MLVe++9hy1btiAtLU23E0p3efXVV/Hggw/+pmvou7308d+YiIwbwygRdQuZTIaVK1di5MiRCAsLw7Zt23TPpaamYvz48RgyZAji4uKwc+dOAEBeXh569+6NV199FaNGjcIHH3wApVKJZcuWYfDgwYiOjsaUKVN013nnnXcwbNgwxMXFYdq0aSgoKADQ0tP3yCOP4J577kFkZCTGjx+PqqoqlJWVYcWKFdi3bx9iYmKwaNEiAMCcOXMwZMgQREVFYfr06SgrK9O9xyuvvIL+/fsjISEBy5Ytw5AhQ3TPbdmyBQkJCYiLi8OYMWNw9uzZTtvi+PHjSExMRFRUFIYNG4b//ve/AICZM2ciOzsbjz76KGbOnNnhNQqFAu7u7lCpVACAuLg43RahOTk5CAkJAdCyl/3y5csxbNgwxMTE4KGHHtLtbDV//nx8+OGHAIDa2lrMmDEDERERGD9+PB599NEOvaHFxcW31F6NjY148MEHERkZiejoaNx1112dfuaxY8fiueeew9ixYxEaGoply5bp9rQeO3YsXnnlFUyYMAGTJ0/W/Xdv/9nb74Utk8nw5ptvIiEhAf369cOGDRt0z2VlZeHuu+/G0KFDER0djY8++kj33K5duxAREYHExES89tprndZJRAZI78vqE5FZACBWrlwphGjZUcjDw0Pk5+eL6upqERsbK4qLi4UQQpSXl4uAgABx5coVkZubKwCIzz77THedlStXivvuu080NTUJIYQoKysTQgjx2WefiSeeeEKo1WohhBCbN28WycnJQggh/vKXv4jg4GBRWVkphBDiwQcfFKtWrRJCCLFhwwYxY8aMDrWWl5frfl69erV4+umnhRBCfP311yIqKkooFAqh0WjEfffdJ+Lj44UQQhw+fFhMmzZNV9fBgwdFVFTUde3Q3Nws/P39xQ8//CCEEOLQoUOiT58+QqFQCCE67sp1raSkJJGSkiIqKipEVFSUCA0NFUIIsXbtWvHkk08KIYR44403xGuvvaZ7zauvviqeffZZIYQQ8+bNEx988IEQQoglS5aIxx9/XAghRFVVlQgKChLPP//8bbfXrl27xKRJk3THba+51pgxY8SkSZOEUqkU9fX1Ij4+XnzxxRe656ZNmyaUSqUQQojc3Fzh4eGhe61cLhft/3cEQLz33ntCCCHOnz8vnJychEqlEmq1WgwZMkRcuHBBCCFEfX29GDx4sDhx4oQoLS0V7u7uIiMjQwghxJtvvikACLlc3mm9RGQ4rCRNwkRkUhYuXAgACA4ORlJSEg4dOgRXV1fk5ORg6tSpuvOEEMjMzERgYCDs7Ozw8MMP65779ttv8e6778LW1hYA4OnpCQD46quvcPz4ccTHxwMANBoNLC0tda+bOnUq3N3dAQCJiYk4c+ZMl3V+9tln2LJlC5qbm9HY2Ig+ffoAAPbv349Zs2bB0dERADBv3jxdD9uePXuQnp6OhIQE3XXKy8uhVCphY2OjeywzMxM2NjaYPHkyACApKQleXl44ffo0EhMTb9h+EydOxL59+1BaWorJkyfjwoULOHv2LPbt24dZs2bp2qGurg47duwAACiVSl2vaXv79+/HBx98AABwc3PDvffe2+H5W22v6OhoZGRkYPHixRgzZgymTZvWZf3z5s2DtbU1rK2tMWfOnA51P/roo7C2tr7h52+vrVd4wIABsLKyQklJCerq6nDu3Dk89NBDuvPkcjnOnz+PwsJCxMXFITw8HADw5JNP4sUXX7zl9yMi6TCMEpHeyGQyCCEQFRWFgwcPXvd8Xl4eHB0dOwzRdkUIgT/96U9YsGBBp8/b2dnpfra0tOzy3sTDhw/jww8/xC+//AJPT098/fXXePXVV3Xv0VUtQggsWLBAd+6N6uzsGrfyGSdOnIhly5ahrKwM999/P3x9ffHjjz8iJSUFa9eu1V3/o48+wvjx4++ojja32l7BwcE4f/48fv75Z+zbtw8vvPAC0tLS4ObmdtPP0/79nZycdD9bWVlBo9Hojpuamm6pPiEEevfujbS0tOvO37Nnz03rISLDxHtGiajbfPrppwBaQubhw4eRlJSEESNGICsrCz///LPuvLS0NCiVyk6vkZycjPfeew/Nzc0AWnof2x7/6KOPUFVVBaDl3slTp07dtCYXFxfU1tbqjqurq+Hi4gJ3d3colUqsW7dO99y4ceOwfft2NDQ0QKvVYsuWLbrn7rnnHmzevFl3n6pWq8Xx48eve7+IiAg0NzfrPu8vv/yCsrIyDB48+Ka1JiQkICMjA/v27cOoUaMwceJE/P3vf4e/v79uslNycjLWrFmDhoYGAEBDQwPOnTt33bXGjRuHTZs2AQBqampuOaxd216FhYWQyWRITk7GO++8AyGErg2utWXLFqjVajQ2NuLzzz/HxIkTOz2vT58+UKvVyMzMBABs3rz5lmoLDw+Hg4NDh/MvXbqEqqoqJCYm4tSpU7h48SIA4JNPPrmlaxKR9BhGiajb2NraYuTIkbjrrrvwwQcfwN/fH25ubvjmm2/w2muvITo6GpGRkVi+fDm0Wm2n13jxxRcREhKC2NhYxMTEYN68eQBahnnnzJmDsWPHIjo6GjExMdi/f/9Na5owYQLq6+sRHR2NRYsWYerUqejfvz8iIiIwefJkxMTE6M5NTk7G5MmTER0djXHjxiEkJAS9evUCAIwePRqrVq3C7373O0RHR2PQoEGdLqNkY2ODnTt34pVXXkFUVBSee+45bN++XTf0fyNWVlZISkpCQEAA7O3tMXDgQKhUqg6hbvny5YiJiUFCQgKioqIwfPjwTnsKV6xYgbKyMkRGRuKRRx7ByJEjdZ/ldtrrzJkzGDFiBKKiohAXF4dHH30UUVFRnb42Li4OEydORFRUFMaMGXPdJK32n/P999/H1KlTMXr0aN0/PG6lfb755ht8+eWXiIqKwsCBA7Fw4UI0NjbCy8sL69evxz333IMRI0bAwoL/eyMyFjIhWqc7EhH9BjKZDHK5vMNwrDGSy+VwdnaGVqvFwoUL4ePjg9dff13qsm6bSqWCRqOBnZ0d6urqkJSUhDVr1nTZW/lbjR07FkuXLsX06dP1cn0iMl28Z5SIqJ25c+ciLy8PjY2NiIuLwwsvvCB1SXekuroaU6dOhUajQWNjI2bPnq23IEpE9FuwZ5SIiIiIJMObaoiIiIhIMgyjRERERCQZhlEiIiIikgzDKBERERFJhmGUiIiIiCTDMEpEREREkmEYJSIiIiLJ/D9HwlHEkNEf0wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 800x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualization of training and testing loss VS percentage of remaining weights-\n",
    "# pruned-\n",
    "fig=plt.figure(figsize=(10, 9), dpi= 80, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.plot(list(plot_loss.keys()), list(plot_loss.values()), label = 'training_loss')\n",
    "plt.plot(list(plot_test_loss.keys()), list(plot_test_loss.values()), label = 'testing_loss')\n",
    "\n",
    "plt.title(\"Conv-2 CNN: Percentage of weights pruned VS. Loss\")\n",
    "plt.xlabel(\"percentage of weights pruned\")\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.legend(loc = 'best')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Plot number of epochs (Early Stopping) VS percentage of weights pruned-\n",
    "\n",
    "# Python 3 dict to hold number of epochs vs % of weights pruned-\n",
    "# Experiment-1\n",
    "plot_num_epochs = {}\n",
    "# plot_num_epochs_test = {}\n",
    "\n",
    "# populate 'plot_num_epochs'-\n",
    "for k in history_main.keys():\n",
    "    num_epochs = len(history_main[k]['accuracy'])\n",
    "    plot_num_epochs[history_main[k]['percentage_wts_pruned']] = num_epochs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x560 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize percentage of weights remaining VS number of epochs (Early Stopping)\n",
    "fig=plt.figure(figsize=(10, 7), dpi= 80, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.plot(list(plot_num_epochs.keys()), list(plot_num_epochs.values()), label = 'number of epochs')\n",
    "# plt.plot(list(plot_test_accuracy.keys()), list(plot_test_accuracy.values()), label = 'validation accuracy')\n",
    "# plt.plot(list(plot_test_loss.keys()), list(plot_test_loss.values()), label = 'validation loss')\n",
    "\n",
    "plt.title(\"Conv-2 CNN: Percentage of weights pruned VS number of epochs (Early Stopping)\")\n",
    "plt.xlabel(\"percentage of weights pruned\")\n",
    "plt.ylabel(\"number of epochs\")\n",
    "plt.legend(loc = 'best')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Try to visualize the STARTING training and testing accuracy VS percentage\n",
    "# of remaining weights-\n",
    "\n",
    "# Python 3 dict for visualization-\n",
    "plot_starting_accuracy = {}\n",
    "plot_starting_test_accuracy = {}\n",
    "\n",
    "# Key:    percentage of weights pruned\n",
    "# Value:  starting accuracy\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "# populate 'plot_starting_accuracy'-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_starting_accuracy[history_main[k]['percentage_wts_pruned']] = history_main[k]['accuracy'][0]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "# populate 'plot_starting_test_accuracy'-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_starting_test_accuracy[history_main[k]['percentage_wts_pruned']] = history_main[k]['val_accuracy'][0]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize starting accuracy VS percentage of weights pruned-\n",
    "fig=plt.figure(figsize=(10, 9), dpi= 80, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.plot(list(plot_starting_accuracy.keys()), list(plot_starting_accuracy.values()), label = 'training starting accuracy')\n",
    "plt.plot(list(plot_starting_test_accuracy.keys()), list(plot_starting_test_accuracy.values()), label = 'testing starting accuracy')\n",
    "\n",
    "plt.title(\"Conv-2 CNN: Percentage of weights pruned VS. Starting Accuracy\")\n",
    "plt.xlabel(\"percentage of weights pruned\")\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.legend(loc = 'best')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Try to visualize the STARTING training and testing loss VS percentage\n",
    "# of remaining weights-\n",
    "\n",
    "# Python 3 dict for visualization-\n",
    "plot_starting_loss = {}\n",
    "plot_starting_test_loss = {}\n",
    "\n",
    "# Key:\t\tpercentage of weights pruned\n",
    "# Value:\tstarting loss\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Populate 'plot_starting_loss' Python 3 dict-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_starting_loss[history_main[k]['percentage_wts_pruned']] = history_main[k]['loss'][0]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Populate 'plot_starting_test_loss' Python 3 dict-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_starting_test_loss[history_main[k]['percentage_wts_pruned']] = history_main[k]['val_loss'][0]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize Starting training & testing loss VS percentage of weights pruned-\n",
    "fig=plt.figure(figsize=(10, 9), dpi= 80, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.plot(list(plot_starting_loss.keys()), list(plot_starting_loss.values()), label = 'training starting loss')\n",
    "plt.plot(list(plot_starting_test_loss.keys()), list(plot_starting_test_loss.values()), label = 'testing starting loss')\n",
    "\n",
    "plt.title(\"Conv-2 CNN: Percentage of weights pruned VS. Starting Loss\")\n",
    "plt.xlabel(\"percentage of weights pruned\")\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.legend(loc = 'best')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Try to visualize: percentage of weights remaining VS BEST accuracy within each iterative\n",
    "# pruning round\n",
    "\n",
    "# Python 3 dict for visualization-\n",
    "plot_best_accuracy = {}\n",
    "plot_best_test_accuracy = {}\n",
    "\n",
    "# Key:\t\tpercentage of weights pruned\n",
    "# Value:\tbest accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Populate 'plot_best_accuracy' Python 3 dict-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['accuracy'])\n",
    "    plot_best_accuracy[history_main[k]['percentage_wts_pruned']] = np.amax(history_main[k]['accuracy'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Populate 'plot_best_test_accuracy' Python 3 dict-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['accuracy'])\n",
    "    plot_best_test_accuracy[history_main[k]['percentage_wts_pruned']] = np.amax(history_main[k]['val_accuracy'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize best accuracy VS percentage of weights pruned-\n",
    "fig=plt.figure(figsize=(10, 9), dpi= 80, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.plot(list(plot_best_accuracy.keys()), list(plot_best_accuracy.values()), label = 'training best accuracy')\n",
    "plt.plot(list(plot_best_test_accuracy.keys()), list(plot_best_test_accuracy.values()), label = 'testing best accuracy')\n",
    "\n",
    "plt.title(\"Conv-2 CNN: Percentage of weights pruned VS. Best Accuracy\")\n",
    "plt.xlabel(\"percentage of weighs pruned\")\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.legend(loc = 'best')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Try to visualize: percentage of weights remaining VS LOWEST loss within each iterative\n",
    "# pruning round\n",
    "\n",
    "# Python 3 dict for visualization-\n",
    "plot_best_loss = {}\n",
    "plot_best_test_loss = {}\n",
    "\n",
    "# Key:\t\tpercentage of weights pruned\n",
    "# Value:\tbest loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Populate 'plot_best_loss' Python 3 dict-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_best_loss[history_main[k]['percentage_wts_pruned']] = np.amin(history_main[k]['loss'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Populate 'plot_best_test_loss' Python 3 dict-\n",
    "for k in history_main.keys():\n",
    "    epoch_length = len(history_main[k]['loss'])\n",
    "    plot_best_test_loss[history_main[k]['percentage_wts_pruned']] = np.amin(history_main[k]['val_loss'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize best loss VS percentage of weights pruned-\n",
    "fig=plt.figure(figsize=(10, 9), dpi= 80, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.plot(list(plot_best_loss.keys()), list(plot_best_loss.values()), label = 'training best loss')\n",
    "plt.plot(list(plot_best_test_loss.keys()), list(plot_best_test_loss.values()), label = 'testing best loss')\n",
    "\n",
    "plt.title(\"Conv-2 CNN: Percentage of weights pruned VS. Best Loss\")\n",
    "plt.xlabel(\"percentage of weights pruned\")\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.legend(loc = 'best')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}