NN_pytorch.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "%matplotlib inline\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_mnist(path, kind='train'):\n",
    "    import os\n",
    "    import struct\n",
    "    import gzip\n",
    "    import numpy as np\n",
    "\n",
    "    \"\"\"Load MNIST data from `path`\"\"\"\n",
    "    labels_path = os.path.join(path,\n",
    "                               '%s-labels-idx1-ubyte.gz'\n",
    "                               % kind)\n",
    "    images_path = os.path.join(path,\n",
    "                               '%s-images-idx3-ubyte.gz'\n",
    "                               % kind)\n",
    "\n",
    "    with gzip.open(labels_path, 'rb') as lbpath:\n",
    "        struct.unpack('>II', lbpath.read(8))\n",
    "        labels = np.frombuffer(lbpath.read(), dtype=np.uint8)\n",
    "\n",
    "    with gzip.open(images_path, 'rb') as imgpath:\n",
    "        struct.unpack(\">IIII\", imgpath.read(16))\n",
    "        images = np.frombuffer(imgpath.read(), dtype=np.uint8).reshape(len(labels), 784)\n",
    "\n",
    "    return images, labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set size: (60000, 784)\n",
      "Testing set size: (10000, 784)\n"
     ]
    }
   ],
   "source": [
    "X_train, y_train = load_mnist('data/fashion', kind='train')\n",
    "X_test, y_test = load_mnist('data/fashion', kind='t10k')\n",
    "y_train,y_test = y_train.astype(np.int64),y_test.astype(np.int64)\n",
    "print('Training set size: {}'.format(X_train.shape))\n",
    "print('Testing set size: {}'.format(X_test.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from matplotlib import pyplot as plt,cm\n",
    "def show_img(im,figsize=None,ax=None,cmap=cm.binary):\n",
    "    if not ax: fig,ax = plt.subplots(figsize = figsize)\n",
    "    ax.imshow(im,cmap=cmap)\n",
    "    return ax"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = X_train.reshape([-1,28,28])\n",
    "X_test = X_test.reshape([-1,28,28])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x864 with 12 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig,axes=plt.subplots(3, 4, figsize=(16, 12))\n",
    "for i,ax in enumerate(axes.flat):\n",
    "    _=show_img(X_train[i],ax=ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Custom FashionMNIST data loader"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pytorch model expects data in minibatches and in torch tensors. We will use pytorch's DataLoader to do this"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torch.autograd import Variable\n",
    "#To download the dataset for torchvision\n",
    "import torchvision\n",
    "from torch.utils.data import DataLoader\n",
    "from torch.utils.data.dataset import Dataset\n",
    "from torchvision import datasets, transforms\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels={\n",
    "    0:'T-shirt/top',\n",
    "    1:'Trouser',\n",
    "    2:'Pullover',\n",
    "    3:'Dress',\n",
    "    4:'Coat',\n",
    "    5:'Sandal',\n",
    "    6:'Shirt',\n",
    "    7:'Sneaker',\n",
    "    8:'Bag',\n",
    "    9:'Ankle boot',\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# custom dataset\n",
    "class FMNIST(Dataset):\n",
    "    def __init__(self,X,y,transforms=None):\n",
    "        self.X = X[:,:,:,None] # n,28,28 to n,28,28,1\n",
    "        self.y = y\n",
    "        self.n = len(X)\n",
    "        self.transforms = transforms\n",
    "    \n",
    "    def __getitem__(self,index):\n",
    "        img = self.X[index]\n",
    "        label = self.y[index]\n",
    "        if self.transforms is not None:\n",
    "            img = self.transforms(img)\n",
    "        return (img,label)\n",
    "    \n",
    "    def __len__(self):\n",
    "        return self.n\n",
    "tsfms = transforms.ToTensor() # convert numpy to tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_ds = FMNIST(X_train,y_train,tsfms)\n",
    "train_dl = DataLoader(train_ds,batch_size=12,shuffle=True,num_workers=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "X,y=next(iter(train_dl))\n",
    "X=X.numpy()\n",
    "y=y.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(12, 1, 28, 28)\n",
      "(12,)\n"
     ]
    }
   ],
   "source": [
    "print(X.shape)\n",
    "print(y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.00392157, 0.        , 0.        ,\n",
       "        0.06666667, 0.11372549, 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.01960784, 0.14509805, 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        ],\n",
       "       [0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.6       ,\n",
       "        0.7882353 , 0.7882353 , 0.90588236, 0.61960787, 0.43137255,\n",
       "        0.70980394, 0.8156863 , 0.76862746, 0.8       , 0.6745098 ,\n",
       "        0.09803922, 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        ],\n",
       "       [0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.21568628, 0.85490197,\n",
       "        0.73333335, 0.7372549 , 0.7647059 , 0.79607844, 0.8       ,\n",
       "        0.8117647 , 0.79607844, 0.7529412 , 0.7254902 , 0.83137256,\n",
       "        0.6901961 , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        ],\n",
       "       [0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.43137255, 0.84705883,\n",
       "        0.7411765 , 0.74509805, 0.72156864, 0.7019608 , 0.70980394,\n",
       "        0.7019608 , 0.70980394, 0.7372549 , 0.73333335, 0.78039217,\n",
       "        0.8156863 , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        ],\n",
       "       [0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.5294118 , 0.84313726,\n",
       "        0.7490196 , 0.7490196 , 0.7254902 , 0.7176471 , 0.7137255 ,\n",
       "        0.70980394, 0.7137255 , 0.7411765 , 0.7294118 , 0.79607844,\n",
       "        0.85882354, 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        ]], dtype=float32)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X[0][0][:5][:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x864 with 12 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig,axes=plt.subplots(3, 4, figsize=(16, 12))\n",
    "for i,ax in enumerate(axes.flat):\n",
    "    ax=show_img(np.squeeze(X[i]),ax=ax)\n",
    "    _=ax.text(0,0, labels[y[i]],\n",
    "        verticalalignment='top', color='black', fontsize=12, weight='bold')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(28, 28)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train[0].shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Add normalized mean and std transformation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.2860405969887955, 0.3530242445149223)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nm_mean = np.mean(train_ds.X)/255 # between 0 and 1\n",
    "nm_std = np.std(train_ds.X)/255\n",
    "nm_mean,nm_std"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "tsfms = transforms.Compose([transforms.ToTensor(),transforms.Normalize((nm_mean,),(nm_std,))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_ds = FMNIST(X_train,y_train,tsfms)\n",
    "train_dl = DataLoader(train_ds,batch_size=32,shuffle=True,num_workers=0)\n",
    "test_ds = FMNIST(X_test,y_test,tsfms)\n",
    "test_dl = DataLoader(test_ds,batch_size=32,shuffle=False,num_workers=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data augmentation \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## using fast.ai library data loader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.transforms import *\n",
    "from fastai.conv_learner import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "sz=28\n",
    "bs=12\n",
    "\n",
    "def gc2RGB(X): # grayscale to 3 channels\n",
    "    X1 = X[:,:,:,None]\n",
    "    return np.tile(X1,3)\n",
    "\n",
    "def get_data_fastai(sz,bs):\n",
    "    X_train1 = gc2RGB(X_train)\n",
    "    X_test1 = gc2RGB(X_test)\n",
    "    stats = np.array([[nm_mean]*3,[nm_std]*3])\n",
    "\n",
    "    tfms = tfms_from_stats(stats,sz,\n",
    "                           aug_tfms=[RandomRotate(10),RandomFlip()],\n",
    "                          crop_type=CropType.NO,\n",
    "                          pad=1)\n",
    "    data = ImageClassifierData.from_arrays('data/fashion',(X_train1,y_train),(X_test1,y_test),\n",
    "                                           bs=bs,tfms = tfms,classes=list(labels.values()))\n",
    "    return data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Train set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = get_data_fastai(sz,bs)\n",
    "ite = iter(data.trn_dl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(12, 3, 28, 28)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(12, 28, 28)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X,y = to_np(next(ite))\n",
    "X.shape\n",
    "X=X[:12,0,:28,:28]\n",
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA6AAAAKvCAYAAAB05KLQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzs3Xl8VeW1//HvE8CQAQiYMA8BB5SqZSqgIKAWnOtQtWi1Tq322sGft962t9bp1ttae1W0tvaqKF6H1qnUiWpxwAFnFEekzCEESMI8h8D+/QGkDGctyMnJzg75vF+vvIC9znP2cw5n7b2fnHPWClEUCQAAAACA+pbV0BMAAAAAADQNLEABAAAAALFgAQoAAAAAiAULUAAAAABALFiAAgAAAABiwQIUAAAAABALFqAAAAAAgFiwAAUAAAAAxIIFKAAAAAAgFs3rMjiEcIKkOyQ1k3RfFEU3e7cvLCyMiouL67LLerdlyxYztmHDBjNWVVVlxqIoqvU8Nm/ebMY2bdqU1r5at25txvLz8/duYkjb1KlTK6MoKopzn/tijtYHL38rKytTbg8hmGO8WDrHg7rcZ7t27VJuz87OTmse+zJyFEg2chRItr3N0bQXoCGEZpL+IGmUpFJJ74cQnomi6AtrTHFxsT744IN0dxmLdevWmbEvv/zSjM2bN8+MeYtaa6G5atUqc8ySJUvMWHV1tRk77rjjzNjRRx9txpAZIYT5Me9vn8zR+rBgwQIzdu+996bcvt9++5ljmjVrZsa8xaJ3rPAWjOvXrzdjF1xwQcrtPXv2NMc0VeQokGzkaGZ45yHvl53AnuxtjtblI7iDJM2KomhOFEVVkv4i6bQ63B+AzCJHgWQjR4FkI0eBelCXBWgXSTu+bVC6bRuAZCBHgWQjR4FkI0eBelCXBWiq9+h3e08/hHBZCOGDEMIHFRUVddgdgFoiR4FkS0SOzps3TyEE86N3e4rvjZEjRyqEoPHjx6d9H0ADSESObnf++ecrhKCbbropZby6uromV0tLS+ttHkBd1WUBWiqp2w7/7iqpbNcbRVF0TxRFA6MoGlhUtPN3UouLi2sSJdXP5MmT6zA9oMmrU46Sn0C9q/N5dFt8p3ydPn16/c04wW644QaFEHTRRRc19FSw72hU59GsrCxdeeWVuvLKK9WqVSv3tkcffbSysrL08MMP7xZ7/PHH1apVK1VVVWnYsGEKIaS8HZCuulTBfV/SQSGEnpIWShoj6bza3MEll1yiZcuWSZLuvvtuVVVV6Zvf/Ka6du0qSTV/7qi6ulrNm9epeG+92Lx5s1t4BGgAdcrRfSk/kzovNHl1Po9K0uuvv6758/9V9+Ghhx7Sr3/964xNEmjCYj+P1kVWVpbGjh3r3mbTpk17PB8+//zzGjVqlFtsD6iLtN8BjaKoWtIPJb0oabqkx6Mo+rw293Hddddp7NixGjt2rHJyciRJP/zhD2u2bf9t5r/927/puOOO03777ad33nlHVVVVuummm9S7d2/l5eWpT58+uvPOO2sqSP7yl79UCEHf/e53JaX+SMKtt96qXr16KTs7W0VFRTrmmGM0c+ZMNWvWTBs2bNC1116rww8/XO3bt9ewYcP097//XTk5OcrJydH111+vAQMG6JZbbtEVV1yhwYMHa+bMmeZvuDybN29O+bNlyxbzZ9OmTebP2rVrzZ8ZM2aYP9XV1eYPGqe65mgS81OS1q5dq5/+9Kc68MADlZ+frwEDBuiZZ56pmff2jyjtOi/PunXrzJ/Vq1en/Fm+fLn5s2zZMvNn6dKl5k+641auXGn+bNiwIeUPGl4mzqOSat6Z6NevnyTp0Ucf3anK5faPv/7nf/6nhg8frtzcXA0dOnSnReuOZs2apU6dOqlZs2b6v//7v5S3KSkp0ZgxY9SlSxcVFBRo9OjR+uyzz/Y417lz52r48OHKy8vTMcccs1MF+U8++UQnnHCCCgsLVVRUpFNPPVUzZsyoiVdUVOi73/2uunfvrtatW2vIkCF64YUXJG199/PGG2+UJD344IMKIWjkyJF7nE8mRVFk/ljn+s2bN5vnXm+Md42AzKnv8+iBBx6425iHH35Yhx56qFq2bKl27drpqKOO0ltvvbXTbSoqKnT66acrNzdXffv21aeffipp5/PpggULFEWRunbtqhCCfv3rX+vQQw9Vfn6+jj76aE2ZMkWS9J3vfEdZWVk1H+vdsmWLXnjhBZ188skaNmxYze0uuOAChRD0q1/9Slu2bNHLL7+sYcOGqU2bNurSpYvOPfdczZ49W+vWrdOqVatq5nH77berR48eKigo0HnnnafZs2dr4cKFWrhwoRYsWGD+rF+/3vxJl3euLykpMX8qKytT/lRVVZk/8NXlI7iKomhiFEUHR1F0QBRF/52pSe3qT3/6k7Zs2aLzzz9frVq10s9//nNde+21Wrt2rcaMGaPy8nJdeeWVuvXWW/fq/r788ktdffXVWrNmjS6++GJ9/etf17x582ram3zve9/TbbfdpoKCAp122mkqKSnR2WefvVtZ7ccee0xRFOnUU09Vy5YtM/64gbqKI0fjzs+LLrpIv/vd71RQUKAzzzxT8+fP1+mnn6433njDnReQRHXN0Y0bN+rJJ5+UtPUXN23bttX8+fP1+uuv73bbW265Rd26dVNhYaHeeust/fKXv9ztNgsWLNBxxx2n8vJyPfDAA/rOd76z223WrVunY489Vo8//riOOOIIjRo1SpMnT9Yxxxxj9s3d7uabb1aPHj3Uq1cvTZ48WWeddZYkadGiRRoxYoRefPFFDRkyRP369dNzzz2nkSNHavny5dqyZYu+8Y1vaNy4cSosLNRpp52mqVOn6uSTT9aUKVM0ZMgQDR48WJJ06KGH6sorr6y5b6Au4rrWlbb+gvWiiy7SggULdMEFF+jkk0/WihUrNHfu3J1u9/vf/15ZWVnq0aOHPv74Y/34xz/e431ff/316tu3r8444wydffbZ6ty5syTp+OOP149//OOa/Hn//fdVUVGhk046Seecc85utxs0aJA++ugjnXDCCXr77bd14oknqmvXrvrLX/6iM844Y7c3Lm6++WYdeeSRatasmf785z/v9bUA9m2N4jNpxxxzjF555RVJW38zc/fdd0vaugAcOnSoTjrpJJ111ln6/e9/r//4j//Y4/1t2rRJ0taPPnzzm99Unz591KVLF23evFkLFizQX//6VzVr1kxDhgxRVlaWDjnkEE2ZMkWPP/64Bg4cWHM/gwcP1gMPPCDJ7xEK7MvizM9FixbpySefVLNmzXTUUUcpKytLffr00RtvvKH//d//3amf7Y7zAvZVzz33nFasWKH27dtrxIgROuWUU/TQQw/p4Ycf1ogRI3a67fe//3394Q9/0AMPPKBLLrlEH3300W739/Wvf12lpaXm4lPa+vG82bNnq0uXLurdu7ckqXv37po9e7aefPJJff/73zfne8UVV+j2229XZWWlOnXqpKlTp+rzzz/X888/rxUrVmjkyJF67rnnJG19R3fatGl64okn1LdvX73zzjvKz8/XG2+8oby8PBUWFmrs2LH6wx/+oEcffVTvvPOO3n33XQ0aNGiPH0MEGtqWLVv07//+7zX/PvLII3XCCSdoy5Yt2n///XX66aerT58+6tmz524940899VT99a9/1aRJkzR69OiUubyrX/7yl7ruuutq/v3EE0+orKxM3/72t3X++efXbH/++efVv39/derUST/+8Y/1+OOPq6ysTOeff77OO2/rp48vv/xyVVdX69JLL9U999yjjRs3qlu3bvr000/15ptvatiwYTX3d/fdd2vw4MF6/vnndfnll+upp57SNddck/bzhn1Do1iADh06tObvixcvrvn42KGHHipJOuSQQyRJpaWluyWppN22HX744bruuut01113afTo0TX39dRTT6m8vLxmzB//+MedxpWUlOz07+0fdwKasjjzc8WKFTVjfv/73+80btasWea8gH3V9o/fnnrqqcrKytIZZ5yhhx56SE888YTuuusuZWdn19x2+zmroKBAkrRmzZrd7u+f//yniouLdeqpp5r73P6x2YULF+qOO+7YKbZrHu5q+3GhsLBQhYWFWrx4sUpLS2vuc3tc2nrsmDZtmubPn18z527duikvL68mLsn8KDGQZFu2bNkpf9asWaNvfetb+sMf/qBf/epXOuWUUyRtfc0/8sgjO/2CdW9yeVd7e058/vnn3fyXtFu+Zmdnq0ePHlq6dOlu18rbf0m1/ePG5eXl1GVA3T6CG5cdT6AdO3as+bjr9ncdt39HpGvXrmrWrFnNyWnVqlWStNv3Uqqrq3Xddddp6dKlmjdvnq6++mpNnz5dY8eOVY8ePSRJLVu21KJFi2q+L7V69erdfqPKl7OBePOzuLhY0tb8XLZsWc33q6qqqmo+hphqXsC+aPny5Zo4caIkady4cQoh6Mwzz5QkrVy5Us8+++xOt99+wefVJvjOd76jefPm6ZRTTtG6detS3mZ7Hg4YMEBbtmypycPly5fv8Z2N7RV6t3+HStp6bNh+nzt+mmj7saNHjx418QULFtTMa8e4pJpCgHwPEo1B8+bNd/qe8H333SdJuvTSS1VWVqaysjLddtttWrBgwW5tV/Yml3e16zkxVb4sWrRIH330kU4++WT3drvma1VVVc3Cs3v37jvtZ3uebv/lVPv27Vl8onG8A7qjrKwsff/739fYsWP1rW99S8cff7yefvppSVu/2C396zdDzz77rK6++urdTsLz5s3TsGHDNHz4cBUVFenNN9+UtPU3SZ06ddLpp5+uv/3tbxo6dKiOPfZYLV26VFOmTNGYMWPcjxYBTV0c+XnmmWfqr3/9q772ta/p61//uiorK/XGG2/oRz/6UcrvtAH7qscff1xVVVVq3bq1jjnmmJrtX3zxhWbOnKmHHnqo1t+DvO+++1RWVqaXXnpJZ511Vk3+7uikk05Sr169NHXqVA0dOlRHHHGESkpKNHnyZE2cONEt/vPHP/5RlZWVmjZtmqqrq9W/f3/16dNHbdu21a9//Wu9+uqr+sY3vqGqqip99NFH6tChg8466ywVFBRo8ODBevfdd3X00UfrK1/5iv785z8rhKArrrhC0tZ3iiTp73//u370ox9p5MiR+uY3v1mrxw80pOrqanXo0EEjR45Up06d9Mknn0j61zudmbQ9X26//XZ9+OGHuvTSS/Xee++pffv2O33dbPvtbrvtNk2dOlWXXHKJLrvsMj3wwAO6//77tWbNGs2dO1eVlZU67LDDdnun9YorrtDIkSM1adIkSar5JRmatkbxDuiubr75Zt14443KycnRo48+qsLCQt166636yU9+ImnrF6V/8IMfKDs7WxMmTNCPfvSjncYXFBRo4MCBeuONN3Tvvfdq0aJFOu+88/SLX/xCknTvvffW3NdDDz2kd999V0ceeaSOOuqoeB8o0AjVd36OHz9eP/3pT2v+/vbbb2vo0KE1H9cFmopHHnlE0tbvY/3tb3+r+bn33nslbV2ILV26tFb32aJFCz311FM64ogj9Pe//10XX3zxThV1JSkvL08vv/yyzj33XJWUlOjBBx/UjBkzdP7559d83M7y85//XPPmzdPs2bM1YsQIPfHEEwohqHPnznr11Vc1evRoTZkyRR988IFOPvlkvfrqq2rXrp2ysrL0zDPP6OKLL1Z5ebkmTJigfv366Zlnnqn5vtnZZ5+t448/XmvXrtVdd92lV199tVaPHWhoWVlZOu644zR16lSNGzdO06dP1ymnnKLf/e53Gd/X1VdfrcMOO0yff/657rzzTs2ePVsTJ07USSedtNM7q//xH/+x2+0GDBigiRMnasiQIXr++edVUlKic845RxMmTFCLFi122s+1116rd955R9XV1frWt75Vcy2Api3semKpTwMHDox2rSSbNKm+o7ad9z2TadOmmTHrY0zS1opnqWyv+JmKd0Hh9SL1PpZ05ZVXmrHtH7VIpTYf/2jqQghToygauOdbNpzGkKP14bbbbjNjt9xyS8rtXh56eeEdc9P96GD79u3N2K4f3drukksuSWtf+3LOk6PYEy9HN27caMa2f389lXTanXmV972q395XExpDbpOjO0v3Gj7VdWlVVZV69Oih3/72tzrxxBNTjquoqDDvc8d2Z5s3b9ZVV10lSbrxxhs1YcKElGO290hN5Wc/+5kZ894Q8p6TyZMnmzFvkT9kyJCU27/97W+bY3b8PvuuWrdubcaKiorMmJe/3uOOM7f3Nkcb5TugAAAAADJj+fLl+uEPf7hTsSOgvrAABQAAAJqwDh066JprrlF+fn5DTwVNQKMrQgQAAAAgmZo1a6Y777yzoaeBBOMdUAAAAABALFiAAgAAAABiwQIUAAAAABALvgO6i6wse03evLn9dO23335mzGuNYt3n+vXrzTEHH3ywGfNKxL/11ltmzGsj06FDBzOWm5trxoDGory8vNZjCgsLzdiufdB25OWo144h3TLqX3zxRUbvD2gs0m1L4I3zctSLeS3eNm3alHK7dz1SWVmZ1jy8Fg/eNQ4aVjrtVrx2fk8//bQZe/bZZ83YjBkzzNiqVavMmHVO9PLC6xfqXZd6r3+vfdr+++9vxqZMmZJy+9tvv22O8Vqt9O/f34xdeOGFZsyrUJyXl2fGkoh3QAEAAAAAsWABCgAAAACIBQtQAAAAAEAsWIACAAAAAGLBAhQAAAAAEAtKnu3Cq4znxbxqsF6VL6varVdNzKtU583DGzdv3jwztmzZsrT2BySJl4clJSVmbM2aNSm310elW2+OXjVtb5xV/Xrx4sXmmI4dO5oxoLFIt9KzN27q1KlmzDsmfPbZZ2bMqszpVcrs1KmTGfPO9d6xwnvcXkVeKmrXP+s5XrlypTnmwQcfNGN33HGHGfP+r72OD23atDFjVhVf77XjVZH1zqNeHnbv3t2MeeftnJyclNu9fNqwYYMZe+mll8yYd272npOjjjrKjCUR74ACAAAAAGLBAhQAAAAAEAsWoAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALGoUxuWEMI8SaslbZZUHUXRwExMqiFZpaL3FPNKMa9bt67WsRUrVphjvFLRBxxwgBkrKCgwY97+ysvLzVjXrl3NGBrevpij6fJaG3mvf6ulgVeqftOmTWbMy1+Pd/zJzs42Y0uWLEm5ffLkyeaYMWPG7PW8UDfkaP3xcsZr/+C1HnvooYfMWMuWLc1YRUWFGbOOP7NmzTLHHHjggWasd+/eZuyQQw4xY9788/LyzJjVjs1rh+EdP5MmyTn65ZdfmrHnn3/ejHmterzXgXet6+Wbdd7zXgfptlrxzrHpxtJpI+O1rPGe/7lz55qxDz/80Iz17dvXjCWxZWIm+oAeE0VRZQbuB0D9IEeBZCNHgWQjR4EMajy/ggIAAAAANGp1XYBGkv4RQpgaQrgsExMCkFHkKJBs5CiQbOQokGF1/Qju0CiKykII7SVNCiF8GUXR6zveYFuyXiZJ3bt3r+PuANQSOQokGzkKJBs5CmRYnd4BjaKobNuf5ZImSBqU4jb3RFE0MIqigUVFRXXZHYBaIkeBZCNHgWQjR4HMS3sBGkLICyG02v53SaMlfZapiQGoG3IUSDZyFEg2chSoH3X5CG4HSRO2lSBuLunRKIpeyMisGlC6LRK8cV5LhqqqqpTbvdYtq1evNmO9evUyY23btjVjVqsGSVqwYIEZ69+/vxlDg9snczRdXt5s2LDBjHnl3i3r169P6/68Y4VX0t0rBb9y5cqU2z/++GNzDG1YYkOO1iMvL7xz9gsv2P8F06ZNM2MbN240Yx06dDBjVrsJr+XC/PnzzZjXKqNjx45mrE2bNmbMa5dhtZTwjmdeGwrv/60BJCJHrefSuz4rLS01Y97/p/f68WJe2xRrf948mjVrlta+0uUdE6zXZLrtGb1x3vWD13Zn6dKlZmyfasMSRdEcSV/N4FwAZBA5CiQbOQokGzkK1A/asAAAAAAAYsECFAAAAAAQCxagAAAAAIBYsAAFAAAAAMSCBSgAAAAAIBZ1acPS5Hglob3yzemUrfZKU5eVlZkxr32LV2J98eLFZswr8w00Fl5OeWX/rZhXBt7Lw1tuucWMXXfddWbM2192drYZs1o9eW0cgH2d1+Jk3LhxZuzzzz83Y4MGDTJjPXr0MGPWtUVlZaU5pnXr1massLDQjJWXl5sx77jVsmVLM2a1lqqoqDDHtGvXzozl5+ebsabKarHx9ttvm2O81mNeGxyv5VdeXp4Z69KlS63n4u3La4/mXc961+Ne3nzxxRdmrGfPnim3e9fOXlumFStWmDHvXO/l79q1a81YEvEOKAAAAAAgFixAAQAAAACxYAEKAAAAAIgFC1AAAAAAQCxYgAIAAAAAYsECFAAAAAAQC9qw7MJr1eC1OoiiyIx5bVgsXhlyr32CV7b98MMPr/U8JGn58uVmzCoX7ZX4BhqC1SpA8l+vVhsWr0S8dzy49NJLzdiLL75oxl577TUzVlBQYMasx+2VbPfaMeTm5poxIEm8PHzllVfMWElJiRlLJ9ck/zxqXSO0bdvWHOO1hfvoo4/MmMdrR+XlvdX2wmtRccopp5ixXr16mbGmasmSJSm3v/fee+YY73rWas8l+ec2r51Qhw4dzFhRUVHK7V47Eu8158W8NiYdO3Y0Y961tdWGqHPnzuYYL+dLS0vNmHcc8catXLnSjCUR74ACAAAAAGLBAhQAAAAAEAsWoAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALGgV8YuvDLk++23nxnzyiZ7rDLZXvlsr63LwoULzdghhxxixryS7l67hqVLl6bc7pXjBhqCl9ve69/KRS8PDz744LT29b3vfc+Mvfzyy2bMe2xW+6hVq1aZY7z2CbRIQGOxYMECM+a1KiksLDRj3vmwoqLCjHnXD1ZrF6/lQqtWrcyYZ9OmTWbMa7/htaGz2j+sX7/eHNOuXTszVlxcbMaaKut1V1ZWZo7x/s+885DXPsRqubOnmHW+8XKma9euZmzRokVmzGtjOH36dDPWp08fM2Y9z16rJ+95TOecLUnLli0zY2vWrDFjScQ7oAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALFgAQoAAAAAiAULUAAAAABALPbYhiWEcL+kUySVR1F02LZt7SQ9JqlY0jxJ50RRZNcb3kd4rVZycnLMmNdSxSrFXFVVZY5p27atGbPaokjSihUrzJhXktsr7bxkyZKU22nDEh9ydO94ediyZUszZrVP8O5v6NChZsw7jgwbNiytcd7xwsptr51EeXm5GaMNS+3tiznqtR/weO0H0rVu3bqU259++mlzzOeff27G0m1L5sWsViuS9I1vfCPl9vfff98c47WY8fblHSu8Y5rHarfiPY/777+/GfOucepLEnLUa5FjtTHxnkevnZZ3Xee1GPPODd7/m/XYvNec93x417NeGxav1ZB1PSvZxxiv9Yx3Pb569Wozlm6rJO+YNnjw4JTbveeqvu3N0Wa8pBN22fZzSS9HUXSQpJe3/RtAwxgvchRIsvEiR4EkGy9yFIjNHhegURS9LmnXzqenSXpw298flHR6hucFYC+Ro0CykaNAspGjQLzS/Q5ohyiKFknStj/bWzcMIVwWQvgghPBBRUVFmrsDUEvkKJBs5CiQbOQoUE/qvQhRFEX3RFE0MIqigUVFRfW9OwC1RI4CyUaOAslGjgK1k+4CdEkIoZMkbfvT/lYygIZAjgLJRo4CyUaOAvVkj1VwDc9IulDSzdv+tEvN7UO8al25ublmzKuSlU4VXK9qlTcPr2pY8+b2S2Hjxo1mzKsahgbVJHPUk+kquF6Fx/79+5sxb1yrVq3MWN++fc3YzJkzzVibNm1qPQ+v0h4yJjE5mm5F20zvy6uQ61WNfOihh1Juf/bZZ80x1dXVZqyystKMWfkkSd27dzdjM2bMMGP33ntvyu2tW7dOax5WxU7Jf9we7xrB4lVa9e4v3Wq89SDWHPWekyOPPDLl9ocfftgcM2fOHDM2ceJEM/bZZ5+ZsdLSUjP2xRdfmDHrdecdD7yKu55Zs2alNa6srMyMpVO926tcP3z4cDM2cuRIM3bQQQeZsaOOOsqMeZ06GsoeszyE8GdJb0vqHUIoDSFcqq3JOCqEMFPSqG3/BtAAyFEg2chRINnIUSBee/yVVhRF5xqh4zI8FwBpIEeBZCNHgWQjR4F4JeZzDgAAAACAfRsLUAAAAABALFiAAgAAAABiwQIUAAAAABCLdNuwNEnptnHw2rCks6/NmzebMa/UslfG3hvntWFZtGiRGQOSxGs7smnTJjNm5ZtXYr1Hjx5pzcOL9enTx4x5LR6sNjIe71iBpiPd9ixeywIv5rUPefDBB83YI488knK7d67Mzs42Y17+rl271oydc845ZiwvL8+M/fa3v025vWvXruYY73n0rjm8Y51n/fr1Zsx6npcuXWqOSaetxb7Oe06sljxeO55evXqZMautiyS99dZbZuyGG24wY14bE6vFjPeYvfOhx8v7dO8znder11bt/PPPN2NjxowxY9753IslMd94BxQAAAAAEAsWoAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALFgAQoAAAAAiAVtWGrBa43ilVv2xlmlkb1S0cuXLzdjXsuUFStWmLG2bduaMa/8eklJiRkDkqSqqiqtmNW2wCorL/ltHLzjgZf3vXv3NmNejlqPzWu1Ulpaasaw77Fe46tWrTLHeK9Vr32Ll2uvvPKKGXvzzTdrfZ9ey5R27dqZsUMOOcSMvfPOO2bsj3/8oxn7yle+Uuu55Ofnm2OWLFlixrz/G69Fi3dMS4fXns6LJbFlRENL5zlJ95rV25fX6sPbXzqqq6vNWLqtntJlHdO8nPHOy95j8/5vvGOrF0tiTvEOKAAAAAAgFixAAQAAAACxYAEKAAAAAIgFC1AAAAAAQCxYgAIAAAAAYsECFAAAAAAQC9qw1IJXYjovLy+tcVa59OzsbHOMV3796KOPNmPl5eVmrGvXrmZs8+ZvOgFkAAAgAElEQVTNZqysrCzldq/8vfdcAfXFay1SUVFhxqwWFStXrjTHeO0TvHLoLVq0MGMjR440Y15rBSt/vXYYXssmJJd3rJ4zZ44ZmzJlSsrty5YtM8cUFBSYMStnJP/8NXXqVDM2f/78Ws8lNzfXHFNZWWnGvHPUb37zGzN2/fXXm7FZs2aZsQEDBqTcvnjxYnOMd4zx2j94j81rDZGTk2PGLN7xzGvDgszwzjVeyw6v1YrXdsTLe+uatrG1DtmVd8z1Xv9eqyRP3O1n6hPvgAIAAAAAYsECFAAAAAAQCxagAAAAAIBYsAAFAAAAAMSCBSgAAAAAIBYsQAEAAAAAsdhjG5YQwv2STpFUHkXRYdu23SDpe5K29y74RRRFE+trkkmRbvsEr9z4mjVrar0vrwy2V0bdG7d06VIz5rWE2bBhQ8rtJSUl5pjevXubMa9lDVIjR/fO8uXLzZjXNshrV5JpXt57bRe8NixWaXzv/vbff38zhtrLZI5GUWQed71WW//93/9txqZPn55ye5cuXcwxbdq0MWNeawLvPOS97rz7tFoifeUrXzHHzJw504y99tprac3Da1Xyta99zYxZ8/daNaTbxsRrseEdB73WLunwjllerL5wHv0X7zzkXWN6129WrCH+r9NhzTOdx7ynWFOxN8/AeEknpNh+exRFfbf97PMJCSTYeJGjQJKNFzkKJNl4kaNAbPa4AI2i6HVJdjdqAA2KHAWSjRwFko0cBeJVl/eAfxhC+CSEcH8IoW3GZgQgU8hRINnIUSDZyFGgHqS7AL1b0gGS+kpaJOlW64YhhMtCCB+EED6oqKiwbgYgs8hRINnIUSDZyFGgnqS1AI2iaEkURZujKNoi6V5Jg5zb3hNF0cAoigYWFRWlO08AtUCOAslGjgLJRo4C9SetBWgIodMO/zxD0meZmQ6ATCBHgWQjR4FkI0eB+rM3bVj+LGmkpMIQQqmk6yWNDCH0lRRJmifp8nqcY6PnlWZPh1dG3Ws10bFjRzPmlcJu1aqVGbPKts+ePdscU1xcbMYy/Vw1BeTo3rFaV0h2qxLJbj+Ql5dnjvHaJ3iiKDJjVsumPe1v48aNKbd77ZW8VkmovUzm6KpVq/TSSy+ljM2bN88c95e//MWM9evXL+V2r9WK13rMi61bt86MtW7d2ox169at1vfpvRPlvf4nTZpkxt5//30z1rNnTzPm5bZ1/m3Xrp05ZtWqVWbMu0awWr5I0ooVK8yY1zbCOv54LXes45LkH6vrC+fRf/FaDXltWLz2Ld7rP5NjpPTPv+lIt2VinHNMqj0uQKMoOjfF5nH1MBcAaSBHgWQjR4FkI0eBeNEJFQAAAAAQCxagAAAAAIBYsAAFAAAAAMSCBSgAAAAAIBYsQAEAAAAAsdhjFVzsHa+kstdaxCoz7d2fVw596dKlZsxrreCVdPfaTVitIUpKSswxq1evNmO0YUF9sdqpSH5p+aqqqpTbvZYLXvl1j9cOadGiRWbMa7tgHUu8FhUHH3ywGUPDWrt2rd57772UsdLSUnOc11ph4cKFKbd756GCggIz5r0ePd44ryVD+/btU26fOXOmOcY7V3qPzWtn5rWK8dqOWHnvtVVbsmSJGfOOdR7v+OM9/1YbK28e06ZNM2PesQ6Z4bUP8c6H3mvEuzZNt6VKOtKdY6Z5xzOvVVVTwTugAAAAAIBYsAAFAAAAAMSCBSgAAAAAIBYsQAEAAAAAsWABCgAAAACIBVVwM8SrKJadnW3GrMqEXvXBli1bphVr166dGausrDRjubm5ZsyqKLZgwQJzjFdx16pmCNSVV1HSq9ZoVdRr3bq1OSbdSnte9cE33njDjHkVTq37LCwsNMd06tTJjKFhNW/e3DxOehXGvcrM1rjZs2ebY7xjtXfO8M6VFRUVZsxjnfe8fXnnPK965eLFi83YypUrzdiKFSvMmJW/+fn55hivqq53HPF4xy2vsqg1f6/SsFcpf+3atWYMe897/Xu859+qCi/FW2HW41Xc9V7H6Txf3r68SrfpVgrflyTj1QIAAAAA2OexAAUAAAAAxIIFKAAAAAAgFixAAQAAAACxYAEKAAAAAIgFC1AAAAAAQCxow5IhXon7dMott2rVKq15eOXXvbL5X3zxhRnzWsJYZbeXLFlijlm2bJkZ80pap1tSHJD88uvpaNOmjRnzcsZ7jXsl7l966SUz5uWGVQq+V69e5hgkV25urvr3758ytmnTJnOc97rLyclJuX3dunXmmLKyMjPmtQ9Jt1WDdx71HpvFa13kxbyWTV7Meo4lqbi4uNbz8J5Hrx2bd43gvX68ti9r1qxJud27LjrwwAPNGO3YGpbXhsV7jcTJO5975990Wfnm5Wg65+WmhHdAAQAAAACxYAEKAAAAAIgFC1AAAAAAQCxYgAIAAAAAYsECFAAAAAAQCxagAAAAAIBY7LENSwihm6T/k9RR0hZJ90RRdEcIoZ2kxyQVS5on6ZwoipbX31STzSv77JVbtso05+XlmWO8cvRe+ezc3Fwz5pV091pDWPvzSnWXl5enta/s7Gwz1pSRo3vHa5HglVK3WK0TpPRyXvJLy8+aNcuMee0OCgoKUm4fNGiQOQaZlckczc3N1eGHH54y5rXKsMZIUmlpqbkvi5dPK1euNGPeudJ7/XsxK3+984l3jvLOh14bE493TLDOo6tWrTLHeM+j1/LFO9Zt2LAho7GioiJzjDdHL1ZfOI/+i3ccSbftiNUqKZ0xe4rFfZ8W77zstUpKp62U1PhaFe7NO6DVkn4SRdGhkoZI+kEIoY+kn0t6OYqigyS9vO3fAOJHjgLJRo4CyUaOAjHa4wI0iqJFURR9uO3vqyVNl9RF0mmSHtx2swclnV5fkwRgI0eBZCNHgWQjR4F41eo7oCGEYkn9JL0rqUMURYukrYkrqX2mJwegdshRINnIUSDZyFGg/u31AjSEkC/pKUn/L4oi+4sJu4+7LITwQQjhg4qKinTmCGAvkKNAsmUiRysrK+tvgkATx3kUiMdeLUBDCC20NSEfiaLor9s2LwkhdNoW7yQpZWWZKIruiaJoYBRFA70vpANIHzkKJFumcrSwsDCeCQNNDOdRID57XICGrWWVxkmaHkXRbTuEnpF04ba/Xyjp6cxPD8CekKNAspGjQLKRo0C89tiGRdJQSRdI+jSEMG3btl9IulnS4yGESyWVSDq7fqbYOHhl273y61ZrkTVr1phjvBYtXvlmr6S71aphT+NWr16dcrtXftprw2Ldn0QbFgc5uhcy3YblgAMOMGNeqyTP3LlzzZhXGt/Ljc6dO6fcPnDgwL2fGOoqYzmalZWl/Pz8lLEjjjjCHPfd737XjI0fPz7l9uXL7W4T3jHeawfgtQ/xxqXTPsQ7L3s5mm6rGG9/3n2uWLEi5Xav5Uu67di844i3v3TaV7Rq1coc47X4aaBzPefRbdatW2fGvDxMp31Iui1HPF6OZlq6rVu813im28Ek1R4XoFEUvSnJesTHZXY6AGqLHAWSjRwFko0cBeJVqyq4AAAAAACkiwUoAAAAACAWLEABAAAAALFgAQoAAAAAiAULUAAAAABALPamDQvqyGvDYpXT91qfeKXNlyxZYsZWrlxpxrzGyd44q+y8Vwbea8PiPW4asKMuqqqq0hpnlT33csZrueCZPHmyGfNaPHhl29u1a5dye8eOHfd6Xmgc2rZta8ZOPvnkWo97++23zTHvvvuuGfNaNXjtQ7z2Y15rBauNmJfzXqsJb19e+wTrfC757U+s87Z3HPGuK9J93N7+vONP69atU24fMWKEOebAAw80Y23atDFj2HvptvPw2vF4scbAe9zptBpKt+VUuq3a9qUWLbwDCgAAAACIBQtQAAAAAEAsWIACAAAAAGLBAhQAAAAAEAsWoAAAAACAWLAABQAAAADEgjYsMfDKLVuxdMspL1++3Ix5LVp69uxpxjZt2mTGrHL1FRUV5pjFixebsRUrVpgxoC681gpevjVvnvowuf/++5tj0m3D8tZbb5kxrw2CF+vdu3etxyDZrNer9zru3LmzGTvllFNSbj/iiCPMMX379jVj3nnIO59MmjTJjJWWlpqxI488stb7+vjjj81Yv379zNi5555rxryc8s5tr7/+esrtc+fONcccdNBBZmzRokVmbObMmWasT58+Zqxbt25mLDc3N+X2E044wRzjtVWzjrmIh3ccSbeNiTUundYndZHu/qz5eznfsmVLM5ZuG5Z9CVcgAAAAAIBYsAAFAAAAAMSCBSgAAAAAIBYsQAEAAAAAsWABCgAAAACIBQtQAAAAAEAsqHUdA68U8+bNm1Nu99o4rFq1yoyVlJSYMa/9Sa9evczYsmXLzFiPHj1qtV2SysvLzdj8+fPNWP/+/c0YsCfr1683Y+vWrTNjVisHq/WAlH6Lk9mzZ5sxr0R8Xl6eGRs0aFCt78+TbosoNCzvNZmTk5Nyu9fqo0uXLmasqqrKjHntkI4//ngztnbtWjPWvn37lNu917jXKqaoqMiMFRcXmzHvsXktYawc9c69bdu2NWOVlZVmzLsO6Nq1qxnr0KGDGbOOCd4cabWSGfXRqsQ6HkhSdna2GfPOsS1atKj1PLzHZl07S/71s3cc9O7Tym1vjt5j9l7/6baMa2x4BxQAAAAAEAsWoAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALHYYxmyEEI3Sf8nqaOkLZLuiaLojhDCDZK+J6li201/EUXRxPqaaGPmVai0Kvt5VQTPOOMMM3buueeascsuu8yMDR8+3Ix5FXKtqnkjR440xyxatMiMHXbYYWYMqZGje8erjOdVlrNiXnXrdCvVzZkzx4x5laUPOOAAM2ZV2PR4lf32pSp8cUl6jlr/p17O5Ofn1/r+9sSrPpsO73VcH9VDvcftVQ/t1q1byu1eVVpP586dzZh3jvUqc3qvBeu5bEzHiqTnaJyqq6vN2Jo1a8yYVwXXGrdx40ZzTH1Uavdi3v7222+/lNu96/uKigoz5nWzqI9jUxLtTR3sakk/iaLowxBCK0lTQwiTtsVuj6Lof+pvegD2AjkKJBs5CiQbOQrEaI8L0CiKFklatO3vq0MI0yXZDcAAxIocBZKNHAWSjRwF4lWr74CGEIol9ZP07rZNPwwhfBJCuD+EYHccBhALchRINnIUSDZyFKh/e70ADSHkS3pK0v+LomiVpLslHSCpr7b+1uhWY9xlIYQPQggfeJ+HBlA35CiQbOQokGzkKBCPvVqAhhBaaGtCPhJF0V8lKYqiJVEUbY6iaIukeyWlrHQRRdE9URQNjKJoYKYLDADYihwFko0cBZKNHAXis8cFaNhaMmqcpOlRFN22w/ZOO9zsDEmfZX56APaEHAWSjRwFko0cBeK1N1Vwh0q6QNKnIYRp27b9QtK5IYS+kiJJ8yRdXi8z3Ad47RqKi4tTbvdKXR9zzDFmzCu//uabb5qxIUOGmLHBgwebMattRGFhoTmmoKAgrRhM5OheuPrqq82Y16LIKoneqVOnlNslKSsrvRbLH330kRlr0aJFWjHrt/HplqpHWva5HG0Mr5G4X+Pp3me6xwuL1zIlXU2gNVOjzNF024p4vFY9F198ccb3lw5vX+nmk/dcWvfptaxp06aNGfPaqtVH/ibR3lTBfVNSqv+VfboPEtBYkKNAspGjQLKRo0C8MvtrNwAAAAAADCxAAQAAAACxYAEKAAAAAIgFC1AAAAAAQCxYgAIAAAAAYrE3bVhQR15p5yOPPDLldq+tQk5OjhnLy8szY3369DFjq1evNmNeuejy8vKU26uqqswx2dnZZmzz5s1mrAmUgUc9ateuXVqxOPXs2bOhpwAAu+Ec23R412jptsqz2ph4LVPSbeuS6bZG3n1617q5ublmzLvGbyp4BxQAAAAAEAsWoAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALFgAQoAAAAAiEVIt8xxWjsLoULS/G3/LJRUGdvOfUmZC/PYWVLmIWVmLj2iKCrKxGTqCzm6R8xjZ0mZh0SONqSkzENKzlyYx84yNQ9yND1JmYeUnLkwj53FmqOxLkB32nEIH0RRNLBBdr6LpMyFeSRzHlKy5hKXJD3mpMyFeSRzHlKy5hKXpDzmpMxDSs5cmEcy5xG3pDzupMxDSs5cmEfDzoOP4AIAAAAAYsECFAAAAAAQi4ZcgN7TgPveVVLmwjx2lpR5SMmaS1yS9JiTMhfmsbOkzENK1lzikpTHnJR5SMmZC/PYWVLmEbekPO6kzENKzlyYx85inUeDfQcUAAAAANC08BFcAAAAAEAsWIACAAAAAGLRIAvQEMIJIYQZIYRZIYSfN8Qcts1jXgjh0xDCtBDCBzHv+/4QQnkI4bMdtrULIUwKIczc9mfbBprHDSGEhduel2khhJNimEe3EMKrIYTpIYTPQwhXbtse63PizCP256QhkaPkaIp5kKMJQo6SoynmQY4mSFPP0aTkpzMXcrQBczT274CGEJpJ+qekUZJKJb0v6dwoir6IdSJb5zJP0sAoimJvABtCGC5pjaT/i6LosG3bbpG0LIqim7cdrNpGUfSzBpjHDZLWRFH0P/W5713m0UlSpyiKPgwhtJI0VdLpki5SjM+JM49zFPNz0lDI0Zp9k6M7z4McTQhytGbf5OjO8yBHE4IcTU5+OnO5QeRog+VoQ7wDOkjSrCiK5kRRVCXpL5JOa4B5NKgoil6XtGyXzadJenDb3x/U1hdDQ8wjdlEULYqi6MNtf18tabqkLor5OXHm0ZSQoyJHU8yDHE0OclTkaIp5kKPJ0eRzNCn56cwlduTovzTEArSLpAU7/LtUDXdgiiT9I4QwNYRwWQPNYUcdoihaJG19cUhq34Bz+WEI4ZNtH1uI5SMS24UQiiX1k/SuGvA52WUeUgM+JzEjR23kqMjRBCBHbeSoyNEEIEdTS1J+SuRog+VoQyxAQ4ptDdULZmgURf0lnSjpB9veood0t6QDJPWVtEjSrXHtOISQL+kpSf8viqJVce13L+bRYM9JAyBHk48cJUd3RY4mCzlKju6KHE0WcrQBc7QhFqClkrrt8O+uksoaYB6Koqhs25/lkiZo60cmGtKSbZ/L3v757PKGmEQURUuiKNocRdEWSfcqpuclhNBCWxPhkSiK/rptc+zPSap5NNRz0kDIURs5So4mATlqI0fJ0SQgR1NLRH5K5GhD52hDLEDfl3RQCKFnCGE/SWMkPRP3JEIIedu+eKsQQp6k0ZI+80fVu2ckXbjt7xdKerohJrE9CbY5QzE8LyGEIGmcpOlRFN22QyjW58SaR0M8Jw2IHLWRo+RoEpCjNnKUHE0CcjS1ROSnRI6mmkesz0kURbH/SDpJW6uDzZZ0TQPNoZekj7f9fB73PCT9WVvf3t6krb8pu1TS/pJeljRz25/tGmgeD0n6VNIn2poUnWKYxzBt/XjKJ5Kmbfs5Ke7nxJlH7M9JQ/6Qo+RoinmQown6IUfJ0RTzIEcT9NPUczQp+enMhRxtwByNvQ0LAAAAAKBpaoiP4AIAAAAAmiAWoAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALFgAQoAAAAAiAULUAAAAABALFiAAgAAAABiwQIUAAAAABALFqAAAAAAgFiwAAUAAAAAxIIFKAAAAAAgFixAAQAAAACxYAEKAAAAAIgFC1AAAAAAQCxYgAIAAAAAYsECFAAAAAAQCxagAAAAAIBYsAAFAAAAAMSCBSgAAAAAIBYsQAEAAAAAsajTAjSEcEIIYUYIYVYI4eeZmhSAzCBHgWQjR4FkI0eBzAtRFKU3MIRmkv4paZSkUknvSzo3iqIvrDGFhYVRcXFxWvtrarz/lxBCWuOWLVtmxrKzs83Yfvvtl1YMO5s6dWplFEVFce2PHN17q1atMmNbtmxJub1Vq1bmmGbNmtV5TrvavHmzGausrDRjrVu3Trk9JyenznPa15CjQLKRo41TdXV1yu3r1683x3jXl941a7q86+eNGzem3L5p0yZzTG5urhmrj2uEpNjbHG1eh30MkjQriqI5khRC+Iuk0ySZSVlcXKwPPvigDrtsOqxklfwXblVVlRl77LHHzJh3sOzRo0daMUu6i+vGLoQwP+ZdNrkcTfe19Y9//MOMrV27NuX24447zhyTl5dnxrKy7A+eWItdyV8k33///WZs1KhRKbcffvjh5hjvefTm39iRo0CykaMNyzs3eOcv6w2QTz75xBzTs2dPM+Zds6Z7jvKuu2fPnp1y++LFi80xX/3qV81YQUHB3k+skdnbHK3LlUQXSQt2+Hfptm0AkoEcBZKNHAWSjRwF6kFdFqCp3k7Y7VcjIYTLQggfhBA+qKioqMPuANQSOQokGzkKJFu95+jkyZMVQnDf1bvooosUQtANN9xQq/sGkqouC9BSSd12+HdXSWW73iiKonuiKBoYRdHAoqLdPxJcXFysEIJCCGrevLk6duyoM844Q3PmzKnD1AAoAzm6PT+zsrKUn5+v4uJinXPOOXr33Xfrd+ZA05CxHOUcCtSLjFzrzpkzR2eeeabat2+vli1bqmvXrjrhhBPMj3buavTo0bryyis1ZMgQ93bjx49XCEEjR47cq/sFGkpdvgP6vqSDQgg9JS2UNEbSeene2SmnnKKePXvqxRdf1N/+9jetXLlSr7zySh2mBzR5GcvRk08+WUVFRZoyZYqeeOIJTZgwQY8++qjOPvvs3W67adMmtWjRok4TB5qIjOUo51CgXmQkR8844wx98sknOvbYY3XwwQertLRUr7/+uhYtWrRX48877zydd56/W68gDpA0ab8DGkVRtaQfSnpR0nRJj0dR9Hm693fppZfqzjvv1P/8z/9Ikr788ktJ0k9+8hMVFxerZcuWys3N1ZAhQzR58uSacWVlZRo9erTy8vI0dOhQXX/99QohqG/fvulOBdgnZDJHL730Ut1///36/PPPNWbMGFVXV+v73/++1q1bV/Puy9ixY9WzZ0/17t1bklRSUqIxY8aoS5cuKigo0OjRo/XZZ5/V3OfYsWN1wAEHqGXLlioqKtLIkSM1Y8YMSdKjjz6qPn36KCcnR+3atdORRx6pN998s47PCJAsmc5RzqFAZmUiR5ctW6ZPPvlEBQUFeumll3T33Xfr2WefVXl5ufr377/TbW+77TZ17NhR7du31+9+97ua7bt+BPeGG25QCEFnnXWWzjnnHOXk5OiRRx7RxRdfLEl67bXX9vixXqAh1eUdUEVRNFHSxExMZNy4cXrllVc0adIkSdI3v/lNSdLcuXM1ePBgFRYWat68eZo4caLOPvtszZkzR61atdJ5552n1157TQcffLB69eqlm2++ORPTiY1VNcyr2GlV5ZS2XkxYWrZsmVbMm4tVNax58zq9tJAhmcxRaev/6/XXX6+//OUvWrZsmaZMmVIT+8UvfqGzzz5bOTk5WrdunY499ljNmTNHxx9/vPLz8/X000/rmGOO0fTp07VixQpdddVVKiws1EUXXaRVq1bpnXfe0aJFi9S9e3dddNFFatGihb797W9r48aN+vDDDzV79mwNGzYs7bl71fs2bNhgxlauXJlyu9cWpT54FQZXr15txviteLJlKkeb6jkUqG91zdFWrVopPz9fK1asUL9+/XTsscdq+PDhGjVq1E4V00tKSnTffffpqKOO0oQJE/Szn/1Mp59+ug466CDzvp966in1799fF1xwgUpKSjRq1ChNmjRJXbp00VlnnaV27drVer5WyxFJWrhwoRnb/kuv2ozz3gH25uHFOnToYMY83n1aVXy963GvxYxXKb+pfIIsMauE5557rubv2dnZGjBggCTpvvvu05NPPql58+bpoIMOUm5uriorK/Xpp5+qe/fueu211yRtbaPQo0cPtWvXTnfeeWeDPAagKdix9U55eXnN3++66y5dcsklkqQnnnhCs2fPVpcuXWreEe3evbtmz56tJ598UiNGjJAkde7cWWeeeab69Omjrl27avPmzVq/fr02b96s9u3b6/TTT1efPn3Uq1ev2Bd8QGPCORRIphYtWmjcuHG67LLL9PHHH+vjjz/W7bffrg4dOujZZ5+tuV1WVpZeeeUVdezYUT169FBJSYk+/vhjdwHaq1cvvfvuuzW/9B8/frwmTZqkAw88UGPHjq33xwakKzEN3SZMmKAtW7bonXfe0ZYtW3TppZdq6tSpOuyww3T55ZfrN7/5je644w6tW7dOklRRUVHzG5WcnJyai+I+ffo02GMAmoL58//V4ql9+/Y1fx86dGjN3+fNmydp628977jjDt1xxx01xRZmzZqlQw89VDfeeKMWLlyo448/Xt26ddMhhxyi6dOnKz8/X3fffbeiKNKpp56qAw44QN26ddMbb7wRzwMEGiHOoUBynXPOOVq8eLFeeOEFXXPNNWrfvr2WLFmiX/3qVzW36dixozp27CjpX30i16xZ497voEGD+MQZGqXELEClrR/1HDBggPLy8rRlyxbNnTtXixYtUlFRkRYvXqyNGzfWJGUURerSZWsrpvXr16u0tFSS/xEAAHVTXV2tG2+8UZLUrl27nRad2dnZNX/f/r2TAQMGaMuWLYqiSFEUafny5brmmmu0efNmXXPNNaqsrNT8+fP1s5/9TDNmzNDtt98uSbrwwgu1cOFClZWV6Y477lBpaelOJ2oAu+McCiTPpk2b9Oabb6ply5Y6/vjjddNNN+k///M/Je389YkdF5LeV592tON5V5KaNWsmyf/KBpAEifm1ybhx4/Tqq6/qww8/1IoVK5Sbm6vtpawrKip01VVXac6cOTv9Nqhr164aMWKEXnvtNY0ePVoDBw7UY4891lAPAdhnjRs3Ts8884ymTJmif/7zn2revLn+9Kc/KTc3N+XtTzrpJPXq1UtTp07V0KFDdcQRR6ikpESTJ+CtdXwAACAASURBVE/WxIkTVVxcrMGDB2v48OFq3759zXdJt18cd+jQQSNHjlTnzp316aef7hQDsDvOoUAybdy4UUcffbQOPfRQ9evXT7m5uZowYYIkadSoURndV7duWzvGTJ06VVdccYX69eun733vexndB5AJiXkH9LnnntOdd96pzz77TMOGDdOzzz6rESNG6JprrlHbtm01adIknXvuuTW/sd3ukUce0ahRozR//nzNnj1bV111laTdfysEIH3PP/+8HnvsMW3cuFHnnHOOpkyZkrIFy3Z5eXl6+eWXde6556qkpEQPPvigZsyYofPPP1+9e/dW69atNWjQIE2ZMkX33nuvysrKNGbMGP3yl7+UtPWk/OGHH2rcuHH6/PPPdfLJJ+vWW2+N6+ECjQ7nUCCZWrZsqauuukrZ2dmaOHGiHnroIRUUFOjaa6/VT3/604zua/jw4TrvvPPUrFkz3X333Xr66aczev9ApjT4O6Dbvytmuemmm3TTTTfV/PvKK6/cKZ6fn69//OMfNf++/PLLJUkHH3xw5iYJNFF7yk/Jri5bXFysRx991By3Y/GFXT3xxBN73C8AzqFA0jVv3ly33XabGR85cuRu59Fp06bt9O/x48dr/PjxNf++4YYbalqy7CgrK0uPPPJIneYLxKHBF6B19cADD+jpp5/WiBEjNHfuXD388MPKysrSD37wg4aeWg2v/YP1Of0VK1aYY6ZPn27Gtn+PJ5Vd+03tyCvV7ZXJttqwdO/e3Ryz/TsKQKZ5ueb54osvzNjEiamr7w8ePNgc06pVKzPmfTfHq/TrHRO8j00efvjhKbf369fPHOPNMSsrMR+cQQY0hnOo5Of23n5fbkdeiwSv8Mv2Ik6pWOdDSWrdurUZ2/5RaSCTvJzxzjXbK2On8re//c2MTZ061YxZbVi8eVhf8ZG2/uLMsmNtil1556/KykozZrUq9Ob43nvvmbHhw4ebsa997WtmzGvf0tjOzY1+Adq7d28tXrxYv/3tb5WXl6ejjz5a1157rYYMGdLQUwMAINE4hwIA4tboF6AnnniiTjzxxIaeBgAAjQ7nUABA3BrX+7UAAAAAgEaLBSgAAAAAIBYsQAEAAAAAsWABCgAAAACIRaMvQpQUXrlrL2aVe589e7Y5xmvHMGDAADN2wAEHmDGvfPzq1avN2LJly1Ju90pkeyXnM11qH9jOe/14uWG1a/BKrHsKCgrM2MaNG83Y22+/bcb2228/M9azZ8+9mxiQYOm0WJoxY4YZmzJlihn79NNPzdiXX35Z63lI0le/+lUzdvXVV6fcXh/tWTjHNh3e/7XXP/iVV14xY2+88YYZq6ioMGPpvLa8Fi1eqzCr5YsklZWVmbGqqiozZrV2WbVqlTnmxRdfNGPeccS75h45cqQZ81o9JTG3eQcUAAAAABALFqAAAAAAgFiwAAUAAAAAxIIFKAAAAAAgFixAAQAAAACxoApuLaRThU+yq2hK0ty5c1Nu96p/eZVu27dvb8a8KlgtWrRI6z6tqp1eZbA2bdqYMa+aJ7An6eao95osLCxMuf1Pf/qTOWbixIlmrEuXLmbMq0Y9ffp0M+ZVv1u6dKkZSwdVNFFf0n1tzZw5M+X2//qv/zLHLF++3Ix17tzZjOXm5pqxHj16mDGvsmiHDh1Sbv/3f/93c0y6xzqPV1k0K4v3KxqbTZs2mbHPP//cjHnXb127djVjXvVW69zm5bUXa9asmRnzrlm96vXNm9tLonXr1qXcXlxcbI55//3304q1bdvWjHkV+/v06WPGLA15zuaIAgAAAACIBQtQAAAAAEAsWIACAAAAAGLBAhQAAAAAEAsWoAAAAACAWLAABQAAAADEok5tWEII8yStlrRZUnUURQMzMamkSrdE/Pz5882Y1SLh4IMPNsd4Jaa9UunplrvOycmp9VwqKirMMV5bGq8dDC0eaq+p5Wi6rxGv/Yn1mrTas0hSSUmJGZsxY4YZy8/PN2Ne3q9du9aMffrppym3f/3rXzfHpHscQe01tRz1Wn14r605c+aYMasl0uLFi80xd911lxk79NBDzZg3x7KyMjM2efJkM/bEE0+k3D5q1ChzzGGHHWbG6qNVUlNuv9RYc3TDhg1mLN32XC1btjRjrVq1MmMrVqxIud1rp+Kx7k+SunXrZsa6d+9uxj7++GMzZrWt8doyeTGv5aB3Hew9bi9Hk9hGKRN9QI+JoqgyA/cDoH6Qo0CykaNAspGjQAYlb0kMAAAAANgn1XUBGkn6RwhhagjhslQ3CCFcFkL4IITwgfexTAD1ghwFko0cBZKNHAUyrK4L0KFRFPWXdKKkH4QQhu96gyiK7omiaGAURQOLiorquDsAtUSOAslGjgLJRo4CGVanBWgURWXb/iyXNEHSoExMCkBmkKNAspGjQLKRo0Dmpb0ADSHkhRBabf+7pNGSPsvUxADUDTkKJBs5CiQbOQrUj7pUwe0gacK28tvNJT0aRdELGZlVA/JKxHsljisr7eJoXmn2Ll26pNwed6sVj7e/vLy8lNu9UtFezGtDUR+PbR+3T+aoJ93Xgdc2ZcmSJSm3Z2dnm2OaN7cPre3atTNjXml273tFa9asMWOffZb6WindtgpNuR1DPdgnc9R7jXgx7/z72GOPmbEPP/ww5Xav1UR9tCPx2r547dh69+6dcvs999xjjrnmmmvMmHf94D3HtF9KKdE56r0eV61aZcas85rkn0+814/Xmqyqqirl9nXr1pljvJj3Gu/cubMZO+igg8zYe++9Z8asY8lHH31kjlm5cqUZS/caf9OmTWbMayeXbrub+pT2AjSKojmSvprBuQDIIHIUSDZyFEg2chSoH7RhAQAAAADEggUoAAAAACAWLEABAAAAALFgAQoAAAAAiAULUAAAAABALOrShmWf5JUa98ofz5s3z4y1aNHCjHXt2jXldq8dg1eiOW7WY8vJyTHHeKXxG1sZaTQe6ea2ZePGjWnNw3sde+Xv165da8bSaZvildP35kgbFuxJuq+RadOmmbHXX3/djFm56LVx2Lx5sxnzeOcor41DmzZtzNh3vvOdlNsfeeQRc8wLL9idQC688EIz5vH+3zzkfcPxXo9eqxWvrdfq1avTinmKiopSbk+3HYwXmzlzphnzjgnedfeGDRtSbvdaL3nXCN7jnjNnjhnzrgMam+SsZAAAAAAA+zQWoAAAAACAWLAABQAAAADEggUoAAAAACAWLEABAAAAALFgAQoAAAAAiAVtWHbhlU0uKSkxY15J6F69epmx3NzclNuTVNbcm4sVa9myZVr7WrVqlRmzynhLtIbAnnmvkYKCAjNmtRpav369OcZrQ+QdY7z79FqjeC1VrMeWbjsn8gl7ku5r5OWXXzZjK1euNGOdOnVKub28vNwc06pVKzNWVVVlxry2F88995wZ844Jp512Wsrtn376qTnmwQcfNGN9+vQxY1/72tfMmPfYvOOndWzyXgdJaifXmHnnk0WLFpkxr52Kd5+lpaV7N7Fd5OXl1XqMlzNejj7zzDNmzGu/lJ+fb8aaN0+9XPKudb3n0Wun4l0HeDnqXQckEUcAAAAAAEAsWIACAAAAAGLBAhQAAAAAEAsWoAAAAADw/9m79zir6nr/4+8vCMNtYMAZ7shNQBFMFLFUvCMeL6lYpqVhWWaP7GTlMX+aKSfz4TnlLS0T81rq0UhSSjMvycUbAiqCCgJyH2CG+30Y+P7+YCCw/fkMs9izZg28no8HD2B95rv3d++9Pmut7+y9Px+kggUoAAAAACAVLEABAAAAAKnYL9uweOXEly9fbsZmzZplxrwWIW3btjVjScrVZ6nliHV/BQUF5hirnLUkbdiwocb3BeyQNDesdkiSXy49yRivDLxXdj7p/XXt2jXndq8NQpaOMcgPb7+zXu+k+4inrKzMjE2aNMmMnXjiiWZs5cqVNZ6Htx97LY+8FglvvvmmGbvsssvMWLt27XJu/8Y3vmGOue2228zYqFGjzJjXFq5NmzZmLMkxwXuO61vLiKzy2pF4bYi8dh4er32L93pbxx9v/t4+t2XLFjO2atUqM+bNsXHjxjW+P+++vPaMXssXb47eNbL3XCa9tqhNvAMKAAAAAEgFC1AAAAAAQCpYgAIAAAAAUsECFAAAAACQChagAAAAAIBUVLsADSE8FEJYFkKYtsu2NiGEl0IIn1T93bp2pwnAQo4C2UaOAtlGjgLp2pM2LI9IulfSY7tsu07SKzHG20II11X9/yf5n17t8EoVe+WPe/fubcaKi4vNWJMmTfZsYrtIWuI+aWnzfLdW8MrYJ23D4rWaaN68+Z5NbN/0iPaxHE2bV37d4rW18NpXePuql79Jy9Vv3rzZjCE1j6iOczRJSxXvvJD0nPHkk0+asXfeeceMXXTRRTW+zZYtW+75xHbhtah4//33zZj3HBcVFZmxsWPH5tzutbzo0KGDGXv55ZfN2Ny5c83YUUcdZca89i3HHntszu0dO3Y0x2SwndMjqofnUa8tUGlpqRnzrqe8c0ajRo0SxazzZdJ2MF4bE+88mqStmmTvr97teefs1q3t32V418j7kmrfAY0xjpO04jObz5X0aNW/H5V0Xp7nBWAPkaNAtpGjQLaRo0C6kn4HtF2MsVSSqv5um78pAcgDchTINnIUyDZyFKgltV6EKIRwRQhhUghhUllZWW3fHYAaIkeBbCNHgWwjR4GaSboAXRpC6CBJVX8vs34wxjgyxjgwxjiwpKQk4d0BqCFyFMg2chTINnIUqCVJF6DPSRpe9e/hkp7Nz3QA5Ak5CmQbOQpkGzkK1JJqSy2FEJ6UdJKk4hDCQkk3SbpN0tMhhMslzZf05dqcZL6tWPHZ75n/y9SpU82YV9nVq4RVWFhoxqyKmF7lXK9ipxdLUgUxKa8amvc8zp8/34x5VQT35yq4+2KOps2ryGjloldpz8s1r+pfQUFBottcs2aNGZs5c6YZs2SwQmW9llaOesdx77y3ePHinNu7d+9ujvEqbHqxN954w4x5570pU6aYsYULF+bc7lUIveKKK8zYqlWrzNiWLVvMWPv27c3YY489ZsZ++9vf5tzuHWO819o7x3qVdT/++GMz5r02VhcAr1L4yJEjzZj3PNaW+noe9Y79Vl5LfoVWbx/x9gNvn7TObd71sXdd3a5dOzPmVXoeMGCAGfv000/NmHdMsHjXwd5xxDtueedm7xohi6pdgMYYLzZCp+Z5LgASIEeBbCNHgWwjR4F01a/lMgAAAACg3mIBCgAAAABIBQtQAAAAAEAqWIACAAAAAFLBAhQAAAAAkIpqq+Dui2bPnm3G7r77bjO2ZMkSM9ahQwcz5pUU79SpU87tXolp7/a8BsitWrUyYwccYO8KXtltq/2Jd3te+emWLVuaMa+0vFfu3RuHfUvSdkJeifWlS5fm3O61U/HaJ3i8EuteTnnjrHwjZ/Y9ixYtMmNDhgwxY1a7AK+9lddqxWsj0KZNGzPmnWtGjRplxqz2Y82aNTPHeG1pvHZIXq55LTG8/O3atWuN78s7xnjtzMrKysxYt27dzJh3TFi5cmXO7V47jKeeesqMefsIdue1TCkvLzdjmzdvNmPW6ykl3yet+/Ou+bzjyIIFC8yY177liCOOMGObNm0yY8uXL8+53TvGHHTQQWZs9erVZszLUe9aJd/tFGsb74ACAAAAAFLBAhQAAAAAkAoWoAAAAACAVLAABQAAAACkggUoAAAAACAVLEABAAAAAKnYL9uwrFu3zox5rVa8EvdeSXqv7UjSdg1JJG3t4rWY6dixY87tXvl+rzR4UVGRGbvgggvMWOvWrc0YLSX2H0lfay/vCwsLa3xfXouTBg2S/d7PK5vvtY2wjk1ei4QePXqYMa/Ue9LHhj1XWVlplun3jq1JLFu2zIx57QesnJH8868X827TOo96rQ68nPFaplgtXyS/RcWGDRvMmNUiLenxzDufW22lJL/9hpfb1nOZ9PnwXhvszjvXeG1MvFzzrku918bbR6zzkJdPSc+xVlspSXrppZfMmNeGxbqO99obHnzwwWZs6tSpZsw713vPsfdcZhFXCwAAAACAVLAABQAAAACkggUoAAAAACAVLEABAAAAAKlgAQoAAAAASAULUAAAAABAKvbZNiyVlZVmzCtN7ZU4btq0qRlLWi7dKnddUVFR4zGS3yLBKt0v+eX2P/jgg0RzsXivzdFHH23Gzj77bDPmlc0HJL/s/7x588yYtY97ZeC9/dFry+QdR7yYl1NWbk+aNMkc07NnTzPmtVZA7duwYYPefffdnLGkbcT69euXc7vXzsA79nvjvH111apVZsw7f1lz8doSeOfY9evXmzHvOsA7JnjHn/nz5+fc3rJlS3OMl4feuE6dOpmxFStWmDGP1QbN2k8l6dBDDzVj3hz3V9a+tXz5cnOM12rFO594LZY83rW1lTdeyxHvWJG0LZnHa1FkHUu8vF68eLEZ89YaXm57LQ5pwwIAAAAAQA4sQAEAAAAAqWABCgAAAABIBQtQAAAAAEAqWIACAAAAAFJR7QI0hPBQCGFZCGHaLttuDiEsCiG8V/XnzNqdJgALOQpkGzkKZBs5CqRrT3pXPCLpXkmPfWb7nTHGX+V9RnmyadMmM7Z27Voz5rVP8FqceDGvJLRVitkr9e7xyt97Zau9kvpJ274kGdOuXTsz1qJFCzPmlbTeDzyiepijafNarXhtiKyc8loueDnjjUu6H3vjrHz78MMPE90XEnlEecrREILZQsHbD7yWBk2aNMm53dtXV69ebca8dh7WfUlSUVGRGfNYrSi8c73HO2d7bRe8c5QXs9oueNcj3mvjtXryeK05vGOa9Tx7+6PXcq1Hjx5mrBY9ojo+j3rXRtY+vmDBAnNM27ZtzVh5ebkZ8/Yfbz9Ys2aNGbNahHiP2btm7dq1qxnz2jl16dLFjHnPiZVvXssUr/WV1+rGu03vGr++XQdXO9sY4zhJyZpDAah15CiQbeQokG3kKJCuvVkuXxVCmFr1sYXcXYgB1CVyFMg2chTINnIUqAVJF6D3Seop6QhJpZJut34whHBFCGFSCGFSWVlZwrsDUEPkKJBtiXLU+3gZgLziPArUkkQL0Bjj0hjj1hjjNkkPSBrk/OzIGOPAGOPAkpKSpPMEUAPkKJBtSXM06fckAdQM51Gg9iRagIYQOuzy3/MlTbN+FkD6yFEg28hRINvIUaD2VFsFN4TwpKSTJBWHEBZKuknSSSGEIyRFSXMlfacW5wjAQY4C2UaOAtlGjgLpqnYBGmO8OMfmB2thLnnllUpfsmRJonEVFRVmLGn5Y6vMulcO3Stb7ZXat8pgVzfO483T4pXx9kr0e6Wp92f1NUfT5rWN2LhxoxlLst8lbV3k5aE3zmvXYN3m0qVLE90XeVhz+czRyspKs83JypUrzXFea7J169bl3O61+vDOh0n3cS9vvHNsYWFhzu3evurd3le/+lUz9sc//tGMeW0jvPYJ1vPcvHlzc4zXqs17bazXWvKPg94xxro/63WR/Odq/fr1Zqy2ZOE8mqTFiddCzIt53yX35uHtB941pnWt6O0HSduRePuxd0zzWqNY3+stLi42x3h5mPRxe+rbebt+NY0BAAAAANRbLEABAAAAAKlgAQoAAAAASAULUAAAAABAKliAAgAAAABSwQIUAAAAAJCKatuw1Fdr1641Y7NmzTJjXtnqpG1YtmzZUuNx3u15rU+8mHebXmltL2aV6/bKcRcUFJix1q1bmzHvNoHqlJaWmjGvJLqVN0lLnnt56N1m0tYu1ly8MvxWyX9JatWqlRlD7auoqNCiRYtyxrzzl7ePlJeX59zutcPwzodeiwSP16LFO7dZ+7g3x65du5qxc88914x5bUyee+45M+bNxWo/5h1HvNfTayfhtW9J0mpFsq9xvPP5ggULEt3X/sra/9u0aWOO8fYfL7e9fcS7nvX2O4t3re7t4951vHeOOuigg8zYJ598Ysas86/3HCdtfdKyZUszlrTFVdJjcm3iHVAAAAAAQCpYgAIAAAAAUsECFAAAAACQChagAAAAAIBUsAAFAAAAAKRin62C61X48ioFelWkklao9GKWpJW1klbBTXqbVtW8pNW42rZta8ZatGhhxoDqrFixwox5VXCtfdyrBujt/14eeuM8SfLXez6sKqsSVXCzYOvWrTm3e+cor8KpdUz2qmF61cyTnketx1XdOO+xWbycefjhh83YtGnTzJj3nHiVXTdt2pRzu/eYvfOoVVVX8l9Tz8aNG82YNc/CwkJzjPd8bNiwYc8ntg/xXm/rmnb16tXmGO959O7LO0d5t+l1KrBy1DsPeTnqXVd71Ze9vElyTvfWGl4eescKb5x3f16OUgUXAAAAALDfYgEKAAAAAEgFC1AAAAAAQCpYgAIAAAAAUsECFAAAAACQChagAAAAAIBU7LNtWLxyxF7ZZ49XEjpNXolsj1diOun9WaXUvVLRXlsXr8VD0vLxgOTvk0nKvSdtJ+HxbjPpOOuxrVmzxhzjtarq27fvnk8MeRdCMEvqe6X2vX3S2keStkHwePPwzjXeecNqkeC1M1iyZIkZGz16tBnznhPvHJVk/l57KK/VjdXWRfLbnyRtLWWdt5cuXWqO8Vo9eY97X+Y97pkzZ+bcPnHiRHNMWVmZGfPOGd71s7dvJWln5h1HkrYVXLVqlRl77733zJjXYibJMdK7nvWeqzRbPtYl3gEFAAAAAKSCBSgAAAAAIBUsQAEAAAAAqWABCgAAAABIBQtQAAAAAEAqql2AhhC6hBD+GUL4KIQwPYTwg6rtbUIIL4UQPqn6u3XtTxfAZ5GjQLaRo0C2kaNAuvakDUulpB/HGKeEEAolTQ4hvCTpMkmvxBhvCyFcJ+k6ST+pvanWzLp168yYV5o6K7wyzF45dK80tXebXrnrJGWyk5T8l6TCwkIz1rhxYzO2n6uXOVobvP2uvLzcjHml5a18q41y6ElzO0mOem1YFi5caMaQSN5ytHHjxurSpUvOmNXqQErWqicpbz/2zjXePNJsg+a1T6iNVgdW+xzv9fR4x7OVK1eaMe+1admypRmz2ldccskl5pgvfelLie6rFtX5edQ7f1mtsbzr2aT7qtciJGnbFKtFkdf6pFmzZmbMyw2vjczy5cvNmPfYCgoKcm5v3dr+fYQ3f+/86x2rW7RoYca85z+Lqn0HNMZYGmOcUvXvtZI+ktRJ0rmSHq36sUclnVdbkwRgI0eBbCNHgWwjR4F01eg7oCGEbpIGSHpbUrsYY6m0PXEltc335ADUDDkKZBs5CmQbOQrUvj1egIYQWkj6s6SrY4z2e8f/Pu6KEMKkEMKk+vDRV6C+IkeBbMtHjnof3QKwdziPAunYowVoCKGRtifk4zHGZ6o2Lw0hdKiKd5CU8wPqMcaRMcaBMcaBJSUl+ZgzgM8gR4Fsy1eO1tH344B9HudRID17UgU3SHpQ0kcxxjt2CT0naXjVv4dLejb/0wNQHXIUyDZyFMg2chRI156UWDtO0qWSPgghvFe17XpJt0l6OoRwuaT5kr5cO1MEUA1yFMg2chTINnIUSFG1C9AY4wRJVv3zU/M7nZqzyr2vXr3aHOOVaPbKH3sxT5rl45OW3c53afmkz1Xz5s3NmFUGe3+X9RxN0/r1683YokWLzJjXtiBpKwRL0nYqSVssWbzjYGlpaY1vD7Z85mjz5s01aNCgnLH58+eb47zjp9V2wdvnvFjSViteruW7xYDXKsY7Hng56j027/m3xnk56n3Ms0+fPmbMauEjST169Eh0f9Zzcs4555hj2rRpY8bSvGbaIQvnUW+/s1pjeec8j9fyLmluePlrvabeuauiosKMbdmyxYwlPY96bXCs9ife8+jx9nHvOti7P6udU1bVqAouAAAAAABJsQAFAAAAAKSCBSgAAAAAIBUsQAEAAAAAqWABCgAAAABIBQtQAAAAAEAq8ttfoA5Y5ePLysrMMV5p56RtWLySyknH7au8cvTNmjUzY/luh4F9z5o1a8xY0nL1SfY7L+e9Uu9eLEkbB28uXqn98vJyM4a61ahRI3Xq1ClnrGnTpua4iy++2Iz97W9/y7ndOx57GjdubMa83PDaJ1jnesluDeG1XFi3bp0Zs55fyW9j0rt3bzN2yCGHmLEOHTrk3O49H14bk7Zt29b4vqobl6QNGtc+NeNdmy5evDjndm8f8Vp2dO3a1YxZLV8k/zX1zilWa0RvjBfz5uHt417+rlixosb3543x9nHvsVktXyT/mFzfcop3QAEAAAAAqWABCgAAAABIBQtQAAAAAEAqWIACAAAAAFLBAhQAAAAAkAoWoAAAAACAVNT7vhZW+fVly5aZY5KWBm/YsGGicUlbu6Qp3y1mvDEtW7Y0Y02aNEk0D0Dy27B4Je69dg1W3ifNGa9svhfz5ui1b7Fu05v/hg0banx7kj9H1L5WrVqZsf/8z/80Y1aLhLfeessc4+1z1nlZ8tufeG0jDjroIDN25JFH1mi7JBUXF5sxr8VJ0vYn3v15bReSSNrqxpNkXNLrov1VkvOG1wakc+fOZqxXr15mzMvRpUuXmjGPdf71rquTtmjxzkPeNaa3T65du9aMJbkvr62Rd6ygDQsAAAAAADXEAhQAAAAAkAoWoAAAAACAVLAABQAAAACkggUoAAAAACAVLEABAAAAAKmo921YrNLO5eXl5hir5Lzkl49v1KiRGUvSxsGLeWMOOCDZy5a0bUSSUurec+WVsW/atKkZA6rjlYhfuXJlotv02k1YkpZD93K7cePGZmzz5s1mzDpGem1pZs+ebcYWL15sxryy/6h93nnjsMMOM2M33HBDzu1PP/20OWbevHlmrF+/fmasZ8+eZqx9+/ZmzGvRUlJSknN72u0MkrZcs44x3nVF0jmm2SqpNp7HfVnz5s3NWP/+/XNuf/vtt80xXj4N7JyDRAAAIABJREFUGjTIjHXt2tWMffTRR2bMu36eMWNGjcesX78+0X15bRi960+PNa5Hjx7mGG/+XszbD7xjfH1rg1a/ZgsAAAAAqLdYgAIAAAAAUsECFAAAAACQChagAAAAAIBUsAAFAAAAAKSi2nKqIYQukh6T1F7SNkkjY4x3hxBulvRtSWVVP3p9jPH52pqoxaqMN3ToUHOMV9XSq/64fPlyM7Zu3ToztnHjRjO2YcOGGm2XpFWrVpkxr3qcV2HTq6zlVbS1Yi1btjTHdOjQwYy1aNHCjCG3rOdomrxc82JeZdckVWS9Y4yXo1412yVLliQa16RJk5zbW7VqZY7xqifurxUq90YWctSrovy5z30u53avGqaXT97x36s+651r6oN8V5NPWkU26bisqIv5ZyFHveqn559/fs7tffr0MccUFBSYMa8qtpejXt5v27atxuO8MZWVlWbMO/8uXLjQjHnnL2+cdf3sdW4YM2aMGXvjjTfMmHes9u7Pu47Poj3p51Ep6ccxxikhhEJJk0MIL1XF7owx/qr2pgdgD5CjQLaRo0C2kaNAiqpdgMYYSyWVVv17bQjhI0mdantiAPYMOQpkGzkKZBs5CqSrRt8BDSF0kzRA0o7Ot1eFEKaGEB4KIbQ2xlwRQpgUQphUVlaW60cA5Ak5CmQbOQpkGzkK1L49XoCGEFpI+rOkq2OMayTdJ6mnpCO0/bdGt+caF2McGWMcGGMcWFJSkocpA8iFHAWyjRwFso0cBdKxRwvQEEIjbU/Ix2OMz0hSjHFpjHFrjHGbpAckDaq9aQLwkKNAtpGjQLaRo0B6ql2Ahu3lyB6U9FGM8Y5dtu9ayvR8SdPyPz0A1SFHgWwjR4FsI0eBdO1JFdzjJF0q6YMQwntV266XdHEI4QhJUdJcSd+plRlWo0GD3Gtor51H27ZtzZhX2rm4uNiMeW0X8s0rI71p0yYz5rVvWb9+vRnzWsJYpbW9cuLdu3c3Y61b5/x6BXyZztE0nXzyyWbs8MMPN2Neblil2RcsWJDo9rwc9Vqj9OjRw4x5xzsrp4qKihLdnjcOpkznqNViwDvnIX/qe9uUfUSd56jXRqNz5845t3fp0iXv92VdV0vZOf5718H9+vUzY17blyTX8d65fs2aNWbMe/779+9vxrz1S307juxJFdwJknI9qn26nyBQX5CjQLaRo0C2kaNAumpUBRcAAAAAgKRYgAIAAAAAUsECFAAAAACQChagAAAAAIBUsAAFAAAAAKRiT9qwZJpVitlrdeDFPFap+upiXmlkqxSzN8Yrkd2yZUszVlJSkug2vblUVlbm3N6xY0dzzLnnnmvGvBLTQHWaNm1qxjp16mTGCgsLzZh1jPFaNjVq1MiMWa2LJKlNmzZmzCu37+V9kjYsAIB0ea1FrGs079pzX+Zdl6b5nDRp0sSMfetb3zJjX//6182Y1w7Gu7/6hndAAQAAAACpYAEKAAAAAEgFC1AAAAAAQCpYgAIAAAAAUsECFAAAAACQChagAAAAAIBUBK/sc97vLIQySfOq/lssqTy1O/dlZS7MY3dZmYeUn7l0jTHavXAygBytFvPYXVbmIZGjdSkr85CyMxfmsbt8zYMcTSYr85CyMxfmsbtUczTVBehudxzCpBjjwDq588/IylyYRzbnIWVrLmnJ0mPOylyYRzbnIWVrLmnJymPOyjyk7MyFeWRzHmnLyuPOyjyk7MyFedTtPPgILgAAAAAgFSxAAQAAAACpqMsF6Mg6vO/PyspcmMfusjIPKVtzSUuWHnNW5sI8dpeVeUjZmktasvKYszIPKTtzYR67y8o80paVx52VeUjZmQvz2F2q86iz74ACAAAAAPYvfAQXAAAAAJCKOlmAhhDOCCHMCCHMCiFcVxdzqJrH3BDCByGE90IIk1K+74dCCMtCCNN22dYmhPBSCOGTqr9b19E8bg4hLKp6Xt4LIZyZwjy6hBD+GUL4KIQwPYTwg6rtqT4nzjxSf07qEjlKjuaYBzmaIeQoOZpjHuRohuzvOZqV/HTmQo7WYY6m/hHcEEJDSTMlDZG0UNI7ki6OMX6Y6kS2z2WupIExxtT774QQTpC0TtJjMcZ+Vdv+V9KKGONtVQer1jHGn9TBPG6WtC7G+KvavO/PzKODpA4xxikhhEJJkyWdJ+kypficOPO4UCk/J3WFHN153+To7vMgRzOCHN153+To7vMgRzOCHM1OfjpzuVnkaJ3laF28AzpI0qwY45wYY4Wk/5N0bh3Mo07FGMdJWvGZzedKerTq349q+85QF/NIXYyxNMY4perfayV9JKmTUn5OnHnsT8hRkaM55kGOZgc5KnI0xzzI0ezY73M0K/npzCV15Oi/1MUCtJOkBbv8f6Hq7sAUJf0jhDA5hHBFHc1hV+1ijKXS9p1DUts6nMtVIYSpVR9bSOUjEjuEELpJGiDpbdXhc/KZeUh1+JykjBy1kaMiRzOAHLWRoyJHM4AczS1L+SmRo3WWo3WxAA05ttVVKd7jYoxHSvoPSd+reose0n2Seko6QlKppNvTuuMQQgtJf5Z0dYxxTVr3uwfzqLPnpA6Qo9lHjpKjn0WOZgs5So5+FjmaLeRoHeZoXSxAF0rqssv/O0taXAfzUIxxcdXfyySN1vaPTNSlpVWfy97x+exldTGJGOPSGOPWGOM2SQ8opeclhNBI2xPh8RjjM1WbU39Ocs2jrp6TOkKO2shRcjQLyFEbOUqOZgE5mlsm8lMiR+s6R+tiAfqOpF4hhO4hhMaSLpL0XNqTCCE0r/rirUIIzSWdLmmaP6rWPSdpeNW/h0t6ti4msSMJqpyvFJ6XEEKQ9KCkj2KMd+wSSvU5seZRF89JHSJHbeQoOZoF5KiNHCVHs4AczS0T+SmRo7nmkepzEmNM/Y+kM7W9OthsSTfU0Rx6SHq/6s/0tOch6Ultf3t7i7b/puxySQdKekXSJ1V/t6mjefxB0geSpmp7UnRIYR7Ha/vHU6ZKeq/qz5lpPyfOPFJ/TuryDzlKjuaYBzmaoT/kKDmaYx7kaIb+7O85mpX8dOZCjtZhjqbehgUAAAAAsH+qi4/gAgAAAAD2QyxAAQAAAACpYAEKAAAAAEgFC1AAAAAAQCpYgAIAAAAAUsECFAAAAACQChagAAAAAIBUsAAFAAAAAKSCBSgAAAAAIBUsQAEAAAAAqWABCgAAAABIBQtQAAAAAEAqWIACAAAAAFLBAhQAAAAAkAoWoAAAAACAVLAABQAAAACkggUoAAAAACAVLEABAAAAAKlgAQoAAAAASAULUAAAAABAKliAAgAAAABSwQIUAAAAAJCKA/ZmcAjhDEl3S2oo6fcxxtu8ny8uLo7dunXbm7vMrMrKSjO2bt26Gt9eCCHRPNauXWvGDjzwQDNWUVFhxrZu3WrGGjZsmHN7q1atzDH7q8mTJ5fHGEvSvE9ydO/FGGs8ZtWqVWastLTUjB1wgH1I9o4JXszKXy+vvdvzjiMlJfbunfSYliZydN/i5dqWLVvMWMeOHc2Yl6MbN240Y/PnzzdjBQUFObcfdNBB5pgGDfbP9w/I0fxIcl6T6sdxHHVrT3M08QI0hNBQ0m8kDZG0UNI7IYTnYowfWmO6deumSZMmJb3LTCsrKzNjr7/+eo1vzzohSf6F4/jx483Y1772NTPmnRzXrFljxlq3bp1z+xlnnGGO2V8PYCGEeSnfX73NUe/kmHT/8W5z27ZtZsz65ZJ3e88++6wZ+5//+R8zVlRUZMYaNWpkxho3bmzGrPz1flnlPcfDhw83Y9/97nfNmDf/JLzXLOnFOTm6b/n5z39uxrzF6c0332zGvF+yfPDBB2bsqquuMmMHH3xwzu2//vWvzTEtWrQwY/sycnR3SReS3i9gPN5xfH+9tsPu9jRH9+ZXaIMkzYoxzokxVkj6P0nn7sXtAcgvchTINnIUyDZyFKgFe7MA7SRpwS7/X1i1bTchhCtCCJNCCJO8dwkB5B05CmQbOQpkGzkK1IK9WYDmeq/93z4LEGMcGWMcGGMc6H10BUDe1XmOXnbZZQoh7PxI280336wQgi677LK83g9QT9V5jgJw1cscnTt3rho0aKAGDRq4NQl2NWTIEBUUFOixxx6r5dkBe7cAXSipyy7/7yxpcU1uoFu3bgohmH9ee+21vZgesN/Le44WFxdr6NChfAcNyI+85WiDBg3UokULdevWTRdeeKHefvvtvE4U2E/tdY5K0pw5czRs2DC1bdtWTZo0UefOnXXGGWdo9uzZeZsoUJ/sTRXcdyT1CiF0l7RI0kWSvlqTG/jmN7+pFStWSJLuu+8+VVRU6IILLlDnzp0laeffu6qsrHSr0NWVrM4L+7W9ztEdzj77bHXv3l1jx47VP/7xD73zzjv6+OOP1bZt23zONxVbtmzJe0EcIKG85ehZZ52lkpISvf766/rTn/6k0aNH64knntCXv/zlf/tZcgDYY3nJ0fPPP19Tp07VKaecot69e2vhwoUaN26cSktL1bNnz3zPGci8xCumGGNlCOEqSS9qe2nqh2KM02tyGz/72c92/vuRRx5RRUWFrrrqKp100kmSpEsuuUSPP/64rrzySs2cOVPjx4/Xq6++qkGDBul///d/9Yc//EELFy5U165ddeWVV+qqq65SgwYN9NOf/lS/+MUvdPnll+v3v/+9Kisrd55sFyxYoM6dO+v222/Xb37zGy1atEgtW7ZUv379NHLkSPXq1Uvr16/XiBEj9Mwzz2jJkiXq06ePbrrpJn3xi1805/XMM8/o85//fM7HuWnTJvM5sCqReRXKvCq4XmXL8vJyM+ZVBFy+fLkZ27BhQ87tq1evNsd4lT6RP/nI0R0uv/xynXfeeSovL1dJSYlWrlypN998U+edd566du2quXPnSpJOOukkjR07Vg8//PAef8x29OjRuvXWW/Xxxx+rTZs2Ovvss/WLX/xCRUVFOvHEEzVu3Dj9+c9/1rBhw3a7jyeffFLDhg3T66+/rptvvllTp05VQUGBTj31VN18881q06aN5s+fryOPPFKSdMstt+jXv/61evTooSeffHK3OcyZM8ecn1XF2mtHMmPGDDPm5Wjz5s3NmPcLLu94YcW8KrJNmzY1Y171TastE3KrjRytrKzUpZdeqv/7v//TlVdeqbPOOmvnfnXnnXfq7rvvVghBc+bM0fz583Xttddq/PjxWr9+vQYNGqQ77rhD/fr1kyTddddduueee7Ro0SIVFhbqsMMO0/33368+ffroiSee0C233KJPP/1UTZs2VZ8+ffTLX/5Sxx9/fJ6enT23ZMmSnNu9c5fXHs2rCu+d25YuXWrGvHwrLCzMuZ2qonUvHzm6YsUKTZ06VUVFRXr55Zd3vq6bN2/W1q1bVVpaqq985Sv68MMPtXr1ahUVFWno0KG69957VVRUpLlz56p79+6SpN///vcaMWKE1q5dq+HDh+uOO+6QtL2V3tVXX62nnnpKBx54oK655pqd979x40YVFBTouuuu07PPPrtzP+3Zs6euvPJKDRgwQNK/KqZPmzZN//jHP3I+Fus6V5KKi4tr8rRUqzaq4SM79qqRVIzx+Rhj7xhjzxjjL/I1qc/63e9+p23btumSSy5RYWGhrrvuOt14441av369LrroIi1btkw/+MEPdPvtt+/R7X388ce65pprtG7dOn3jG9/Qaaedprlz5+5Myssuu0y//OUvVVRUpGHDhmnevHk677zz/q3Fya7z2l9LoiPb8pmj27Zt09ixY3f+f9q0aXs9vxdeeEHDhg3T1KlTNWzYMBUWFuq3v/2tLrroIknS17/+dUnSU089JWn7Bd748ePVunVrnXXWWZo+fbrOOOMMTZkyRaeffrp69uypJ554Qt/85jf/7eT1q1/9SieddNLOBSmQBfk+jx5wwAG66aabJG2/8N31FyjXX3+9TjjhBJ1++unasGGDTjnlFD399NM6/PDDNWTIEL322ms6+eSTVV5erlmzZumHP/yh1qxZo8suu0xDhgzR/PnzVVpaqo0bN+qyyy7TvHnz9LWvfU1nnXWW1qxZw8cJsU/a2xwtLCxUixYttGrVKg0YMEA/+tGP9Je//EWVlZVq1qyZ1q5dq40bN+qcc87Rt7/9bbVu3VqPP/64rrvuun+7rREjRmjw4MFas2aN7r77br3yyiuSpFtvvVX333+/GjRooMGDB+u///u//23svHnzdPTRR2v48OEaMGCAPvzwQ914443mGwlAbaoXnxk9+eST9eqrr0rafhF83333Sdp+UXrcccfpzDPP1Je+9CXdc889+q//+q9qb2/HOwKdO3fWBRdcoL59+6pTp047fxM1atQoNWzYUMcee6waNGigvn37avz48br//vs1ePDgnPPy3l0E6rvzzz9/t/+fc845+sIXvrDXt3vPPfdI2n5hfNNNN6m8vFwdOnTQiy++qJkzZ+rLX/6yvv/97+uvf/2r1q9fr1GjRmnbtm264IILVFBQoPvvv18VFRU64ogj1LZtW7Vu3VqTJ0/WhAkTNGvWrN366f7mN7/Rscceu9dzBrKua9euO/+9bNmynf++99579c1vflOS9Kc//UmzZ89Wp06d1KdPH0nSQQcdpNmzZ2vUqFE68cQTJUkdO3bUsGHD1LdvX3Xu3Flbt27Vxo0btXXrVrVt21bnnXee+vbtqx49erifzgH2V40aNdKDDz6oK664Qu+//77ef/993XnnnWrXrp3GjBmjo48+WiNHjtRLL72kZcuW6bDDDtMnn3yy8/pyV6NGjdLRRx+98yO87733nk499VQ98cQTkqQ77rhDl156qcaMGaPzzjtvt7G/+c1v9Je//EXz5s1Tly5d9N5772n16tWaPXu2+vfvn8pzAexQLxagxx133M5/L1myZOdHWg899FBJ0iGHHCJJWrhwYc4T4Ge39e/fXz/72c9077336vTTT995W3/+8593VgvbunXrzovjHWbNmmXOC9iXnX322Tr44IN14IEH6qijjtIZZ5yx27uhO9T0AnTHR3d35HJxcbGKi4u1ZMkSzZs3T71799Z5552nJ598Un/961/19NNPS5IuvfRSSdt/oytJEydO1MSJE3e77Tlz5uy8XUk66qijajQ3oL7akReSdvue9q7nrB25t2jRIt199927jZ81a5auvPJKjRgxQr/+9a81dOhQSVKfPn00atQo9evXT/fdd59GjBihc845R9L2X+j+4Q9/2PkVGgD/cuGFF+qLX/yixo4dq/Hjx+uBBx7Q0qVL9fOf/1wXX3yxvvrVf/9aaa52Ljs+Lrvjq0w7Pkq+aNEiSdr5y6TevXvvNm758uUaNGhQzo+r72mVXCCf9uojuGnZ9V2M9u3bq0mTJpK2f5RW+tf3rTp37qyGDRvu/L7Lrp9n31VlZaV+9rOfafny5Zo7d66uueYaffTRR7rrrrvUrVs3SVKTJk20YsUKxRgVY1RFRYVGjRplzgvYl11++eW688479dOf/lT/8R//oRDCzjzb8Z3GLVu2aObMmTW63R35tiOXly9fvvPTBDvexdnxMdx77rlHEyZM0MEHH7zzeyg7fuYHP/iBNm/erPLycpWXl2vy5Mk7L5p3IF+xP6isrNSIESMkSW3atNlt0blrDuzIvaOOOkrbtm3bea5buXKlbrjhBm3dulU33HCDysvLNW/ePP3kJz/RjBkzdOedd0qShg8frkWLFmnx4sW6++67tXDhQv385z9P74EC9cSWLVs0YcIENWnSREOHDtUtt9yi//f//p+k7efPHV8x+da3vqXNmzfv/H+u70DuqAXw2e9Aduq0vTXpjuvhz56L33jjDS1ZskTFxcWaM2eOXn311Z1fHfO+awnUlnrxDuiuGjRooCuvvFJ33XWXvvKVr2jo0KF69tlnJUlXXXWVpH/9hmjMmDG65pprNGbMmN1uY+7cuTr++ON1wgknqKSkRBMmTJC0/TdKHTp00LBhw/TMM8/o6KOP1mmnnaby8nKNHz9e3//+9/XTn/40xUcLZFfv3r3VrFkzrVixQl//+te1ZMmS3T7utye+973v6YUXXtCtt96qOXPmaPLkyaqsrNSQIUN2/gZ3yJAh6tChw87vsn3ta1/bOf5b3/qWHnroId1777369NNP1bJlS33yySeaOHFizt8eA/uqBx98UM8995xef/11zZw5UwcccIB+97vfqVmzZjl//swzz1SPHj00efJkHXfccTr88MM1f/58vfbaa3r++efVrVs3HXPMMTrhhBPUtm3bnfm3452Xdu3a6aSTTlLHjh31wQcf7BYD8C+bN2/W4MGDdeihh2rAgAFq1qyZRo8eLWn7+W3HJxZeeOEFffe739Xzzz9f4/u4+OKLdcstt+hHP/qRxo4dqxdeeGG3+I5PQpSXl+snP/mJpk+fro0bN+7lIwOSqxfvgH7WbbfdphEjRqhp06Z64oknVFxcrNtvv10//vGPJUlDhw7V9773PRUUFGj06NH6/ve/v9v4oqIiDRw4cOfHIEpLS/XVr35V119/vaTtFXmvvfbanf9+8803ddxxx+38uC4AqVWrVvrd736njh076u9//7t69uzpVsjL5ayzztLTTz+tww47TKNGjdLq1av1ne98Z+dvgKXtlVV3LDpDCLt9VOnwww/XCy+8oMGDB2vChAkaPXq01q1bpx/84Af5eZBAPfG3v/1NTz31lDZv3qwLL7xQr7/+es4WLDs0b95cr7zyii6++GLNnz9fjz76qGbMmKFLLrlEffr0UcuWLTVo0CC9/vrreuCBB7R48WJddNFFO38JO2TIEE2ZMkUPPvigpk+frrPOOmuPCwEC+5MmTZrohz/8oQoKCvT888/rD3/4g4qKinTjjTfq2muv1U033aSTTz5Zy5cv1+TJk3dei9bE9ddfr29/+9vaunWr/vnPf+58h3WHY445Rtdee61at26tf/7znzrttNNUUlKSr4cI1FhI8633gQMHxn21gb1XEfShhx4yY1b5+MaNG5tjdvROzcVr/+AVjVm5cqUZ8z66+NnvGeyQ6/sM1Y2R/FYT9V0IYXKMcWBdz8OTlRzdvHmzGdvxcd1cnnvuOTP22SrWu/K+A2NVCPT6KFZWVpoxr41D0hYn3ndvrd9ye+2h2rdvb8Z2FLHJZcf3AXPxSvR77Wescvu1UaKfHM0uL6e+973v5dy+4zvjuSQ913gFBw866CAztuMjkrlYH13O1Qt9B693JOfRulUbOWod77z2Prt+F/yzdrRvyeXNN980Y1dffbUZ21G4LBdrn/TOedanN7zbQ93b0xytl++AAgAAAADqHxagAAAAAIBUsAAFAAAAAKSCBSgAAAAAIBUsQAEAAAAAqWABCgAAAABIBXWM88RrY+KVi7baHXgtHtq0aWPGBgwYYMY83bp1M2NeSxirhPaaNWvMMV45fUprQ5KWLl1qxrxWK0888YQZ8/atM88804xZLWGeffZZc4xXWv7cc881Y9OnTzdjXmn8wsJCM2a1cvDamKxfv96Mvfjii2bMawM1cKBdlX3w4MFmrEuXLjm3e8dI7Hu89mNWSyGvjYl3HvJaBjVoYP/efubMmWasa9euZmzixIk5t999993mmJEjR5qxdu3ambGkLYqQTV5br7feesuMffDBB2Zszpw5Zszr9funP/3JjLVu3Trndu/YP3ToUDPmtTWqjX2cvMk/3gEFAAAAAKSCBSgAAAAAIBUsQAEAAAAAqWABCgAAAABIBQtQAAAAAEAqKDmaJ1YVPsmvvmlV1vIqVG7bts2MedVDmzVrZsa8qp3r1q2r8VxWrVpljtmyZYsZa9KkiRnDvsXbjxcuXGjGJk2aZMa8PPSqOb/77rtmzGJVx5Xsin+SVFJSYsYOOuggM+bN8cADDzRjp512Ws7tGzduNMf8/e9/N2MffvihGSsrKzNjXhXfyZMnm7HLL7885/Z+/fqZY6imve/xquBu2LAh53avursX844x3rnNy6mXX37ZjI0ZMybn9kGDBiW6L+x7rPONd+z84x//aMY+/fRTM3bIIYeYMa9C+tixY82YdU3rVU73Kk571eS98+HWrVvNmHfd3bBhw5zbqY6bHO+AAgAAAABSwQIUAAAAAJAKFqAAAAAAgFSwAAUAAAAApIIFKAAAAAAgFSxAAQAAAACp2Kta9SGEuZLWStoqqTLGODAfk6qPkpZitlqSeC0qKioqEsW8ktaNGjUyY1b5aUmqrKzMud0rEe+1YUF+ZTlHvZYps2fPNmPz5883Y+3atTNjBQUFZuz99983Y23atMm5/aijjjLHzJw504zde++9Zsw7jhQVFZkxL9+StJhZs2aNGevdu7cZu/DCC83Y1KlTzZhXvr9jx445t3fo0MEc07ZtWzOWtbL5Wc7RLPHaHlntx7y2ZL/61a/M2KOPPmrGvP3YayNWXl5uxg4++GAzZlm0aJEZ89o5eedz5JZWjnotQj755JOc271WK2+99ZYZ+9znPmfGrrvuuhrPQ5Luv/9+M2btr95+/OSTT5qxBQsWmDHv3OzlRvv27c2Y1VotacuvrJ2H6kI+mqWdHGO0j6wA6ho5CmQbOQpkGzkK5BEfwQUAAAAApGJvF6BR0j9CCJNDCFfkY0IA8oocBbKNHAWyjRwF8mxvP4J7XIxxcQihraSXQggfxxjH7foDVcl6heR/9hpArSBHgWwjR4FsI0eBPNurd0BjjIur/l4mabSkQTl+ZmSMcWCMcWBJScne3B2AGiJHgWwjR4FsI0eB/Eu8AA0hNA8hFO74t6TTJU3L18QA7B1yFMg2chTINnIUqB178xHcdpJGV5USPkDSEzHGv+dlVvWQV37da3FitTHxNG7c2Ixt2LDBjC1fvjzRbXqPzeI9riSPGYlkOkdXrVplxqZPn27GlixZYsb69u1rxrp27WrG5s2bZ8astiPnnnuuOeahhx4yY16LlpNPPtnLxIdFAAAf4ElEQVSMnXHGGWbsiSeeMGMTJ07Mud1rVdK5c2cztnbtWjP28ssvm7GVK1easWXLlpmx8ePH59w+ZMgQc8yBBx5oxpKWza8lmc7RLPHaNsUYc2739tWXXnop0TxuvfVWM3b11VebMS/fBg8enHP7hx9+aI7x9mNaPORVajm6YsUKM/bUU0/l3P7888+bY7zrQa8Ny7HHHmvGunTpYsa89l0ffPBBzu3evuqdMx5++GEz9vjjj5uxww47zIx94QtfMGOnnnpqzu3e85ixc03mJH52YoxzJNnPPIA6RY4C2UaOAtlGjgK1gzYsAAAAAIBUsAAFAAAAAKSCBSgAAAAAIBUsQAEAAAAAqWABCgAAAABIBTWC86RZs2ZmrGHDhjW+Pa98s9cyZfHixWbMa3XgtS3w2rA0aFDz32Fs27atxmNQf1ltd7x91WtVsnXrVjM2Z84cM+aVdN+8ebMZs9oXvfvuu+YYr/2DV3a+e/fuZqy4uNiMefk7Y8aMnNsPOeQQc4zX4sQrce+1tmjevLkZ817Tjz76KOd2ry0HbSj2PV7bJqsNi7Vdkn7729+asQsuuMCMtW/f3ox9+umnZsxq5yRJ48aNy7m9R48e5pgWLVqYMdStrVu3avXq1Tlj3j7pnVOs4+66devMMd61YmFhoRkrKCgwYwcffLAZGz58uBkrLy/Pud079m/ZssWMffLJJ2Zs6tSpZmzChAlmbPTo0WbM4rVh8V5rz/5y/uIdUAAAAABAKliAAgAAAABSwQIUAAAAAJAKFqAAAAAAgFSwAAUAAAAApIIFKAAAAAAgFbRhyZOmTZuasUaNGpkxq0WF14YlaTuD999/34wdeeSRZqxNmzZmzGrD4j1m77Fh31NRUZFz+9y5c80xpaWlZsxra7RmzRoztn79ejPmtROaPXt2zu0LFiwwx3htXbz2Ca+99poZmzx5shnz2r5YJfW9x2y9Zt7tVRdL0o5K8o9plv2ljP3+xDsmWG2IrBZKkr+veuNuu+02M+ad98rKysyYdWzycnT69OlmrG/fvmYMtW/9+vV6++23c8a8Y9Orr75qxqzzZatWrcwx3nHca0fy1ltvmbHjjz/ejHXq1ClRLIljjjnGjHktm7xzrNf+zWqJlPS8xjmKd0ABAAAAAClhAQoAAAAASAULUAAAAABAKliAAgAAAABSwQIUAAAAAJAKFqAAAAAAgFTQDyNPvNYKRUVFZmzp0qU5t3utB7Zs2WLG+vXrZ8ZmzZplxpYsWWLGunTpYsYsXquVpGWrUT9t2LAh53Zvf1y5cmWi+/L2O6/sudfuIMaYc7vXasXjtWzy2sh4peW9nLLaTcybN88cs2nTJjNWXl5uxrw2FNu2bTNjHmv+3jHLO0Z67TeQXYsXLzZj3bt3z7ndap0g+fn73nvvmTGvHcyIESPM2D//+U8z1r9//5zbvVYr1rWDZB+zkI6NGzdq2rRpOWPetd0777xT4/vyjv3e8XjKlClm7IEHHjBjJSUlZuyQQw4xYxbvvJx0P/ZaFZ5yyilm7LjjjjNj1nNJO5XkeAcUAAAAAJAKFqAAAAAAgFSwAAUAAAAApIIFKAAAAAAgFSxAAQAAAACpqHYBGkJ4KISwLIQwbZdtbUIIL4UQPqn6u3XtThOAhRwFso0cBbKNHAXStSdtWB6RdK+kx3bZdp2kV2KMt4UQrqv6/0/yP736w2vDUlxcbMY+/vjjnNu98tNWWwtJ6tWrlxn77ne/a8ZeeOEFM+aVq7faXlRWVppjKFudd48owzm6fv36nNs//fRTc8zGjRsT3VfSsu3eOGt/9Urce/u/d19eOxhPkvl7rVa8WLdu3cxYmzZtzNiCBQvMmHdMs44xZWVl5hiv1UEdeUQZztH64MwzzzRjVmsdrx3Diy++aMbmzp1rxoYNG2bGvDZQhx12mBm75JJLcm732lE1adLEjCU9juznHlGecrRx48Y66KCDcsaWL19ujvNabXnnG4u3H3itqiZMmGDGvGtFr0WLdW7wrgeTtmjxWtM0a9bMjHnX8eRU/lX7jMYYx0la8ZnN50p6tOrfj0o6L8/zArCHyFEg28hRINvIUSBdSZf07WKMpZJU9Xfb/E0JQB6Qo0C2kaNAtpGjQC2p9feUQwhXhBAmhRAmeR+ZAlA3yFEg28hRINt2zdE1a9bU9XSAzEu6AF0aQuggSVV/L7N+MMY4MsY4MMY40Pt8OIC8IkeBbCNHgWxLlKMtW7ZMbYJAfZV0AfqcpOFV/x4u6dn8TAdAnpCjQLaRo0C2kaNALam2Cm4I4UlJJ0kqDiEslHSTpNskPR1CuFzSfElfrs1J1gdWFT5JKioqqvHtedW/vOpl3m/ejjjiCDM2btw4M7Zt2zYzZvHm6FUKRM1lPUetCqdeVVSviqmXG964pPudVf0uSV7UliTVAles+Gy9jX9Ztsz8Rb9bjdGrPpj0+bIqCnuVkpNWQ64tWc/R+qB///5mzMpRLy+6du1qxs444wwz5lUxveGGG8zYY489ZsasSvlJ3+2m0nzN5TNHCwsLdcopp+SMeee9hx56yIxZx9akFb+986FXfdmrkHvssceasSOPPDLndu98krRCbtIY0lXtFVmM8WIjdGqe5wIgAXIUyDZyFMg2chRIF41tAAAAAACpYAEKAAAAAEgFC1AAAAAAQCpYgAIAAAAAUsECFAAAAACQCvph5IlX0rpJkyZmzCqh7bUx8doZtGjRwoy1b9/ejDVu3NiMbd68ucb355UG9+4L9ZPX9mLNmjU5t8+fP98cY7VVkPx9y8uNpOXXk7QPyVIbEOtxe3mYtJ2K1TJF8l9T77VZv359zu20c9q/ePtPEqWlpWZsypQpZqxLly5m7IQTTjBjH374oRkbPHiwGUP9c8ABB6h169Y5Y5s2bTLHedeKXrsSS9LzYUVFhRmbPHmyGZs2bZoZ69u3b87tSR7X3kizDYt3HUA7GN4BBQAAAACkhAUoAAAAACAVLEABAAAAAKlgAQoAAAAASAULUAAAAABAKliAAgAAAABSQR37PPFKxHtlpq12KwUFBeYYr4x306ZNzVizZs3M2JIlS8xYx44dzZjVEsMr/5122W3UPu/1Lisry7l90aJF5piWLVuasaQtTvbXsudJHrc3xnv+vRY5SdtoWG2gWrVqZY7x2sigfsp3/s6aNcuMeedDb99avHixGVuxYsWeTQz7BGt/9a7RevfubcbeeuutnNuTXismbR24cuVKM7Z8+XIzZl0j7MutSur7/Gsb74ACAAAAAFLBAhQAAAAAkAoWoAAAAACAVLAABQAAAACkggUoAAAAACAVLEABAAAAAKmgDUueeCWt27Vrl9f7qqioMGOrVq0yY4WFhWasqKjIjFVWVpoxq+2C144B9ZNXLn3dunVmbM6cOTm3Wy2IpOQtO7w89Nr/eG1kvNi+KmlpfG9c0vY51vGuS5cu5hhaPe1frGOJdzx4+eWXzVivXr3MmHdu89pQlJeXmzHsP7xz26GHHmrGrPYtRx55pDnGu1acMWNGonHe8d9rQ7Rhw4ac273r0n25RQt4BxQAAAAAkBIWoAAAAACAVLAABQAAAACkggUoAAAAACAVLEABAAAAAKmodgEaQngohLAshDBtl203hxAWhRDeq/pzZu1OE4CFHAWyjRwFso0cBdK1J21YHpF0r6THPrP9zhjjr/I+o31QcXGxGWvRokXO7V5bC68txKZNm8yYVz6+U6dOZsxrw2LNxbuvzZs3m7EmTZqYMZgeUQo56pVEX7lypRn79NNPc273WiTUhqTtVKyy+fW9PYvXDsB7bEnL3+f7+fJaRzVs2DCv95UHj4jzaK1J0nbHa3nhtWGZOHGiGTvjjDPMmNd2CpnwiFLIUW9fta4HJWnw4ME5t994443mmP79+5uxv/3tb2bsrrvuMmNWWzVJKigoMGPWecM7D9FqZd9W7TugMcZxklakMBcACZCjQLaRo0C2kaNAuvbmO6BXhRCmVn1soXXeZgQgX8hRINvIUSDbyFGgFiRdgN4nqaekIySVSrrd+sEQwhUhhEkhhEllZWUJ7w5ADZGjQLaRo0C2kaNALUm0AI0xLo0xbo0xbpP0gKRBzs+OjDEOjDEOLCkpSTpPADVAjgLZRo4C2UaOArUn0QI0hNBhl/+eL2ma9bMA0keOAtlGjgLZRo4CtafaMpQhhCclnSSpOISwUNJNkk4KIRwhKUqaK+k7tThHAA5yFMg2chTINnIUSFe1C9AY48U5Nj9YC3PZZzVv3tyMtW6d+zvtXluLiooKM+a1YfHKwLdq1cqMrVq1yoxZ7Va8lh3ePGjDUnNp5aj3mnr7SGlpac7tXol1r41PbbQI2ZdZr5vXXsnjPcdezHvdvH3LamOVpD2U5Jf9ry2cR+uGt1996UtfMmNTpkwxYxs2bDBj7777rhnr27evGUPdSytHvXObt28deeSRObf37NnTHNO0aVMzNnToUDPmHccffvhhM7Z27doax9q1a2eO8XCur//SPxMDAAAAAPZLLEABAAAAAKlgAQoAAAAASAULUAAAAABAKliAAgAAAABSwQIUAAAAAJCKatuwYO95Zf/btGmTc/ucOXPMMUnbGXjlvzt37mzGvBYbVksV7768NizILu81XbJkiRmz2rB4eeG12PD2ce82892aw5tHUt5temXnk4yrL/O3jpEbN240x2StDQvqhnfMGjNmjBlbtmyZGTvwwAPN2FtvvWXGDj30UDOG/Yd3bisvLzdjHTt2zLk96TmvWbNmZmzw4MFmrKyszIyNGzfOjFmtjdq3b2+O8VoYNmzY0IyhfuBMDAAAAABIBQtQAAAAAEAqWIACAAAAAFLBAhQAAAAAkAoWoAAAAACAVFAFt44VFRXl3O5VjPQqm3lV/7xKk1aFNUmaMWOGGbMqunn3VVFRYcaQXd6+tXjxYjNmVZQ84AD78OPtPwUFBWYs6f7v5VQSaVeYzfc4r7KiV5nQ4+W99/y3bNky53bvcdXG84/6x9sPvPOaVxXeqsosScXFxWascePGZgz7D+8ctXbtWjPWtm3bnNu9arBJz2vWdakknXTSSWbMy6mJEyfm3H7EEUeYY3r16mXGUP/xDigAAAAAIBUsQAEAAAAAqWABCgAAAABIBQtQAAAAAEAqWIACAAAAAFLBAhQAAAAAkArasNSxpk2b5tzutaiwWp9I0ubNmxON81oreC1hrNYKXvl7bx7ILq+ke3l5uRlbtWpVzu3efuW1WikpKUk0j40bN5oxby5JWnqk3QYk3/Nv1KiRGWvfvr0ZKysrSxTzWhM0adIk53bvGAlUp1mzZmZs6tSpZqx79+5m7MQTTzRj+W7D4uW1dzxA3fKOdStXrjRjrVu3zrndO1Z752zv+Om14erWrZsZ+9znPmfGxowZk3P7vHnzzDE9evQwY94c09z/a6M9mjf/fSm3eQcUAAAAAJAKFqAAAAAAgFSwAAUAAAAApIIFKAAAAAAgFSxAAQAAAACpYAEKAAAAAEhFtXXsQwhdJD0mqb2kbZJGxhjvDiG0kfSUpG6S5kq6MMZo15BGTla7iaRlmNetW2fGtmzZYsasdjCSX+5606ZNObd7pca90uCoubRy1Gq5I0krVqwwY2vXrs253dvnCgsLzVivXr0SzcPbJ719PImkpdK93Mh3+XXvvrznynvdkraD8eZitRlIesyqC5xH64a3X23YsMGMee3MWrZsacY+/vhjM5bvNizIr7Ry1LpmkuxzpSS1aNEi5/ak7ai8Y6SXN95+/PnPf96M/eUvf8m5ffr06eaYY4891ox586+NFl3W9Y93ze21evLa5+xLrVY8e3KWrpT04xjjoZI+L+l7IYS+kq6T9EqMsZekV6r+DyB95CiQbeQokG3kKJCiahegMcbSGOOUqn+vlfSRpE6SzpX0aNWPPSrpvNqaJAAbOQpkGzkKZBs5CqSrRp9TCiF0kzRA0tuS2sUYS6XtiSuprTHmihDCpBDCpLKysr2bLQAXOQpkGzkKZBs5CtS+PV6AhhBaSPqzpKtjjGv2dFyMcWSMcWCMcWBJSUmSOQLYA+QokG3kKJBt5CiQjj1agIYQGml7Qj4eY3ymavPSEEKHqngHSctqZ4oAqkOOAtlGjgLZRo4C6al2ARq2l2N6UNJHMcY7dgk9J2l41b+HS3o2/9MDUB1yFMg2chTINnIUSNee1Co+TtKlkj4IIbxXte16SbdJejqEcLmk+ZK+XDtT3LdZrQS8lgWejRs3mrHKykoz1qRJEzPmlYu2SlB79+XFkEhec9Ta97wS8cuXL6/BdLfzyqi3bZvzazaSpHbt2pmxhg0b1ngeUvJ8y/d9ebGkLU4sSVsleccK7/n35pikJUDz5s3NMRksY895tA545xqvrdTJJ59sxubMmWPGZsyYYcaOOeYYM4ZMyGuOWsct7/uhXmspq7VIbRzrkt5mp06dzFj37t1zbn/77bfNMRdddJEZ81qcJD2fe+dEq8XS6NGjzTFeyzivZU2HDh3MmLeP1DfVLkBjjBMkWXvjqfmdDoCaIkeBbCNHgWwjR4F0ZatbNwAAAABgn8UCFAAAAACQChagAAAAAIBUsAAFAAAAAKSCBSgAAAAAIBV70oYFtahFixY5t3vtDLxS0Zs2bUo0j4KCAjPmleS2ytx7c0S2WSXMrZY7kt/uwNq3vDLqnTt3NmPePLwS90lLs1v7v9c6JMntScnbyOR7Ht7xoGXLlmbMa43i8Vo9Wc+z1w4m6WuD/YeXa1bLBUnq37+/GRs0aJAZe//99/dsYrtIsz0U8ifGaF4DLVmyxBznHVutNiy1IWnLL+84ftRRR+Xc/t577+XcLknTp083Y147Nm/+3rWp107u1Vdfzbl91KhR5pjWrVubsTfeeMOMee1njj/+eDOWtI1bXeEsDQAAAABIBQtQAAAAAEAqWIACAAAAAFLBAhQAAAAAkAoWoAAAAACAVLAABQAAAACkgjYsdaywsDDndq/FgMcrtey1JkjSBkGy2754bTlQP3ltC7zy39u2bcu53Ss5P2DAADPmtfooLS01Y2vXrjVjW7ZsMWPr16/PuX3Dhg01HiNJGzduNGPWcyX5eejlfZLy69ZxSZI6depkxsrLy82Yt494bV+s1zvNljWon7x932sDNWHCBDNmtZOQpOHDh5uxcePGmTHr2GS1aUP2We0+vFYf3nVYktZ2abfe8Obfq1evnNu964CxY8easRNOOMGMeY/bO9fPnj3bjE2aNCnndu86wGsZV1FRYcZOOeUUM+adR5O0barL9iy8AwoAAAAASAULUAAAAABAKliAAgAAAABSwQIUAAAAAJAKFqAAAAAAgFSwAAUAAAAApII2LHXMKrPutUHwWjV47U+8Es1eawuvJYx1f94caZ+QbVZZ7jZt2phjevbsacZ69OiRc7vXzqN///5m7IgjjjBjRx99tBnzWoR4pfGt2LJly8wxCxYsMGNLly41Y177FqvlkSRt3rzZjFm56I0pKioyY02bNjVjXkl37/jjHROsY+EBB3D6gs9rXeQds7z8feGFF8yY11rBa/9wySWXmDHUT9YxzdvvvP21VatWNZ5D2i02vPl3794953av5dqLL75oxg455BAz9sUvftGMeeeasrIyM2adt73zqHc93rhx40Qx7xjjnROtfSFJ65Z84R1QAAAAAEAqWIACAAAAAFLBAhQAAAAAkAoWoAAAAACAVLAABQAAAACkotoygiGELpIek9Re0jZJI2OMd4cQbpb0bUk7ykZdH2N8vrYmuq8qKCjIud2rZuVVyly7dq0Z86rZevfnVT+1qnx51dC8Cl+ouXznqFUtzauUfPDBB5ux008/vcZj+vbta8a8/dGrFLh169ZEMWsf96paepXqvNiqVavMmFc9d8WKFWbMqp7rVar+/+3dT4idVx3G8edHiCG0E8iM+TNpJo4W8z8hldBNxZ0Ss2kKQexCKgh1YcGCC0U3WQWRti4HWjpQpeimFbsKFiMJQihNhphJHXRqiRg7ySQ0oZ1ForY/F/dNnTu858zk5s45PzPfDwyZeU9u3ue+uc+9czI352zcuDE5lnv+OXXqVHKs1xUeR0ZGWo//Pz2P8Dp673KrNaaes3IrVE5MTCTHcqti54yNjSXHDh06lBxbvXr1XZ+r5uqV96N+dtTMkt9T7dmzJ3m7Xbt2JcdSOybkVrpdjlVwe13pfGhoqPX40aNHk7e5cuVKcmx8fDw5duLEieTYli1bkmO519jUyti555jca97c3FxyLPc6euPGjeRY7r5t27at9fimTZuSt0k95qT8fVuqpaxj/x9JP3D3CTMbkHTOzN5sxn7u7s/dcwoA94KOArHRUSA2OgoUtOgE1N1nJM00n39kZlOS0hv4ASiKjgKx0VEgNjoKlHVXP0M1s1FJj0h6qzn0jJldMLNxM1vf52wA7hIdBWKjo0BsdBRYfkuegJrZg5Jek/Ssu38oaUzSw5IOqPOvRs8nbve0mZ01s7PXrl1r+y0A+oCOArHRUSA2OgqUsaQJqJmtVqeQr7r765Lk7lfd/WN3/0TSS5Iebbutu7/o7gfd/eCGDRv6lRvAPHQUiI2OArHRUaCcRSeg1lny6mVJU+7+wrzjw/N+2xOSLvY/HoDF0FEgNjoKxEZHgbKWsgruY5K+JWnSzM43x34s6UkzOyDJJV2S9N1lSXifGxgYaD2+c+fO5G0mJyeTY7mtVjZv3pwcS201IUl79+5Njp05c6b1eG6J5l6WnEdWkY7mlljPbWOS2r4lt2R4bluOnFWrViXHclsNRZHackTKL9Gfu/69yPX39OnTPeXIbVuzbt265NiOHTtaj69duzZ5m4B4Ha0gt9VQbsuy3FZJ27dv7ynL/v37e7odiinS0dz2Ubnv31LPybnHeGm5LVpSr82517Xjx48nx3LbKOVeo6anp5Nj169fT46lXqNy26nMzs4mx3LPMSdPnkyO5fLntk1JzSlGR0eTt8l9r57blm+plrIK7h8ltT2q2KsMCICOArHRUSA2OgqUde87iQIAAAAAsARMQAEAAAAARTABBQAAAAAUwQQUAAAAAFAEE1AAAAAAQBHx9yW4z6WW5B4eHm49LuWX6s5tddDr9ie5LRJSbt++nRzLbeeBuHrd0iD19517HOfG7me57U9yy/fn/m76vXz/7t27k2NHjhxJjg0ODibH9u3bd9fny10PrCz9fk3J9TC3nVPuNXbNmjX3lAn3h9zzbm4bk170+89bDrmt03KvGbntW27dupUcy207kvu+I7VlYur4YmO579Vv3rzZ01jufm/durX1+NDQUPI2uW3VcudaKn4CCgAAAAAoggkoAAAAAKAIJqAAAAAAgCKYgAIAAAAAimACCgAAAAAoggkoAAAAAKAIK7klhpldk/T35svPSrpe7OR5UbKQo1uUHFJ/snzO3Tf0I8xyoaOLIke3KDkkOlpTlBxSnCzk6NavHHS0N1FySHGykKNb0Y4WnYB2ndjsrLsfrHLyBaJkIUfMHFKsLKVEus9RspAjZg4pVpZSotznKDmkOFnIETNHaVHud5QcUpws5Kibg7fgAgAAAACKYAIKAAAAACii5gT0xYrnXihKFnJ0i5JDipWllEj3OUoWcnSLkkOKlaWUKPc5Sg4pThZydIuSo7Qo9ztKDilOFnJ0K5qj2v8BBQAAAACsLLwFFwAAAABQRJUJqJkdMrO/mNm7ZvajGhmaHJfMbNLMzpvZ2cLnHjezWTO7OO/YoJm9aWbTza/rK+U4Zmb/bK7LeTM7XCDHiJn9wcymzOwdM/t+c7zoNcnkKH5NaqKjdLQlBx0NhI7S0ZYcdDSQld7RKP3MZKGjFTta/C24ZrZK0l8lfVXSZUlvS3rS3f9cNEgnyyVJB929+P47ZvYVSXOSfuHue5tjP5P0gbv/tHmyWu/uP6yQ45ikOXd/bjnPvSDHsKRhd58wswFJ5yQdkfRtFbwmmRzfUOFrUgsd/fTcdLQ7Bx0Ngo5+em462p2DjgZBR+P0M5PlmOhotY7W+Anoo5Ledff33P1fkn4t6fEKOapy99OSPlhw+HFJrzSfv6LOg6FGjuLcfcbdJ5rPP5I0JekhFb4mmRwrCR0VHW3JQUfjoKOioy056GgcK76jUfqZyVIcHf2fGhPQhyT9Y97Xl1Xvickl/c7MzpnZ05UyzLfJ3WekzoND0saKWZ4xswvN2xaKvEXiDjMblfSIpLdU8ZosyCFVvCaF0dE0Oio6GgAdTaOjoqMB0NF2kfop0dFqHa0xAbWWY7WW4n3M3b8k6euSvtf8iB7SmKSHJR2QNCPp+VInNrMHJb0m6Vl3/7DUeZeQo9o1qYCOxkdH6ehCdDQWOkpHF6KjsdDRih2tMQG9LGlk3tdbJb1fIYfc/f3m11lJv1HnLRM1XW3el33n/dmzNUK4+1V3/9jdP5H0kgpdFzNbrU4RXnX315vDxa9JW45a16QSOppGR+loBHQ0jY7S0QjoaLsQ/ZToaO2O1piAvi3pi2b2eTP7jKRvSnqjdAgze6D5j7cyswckfU3Sxfytlt0bkp5qPn9K0m9rhLhTgsYTKnBdzMwkvSxpyt1fmDdU9JqkctS4JhXR0TQ6SkcjoKNpdJSORkBH24Xop0RH23IUvSbuXvxD0mF1Vgf7m6SfVMrwBUl/aj7eKZ1D0q/U+fH2v9X5l7LvSBqS9HtJ082vg5Vy/FLSpKQL6pRiuECOL6vz9pQLks43H4dLX5NMjuLXpOYHHaWjLTnoaKAPOkpHW3LQ0UAfK72jUfqZyUJHK3a0+DYsAAAAAICVqcZbcAEAAAAAKxATUAAAAABAEUxAAQAAAABFMAEFAAAAABTBBBQAAAAAUAQTUAAAAABAEUxAAQAAAABFMAEFAAAAABTxX77R8N86LxHGAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1152x864 with 12 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "fig,axes=plt.subplots(3, 4, figsize=(16, 12))\n",
    "for i,ax in enumerate(axes.flat):\n",
    "    ax=show_img(X[i],ax=ax)\n",
    "    _=ax.text(0,0, labels[y[i]],\n",
    "        verticalalignment='top', color='black', fontsize=12, weight='bold')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Validation set (no augmentation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = get_data(sz,bs)\n",
    "ite = iter(data.val_dl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x864 with 12 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X,y = to_np(next(ite))\n",
    "X=X[:12,0,:28,:28]\n",
    "fig,axes=plt.subplots(3, 4, figsize=(16, 12))\n",
    "for i,ax in enumerate(axes.flat):\n",
    "    ax=show_img(X[i],ax=ax)\n",
    "    _=ax.text(0,0, labels[y[i]],\n",
    "        verticalalignment='top', color='black', fontsize=12, weight='bold')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## using custom data loader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "sz=28\n",
    "bs=12\n",
    "\n",
    "class FMNIST_3C(Dataset):\n",
    "    def __init__(self,X,y,transforms=None):\n",
    "        self.X = X\n",
    "        self.y = y\n",
    "        self.n = len(X)\n",
    "        self.transforms = transforms\n",
    "    \n",
    "    def __getitem__(self,index):\n",
    "        img = self.X[index]\n",
    "        label = self.y[index]\n",
    "        if self.transforms is not None:\n",
    "            img = self.transforms(img)\n",
    "        return (img,label)\n",
    "    \n",
    "    def __len__(self):\n",
    "        return self.n\n",
    "\n",
    "\n",
    "def get_data(bs,X,y,shuffle=True):\n",
    "    X1 = gc2RGB(X)\n",
    "    tsfms = transforms.Compose([transforms.ToTensor(),\n",
    "                                transforms.Normalize(tuple([nm_mean]*3),tuple([nm_std]*3)),\n",
    "                                transforms.ToPILImage(None),\n",
    "                                transforms.Pad(1),\n",
    "                                transforms.RandomRotation(10),\n",
    "                                transforms.ToTensor()\n",
    "                               ]\n",
    "                          )\n",
    "    ds = FMNIST_3C(X1,y,tsfms)\n",
    "    dl = DataLoader(ds,batch_size=bs,shuffle=shuffle,num_workers=0)\n",
    "    \n",
    "    \n",
    "    return dl"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dl = get_data(12,X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x864 with 12 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X,y = next(ite)\n",
    "X = X.numpy()\n",
    "y=y.numpy()\n",
    "X=X[:12,0,:28,:28]\n",
    "fig,axes=plt.subplots(3, 4, figsize=(16, 12))\n",
    "for i,ax in enumerate(axes.flat):\n",
    "    ax=show_img(X[i],ax=ax)\n",
    "    _=ax.text(0,0, labels[y[i]],\n",
    "        verticalalignment='top', color='black', fontsize=12, weight='bold')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Those weird artifacts appear to be the result of converting to PIL and back to torch Tensor\n",
    "\n",
    "Currently there is no easy way to do data augmentation efficiently with torchvision library. This issue has also been raised here: \n",
    "\n",
    "https://github.com/pytorch/vision/issues/448"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Simple NN\n",
    "\n",
    "Using custom data loader with no data augmentation such as random rotation and cropping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_ds = FMNIST(X_train,y_train,tsfms)\n",
    "train_dl = DataLoader(train_ds,batch_size=32,shuffle=True,num_workers=0)\n",
    "test_ds = FMNIST(X_test,y_test,tsfms)\n",
    "test_dl = DataLoader(test_ds,batch_size=32,shuffle=False,num_workers=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.nn.init import kaiming_uniform, kaiming_normal\n",
    "class SimpleNN(nn.Module):\n",
    "    def __init__(self,layers,outp,drops):\n",
    "        super().__init__()\n",
    "        self.weights = nn.ModuleList([nn.Linear(layers[i],layers[i+1]) for i in range(len(layers)-1)])\n",
    "        self.outp = nn.Linear(layers[-1],outp)\n",
    "        self.bns = nn.ModuleList([nn.BatchNorm1d(l) for l in layers[1:]])\n",
    "        self.drops = nn.ModuleList([nn.Dropout(drop) for drop in drops])\n",
    "        \n",
    "        for w in self.weights:\n",
    "            kaiming_normal(w.weight.data)\n",
    "        kaiming_normal(self.outp.weight.data)\n",
    "    def forward(self,x):\n",
    "        # flatten\n",
    "        x = x.view(x.size(0),-1)\n",
    "        for w,bn,dr in zip(self.weights, self.bns,self.drops):\n",
    "#             x= F.relu(dr(bn(w(x))))\n",
    "            x = F.relu(w(x))\n",
    "#             x= F.relu(bn(w(x)))\n",
    "        x = self.outp(x) # (bs,10)\n",
    "        x = F.log_softmax(x,dim=-1)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "net = SimpleNN([28*28,100],10,[.1])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam(net.parameters(),lr=0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(nb_epoch):\n",
    "    best=[0,100,0]\n",
    "    for epoch in range(nb_epoch):\n",
    "        running_trn_loss = 0.\n",
    "        trn_corrects = 0\n",
    "        print(f'Epoch {epoch+1}:')\n",
    "        for ba in train_dl:\n",
    "            inputs,labels = ba\n",
    "            inputs, labels = Variable(inputs), Variable(labels)\n",
    "            optimizer.zero_grad()\n",
    "            #Compute the outputs given by our model at this stage.\n",
    "            outputs = net(inputs)\n",
    "            _,preds = torch.max(outputs.data,1)\n",
    "            #Compute the loss\n",
    "            loss = F.nll_loss(outputs, labels)\n",
    "            running_trn_loss += loss.data[0] * inputs.size(0)\n",
    "            trn_corrects += torch.sum(labels.data == preds)\n",
    "            #Backpropagate the computation of the gradients\n",
    "            loss.backward()\n",
    "            #Do the step of the SGD\n",
    "            optimizer.step()\n",
    "        print(f'Training Loss: {running_trn_loss/len(train_ds)}  Accuracy: {100.*trn_corrects/len(train_ds)}')\n",
    "        \n",
    "        \n",
    "        running_val_loss = 0.\n",
    "        val_corrects = 0    \n",
    "        for ba in test_dl:\n",
    "            inputs,labels = ba\n",
    "            inputs, labels = Variable(inputs), Variable(labels)\n",
    "            outputs = net(inputs)\n",
    "            _,preds = torch.max(outputs.data,1)\n",
    "            #Compute the loss\n",
    "            loss = F.nll_loss(outputs, labels)\n",
    "            running_val_loss += loss.data[0] * inputs.size(0)\n",
    "            val_corrects += torch.sum(labels.data == preds)\n",
    "        print(f'Validation Loss: {running_val_loss/len(test_ds)}  Accuracy: {100.*val_corrects/len(test_ds)}')\n",
    "        \n",
    "        val_loss = running_val_loss/len(test_ds)\n",
    "        val_acc = 100.*val_corrects/len(test_ds)\n",
    "        if val_loss < best[1]:\n",
    "            best[1]= val_loss\n",
    "            best[2]= val_acc\n",
    "            best[0] = epoch+1\n",
    "    print(f'Best iteration: {best[0]} with val loss: {best[1]}, val acc: {best[2]} ')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## without batchnorm and dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1:\n",
      "Training Loss: 0.4701891098221143  Accuracy: 83.115\n",
      "Validation Loss: 0.426979985165596  Accuracy: 84.86\n",
      "Epoch 2:\n",
      "Training Loss: 0.3604118819355965  Accuracy: 86.88666666666667\n",
      "Validation Loss: 0.37660377769470216  Accuracy: 86.48\n",
      "Epoch 3:\n",
      "Training Loss: 0.3251252915918827  Accuracy: 87.94666666666667\n",
      "Validation Loss: 0.39630649839639664  Accuracy: 85.34\n",
      "Epoch 4:\n",
      "Training Loss: 0.30300698330402376  Accuracy: 88.82166666666667\n",
      "Validation Loss: 0.36805234270095827  Accuracy: 87.15\n",
      "Epoch 5:\n",
      "Training Loss: 0.28481361074844996  Accuracy: 89.425\n",
      "Validation Loss: 0.3544987232208252  Accuracy: 87.9\n",
      "Epoch 6:\n",
      "Training Loss: 0.2722178290585677  Accuracy: 89.84333333333333\n",
      "Validation Loss: 0.3632378236413002  Accuracy: 87.49\n",
      "Epoch 7:\n",
      "Training Loss: 0.25797976875305173  Accuracy: 90.33\n",
      "Validation Loss: 0.36197250916957857  Accuracy: 87.54\n",
      "Epoch 8:\n",
      "Training Loss: 0.24535473074018954  Accuracy: 90.89\n",
      "Validation Loss: 0.356014321154356  Accuracy: 87.85\n",
      "Epoch 9:\n",
      "Training Loss: 0.2385104939242204  Accuracy: 91.02333333333333\n",
      "Validation Loss: 0.36032191736102104  Accuracy: 88.21\n",
      "Epoch 10:\n",
      "Training Loss: 0.22561907813052337  Accuracy: 91.43166666666667\n",
      "Validation Loss: 0.3686384593963623  Accuracy: 87.84\n",
      "Epoch 11:\n",
      "Training Loss: 0.2200901007473469  Accuracy: 91.83666666666667\n",
      "Validation Loss: 0.3703323360383511  Accuracy: 87.74\n",
      "Epoch 12:\n",
      "Training Loss: 0.21343959808945656  Accuracy: 92.005\n",
      "Validation Loss: 0.3667078548967838  Accuracy: 88.02\n",
      "Epoch 13:\n",
      "Training Loss: 0.2060954282095035  Accuracy: 92.27166666666666\n",
      "Validation Loss: 0.37555275305509567  Accuracy: 88.17\n",
      "Epoch 14:\n",
      "Training Loss: 0.19966280966848135  Accuracy: 92.40833333333333\n",
      "Validation Loss: 0.3845762151122093  Accuracy: 88.15\n",
      "Epoch 15:\n",
      "Training Loss: 0.19324996203432482  Accuracy: 92.815\n",
      "Validation Loss: 0.4041687335968018  Accuracy: 88.1\n",
      "Epoch 16:\n",
      "Training Loss: 0.18887905380378167  Accuracy: 92.88166666666666\n",
      "Validation Loss: 0.4010792544066906  Accuracy: 87.78\n",
      "Epoch 17:\n",
      "Training Loss: 0.18344330060631037  Accuracy: 92.98166666666667\n",
      "Validation Loss: 0.4314974734991789  Accuracy: 87.24\n",
      "Epoch 18:\n",
      "Training Loss: 0.17565666512474418  Accuracy: 93.22333333333333\n",
      "Validation Loss: 0.4101353040933609  Accuracy: 88.21\n",
      "Epoch 19:\n",
      "Training Loss: 0.17373821822603544  Accuracy: 93.36\n",
      "Validation Loss: 0.420098881945014  Accuracy: 88.13\n",
      "Epoch 20:\n",
      "Training Loss: 0.16728267382935932  Accuracy: 93.71333333333334\n",
      "Validation Loss: 0.42805914930701255  Accuracy: 88.18\n",
      "Best iteration: 5 with val loss: 0.3544987232208252, val acc: 87.9 \n"
     ]
    }
   ],
   "source": [
    "train(20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## with bn no dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1:\n",
      "Training Loss: 0.5020102681438128  Accuracy: 82.66833333333334\n",
      "Validation Loss: 0.4268017381668091  Accuracy: 84.59\n",
      "Epoch 2:\n",
      "Training Loss: 0.37354170792301494  Accuracy: 86.465\n",
      "Validation Loss: 0.397764958524704  Accuracy: 85.54\n",
      "Epoch 3:\n",
      "Training Loss: 0.3364966302990913  Accuracy: 87.735\n",
      "Validation Loss: 0.38506978566646577  Accuracy: 86.09\n",
      "Epoch 4:\n",
      "Training Loss: 0.31336213549574216  Accuracy: 88.61166666666666\n",
      "Validation Loss: 0.36150245010852816  Accuracy: 87.47\n",
      "Epoch 5:\n",
      "Training Loss: 0.29700259143710134  Accuracy: 89.215\n",
      "Validation Loss: 0.3636846319437027  Accuracy: 87.32\n",
      "Epoch 6:\n",
      "Training Loss: 0.2783734321117401  Accuracy: 89.89\n",
      "Validation Loss: 0.3491296519756317  Accuracy: 87.42\n",
      "Epoch 7:\n",
      "Training Loss: 0.2697597583323717  Accuracy: 90.09166666666667\n",
      "Validation Loss: 0.3469563156604767  Accuracy: 87.79\n",
      "Epoch 8:\n",
      "Training Loss: 0.26066138525605204  Accuracy: 90.63833333333334\n",
      "Validation Loss: 0.34567423810958864  Accuracy: 87.84\n",
      "Epoch 9:\n",
      "Training Loss: 0.24696346764564514  Accuracy: 90.90333333333334\n",
      "Validation Loss: 0.35213400201797485  Accuracy: 87.72\n",
      "Epoch 10:\n",
      "Training Loss: 0.24091521470844746  Accuracy: 91.14666666666666\n",
      "Validation Loss: 0.3532409568071365  Accuracy: 87.66\n",
      "Epoch 11:\n",
      "Training Loss: 0.23352146077553432  Accuracy: 91.43833333333333\n",
      "Validation Loss: 0.3577724233031273  Accuracy: 87.74\n",
      "Epoch 12:\n",
      "Training Loss: 0.22752601362665495  Accuracy: 91.62833333333333\n",
      "Validation Loss: 0.3516536557197571  Accuracy: 87.99\n",
      "Epoch 13:\n",
      "Training Loss: 0.2168505831172069  Accuracy: 92.11666666666666\n",
      "Validation Loss: 0.3554378990650177  Accuracy: 87.95\n",
      "Epoch 14:\n",
      "Training Loss: 0.21235097715059917  Accuracy: 92.18833333333333\n",
      "Validation Loss: 0.3656087178945541  Accuracy: 87.71\n",
      "Epoch 15:\n",
      "Training Loss: 0.2072366803407669  Accuracy: 92.35833333333333\n",
      "Validation Loss: 0.3590160031080246  Accuracy: 87.91\n",
      "Epoch 16:\n",
      "Training Loss: 0.2038482510517041  Accuracy: 92.57666666666667\n",
      "Validation Loss: 0.35332375901937485  Accuracy: 88.08\n",
      "Epoch 17:\n",
      "Training Loss: 0.19465444368819396  Accuracy: 92.88\n",
      "Validation Loss: 0.3541595380187035  Accuracy: 88.32\n",
      "Epoch 18:\n",
      "Training Loss: 0.1913457138578097  Accuracy: 92.885\n",
      "Validation Loss: 0.36515003547668456  Accuracy: 87.99\n",
      "Epoch 19:\n",
      "Training Loss: 0.18722141644457976  Accuracy: 93.03333333333333\n",
      "Validation Loss: 0.3705548891544342  Accuracy: 87.75\n",
      "Epoch 20:\n",
      "Training Loss: 0.18422935107946395  Accuracy: 93.11833333333334\n",
      "Validation Loss: 0.36508447090387347  Accuracy: 88.36\n",
      "Best iteration: 8 with val loss: 0.34567423810958864, val acc: 87.84 \n"
     ]
    }
   ],
   "source": [
    "train(20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## with bn + dropouts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1:\n",
      "Training Loss: 0.5135111429055532  Accuracy: 82.395\n",
      "Validation Loss: 0.43876418895721436  Accuracy: 84.27\n",
      "Epoch 2:\n",
      "Training Loss: 0.39121287387212117  Accuracy: 85.85833333333333\n",
      "Validation Loss: 0.4129725981235504  Accuracy: 85.18\n",
      "Epoch 3:\n",
      "Training Loss: 0.356039269498984  Accuracy: 87.12\n",
      "Validation Loss: 0.4003729352712631  Accuracy: 85.65\n",
      "Epoch 4:\n",
      "Training Loss: 0.3348866423567136  Accuracy: 87.75666666666666\n",
      "Validation Loss: 0.3877351433753967  Accuracy: 85.94\n",
      "Epoch 5:\n",
      "Training Loss: 0.31816098250945407  Accuracy: 88.53666666666666\n",
      "Validation Loss: 0.37666326681375506  Accuracy: 86.6\n",
      "Epoch 6:\n",
      "Training Loss: 0.3064735544085503  Accuracy: 88.695\n",
      "Validation Loss: 0.3726346318721771  Accuracy: 86.73\n",
      "Epoch 7:\n",
      "Training Loss: 0.2937574561615785  Accuracy: 89.17\n",
      "Validation Loss: 0.3658920074701309  Accuracy: 87.05\n",
      "Epoch 8:\n",
      "Training Loss: 0.2846807190577189  Accuracy: 89.64833333333333\n",
      "Validation Loss: 0.3674471758365631  Accuracy: 86.9\n",
      "Epoch 9:\n",
      "Training Loss: 0.2740944677670797  Accuracy: 89.89333333333333\n",
      "Validation Loss: 0.36631999120712283  Accuracy: 87.11\n",
      "Epoch 10:\n",
      "Training Loss: 0.2683616222023964  Accuracy: 90.15166666666667\n",
      "Validation Loss: 0.36482748489379885  Accuracy: 87.09\n",
      "Epoch 11:\n",
      "Training Loss: 0.26138698123892146  Accuracy: 90.31833333333333\n",
      "Validation Loss: 0.3672550737142563  Accuracy: 87.25\n",
      "Epoch 12:\n",
      "Training Loss: 0.25683154104153316  Accuracy: 90.48333333333333\n",
      "Validation Loss: 0.37047415314912796  Accuracy: 87.22\n",
      "Epoch 13:\n",
      "Training Loss: 0.24966281658510367  Accuracy: 90.87\n",
      "Validation Loss: 0.3707993052244186  Accuracy: 87.19\n",
      "Epoch 14:\n",
      "Training Loss: 0.24579233148097993  Accuracy: 90.82\n",
      "Validation Loss: 0.37646919101476667  Accuracy: 87.04\n",
      "Epoch 15:\n",
      "Training Loss: 0.24055394144654274  Accuracy: 91.16333333333333\n",
      "Validation Loss: 0.37289145069122315  Accuracy: 87.26\n",
      "Epoch 16:\n",
      "Training Loss: 0.2361799540023009  Accuracy: 91.275\n",
      "Validation Loss: 0.3690119399547577  Accuracy: 87.15\n",
      "Epoch 17:\n",
      "Training Loss: 0.2313108193953832  Accuracy: 91.405\n",
      "Validation Loss: 0.3659739604592323  Accuracy: 87.7\n",
      "Epoch 18:\n",
      "Training Loss: 0.2294640877187252  Accuracy: 91.505\n",
      "Validation Loss: 0.3736760178089142  Accuracy: 87.59\n",
      "Epoch 19:\n",
      "Training Loss: 0.2244130203882853  Accuracy: 91.685\n",
      "Validation Loss: 0.3760404171943664  Accuracy: 87.5\n",
      "Epoch 20:\n",
      "Training Loss: 0.21936514503161111  Accuracy: 91.78666666666666\n",
      "Validation Loss: 0.3782915353894234  Accuracy: 87.34\n",
      "Epoch 21:\n",
      "Training Loss: 0.21598267135322094  Accuracy: 92.03166666666667\n",
      "Validation Loss: 0.37709784047603606  Accuracy: 87.29\n",
      "Best iteration: 10 with val loss: 0.36482748489379885, val acc: 87.09 \n"
     ]
    }
   ],
   "source": [
    "train(21)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Beside regularization from Adam optimizer, extra regularization (bn and drop out) does not work that well for such small dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Learning rate finder\n",
    "\n",
    "Source : https://sgugger.github.io/how-do-you-find-a-good-learning-rate.html"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The idea is to use a small learning rate (10^-8) and for each minibatch, increase it by multiplying by some factor. We record loss for each learning rate and plot the loss function with respect to learning rate, in log scale. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1875"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_dl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SimpleNN(\n",
       "  (weights): ModuleList(\n",
       "    (0): Linear(in_features=784, out_features=100, bias=True)\n",
       "  )\n",
       "  (bns): ModuleList(\n",
       "    (0): BatchNorm1d(100, eps=1e-05, momentum=0.1, affine=True)\n",
       "  )\n",
       "  (outp): Linear(in_features=100, out_features=10, bias=True)\n",
       "  (drops): ModuleList(\n",
       "    (0): Dropout(p=0.1)\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "def find_lr(lr_start = 1e-8,lr_end=10, beta = .95):\n",
    "    num = len(train_dl) # number of steps (minibatch) \n",
    "    mult = (lr_end / lr_start) ** (1/num)\n",
    "    lr = lr_start\n",
    "    optimizer.param_groups[0]['lr'] = lr # set lr for optimizer\n",
    "    avg_loss = 0. # old avg loss\n",
    "    best_loss = 0. #lowest loss. Used for stopping when loss starts to explode: \n",
    "    # current smoothed loss > 4 * minimum smoothed loss\n",
    "    batch_num = 0\n",
    "    \n",
    "    # for plotting\n",
    "    losses = []\n",
    "    log_lrs = []\n",
    "    for data in train_dl:\n",
    "        batch_num += 1\n",
    "        #As before, get the loss for this mini-batch of inputs/outputs\n",
    "        inputs,labels = data\n",
    "        inputs, labels = Variable(inputs), Variable(labels)\n",
    "        optimizer.zero_grad()\n",
    "        outputs = net(inputs)\n",
    "        loss = F.nll_loss(outputs, labels)\n",
    "        #Compute the smoothed loss\n",
    "        avg_loss = beta * avg_loss + (1-beta) *loss.data[0] # exp weighted averages on log loss\n",
    "        smoothed_loss = avg_loss / (1 - beta**batch_num) # bias-corrected of exp weighted averages\n",
    "        #Stop if the loss is exploding\n",
    "        if batch_num > 1 and smoothed_loss > 4 * best_loss:\n",
    "            return log_lrs, losses\n",
    "        #Record the best loss\n",
    "        if smoothed_loss < best_loss or batch_num==1:\n",
    "            best_loss = smoothed_loss\n",
    "        #Store the values\n",
    "        losses.append(smoothed_loss)\n",
    "        \n",
    "        #taking log lr to have evenly spaced points on the plotted curve\n",
    "        log_lrs.append(math.log10(lr))\n",
    "        \n",
    "        #Do the SGD step\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        #Update the lr for the next step\n",
    "        lr *= mult\n",
    "        optimizer.param_groups[0]['lr'] = lr\n",
    "    return log_lrs, losses"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Test with SGD optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "net = SimpleNN([28*28,100],10,[.1])\n",
    "optimizer = optim.SGD(net.parameters(),lr=0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "log_lrs,losses = find_lr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f2d18da1cf8>]"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(log_lrs,losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f2d18c14908>]"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(log_lrs[10:-5],losses[10:-5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "net = SimpleNN([28*28,100],10,[.1])\n",
    "optimizer = optim.SGD(net.parameters(),lr=0.05)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1:\n",
      "Training Loss: 0.4815035904288292  Accuracy: 82.77166666666666\n",
      "Validation Loss: 0.42095525534152983  Accuracy: 84.96\n",
      "Epoch 2:\n",
      "Training Loss: 0.36804150743087133  Accuracy: 86.62166666666667\n",
      "Validation Loss: 0.3863409427881241  Accuracy: 85.76\n",
      "Epoch 3:\n",
      "Training Loss: 0.33324395702878634  Accuracy: 87.895\n",
      "Validation Loss: 0.3711943527460098  Accuracy: 86.72\n",
      "Epoch 4:\n",
      "Training Loss: 0.3106721770465374  Accuracy: 88.725\n",
      "Validation Loss: 0.36010925846099856  Accuracy: 87.12\n",
      "Epoch 5:\n",
      "Training Loss: 0.293251091097792  Accuracy: 89.34666666666666\n",
      "Validation Loss: 0.35338898152112963  Accuracy: 87.08\n",
      "Best iteration: 5 with val loss: 0.35338898152112963, val acc: 87.08 \n"
     ]
    }
   ],
   "source": [
    "train(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1:\n",
      "Training Loss: 0.47229704044262566  Accuracy: 82.91333333333333\n",
      "Validation Loss: 0.4190855041027069  Accuracy: 84.87\n",
      "Epoch 2:\n",
      "Training Loss: 0.36162473235527676  Accuracy: 86.83833333333334\n",
      "Validation Loss: 0.40521939635276794  Accuracy: 85.65\n",
      "Epoch 3:\n",
      "Training Loss: 0.3273186469376087  Accuracy: 88.07833333333333\n",
      "Validation Loss: 0.3733721847772598  Accuracy: 86.89\n",
      "Epoch 4:\n",
      "Training Loss: 0.3034215909818808  Accuracy: 88.92833333333333\n",
      "Validation Loss: 0.3651914852142334  Accuracy: 86.9\n",
      "Epoch 5:\n",
      "Training Loss: 0.28775047983924545  Accuracy: 89.315\n",
      "Validation Loss: 0.3698055908083916  Accuracy: 86.94\n",
      "Best iteration: 4 with val loss: 0.3651914852142334, val acc: 86.9 \n"
     ]
    }
   ],
   "source": [
    "net = SimpleNN([28*28,100],10,[.1])\n",
    "optimizer = optim.SGD(net.parameters(),lr=10**-1)\n",
    "\n",
    "train(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Selecting minimum learning rate (0.1) based on LR finder results in a worse validation loss since LR here is still high and loss is on the verge of increasing. A LR lower than it (in this case it's 0.03 or 10^-1.5) is more safe and effective"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Adam"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "net = SimpleNN([28*28,100],10,[.1])\n",
    "optimizer = optim.Adam(net.parameters(),lr=0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "log_lrs,losses = find_lr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f2d18b825f8>]"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(log_lrs[10:-5],losses[10:-5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1:\n",
      "Training Loss: 0.46227240395943325  Accuracy: 83.28\n",
      "Validation Loss: 0.40756298243999484  Accuracy: 85.28\n",
      "Epoch 2:\n",
      "Training Loss: 0.3565093750715256  Accuracy: 86.95333333333333\n",
      "Validation Loss: 0.39446539452075957  Accuracy: 85.44\n",
      "Epoch 3:\n",
      "Training Loss: 0.3195662118256092  Accuracy: 88.16333333333333\n",
      "Validation Loss: 0.39533088693618773  Accuracy: 86.07\n",
      "Epoch 4:\n",
      "Training Loss: 0.3002463410238425  Accuracy: 88.87833333333333\n",
      "Validation Loss: 0.36748897948265075  Accuracy: 86.55\n",
      "Epoch 5:\n",
      "Training Loss: 0.28221147474249203  Accuracy: 89.50666666666666\n",
      "Validation Loss: 0.3491303251862526  Accuracy: 87.34\n",
      "Best iteration: 5 with val loss: 0.3491303251862526, val acc: 87.34 \n"
     ]
    }
   ],
   "source": [
    "net = SimpleNN([28*28,100],10,[.1])\n",
    "optimizer = optim.Adam(net.parameters(),lr=10**-3)\n",
    "\n",
    "train(5)"
   ]
  }
 ],
 "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": "3.7.0"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}